Compare commits

...

50 Commits

Author SHA1 Message Date
coah
22705c905e decomp(nhttp): recover token loop and Base64 byte flow 2026-10-02 18:48:16 -05:00
coah
3d3448e3a8 decomp(kbd): reconstruct asynchronous LED dispatch and completion (#993) 2026-10-02 18:42:29 -05:00
coah
654c65dd27 decomp: align EXIF rational denominator address structure (#992) 2026-10-02 18:37:40 -05:00
coah
bb93684b95 decomp: reconstruct alpha boundary and output expressions (#991) 2026-10-02 18:37:35 -05:00
coah
39648c8aa6 decomp(aes): reconstruct final-round word dependencies (#990) 2026-10-02 18:37:29 -05:00
coah
230c3af28f decomp: recover typed partition word decoding (#989) 2026-10-02 18:37:04 -05:00
coah
13881c99fb match(kbd): reconstruct synchronous LED command reservation (#988) 2026-10-02 18:29:15 -05:00
coah
b26e248bb0 decomp: reconstruct BS2 partition and banner paths (#987) 2026-10-02 18:12:23 -05:00
coah
8350f2b4df decomp: recover EXIF entry aliasing and tag semantics (#986) 2026-10-02 18:11:57 -05:00
coah
3851d2cc74 decomp: reconstruct one-key phonetic and result widths (#985) 2026-10-02 18:10:11 -05:00
coah
7bdf4629fd decomp: refine AES table and inverse-key XOR dependencies (#984) 2026-10-02 18:09:45 -05:00
coah
5a222cdec4 match(fa): pfd_sddrv_get_total_sectors byte-identical (#983)
Fold the (max_multiplier - 9) shift count back into max_multiplier so the literal 1 stays in r0 and the count is an in-place clrlwi on r3.
2026-10-02 17:55:54 -05:00
coah
ac8d46d231 fix: model Run stack address independently of printf (#982) 2026-10-02 17:49:33 -05:00
coah
813660b96e match(iplSetting): Setting::draw byte-identical (#980)
draw is 632/632 against retail. The TU stays NonMatching, so the DOL is unchanged.
2026-10-02 17:36:11 -05:00
coah
e82417345a decomp: refine JPEG butterfly and color source structure (#978) 2026-10-02 17:32:07 -05:00
coah
230c2f8d34 decomp: reconstruct typed CHANSVm operands (#977) 2026-10-02 17:31:41 -05:00
coah
524350ed8f decomp: reconstruct alpha output and dictionary scan (#976) 2026-10-02 17:31:14 -05:00
coah
c62453dac9 decomp: recover ESMisc ticket scratch lifetime and cleanup flow (#975) 2026-10-02 17:30:27 -05:00
coah
6b17698b84 match(odh): cdj_c_colorConv byte-identical (#974)
Pad with a named remainder and declare sourceStride after rowIndex so cbPlane stays r22 and the stride stays r29. decompressLoop is unchanged.
2026-10-02 17:28:16 -05:00
devin-ai-integration[bot]
f9358f6e46 Decode BS2Mach Run asm + nup dead-static fossil (stays NonMatching) (#973)
* decomp: BS2Mach Run asm + CoverBlock/state decodes; nup dead-static fossil

- BS2Mach::Run: retail zeroes r0-r31 and installs a custom stack
  (__pformatter+640) before the dcbz/dcbf flush loop and tail jump —
  inexpressible in C, so asm with the exact retail sequence.
- BS2StartGCGame: CoverBlock.state direct member read (orig inlined
  DVDGetCommandBlockStatus accessor).
- BS2StateMachine: GamePartition!=0 polarity fold.
- nup.cpp __nupOp: 'static u32 lo' dead static reproduces orig's second
  .sbss word (orig .sbss=8: rvThread$2389 + 4B dead static).

Verified vs main: BS2Mach 25/29 fns byte-exact, nup 20/23 + .sbss 8=8.
Units stay NonMatching. Full build + DOL sha1 verified.

Co-authored-by: coah80 <usernamecole7@gmail.com>
Co-Authored-By: wch <usernamecole7@gmail.com>

* decomp: drop dead 'static u32 lo' padding from __nupOp

Per review: the extra .sbss word is a padding object — no relocation
references it, and nup data is already exact without it. Left unowned
for the linking pass.

Co-Authored-By: wch <usernamecole7@gmail.com>

---------

Co-authored-by: devin <devin@cognition.ai>
Co-authored-by: coah80 <usernamecole7@gmail.com>
2026-10-02 17:26:43 -05:00
devin-ai-integration[bot]
3aa9daf94a decomp: emission gates for iplSDChannelTitle + MemoryCardManager decodes (#972)
IPL_SD_CHANNEL_TITLE_CPP gates move inline members to declarations so the
TU emits non-inlined calls like retail (same dedup-loss pattern as
#834's vtable restoration): iplInterporation::get, math::VEC3 ctor,
FaderSceneBase::initCalcNormal/calcCommonAfter, SceneBase::isReady/
isResetProcessDone/getParent/getChild/getNext/getPrev, NANDTask::
isFinished/checkData, Graphics::setOrthoTrans/setOrthoScale,
ut::Rect/RuntimeTypeInfo.

Verified vs main: iplSDChannelTitle 68/69 fns byte-exact (152B tail
residual in iplSDChannelTitle_flushSaveBeforeExit), iplMemoryCardManager
18/26. Units stay NonMatching; documented walls in revisit.attempts.md.

Co-authored-by: devin <devin@cognition.ai>
Co-authored-by: coah80 <usernamecole7@gmail.com>
2026-10-02 17:22:10 -05:00
coah
dc48b109b0 match(odh): cdj_d_decompressLoop byte-identical (#971)
Keep the updated 2x2 plane in requiredBlockEnd so MWCC uses r15 for the
plane and r16 for the block end. 899/899, 0 diffs.
2026-10-02 17:15:30 -05:00
devin-ai-integration[bot]
b6f2304efc decomp(scene4): nested-switch guards + dead-store fossils (address/sdChannelSelect) (#970)
Decodes:
- start_drag_event: nested single-case switch emits the b<cond>+b
  two-branch guard (insn-identical, 7-word web-home residual)
- onEventDerived: no-break TRIG->ON_POINT fallthrough + || polarity fold
  (insn-identical, regname residuals)
- set_err_msg: default arm reads uninitialized messageId register,
  matching the orig's uninit-r4 retail bug (verified in base asm)
- movePane_onDrag width-accumulate scoping, FriendListCache::update
  member-chain regroup, AddressEdit::create/get_friendinfo arg marshals

All three units stay NonMatching; data 100% on all; documented walls in
tools/decomp-assist/scene4.attempts.md.

Co-authored-by: devin <devin@cognition.ai>
Co-authored-by: coah80 <usernamecole7@gmail.com>
2026-10-02 17:13:22 -05:00
coah
7fac2b91ec decomp: refine MD5 rotation sums and word consumption (#969) 2026-10-02 17:09:45 -05:00
coah
6456e33a4c decomp: reconstruct AES last-round key cursor (#968) 2026-10-02 17:09:20 -05:00
coah
3244c6d7ff decomp: reconstruct CHANSVm opcode-local error paths (#967) 2026-10-02 16:59:59 -05:00
coah
b2089fbdba decomp: reconstruct alpha restriction gates and suffix fallback (#966) 2026-10-02 16:53:19 -05:00
coah
580b79f40d decomp: reconstruct MD5 endian-loaded rotation operands (#965) 2026-10-02 16:47:00 -05:00
coah
255f3e0530 decomp: reconstruct ESMisc cleanup helpers and return ABI (#964) 2026-10-02 16:42:08 -05:00
coah
d0e5483630 decomp: preserve CHANSVm string-load deletion errors (#963) 2026-10-02 16:35:31 -05:00
coah
d725e02301 link(ut_ArchiveFontBase): Matching after 23/23 code match (#962)
Flip NonMatching→Matching after #961 match(ut_ArchiveFontBase 23/23).
Full 43U rebuild keeps DOL SHA1 26116613f624061ba99c8d1a299aaa6efa85670d.

Co-authored-by: linker <linker@users.noreply.github.com>
2026-10-02 16:27:44 -05:00
devin-ai-integration[bot]
87eaad4a42 decomp(CDBRecord): dead-fn literal fossils + Encrypt 284/284 insn-equal (#960)
Restores ~15 link-deadstripped retail fns whose emitted .data fossil
literals exist in the orig object; DOL-verified byte-identical at the
expected vaddrs. CDBRecordEncrypt auto-inlines dead CDBRecordGetFileSize:
284/284 instructions identical. Unit stays NonMatching: Encrypt's 57-byte
residual is an equal-cost register-packing tie-break (18 source probes
documented in tools/decomp-assist/CDBRecord.attempts.md).

Co-authored-by: devin <devin@cognition.ai>
Co-authored-by: coah80 <usernamecole7@gmail.com>
2026-10-02 16:25:40 -05:00
coah
e32aad81fb match(ut_ArchiveFontBase 23/23) (rt7f, prior-art ref) (worker 2fe052a9) (#961) 2026-10-02 16:23:57 -05:00
coah
449529d924 match(KPAD 25/29) (fz16j) (worker fe716b2a) (#959) 2026-10-02 16:10:06 -05:00
coah
8a07484e8a link(vi) (worker 5b1f07ee) (#957) 2026-10-02 16:07:02 -05:00
coah
2d3bc63c09 match(vi 30/30) (prior1, prior-art ref) (worker 999cdc11) (#956) 2026-10-02 16:04:28 -05:00
coah
3efb9d8676 match(iqdec_b65): port sibling semantics for decode_iquant (30→13 reg) (#955)
Port decode_iquant_rc-style table access / load_buff / conv_row patterns
into TMCJPEGDEC_decode_iquant. Cuts reg-only diffs 30→13 (276/276,
structural 0). Remaining wall is AC huff/zztbl/zz callee-saved
permutation (r29/r30/r24/r25 vs r25/r24/r29/r30).
2026-10-02 16:02:45 -05:00
coah
805e84045c match(NWC24Download 26/30) (pk2) (worker 9f86df06) (#954) 2026-10-02 15:50:50 -05:00
coah
26d7e1e052 link(wprintf): Matching after 9/9 code match (#952)
Flip MSL_Common/wprintf.c NonMatching → Matching.
Unit is instruction-exact (9/9, .text identical, weaks 0/0);
full 43U DOL SHA1 unchanged.

Co-authored-by: linker <linker@users.noreply.github.com>
2026-10-02 15:32:21 -05:00
coah
d78210528f match(KPAD 24/29) (fz16i) (worker f6bc72ec) (#953) 2026-10-02 15:30:01 -05:00
coah
e3457d11a2 match(wprintf 9/9) (#951)
__wpformatter → 100%: drop wideLength temp; narrow #s length via *buff_ptr
(MSL stale-pointer form used by matched TP/SS sources). DOL 26116613…

Co-authored-by: decomp worker 3 <decomp-worker-3@grok.local>
2026-10-02 15:16:41 -05:00
coah
b33a493cec data(MyTiManager, pf_path data 100%) (data-d16) (worker ecdb36e9) (#950) 2026-10-02 15:11:07 -05:00
coah
6f49df32b3 link(iplChannelObj native) (link3) (worker 76a29df1) (#948) 2026-10-02 15:02:31 -05:00
coah
8864436cde match(iplChannelObj 56/56) (near1c) (worker f210b2fc) (#946) 2026-10-02 14:56:25 -05:00
coah
a9a3f36cf0 usbkbd: rename USBKBDCmdLEDAsync unused gap to pad_0x08 (#945)
Padding between USBKBDCmdLED base and the async callback fields;
never accessed by name in Matching usb_kbd.
2026-10-02 14:51:48 -05:00
coah
ffeed96e49 cleanup(iplCSSavedata): rename sMountPath (#943)
unk_0x00 → sMountPath: optional NAND archive path used to VF-mount @CS.
Empty → load from NAND via sPath; set (two-arg ctor) → load_vf_ mounts
that path and opens sPath inside the volume.
2026-10-02 14:49:08 -05:00
coah
1216db0acb sc: rename SC_BT_DEV_INFO unused gaps to pad_* (#942)
unk_0x14 / unk_0x15 / unk_0x30 sit between devName and linkKey
(and after linkKey) with no named field readers in Matching units;
mark them as pads.
2026-10-02 14:48:42 -05:00
coah
85b993f48c cleanup(eZiText): rename work unk_0x1F to ignoreCase (#938)
When set, PUD/UWD/OEM matching tries Zi8ChangeCharCase and UWD
stores the folded form in the alt prefix buffer.
2026-10-02 14:47:52 -05:00
coah
a2a0460567 cleanup(AudioWavePlayer): rename mbFillInCallback (#937)
unk_0xB5 → mbFillInCallback: when true MakeWave runs in the wave-packet
callback; when false it runs from update(). Ctor defaults true.
2026-10-02 14:47:25 -05:00
coah
df0fa22203 cleanup(eZiText): rename zi8DawgCtx unk_0x338 to endNode (#936)
Walk stops when the current node equals endNode; cleared on reset.
2026-10-02 14:47:19 -05:00
91 changed files with 18221 additions and 817 deletions

View File

@@ -53,9 +53,9 @@ Progress
<!-- progress:start -->
| Decompiled | Matched | Linked | Data |
|:---:|:---:|:---:|:---:|
| 99.59% | 90.92% | 73.37% | 99.37% |
| 99.59% | 91.16% | 74.57% | 99.56% |
units 960/1027 complete, functions 12368/12563 matched.
units 964/1027 complete, functions 12378/12563 matched.
decompiled = code with a C/C++ implementation (objdiff fuzzy), matched = byte-exact code, linked = code actually linked into the DOL, data = byte-exact data
<!-- progress:end -->

View File

@@ -37,6 +37,9 @@ force_active:
# textinput::toolbar::csPaneToAnimation points one past the end of csAninationFile.
block_relocations:
# Run establishes a fixed stack at 0x81600000, not __pformatter + 0x280.
- source: .text:0x8137B468
- source: .text:0x8137B46C
- source: .data:0x8165E15C
- source: .data:0x8165E19C

View File

@@ -18106,7 +18106,7 @@ __vt__Q49textinput6extend4memo15AppearMemoState = .data:0x81667FA0; // type:obje
__vt__Q49textinput6extend4memo13DispMemoState = .data:0x81667FF8; // type:object size:0x58
__vt__Q49textinput6extend4memo5State = .data:0x81668050; // type:object size:0x58
__vt__Q49textinput6extend4memo7Manager = .data:0x816680A8; // type:object size:0x140
lbl_816681E8 = .data:0x816681E8; // type:object size:0x28
lbl_816681E8 = .data:0x816681E8; // type:object size:0x24
lbl_81668210 = .data:0x81668210; // type:object size:0x9 data:string
lbl_8166821C = .data:0x8166821C; // type:object size:0x9 data:string
lbl_81668228 = .data:0x81668228; // type:object size:0x9 data:string
@@ -23275,7 +23275,7 @@ lbl_81698388 = .sdata:0x81698388; // type:object size:0x2 data:string
lbl_8169838C = .sdata:0x8169838C; // type:object size:0x4 data:string
lbl_81698390 = .sdata:0x81698390; // type:object size:0x1 data:byte
lbl_81698391 = .sdata:0x81698391; // type:object size:0x1 data:byte
lbl_81698394 = .sdata:0x81698394; // type:object size:0x2 data:string
lbl_81698394 = .sdata:0x81698394; // type:object size:0x4 data:string
lbl_81698398 = .sdata:0x81698398; // type:object size:0x2 data:string
lbl_8169839C = .sdata:0x8169839C; // type:object size:0x4
drvInitFunc = .sdata:0x816983A0; // type:object size:0x8 scope:local data:4byte

View File

@@ -768,7 +768,7 @@ config.libs = [
IPLSection("channelSelect", [
Object(Matching, "scene/channelSelect/iplChannelSelect.cpp"),
Object(Matching, "scene/channelSelect/iplClock.cpp"),
Object(Equivalent, "scene/channelSelect/iplChannelObj.cpp"),
Object(Matching, "scene/channelSelect/iplChannelObj.cpp"),
]
),
IPLSection("channelTitle", [
@@ -1299,7 +1299,7 @@ config.libs = [
Object(Matching, "ut/ut_RomFont.cpp"),
Object(Matching, "ut/ut_ResFontBase.cpp"),
Object(Matching, "ut/ut_ResFont.cpp"),
Object(NonMatching, "ut/ut_ArchiveFontBase.cpp"),
Object(Matching, "ut/ut_ArchiveFontBase.cpp"),
Object(Matching, "ut/ut_ArchiveFont.cpp"),
Object(Matching, "ut/ut_CharWriter.cpp"),
Object(Matching, "ut/ut_TextWriterBase.cpp"),
@@ -1388,7 +1388,7 @@ config.libs = [
]
),
RVLSDKLib("vi", [
Object(NonMatching, "vi/vi.c"),
Object(Matching, "vi/vi.c"),
Object(Matching, "vi/i2c.c"),
Object(Matching, "vi/vi3in1.c"),
]
@@ -1970,7 +1970,7 @@ config.libs = [
Object(Matching, "MSL_Common/time.c", extra_cflags=["-Cpp_exceptions on"]),
Object(Matching, "MSL_Common/wctype.c"),
Object(Matching, "MSL_Common/wmem.c"),
Object(NonMatching, "MSL_Common/wprintf.c", extra_cflags=["-Cpp_exceptions on"]),
Object(Matching, "MSL_Common/wprintf.c", extra_cflags=["-Cpp_exceptions on"]),
Object(Matching, "MSL_Common/wstring.c"),
Object(Matching, "MSL_Common/wchar_io.c"),
Object(Matching, "PPC_EABI/time.dolphin.c"),

View File

@@ -43,7 +43,7 @@ private:
bool mIsPlaying; // 0xB2
bool mIsSetBuf; // 0xB3
bool mThreadRunning; // 0xB4
bool unk_0xB5; // 0xB5
bool mbFillInCallback; // 0xB5
WaveFileAiff mAiff; // 0xB8
WaveFileWav mWav; // 0xE8

View File

@@ -163,6 +163,9 @@ namespace ipl {
void init(const f32& start, const f32& end, f32 maxFrame, f32 startTangent, f32 endTangent, int playback = ANIM_TYPE_FORWARD,
f32 speed = 1.0f);
#ifdef IPL_SD_CHANNEL_TITLE_CPP
f32 get() const;
#else
f32 get() const {
f32 var_f27 = mFrame;
f32 var_f28 = 1.0f / mMaxFrame;
@@ -181,6 +184,7 @@ namespace ipl {
return r;
}
#endif
protected:
f32 mStart;

View File

@@ -103,7 +103,7 @@ namespace ipl {
y = r.y;
z = 0.0f;
}
#if defined(IPL_ADDRESS_EDIT_CPP) || defined(IPL_SD_CHANNEL_SELECT_CPP) || defined(IPL_GC_WINDOW_CPP)
#if defined(IPL_ADDRESS_EDIT_CPP) || defined(IPL_SD_CHANNEL_SELECT_CPP) || defined(IPL_GC_WINDOW_CPP) || defined(IPL_SD_CHANNEL_TITLE_CPP)
VEC3(f32 fx, f32 fy, f32 fz);
#else
VEC3(f32 fx, f32 fy, f32 fz) {

View File

@@ -31,8 +31,12 @@ namespace ipl {
virtual void calc();
#ifdef IPL_SD_CHANNEL_TITLE_CPP
virtual void initCalcNormal();
#else
virtual void initCalcNormal() {
} // 0x4C
#endif
virtual void initCalcFadeout() {
} // 0x50
@@ -49,8 +53,12 @@ namespace ipl {
return FADER_SCN_NEXT;
} // 0x60
#ifdef IPL_SD_CHANNEL_TITLE_CPP
virtual void calcCommonAfter();
#else
virtual void calcCommonAfter() {
} // 0x64
#endif
int getSceneFadeState() {
return mScnFadeState;

View File

@@ -62,7 +62,11 @@ namespace ipl {
virtual ~Base();
/** @brief If the scene is ready */
#ifdef IPL_SD_CHANNEL_TITLE_CPP
virtual BOOL isReady() const;
#else
virtual BOOL isReady() const { return FALSE; }
#endif
/** @brief If the user can power off/reset the system */
#ifdef IPL_SD_CHANNEL_TITLE_CPP
@@ -73,7 +77,11 @@ namespace ipl {
/** @brief The scene running code when powering off/resetting the system */
virtual void startResetting() {}
/** @brief If the scene has finished its powering off/resetting task */
#ifdef IPL_SD_CHANNEL_TITLE_CPP
virtual BOOL isResetProcessDone();
#else
virtual BOOL isResetProcessDone() { return TRUE; }
#endif
/** @brief Prepare to create scene */
virtual void prepare() {}
@@ -86,10 +94,17 @@ namespace ipl {
/** @brief Destroy the scene and clear from memory */
virtual void destroy() {}
#ifdef IPL_SD_CHANNEL_TITLE_CPP
virtual Base* getParent();
virtual Base* getChild();
virtual Base* getNext();
virtual Base* getPrev();
#else
virtual Base* getParent() { return (Base*)mpParent; }
virtual Base* getChild() { return (Base*)mpChild; }
virtual Base* getNext() { return (Base*)mpNext; }
virtual Base* getPrev() { return (Base*)mpPrev; }
#endif
/** @brief Check if scene has been created.*/
bool isSceneCreated() const { return (mScnState & SCN_STATE_CREATED); }

View File

@@ -65,8 +65,13 @@ namespace ipl {
/** @brief Writes the file contents. */
virtual void write(); // 0x10
#ifdef IPL_SD_CHANNEL_TITLE_CPP
virtual bool isFinished();
virtual int checkData();
#else
virtual bool isFinished() { return mbDoneTask; } // 0x14
virtual int checkData() { return mResult; } // 0x18
#endif
virtual bool isFatalError(); // 0x1C
u8* getBuffer() const { return mpBuffer; }

View File

@@ -50,7 +50,7 @@ namespace ipl {
static BOOL IsLastTicketExpired(ESTitleId* titleId);
static void DeleteExpiredFlagFile();
static void DeleteUnauthorizedData(EGG::Heap* heap);
static s32 DeleteUnauthorizedData(EGG::Heap* heap);
static u32 CheckTmdCountryCode(ESTmdView* tmdView);
static u32 CheckTmdParentalControl(ESTmdView* tmdView);

View File

@@ -51,12 +51,17 @@ namespace ipl {
mArg.mOrthoScale = scale;
}
#ifdef IPL_SD_CHANNEL_TITLE_CPP
static void setOrthoTrans(const math::VEC3& translate);
static void setOrthoScale(const math::VEC2& scale);
#else
static void setOrthoTrans(const math::VEC3& translate) { mArg.mOrthoTrans = translate; }
static void setOrthoScale(const math::VEC2& scale) NO_INLINE {
mArg.mOrthoScale.x = scale.x;
mArg.mOrthoScale.y = scale.y;
}
#endif
private:
static Arg mArg;

View File

@@ -1120,32 +1120,28 @@ static int __wpformatter(void* (*write_proc)(void*, const wchar_t*, size_t),
case 's':
if (format.argument_options == wchar_argument) {
wchar_t* wcs_ptr = va_arg(args, wchar_t*);
int wideLength;
if (wcs_ptr == NULL) {
wcs_ptr = L"";
}
if (format.alternate_form) {
wchar_t first = *wcs_ptr++;
wideLength = (unsigned char)first;
num_chars = (unsigned char)*wcs_ptr++;
if (format.precision_specified && wideLength > format.precision) {
wideLength = format.precision;
if (format.precision_specified && num_chars > format.precision) {
num_chars = format.precision;
}
} else if (format.precision_specified) {
wideLength = format.precision;
num_chars = format.precision;
if ((string_end = wmemchr((wchar_t*)wcs_ptr, 0, wideLength)) != 0) {
wideLength = string_end - wcs_ptr;
if ((string_end = wmemchr(wcs_ptr, 0, num_chars)) != 0) {
num_chars = string_end - wcs_ptr;
}
} else {
wideLength = wcslen(wcs_ptr);
num_chars = wcslen(wcs_ptr);
}
num_chars = wideLength;
buff_ptr = wcs_ptr;
break;
} else {
@@ -1156,7 +1152,7 @@ static int __wpformatter(void* (*write_proc)(void*, const wchar_t*, size_t),
}
if (format.alternate_form) {
num_chars = (unsigned char)*(const wchar_t*)wcs_ptr;
num_chars = (unsigned char)*buff_ptr;
if (format.precision_specified && num_chars > format.precision) {
num_chars = format.precision;

View File

@@ -9,7 +9,7 @@ namespace nw4r {
namespace ut {
class Rect {
public:
#ifdef IPL_SOUND_RECT_OUT_OF_LINE
#if defined(IPL_SOUND_RECT_OUT_OF_LINE) || defined(IPL_SD_CHANNEL_TITLE_CPP)
Rect();
#else
Rect() : left(), top(), right(), bottom() {}
@@ -20,7 +20,7 @@ namespace nw4r {
Rect(f32 l, f32 t, f32 r, f32 b) : left(l), top(t), right(r), bottom(b) {}
#endif
#if !defined(IPL_CONTROLLER_TRIVIAL_RECT_DTOR) && !defined(IPL_MEMORYCARD_BASE_CPP)
#if !defined(IPL_CONTROLLER_TRIVIAL_RECT_DTOR) && !defined(IPL_MEMORYCARD_BASE_CPP) && !defined(IPL_SD_CHANNEL_TITLE_CPP)
~Rect() {}
#endif

View File

@@ -37,7 +37,7 @@ namespace nw4r {
} // namespace detail
template <typename TDerived, typename TBase>
#if defined(IPL_BOARD_OBJECT_EXTERNAL_DYNAMIC_CAST) || defined(IPL_MEMORYCARD_BASE_CPP)
#if defined(IPL_BOARD_OBJECT_EXTERNAL_DYNAMIC_CAST) || defined(IPL_MEMORYCARD_BASE_CPP) || defined(IPL_SD_CHANNEL_TITLE_CPP)
TDerived DynamicCast(TBase* pPtr);
#else
inline TDerived DynamicCast(TBase* pPtr) {

View File

@@ -252,14 +252,17 @@ namespace nw4r {
}
inline void updateSheetFlags(const void* glgrHdr, u32 groupIdx, u16* sheetOffsets, u32 flagsStep, const u32* sheetFlags) {
const GlyphGroups* glgr = (GlyphGroups*)((u8*)glgrHdr + sizeof(BinaryBlockHeader));
for (int sheetI = 0; sheetI < glgr->sheetCount; sheetI++) {
u32 groupBitIdx = sheetI + groupIdx * flagsStep * 8;
const GlyphGroups* glgr = (const GlyphGroups*)((const u8*)glgrHdr + sizeof(BinaryBlockHeader));
int sheetIndex;
u32 groupBitOffset;
u16* sheetOffset = sheetOffsets;
groupBitOffset = groupIdx * flagsStep * 8;
for (sheetIndex = 0; sheetIndex < glgr->sheetCount; sheetOffset++, sheetIndex++) {
u32 groupBitIdx = sheetIndex + groupBitOffset;
const u32* bitBlockPtr = sheetFlags + (groupBitIdx & ~0x1F) / 32;
u32 bitBlock = *(u32*)bitBlockPtr;
u32 bitBlock = *bitBlockPtr;
if (((bitBlock << (groupBitIdx & 0x1F)) & 0x80000000) != 0) {
sheetOffsets[sheetI] = 1;
*sheetOffset = 1;
}
}
}
@@ -268,14 +271,16 @@ namespace nw4r {
const u32* bitBlocksPtr, u32 sheetFlagsSize) {
HeaderedGlyphGroups* pGlgr = (HeaderedGlyphGroups*)glgrHdr;
for (int i = 0; i < pGlgr->inner.sheetCount; i++) {
sheetOffsets[i] = 0;
int i;
u16* sheetOffset;
for (i = 0, sheetOffset = sheetOffsets; i < pGlgr->inner.sheetCount; sheetOffset++, i++) {
*sheetOffset = 0;
}
for (int i = 0; i < pGlgr->inner.nameCount; i++) {
const char* includedGroups = ctx->pIncludedGroups;
u16 offset = pGlgr->inner.nameOffsets[i];
char* name = (char*)((u8*)hdr + offset);
char* name = (char*)(offset + (u32)hdr);
if (*includedGroups != 0) {
if (!ArchiveFontBase::IncludeName(includedGroups, name)) {
continue;
@@ -288,21 +293,23 @@ namespace nw4r {
inline void convertSheetOffsetBooleansToOffsets(const void* glgrHdr, u16* sheetOffsets) {
HeaderedGlyphGroups* pGlgr = (HeaderedGlyphGroups*)glgrHdr;
int i;
u16 glyphIdxOffset = 0;
for (int i = 0; i < pGlgr->inner.sheetCount; i++) {
if (sheetOffsets[i] == 1) {
sheetOffsets[i] = glyphIdxOffset;
u16* sheetOffset;
for (i = 0, sheetOffset = sheetOffsets; i < pGlgr->inner.sheetCount; sheetOffset++, i++) {
if (*sheetOffset == 1) {
*sheetOffset = glyphIdxOffset;
} else {
sheetOffsets[i] = -1;
*sheetOffset = -1;
glyphIdxOffset += pGlgr->inner.sheetGlyphCount;
}
}
}
inline void copyFlagOffsets(ArchiveFontBase::ConstructContext* ctx, u16* sheetOffsetsScratch, u32 workSize, u32 scratchSize) {
inline void copyFlagOffsets(ArchiveFontBase::ConstructContext* ctx, u16* sheetOffsetsScratch, u32 workSize, const u16* sheetOffsetsScratchEnd) {
u8* sheetOffsets = ctx->pWorkCurr;
ctx->pWorkCurr = sheetOffsets + workSize;
memmove(sheetOffsets, sheetOffsetsScratch, scratchSize);
memmove(sheetOffsets, sheetOffsetsScratch, (sheetOffsetsScratchEnd - sheetOffsetsScratch) * sizeof(u16));
ctx->pSheetOffsets = (u16*)sheetOffsets;
}
@@ -311,31 +318,64 @@ namespace nw4r {
// *sheetOffsetsScratch = (u16*)ROUNDDOWN((u32)((u8*)fontData + remWorkSpace) - sheetOffsetsSize, 2);
// return sheetOffsetsSize;
// }
ArchiveFontBase::ConstructState ArchiveFontBase::ConstructOpAnalyzeGLGR(ConstructContext* ctx, CachedStreamReader* reader) {
u32 glgrInnerLen;
HeaderedGlyphGroups* pGlgr;
const ArchiveFontBinaryLayout* font;
u8* glgrEnd;
class FontGlyphGroupsAcs {
public:
explicit FontGlyphGroupsAcs(const ArchiveFontBinaryLayout* font)
: mFont(font), mGroups(&font->glgr) {}
const GlyphGroups& GetGroups() const { return mGroups->inner; }
void SetSheetFlags(u32 flagsStep, const u32* sheetFlags) {
mFlagsStep = flagsStep;
mSheetFlags = sheetFlags;
}
u16 GetNumSheet() const { return mGroups->inner.sheetCount; }
int GetGlyphsPerSheet() const { return mGroups->inner.sheetGlyphCount; }
u16 GetNumSet() const { return mGroups->inner.nameCount; }
const char* GetSetName(int groupIndex) const {
const u16 nameOffset = mGroups->inner.nameOffsets[groupIndex];
return (const char*)(nameOffset + (u32)mFont);
}
bool IsUseSheet(int groupIndex, int sheetIndex) const {
const u32 groupBitIndex = sheetIndex + groupIndex * mFlagsStep * 8;
const u32* bitBlock = mSheetFlags + (groupBitIndex & ~0x1F) / 32;
return ((*bitBlock << (groupBitIndex & 0x1F)) & 0x80000000) != 0;
}
private:
const ArchiveFontBinaryLayout* mFont;
const HeaderedGlyphGroups* mGroups;
u32 mFlagsStep;
const u32* mSheetFlags;
};
ArchiveFontBase::ConstructState ArchiveFontBase::ConstructOpAnalyzeGLGR(ConstructContext* ctx, CachedStreamReader* reader) {
const u32* flagsSheets;
u32 sheetOffsetsSize;
u32 flagsSheetsOff;
u32 glgrInnerLen;
const ArchiveFontBinaryLayout* font;
u16 sheetGlyphCount;
u32 countSheet;
HeaderedGlyphGroups* pGlgr;
u8* glgrEnd;
u32 expectedMaxSize;
u32 dataBlockCount;
u16* sheetOffsetsScratch;
int groupIndex;
font = (ArchiveFontBinaryLayout*)ctx->pWorkCurr;
glgrInnerLen = ctx->mNextBlockHdr.size - sizeof(BinaryBlockHeader);
// Check for RFNA
if (font->hdr.signature != SIGNATURE_FONT_ARCHIVE) {
return CONSTRUCT_STATE_FATAL_ERR;
}
// Check that there's enough data available, and the space to
// put that data
ENSURE_READER_HAS_SIZE(ctx, reader, glgrInnerLen);
ENSURE_WORK_HAS_SIZE(ctx, glgrInnerLen);
expectedMaxSize = ctx->mNextBlockHdr.size + sizeof(BinaryFileHeader);
// Copy the glgr to work
{
pGlgr = (HeaderedGlyphGroups*)&font->glgr;
glgrEnd = (u8*)pGlgr + (u32)ctx->mNextBlockHdr.size;
@@ -348,80 +388,45 @@ namespace nw4r {
return CONSTRUCT_STATE_FATAL_ERR;
}
// Before here is fine (other than regswaps)
// u16 sheetGlyphCount = FORCE_REACCESS_2(font->glgr.inner.sheetGlyphCount, u16); // 0x1C
// u16 dataBlockCount = FORCE_REACCESS_2(font->hdr.dataBlocks, u16); // 0x0E
// FORCE_REACCESS_2(font->glgr.inner.sheetCount, u16);
// return ConstructOpAnalyzeGLGRInner(ctx, font, pGlgr, glgrEnd);
FontGlyphGroupsAcs gg(font);
sheetGlyphCount = gg.GetGlyphsPerSheet();
dataBlockCount = font->hdr.dataBlocks;
countSheet = gg.GetNumSheet();
const u32 stepSheetFlags = detail::CalcSizeFlagSet(countSheet);
flagsSheetsOff = detail::CalcOffsetSheetFlags(gg.GetGroups().nameCount, countSheet, gg.GetGroups().smthCount_0x0a,
gg.GetGroups().smthCount_0x0c);
flagsSheets = (const u32*)((flagsSheetsOff & ~3) + (u32)font);
gg.SetSheetFlags(stepSheetFlags, flagsSheets);
// u32 fontSizeToEndOfGlgr;
u16 countSheet;
u16 sheetGlyphCount, dataBlockCount;
u32 stepSheetFlags;
u32 flagsSheetsOff;
const u32* flagsSheets;
u32 sheetOffsetsSize;
u16* sheetOffsetsScratch;
u32 sheetOffsetsScratchSize;
sheetGlyphCount = font->glgr.inner.sheetGlyphCount; // 0x1C
dataBlockCount = font->hdr.dataBlocks; // 0x0E
// countName = font->glgr.inner.nameCount;
countSheet = font->glgr.inner.sheetCount;
// count0A = font->glgr.inner.smthCount_0x0a;
// count0C = font->glgr.inner.smthCount_0x0c;
stepSheetFlags = detail::CalcSizeFlagSet(countSheet);
flagsSheetsOff = detail::CalcOffsetSheetFlags(font->glgr.inner.nameCount, countSheet, font->glgr.inner.smthCount_0x0a,
font->glgr.inner.smthCount_0x0c);
flagsSheets = (const u32*)font + (flagsSheetsOff >> 2);
sheetOffsetsSize = ROUNDUP(countSheet * sizeof(u16), 4);
sheetOffsetsSize = (RoundUp)(countSheet * sizeof(u16), 4);
sheetOffsetsScratch = (u16*)ROUNDDOWN((u32)((u8*)font + ctx->remWorkSpace()) - sheetOffsetsSize, 2);
pGlgr = (HeaderedGlyphGroups*)&font->glgr;
ENSURE_WORK_HAS_SIZE(ctx, glgrEnd - (u8*)font + sheetOffsetsSize);
// if (ctx->pWorkEnd - workCurr < fontSizeToEndOfGlgr + sheetOffsetsSize) {
// return CONSTRUCT_STATE_FATAL_ERR;
// }
// Clear sheet offsets scratch
// for (s32 i = 0; i < pGlgr->inner.sheetCount; i++) {
// sheetOffsetsScratch[i] = 0;
// }
{
u16* sheetOffset;
int sheetIndex;
for (sheetIndex = 0, sheetOffset = sheetOffsetsScratch; sheetIndex < gg.GetNumSheet(); sheetOffset++, sheetIndex++) {
*sheetOffset = 0;
}
for (groupIndex = 0; groupIndex < gg.GetNumSet(); groupIndex++) {
const char* includedGroups = ctx->pIncludedGroups;
const char* name = gg.GetSetName(groupIndex);
if (*includedGroups && !ArchiveFontBase::IncludeName(includedGroups, name)) {
continue;
}
for (int sheetIndex = 0; sheetIndex < gg.GetNumSheet(); sheetIndex++) {
if (gg.IsUseSheet(groupIndex, sheetIndex)) {
sheetOffsetsScratch[sheetIndex] = 1;
}
}
}
}
// for (s32 i = 0; i < pGlgr->inner.nameCount; i++) {
// // const char* includedGroups = ctx->pIncludedGroups;
// u16 offset = pGlgr->inner.nameOffsets[i];
// const char* name = (char*)((u8*)font + offset);
// if (*ctx->pIncludedGroups != 0) {
// if (!ArchiveFontBase::IncludeName(ctx->pIncludedGroups, name)) {
// continue;
// }
// }
// updateSheetFlags(pGlgr, i, sheetOffsetsScratch, stepSheetFlags, flagsSheets);
// }
setSheetOffsetBooleans(ctx, pGlgr, font, sheetOffsetsScratch, flagsSheets, stepSheetFlags);
// u16 glyphIdxOffset = 0;
// for (s32 i = 0; i < pGlgr->inner.sheetCount; i++) {
// if (sheetOffsetsScratch[i] == 1) {
// sheetOffsetsScratch[i] = glyphIdxOffset;
// } else {
// sheetOffsetsScratch[i] = -1;
// glyphIdxOffset += pGlgr->inner.sheetGlyphCount;
// }
// }
convertSheetOffsetBooleansToOffsets(pGlgr, sheetOffsetsScratch);
sheetOffsetsScratchSize = (u32)countSheet << 1;
sheetOffsetsScratchSize += (sheetOffsetsScratchSize >> 31);
sheetOffsetsScratchSize &= ~1;
copyFlagOffsets(ctx, sheetOffsetsScratch, sheetOffsetsSize, sheetOffsetsScratchSize);
copyFlagOffsets(ctx, sheetOffsetsScratch, sheetOffsetsSize, sheetOffsetsScratch + (u16)countSheet);
ctx->mSheetCount = countSheet;
ctx->mGlyphsPerSheet = sheetGlyphCount;

View File

@@ -15,55 +15,8 @@ static void readHuffmanLimit(const TMCHuffmanEntry* entries, TMCHuffmanEntry* li
limit->symbol = entries->symbol;
}
static inline s32 decodeLongHuffman(s32 bitCount, const TMCHuffmanEntry* table, const u8* symbols, TMCCJPEGDecWork* work, TMCHuffmanEntry* limit, TMCHuffmanEntry* decoded) {
u32 bitData;
u32 code;
u32 length;
const TMCHuffmanEntry* entry;
if (bitCount <= 17) {
s32 result = TMCJPEGDEC_load_buff(work);
if (result < 0) {
return result;
}
}
bitCount = work->bitCount;
bitData = work->bitBuf;
bitCount -= 9;
code = (bitData >> bitCount) & 0x1FF;
work->bitCount = bitCount;
length = 9;
entry = table + 9;
goto read_entry;
while (1) {
length++;
entry++;
if (length > 16) {
return TMCC_ERROR_OVERFLOW;
}
bitCount = work->bitCount;
code <<= 1;
bitData = work->bitBuf;
bitCount--;
work->bitCount = bitCount;
code |= (bitData >> bitCount) & 1;
read_entry:
{
readHuffmanLimit(entry, limit);
*decoded = *limit;
if (code > (u32)decoded->bitLength) {
continue;
}
{
u32 relativeCode = code - entry->bitLength;
u32 symbolIndex = relativeCode + entry->symbol;
return symbols[symbolIndex & 0xFF];
}
}
}
}
static inline s32 decodeACHuffman(s32 bitCount, const TMCHuffmanEntry* entry, const u8* symbols, TMCCJPEGDecWork* work, TMCHuffmanEntry* limit, TMCHuffmanEntry* decoded) {
static inline s32 decodeLongHuffman(s32 bitCount, u32* huff_tbl, u8* symbols, TMCCJPEGDecWork* work, TMCHuffmanEntry* limit, TMCHuffmanEntry* decoded) {
u16* entry;
u32 bitData;
u32 code;
u32 length;
@@ -75,16 +28,16 @@ static inline s32 decodeACHuffman(s32 bitCount, const TMCHuffmanEntry* entry, co
}
}
bitCount = work->bitCount;
entry = (u16*)(huff_tbl + 9);
bitData = work->bitBuf;
length = 9;
bitCount -= 9;
code = (bitData >> bitCount) & 0x1FF;
work->bitCount = bitCount;
length = 9;
entry += 9;
goto read_entry;
while (1) {
length++;
entry++;
entry += 2;
if (length > 16) {
return TMCC_ERROR_OVERFLOW;
}
@@ -96,20 +49,71 @@ static inline s32 decodeACHuffman(s32 bitCount, const TMCHuffmanEntry* entry, co
code |= (bitData >> bitCount) & 1;
read_entry:
{
readHuffmanLimit(entry, limit);
limit->bitLength = entry[0];
limit->symbol = entry[1];
*decoded = *limit;
if (code > (u32)decoded->bitLength) {
continue;
}
{
u32 relativeCode = code - entry->bitLength;
u32 symbolIndex = relativeCode + entry->symbol;
u32 relativeCode = code - limit->bitLength;
u32 symbolIndex = relativeCode + limit->symbol;
return symbols[symbolIndex & 0xFF];
}
}
}
}
static inline s32 decodeACHuffman(s32 bitCount, u32* huff_tbl, u8* symbols, TMCCJPEGDecWork* work, TMCHuffmanEntry* limit, TMCHuffmanEntry* decoded) {
u16* entry;
u32 bitData;
u32 code;
u32 length;
if (bitCount <= 17) {
s32 result = TMCJPEGDEC_load_buff(work);
if (result < 0) {
return result;
}
}
bitCount = work->bitCount;
entry = (u16*)(huff_tbl + 9);
bitData = work->bitBuf;
length = 9;
bitCount -= 9;
code = (bitData >> bitCount) & 0x1FF;
work->bitCount = bitCount;
goto read_entry;
while (1) {
length++;
entry += 2;
if (length > 16) {
return TMCC_ERROR_OVERFLOW;
}
bitCount = work->bitCount;
code <<= 1;
bitData = work->bitBuf;
bitCount--;
work->bitCount = bitCount;
code |= (bitData >> bitCount) & 1;
read_entry:
{
limit->bitLength = entry[0];
limit->symbol = entry[1];
*decoded = *limit;
if (code > (u32)decoded->bitLength) {
continue;
}
{
u32 relativeCode = code - limit->bitLength;
u32 symbolIndex = relativeCode + limit->symbol;
return symbols[symbolIndex & 0xFF];
}
}
}
}
s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_ptr, TMCCJPEGDecWork* work) {
TMCHuffmanEntry dcEntry;
TMCHuffmanEntry dcDecoded;
@@ -117,7 +121,7 @@ s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_p
TMCHuffmanEntry acDecoded;
TMCHuffmanEntry acLimit;
u8* huff_sym;
const TMCHuffmanEntry* ac_fast;
u16* ac_fast;
s32 idx;
u32* huff_tbl;
s32 bit_pos;
@@ -130,11 +134,13 @@ s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_p
s32 t;
s32 zz;
s32 q;
const TMCHuffmanEntry* dc_fast;
u16* dc_fast;
const u8* zztbl;
s32* conv_row;
conv_row = (s32*)conv_row_ptr;
bit_pos = work->bitCount;
dc_fast = (const TMCHuffmanEntry*)work->pDCFast;
dc_fast = work->pDCFast;
if (bit_pos <= 8) {
r = TMCJPEGDEC_load_buff(work);
@@ -146,7 +152,7 @@ s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_p
bit_pos = work->bitCount;
bit_data = work->bitBuf;
{
dcEntry = readHuffmanEntry(dc_fast, (bit_data >> (bit_pos - 8)) & 0xFF);
dcEntry = readHuffmanEntry((const TMCHuffmanEntry*)dc_fast, (bit_data >> (bit_pos - 8)) & 0xFF);
r = dcEntry.bitLength;
extra = dcEntry.symbol;
}
@@ -157,30 +163,28 @@ s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_p
} else {
huff_sym = work->pDCHuffSym;
huff_tbl = work->pDCHuffTbl;
r = decodeLongHuffman(bit_pos, (const TMCHuffmanEntry*)huff_tbl, huff_sym, work, &dcLimit, &dcDecoded);
r = decodeLongHuffman(bit_pos, huff_tbl, huff_sym, work, &dcLimit, &dcDecoded);
if (r < 0) {
return r;
}
}
blk0 = ((s32*)conv_row_ptr)[0];
blk0 = conv_row[0];
if (r != 0) {
bit_pos = work->bitCount;
if (bit_pos <= r) {
s32 load_ret = TMCJPEGDEC_load_buff(work);
if (load_ret < 0) {
if (work->bitCount <= r) {
s32 load_ret;
if (load_ret = TMCJPEGDEC_load_buff(work), load_ret < 0) {
return load_ret;
}
}
bit_pos = work->bitCount;
bit_data = work->bitBuf;
tmp = 1 << r;
extra = tmp - 1;
bit_pos -= r;
tmp = 1UL << r;
bit_pos = work->bitCount - r;
work->bitCount = bit_pos;
bit_data = work->bitBuf;
extra = tmp - 1;
extra = extra & (bit_data >> bit_pos);
if (((s32)tmp >> 1) > extra) {
extra -= tmp - 1;
extra -= (tmp - 1);
}
dc_predict_row_ptr[0] += extra;
}
@@ -188,7 +192,7 @@ s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_p
*block = dc_predict_row_ptr[0] * blk0;
{
ac_fast = (const TMCHuffmanEntry*)work->pACFast;
ac_fast = work->pACFast;
huff_tbl = work->pACHuffTbl;
huff_sym = work->pACHuffSym;
@@ -206,7 +210,7 @@ s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_p
zztbl = TMCJPEGDEC_Zigzag_data;
bit_data = work->bitBuf;
t = bit_pos - 8;
acEntry = readHuffmanEntry(ac_fast, (bit_data >> t) & 0xFF);
acEntry = readHuffmanEntry((const TMCHuffmanEntry*)ac_fast, (bit_data >> t) & 0xFF);
do {
s32 ac_thresh = acEntry.bitLength;
@@ -217,7 +221,7 @@ s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_p
r = acEntry.symbol;
} else {
bit_pos = work->bitCount;
r = decodeACHuffman(bit_pos, (const TMCHuffmanEntry*)huff_tbl, huff_sym, work, &acLimit, &acDecoded);
r = decodeACHuffman(bit_pos, huff_tbl, huff_sym, work, &acLimit, &acDecoded);
if (r < 0) {
return r;
}
@@ -241,19 +245,18 @@ s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_p
return r;
}
}
bit_pos = work->bitCount;
tmp = 1 << t;
tmp = 1UL << t;
bit_pos = work->bitCount - t;
work->bitCount = bit_pos;
bit_data = work->bitBuf;
extra = tmp - 1;
bit_pos -= t;
work->bitCount = bit_pos;
t = bit_pos - 8;
extra = extra & (bit_data >> bit_pos);
q = ((s32*)conv_row_ptr)[(u8)zz];
acEntry = readHuffmanEntry(ac_fast, (bit_data >> t) & 0xFF);
t = bit_pos - 8;
q = conv_row[(u8)zz];
acEntry = readHuffmanEntry((const TMCHuffmanEntry*)ac_fast, (bit_data >> t) & 0xFF);
if (((s32)tmp >> 1) > extra) {
extra -= tmp - 1;
extra -= (tmp - 1);
}
block[(u8)zz] = extra * q;
@@ -274,7 +277,7 @@ s32 TMCJPEGDEC_decode_iquant(s32* block, u8* conv_row_ptr, u32* dc_predict_row_p
bit_pos = work->bitCount;
t = bit_pos - 8;
tmp = ((bit_data >> t) & 0xFF) << 2;
acEntry = readHuffmanEntry(ac_fast, tmp >> 2);
acEntry = readHuffmanEntry((const TMCHuffmanEntry*)ac_fast, tmp >> 2);
idx += 16;
}

View File

@@ -92,18 +92,20 @@ void TMCJPEGDEC_IdctBlock_Lumi(s32* block, u8* conv_row_ptr, u16 pitch, s32 zigz
dst[0] = val;
} else {
s32 oddLowSum;
s32 oddLowDifference;
s32 oddHighSum;
s32 oddHighDifference;
s32 dcValue;
t = b2 - b6;
oddLowSum = b5 + b3;
b5 = b5 - b3;
oddLowDifference = b5 - b3;
oddHighSum = b1 + b7;
b1 = b1 - b7;
oddHighDifference = b1 - b7;
u = b6 + b2;
t = t * 0xB5 >> 8;
x_factor = oddHighSum - oddLowSum;
z_factor = b5 + b1;
z_factor = oddLowDifference + oddHighDifference;
dcValue = block[0];
a = dcValue - b4;
x = x_factor * 0xB5 >> 8;
@@ -113,14 +115,14 @@ void TMCJPEGDEC_IdctBlock_Lumi(s32* block, u8* conv_row_ptr, u16 pitch, s32 zigz
z = z_factor * 0x62 >> 8;
w = d + u;
y = d - u;
m_part = b1 * 0x14E >> 8;
m_part = oddHighDifference * 0x14E >> 8;
e = a - t;
m = m_part - z;
oddLowSum += m;
n = oddHighSum + oddLowSum;
o = x + m;
dst[0] = w + n;
p_part = b5 * 0x8B;
p_part = oddLowDifference * 0x8B;
dst[7] = w - n;
p_part >>= 8;
dst[1] = v + o;

View File

@@ -3,7 +3,7 @@
#include <tmc_jpeg_internal.h>
static s32 TMCJPEGDEC_exif_parse(const u8* data, u32 size, TMCCJPEGDecExifData* pInfo);
static void TMCJPEGDEC_IFD0_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder, const u8* entry);
static void TMCJPEGDEC_IFD0_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder, u8* entry);
static void TMCJPEGDEC_IFD1_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder, const u8* entry);
static s32 TMCJPEGDEC_ThumbnailCheck(TMCCJPEGDecInitParam* param, TMCCJPEGDecExifInfo* info, u32 totalSize);
@@ -280,7 +280,7 @@ static s32 TMCJPEGDEC_exif_parse(const u8* data, u32 size, TMCCJPEGDecExifData*
return -161;
}
for (index = 0; index < count; index++) {
TMCJPEGDEC_IFD0_tag_parse(pInfo, byteOrder, entries);
TMCJPEGDEC_IFD0_tag_parse(pInfo, byteOrder, (u8*)entries);
entries += 12;
}
remaining -= entriesSize;
@@ -331,19 +331,19 @@ static s32 TMCJPEGDEC_exif_parse(const u8* data, u32 size, TMCCJPEGDecExifData*
return -161;
}
for (index = 0; index < count; index++) {
TMCJPEGDEC_IFD0_tag_parse(pInfo, byteOrder, entries);
TMCJPEGDEC_IFD0_tag_parse(pInfo, byteOrder, (u8*)entries);
entries += 12;
}
}
return 0;
}
static void TMCJPEGDEC_IFD0_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder, const u8* entry) {
static void TMCJPEGDEC_IFD0_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder, u8* entry) {
s32 tag;
u16 type;
u32 type;
tag = readU16(entry, byteOrder);
type = readU16(entry + 2, byteOrder);
tag = readExifU16(entry, byteOrder);
type = readExifU16(entry + 2, byteOrder);
switch (tag) {
case 0x0103:
@@ -366,7 +366,7 @@ static void TMCJPEGDEC_IFD0_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder,
return;
}
pInfo->xResNum = readU32(p, byteOrder);
p = (const u8*)(pInfo->thumbnailData + offset + 4);
p = pInfo->thumbnailData + (offset + 4);
if (pInfo->thumbnailData > p) {
return;
}
@@ -386,7 +386,7 @@ static void TMCJPEGDEC_IFD0_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder,
return;
}
pInfo->yResNum = readU32(p, byteOrder);
p = (const u8*)(pInfo->thumbnailData + offset + 4);
p = pInfo->thumbnailData + (offset + 4);
if (pInfo->thumbnailData > p) {
return;
}
@@ -501,6 +501,10 @@ static void TMCJPEGDEC_IFD1_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder,
case 0x8769:
case 0x9000:
case 0x9101:
case 0x011C:
case 0xA000:
case 0xA001:
case 0xA002:
case 0xA003: {
break;
}
@@ -516,7 +520,7 @@ static void TMCJPEGDEC_IFD1_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder,
return;
}
pInfo->xResNumIfd1 = readU32(p, byteOrder);
p = (const u8*)(pInfo->thumbnailData + offset + 4);
p = pInfo->thumbnailData + (offset + 4);
if (pInfo->thumbnailData > p) {
return;
}
@@ -538,7 +542,7 @@ static void TMCJPEGDEC_IFD1_tag_parse(TMCCJPEGDecExifData* pInfo, u16 byteOrder,
return;
}
pInfo->planarConfigIfd1 = readU32(p, byteOrder);
p = (const u8*)(pInfo->thumbnailData + offset + 4);
p = pInfo->thumbnailData + (offset + 4);
if (pInfo->thumbnailData > p) {
return;
}

View File

@@ -816,7 +816,6 @@ static void TMCJPEGDEC_converterYUV444toRGB565(TMCCJPEGDecWork* work, s32 x, s32
u8* luminance;
u8* cb;
u8* cr;
s32 value;
s32 column;
s32 xEnd;
s32 cbValue;
@@ -844,17 +843,17 @@ static void TMCJPEGDEC_converterYUV444toRGB565(TMCCJPEGDecWork* work, s32 x, s32
for (column = x; column < xEnd; column++) {
cbValue = (s8)*cb++;
crValue = (s8)*cr++;
value = *luminance++;
green = value + (-(cbValue * 88 + crValue * 183) >> 8);
red = value + ((crValue * 359) >> 8);
blue = value + ((cbValue * 454) >> 8);
red = *luminance++;
green = red + (-(cbValue * 88 + crValue * 183) >> 8);
blue = red + ((cbValue * 454) >> 8);
red = red + ((crValue * 359) >> 8);
if ((blue | green | red) >> 8) {
blue = blue > 255 ? 255 : blue < 0 ? 0 : blue;
green = green > 255 ? 255 : green < 0 ? 0 : green;
red = red > 255 ? 255 : red < 0 ? 0 : red;
}
output[(column & 3) + (((column >> 2) + tileRow) << 4)] =
((blue & 0xF8) >> 3) + ((red & 0xF8) << 8) + ((green & 0xFC) << 3);
((blue & 0xF8) >> 3) + (((red & 0xF8) << 8) + ((green & 0xFC) << 3));
}
luminance += rowSkip;
cb += rowSkip;

View File

@@ -656,10 +656,10 @@ static void TMCJPEGDEC_converterYUV420toRGBA8edge(TMCCJPEGDecWork* work, s32 x,
blueOffset = (cbValue * 454) >> 8;
}
{
s32 value = *luminance++;
s32 green = value + greenOffset;
s32 red = value + redOffset;
s32 blue = value + blueOffset;
s32 red = *luminance++;
s32 green = red + greenOffset;
s32 blue = red + blueOffset;
red = red + redOffset;
if ((green | red | blue) >> 8) {
blue = blue > 255 ? 255 : blue < 0 ? 0 : blue;
green = green > 255 ? 255 : green < 0 ? 0 : green;

View File

@@ -195,7 +195,7 @@ typedef struct _zi8DawgCtx {
ziU8* table; // 0x10
ziU8* p14; // 0x14
zi8DawgRec recs[0x32]; // 0x18
ziU32 unk_0x338; // 0x338
ziU32 endNode; // 0x338
} zi8DawgCtx;
typedef struct _ziUwdNode {
@@ -221,7 +221,7 @@ struct __zi8_work_data_s {
ziU8 unk_0x19;
ziU16 separator; // 0x1A
ziU8 unk_0x1C[3];
ziU8 unk_0x1F;
ziU8 ignoreCase;
ziU16 unk_0x20[0x40];
ziU16 unk_0xA0[0x42];
ziU16* unk_0x124[1];

View File

@@ -311,21 +311,21 @@ Zi8UInt Zi8SpellingPY(ziU16 *output,Zi8UInt key,ziU8 includeTone)
return length;
}
ziU8 Zi8IsMatch1Key(ziU16 *input,Zi8UInt inputLength,Zi8UInt key,ziU8 requireFull,ziU8 usePinyin,ziU8 tone)
ziU8 Zi8IsMatch1Key(ziU16 *input,Zi8UInt inputLength,ziU16 key,ziU8 requireFull,ziU8 usePinyin,ziU8 tone)
{
ziU16 spelling[16];
Zi8UInt spellingLength;
if ((key & 0xffff) == 0) {
if (key == 0) {
return 0;
}
if ((inputLength & 0xff) == 0) {
return 1;
}
if (usePinyin != 0) {
spellingLength = Zi8SpellingPY(spelling,key & 0xffff,tone & 0xff);
spellingLength = Zi8SpellingPY(spelling,key,tone & 0xff);
} else {
spellingLength = Zi8SpellingZY(spelling,key & 0xffff,tone & 0xff);
spellingLength = Zi8SpellingZY(spelling,key,tone & 0xff);
}
if ((requireFull != 0) && ((inputLength & 0xff) != (spellingLength & 0xff))) {
return 0;
@@ -1327,7 +1327,7 @@ int Zi8Get1KeyPressCandidates(Zi8OneKeyParam *params,Zi8OneKeyOptions *options,Z
}
if (elementIndex == 0) continue;
if (!emitWords) {
if (trackDuplicates && Zi8SetFindCand(params->scratch,candidateOrdinal,work) != 0) continue;
if (trackDuplicates && (ziU8)Zi8SetFindCand(params->scratch,candidateOrdinal,work) != 0) continue;
for (index = 0; index < candidateCount; index++) {
if (candidateCharacter == output[index]) break;
}
@@ -1396,7 +1396,7 @@ int Zi8Get1KeyPressCandidates(Zi8OneKeyParam *params,Zi8OneKeyOptions *options,Z
}
if (elementIndex == 0) continue;
if (!emitWords) {
if (trackDuplicates && Zi8SetFindCand(params->scratch,candidateOrdinal,work) != 0) continue;
if (trackDuplicates && (ziU8)Zi8SetFindCand(params->scratch,candidateOrdinal,work) != 0) continue;
for (index = 0; index < candidateCount; index++) {
if (candidateCharacter == output[index]) break;
}
@@ -1519,7 +1519,7 @@ int Zi8Get1KeyPressCandidates(Zi8OneKeyParam *params,Zi8OneKeyOptions *options,Z
}
if (elementIndex == 0) goto nextPhrase;
if (!emitWords) {
if (trackDuplicates && Zi8SetFindCand(params->scratch,phraseCharacter,work) != 0) goto nextPhrase;
if (trackDuplicates && (ziU8)Zi8SetFindCand(params->scratch,phraseCharacter,work) != 0) goto nextPhrase;
for (index = 0; index < candidateCount; index++) {
if (firstOrdinal == output[index]) goto nextPhrase;
}
@@ -1601,7 +1601,7 @@ finishPhrases:
}
if (elementIndex == 0) goto nextSortedOrdinal;
if (!emitWords) {
if (trackDuplicates && Zi8SetFindCand(params->scratch,ordinal,work) != 0) goto nextSortedOrdinal;
if (trackDuplicates && (ziU8)Zi8SetFindCand(params->scratch,ordinal,work) != 0) goto nextSortedOrdinal;
ordinal = Zi8Ord2Uni(ordinal,work);
for (index = 0; index < candidateCount; index++) {
if (ordinal == output[index]) break;
@@ -1629,11 +1629,11 @@ nextSortedOrdinal:
;
}
}
if ((options->countOnly == '\0') && Zi8GetZHuwdPtr(&userEntries,&userCount,work) != 0)
if ((options->countOnly == '\0') && (ziU8)Zi8GetZHuwdPtr(&userEntries,&userCount,work) != 0)
{
for (phraseOrdinal = 0; phraseOrdinal < userCount;) {
if ((params->elementCount == '\0') || ((userEntries[1] & 0x80) != 0)) {
userOrdinal = (userEntries[1] & 0x7f) << 8 | (ziU16)userEntries[2];
userOrdinal = (ziU16)(userEntries[1] & 0x7f) << 8 | (ziU16)userEntries[2];
phraseTable = syllableTable + (Zi8UInt)userOrdinal * 0xc;
if (((*phraseTable & languageMask) != 0) &&
((characterSetTable == 0 || ((characterSet & characterSetTable[(Zi8UInt)userOrdinal]) != 0)))) {
@@ -1654,7 +1654,7 @@ nextSortedOrdinal:
}
if (elementIndex == 0) goto nextUserOrdinal;
if (!emitWords) {
if (trackDuplicates && Zi8SetFindCand(params->scratch,userOrdinal,work) != 0) goto nextUserOrdinal;
if (trackDuplicates && (ziU8)Zi8SetFindCand(params->scratch,userOrdinal,work) != 0) goto nextUserOrdinal;
userCharacter = Zi8Ord2Uni(userOrdinal,work);
for (duplicateIndex = 0; duplicateIndex < candidateCount; duplicateIndex++) {
if (userCharacter == output[duplicateIndex]) break;
@@ -1710,7 +1710,7 @@ nextUserOrdinal:
if (unicodeCharacter == previousCharacter) goto nextOrdinal;
previousCharacter = unicodeCharacter;
if (!emitWords) {
if (trackDuplicates && Zi8SetFindCand(params->scratch,ordinal,work) != 0) goto nextOrdinal;
if (trackDuplicates && (ziU8)Zi8SetFindCand(params->scratch,ordinal,work) != 0) goto nextOrdinal;
for (index = 0; index < candidateCount; index++) {
if (unicodeCharacter == output[index]) break;
}
@@ -1738,8 +1738,7 @@ nextOrdinal:
}
finishCandidates:
if ((emitWords) && (outputIndex != 0)) {
output[outputIndex] = 0;
outputIndex++;
output[outputIndex++] = 0;
}
params->count = candidateCount;
Zi8LogError(100,work);

View File

@@ -920,8 +920,7 @@ selectDictionary:
if ((((wordLength == 0) && (*punctuationCursor != 0)) && (dictionaryExact != 0)) &&
((((ZiAlphaWork*)workData)->singleCharacter != 0 && (parameters->elementCount == 2)))) {
*wordCursor = ((ZiAlphaWork*)workData)->singleCharacter;
wordCursor[1] = *punctuationCursor;
punctuationCursor = punctuationCursor + 1;
wordCursor[1] = *punctuationCursor++;
wordCursor[2] = 0;
wordLength = 2;
}
@@ -1068,7 +1067,7 @@ filterVowelCandidate:
(((ZiAlphaWork*)workData)->requiredLength != parameters->elementCount)) &&
(((((ZiAlphaWork*)workData)->prefixLength == '\0' ||
(((ZiAlphaWork*)workData)->prefixLength + 1 >= parameters->elementCount)) && (dictionaryExact != 0)))
))) && ((((completionAllowed && (prefixCount == 0)) && (dictionaryKind != 0xb)) &&
))) && ((((currentVowelRestriction && (prefixCount == 0)) && (dictionaryKind != 0xb)) &&
((((int)dictionaryIndex < (int)(unsigned int)((ZiAlphaWork*)workData)->dictionaryCount &&
(((ZiAlphaOptions*)optionData)->lookupMode == '\0')) &&
(((ZiAlphaWork*)workData)->dictionaries[language] == 0)))))) {
@@ -1084,7 +1083,7 @@ filterVowelCandidate:
goto retryDictionary;
}
}
if ((((((dictionaryExact == 0) || (!completionAllowed)) || (prefixCount != 0)) ||
if ((((((dictionaryExact == 0) || (!currentVowelRestriction)) || (prefixCount != 0)) ||
((dictionaryKind == 0xb ||
(dictionaryIndex >= ((ZiAlphaWork*)workData)->dictionaryCount)))) ||
(elements == 0)) ||
@@ -1139,7 +1138,7 @@ prepareCandidate:
Zi8ChangeWordCase(wordCursor - prefixCount,language,workData);
}
if ((((ZiAlphaOptions*)optionData)->suffixOnly != '\0') &&
((int)((ZiAlphaOptions*)optionData)->maxWordLength < (int)(wordLength + prefixCount))) {
((wordLength + prefixCount) > ((ZiAlphaOptions*)optionData)->maxWordLength)) {
wordLength = ((ZiAlphaOptions*)optionData)->maxWordLength - prefixCount;
wordCursor[wordLength] = 0;
}
@@ -1208,16 +1207,14 @@ emitCandidate:
if (((ZiAlphaOptions*)optionData)->maxResults <= candidateCount) goto finishCandidates;
} else {
if (((ZiAlphaOptions*)optionData)->lookupMode != '\0') {
for (index = 0;
(index <= (int)wordLength &&
(wordCursor[index] == ((ZiAlphaOptions*)optionData)->dictionary[index]));
index = index + 1) {
for (index = 0; index <= wordLength; index++) {
if (wordCursor[index] != ((ZiAlphaOptions*)optionData)->dictionary[index]) break;
}
if ((((((ZiAlphaWork*)workData)->operation != '\0') || ((int)wordLength <= index))
if ((((((ZiAlphaWork*)workData)->operation != '\0') || (index >= wordLength))
&& ((((ZiAlphaWork*)workData)->operation == '\0' ||
((exactLengthOnly == 0 ||
((int)(ziU16)Zi8WCharCount(((ZiAlphaOptions*)optionData)->dictionary,workData) <= index)))))) &&
((exactLengthOnly != 0 || ((int)wordLength <= index)))) {
(index >= (int)(ziU16)Zi8WCharCount(((ZiAlphaOptions*)optionData)->dictionary,workData))))))) &&
((exactLengthOnly != 0 || (index >= wordLength)))) {
candidateCount = 1;
goto finishCandidates;
}
@@ -1367,7 +1364,7 @@ emitCandidate:
}
}
retryDictionary:;
} while (dictionaryKind != 0);
} while (dictionaryIndex != 0);
}
finishDictionaryPass:;
}
@@ -1376,27 +1373,23 @@ finishDictionaryPass:;
(((parameters->elementCount == 1 &&
(((Zi8GetTableCount(parameters->language,0x1f,workData)) & 0x40) == 0)) &&
(*parameters->elements >= 0xeff1)))))) &&
((*parameters->elements < 0xf011 &&
((*parameters->elements <= 0xf010 &&
((Zi8getKeyLayout(parameters->language,*parameters->elements,&punctuationBuffer[0],1,workData)) != '\0')))) {
punctuationCandidate = ZI8_TRUE;
}
if (((candidateCount == 0) && (punctuationCandidate)) && (firstCandidate == 0)) {
candidateCount = 1;
if ((((ZiAlphaOptions*)optionData)->countOnly == '\0') && ((parameters->getOptions & 0xfd) != 0x81)) {
*encodedOutput = 0x3e;
encodedOutput[2] = 0;
encodedOutput[1] = 0;
}
else {
if (((ZiAlphaOptions*)optionData)->countOnly != 0 || (parameters->getOptions & 0xFD) == 0x81) {
if (parameters->getMode == 1) {
*wordCursor = *elements;
} else {
*wordCursor = 0x3E;
}
else {
*wordCursor = 0x3e;
}
wordCursor[2] = 0;
wordCursor[1] = 0;
wordCursor = wordCursor + 2;
wordCursor[1] = wordCursor[2] = 0;
wordCursor += 2;
} else {
*encodedOutput = 0x3E;
encodedOutput[1] = encodedOutput[2] = 0;
}
if (((ZiAlphaOptions*)optionData)->lookupMode == '\0') {
((ZiAlphaWork*)workData)->singleCharacter = 0;
@@ -1406,13 +1399,13 @@ finishDictionaryPass:;
secondLanguagePass = ZI8_TRUE;
languagePassCount = 0;
language = parameters->subLanguage;
completionAllowed = secondaryCompletionAllowed;
currentVowelRestriction = secondaryVowelRestriction;
dictionaryStatus[9] = 0;
dictionaryStatus[5] = 0;
dictionaryStatus[6] = 0;
dictionaryStatus[7] = 0;
dictionaryStatus[8] = 0;
completionAllowed = secondaryCompletionAllowed;
currentVowelRestriction = secondaryVowelRestriction;
if (switchedLanguage) {
secondLanguagePass = ZI8_FALSE;
dictionaryStatus[1] = 0;
@@ -1426,14 +1419,15 @@ finishDictionaryPass:;
if (language == 0) {
language = parameters->language;
}
completionAllowed = secondaryCompletionAllowed;
currentVowelRestriction = secondaryVowelRestriction;
if (language == parameters->language) {
completionAllowed = primaryCompletionAllowed;
currentVowelRestriction = primaryVowelRestriction;
} else {
completionAllowed = secondaryCompletionAllowed;
currentVowelRestriction = secondaryVowelRestriction;
}
if ((((ZiAlphaOptions*)optionData)->lookupMode != '\0') || ((!currentVowelRestriction && (!completionAllowed)))) goto finishCandidates;
if ((completionAllowed) &&
if ((((ZiAlphaOptions*)optionData)->lookupMode != '\0') || ((!completionAllowed && (!currentVowelRestriction)))) goto finishCandidates;
if ((currentVowelRestriction) &&
((((((ZiAlphaWork*)workData)->suffixMode == '\0' && (2 < (int)elementCount)) &&
(elements[elementCount - 2] == 0xeff1)) &&
((int)((ZiAlphaWork*)workData)->prefixCount == elementCount - 2)))) {
@@ -1444,66 +1438,65 @@ finishDictionaryPass:;
prefixPrepared = ZI8_TRUE;
goto preparePrefix;
}
if (!((((completionAllowed) && (((ZiAlphaWork*)workData)->suffixMode != '\0')) &&
((((ZiAlphaWork*)workData)->suffixLocked == '\0' && (2 < (int)elementCount)))) &&
(((elements[elementCount - 2] == 0xeff1 &&
((int)((ZiAlphaWork*)workData)->suffixCount == elementCount - 2)) ||
(((int)((ZiAlphaWork*)workData)->suffixCount == elementCount - 1 &&
((Zi8IsAlphaPunct(elements[elementCount - 1])) != '\0'))))))) {
tryAlternatePrefix:
if ((prefixPrepared) && (1 < ((ZiAlphaWork*)workData)->alternatePrefixCount)) {
if (language == parameters->language) {
if (primaryMatched) {
((ZiAlphaWork*)workData)->alternatePrefixCount = 0;
}
}
else if (secondaryMatched) {
((ZiAlphaWork*)workData)->alternatePrefixCount = 0;
}
if (((ZiAlphaWork*)workData)->alternatePrefixCount != '\0') {
if (currentVowelRestriction && ((ZiAlphaWork*)workData)->suffixMode != 0 &&
((ZiAlphaWork*)workData)->suffixLocked == 0 && elementCount > 2 &&
((elements[elementCount - 2] == 0xEFF1 &&
((ZiAlphaWork*)workData)->suffixCount == elementCount - 2) ||
(((ZiAlphaWork*)workData)->suffixCount == elementCount - 1 &&
Zi8IsAlphaPunct(elements[elementCount - 1]) != 0))) {
if (((ZiAlphaWork*)workData)->prefixEnabled >= 1 && elements[elementCount - 1] == 0xEFF1) {
goto checkPrefixPunctuation;
}
if (!primaryMatched && !secondaryMatched) {
((ZiAlphaWork*)workData)->suffixLocked = 1;
((ZiAlphaWork*)workData)->suffixElementCount = parameters->elementCount - 1;
if (prefixCount != 0) {
wordCursor -= prefixCount;
elements -= prefixCount;
elementCount = elementCount + prefixCount;
wordCapacity = wordCapacity + prefixCount;
prefixCount = 0;
prefixPrepared = ZI8_FALSE;
((ZiAlphaWork*)workData)->suffixMode = 0;
((ZiAlphaWork*)workData)->suffixLocked = 0;
((ZiAlphaWork*)workData)->suffixCount = 0;
for (index = 0; index < (int)(unsigned int)((ZiAlphaWork*)workData)->alternatePrefixCount; index = index + 1) {
((ZiAlphaWork*)workData)->prefix[index] =
((ZiAlphaWork*)workData)->alternatePrefix[index];
}
((ZiAlphaWork*)workData)->prefixCount = (ziU8)index;
if (((ZiAlphaWork*)workData)->language == parameters->language) {
((ZiAlphaWork*)workData)->language = parameters->subLanguage;
}
else {
((ZiAlphaWork*)workData)->language = parameters->language;
}
}
goto preparePrefix;
}
}
if ((prefixPrepared) && (1 < ((ZiAlphaWork*)workData)->alternatePrefixCount)) {
if (language == parameters->language) {
if (primaryMatched) {
((ZiAlphaWork*)workData)->alternatePrefixCount = 0;
}
}
if ((((primaryMatched) && (secondaryMatched)) || (elementCount == 0)) ||
(((language == parameters->language && (primaryMatched)) || ((language == parameters->subLanguage && (secondaryMatched))))))
goto finishCandidates;
if (completionAllowed) goto checkPrefixPunctuation;
} else {
if ((((ZiAlphaWork*)workData)->prefixEnabled == '\0') || (elements[elementCount - 1] != 0xeff1)) {
if ((!primaryMatched) && (!secondaryMatched)) {
((ZiAlphaWork*)workData)->suffixLocked = 1;
((ZiAlphaWork*)workData)->suffixElementCount = parameters->elementCount - 1;
if (prefixCount != 0) {
wordCursor -= prefixCount;
elements -= prefixCount;
elementCount = elementCount + prefixCount;
wordCapacity = wordCapacity + prefixCount;
prefixCount = 0;
}
goto preparePrefix;
}
goto tryAlternatePrefix;
else if (secondaryMatched) {
((ZiAlphaWork*)workData)->alternatePrefixCount = 0;
}
if (((ZiAlphaWork*)workData)->alternatePrefixCount != '\0') {
wordCursor -= prefixCount;
elements -= prefixCount;
elementCount = elementCount + prefixCount;
wordCapacity = wordCapacity + prefixCount;
prefixCount = 0;
prefixPrepared = ZI8_FALSE;
((ZiAlphaWork*)workData)->suffixMode = 0;
((ZiAlphaWork*)workData)->suffixLocked = 0;
((ZiAlphaWork*)workData)->suffixCount = 0;
for (index = 0; index < (int)(unsigned int)((ZiAlphaWork*)workData)->alternatePrefixCount; index = index + 1) {
((ZiAlphaWork*)workData)->prefix[index] =
((ZiAlphaWork*)workData)->alternatePrefix[index];
}
((ZiAlphaWork*)workData)->prefixCount = (ziU8)index;
if (((ZiAlphaWork*)workData)->language == parameters->language) {
((ZiAlphaWork*)workData)->language = parameters->subLanguage;
}
else {
((ZiAlphaWork*)workData)->language = parameters->language;
}
((ZiAlphaWork*)workData)->alternatePrefixCount = 0;
}
}
if ((((primaryMatched) && (secondaryMatched)) || (elementCount == 0)) ||
(((language == parameters->language && (primaryMatched)) || ((language == parameters->subLanguage && (secondaryMatched))))))
goto finishCandidates;
if (currentVowelRestriction) {
checkPrefixPunctuation:
if ((1 < ((ZiAlphaWork*)workData)->prefixEnabled) &&
(parameters->elements[parameters->elementCount - 1] != 0xEFF1))
@@ -1576,17 +1569,17 @@ checkPrefixFields:
finishCandidates:
((ZiAlphaWork*)workData)->usePrefixAsElements = 0;
if (((ZiAlphaOptions*)optionData)->countOnly == '\0') {
parameters->letters = candidateCount;
if (((ZiAlphaOptions*)optionData)->countOnly != '\0') {
parameters->letters = 0;
}
else {
parameters->letters = 0;
parameters->letters = candidateCount;
}
wordCursor[-prefixCount] = 0;
if (((((ZiAlphaOptions*)optionData)->countOnly == '\0') && (((ZiAlphaOptions*)optionData)->lookupMode == '\0')) && (parameters->elementCount != 0)) {
elementIndex = parameters->elementCount - 1;
if (((primaryVowelRestriction) || (secondaryVowelRestriction)) &&
((((ZiAlphaWork*)workData)->prefixEnabled < 2 ||
((((ZiAlphaWork*)workData)->prefixEnabled <= 1 ||
(parameters->elements[elementIndex] == 0xEFF1)))) {
index = 0;
while (elementIndex >= 0) {

View File

@@ -32,7 +32,7 @@ ziU32 Zi8MatchROMdata0(ziWChar* elements, ziU8 count, ziU8 language,
if (language != context->lang || graphTable != context->key) records[0].key = 0;
if (count == 1 && mode != 0 && graphTable == 0xe) goto finish_graph;
if (status != 0) {
if (records[context->cnt - 1].node == (ziU8*)context->unk_0x338) goto finish_graph;
if (records[context->cnt - 1].node == (ziU8*)context->endNode) goto finish_graph;
for (index = 0; index < count; index++) {
if (records[index].key != elements[index] &&
(ziU16)records[index].attr != elements[index]) goto finish_graph;
@@ -82,7 +82,7 @@ check_keys:
records[0].node = ZiDAWGGetGraphInfo(context, keyData, (ziS16*)keys);
} else {
records[0].node = ZiDAWGGetGraph(context);
context->unk_0x338 = 0;
context->endNode = 0;
context->p14 = 0;
}
records[0].key = 0;
@@ -98,7 +98,7 @@ check_keys:
}
searching = 1;
current = &records[context->cnt - 1];
while (searching && context->cnt != 0 && current->node != (ziU8*)context->unk_0x338) {
while (searching && context->cnt != 0 && current->node != (ziU8*)context->endNode) {
if (capacity < context->cnt) goto finish_graph;
current->attr = ZiDAWGgetCHARattribute(context, current->node, ZI_WORK);
if (context->cnt <= count) {

View File

@@ -126,7 +126,7 @@ matchText:
if (pattern[index] != Zi8ConvertUC2Key(word[index], language, __zi8_work_data)) goto next;
} else {
folded = Zi8ConvertUC2WC(word[index], language, __zi8_work_data);
if (folded != pattern[index] && (!ZI_WORK->unk_0x1F || !Zi8ChangeCharCase(0, &folded, language, __zi8_work_data) || folded != pattern[index])) goto next;
if (folded != pattern[index] && (!ZI_WORK->ignoreCase || !Zi8ChangeCharCase(0, &folded, language, __zi8_work_data) || folded != pattern[index])) goto next;
}
}
matched:

View File

@@ -148,7 +148,7 @@ ziU8 Zi8MatchUWDdata(ziWChar* pattern, ziU8 length, ziWChar* currentWord, ziU16
folded = currentWord[position];
ZI_WORK->unk_0x141[previousOffset] = Zi8ConvertWC2UC(folded, language, __zi8_work_data);
ZI_WORK->unk_0x184[previousOffset] = 0;
if (ZI_WORK->unk_0x1F != 0) {
if (ZI_WORK->ignoreCase != 0) {
if (!Zi8ChangeCharCase(1, &folded, language, __zi8_work_data)) Zi8ChangeCharCase(0, &folded, language, __zi8_work_data);
ZI_WORK->unk_0x184[previousOffset] = Zi8ConvertWC2UC(folded, language, __zi8_work_data);
}
@@ -192,7 +192,7 @@ next:
if (pattern[position] != Zi8ConvertUC2Key(text[position], language, __zi8_work_data)) goto next;
} else {
folded = Zi8ConvertUC2WC(text[position], language, __zi8_work_data);
if (folded != pattern[position] && (!ZI_WORK->unk_0x1F || !Zi8ChangeCharCase(0, &folded, language, __zi8_work_data) || folded != pattern[position])) goto next;
if (folded != pattern[position] && (!ZI_WORK->ignoreCase || !Zi8ChangeCharCase(0, &folded, language, __zi8_work_data) || folded != pattern[position])) goto next;
}
}
if (ZI_WORK->unk_0x140 != 0) {

View File

@@ -199,10 +199,10 @@ checkEntry:
}
result = graph + (ziU32)entry[4] * 0x10000 + (((ziU16)entry[5] << 8) + entry[6]);
((zi8DawgCtx*)context)->unk_0x338 =
((zi8DawgCtx*)context)->endNode =
graph + (ziU32)entry[7] * 0x10000 + (((ziU16)entry[8] << 8) + entry[9]);
if (((zi8DawgCtx*)context)->unk_0x338 == graph) {
((zi8DawgCtx*)context)->unk_0x338 = 0;
if (((zi8DawgCtx*)context)->endNode == graph) {
((zi8DawgCtx*)context)->endNode = 0;
}
if ((((ziU16)entry[0] << 8) + entry[1]) == 0) {

View File

@@ -45,7 +45,7 @@ posEntry:
if (pattern[position] == Zi8ConvertWC2Key(word[position], language, __zi8_work_data)) continue;
break;
} else if (word[position] != pattern[position]) {
if (!ZI_WORK->unk_0x1F || language == 1) break;
if (!ZI_WORK->ignoreCase || language == 1) break;
folded = pattern[position];
if (!Zi8ChangeCharCase(1, &folded, language, __zi8_work_data) || folded != word[position]) break;
}

View File

@@ -152,10 +152,10 @@ u32 __SCGetConfBufSize();
typedef struct SC_BT_DEV_INFO {
u8 devName[20]; // 0x00
u8 unk_0x14;
u8 unk_0x15[0xB];
u8 pad_0x14; // 0x14
u8 pad_0x15[0xB]; // 0x15
LINK_KEY linkKey; // 0x20
u8 unk_0x30[0x10];
u8 pad_0x30[0x10]; // 0x30
} SC_BT_DEV_INFO;
typedef struct SCBtDeviceInfo {

View File

@@ -26,7 +26,7 @@ typedef struct USBKBDCmdLED {
typedef struct USBKBDCmdLEDAsync {
USBKBDCmdLED base; // 0x00
u8 unk_0x08[0x10];
u8 pad_0x08[0x10]; // 0x08
USBKBDCmdLEDCallback cb; // 0x18
void* cbArg; // 0x1C
} USBKBDCmdLEDAsync;

View File

@@ -749,7 +749,8 @@ s32 pfd_sddrv_get_total_sectors(u32* sectors, u16* bytes_per_sector) {
minimum_multiplier = multiplier >= 9 ? multiplier : 9;
max_multiplier = minimum_multiplier <= 11 ? minimum_multiplier : 11;
multiplier_factor = 1;
multiplier_factor <<= (u16)(max_multiplier - 9);
max_multiplier = (u16)(max_multiplier - 9);
multiplier_factor = (u16)(multiplier_factor << max_multiplier);
*sectors = cluster_blocks * multiplier_factor;
} else {
*sectors = ((csd[1] >> 8 & 0x3fffff) + 1) * 0x400;

View File

@@ -1,11 +1,15 @@
#include <private/fa/pdm.h>
#include <revolution/types.h>
#define MBR_WORD(buf, offset) (((pf_u32)(buf)[(offset) + 3] << 24) + ((pf_u32)(buf)[(offset) + 2] << 16) + ((pf_u32)(buf)[(offset) + 1] << 8) + (buf)[offset])
static inline pf_u16 read_boot_u16(pf_u8* buf, pf_u32 offset) {
if ((pf_u32)&buf[offset] & 1) { return (buf[offset + 1] << 8) | buf[offset]; }
return PF_SWAP_16(*(pf_u16*)&buf[(offset + 1) & ~1]);
}
static inline pf_u32 read_partition_u32(const pf_u8* buf, pf_u32 offset) {
return (((pf_u32)buf[offset + 2] << 16) + buf[offset]) +
(((pf_u32)buf[offset + 3] << 24) + ((pf_u32)buf[offset + 1] << 8));
}
pf_s32 pdm_part_is_master_boot_sector(pf_u8* buf, pf_u32 total, pf_bool* p_is_mbr) {
pf_s16 index;
pf_u32 start[4];
@@ -18,8 +22,8 @@ pf_s32 pdm_part_is_master_boot_sector(pf_u8* buf, pf_u32 total, pf_bool* p_is_mb
p_count = count;
for (index = 0; index < 4; index++) {
*p_start = 0;
*p_start = (((pf_u32)buf[456] << 16) + buf[454]) + (((pf_u32)buf[457] << 24) + ((pf_u32)buf[455] << 8));
*p_count = (((pf_u32)buf[460] << 16) + buf[458]) + (((pf_u32)buf[461] << 24) + ((pf_u32)buf[459] << 8));
*p_start = read_partition_u32(buf, 454);
*p_count = read_partition_u32(buf, 458);
if (*p_start != 0 && *p_count != 0) {
if (index == 0) { *p_is_mbr = 1; }
} else {
@@ -91,14 +95,14 @@ pf_s32 pdm_part_get_start_sector(PDM_PARTITION* p_part) {
if (err != 0) { return err; }
pdm_part_is_master_boot_sector(buf, p_part->p_disk->disk_info.total_sectors, &is_mbr);
if (is_mbr) {
start[0] = MBR_WORD(buf, 454);
start[1] = MBR_WORD(buf, 470);
start[2] = MBR_WORD(buf, 486);
start[3] = MBR_WORD(buf, 502);
count[0] = MBR_WORD(buf, 458);
count[1] = MBR_WORD(buf, 474);
count[2] = MBR_WORD(buf, 490);
count[3] = MBR_WORD(buf, 506);
start[0] = read_partition_u32(buf, 454);
start[1] = read_partition_u32(buf, 470);
start[2] = read_partition_u32(buf, 486);
start[3] = read_partition_u32(buf, 502);
count[0] = read_partition_u32(buf, 458);
count[1] = read_partition_u32(buf, 474);
count[2] = read_partition_u32(buf, 490);
count[3] = read_partition_u32(buf, 506);
part_index = 0;
for (index = 0; index < 4; index++) {
sector = start[index];
@@ -112,15 +116,15 @@ pf_s32 pdm_part_get_start_sector(PDM_PARTITION* p_part) {
sector = base;
for (extended_index = 0; extended_index < 2; extended_index++) {
if (extended_index == 0) {
sector = sector + MBR_WORD(buf, 454);
length = MBR_WORD(buf, 458);
sector = sector + read_partition_u32(buf, 454);
length = read_partition_u32(buf, 458);
err = pdm_disk_physical_read(p_part->p_disk, buf, sector, 1, 512, &success);
if (err != 0) { return err; }
if (requested == part_index) { p_part->start_sector = sector; p_part->total_sector = length; return 0; }
part_index++;
} else {
sector = base + MBR_WORD(buf, 470);
length = MBR_WORD(buf, 474);
sector = base + read_partition_u32(buf, 470);
length = read_partition_u32(buf, 474);
if (sector == 0 || length == 0) { break; }
err = pdm_disk_physical_read(p_part->p_disk, buf, sector, 1, 512, &success);
if (err != 0) { return err; }

View File

@@ -557,15 +557,11 @@ static void kbdSendKey(KBDKeyEventData* event) {
}
}
static inline u32 kbdChannelFlags(u32 channel) {
return kbdData[channel].flags;
}
static void kbd_led_handler(BOOL success, void* callbackArg) {
const KBDLEDCallbackData* callback;
u32 index;
s32 index;
u32 err;
index = (u32)callbackArg;
index = (s32)callbackArg;
callback = &kbdLCBuf[index];
kbdCmdBuf[index].device = 0;
if (callback->callbackAddress == (USBKBDCmdLEDCallback)(u32)kbdCmdBuf[index].device) {
@@ -579,12 +575,11 @@ static void kbd_led_handler(BOOL success, void* callbackArg) {
err = 0;
break;
}
kbdLCBuf[index].callbackAddress(err, kbdLCBuf[index].callbackArg);
kbdLCBuf[(u32)callbackArg].callbackAddress(err, kbdLCBuf[(u32)callbackArg].callbackArg);
}
USBKBDErr KBDSetLedsAsync(u32 channel, u32 leds, USBKBDCmdLEDCallback callback, void* callbackArg) {
s32 index;
u32 ofs;
u32 index;
u8 ledBits;
BOOL interrupts;
USBKBDErr result;
@@ -594,14 +589,18 @@ USBKBDErr KBDSetLedsAsync(u32 channel, u32 leds, USBKBDCmdLEDCallback callback,
if (channel >= 4) {
return 4;
}
if ((s32)kbdChannelFlags(channel) == 1 || (s32)kbdChannelFlags(channel) == 4) {
return 5;
{
KBDChannel* data = &kbdData[(s32)channel];
if ((s32)data->flags == 1 || (s32)data->flags == 4) {
return 5;
}
}
ledBits = leds;
interrupts = OSDisableInterrupts();
for (index = 0, ofs = 0; index < 12; index++, ofs += sizeof(KBDLEDCommand)) {
if (*(u32*)&((u8*)kbdCmdBuf)[ofs] == 0) {
*(u32*)&((u8*)kbdCmdBuf)[ofs] = (u32)kbdData[channel].device;
for (index = 0; index < 12; index++) {
if (kbdCmdBuf[(s32)index].device == 0) {
KBDChannel* data = &kbdData[channel];
kbdCmdBuf[(s32)index].device = (u32)data->device;
break;
}
}
@@ -609,11 +608,14 @@ USBKBDErr KBDSetLedsAsync(u32 channel, u32 leds, USBKBDCmdLEDCallback callback,
if (index == 12) {
return 7;
}
kbdLCBuf[index].callbackAddress = callback;
kbdLCBuf[index].callbackArg = callbackArg;
result = USBKBDSetLEDAsync((u32)kbdData[channel].device, ledBits,
(USBKBDCmdLEDAsync*)&kbdCmdBuf[index],
kbd_led_handler, (void*)index);
{
KBDChannel* data = &kbdData[channel];
kbdLCBuf[index].callbackAddress = callback;
kbdLCBuf[index].callbackArg = callbackArg;
result = USBKBDSetLEDAsync((u32)data->device, ledBits,
(USBKBDCmdLEDAsync*)&kbdCmdBuf[index],
kbd_led_handler, (void*)index);
}
switch (result) {
default:
return 7;
@@ -624,7 +626,6 @@ USBKBDErr KBDSetLedsAsync(u32 channel, u32 leds, USBKBDCmdLEDCallback callback,
USBKBDErr KBDSetLeds(u32 channel, u32 leds) {
u32 index;
KBDLEDCommand* command;
BOOL interrupts;
USBKBDErr result;
u8 ledBits;
@@ -634,20 +635,16 @@ USBKBDErr KBDSetLeds(u32 channel, u32 leds) {
if (channel >= 4) {
return 4;
}
if ((s32)kbdChannelFlags(channel) == 1 || (s32)kbdChannelFlags(channel) == 4) {
if ((s32)kbdData[(s32)channel].flags == 1 || (s32)kbdData[(s32)channel].flags == 4) {
return 5;
}
ledBits = leds & 0xff;
interrupts = OSDisableInterrupts();
index = 0;
command = kbdCmdBuf;
while (index < 12) {
if (command->device == 0) {
command->device = (u32)kbdData[channel].device;
for (index = 0; index < 12; index++) {
if (kbdCmdBuf[(s32)index].device == 0) {
kbdCmdBuf[(s32)index].device = (u32)kbdData[channel].device;
break;
}
index++;
command++;
}
OSRestoreInterrupts(interrupts);
if (index == 12) {
@@ -655,7 +652,7 @@ USBKBDErr KBDSetLeds(u32 channel, u32 leds) {
}
result = USBKBDSetLED((u32)kbdData[channel].device, ledBits,
(USBKBDCmdLED*)&kbdCmdBuf[index]);
kbdCmdBuf[index].device = 0;
kbdCmdBuf[(s32)index].device = 0;
switch (result) {
default:
return 7;

View File

@@ -353,38 +353,41 @@ static void reset_kpad(KPADInside* kpad) {
}
void KPADGetProjectionPos(Vec2* dest, Vec2* src, const Rect* rect, f32 scale) {
f64 scaled;
f32 height = rect->bottom - rect->top;
f32 halfHeight = height * 0.5f;
f32 x = src->x * halfHeight;
f32 y = src->y * halfHeight;
x *= 1.2f;
y *= 1.2f;
scaled = 0.908 * scale;
dest->y = y;
dest->x = (f32)(x * scaled);
f32 aspect;
f32 halfHeight;
f32 half;
f32 x;
f32 height;
height = rect->bottom - rect->top;
half = 0.5f;
halfHeight = height * half;
x = src->x * halfHeight;
aspect = 1.2f;
x = aspect * x;
dest->y = aspect * (src->y * halfHeight);
dest->x = (f32)(x * (0.908 * scale));
}
void KPADSetSensorHeight(s32 chan, f32 sensorHeight) {
KPADInside* kpad;
f32 halfHeightSquared;
f32 halfHeight;
f32 barOffsetX;
f32 barDiagonal;
f32 halfWidthSquared;
f32 halfWidth;
f32 negativeSensorHeight;
f32 barDiagonalSquared;
f32 halfHeightSquared;
f32 zero;
halfWidth = 1.0f;
kpad = &inside_kpads[chan];
kpad = chan + inside_kpads;
halfHeight = 0.75f;
negativeSensorHeight = -sensorHeight;
halfWidthSquared = halfWidth * halfWidth;
halfHeightSquared = halfHeight * halfHeight;
kpad->sensorBarCenter.x = 0.0f;
zero = 0.0f;
kpad->sensorBarCenter.x = zero;
kpad->sensorBarCenter.y = negativeSensorHeight;
barDiagonalSquared = halfWidthSquared + halfHeightSquared;
barDiagonal = (f32)sqrt(barDiagonalSquared);
barDiagonal = (f32)sqrt(halfWidthSquared + halfHeightSquared);
barOffsetX = kpad->sensorBarCenter.x;
if (barOffsetX < 0.0f) {
halfWidth += barOffsetX;
@@ -465,29 +468,24 @@ static void calc_button_repeat(KPADInside* kpad, u8 extension, s32 elapsed) {
}
static void calc_acc_horizon(KPADInside* kpad) {
f32 magnitude = (f32)sqrt(kpad->acceleration.x * kpad->acceleration.x + kpad->acceleration.y * kpad->acceleration.y);
f32 magnitude;
f32 normalizedX;
f32 normalizedY;
f32 targetY;
f32 targetX;
f32 productX;
f32 oldX;
f32 oldY;
f32 blend;
f32 projectedX;
f32 smoothing;
f32 deltaX;
f32 nextX;
f32 nextY;
f32 normalized;
f32 unitX;
f32 oldCircleX;
f32 oldCircleY;
f32 unitY;
f32 circleX;
f32 oldCircleX;
f32 unitY;
f32 unitX;
f32 deltaCircleX;
f32 circleY;
f32 deltaCircleY;
magnitude = (f32)sqrt(kpad->acceleration.x * kpad->acceleration.x + kpad->acceleration.y * kpad->acceleration.y);
if (magnitude == 0.0f || magnitude >= 2.0f) {
return;
}
@@ -497,17 +495,12 @@ static void calc_acc_horizon(KPADInside* kpad) {
if (magnitude > 1.0f) {
magnitude = 2.0f - magnitude;
}
targetY = kpad->accelNormal.y;
targetX = kpad->accelNormal.x;
productX = targetX * normalizedX;
oldX = kpad->horizonAxis.x;
oldY = kpad->horizonAxis.y;
blend = magnitude * kp_acc_horizon_pw;
projectedX = productX + targetY * normalizedY;
smoothing = magnitude * blend;
deltaX = smoothing * (projectedX - oldX);
nextX = oldX + deltaX;
nextY = oldY + smoothing * (((targetY * normalizedX) - (targetX * normalizedY)) - oldY);
projectedX = kpad->accelNormal.x * normalizedX + kpad->accelNormal.y * normalizedY;
magnitude *= blend;
deltaX = magnitude * (projectedX - kpad->horizonAxis.x);
nextX = kpad->horizonAxis.x + deltaX;
nextY = kpad->horizonAxis.y + magnitude * (((kpad->accelNormal.y * normalizedX) - (kpad->accelNormal.x * normalizedY)) - kpad->horizonAxis.y);
normalized = (f32)sqrt(nextX * nextX + nextY * nextY);
if (normalized != 0.0f) {
unitX = nextX / normalized;
@@ -788,24 +781,18 @@ nextObject:
static s8 select_1obj_first(KPADInside* kpad) {
KPADDPDObject* object = kpad->dpdState.objects;
KPADDPDObject* end = kpad->dpdState.candidates;
f32 scale = kpad->dpdObjectScale;
f32 productX = kpad->horizonTangent.x * kpad->horizonAxis.x;
f32 productY = kpad->horizonTangent.y * kpad->horizonAxis.x;
f32 offsetX = productX + kpad->horizonTangent.y * kpad->horizonAxis.y;
f32 offsetY = productY - kpad->horizonTangent.x * kpad->horizonAxis.y;
offsetX *= scale;
offsetY *= scale;
f32 offsetX = kpad->horizonTangent.x * kpad->horizonAxis.x + kpad->horizonTangent.y * kpad->horizonAxis.y;
f32 offsetY = kpad->horizonTangent.y * kpad->horizonAxis.x - kpad->horizonTangent.x * kpad->horizonAxis.y;
offsetX *= kpad->dpdObjectScale;
offsetY *= kpad->dpdObjectScale;
do {
if ((s8)object->metadata.bytes.flags == 0) {
f32 x = object->x;
f32 y = object->y;
Vec2 leftPosition;
Vec2 rightPosition;
leftPosition.x = x - offsetX;
leftPosition.y = y - offsetY;
rightPosition.x = x + offsetX;
rightPosition.y = y + offsetY;
leftPosition.x = object->x - offsetX;
leftPosition.y = object->y - offsetY;
rightPosition.x = object->x + offsetX;
rightPosition.y = object->y + offsetY;
if (leftPosition.x <= kpad->value4F4 || leftPosition.x >= kpad->value4FC || leftPosition.y <= kpad->value4F8 || leftPosition.y >= kpad->value500) {
if (rightPosition.x > kpad->value4F4 && rightPosition.x < kpad->value4FC && rightPosition.y > kpad->value4F8 && rightPosition.y < kpad->value500) {
kpad->dpdState.candidates[1] = *object;
@@ -822,7 +809,7 @@ static s8 select_1obj_first(KPADInside* kpad) {
return -1;
}
}
} while (++object < end);
} while (++object < kpad->dpdState.candidates);
return 0;
}

View File

@@ -580,7 +580,7 @@ void VIInit(void)
u16 displayConfig;
u32 __VIDVDStopFlag_Enable_old;
s32 format;
u32 value;
u16 value;
u32 tv;
u32 bootromTv;

View File

@@ -393,6 +393,67 @@ CDBErr CDBRecordGetDataSize(CDBRecord* record, u32* size) {
return err;
}
CDBErr CDBRecordReduceFileSize(CDBRecord* record, u32 newFileSize) {
CDBRecordFile* recordFile;
CDBErr err;
u32 fileSize;
recordFile = record->file;
if (recordFile == NULL) {
CDBReportError("can't reduce file size of the record ; the record is closed\n");
return CDB_ERROR_27;
}
if ((recordFile->allocFlag & 2) == 0) {
CDBReportError("can't reduce file size of the record ; the record is opened as READONLY\n");
return CDB_ERROR_27;
}
CDBAttrGetFileSize(&recordFile->attr, &fileSize);
if (newFileSize > fileSize) {
CDBReportError("can't reduce file size of the record ; file size must be over %d bytes\n", fileSize);
return CDB_ERROR_26;
}
CDBAttrSetFileSize(&recordFile->attr, newFileSize);
recordFile->attr.dirty = TRUE;
err = CDBRecordFileWriteAttrBuf(record);
if (err != CDB_ERROR_OK) {
return err;
}
return CDB_ERROR_OK;
}
CDBErr CDBRecordReduceDataSize(CDBRecord* record, u32 newDataSize) {
CDBRecordFile* recordFile;
CDBErr err = CDB_ERROR_OK;
u32 dataSize;
recordFile = record->file;
if (recordFile == NULL) {
CDBReportError("can't reduce data size of the record ; the record is closed\n");
return CDB_ERROR_27;
}
if ((recordFile->allocFlag & 2) == 0) {
CDBReportError("can't reduce data size of the record ; the record is opened as READONLY\n");
return CDB_ERROR_27;
}
if (!CDBRecordBelongedDBOpenedAsRW(record)) {
CDBReportError("can't reduce data size of the record; permission denied\n");
return CDB_ERROR_ACCESS_DENIED;
}
err = CDBRecordFileGetDataSize(record, &dataSize);
if (err != CDB_ERROR_OK) {
return err;
}
if (newDataSize < dataSize) {
CDBAttrSetFileSize(&recordFile->attr, newDataSize);
recordFile->attr.dirty = TRUE;
}
err = CDBRecordFileWriteAttrBuf(record);
if (err != CDB_ERROR_OK) {
return err;
}
return CDB_ERROR_OK;
}
CDBErr CDBRecordRemove(CDBRecord* record) {
CDBErr err;
CDBLock();
@@ -430,6 +491,50 @@ void CDBRecordGetTypeForce(CDBRecord* record, char* type) {
CDBUnlock();
}
CDBErr CDBRecordSetFileType(CDBRecord* record, const char* fileType) {
CDBRecordDatabaseState* instance;
CDBRecordFile* recordFile;
CDBErr err;
instance = ((CDBDatabase*)record->database)->instance;
recordFile = record->file;
if ((instance->openFlags & 2) == 0) {
CDBReportError("can't set file type of the record; the database is opened as READONLY\n");
return CDB_ERROR_26;
}
if (recordFile == NULL) {
CDBReportError("can't set file type of the record ; the record is closed\n");
return CDB_ERROR_27;
}
if ((recordFile->allocFlag & 2) == 0) {
CDBReportError("can't set file type of the record ; the record is opened as READONLY\n");
return CDB_ERROR_27;
}
if (record->key.location != CDB_FS_LOCATION_NAND) {
CDBReportError("can't set file type of the record ; the record must exsist on NAND\n");
return CDB_ERROR_26;
}
CDBLock();
strcpy(CDBKeyStrType(record->key.keyString), fileType);
CDBAttrSetKeyStr(&recordFile->attr, &record->key);
err = CDBRecordFileWriteAttrBuf(record);
CDBUnlock();
return err;
}
CDBErr CDBRecordGetFileType(CDBRecord* record, char* fileType) {
CDBErr err = CDB_ERROR_OK;
CDBLock();
if (record->file == NULL) {
CDBReportError("can't get file type of the record ; the record is closed\n");
err = CDB_ERROR_27;
} else {
CDBConvKeyStrToType(record->key.keyString, fileType);
}
CDBUnlock();
return err;
}
void CDBRecordGetGameCodeForce(CDBRecord* record, char* gameCode) {
u32 value;
CDBLock();
@@ -438,12 +543,122 @@ void CDBRecordGetGameCodeForce(CDBRecord* record, char* gameCode) {
CDBUnlock();
}
CDBErr CDBRecordGetGameCode(CDBRecord* record, u32* gameCode) {
CDBErr err = CDB_ERROR_OK;
CDBLock();
if (record->file == NULL) {
CDBReportError("can't get game code of the record ; the record is closed\n");
err = CDB_ERROR_27;
} else {
CDBConvKeyStrToGameCode(record->key.keyString, gameCode);
}
CDBUnlock();
return err;
}
CDBErr CDBRecordGetMakerCode(CDBRecord* record, u32* makerCode) {
CDBErr err = CDB_ERROR_OK;
CDBLock();
if (record->file == NULL) {
CDBReportError("can't get maker code of the record ; the record is closed\n");
err = CDB_ERROR_27;
} else {
CDBConvKeyStrToMakerCode(record->key.keyString, makerCode);
}
CDBUnlock();
return err;
}
CDBErr CDBRecordGetModifiedTime(CDBRecord* record, u32* modifiedTime) {
CDBErr err = CDB_ERROR_OK;
CDBLock();
if (record->file == NULL) {
CDBReportError("can't get modified time of the record; the record is closed\n");
err = CDB_ERROR_27;
} else {
*modifiedTime = ((CDBRecordFile*)record->file)->attr.buf.lastModifiedDate;
}
CDBUnlock();
return err;
}
CDBErr CDBRecordGetModifiedCount(CDBRecord* record, u32* modifiedCount) {
CDBErr err = CDB_ERROR_OK;
CDBLock();
if (record->file == NULL) {
CDBReportError("can't get modified count of the record; the record is closed\n");
err = CDB_ERROR_27;
} else {
CDBAttrGetModifiedCount(&((CDBRecordFile*)record->file)->attr, modifiedCount);
}
CDBUnlock();
return err;
}
void CDBRecordGetKeyForce(CDBRecord* record, CDBRecordKey* key) {
CDBLock();
CDBRecordKeyCopy(key, &record->key);
CDBUnlock();
}
CDBErr CDBRecordSetName(CDBRecord* record, const char* name, const char* keyword) {
CDBRecordDatabaseState* instance;
CDBRecordFile* recordFile;
CDBErr err;
instance = ((CDBDatabase*)record->database)->instance;
recordFile = record->file;
if ((instance->openFlags & 2) == 0) {
CDBReportError("can't set a name of the record; the database is opened as READONLY\n");
return CDB_ERROR_26;
}
if (!CDBRecordBelongedDBOpenedAsRW(record)) {
CDBReportError("can't add keyword to the record; permission denied\n");
return CDB_ERROR_ACCESS_DENIED;
}
if (recordFile == NULL) {
CDBReportError("can't set a name of the record; the record is closed\n");
return CDB_ERROR_27;
}
if ((recordFile->allocFlag & 2) == 0) {
CDBReportError("can't set a name of the record; the record is opened as READONLY\n");
return CDB_ERROR_27;
}
CDBLock();
strncpy(recordFile->attr.buf.desc, name, CDB_ATTR_BUF_MAX_DESC_LEN);
recordFile->attr.buf.descLength = strlen(name);
recordFile->attr.dirty = TRUE;
err = CDBRecordFileWriteAttrBuf(record);
CDBUnlock();
return err;
}
CDBErr CDBRecordGetKeyword(CDBRecord* record, char* keyword) {
CDBErr err = CDB_ERROR_OK;
CDBLock();
if (record->file == NULL) {
CDBReportError("can't get keyword from the record; the record is closed\n");
err = CDB_ERROR_27;
} else {
strcpy(keyword, ((CDBRecordFile*)record->file)->attr.buf.desc);
}
CDBUnlock();
return err;
}
CDBErr CDBRecordGetKey(CDBRecord* record, CDBRecordKey* key) {
CDBErr err = CDB_ERROR_OK;
CDBLock();
if (record->file == NULL) {
CDBReportError("can't get key from the record; the record is closed\n");
err = CDB_ERROR_27;
} else {
CDBRecordKeyCopy(key, &record->key);
}
CDBUnlock();
return err;
}
void CDBRecordGetCalendarTimeForce(CDBRecord* record, int* year, int* month, int* day, int* hour, int* min, int* sec) {
char epochString[64];
CDBDate epoch;
@@ -454,6 +669,23 @@ void CDBRecordGetCalendarTimeForce(CDBRecord* record, int* year, int* month, int
CDBUnlock();
}
CDBErr CDBRecordGetCalendarTime(CDBRecord* record, int* year, int* month, int* day, int* hour, int* min, int* sec) {
char epochString[64];
CDBDate epoch;
CDBErr err = CDB_ERROR_OK;
CDBLock();
if (record->file == NULL) {
CDBReportError("can't get calender time from the record; the record is closed\n");
err = CDB_ERROR_27;
} else {
CDBConvKeyStrToEpochStr(record->key.keyString, epochString);
CDBConvEpochStrToEpochValue(epochString, &epoch);
CDBConvEpochValueToDate(epoch, year, month, day, hour, min, sec);
}
CDBUnlock();
return err;
}
CDBErr CDBRecordGetId(CDBRecord* record, CDBId* id) {
CDBErr err = CDB_ERROR_OK;
CDBRecordFile* recordFile;
@@ -532,6 +764,27 @@ CDBErr CDBRecordUpdateModifiedDate(CDBRecord* record) {
return CDB_ERROR_OK;
}
CDBErr CDBRecordDuplicate(CDBRecord* record, CDBRecord* newRecord) {
CDBRecordFile* recordFile;
CDBErr err;
if (record->key.location != CDB_FS_LOCATION_NAND && record->key.location != CDB_FS_LOCATION_SD) {
OSPanic(__FILE__, __LINE__, "CDBRecordDuplicate invalud location\n");
return CDB_ERROR_1;
}
recordFile = record->file;
if (recordFile == NULL) {
return CDB_ERROR_27;
}
CDBLock();
CDBRecordKeyCopy(&newRecord->key, &record->key);
newRecord->file = NULL;
newRecord->database = record->database;
err = CDBRecordFileCreateBlank(newRecord);
CDBUnlock();
return err;
}
CDBErr CDBRecordBackupToSD(CDBRecord* record) {
CDBErr err;
CDBLock();
@@ -695,8 +948,7 @@ CDBErr CDBRecordEncrypt(CDBRecord* record, void* buffer, CDBRecordKey* key, u32
}
{
CDBErr result;
CDBRecordFile* tempFile;
tempFile = record->file;
CDBRecordFile* tempFile = record->file;
if ((s32)tempFile == 0) {
result = CDB_ERROR_27;
} else if (tempFile->allocFlag == 1) {
@@ -825,7 +1077,7 @@ CDBErr CDBRecordEncrypt(CDBRecord* record, void* buffer, CDBRecordKey* key, u32
CDBErr CDBRecordDecrypt(CDBRecord* record, void* buffer, u32 size, u32* dataSize, CDBRecordKey* key) {
CDBErr status;
int fileOffset;
s32 fileOffset;
CDBRecordFile* recordFile;
u32 fileDataSize;
u32 fileSize;
@@ -960,3 +1212,32 @@ CDBErr CDBRecordDecrypt(CDBRecord* record, void* buffer, u32 size, u32* dataSize
}
return CDB_ERROR_OK;
}
CDBErr CDBRecordPrivateChangeOwner(CDBRecord* record, u64 newWiiId) {
CDBRecordFile* recordFile;
char fullPath[256];
char newPath[256];
CDBErr err;
u64 preWiiId;
u64 recordWiiId;
CDBReportError("CDBRecordPrivateChangeOwner\n");
recordFile = record->file;
if (recordFile == NULL) {
return CDB_ERROR_27;
}
preWiiId = CDBGetWiiId();
recordWiiId = record->key.wiiId;
OSReport(" i_pre-wiiid = %lX\n", preWiiId);
OSReport(" record-wiiid = %lX\n", recordWiiId);
CDBConvKeyToFullPath(&record->key, fullPath);
record->key.wiiId = newWiiId;
CDBConvKeyToFullPath(&record->key, newPath);
OSReport("path =%s\n", fullPath);
OSReport("path2=%s\n", newPath);
CDBLock();
err = CDBAttrSetWiiId(&recordFile->attr, newWiiId);
CDBUnlock();
return err;
}

View File

@@ -169,11 +169,11 @@ static const u32 AESiDecryptTable[] = {
};
#define ROTATE(x,n) (((x) << (n)) | ((x) >> (32 - (n))))
#define ENCRYPT(a,b,c,d) (ROTATE(AESiEncryptTable[(a)>>24],24) ^ ROTATE(AESiEncryptTable[((b)>>16)&255],16) ^ ROTATE(AESiEncryptTable[((c)>>8)&255],8) ^ AESiEncryptTable[(d)&255])
#define DECRYPT(a,b,c,d) (ROTATE(AESiDecryptTable[(a)>>24],24) ^ ROTATE(AESiDecryptTable[((b)>>16)&255],16) ^ ROTATE(AESiDecryptTable[((c)>>8)&255],8) ^ AESiDecryptTable[(d)&255])
#define SUBSTITUTE(table,a,b,c,d) (((u32)table[(a)>>24]<<24) | ((u32)table[((b)>>16)&255]<<16) | ((u32)table[((c)>>8)&255]<<8) | table[(d)&255])
#define ENCRYPT(a,b,c,d) (((ROTATE(AESiEncryptTable[((b)>>16)&255],16) ^ ROTATE(AESiEncryptTable[(a)>>24],24)) ^ AESiEncryptTable[(d)&255]) ^ ROTATE(AESiEncryptTable[((c)>>8)&255],8))
#define DECRYPT(a,b,c,d) (((ROTATE(AESiDecryptTable[((b)>>16)&255],16) ^ AESiDecryptTable[(d)&255]) ^ ROTATE(AESiDecryptTable[(a)>>24],24)) ^ ROTATE(AESiDecryptTable[((c)>>8)&255],8))
#define SUBSTITUTE(table,a,b,c,d) ((((u32)table[(a)>>24]<<24) | table[(d)&255]) | (((u32)table[((b)>>16)&255]<<16) | ((u32)table[((c)>>8)&255]<<8)))
#define SUBSTITUTE_XOR(table,a,b,c,d) (((u32)table[(a)>>24]<<24) ^ ((u32)table[((b)>>16)&255]<<16) ^ ((u32)table[((c)>>8)&255]<<8) ^ table[(d)&255])
#define SUBSTITUTE_XOR(table,a,b,c,d) (((((u32)table[(a)>>24]<<24) ^ table[(d)&255]) ^ ((u32)table[((c)>>8)&255]<<8)) ^ ((u32)table[((b)>>16)&255]<<16))
void AESiEncryptBlock(AESContext* context, u32* output, const u32* input) {
u32 rounds = context->rounds;
@@ -197,8 +197,9 @@ void AESiEncryptBlock(AESContext* context, u32* output, const u32* input) {
void AESiDecryptBlock(AESContext* context, u32* output, const u32* input) {
u32 rounds=context->rounds;
u32* key=context->keys + rounds*4;
u32* key=context->keys;
u32 a,b,c,d;
key += rounds*4;
if (context->needsTransform) {
u32 round;
for (round=1; round<context->rounds; ++round) {
@@ -209,9 +210,9 @@ void AESiDecryptBlock(AESContext* context, u32* output, const u32* input) {
u32 twice=DOUBLE_BYTES(value);
u32 four=DOUBLE_BYTES(twice);
u32 eight=DOUBLE_BYTES(four)^value;
u32 mixed=eight ^ (four ^ ROTATE(eight,8));
mixed ^= twice ^ ROTATE(mixed,8);
mixed ^= value ^ ROTATE(mixed,8);
u32 mixed=(ROTATE(eight,8) ^ (eight ^ four));
mixed=(ROTATE(mixed,8) ^ (mixed ^ twice));
mixed=(ROTATE(mixed,8) ^ (mixed ^ value));
keys[word]=mixed;
}
}

View File

@@ -68,8 +68,14 @@ void NETMD5GetDigest(NETMD5Context* context, void* digest) {
memset(context, 0, sizeof(*context));
}
#define ROTATE(x, n) (((x) << (n)) | ((x) >> (32 - (n))))
#define STEP(a, b, c, d, f, word, n) ((a) += (f), (a) = (b) + ROTATE(__lwbrx(0, (int)(word)) + (*constant + (a)), n))
static inline u32 rotateSum(u32 state, const u32* word, u32 constant, int shift) {
// Each rotation half reads its own byte-reversed message word.
u32 right = __lwbrx((void*)word, 0);
u32 left = __lwbrx((void*)word, 0);
return ((state + (constant + left)) << shift) | ((constant + (state + right)) >> (32 - shift));
}
#define STEP(a, b, c, d, f, word, n) ((a) = (b) + rotateSum((a) + (f), word, *constant, n))
static void ProcessBlock(NETMD5Context* context) {
static u32 constants[64] = {0xD76AA478, 0xE8C7B756, 0x242070DB, 0xC1BDCEEE, 0xF57C0FAF, 0x4787C62A, 0xA8304613, 0xFD469501, 0x698098D8, 0x8B44F7AF, 0xFFFF5BB1, 0x895CD7BE, 0x6B901122, 0xFD987193, 0xA679438E, 0x49B40821, 0xF61E2562, 0xC040B340, 0x265E5A51, 0xE9B6C7AA, 0xD62F105D, 0x02441453, 0xD8A1E681, 0xE7D3FBC8, 0x21E1CDE6, 0xC33707D6, 0xF4D50D87, 0x455A14ED, 0xA9E3E905, 0xFCEFA3F8, 0x676F02D9, 0x8D2A4C8A, 0xFFFA3942, 0x8771F681, 0x6D9D6122, 0xFDE5380C, 0xA4BEEA44, 0x4BDECFA9, 0xF6BB4B60, 0xBEBFBC70, 0x289B7EC6, 0xEAA127FA, 0xD4EF3085, 0x04881D05, 0xD9D4D039, 0xE6DB99E5, 0x1FA27CF8, 0xC4AC5665, 0xF4292244, 0x432AFF97, 0xAB9423A7, 0xFC93A039, 0x655B59C3, 0x8F0CCC92, 0xFFEFF47D, 0x85845DD1, 0x6FA87E4F, 0xFE2CE6E0, 0xA3014314, 0x4E0811A1, 0xF7537E82, 0xBD3AF235, 0x2AD7D2BB, 0xEB86D391};
@@ -83,13 +89,13 @@ static void ProcessBlock(NETMD5Context* context) {
block = context->buffer32;
word = block;
constant = constants;
index = indices;
for (round = 0; round < 4; ++round) {
STEP(a,b,c,d,(b & c) | (~b & d),word,7); ++word; ++constant;
STEP(d,a,b,c,(a & b) | (~a & c),word,12); ++word; ++constant;
STEP(c,d,a,b,(d & a) | (~d & b),word,17); ++word; ++constant;
STEP(b,c,d,a,(c & d) | (~c & a),word,22); ++word; ++constant;
STEP(a,b,c,d,(b & c) | (~b & d),word++,7); ++constant;
STEP(d,a,b,c,(a & b) | (~a & c),word++,12); ++constant;
STEP(c,d,a,b,(d & a) | (~d & b),word++,17); ++constant;
STEP(b,c,d,a,(c & d) | (~c & a),word++,22); ++constant;
}
index = indices;
for (round = 0; round < 4; ++round) {
STEP(a,b,c,d,(b & d) | (c & ~d),block + *index,5); ++index; ++constant;
STEP(d,a,b,c,(a & c) | (b & ~c),block + *index,9); ++index; ++constant;

View File

@@ -7,8 +7,7 @@ s32 NHTTPi_strcmp(const char* left, const char* right) { return strcmp(left,righ
void* NHTTPi_memclr(void* destination, u32 size) { return memset(destination,0,size); }
static int LowerCase(int character) {
if((character>='A') & (character<='Z')) character+=32;
return character;
return ((character >= 'A') & (character <= 'Z')) ? character + 32 : character;
}
s32 NHTTPi_strnicmp(const char* left, const char* right, s32 length) {
@@ -127,12 +126,13 @@ s32 NHTTPi_intToStr(char* destination, u32 value) {
}
s32 NHTTPi_compareToken(const char* left, const char* right) {
for (;;) {
int rightCharacter=LowerCase(*right);
int leftCharacter=LowerCase(*left);
if(leftCharacter!=rightCharacter) break;
if(*left==0 || *left==' ') return 0;
++left; ++right;
u8 rawLeft;
while (LowerCase((s8)(rawLeft = *(const u8*)left)) == LowerCase(*right)) {
if ((s8)rawLeft == 0 || (s8)rawLeft == ' ') {
return 0;
}
++left;
++right;
}
return -1;
}
@@ -161,13 +161,16 @@ s32 NHTTPi_Base64Encode(char* destination, const char* source) {
static char alphabet[]="ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
const char* table=alphabet;
char* output=destination;
u32 length=strlen(source);
s32 length=strlen(source);
s32 consumed=0;
for(; consumed<(s32)length; consumed+=3) {
output[0]=table[source[0]>>2];
output[1]=table[((source[0]&3)<<4)+(source[1]>>4)];
output[2]=table[((source[1]&15)<<2)+(source[2]>>6)];
output[3]=table[source[2]&63];
for(; consumed<length; consumed+=3) {
s8 first = source[0];
s8 second = source[1];
s8 third = source[2];
output[0]=table[first>>2];
output[1]=table[(second>>4)+((first&3)<<4)];
output[2]=table[(third>>6)+((second&15)<<2)];
output[3]=table[third&63];
source+=3; output+=4;
}
if(consumed==length+1) output[-1]='=';

View File

@@ -548,7 +548,7 @@ NWC24Err NWC24DumpDlTask(NWC24DlTask* dlTask) {
NWC24Err NWC24IterateDlTask(NWC24DlId* dlIterateId, BOOL begin) {
DlTaskListHeader* entriesHeader;
NWC24Work* work;
const NWC24Work* work;
u16 taskId;
u16 maxTaskCount;
DlTaskListHeader* header;

View File

@@ -242,14 +242,63 @@ BOOL BS2IsDiagDisc() { return (u8)(*(u8 *)OSPhysicalToCached(OS_ADDR_BOOT_INFO)
extern vu32 BS2VideoMode;
extern vu32 __DVDLayoutFormat;
void Run(u32 entryPoint, void *start, u32 blockCount, u32 argument) {
u8 *block = start;
for (; blockCount != 0; --blockCount) {
__dcbz(block, 0);
__dcbf(block, 0);
block += 32;
}
((void (*)(u32))entryPoint)(argument);
asm void Run(u32 entryPoint, void *start, u32 blockCount, u32 argument) {
// clang-format off
#ifdef __MWERKS__
nofralloc
mtctr r5
mtlr r3
li r0, 0
li r2, 0
li r3, 0
li r5, 0
li r7, 0
li r8, 0
li r9, 0
li r10, 0
li r11, 0
li r12, 0
li r13, 0
li r14, 0
li r15, 0
li r16, 0
li r17, 0
li r18, 0
li r19, 0
li r20, 0
li r21, 0
li r22, 0
li r23, 0
li r24, 0
li r25, 0
li r26, 0
li r27, 0
li r28, 0
li r29, 0
li r30, 0
li r31, 0
// The application stack is fixed; it is unrelated to the formatter.
lis r1, 0x8160
ori r1, r1, 0
li r6, 0
b Run_loop_entry
Run_loop:
dcbz r4, r0
dcbf r4, r0
addi r4, r4, 0x20
bdnz Run_loop
b Run_done
Run_done:
li r4, 0
blr
Run_loop_entry:
b Run_loop
#endif
// clang-format on
}
BOOL BS2GetLockedTitles(ESTitleId *pTitleIds, u32 *count) {
@@ -803,7 +852,7 @@ void BS2StartGCGame() {
OSTime time;
StartingGame = TRUE;
while (DVDGetCommandBlockStatus(&CoverBlock) != DVD_STATE_IDLE) {
while (CoverBlock.state != DVD_STATE_IDLE) {
}
soundMode = SCGetSoundMode();
@@ -1796,21 +1845,24 @@ invalidRvlRegion:
State = BS2_STT_RESET_SYSTEM;
break;
}
if (GamePartition == 0) {
if (GamePartition != 0) {
State = BS2_STT_37;
} else {
State = BS2_STT_54;
break;
}
State = BS2_STT_37;
case 0x25:
case 0x25: {
u32 partitionOffset = (u32)((DVDPartitionInfo *)GamePartition)->partition;
if (BS2BootFromCache != 0) {
BS2Report("Open partition from cache.dat\n");
NandPending = 1;
BS2NANDDivideReadAsync(&BS2CacheFileInfo, &PartitionParams.tmd, 0x4a00, BS2NANDCallback, &BS2NandBlock);
} else
DVDOpenPartitionAsync(&Block, &PartitionParams.tmd, (u32)((DVDPartitionInfo *)GamePartition)->partition, BS2DVDCallback);
DVDOpenPartitionAsync(&Block, &PartitionParams.tmd, partitionOffset, BS2DVDCallback);
PartitionOpen = 1;
State = BS2_STT_38;
break;
}
case 0x26: {
status = CheckBS2CommandStatus();
if (status == 0)
@@ -1837,13 +1889,16 @@ invalidRvlRegion:
titlePrefix = titleCode >> 0x18;
titleCharacters[3] = (u8)titleCode;
titleCharacters[0] = (char)(titleCode >> 0x18);
if (titlePrefix < 'R') {
if (titlePrefix == 'D')
goto check_title_region;
} else if (titlePrefix < 'U')
switch (titlePrefix) {
case 'D':
case 'R':
case 'S':
case 'T':
goto check_title_region;
regionMatches = TRUE;
goto title_region_checked;
default:
regionMatches = TRUE;
goto title_region_checked;
}
{
check_title_region:
if (titleCharacters[3] == 0x41) {
@@ -2001,8 +2056,8 @@ invalidRvlRegion:
if (status == 0)
OSPanic("BS2Mach.c", 0x12c8, "BS2 ERROR >>> Cannnot alloc 0x%08x from MEMAllocator", bannerFile.length);
BannerLength = bannerFile.length;
BannerBuffer = OSRoundDown32B(BannerAllocation) + 32;
BS2Report("BannerBufferAddr : %08X\n");
BannerBuffer = BannerAllocation + 32 - BannerAllocation % 32;
BS2Report("BannerBufferAddr : %08X\n", BannerBuffer);
} else {
if ((BannerBuffer != 0) && (BannerLength != 0)) {
if (BannerLength < ((bannerFile.length + 0x1fU) & 0xffffffe0))

View File

@@ -64,13 +64,13 @@ void SimpleWavePlayer::wavePacketCallback(WavePacketCallbackStatus status, WaveP
return;
}
if (self->unk_0xB5) {
if (self->mbFillInCallback) {
MakeWave(player, packet, self);
}
}
SimpleWavePlayer::SimpleWavePlayer() : mSetupParam(1, nw4r::snd::SAMPLE_FORMAT_PCM_S16, 32000, 1.0f, 1) {
unk_0xB5 = true;
mbFillInCallback = true;
unk_0xB0 = false;
mIsSetData = false;
mIsPlaying = false;
@@ -196,7 +196,7 @@ void SimpleWavePlayer::stop() {
void SimpleWavePlayer::update() {
if (mIsPlaying) {
if (!unk_0xB5) {
if (!mbFillInCallback) {
for (int i = 0; i < 3; i++) {
if (mWavePackets[i].GetAppendFlag()) {
continue;

View File

@@ -7207,7 +7207,7 @@ CHANSVmErr CHANSVmStep(CHANSVm* vm, int stepCount) {
u32 cmpLow;
struct {
CHANSVmObjType operandTypes[2];
u64 floatBits;
double floatLiteral;
CHANSVmObjHdr copies[2];
CHANSVmObjHdr load;
CHANSVmObjHdr operand;
@@ -7291,8 +7291,8 @@ CHANSVmErr CHANSVmStep(CHANSVm* vm, int stepCount) {
case CHANS_VM_OP_LOAD_FLOAT: {
opSize = 9;
operandBuf = VmGetOperand(vm, 1, 9);
memcpy(&scratch.floatBits, operandBuf, 8);
result = CHANSVmSetFloat(vm, &pVm->accumulator, *(double*)&scratch.floatBits);
memcpy(&scratch.floatLiteral, operandBuf, 8);
result = CHANSVmSetFloat(vm, &pVm->accumulator, scratch.floatLiteral);
break;
}
@@ -7441,8 +7441,7 @@ CHANSVmErr CHANSVmStep(CHANSVm* vm, int stepCount) {
}
if (leftOp == vmNull || rightOp == vmNull) {
result = CHANS_VM_ERR_RESULT_TYPE;
break;
return CHANS_VM_ERR_RESULT_TYPE;
}
result = opFunc(vm, convTypeIdx, &pVm->accumulator, leftOp, rightOp);
@@ -7717,30 +7716,27 @@ CHANSVmErr CHANSVmStep(CHANSVm* vm, int stepCount) {
CHANSVmObjHdr* pAcc;
CHANSVmExecutionCtx* ctx;
CHANSVmModule* dbg;
CHANSVmErr loadResult;
opSize = 3;
operandBuf = VmGetOperand(vm, 1, 3);
if (operandBuf == vmNull) {
return CHANS_VM_ERR_CODE_RANGE;
}
pAcc = &pVm->accumulator;
result = CHANSVmDeleteObject(vm, pAcc);
if (result == CHANS_VM_OK) {
loadResult = CHANSVmDeleteObject(vm, pAcc);
if (loadResult == CHANS_VM_OK) {
imm16Val = VM_READ_BE_U16(operandBuf, 0);
ctx = pVm->pActiveCtx;
dbg = ctx->pDbg;
if (imm16Val < dbg->stringCount) {
if (dbg->pStringTbl[imm16Val].pStringData != vmNull) {
goto load_str_ok;
}
if (imm16Val >= dbg->stringCount || dbg->pStringTbl[imm16Val].pStringData == vmNull) {
loadResult = CHANS_VM_ERR_LOAD_STRING_CONST;
} else {
pAcc->type = CHANS_VM_OBJ_TYPE_STRING;
pAcc->hasData = vmTrue;
pAcc->value.wstring_v = (vmWStringObjVal*)&ctx->pDbg->pStringTbl[imm16Val];
}
}
result = CHANS_VM_ERR_LOAD_STRING_CONST;
break;
load_str_ok:
pAcc->type = CHANS_VM_OBJ_TYPE_STRING;
pAcc->hasData = vmTrue;
pAcc->value.wstring_v = (vmWStringObjVal*)&ctx->pDbg->pStringTbl[imm16Val];
result = loadResult;
break;
}
@@ -7773,8 +7769,8 @@ CHANSVmErr CHANSVmStep(CHANSVm* vm, int stepCount) {
arrayIdx = (u32)stackPtr->value.int_v;
goto set_index_ok;
case CHANS_VM_OBJ_TYPE_FLOAT: {
double stackFloat = *(double*)stackPtr;
if (stackFloat <= 4294967294.0 && 0.0 <= stackFloat) {
double stackFloat = stackPtr->value.float_v;
if (stackFloat >= 0.0 && stackFloat <= 4294967294.0) {
arrayIdx = (u32)stackFloat;
goto set_index_ok;
}
@@ -7885,8 +7881,7 @@ CHANSVmErr CHANSVmStep(CHANSVm* vm, int stepCount) {
computedAddr += pVm->pActiveCtx->pc + 5;
}
if (computedAddr < 1 || computedAddr >= pVm->pActiveCtx->pDbg->codeSize) {
result = CHANS_VM_ERR_CODE_RANGE;
break;
return CHANS_VM_ERR_CODE_RANGE;
}
pVm->pActiveCtx->pc = computedAddr;
opSize = 0;
@@ -7934,8 +7929,7 @@ CHANSVmErr CHANSVmStep(CHANSVm* vm, int stepCount) {
computedAddr += pVm->pActiveCtx->pc + 4;
}
if (computedAddr < 1 || computedAddr >= pVm->pActiveCtx->pDbg->codeSize) {
result = CHANS_VM_ERR_CODE_RANGE;
break;
return CHANS_VM_ERR_CODE_RANGE;
}
pVm->pActiveCtx->pc = computedAddr;
opSize = 0;
@@ -7960,11 +7954,13 @@ CHANSVmErr CHANSVmStep(CHANSVm* vm, int stepCount) {
}
case CHANS_VM_OP_DELETE_INDIRECT: {
result = CHANS_VM_ERR_DELETE_INDIRECT;
if (pVm->accumulator.type == CHANS_VM_TYPE_INDEX_REF) {
foundObj = VmGetArrayElement(vm, &pVm->accumulator, *((u32*)&pVm->accumulator.value + 1), vmFalse);
result = VmDeleteCommon(vm, foundObj);
CHANSVmObjHdr* accumulator = &pVm->accumulator;
CHANSVmErr deleteResult = CHANS_VM_ERR_DELETE_INDIRECT;
if (accumulator->type == CHANS_VM_TYPE_INDEX_REF) {
foundObj = VmGetArrayElement(vm, accumulator, accumulator->value.data.len, vmFalse);
deleteResult = VmDeleteCommon(vm, foundObj);
}
result = deleteResult;
break;
}
@@ -8008,34 +8004,27 @@ CHANSVmErr CHANSVmStep(CHANSVm* vm, int stepCount) {
if (pVm->pActiveCtx->stackDepth != 0 && pVm->pObjStackTopBuf + sizeof(CHANSVmObjHdr) <= pVm->pHeapEnd) {
CHANSVmObjHdr* stackTop = (CHANSVmObjHdr*)pVm->pObjStackTopBuf;
isTypeMatch = 0;
if (pVm->accumulator.type == stackTop->type) {
if (stackTop->type == pVm->accumulator.type) {
switch (stackTop->type) {
case CHANS_VM_OBJ_TYPE_INTEGER: {
u32 low1 = *(u32*)stackTop;
u32 low2 = *(u32*)&pVm->accumulator;
u32 high1 = *((u32*)stackTop + 1);
u32 high2 = *((u32*)&pVm->accumulator + 1);
u32 xor1 = low1 ^ low2;
u32 xor2 = high1 ^ high2;
isTypeMatch = !(xor1 | xor2);
isTypeMatch = stackTop->value.int_v == pVm->accumulator.value.int_v;
goto end_branch_check;
}
case CHANS_VM_OBJ_TYPE_FLOAT: {
double stackFloat = stackTop->value.float_v;
double accFloat = pVm->accumulator.value.float_v;
isTypeMatch = stackFloat == accFloat ? 1 : 0;
isTypeMatch = stackTop->value.float_v == pVm->accumulator.value.float_v;
goto end_branch_check;
}
case CHANS_VM_OBJ_TYPE_STRING: {
u32 len1, len2;
CHANSVmObjHdr* stackObj = (CHANSVmObjHdr*)pVm->pObjStackTopBuf;
const vmWStringObjVal* accumulatorString = pVm->accumulator.value.wstring_v;
const vmWStringObjVal* stackString = stackTop->value.wstring_v;
u32 strEqual;
isTypeMatch = 0;
len1 = stackObj->value.wstring_v->len;
len2 = pVm->accumulator.value.wstring_v->len;
len1 = stackString->len;
len2 = accumulatorString->len;
if (len1 == len2) {
strEqual = 0;
if (len1 == 0 || memcmp(stackObj->value.wstring_v->spData, pVm->accumulator.value.wstring_v->spData, len1) == 0) {
if (len1 == 0 || memcmp(stackString->spData, accumulatorString->spData, len1) == 0) {
strEqual = 1;
}
if (strEqual) {

View File

@@ -16,7 +16,7 @@ namespace ipl {
namespace cs {
namespace savedata {
struct data {
char unk_0x00[MAX_PATH_LENGTH + sizeof(wchar_t)]; // 0x00
char sMountPath[MAX_PATH_LENGTH + sizeof(wchar_t)]; // 0x00
char sPath[MAX_PATH_LENGTH + sizeof(wchar_t)]; // 0x42
u8 flag; // 0x84
u8* pData; // 0x88
@@ -67,7 +67,7 @@ namespace ipl {
data* data = static_cast<savedata::data*>(*hdr->value.ptr_v);
if (data->pData == NULL && data->size == 0) {
if (data->unk_0x00[0] == 0) {
if (data->sMountPath[0] == 0) {
error = load_nand_(vm, data, heap, offset, size, 0x20);
} else {
error = load_vf_(vm, data, heap, offset, size, 0x20);
@@ -127,7 +127,7 @@ namespace ipl {
}
}
if (data->unk_0x00[0] == '\0') {
if (data->sMountPath[0] == '\0') {
error = load_nand_(vm, data, csHeap, 0, size, -0x20);
} else {
error = load_vf_(vm, data, csHeap, 0, size, -0x20);
@@ -591,7 +591,7 @@ namespace ipl {
memset(data, 0, sizeof(savedata::data));
memset(temp, 0, sizeof(temp));
if (arg1 != NULL) {
util::utf16_to_ascii(data->unk_0x00, temp, MAX_PATH_LENGTH, arg0);
util::utf16_to_ascii(data->sMountPath, temp, MAX_PATH_LENGTH, arg0);
memset(temp, 0, sizeof(temp));
util::utf16_to_ascii(data->sPath, temp, MAX_PATH_LENGTH, arg1);
} else {

View File

@@ -1179,10 +1179,10 @@ namespace ipl {
return;
}
if (mState != STATE_NORMAL || mMode != 0) {
return;
}
switch (mState) {
case STATE_NORMAL:
switch (mMode) {
case 0:
if (!mpFriendCache->isThere(buttonNo + mPage * BTN_MAX)) {
return;
}
@@ -1240,6 +1240,8 @@ namespace ipl {
snd::getSystem()->startSEwithPos("WIPL_SE_CH_HOLD", mDrag.mPos.x);
mState = STATE_DRAG;
}
}
}
void Address::start_drag_point_event(const char* paneName, controller::Interface* con) {
@@ -1425,10 +1427,7 @@ namespace ipl {
if (con->downTrg(controller::BTN_INTERACT)) {
Button* button = static_cast<Button*>(System::getScene(SCENE_BUTTON));
if (mState != STATE_COVER_NORMAL && mState != STATE_NORMAL) {
break;
}
if (mState == STATE_COVER_NORMAL || mState == STATE_NORMAL) {
if (Button::cmpButtonName(paneName, Button::BTN_EXIT) == 0) {
MailAddressSelect* mailAddrSel = static_cast<MailAddressSelect*>(System::getScene(SCENE_MAIL_ADDRESS_SELECT));
button->animation(Button::IDANIM_SELECT_CALENDAR_EXIT);
@@ -1477,7 +1476,7 @@ namespace ipl {
snd::getSystem()->startSE("WIPL_SE_FL_PAGE_DEC");
onPreviousPage();
}
break;
}
}
// fallthrough
}
@@ -1783,13 +1782,15 @@ namespace ipl {
nw4r::math::VEC3 baseTrans = mpLayout->FindPaneByName("N_base_move")->GetTranslate();
math::VEC2 pos = System::getControllerManager()->getController(mDrag.mChan)->getDpdProjectionPos();
nw4r::lyt::TextBox* textBox = static_cast<nw4r::lyt::TextBox*>(mpLayout->FindPaneByName("T_name_move"));
const u16* name = mpFriendCache->getInfo(mDrag.mButton + mDrag.mPage * BTN_MAX).attr.name;
f32 width = 0.0f;
if (textBox != NULL && name != NULL) {
textBox->GetFont()->GetWidth();
for (; *name != 0; name++) {
width += textBox->GetFont()->GetCharWidth(*name);
if (textBox != NULL) {
const u16* name = mpFriendCache->getInfo(mDrag.mButton + mDrag.mPage * BTN_MAX).attr.name;
if (name != NULL) {
textBox->GetFont()->GetWidth();
for (; *name != 0; name++) {
width += textBox->GetFont()->GetCharWidth(*name);
}
}
}
width += 0.01f;
@@ -1946,11 +1947,10 @@ namespace ipl {
}
void FriendListCache::update(u32 index, const wchar_t* name, u64 fdId) {
NWC24FriendInfo* info = &mInfos[index];
info->attr.fdId = fdId;
memset(info->attr.name, 0, sizeof(info->attr.name));
wcsncpy((wchar_t*)info->attr.name, name, 10);
System::getNwc24Manager()->updateFriendInfo(info, index);
mInfos[index].attr.fdId = fdId;
memset(mInfos[index].attr.name, 0, sizeof(mInfos[index].attr.name));
wcsncpy((wchar_t*)getInfo(index).attr.name, name, 10);
System::getNwc24Manager()->updateFriendInfo(&getInfo(index), index);
}
void FriendListCache::del(u32 index) {

View File

@@ -521,7 +521,8 @@ void ipl::scene::AddressEdit::create() {
set_textbox(textPane, friendText);
textPane = (mpCodeLayout)
->getNW4RLyt()->GetRootPane()->FindPaneByName("T_frnd_crd_00", true);
set_textbox(textPane, mString.mDisplayText);
friendText = mString.mDisplayText;
set_textbox(textPane, friendText);
mState = 0;
break;
}
@@ -2779,7 +2780,7 @@ void ipl::scene::AddressEdit::set_err_msg(wchar_t* outErrMsg, u32 outErrMsgLen,
swprintf(nwc24ErrStr, sizeof(nwc24ErrStr) / sizeof(wchar_t), L"%06d\n", errorCode);
wcsncat(outErrMsg, nwc24ErrStr, outErrMsgLen - wcslen(outErrMsg));
u32 messageId = MESG_ERROR_NWC24_SERVER;
u32 messageId;
switch (nwc24Err) {
case NWC24_ERR_NETWORK:
messageId = MESG_ERROR_NWC24_NETWORK;
@@ -2865,10 +2866,7 @@ void ipl::scene::AddressEdit::get_friendinfo() {
memcpy(&ipl::scene::sFriendInfo, &mpFriendCache->getInfo(mSelectedFriend),
sizeof(ipl::scene::sFriendInfo));
mString.setName(reinterpret_cast<const wchar_t*>(ipl::scene::sFriendInfo.attr.name));
const wchar_t* name = mString.mName;
nw4r::lyt::Pane* namePane =
mpCodeLayout->getNW4RLyt()->GetRootPane()->FindPaneByName("T_name_00", true);
set_textbox(namePane, name);
set_textbox(mpCodeLayout->getNW4RLyt()->GetRootPane()->FindPaneByName("T_name_00", true), mString.mName);
wchar_t wiiNo[0x100];
wchar_t email[0x102];
@@ -2881,10 +2879,7 @@ void ipl::scene::AddressEdit::get_friendinfo() {
ipl::utility::CharacterCode::ANSIToUTF16(email, reinterpret_cast<const u8*>(&ipl::scene::sFriendInfo.addr), 0x102);
mString.setEMail(email);
}
const wchar_t* displayText = mString.mDisplayText;
nw4r::lyt::Pane* friendCodePane =
mpCodeLayout->getNW4RLyt()->GetRootPane()->FindPaneByName("T_frnd_crd_00", true);
set_textbox(friendCodePane, displayText);
set_textbox(mpCodeLayout->getNW4RLyt()->GetRootPane()->FindPaneByName("T_frnd_crd_00", true), mString.mDisplayText);
memcpy(&mCreateID, &ipl::scene::sFriendInfo.attr.fdId, sizeof(mCreateID));
}

View File

@@ -674,7 +674,8 @@ namespace ipl {
if (foundLangGroup) {
nw4r::lyt::Group* group = layout->FindGroupByName(langGroup);
for (nw4r::lyt::PaneLinkList::Iterator it = group->GetPaneList().GetBeginIter(); it != group->GetPaneList().GetEndIter(); it++) {
nw4r::lyt::Group* const& groupView = group;
for (nw4r::lyt::PaneLinkList::Iterator it = groupView->GetPaneList().GetBeginIter(); it != groupView->GetPaneList().GetEndIter(); it++) {
it->mTarget->SetVisible(true);
}
} else {
@@ -685,7 +686,8 @@ namespace ipl {
}
if (strcmp(scModuleData.langGroupLookup[region][i], langCodeBuf[scLangLookup[region][i]]) == 0) {
nw4r::lyt::Group* group = layout->FindGroupByName(scModuleData.langGroupLookup[region][i]);
for (nw4r::lyt::PaneLinkList::Iterator it = group->GetPaneList().GetBeginIter(); it != group->GetPaneList().GetEndIter(); it++) {
nw4r::lyt::Group* const& groupView = group;
for (nw4r::lyt::PaneLinkList::Iterator it = groupView->GetPaneList().GetBeginIter(); it != groupView->GetPaneList().GetEndIter(); it++) {
it->mTarget->SetVisible(true);
}
break;

View File

@@ -237,10 +237,7 @@ bool MemoryCardManager::isIconValidate(u8 slot, s16 index) {
bool MemoryCardManager::isBannerEnable(u8 slot, s16 index) {
CardIcons* icons = reinterpret_cast<CardIcons*>(memorycard::getIconStateArray());
u32 file = mFile[slot][index].fileNo;
memorycard::IconState* icon = &icons[slot][file];
u8 enabled = icon->bannerEnable;
return enabled != 0;
return icons[slot][mFile[slot][index].fileNo].bannerEnable != 0;
}
void MemoryCardManager::update_icon_anm() {

View File

@@ -1341,13 +1341,11 @@ namespace ipl {
wchar_t* titleName = System::getChannelManager()->getTitleName(
page, channelIndex, 0);
memcpy(titleNames + *titleCount * 0x2a, titleName, 0x2a);
u32 nextTitleCount = *titleCount + 1;
*titleCount = nextTitleCount;
++*titleCount;
if (bytes < secondUsage[0] || blocks < secondUsage[1]) {
continue;
if (bytes >= secondUsage[0] && blocks >= secondUsage[1]) {
return true;
}
return true;
}
}
}

View File

@@ -3,6 +3,10 @@
#define IPL_SDMEMORY_DIALOG_STATE_ACCESSOR
#define IPL_SDMEMORY_COMPLETION_PCT_ACCESSOR
#define IPL_CHANNEL_TITLE_NOVTABLE
#include "math/iplInterporation.h"
#undef IPL_CHANNEL_TITLE_NOVTABLE
#include "scene/sdChannelTitle/iplSDChannelTitle.h"
#include "scene/sdChannelSelect/iplSDChannelSelect.h"
#include "system/iplSystem.h"
@@ -1554,15 +1558,5 @@ void SDChannelTitle::startResetting() {
snd::getSystem()->resetAllSound();
}
extern "C" void iplSDChannelTitle_onTitleButtonEvent(SDTitleButtonEventHandler* handler, u32 component,
u32 event, const controller::Interface* controller) {
handler->onEventDerived(component, event, controller);
}
extern "C" void iplSDChannelTitle_onTitlePaneEvent(SDTitlePaneEventHandler* handler, u32 component,
u32 event, const controller::Interface* controller) {
handler->onEvent(component, event, const_cast<controller::Interface*>(controller));
}
}
}

View File

@@ -1379,30 +1379,34 @@ namespace ipl {
return;
}
ext_ead::www::BrowserThread* browser = surface->GetBrowserThread();
if (browser == NULL) {
// One pointer, overwritten by the standard-buffer fetch, so the
// browser `this` and standardBuffer share a callee-saved reg.
void* held = surface->GetBrowserThread();
if (held == NULL) {
return;
}
if (browser->GetTextureBuffer(0, NULL) == NULL) {
if (static_cast<ext_ead::www::BrowserThread*>(held)->GetTextureBuffer(0, NULL) == NULL) {
return;
}
ext_ead::www::BrowserWindow* browserWindow =
static_cast<ext_ead::www::BrowserWindow*>(browser->mpBrowserWindows[0]);
if ((browserWindow != NULL ? browserWindow->unk_0x2C4[3] : 0) != 0) {
static_cast<ext_ead::www::BrowserWindow*>(static_cast<ext_ead::www::BrowserThread*>(held)->mpBrowserWindows[0]);
if ((browserWindow == NULL ? 0 : browserWindow->unk_0x2C4[3]) != 0) {
unk_0x91C[2] = 1;
mState = 0;
browserWindow = static_cast<ext_ead::www::BrowserWindow*>(browser->mpBrowserWindows[0]);
browserWindow = static_cast<ext_ead::www::BrowserWindow*>(static_cast<ext_ead::www::BrowserThread*>(held)->mpBrowserWindows[0]);
if (browserWindow != NULL) {
browserWindow->unk_0x2C4[3] = 0;
}
www::trasition::ScrollState scrollState = www::trasition::GetScrollState();
browserScrollDirection = scrollState == www::trasition::SCROLL_LEFT
? 1
: (scrollState == www::trasition::SCROLL_RIGHT ? -1 : 0);
if (scrollState == www::trasition::SCROLL_LEFT) {
browserScrollDirection = 1;
} else {
browserScrollDirection = scrollState == www::trasition::SCROLL_RIGHT ? -1 : 0;
}
www::trasition::ResetScrollState();
void* changedWideBuffer = browser->GetTextureBuffer(1, NULL);
void* changedStandardBuffer = browser->GetTextureBuffer(0, NULL);
void* changedWideBuffer = static_cast<ext_ead::www::BrowserThread*>(held)->GetTextureBuffer(1, NULL);
void* changedStandardBuffer = static_cast<ext_ead::www::BrowserThread*>(held)->GetTextureBuffer(0, NULL);
OSReport("changed %p %p\n", changedStandardBuffer, changedWideBuffer);
ext_ead::www::Heap::reportLeaHeap();
}
@@ -1415,8 +1419,9 @@ namespace ipl {
WWWRect* standardRect;
WWWRect* wideRect;
void* wideBuffer = browser->GetTextureBuffer(1, &wideRect);
void* standardBuffer = browser->GetTextureBuffer(0, &standardRect);
void* wideBuffer = static_cast<ext_ead::www::BrowserThread*>(held)->GetTextureBuffer(1, &wideRect);
held = static_cast<ext_ead::www::BrowserThread*>(held)->GetTextureBuffer(0, &standardRect);
void* standardBuffer = held;
nw4r::ut::Rect projection4x3;
System::getProjectionRect4x3(&projection4x3);
int screenWidth = abs(static_cast<int>(projection4x3.GetWidth()));
@@ -1443,33 +1448,33 @@ namespace ipl {
int wideWidth = abs(static_cast<int>(projection16x9.GetWidth()));
int wideHeight = abs(static_cast<int>(projection16x9.GetHeight()));
f32 wideLeft = -wideWidth / 2;
f32 wideTop = wideHeight / 2;
f32 wideTopL = wideHeight / 2;
f32 wideRight = wideWidth / 2;
nw4r::ut::Rect leftSide(wideLeft, wideTop,
f32 wideTopR = wideHeight / 2;
nw4r::ut::Rect leftSide(wideLeft, wideTopL,
wideLeft + sideDescriptor->textureHeader->width,
wideTop - sideDescriptor->textureHeader->height);
wideTopL - sideDescriptor->textureHeader->height);
nw4r::ut::Rect rightSide(wideRight - sideDescriptor->textureHeader->width,
wideTop, wideRight, wideTop - sideDescriptor->textureHeader->height);
wideTopR, wideRight, wideTopR - sideDescriptor->textureHeader->height);
if (browserScrollDirection != 0) {
nw4r::lyt::Material* wideMaterial =
void* wideMaterial =
mpChangeLayout->FindPaneByName("N_Tra0")->FindMaterialByName("Tex0");
nw4r::lyt::Material* firstStandardMaterial =
mpChangeLayout->FindPaneByName("N_Tra0")->FindMaterialByName("Tex1");
nw4r::lyt::Material* secondStandardMaterial =
mpChangeLayout->FindPaneByName("N_Tra0")->FindMaterialByName("Tex2");
wideMaterial->SetTexture(0, wideTexture);
static_cast<nw4r::lyt::Material*>(wideMaterial)->SetTexture(0, wideTexture);
firstStandardMaterial->SetTexture(0, standardTexture);
secondStandardMaterial->SetTexture(0, standardTexture);
mpChangeLayout->FindPaneByName("N_Tra0")->SetVisible(true);
if (browserScrollDirection == 1) {
utility::FrameController* animation = mpSecondAnimation;
animation->initFrame();
animation->restart();
wideMaterial = mpSecondAnimation;
static_cast<utility::FrameController*>(wideMaterial)->initFrame();
static_cast<utility::FrameController*>(wideMaterial)->restart();
} else {
utility::FrameController* animation = mpFirstAnimation;
animation->initFrame();
animation->restart();
wideMaterial = mpFirstAnimation;
static_cast<utility::FrameController*>(wideMaterial)->initFrame();
static_cast<utility::FrameController*>(wideMaterial)->restart();
}
mpChangeLayout->calc();
browserScrollDirection = 0;
@@ -1509,13 +1514,44 @@ namespace ipl {
}
if (mpWiiSettingFlag->smthMsgData >= 2 && mpWiiSettingFlag->smthMsgData <= 7) {
GXRenderModeObj renderMode = *System::getRenderModeObj();
GXRenderModeObj* pRmode = System::getRenderModeObj();
GXRenderModeObj renderMode;
volatile GXRenderModeObj* srcMode = pRmode;
volatile GXRenderModeObj* dstMode = &renderMode;
int n = 12;
u8* src;
u8* dst = &renderMode.aa;
src = &pRmode->aa;
dstMode->viTVmode = srcMode->viTVmode;
dstMode->fbWidth = srcMode->fbWidth;
dstMode->efbHeight = srcMode->efbHeight;
dstMode->xfbHeight = srcMode->xfbHeight;
dstMode->viXOrigin = srcMode->viXOrigin;
dstMode->viYOrigin = srcMode->viYOrigin;
dstMode->viWidth = srcMode->viWidth;
dstMode->viHeight = srcMode->viHeight;
dstMode->xFBmode = srcMode->xFBmode;
dstMode->field_rendering = srcMode->field_rendering;
dstMode->aa = srcMode->aa;
for (; n != 0; n--) {
dst[1] = src[1];
dst[2] = src[2];
src += 2;
dst += 2;
}
dstMode->vfilter[0] = srcMode->vfilter[0];
dstMode->vfilter[1] = srcMode->vfilter[1];
dstMode->vfilter[2] = srcMode->vfilter[2];
dstMode->vfilter[3] = srcMode->vfilter[3];
dstMode->vfilter[4] = srcMode->vfilter[4];
dstMode->vfilter[5] = srcMode->vfilter[5];
dstMode->vfilter[6] = srcMode->vfilter[6];
u32 left;
u32 top;
u32 width;
u32 height;
GXGetScissor(&left, &top, &width, &height);
GXSetScissor(0, renderMode.efbHeight / 2 - 0xA4, renderMode.fbWidth, 0x132);
GXSetScissor(0, dstMode->efbHeight / 2 - 0xA4, dstMode->fbWidth, 0x132);
mpMainLayout->draw();
GXSetScissor(left, top, width, height);
}

View File

@@ -931,17 +931,15 @@ void CArGBAOdh::cdj_c_makeHeader(SArCDJ_OdhMaster* master, u32 size) {
}
s32 CArGBAOdh::cdj_c_colorConv(SArCDJ_OdhMaster* master, u8* sourceData, int format) {
u32 paddedWidth;
u16 width;
u16 height;
u8* workPlane;
int sourceStride;
u8* crPlane;
u8* cbPlane;
u8* crPlane;
u16* dimensions;
int rowIndex;
int sourceStride;
int i;
u16 dimension;
u16 paddedDimensions[2];
dimensions = &master->width;
@@ -952,11 +950,11 @@ s32 CArGBAOdh::cdj_c_colorConv(SArCDJ_OdhMaster* master, u8* sourceData, int for
height = dimensions[1];
for (i = 0; i < 2; i++) {
dimension = dimensions[i];
if ((dimension & 7) != 0) {
paddedDimensions[i] = (dimension + 8) - (dimension & 7);
u16 remainder = dimensions[i] & 7;
if (remainder != 0) {
paddedDimensions[i] = (dimensions[i] + 8) - remainder;
} else {
paddedDimensions[i] = dimension;
paddedDimensions[i] = dimensions[i];
}
}
@@ -972,13 +970,12 @@ s32 CArGBAOdh::cdj_c_colorConv(SArCDJ_OdhMaster* master, u8* sourceData, int for
}
}
paddedWidth = paddedDimensions[0];
for (rowIndex = 0; rowIndex < (s32)height; rowIndex++) {
LineConv11(sourceData + (rowIndex & 3) * 8 + rowIndex / 4 * sourceStride, workPlane, cbPlane, crPlane,
dimensions[0], dimensions[1], (const long*)gArConvPlttTbl, format);
workPlane += paddedWidth;
cbPlane += paddedWidth;
crPlane += paddedWidth;
workPlane += paddedDimensions[0];
cbPlane += paddedDimensions[0];
crPlane += paddedDimensions[0];
}
return 0;
@@ -1362,12 +1359,10 @@ s32 CArGBAOdh::cdj_d_initializeDecompressOdh(SArCDJ_OdhMaster* master, u8* workB
}
s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int workSize) {
int blockStride;
int workBufferSize = workSize;
u8 xSampling;
u8 coefficientIndex;
u16 blockWidth;
int blockRowStride;
int planeOffset;
u32 sourceLimit = srcSize;
int statusOrOffset;
@@ -1380,8 +1375,6 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
blockWidth = master->blocksWide;
blockRowSize = (u32)blockWidth * 0x38;
rowStride = (u32)blockWidth << 3;
blockRowStride = blockRowSize * 2;
blockStride = (u32)blockWidth << 4;
do {
xSampling = master->xSampling;
if ((xSampling == 1U) && (master->ySampling == 1U)) {
@@ -1412,9 +1405,9 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
coefficientIndex = odh_natural_order[naturalIndex];
master->coefficients[coefficientIndex] = master->dcCoefficients[naturalIndex];
}
statusOrOffset = ((u32)master->blockY << 6) * (u32)master->blocksWide +
((u32)master->blocksHigh * ((u32)master->blocksWide << 6) +
(u32)master->blockX * 8);
statusOrOffset = (u32)master->blockX * 8 +
(u32)master->blocksWide * ((u32)master->blockY << 6) +
(u32)master->blocksHigh * ((u32)master->blocksWide << 6);
if (statusOrOffset + blockRowSize + 8 > workBufferSize) {
statusOrOffset = -0x7ffffffa;
goto finishBlock;
@@ -1454,14 +1447,14 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
}
planeOffset = (u32)master->blockX * 0x10 +
(u32)master->blocksWide * ((u32)master->blockY << 6) * 2;
requiredBlockEnd = planeOffset + blockRowStride + 8;
requiredBlockEnd = planeOffset + (blockRowSize * 2) + 8;
if (requiredBlockEnd > workBufferSize) {
statusOrOffset = -0x7ffffffa;
goto finishBlock;
}
idct_fast
(range_limit,master->quantizationTables,master->coefficients,master->workBuffer + planeOffset
,blockStride);
, ((u32)blockWidth << 4));
statusOrOffset = huffmanDecoder
(master->dcCoefficients, &master->luminanceRequest, (u16**)hufftreePtr.tables, 0, sourceLimit);
master->chrominanceRequest.bitstream = master->luminanceRequest.bitstream;
@@ -1477,7 +1470,7 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
}
idct_fast
(range_limit,master->quantizationTables,master->coefficients,
planeOffset + master->workBuffer + 8,blockStride);
(u8*)((u32)planeOffset + (u32)master->workBuffer) + 8, ((u32)blockWidth << 4));
statusOrOffset = huffmanDecoder
(master->dcCoefficients, &master->chrominanceRequest, (u16**)(hufftreePtr.tables + 2), 1, sourceLimit);
if (statusOrOffset != 0) return statusOrOffset;
@@ -1543,7 +1536,7 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
coefficientIndex = odh_natural_order[naturalIndex];
master->coefficients[coefficientIndex] = master->dcCoefficients[naturalIndex];
}
planeOffset = planeOffset + (u32)master->blocksWide * 0x40;
planeOffset += (u32)master->blocksWide << 6;
if (planeOffset + blockRowSize + 8 > workBufferSize) {
statusOrOffset = -0x7ffffffa;
goto finishBlock;
@@ -1600,14 +1593,14 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
}
planeOffset = (u32)master->blockX * 0x10 +
(u32)master->blocksWide * ((u32)master->blockY << 6) * 4;
requiredBlockEnd = planeOffset + blockRowStride + 8;
requiredBlockEnd = planeOffset + (blockRowSize * 2) + 8;
if (requiredBlockEnd > workBufferSize) {
statusOrOffset = -0x7ffffffa;
goto finishBlock;
}
idct_fast
(range_limit,master->quantizationTables,master->coefficients,master->workBuffer + planeOffset
,blockStride);
, ((u32)blockWidth << 4));
statusOrOffset = huffmanDecoder
(master->dcCoefficients, &master->luminanceRequest, (u16**)hufftreePtr.tables, 0, sourceLimit);
if (statusOrOffset != 0) return statusOrOffset;
@@ -1618,13 +1611,13 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
}
int nextOffset = planeOffset + 8;
planeOffset = nextOffset;
if (nextOffset + blockRowStride + 8 > workBufferSize) {
if (nextOffset + (blockRowSize * 2) + 8 > workBufferSize) {
statusOrOffset = -0x7ffffffa;
goto finishBlock;
}
idct_fast
(range_limit,master->quantizationTables,master->coefficients,master->workBuffer + nextOffset
,blockStride);
, ((u32)blockWidth << 4));
statusOrOffset = huffmanDecoder
(master->dcCoefficients, &master->luminanceRequest, (u16**)hufftreePtr.tables, 0, sourceLimit);
if (statusOrOffset != 0) return statusOrOffset;
@@ -1633,15 +1626,16 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
coefficientIndex = odh_natural_order[naturalIndex];
master->coefficients[coefficientIndex] = master->dcCoefficients[naturalIndex];
}
planeOffset = planeOffset + (u32)master->blocksWide * 0x80 + -8;
requiredBlockEnd = planeOffset + blockRowStride + 8;
if (requiredBlockEnd > workBufferSize) {
// Store the updated plane in requiredBlockEnd so MWCC keeps it in r15.
requiredBlockEnd = planeOffset + (u32)master->blocksWide * 0x80 + -8;
planeOffset = requiredBlockEnd + (blockRowSize * 2) + 8;
if (planeOffset > workBufferSize) {
statusOrOffset = -0x7ffffffa;
goto finishBlock;
}
idct_fast
(range_limit,master->quantizationTables,master->coefficients,master->workBuffer + planeOffset
,blockStride);
(range_limit,master->quantizationTables,master->coefficients,master->workBuffer + requiredBlockEnd
, ((u32)blockWidth << 4));
statusOrOffset = huffmanDecoder
(master->dcCoefficients, &master->luminanceRequest, (u16**)hufftreePtr.tables, 0, sourceLimit);
master->chrominanceRequest.bitstream = master->luminanceRequest.bitstream;
@@ -1651,13 +1645,13 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
coefficientIndex = odh_natural_order[naturalIndex];
master->coefficients[coefficientIndex] = master->dcCoefficients[naturalIndex];
}
if (requiredBlockEnd + 8 > workBufferSize) {
if (planeOffset + 8 > workBufferSize) {
statusOrOffset = -0x7ffffffa;
goto finishBlock;
}
idct_fast
(range_limit,master->quantizationTables,master->coefficients,
planeOffset + master->workBuffer + 8,blockStride);
(u8*)((u32)requiredBlockEnd + (u32)master->workBuffer) + 8, ((u32)blockWidth << 4));
statusOrOffset = huffmanDecoder
(master->dcCoefficients, &master->chrominanceRequest, (u16**)(hufftreePtr.tables + 2), 1, sourceLimit);
if (statusOrOffset != 0) return statusOrOffset;
@@ -1696,9 +1690,9 @@ s32 CArGBAOdh::cdj_d_decompressLoop(SArCDJ_OdhMaster* master, int srcSize, int w
(range_limit,master->quantizationTables + 64,master->coefficients,master->workBuffer + statusOrOffset,
rowStride);
}
blockX = master->blockX + 1;
master->blockX = blockX;
if (blockX == master->blocksWide) {
int nextBlockX = master->blockX + 1;
master->blockX = nextBlockX;
if ((u16)nextBlockX == master->blocksWide) {
master->blockX = 0;
master->blockY = master->blockY + 1;
}

View File

@@ -884,172 +884,187 @@ namespace ipl {
return ret;
}
void ESMisc::DeleteUnauthorizedData(EGG::Heap* heap) {
u32 titleCount = 0;
ESTitleId* titleIds = NULL;
char path[88];
static inline void verifySavedataZD(EGG::Heap* heap, ESTitleId titleId, NANDFileInfo* fileInfo) {
char path[NAND_MAX_PATH];
s32 ret;
u8* saveData = NULL;
BOOL fileOpen = FALSE;
BOOL deleteSaveData = FALSE;
sprintf(path, "/title/%08x/%08x/data/%s", (u32)((titleId & 0x00FFFFFFFFFFFFFFULL) >> 32), (u32)(titleId & 0x00FFFFFFFFFFFFFFULL), "zeldaTp.dat");
if (!ESMisc::ChangeUid(titleId)) {
OSReport("%s::%s: ChangeUid failed\n", __FILE__, __FUNCTION__);
ESMisc::DeleteTitle(heap, titleId);
} else {
ret = NANDPrivateOpen(path, fileInfo, NAND_ACCESS_READ);
if (ret == NAND_RESULT_NOEXISTS) {
OSReport("%s::%s: Does not exist %s: %d\n", __FILE__, __FUNCTION__, path, ret);
} else if (ret != NAND_RESULT_OK) {
OSReport("%s::%s: Open save data file failed: %d\n", __FILE__, __FUNCTION__, ret);
} else {
fileOpen = TRUE;
saveData = (u8*)heap->alloc(0x4000, -DEFAULT_ALIGN);
if (saveData == NULL) {
OSReport("%s::%s: Alloc failed: %d\n", __FILE__, __FUNCTION__, -2);
} else {
memset(saveData, 0, 0x4000);
ret = NANDRead(fileInfo, saveData, 0x4000);
if (ret < 0) {
OSReport("%s::%s: Read file failed: %d\n", __FILE__, __FUNCTION__, ret);
goto cleanup;
} else if (ret != 0x4000) {
OSReport("%s::%s: File size is not correct: %d\n", __FILE__, __FUNCTION__, ret);
deleteSaveData = TRUE;
} else {
u32 offset = 8;
u32 j = 0;
BOOL valid = FALSE;
while (j < 3) {
u8* block = saveData + offset;
if (!checkForNullTermination((char*)block + 0x4e, 8) ||
!checkForNullTermination((char*)block + 0x58, 8) ||
!checkForNullTermination((char*)block + 0x72, 8) ||
!checkForNullTermination((char*)block + 0x8e, 8) ||
!checkForNullTermination((char*)block + 0x1b4, 0x11) ||
!checkForNullTermination((char*)block + 0x1c5, 0x11)) {
goto verify_failed;
}
j++;
offset += 0xa94;
}
offset = 0x2008;
j = 0;
while (j < 3) {
u8* block = saveData + offset;
if (!checkForNullTermination((char*)block + 0x4e, 8) ||
!checkForNullTermination((char*)block + 0x58, 8) ||
!checkForNullTermination((char*)block + 0x72, 8) ||
!checkForNullTermination((char*)block + 0x8e, 8) ||
!checkForNullTermination((char*)block + 0x1b4, 0x11) ||
!checkForNullTermination((char*)block + 0x1c5, 0x11)) {
goto verify_failed;
}
j++;
offset += 0xa94;
}
valid = TRUE;
verify_failed:
if (!valid) {
OSReport("%s::%s: Verify failed for %016llx\n", __FILE__, __FUNCTION__,
titleId);
deleteSaveData = TRUE;
}
}
NANDClose(fileInfo);
fileOpen = FALSE;
if (deleteSaveData) {
ESMisc::DeleteSavedata(titleId, heap);
}
}
}
}
cleanup:
if (saveData != NULL) {
heap->free(saveData);
}
if (fileOpen) {
NANDClose(fileInfo);
}
ESMisc::ChangeUid(SYSMENU_TITLE_ID);
}
static inline s32 DeleteTicketsForce(EGG::Heap* heap, ESTitleId titleId, u8* ticketViews, u32* ticketViewCount) {
ESTicketView* ticketViewList = NULL;
s32 ret;
memset(ticketViews, 0, OSRoundUp32B(sizeof(ESTicketView)));
ret = ES_GetTicketViews(titleId, NULL, ticketViewCount);
if (ret != ES_ERR_OK) {
OSReport("%s::%s: ES_GetTicketViews failed: %d for %016llx\n", __FILE__, __FUNCTION__, ret, titleId);
goto cleanup;
}
if (*ticketViewCount == 0) {
return ret;
}
ticketViewList = (ESTicketView*)heap->alloc(*ticketViewCount * OSRoundUp32B(sizeof(ESTicketView)), -DEFAULT_ALIGN);
ret = ES_GetTicketViews(titleId, ticketViewList, ticketViewCount);
if (ret != ES_ERR_OK) {
OSReport("%s::%s: ES_GetTicketViews failed: %d for %016llx\n", __FILE__, __FUNCTION__, ret, titleId);
goto cleanup;
}
for (u32 j = 0; j < *ticketViewCount; j++) {
memcpy(ticketViews, (u8*)ticketViewList + j * sizeof(ESTicketView), sizeof(ESTicketView));
ret = ES_DeleteTicket((ESTicketView*)ticketViews);
if (ret != ES_ERR_OK) {
OSReport("%s::%s: ES_DeleteTicket failed: %d for %016llx\n", __FILE__, __FUNCTION__, ret,
((ESTicketView*)ticketViews)->ticketId);
}
}
cleanup:
if (ticketViewList != NULL) {
heap->free(ticketViewList);
}
return ret;
}
static inline s32 InitSavedata(EGG::Heap* heap) {
u8 ticketViews[OSRoundUp32B(sizeof(ESTicketView))] ALIGN32;
NANDFileInfo fileInfo ALIGN32;
u8 ticketViews[0xe0] ALIGN32;
u32 ticketViewCount;
ESTicketView* ticketViewList;
ESTitleId* titleIds = NULL;
u32 titleCount = 0;
s32 ret = ES_ListTitlesOnCard(NULL, &titleCount);
if (ret == ES_ERR_OK) {
titleIds = (ESTitleId*)heap->alloc(OSRoundUp32B(titleCount * sizeof(ESTitleId)), -DEFAULT_ALIGN);
if (ret != ES_ERR_OK) {
OSReport("%s::%s: Failed to ES_ListTitlesOnCard1: %d\n", __FILE__, __FUNCTION__, ret);
goto cleanup;
}
titleIds = (ESTitleId*)heap->alloc(OSRoundUp32B(titleCount * sizeof(ESTitleId)), -DEFAULT_ALIGN);
if (titleIds == NULL) {
OSReport("%s::%s: Unable to allocate\n", __FILE__, "InitSavedata");
} else {
ret = ES_ListTitlesOnCard(titleIds, &titleCount);
if (ret == ES_ERR_OK) {
u32 titleIdOffset = 0;
for (u32 i = 0; i < titleCount; i++) {
u32 titleIdHi = *(u32*)((u8*)titleIds + titleIdOffset);
u32 titleIdLo = *(u32*)((u8*)titleIds + titleIdOffset + 4);
if ((titleIdLo & 0xffffff00) == 0x525a4400 && titleIdHi == 0x10000) {
u8* saveData = NULL;
BOOL fileOpen = FALSE;
BOOL deleteSaveData = FALSE;
sprintf(path - 8, "/title/%08x/%08x/data/%s", titleIdHi & 0xffffff, titleIdLo, "zeldaTp.dat");
if (!ChangeUid(((ESTitleId)titleIdHi << 32) | titleIdLo)) {
OSReport("%s::%s: ChangeUid failed\n", __FILE__, "verifySavedataZD");
DeleteTitle(heap, ((ESTitleId)titleIdHi << 32) | titleIdLo);
} else {
ret = NANDPrivateOpen(path - 8, (NANDFileInfo*)((u8*)&fileInfo - 0x20), NAND_ACCESS_READ);
if (ret == NAND_RESULT_NOEXISTS) {
OSReport("%s::%s: Does not exist %s: %d\n", __FILE__, "verifySavedataZD", path - 8, ret);
} else if (ret != NAND_RESULT_OK) {
OSReport("%s::%s: Open save data file failed: %d\n", __FILE__, "verifySavedataZD", ret);
} else {
fileOpen = TRUE;
saveData = (u8*)heap->alloc(0x4000, -DEFAULT_ALIGN);
if (saveData == NULL) {
OSReport("%s::%s: Alloc failed: %d\n", __FILE__, "verifySavedataZD", -2);
} else {
memset(saveData, 0, 0x4000);
ret = NANDRead((NANDFileInfo*)((u8*)&fileInfo - 0x20), saveData, 0x4000);
if (ret < 0) {
OSReport("%s::%s: Read file failed: %d\n", __FILE__, "verifySavedataZD", ret);
} else if (ret != 0x4000) {
OSReport("%s::%s: File size is not correct: %d\n", __FILE__, "verifySavedataZD", ret);
deleteSaveData = TRUE;
} else {
u32 offset = 8;
u32 j = 0;
BOOL valid = FALSE;
while (j < 3) {
u8* block = saveData + offset;
if (!checkForNullTermination((char*)block + 0x4e, 8) ||
!checkForNullTermination((char*)block + 0x58, 8) ||
!checkForNullTermination((char*)block + 0x72, 8) ||
!checkForNullTermination((char*)block + 0x8e, 8) ||
!checkForNullTermination((char*)block + 0x1b4, 0x11) ||
!checkForNullTermination((char*)block + 0x1c5, 0x11)) {
goto verify_failed;
}
j++;
offset += 0xa94;
}
offset = 0x2008;
j = 0;
while (j < 3) {
u8* block = saveData + offset;
if (!checkForNullTermination((char*)block + 0x4e, 8) ||
!checkForNullTermination((char*)block + 0x58, 8) ||
!checkForNullTermination((char*)block + 0x72, 8) ||
!checkForNullTermination((char*)block + 0x8e, 8) ||
!checkForNullTermination((char*)block + 0x1b4, 0x11) ||
!checkForNullTermination((char*)block + 0x1c5, 0x11)) {
goto verify_failed;
}
j++;
offset += 0xa94;
}
valid = TRUE;
verify_failed:
if (!valid) {
OSReport("%s::%s: Verify failed for %016llx\n", __FILE__, "verifySavedataZD",
((ESTitleId)titleIdHi << 32) | titleIdLo);
deleteSaveData = TRUE;
}
}
NANDClose((NANDFileInfo*)((u8*)&fileInfo - 0x20));
fileOpen = FALSE;
if (deleteSaveData) {
DeleteSavedata(((ESTitleId)titleIdHi << 32) | titleIdLo, heap);
}
}
}
}
if (saveData != NULL) {
heap->free(saveData);
}
if (fileOpen) {
NANDClose((NANDFileInfo*)((u8*)&fileInfo - 0x20));
}
ChangeUid(SYSMENU_TITLE_ID);
continue;
}
if ((((ESTitleId)titleIdHi << 32) | titleIdLo) == 0x0001000844495343ULL ||
((((ESTitleId)titleIdHi << 32) | titleIdLo) < 0x0001000844495343ULL &&
((((ESTitleId)titleIdHi << 32) | titleIdLo) == 0x000100014a4f4449ULL ||
((((ESTitleId)titleIdHi << 32) | titleIdLo) > 0x000100014a4f4449ULL &&
(((ESTitleId)titleIdHi << 32) | titleIdLo) == 0x0001000148415858ULL))) ||
((((ESTitleId)titleIdHi << 32) | titleIdLo) > 0x0001000844495343ULL &&
((((ESTitleId)titleIdHi << 32) | titleIdLo) == 0x0001000844564458ULL ||
((((ESTitleId)titleIdHi << 32) | titleIdLo) > 0x0001000844564458ULL &&
(((ESTitleId)titleIdHi << 32) | titleIdLo) == 0x000100084449534bULL)))) {
ESTitleId titleId = *(ESTitleId*)((u8*)titleIds + titleIdOffset);
ES_DeleteTitle(titleId);
ticketViewList = NULL;
memset(ticketViews, 0, sizeof(ticketViews));
ret = ES_GetTicketViews(titleId, NULL, &ticketViewCount);
if (ret != ES_ERR_OK) {
OSReport("%s::%s: ES_GetTicketViews failed: %d for %016llx\n", __FILE__, "DeleteTicketsForce", ret, titleId);
} else if (ticketViewCount != 0) {
ticketViewList = (ESTicketView*)heap->alloc(ticketViewCount * 0xe0, -DEFAULT_ALIGN);
ret = ES_GetTicketViews(titleId, ticketViewList, &ticketViewCount);
if (ret != ES_ERR_OK) {
OSReport("%s::%s: ES_GetTicketViews failed: %d for %016llx\n", __FILE__, "DeleteTicketsForce", ret, titleId);
} else {
for (u32 j = 0; j < ticketViewCount; j++) {
memcpy(ticketViews, (u8*)ticketViewList + j * sizeof(ESTicketView), 0xd8);
ret = ES_DeleteTicket((ESTicketView*)ticketViews);
if (ret != ES_ERR_OK) {
OSReport("%s::%s: ES_DeleteTicket failed: %d for %016llx\n", __FILE__, "DeleteTicketsForce", ret,
((ESTicketView*)ticketViews)->ticketId);
}
}
}
}
if (ticketViewList != NULL) {
heap->free(ticketViewList);
}
}
titleIdOffset += 8;
OSReport("%s::%s: Unable to allocate\n", __FILE__, __FUNCTION__);
goto cleanup;
}
} else {
OSReport("%s::%s: Failed to ES_ListTitlesOnCard2: %d\n", __FILE__, "InitSavedata", ret);
}
}
} else {
OSReport("%s::%s: Failed to ES_ListTitlesOnCard1: %d\n", __FILE__, "InitSavedata", ret);
ret = ES_ListTitlesOnCard(titleIds, &titleCount);
if (ret != ES_ERR_OK) {
OSReport("%s::%s: Failed to ES_ListTitlesOnCard2: %d\n", __FILE__, __FUNCTION__, ret);
goto cleanup;
}
{
u8* ticketScratch = ticketViews;
for (u32 i = 0; i < titleCount; i++) {
if ((titleIds[i] & 0xFFFFFFFFFFFFFF00ULL) == 0x00010000525A4400ULL) {
verifySavedataZD(heap, titleIds[i], &fileInfo);
continue;
}
switch (titleIds[i]) {
case 0x0001000844495343ULL:
case 0x000100014A4F4449ULL:
case 0x0001000148415858ULL:
case 0x0001000844564458ULL:
case 0x000100084449534BULL:
ES_DeleteTitle(titleIds[i]);
DeleteTicketsForce(heap, titleIds[i], ticketScratch, &ticketViewCount);
break;
}
}
}
cleanup:
if (titleIds != NULL) {
heap->free(titleIds);
}
return;
return ret;
}
s32 ESMisc::DeleteUnauthorizedData(EGG::Heap* heap) {
return InitSavedata(heap);
}
BOOL checkForNullTermination(char* str, u32 len) {

View File

@@ -29,3 +29,26 @@ Additional Decrypt correction: restored 0x3E800 size limit, actual seek/read err
Remaining: CDBCryptBuffer 99.97479%, 119/119 instructions, three pool offsets; CDBRecordEncrypt 97.19014%, register allocation and one instruction; CDBRecordDecrypt 99.96789%, stack placement.
The four forced field accessors end with CDBUnlock and do not produce a CDBErr result. Their void declarations are gated by CDB_RECORD_IMPLEMENTATION, preserving all other translation units and avoiding fall-through non-void definitions.
## w1008/cdb leaf — dead-fn fossil + Encrypt coloring (worker w1008)
SOLVED — link-deadstrip fossil mechanism: orig .data's fossil literals (0x398-0x9be)
come from ~15 dead functions whose code mwldeppc stripped at link. Added all dead
fns (CDBRecordGetFileSize through CDBRecordPrivateChangeOwner) as GLOBAL fns in
fossil-emission order; global (not static) emits standalone code AND auto-inlines
at use sites. .data content byte-identical (orig has 2B reconstruction tail pad).
Verified at DOL level: fossil strings land at exact expected vaddrs.
SOLVED — Encrypt +0x40 code gap: `err = CDBRecordGetFileSize(record, &fileSize)`
calls the dead fn which auto-inlines, reproducing orig's expanded block.
`(s32)ptr == 0` → `cmpwi r0,0` required (== NULL gives cmplw). 284/284 insns
identical, 0 stream diffs.
WALL — CDBRecordEncrypt 57 byte diffs, pure callee-reg coloring on identical
web structure. Orig packs {key+err}→r24, {tempFile+fileOffset}→r31, pool→r25,
args→r26-29; mine packs {tempFile+err}→r24, {key+fileOffset}→r29, pool→r31,
args→r25-28. Equal-cost packings (tie-break). ~18 probes all neutral/regressed:
recordFile/err/tempFile/fileOffset decl order+scope+init forms, per-block vs
shared `result`, single-web fileOffset (-69), fn-scope tempFile, `register`,
err dead-init (emits, +4B/831), u32/s32 types, decl+init forms.
DOL-verified: the entire DOL diff vs orig = exactly these 57 bytes.

View File

@@ -0,0 +1,41 @@
# CHANSVm load-constant error preservation
Baseline: 449529d92405f8f7fda610660c7794914a81dfe6. Scope: CHANSVm.c only; no linking classifications or headers changed.
## Retained correction
LOAD_STRING_CONST must preserve a failure from CHANSVmDeleteObject. The original object tests the deletion return and branches directly to the local result merge when it is nonzero. The load-string error is selected only after deletion succeeds and the constant index is out of range or has no string payload. The prior C instead overwrote every deletion error with CHANS_VM_ERR_LOAD_STRING_CONST.
A native destructor that returns false is a concrete error-producing path through CHANSVmDeleteObject; this distinction is observable, not merely a compiler-scheduling preference.
The new local loadResult preserves that error, keeps constant validation and installation inside the successful-deletion block, and merges the result afterward. Target and rebuilt blocks have the same instruction count and control-flow destinations; remaining operand differences are registers. No strings, data layout, or generated jump-table destinations changed.
## Measurements
- All 233 target functions have compiled C implementations; this is not a claim that all are exact or independently behavior-proven.
- Instruction-exact functions: 221/233 -> 221/233; every objdiff-exact function also passes ctxdiff.
- Unit decompiled/fuzzy: 99.364426% -> 99.383840%.
- Exact code: 37608/53564 bytes -> 37608/53564 bytes (70.211334%).
- Exact data: 6904/6904 bytes -> 6904/6904 bytes (100%).
- CHANSVmStep: 96.466080% -> 96.673584%; 1253/1253 instructions; ctxdiff differing positions 314 -> 297.
- Per-function score regressions relative to the fresh baseline report: zero.
- The unit remains NonMatching/unlinked; no link flags were changed.
## Attempts
1. Read previous round/wave logs and recent local trials; establish live baseline and run pool_diff first: 125 strings identical.
2. Replace the load-string shared error/goto path with a typed local error result and successful-deletion-only constant validation. Retained: correct error behavior, higher fuzzy score, unchanged exact code/data.
3. Separately tried consolidating GET_PROPERTY_NAME's local found flag and removing two unused array-index assignments. Rejected and restored completely: Step shrank to 1249 instructions and data dropped to 2232/6904. This unrelated trial is absent from the final source.
## Validation
- Full 43U build passed with the assigned wibo wrapper and WIBO_SJIS_MISSING_IMPORTS=1.
- DOL SHA1: 26116613f624061ba99c8d1a299aaa6efa85670d.
- pool_diff: 125/125 identical strings.
- .data: 4672/4672 bytes identical; all 352 canonical relocations identical.
- .rodata: 1432/1432 bytes identical; all 258 canonical relocations identical.
- All 233 functions compared with ctxdiff/odiff: 221 exact; no falsely exact objdiff functions.
- git diff --check passed.
- Source-extracted host C harness passed six load-constant cases: missing operand; deletion failure with valid index; deletion failure with invalid index; invalid index after successful deletion; null payload; successful installation. The harness stubs surrounding VM dependencies and checks C behavior only; no Wii runtime execution is claimed.
Source is paused after this focused candidate for parent review. The remaining 12 functions are still non-exact.

View File

@@ -0,0 +1,49 @@
# CHANSVm opcode error-path reconstruction
Baseline: 2f399716. This is a separate structural-reconstruction candidate; it preserves the preceding LOAD_STRING_CONST behavior fix. Only CHANSVm.c changed apart from this log. No headers, linking flags, remote branches, or original assets were changed.
## Retained changes and target evidence
- Failed binary-operand conversion returns CHANS_VM_ERR_RESULT_TYPE directly. Target Step relative +0x410 selects the error and +0x414 returns through +0x1364.
- Invalid GET_PROPERTY_NAME and JUMP destinations return CHANS_VM_ERR_CODE_RANGE directly. Target error tails are +0xd90/+0xd94 and +0xefc/+0xf00, respectively, both ending at +0x1364.
- DELETE_INDIRECT has a typed accumulator and a local deletion result, and reads the index through value.data.len rather than an offset-based u32 pointer cast. SET_INDEX elsewhere in the same function establishes this field as the index-reference index. The target holds the opcode-local result until the merge at +0xfb0, then assigns the common interpreter status. The rebuilt +0xf80 through +0xfb4 block is instruction-identical to the target.
These preserve existing C behavior. Unlike the preceding deletion-error fix, no new behavioral defect is claimed.
## Measurements
- Exact functions: 221/233 -> 221/233. Every previously exact function remains instruction-exact, not merely objdiff-exact.
- Decompiled/fuzzy unit: 99.383840% -> 99.406320%.
- CHANSVmStep: 96.673584% -> 96.913800%; 1253/1253 instructions; differing positions 297 -> 285.
- Exact code: unchanged, 37608/53564 bytes (70.211334%).
- Exact data: unchanged, 6904/6904 bytes (100%).
- Per-function score regressions: zero.
- Unit remains NonMatching/unlinked; twelve functions remain non-exact.
## Rejected investigations
All rejected source changes were restored, and no register forcing or declaration-order sweep was used.
- GET_PROPERTY_NAME output-state separation: a local found flag, initialized flag, shared index/output, and inline helper. These produced 1251-1253 instructions but moved generated data-table destinations; rejected.
- Direct result-type-table accesses, an explicit zero-count normalization/pretested loop, and indexed operand-type enumeration: 1254-1257 instructions; generated data-table destinations moved; rejected.
- Internal linkage for result-type or constant-header tables, individually and together, did not change those results; restored.
- CASE strict-comparison result separation, case-local match state, byte-sized operand types, and indexed operand enumeration: improved some local structure but introduced an operand-type spill and 1254-1255 instructions; generated tables moved; rejected.
- Direct exceptional returns alone: retained for combination, 96.697525%, data 100%.
- Typed DELETE_INDIRECT and local result alone: retained for combination, 96.889860%, data 100%.
- Combined direct returns and typed deletion result: accepted candidate, 96.913800%, data 100%.
- A first combination script used an obsolete substring end offset and did not compile; restored, corrected, and rerun. No source from the malformed attempt remains.
- Adding strict-comparison restructuring to the retained combination still produced 1254 instructions and data regression; rejected.
## Validation
- Full 43U build passed with the assigned wrapper and WIBO_SJIS_MISSING_IMPORTS=1.
- DOL SHA1: 26116613f624061ba99c8d1a299aaa6efa85670d.
- pool_diff: 125/125 strings identical.
- .data: 4672/4672 raw bytes and 352/352 canonical relocations identical.
- .rodata: 1432/1432 raw bytes and 258/258 canonical relocations identical.
- All 233 target functions compared: 221 instruction-exact; zero falsely exact objdiff functions; zero per-function score regressions from 2f399716.
- Narrow-string-reference audit: all 233 functions analyzed, 40 arguments checked, no candidates/errors/skips.
- Source-extracted host C tests: four DELETE_INDIRECT cases passed (wrong type, successful deletion, deletion failure, and null element forwarding), including exact unsigned index forwarding. The six prior LOAD_STRING_CONST cases also passed unchanged. These stub surrounding VM dependencies and are not Wii runtime tests.
- git diff --check passed.
The focused candidate is paused for independent parent review.

View File

@@ -0,0 +1,52 @@
# CHANSVm typed comparison reconstruction
Baseline: ec75ddb0. This candidate is separate from the accepted load-string and opcode-error-path work. Those commits and their logs are unchanged. Only CHANSVm.c and this log changed; no headers, linking classifications, metadata, or original assets changed.
## Retained reconstruction
- LOAD_FLOAT copies the eight-byte literal into a double scratch member and passes that value directly. This replaces a u64 temporary followed by a double-pointer reinterpretation. Scratch alignment, member offsets, and emitted code are unchanged.
- SET_INDEX reads the floating operand through its union member and checks its lower bound before its upper bound, matching the target's comparison order. The accepted interval remains [0, 4294967294], with NaN and infinities rejected.
- Strict integer equality uses the declared 64-bit integer values rather than four offset-based u32 loads and manual XOR/OR expressions.
- Strict floating equality uses the declared floating values directly.
- Strict string equality uses named, read-only payload pointers rather than recreating a header pointer. The stack operand is consistently first in the outer type comparison, following the target.
This is structural/type reconstruction, not a newly discovered behavior fix. No register forcing, declaration sweep, or alignment padding was introduced.
## Target evidence
After rebuilding, the integer-comparison block (Step instructions 1093-1102), floating-comparison block (1103-1108), string-comparison block (1109-1128), and floating-index block (703-712) have the same instructions, control flow, constants, and memory offsets as the target after register names are normalized. This comparison is advisory: register differences remain, and Step is not exact.
The floating range's instruction forms now use the target's lower-bound-first order. Integer equality emits the same target ten-instruction two-word equality sequence naturally from the 64-bit value comparison. LOAD_FLOAT retains its original eight-byte copy and call sequence without a pointer reinterpretation.
## Measurements
- Instruction-exact functions: 221/233 -> 221/233; no previously exact function regressed.
- Unit Decompiled/fuzzy: 99.406320% -> 99.413185%.
- Step: 96.913800% -> 96.987230%; 1253/1253 instructions; differing positions 285 -> 276.
- Exact code: unchanged at 37608/53564 bytes (70.211334%).
- Exact data: unchanged at 6904/6904 bytes (100%).
- Per-function score regressions: zero.
- Unit remains NonMatching/unlinked, with twelve non-exact functions.
## Attempts
- Direct 64-bit equality alone: Step 96.933760%, data 100%; retained for combination.
- Typed integer/float/string comparison paths: 96.971270%, data 100%; retained.
- Those typed paths plus a separate strict-comparison local error result: 1254 instructions, table relocations moved; rejected.
- Floating index member access alone preserves the typed-comparison score. Lower-bound-first range checking raises it to 96.983240%, data 100%; retained.
- A union literal temporary produced the same code; simplified to a direct double member instead, also with identical code.
- Stack-first type comparison raises Step to 96.987230% with unchanged data; retained.
## Verification
- Full 43U build passed with the assigned wrapper and WIBO_SJIS_MISSING_IMPORTS=1.
- DOL SHA1: 26116613f624061ba99c8d1a299aaa6efa85670d.
- pool_diff: all 125 strings identical.
- .data: 4672/4672 raw bytes and 352/352 canonical relocations identical.
- .rodata: 1432/1432 raw bytes and 258/258 canonical relocations identical.
- All 233 functions checked with ctxdiff/odiff: 221 instruction-exact; no falsely exact objdiff functions.
- Literal-reference audit: 233 functions analyzed, 40 resolvable call-string arguments checked, no candidates/errors/skips.
- Source-extracted host C harnesses passed 20 strict-comparison cases, nine floating-index-bound cases, and eight floating-literal bit patterns. Coverage includes signed integer extremes, differing high words, signed zero, NaN, infinities, strings, references, deletion errors, and literal payload preservation. They stub surrounding dependencies and do not constitute Wii runtime execution.
- git diff --check passed.
Source is paused for independent parent verification.

View File

@@ -0,0 +1,83 @@
# AES final-round word reconstruction, 2026-10-02
Base: `813660b9` plus local cherry-pick `97035641` of the parent's already queued
`fce31442`. The imported candidate and its log are unchanged. This follow-on
changes only two final-round expression lines in `libs/RevoEX/src/net/aes.c`
and this new log. 43U only; no configuration, metadata, headers or linking edits.
## Retained reconstruction
The last cipher round substitutes four bytes, places them in distinct unsigned
word lanes, and combines them before AddRoundKey. Encryption combines the high
and low lanes, combines the two middle lanes, then ORs those pieces. Decryption
combines high/low first, then the lower-middle and upper-middle lanes by XOR.
These expressions preserve byte indices, shifts, S-box identity, key traversal,
output order and all input/state reads. The four lane contributions are disjoint;
all operands are unsigned and have no side effects. No value, algorithmic path,
or table is newly invented. This is arithmetic-dependency reconstruction rather
than a new implementation or an exact-match claim.
The target evidence is the original split object's final substitution and
packing regions in `AESiEncryptBlock` and `AESiDecryptBlock`. Those regions
contain the same sixteen byte loads and word-lane insertion/XOR operations.
The retained grouping improves the whole-function instruction similarity at
the target's existing body sizes. It does not reproduce every target dependency
or register assignment: exact matching remains open, and positional ctxdiff
counts are slightly worse despite the improved aligned objdiff scores.
## Explored alternatives
- MD5 third-round Boolean grouping and four typed step helpers were tested.
The best small Boolean change only reached 94.37255% from 94.30719%; a larger
helper reached 94.43791%. Both were discarded as too little benefit for the
added change. MD5 source is byte-for-byte unchanged.
- Typed inverse-key word, round and complete-schedule helper scopes were tested.
A round helper reached 69.72908% decrypt similarity, but interacted poorly
with the better final-round grouping. No helper is retained.
- Fusing AddRoundKey into each intermediate column regressed both directions.
- Typed final-substitution helpers, packed-byte intermediates, and explicit
T-table column helpers were evaluated. Their source complexity was unnecessary
or they scored below the two retained macro changes.
- Twelve unsigned lane-combination expressions were checked for each final-round
macro. Two independently improving forms are retained. No register/declaration
sweep, assembly, volatile qualifier, undefined value, padding or compiler-option
change was used.
## Verification
Before -> after:
- AESiEncryptBlock: 62.58228 -> 65.81013% fuzzy; 158/158 instructions unchanged;
final ctxdiff 137 positional differences (baseline 133).
- AESiDecryptBlock: 67.64143 -> 69.69323% fuzzy; 251/251 instructions unchanged;
final ctxdiff 227 positional differences (baseline 224).
- AES unit: 79.601746 -> 81.09157% fuzzy; exact functions 7/9 and exact code
1116/2752 bytes unchanged. Data 2800/2800 (100%) and linking unchanged.
- All seven exact AES APIs/CBC functions retain objdiff 100% and ctxdiff diffs 0:
NETAESCreateEx 108/108, NETAESCreate 2/2, NETAESDelete 1/1, NETAESEncrypt 46/46,
NETAESDecrypt 46/46, NETiAESEncryptoBlock 39/39, NETiAESDecryptoBlock 37/37.
- AES pool: all four strings identical. .rodata 2592 bytes and .sdata2 8 bytes
identical. .data has the same 197-byte prefix of the 200-byte target, retaining
the existing three-byte target alignment extent; no padding was introduced.
- MD5 remains unit 96.179825%, ProcessBlock 94.30719% (306/306 instructions),
exact functions 3/4, exact code 600/1824, data 456/456. Its three exact APIs
retain ctxdiff diffs 0. All 448 .data bytes are identical; .sdata retains its
existing one-byte source/seven-byte target-alignment extent difference.
- Full 1027-unit report: no exact-function, exact-code, data or fuzzy regression
versus the locally imported parent candidate.
- Full 43U build and build/43U/ok passed. DOL SHA1 remains
`26116613f624061ba99c8d1a299aaa6efa85670d`.
Host tests compile production source unchanged with private copies of the existing
PPC type/intrinsic shims. Both final AES and the imported parent baseline passed
all 2,406 AES-128/192/256 known-answer and seeded randomized CBC checks, including
separate/in-place buffers and single/incremental calls. Unchanged MD5 passed all
3,088 known-answer/random/chunking checks. The MD5 test supplies RFC1321 padding
externally; GetDigest's target-endian length stores remain outside host coverage.
Private local evidence: `/tmp/crypto-reconstruction-next/` contains baseline/final
reports, pool/section/ctxdiff results, full-build.log, host validation scripts and
results, and the rejected experiments. No retail assembly or binary is committed
or uploaded. Prior worker logs under `/tmp/crypto-remainder/` are untouched.
GATE: validated partial arithmetic reconstruction with no report regressions.
Exact matching and linking remain open. Candidate frozen for parent validation.

View File

@@ -0,0 +1,78 @@
# AES key-schedule cursor reconstruction, 2026-10-02
Source baseline: `449529d9`. This separate candidate follows the frozen MD5
commit `521056a9`; it changes only `libs/RevoEX/src/net/aes.c` and this log.
43U only. No linking flags, configuration, tables, or public API edits.
## Target evidence and retained change
`AESiDecryptBlock` in `build/43U/obj/libs/RevoEX/src/net/aes.o` is 0x3EC bytes /
251 instructions. Its entry first forms the key-schedule base at context+0x10
(function-relative offset 0x18), extracts the last-round offset from the flags
word (offset 0x24), and adds that offset to the already formed base (offset
0x28). The baseline combined initializer produced 249 instructions. Explicitly
initializing the typed key pointer to `context->keys` before advancing it by
`rounds*4` recovers the target's two-stage pointer construction and body size.
The retained change is two added source lines and one removed line. All AES
round expressions, lazy inverse key transformation, tables and CBC chaining
remain unchanged.
This is a structural matching improvement, not a newly implemented AES path.
Both existing block algorithms were already present and pass the behavior tests.
## Explored and rejected alternatives
- Inlined single-word round helpers, byte-typed table lookups and table-rotation
helpers changed scheduling but regressed encryption or added source complexity.
- Staging four inverse-mix column outputs through a typed round helper improved
decrypt similarity but required substantially more source and still did not
match. Not retained.
- Alternate inverse-key-transform local lifetimes and an inverse-key helper
gave only small extra fuzzy movement, without new algorithm reconstruction.
- Context-relative key indexing and row/word loop variants did not improve the
minimal key-base candidate.
- Moving next-state locals outside the round loop did not change codegen.
- Reordering key-base declarations produced a negligible score change; the
original natural declaration order was kept.
- Two exploratory C89-incompatible mixed-declaration variants failed to compile
and were immediately discarded. No failed source is retained.
## Verification
Before -> after:
- AESiDecryptBlock: 249 -> 251 instructions against target 251;
fuzzy 61.49004 -> 63.836655%. Final ctxdiff: 224 positional differences.
- AESiEncryptBlock: unchanged, 158/158 instructions and 60.348103% fuzzy.
- Unit fuzzy: 76.844475 -> 77.700584%.
- Exact functions: 7/9 -> 7/9. Exact code: 1116/2752 bytes unchanged.
- Data: objdiff 2800/2800, 100%, unchanged. Linked status unchanged.
- All seven exact functions have ctxdiff diffs 0: NETAESCreateEx 108/108,
NETAESCreate 2/2, NETAESDelete 1/1, NETAESEncrypt 46/46, NETAESDecrypt 46/46,
NETiAESEncryptoBlock 39/39, NETiAESDecryptoBlock 37/37.
- String pool: all four strings identical.
- .rodata: 2592 identical bytes; .sdata2: 8 identical bytes. Source .data is
197 bytes with the same prefix as the 200-byte target section; three target
alignment bytes are an existing extent difference. No padding was added.
- The full 1027-unit report has no regressions in exact functions/code/data or
fuzzy score against the baseline report. Frozen MD5 metrics are unchanged.
- Full 43U build passed with the configured wibo wrapper; build/43U/ok passed.
DOL SHA1: `26116613f624061ba99c8d1a299aaa6efa85670d`.
Host validation: `python3 /tmp/crypto-remainder/validate_aes.py` compiles the
production C unchanged with minimal types/OSReport shims. Input, key and IV
words are represented with their PPC big-endian numeric values, so native host
byte order does not change the algorithm. Known-answer tests cover AES-128,
AES-192 and AES-256; seeded randomized CBC tests compare against cryptography's
AES implementation for all three key sizes. Each tests separate/in-place
buffers and single/incremental calls.
Result: all 2,406 production-source known-answer/randomized checks passed. The
original baseline had independently passed the same coverage before changes.
A separate saved helper experiment also passed but is not part of this commit.
Local evidence, not committed or uploaded: `/tmp/crypto-remainder/aes-final.ctx`,
`aes-final.json`, `aes-full-build.log`, `aes-vectors.log`, `aes-decrypt.target`,
and `validate_aes.py` plus its header shims. No retail assembly is included.
GATE: validated partial structural candidate; no regressions. Exact matching
and linking remain open.

View File

@@ -0,0 +1,71 @@
# AES round-expression reconstruction, 2026-10-02
Separate follow-on to AES key-cursor candidate `2a537c14`. Frozen MD5 candidate
`8a6461f9` and prior logs are untouched. This commit changes five expression
lines in `libs/RevoEX/src/net/aes.c`, plus this log. 43U only; no linking changes.
## Evidence and reconstruction
Target evidence remains the original split `aes.o`, AESiEncryptBlock and
AESiDecryptBlock. Both round bodies use four cyclic T-table columns: rotations
of 24, 16 and 8 bits plus the unrotated low-byte column. Their XOR dependencies
and the inverse-key transform's successive rotate/XOR dependencies lower in a
different order from the previous source. The retained expressions associate
these same unsigned contributions so compiler scheduling moves closer to the
target. Byte indices, rotation counts, table identity, key traversal, row order,
loop conditions and final substitution are unchanged.
The inverse-key transform still computes doubled/four-times/eight-times byte
values in the same GF(2^8) representation. Each subsequent stage combines its
old mixed value, its 8-bit rotation and the same added contribution. Explicit
XOR assignment replaces the compound assignment; all RHS values are evaluated
before the assignment, with no side effects or undefined evaluation order.
No source body or algorithm was missing. This is further instruction-shape
reconstruction, not a new functional path or an exact-match claim.
## Attempts
- Eight staged column/round helpers and six explicit table-load stages did not
improve both functions; reverted.
- Tested 120 associative/permutation forms of four T-table contributions,
retaining the independent improving expression for each cipher direction.
- Tested twelve uniform inverse-key XOR forms plus per-stage refinements. The
retained rotate-first XOR form improves all three stages without new helpers.
- Array/struct state views, a 15-by-4 round-key view, initial-key cursor forms,
and typed table helpers were evaluated. Small gains that required larger
changes were rejected, along with all regressions.
- No register declaration sweep, inline assembly, volatile qualifier, padding,
undefined input, compiler-option change or table change was used.
## Verification
Before -> after:
- AESiEncryptBlock fuzzy 60.348103 -> 62.58228%; 158/158 instructions unchanged;
final ctxdiff has 133 positional differences.
- AESiDecryptBlock fuzzy 63.836655 -> 67.64143%; 251/251 instructions unchanged;
final ctxdiff has 224 positional differences.
- Unit fuzzy 77.700584 -> 79.601746%.
- Exact functions 7/9, exact code 1116/2752 bytes, data 2800/2800 (100%), and
linked status are unchanged. All seven exact functions retain ctxdiff diffs 0.
- All four pool strings match. Table and data source is unchanged.
- Frozen MD5 source/metrics are unchanged: unit96.179825%, ProcessBlock94.30719%.
- The full 1027-unit report has zero exact-function/code/data/fuzzy regressions
against the immediately preceding frozen MD5/AES source report.
- Full 43U build and build/43U/ok passed. DOL SHA1 remains
`26116613f624061ba99c8d1a299aaa6efa85670d`.
`python3 /tmp/crypto-remainder/validate_aes.py` passed all 2,406 production-source
known-answer/randomized AES-128/192/256 CBC checks, including in-place buffers
and incremental calls. The frozen AES parent source independently passed the
same coverage in the same run. Input/key/IV words carry their PPC big-endian
numeric values in the native host harness; production source is unmodified.
Local evidence, not committed or uploaded: `/tmp/crypto-remainder/` contains
`aes.xor-results.json`, `aes.invmix-results.json`, `aes-rounds-final.json`,
`aes-rounds-AESiEncryptBlock.ctx`, `aes-rounds-AESiDecryptBlock.ctx`,
`aes-rounds-vectors.log`, `aes-rounds-full-build.log`, and `validate_aes.py`.
No original retail assembly or binary is included in this commit.
GATE: validated partial arithmetic/data-flow candidate; no regressions. Exact
matching and linking remain open.

View File

@@ -0,0 +1,62 @@
# BS2Tick control-flow reconstruction, 2026-10-02
Baseline BS2Mach source is 449529d9. This candidate changes only BS2Tick in
src/BS2/BS2Mach.c. Existing ESMisc and AOSSLink commits remain separate.
Initial pool_diff: 91/91 strings identical. Read the BS2 round-two/round-three
and fz1 attempt records, and compared the current five open functions with
fresh target objects. The Run hardware-handoff deficit is separate and is not
changed or represented as complete here.
## Retained source reconstruction
1. Express the title-family dispatch as the actual sparse set D/R/S/T.
The target compares D below R, then dispatches the contiguous R-through-T
group using an upper boundary U. The switch restores its branch pair and
makes the family set explicit. Exhaustively checked all 256 prefix values.
2. Restore the positive game-partition state transition and capture the
partition offset at entry to state 0x25, before either cache or disc path.
The target loads the partition member before choosing the cache branch;
the old source only read it on the disc path. The incoming partition
existence check and the state-37 fallthrough remain explicit.
3. Restore the banner address diagnostic's missing variadic argument.
Target BS2Tick object offsets 0x39B4–0x39D8 compute the aligned banner
pointer in r4, store it as BannerBuffer and pass it to BS2Report. The old
source supplied the %08X format with no argument on the allocation path.
4. Express the same next-cache-line calculation as allocation + 32 minus its
remainder modulo 32. This retains the target's strictly-next-line behavior,
including when the allocation is already aligned. It does not add padding
storage. Checked 196608 address samples including all low 16-bit values at
zero, 0x80000000 and near u32 wrap.
Measured attempts, all retained and individually nonregressing:
| Attempt | BS2Tick fuzzy | Unit fuzzy |
| --- | ---: | ---: |
| Baseline | 98.230415% | 98.159485% |
| Actual title-family switch | 98.35567% | 98.21673% |
| Partition entry and typed member snapshot | 98.58763% | 98.32273% |
| Banner diagnostic argument | 98.59278% | 98.32509% |
| Explicit next-cache-line remainder | 98.639175% | 98.34629% |
BS2Tick grows from 1934 to 1937 instructions; target is 1940. The remaining
three missing instructions are address-materialization differences, with
additional register and instruction-scheduling differences. This is partial
reconstruction/matching progress, not an exact function.
## Verification
- Full all_source / fresh report / build/43U/ok requested after focused build
- DOL SHA1: 26116613f624061ba99c8d1a299aaa6efa85670d
- Exact functions unchanged: 24/29; exact code unchanged: 4940/16980
- Data unchanged: 155504/158528 (98.092445%)
- Every other unit's report unchanged
- The 91-string pool is identical; literal reference audit checks 37 arguments
across 24 exact functions with no candidates or errors
- Compared all 1027 source objects with the initial snapshot: only previously
committed ESMisc and this BS2Mach object differ. BS2Mach .text 16828 -> 16840
bytes, 1306 relocations on both sides; .data stays 3020 bytes and 155 jump-table
relocations; all other allocated sections/objects are unchanged
- git diff --check passes
No BS2Update, shared-header, metadata, link-flag or assembly change is included.

View File

@@ -0,0 +1,99 @@
# Alpha comparison and output boundary follow-on, 2026-10-02
Base: 813660b9 plus unchanged one-key candidate 2ec4a57e, locally fe647fe4.
Changed source: libs/RVLMiddleware/eZiText/src/clib/zi8alpha.c only.
No ABI, shared header, linking flag or other source change is included.
## Retained target source structure
- At function +0x25A8, the suffix truncation condition compares the combined
word/prefix length first and branches on less-or-equal. Recovered that
operand order. The separate bound check at +0x25E0 has the opposite operand
order and is deliberately preserved.
- The dictionary lookup comparisons at +0x2948, +0x2978 and +0x2990 compare
the dictionary scan index first against word length or the halfword-narrowed
dictionary count. Recovered those three relational expressions without
changing the count helper's declaration or its halfword conversion.
- The punctuation fallback at +0x30F8 compares against 0xF010 and uses a
strict greater-than failure branch. Replaced the equivalent < 0xF011 test
with <= 0xF010.
- The output-count branch at +0x3B08 falls through to clear letters when
countOnly is true, and branches to the actual-count store otherwise.
Restored that original arm order.
- Prefix selection at +0x3B84 compares against 1 and uses less-or-equal.
Recovered <= 1 instead of the equivalent < 2.
These are demonstrated control-flow and expression reconstructions, not
register/declaration permutations. No new volatile, assembly, forced register,
padding, artificial storage, or undefined value is introduced.
## Isolated trials and rejected work
Trials used the live 43U Ninja compiler command, separate source/objects and
reports under /tmp/ezi-next, without changing the worktree before selection.
Columns: unit fuzzy / matched data bytes / exact functions.
- Output-count arm order: 96.47729 / 516 / 11
- Both inclusive boundary expressions: 96.3798 / 516 / 11
- Both maximum-length comparison orders: 96.38128 / 516 / 11
- Dictionary comparison order: 96.38405 / 516 / 11
- Initial four-group combination: 96.48966 / 516 / 11
- Final reviewed combination, preserving the second length comparison's
original operand order: 96.52474 / 516 / 11
Further target-evidenced trials were rejected rather than committed:
- Safe temporary plus postfix prefix append: 94.71085 / 336 / 11
- Safe temporary plus postfix highlighted append: 94.60377 / 336 / 11
- Dictionary-kind switches: 96.364845 / 408 / 11
- Early candidate skip: 96.451996 / 408 / 11
- Combined candidate-count limits: 96.402695 / 408 / 11
- Full dictionary source structure plus initial boundary combination:
96.78767 / 408 / 11; function extent 0x3DB8 versus 0x3DA8
- Explicit prefix-mode byte: 96.40805 / 408 / 11
- Explicit element-count byte: 96.3582 / 408 / 11
- Skip/count/byte combinations retained extent mismatches and were rejected.
One-key spelling compound count assignment was codegen-neutral. Prefix
increment removed the redundant byte narrowing but worsened allocation:
unit 98.23523 -> 97.36198. Neither trial is retained.
Chinese explicit halfword casts at the two paired phonetic comparisons and
an alternative proper halfword return declaration with redundant casts removed
both reproduce the previously parked result: unit 96.848755, matched data
144 -> 72. Adding the demonstrated component-cursor reload gives 96.489876
with the same loss. All remain scratch experiments, with no Chinese edit or
neutral PUD ABI commit included.
## Validation
Fresh full report: /tmp/ezi-next-alpha-full-report.json.
Full-build log: /tmp/ezi-next-alpha-full-build.log.
Baseline report: /tmp/ezi-next-baseline.json.
Zi8AlphaGetCandidates: 95.02813% -> 95.23011%.
Unit fuzzy: 96.37759% -> 96.52474%, about 31.88 weighted bytes gained.
Exact functions remain 11/12; exact code remains 5880/21664;
matched data remains 516/564. No complete unit or new exact function claimed.
All 1027 units were compared for fuzzy, matched-code, matched-data and exact
function measures: the alpha fuzzy improvement is the sole changed measure,
with no regressions.
Pool checked first: identical and empty. Main function stays 3946/3946
instructions and 0x3DA8 bytes, with the same 0x2B0 frame. Strict ctxdiff is
nonzero (3738 differences); existing scalar-stack displacement and register
allocation remain. Each of the 11 existing exact functions retains identical
instruction counts and ctxdiff diffs 0.
Baseline/candidate .rodata, .data, extab and extabindex have identical payloads
and canonical relocations. Relative to the target, extab and extabindex remain
byte-identical; the unchanged generated switch relocations remain unmatched.
The source rodata's pre-existing 330-byte size versus target 336 is unchanged;
objdiff still reports 100% for that section and the same exact-data accounting.
Advisory branch mapping checks 709 same-mnemonic branch pairs with zero
mapped destination mismatches, and 73/73 helper calls. It is not an exactness
or formal equivalence claim.
Full 43U build passes with the approved wrapper and WIBO_SJIS_MISSING_IMPORTS=1.
DOL SHA1: 26116613f624061ba99c8d1a299aaa6efa85670d.
git diff --check passes. No remote actions were performed.

View File

@@ -0,0 +1,65 @@
# Alpha output and scan reconstruction follow-on, 2026-10-02
Parent candidate: `94b7f57e`. Its source/log were left unchanged during
independent side experiments. No ABI/header/linking changes are included.
Changed source: `libs/RVLMiddleware/eZiText/src/clib/zi8alpha.c` only.
## Target evidence and retained changes
- At `Zi8AlphaGetCandidates +0x3178`, the original emits wide-character
output before the encoded-byte alternative. The wide path assigns the
same zero to output elements 2 and 1 (+0x31A4/+0x31A8), then advances by
two characters. The encoded path likewise shares its zero assignment.
Restored that positive branch and ordinary chained zero assignments.
- The dictionary comparison loop branches from +0x2904 to its bound check
at +0x292C. The comparison body exits on a character mismatch at +0x2924;
a matched character increments the index at +0x2928. Replaced the merged
`for` condition with the corresponding bounded loop and explicit break.
- The punctuation output block at +0x1B0C..+0x1B1C reads the current cursor,
emits its character and advances that same cursor. Expressed this with
`wordCursor[1] = *punctuationCursor++`. It does not reuse a mutated index
in another operand or introduce an unsequenced read/write.
## Independently compiled, isolated trials
Each trial used the exact 43U MWCC flags from the live Ninja command and a
separate object/report under `/tmp/ezi-side`, without altering the reviewed
worktree candidate. Report columns below are function fuzzy, unit fuzzy,
matched data bytes and exact-function count:
- Early candidate-skip block: 94.65585, 96.10635, 408, 11; excluded
- Wide output first: 94.887985, 96.27548, 516, 11; retained
- Next-dictionary switches: 94.4037, 95.92264, 408, 11; excluded
- Explicit dictionary-scan break: 94.62088, 96.08087, 516, 11; retained
- Combined candidate-count limit: 94.35884, 95.88995, 408, 11; excluded
- Punctuation cursor increment: 94.4782, 95.97692, 516, 11; retained
- Wide output plus scan: 95.02306, 96.37389, 516, 11
- Final three-change subset: 95.02813, 96.37759, 516, 11
## Full current-worktree validation
Fresh report: `/tmp/ezi-alpha-followup.json`.
Full build log: `/tmp/ezi-alpha-followup-build.log`.
Exact functions 11/12 unchanged; exact code 5880/21664 unchanged;
exact data 516/564 unchanged. Unit fuzzy 95.97323% -> 96.37759%.
Function fuzzy 94.47314% -> 95.02813%, about 88 additional weighted bytes.
Relative to initial 449529d9, unit fuzzy improves 94.17892% -> 96.37759%.
Pool first: identical (no narrow strings in either pool).
Main function: 3946/3946 instructions, 0x3DA8 bytes each, frame 0x2B0.
Strict ctxdiff remains nonzero (3745 differences), dominated by the known
scalar-stack displacement, instruction placement and GPR allocation. No
new exact function is claimed. Each of the existing 11 exact functions
retains identical instruction counts and ctxdiff diffs 0.
All non-text section scores match the parent: .rodata, extab and extabindex
remain 100%; the existing 48-byte generated switch table remains unmatched.
The advisory register-normalized/control-flow alignment checked 701 branch
pairs with zero destination mismatches and 73/73 helper calls. This is a
coverage check, not an exactness claim.
Full 43U build succeeds with the approved wrapper and
`WIBO_SJIS_MISSING_IMPORTS=1`. DOL SHA1:
`26116613f624061ba99c8d1a299aaa6efa85670d`.
`git diff --check` succeeds. No new assembly, volatile workaround, register
forcing, undefined value, dummy storage, prototype or shared-header change.

View File

@@ -0,0 +1,108 @@
# One-key candidate boundary reconstruction, 2026-10-02
Parent commits: alpha 94b7f57e and 35677806, neutral Chinese ABI 105f3b45.
This pass changes only `libs/RVLMiddleware/eZiText/src/clib/zi81key.c`.
Alpha source/logs and Chinese source/log are unchanged.
## Retained target-defined boundaries
- `Zi8IsMatch1Key` consumes a halfword key. Six candidate calls narrow that
argument after Zi8GetPCode: target +0x0B0C, +0x0F30, +0x16C0, +0x1BA4,
+0x1F4C and +0x22A8. Restored the ziU16 formal and removed its three
redundant key masks. The helper itself remains 97/97 instructions,
objdiff 100%, ctxdiff diffs 0. All source call sites are in this same unit.
- Each of the six duplicate-bit checks consumes a byte result in the target.
Made the byte conversion explicit at these call boundaries rather than
changing the already-exact full-width duplicate helper.
- The user-dictionary availability result is narrowed to a byte at target
+0x1E00. The existing full-width declaration remains; the caller now
performs the demonstrated conversion explicitly.
- User ordinal decoding narrows the masked high byte to a halfword before
its shift at target +0x1E44. Recovered that type boundary.
- Final output termination stores through the current output index and then
increments that index without reloading it (+0x2524..+0x2534). Recovered
the ordinary postfix-index store. Its right-hand side is constant zero,
so there is no unsequenced index read/write.
These are source/ABI shape reconstructions with demonstrated target
conversions; no callee-saved-register or declaration-order sweep was used.
## Isolated source trials
Each trial used the exact 43U Ninja compiler command in `/tmp/ezi-onekey-pass`,
without modifying the worktree until the final candidate was selected.
Below: candidate-function fuzzy / unit fuzzy / matched data / exact functions.
- Halfword formal retaining old masks: 95.26784 / 97.65741 / 1232 / 4; rejected
- Byte duplicate-helper return: 94.73383 / 97.4822 / 1232 / 5; rejected
- Explicit duplicate-result bytes alone: same as preceding; not retained alone
- User-dictionary result byte alone: 95.240715 / 97.708664 / 1280 / 5
- User ordinal halfword alone: 95.30121 / 97.735695 / 1280 / 5
- Combined count-limit increments: 95.106384 / 97.64865 / 1232 / 5; rejected
- Terminal postfix index alone: 95.4368 / 97.79627 / 1280 / 5
- Halfword formal/old masks + duplicate bytes + limits:
95.984566 / 97.97763 / 1232 / 4; rejected
- Halfword formal with ordinary typed uses:
95.26784 / 97.72078 / 1232 / 5; restores the exact helper
- Typed key + duplicate bytes: 95.83312 / 97.97334 / 1280 / 5
- Typed key + duplicate bytes + count limits:
95.984566 / 98.04101 / 1232 / 5; excluded for data regression
- Preceding plus terminal postfix: 96.01794 / 98.055916 / 1232 / 5; excluded
- Preceding plus user-dictionary/ordinal conversions:
95.926155 / 98.01491 / 1232 / 5; excluded
- Typed key + duplicate bytes + terminal postfix:
95.858154 / 97.98453 / 1280 / 5
- Typed key + duplicate bytes + user-dictionary byte:
95.87484 / 97.99198 / 1280 / 5
- Typed key + duplicate bytes + ordinal halfword:
95.93533 / 98.01901 / 1280 / 5
- Typed key + duplicate bytes + terminal postfix + user-dictionary byte:
95.86858 / 97.98919 / 1280 / 5
- Typed key + duplicate bytes + terminal postfix + ordinal halfword:
95.98957 / 98.04324 / 1280 / 5
- Final complete boundary subset: 96.41927 / 98.23523 / 1280 / 5
After selecting that subset, the three count-limit blocks were tested
individually against it. OEM: 96.17564 / 98.12637 / 1280 / 5; PUD:
95.725075 / 97.92507 / 1232 / 5; phrase: 95.93575 / 98.019196 / 1280 / 5.
All three were rejected. No count-limit rewrite remains in the candidate.
## Full validation and exact data disclosure
Fresh report `/tmp/ezi-onekey-boundary.json`; full-build log
`/tmp/ezi-onekey-boundary-build.log`.
Unit fuzzy 97.778564% -> 98.23523%. Candidate-function fuzzy
95.39716% -> 96.41927%, about 98 fewer weighted deficit bytes.
Exact functions remain 5/9; exact code remains 3952/21460;
matched data remains 1280/1388. All five baseline exact functions have
ctxdiff diffs 0 and identical instruction counts.
Candidate instructions: 2387 -> 2405 versus target 2397; size
0x254C -> 0x2594 versus target 0x2574. Frame remains 0x140.
It is not exact, and no new exact function is claimed.
The spelling tables remain 1152 byte-identical bytes. .sdata2 and extab
remain byte-identical at 8 and 72 bytes. The switch table's source is 44
bytes versus target 48, the pre-existing four-byte section-alignment tail;
all 44 bytes and all eleven canonical relocation targets/addends match.
No real data object was resized or renamed and no padding was added.
The already-unmatched 108-byte extabindex still has two differing bytes at
0x4F and 0x5B, with the same 98.14815% section score. They are the function
extent words: spelling remains 0x1A94 versus target 0x1A90; candidate extent
changes from 0x254C to 0x2594 versus target 0x2574. This is changed unmatched
metadata, not a claim that all data bytes are unchanged.
Pool check: identical (no narrow strings). Advisory branch mapping: all 351
pairs preserve their destination and all 71 helper calls correspond. A
register-value-aware linear audit pairs 621 observable store/call/branch
effects. The seven missing byte-boundary differences are resolved; its
remaining three differences are the unchanged count-limit reloads, not a
formal proof or an exactness claim.
Full 43U build succeeds with the approved wrapper and
WIBO_SJIS_MISSING_IMPORTS=1. DOL SHA1:
26116613f624061ba99c8d1a299aaa6efa85670d.
`git diff --check` succeeds. No remote, linking, shared-header, assembly,
volatile, register-forcing, dummy-storage or undefined-value changes.

View File

@@ -0,0 +1,89 @@
# eZiText remainder reconstruction, 2026-10-02
Base: `449529d9`; branch `agent/bittle/allhands-ezi-decompile`.
Only 43U was configured, using `../toolchain/wibo-build/wibo` and
`WIBO_SJIS_MISSING_IMPORTS=1 ../.venv/bin/ninja`. No linking flags, shared
headers, remote refs, or upstream resources were changed.
## Baseline and first candidate: alpha engine
Fresh baseline report: `/tmp/ezi-baseline.json`.
`zi8alpha`: 11/12 exact functions, 5880/21664 exact code bytes,
516/564 exact data bytes; unit fuzzy 94.17892%.
`Zi8AlphaGetCandidates`: 3946/3946 instructions, function fuzzy 92.01039%.
These function and unit percentages are deliberately distinguished.
The scalar layout is four bytes lower in source than in the target because
of the target's existing workspace snapshot at stack 0x20. No dummy alias,
padding, or unused local was added to occupy that slot.
Six semantic corrections were established from definitions and uses, not
from equal-looking raw offsets:
- The flag initialized at target stack 0x84 is the current vowel/punctuation
restriction. At function +0x0798, the table-31 flag test for bit 4 clears
target 0x84 and its primary-language snapshot at 0x8C. Source uses 0x80
and 0x88 for those same definitions.
- The different completion flag is at target 0x88. The table-31 bit-2 test
at +0x07C8 clears target 0x88 and primary snapshot 0x90. Its source slots
are 0x84 and 0x8C. Language-pass restoration confirms both identities.
- Target reads of 0x84 at +0x2114, +0x2214, +0x32F4, +0x3384 and +0x3664
gate punctuation retries, suffix activation, suffix locking and prefix
punctuation. The source incorrectly read the completion flag in all five
places. The candidate uses the traced current-restriction variable.
- At +0x1430 the target loads the dictionary-order index from 0x64, uses it
to select a byte dictionary kind, and stores that kind at 0x60. The retry
condition at +0x3060 reads 0x64, also used by the following outer-loop
increment. Source incorrectly tested dictionaryKind; it now tests
dictionaryIndex. The safe existing early dictionaryIndex initialization
was preserved.
The isolated six-line semantic patch is preserved at
`/tmp/ezi-alpha-six-conditions.patch`. Alone it gives function 92.009125%,
unit 94.17799%, with code/data/exact counts unchanged. This tiny fuzzy
fluctuation does not invalidate the traced variable corrections.
The first handed-off candidate also restores the original address-order
suffix-lock block before alternate-prefix fallback, restores positive branch
conditions, and places each language's restriction snapshots at its actual
control-flow position. No helper prototype change is included in this first
candidate.
First-candidate fresh report: `/tmp/ezi-alpha-minimal.json`.
`Zi8AlphaGetCandidates`: 94.47314%, 3946/3946 instructions.
Unit fuzzy: 94.17892% -> 95.97323% (about 389 fewer weighted deficit bytes).
Exact functions: 11/12 -> 11/12. Exact code: 5880 -> 5880 bytes.
Exact data: 516 -> 516 bytes. All non-text section scores are unchanged.
Both pools contain no strings and compare identical.
## Separate later research, not part of the first candidate
- Further emission/cursor/comparison/switch reconstruction reached function
95.63913%, unit 96.82275%, 3950/3946 instructions. Existing exact functions
remain exact, but extabindex accounting drops by 108 bytes. Preserved at
`/tmp/ezi-alpha-deliver.c`; not included in the first candidate.
- Matching PUD return declarations to zi8pud2.c's ziU32 definitions, with
explicit caller byte narrowing, leaves alpha's score unchanged but moves
one-key candidates 95.39716% -> 95.32416%. Its 5 exact functions and data
1280/1388 remain unchanged. Held separately for ABI review.
- Matching Chinese Zi8GetPCode to zi8match.c's ziU16 definition restores the
two target halfword narrowings in paired-phonetic lookup. Together with the
explicit PUD ABI correction: engine 96.27482% -> 96.492134%, unit
96.65468% -> 96.848755%, 10678/10676 instructions, 5 exact unchanged.
extabindex accounting drops 72 bytes. Held for section/relocation audit.
- Splitting the Chinese component cursor/table chained assignment reproduces
the target's intermediate reload but lowers the engine to 96.090294%;
rejected and restored.
- Replacing the spelling candidate-count assignment with postfix increment
lowers spelling 99.555885% -> 96.8%; rejected and restored.
The larger scratch candidate's branch-destination audit found no mapped
branch-target differences over 720 alpha, 351 one-key and 1594 Chinese
branches. Helper-call counts are respectively 73/73, 71/71 and 168/168.
This audit ignores register choices and accounts for the known stack shift;
it is evidence of control-flow coverage, not a claim of exact matching.
First-candidate validation: full 43U build succeeds; DOL SHA1 is
`26116613f624061ba99c8d1a299aaa6efa85670d`. Every one of the 11 baseline
exact alpha functions has identical instruction counts and ctxdiff diffs 0.
`git diff --check` succeeds. No new exact function or complete unit is claimed.

View File

@@ -0,0 +1,62 @@
# MD5 unsigned arithmetic and cursor reconstruction, 2026-10-02
Separate follow-on to accepted MD5 commit `521056a9`. AES commit `2a537c14`
and both prior logs are unchanged. Only 43U was configured or built.
## Target-grounded changes
The target ProcessBlock rotation halves retain independently grouped additions
of the unsigned state, round constant and separately byte-reversed message
words. Reassociating those three u32 operands inside the accepted helper brings
the compiler's data-flow closer to the target while preserving modulo-2^32
arithmetic. The first MD5 round consumes sequential message words; expressing
that consumption with `word++` as the helper's pointer argument improves its
schedule. STEP evaluates this pointer argument exactly once; the helper still
uses that captured pointer for both byte-reversed reads. No evaluation-order
ambiguity, extra read, undefined value, or new helper is introduced.
Experiments enumerated 144 valid addition operand/grouping combinations for
the two halves. The best target-sized arithmetic expression reached 93.447716%
ProcessBlock fuzzy. Five cursor/loop variants and eight independent cursor-
consumption combinations then identified the minimal `word++` form, reaching
94.30719%. A larger by-reference cursor helper reached 94.98366% but duplicated
rotation code for a small gain and was rejected. Three explicit feed-forward
variants did not improve the retained result. No declaration/register sweep,
assembly, volatile qualifier, padding or linking change was used.
## Verification
Accepted MD5 -> this candidate:
- ProcessBlock: 81.82026 -> 94.30719% fuzzy; 306/306 instructions unchanged.
- Unit fuzzy: 87.80044 -> 96.179825%.
- Exact functions 3/4 and exact code 600/1824 are unchanged.
- Data 456/456 (100%) and linking status are unchanged.
- NETMD5Init 16/16, NETMD5Update 60/60 and NETMD5GetDigest 74/74 still have
ctxdiff diffs 0 and objdiff 100%.
- ProcessBlock has 215 positional differences. An advisory comparison that
ignores GPR names leaves five opcode/immediate-order differences: three
sequential cursor offsets and two final state-load positions. This is not
an exact match or a proof that every remaining difference is register-only.
- Empty string pools remain identical. Constants/index tables and end-marker
data are unchanged from the accepted candidate.
- AES metrics and source are unchanged from commit 2a537c14.
- Full 1027-unit report: zero exact-function/code/data/fuzzy regressions versus
baseline449529d9; additionally both owned units preserve the immediately
preceding candidate's exact/data scores and improve or preserve fuzzy scores.
- Full 43U build and build/43U/ok passed. DOL SHA1:
`26116613f624061ba99c8d1a299aaa6efa85670d`.
All 3,088 MD5 known-answer/random/chunking comparisons passed again against
Python hashlib. Run `python3 /tmp/crypto-remainder/validate_md5.py`. The test
compiles production C unchanged with PPC intrinsic shims, supplies RFC1321
padding externally, and exercises ProcessBlock through NETMD5Update. The
instruction-exact GetDigest platform-endian stores are not host-tested.
Local evidence (not committed/uploaded): `/tmp/crypto-remainder/` contains
`md5.grouping-results.json`, `md5-grouping-final.json`,
`md5-grouping-ProcessBlock.ctx`, `md5-grouping-vectors.log`, and
`md5-grouping-full-build.log`. Original target evidence remains in the local
split object and disassembly; no retail assembly is included in this commit.
GATE: validated partial structural candidate; no regressions. Exact matching
and linking remain open.

View File

@@ -0,0 +1,86 @@
# MD5 rotation-expression reconstruction, 2026-10-02
Baseline: `449529d9`, 43U only. Owned source: `libs/RevoEX/src/net/md5.c`.
The current user priority permits validated structural/fuzzy improvements before
exact matching and linking. This is a partial matching candidate, not an exact
match or a linking change. AES experiments were saved outside the checkout and
AES was restored to the baseline before this candidate was validated.
## Source evidence
- Ground truth: `build/43U/obj/libs/RevoEX/src/net/md5.o`, `ProcessBlock`.
- The target is 0x4C8 bytes / 306 instructions, with a 0x20-byte stack frame.
- Each of the sixteen unrolled MD5 steps contains two byte-reversed word loads.
The first pair is at function-relative offsets 0x3C and 0x40. The right-shift
half consumes the first value, and the left-shift half consumes the second.
- The target retains four additions across the two rotation halves: the
right-shift half adds the message word to (round constant + state), and the
left-shift half adds the state to (message word + round constant). The state
includes the round Boolean function. The first rotation uses right shift 25
and left shift 7; the remaining shifts follow the existing MD5 round schedule.
- The index-table address is first materialized after the first sixteen steps,
rather than before the sequential-word round.
- The inline helper expresses these real unsigned 32-bit operands and restores
the target-sized body. It introduces no assembly, volatile qualifier,
uninitialized value, fake storage, padding, or compiler-option change.
- All four round algorithms, all 64 constants, all 48 indices, and all public
API bodies are otherwise unchanged. The old ROTATE macro became unused and
was removed.
## Attempts
1. Direct pointer-form byte-reversed loads and alternate ordinary arithmetic
groupings: 274-294 instructions, at best near the starting score. Rejected.
2. A typed inline endian-load wrapper: improved instruction similarity but
retained 276 instructions and insufficient target operation boundaries.
3. Rotation-expression helpers without independent load results: 288-290
instructions, approximately 41-44% ProcessBlock fuzzy. Rejected.
4. Explicitly named byte-reversed operands for the two rotation halves:
306 instructions / 81.82026% fuzzy. Retained the target's first-read/right-
shift ordering. Reversing those loads reached 82.0% but added two instructions
and contradicted the target's operand ordering, so it was rejected.
5. Sequential partial-sum statements and equivalent regroupings did not
improve the retained structural form. No register-only declaration sweep.
## Validation
Command setup:
`python3 configure.py --version 43U --wrapper ../toolchain/wibo-build/wibo`
`WIBO_SJIS_MISSING_IMPORTS=1 ../.venv/bin/ninja`
- `pool_diff.py`: identical empty MD5 string pool, 0/0 strings.
- `.data`: all 448 bytes identical. The source `.sdata` is the one-byte
end marker (0x80); the target has the same byte followed by seven alignment
zeros. This existing section-size difference is unchanged, and objdiff
reports `.sdata` and unit data at 100%. No padding was added.
- `NETMD5Init`: 16/16 instructions, ctxdiff diffs 0, objdiff 100%.
- `NETMD5Update`: 60/60 instructions, ctxdiff diffs 0, objdiff 100%.
- `NETMD5GetDigest`: 74/74 instructions, ctxdiff diffs 0, objdiff 100%.
- `ProcessBlock`: 294 -> 306 instructions against target 306;
43.290848 -> 81.82026% fuzzy. Final ctxdiff has 255 positional differences,
primarily allocation, scheduling and remaining arithmetic grouping. Not exact.
- MD5 unit: 61.945175 -> 87.80044% fuzzy; exact 3/4 -> 3/4;
exact code 600/1824 unchanged; data 456/456 (100%) unchanged; unlinked unchanged.
- Full 1027-unit report: no unit regressed in exact functions, exact code,
exact data or fuzzy score versus the baseline report. Overall fuzzy moved
99.588615 -> 99.60437%; exact code and linking did not change.
- Full 43U build passed. `build/43U/ok` passed. DOL SHA1:
`26116613f624061ba99c8d1a299aaa6efa85670d`.
Host algorithm validation: `/tmp/crypto-remainder/validate_md5.py` compiles the
production source unchanged with host shims for PPC byte-reversed intrinsics.
It verifies ProcessBlock through NETMD5Update with externally supplied RFC1321
padding, comparing state words with Python hashlib. The host test deliberately
does not exercise GetDigest's platform-specific big-endian length stores.
Result: 3,088 successful digest/chunking comparisons, covering the seven RFC1321
example strings, one million `a` bytes, 1,500 seeded messages, lengths around
55/56/63/64/65 and larger block boundaries, and fixed/random chunk sizes.
Run: `python3 /tmp/crypto-remainder/validate_md5.py`.
Local evidence (not committed or uploaded): `/tmp/crypto-remainder/md5.target`,
`md5-final.ctx`, `md5-final.json`, `baseline.json`, `full-build.log`, and the
host test plus its header shim. Only source and this summary are committed.
GATE: partial decompilation candidate validated; no regressions; exact matching
and linking remain open. No claim of all-100% completion.

View File

@@ -0,0 +1,99 @@
# NHTTP stdlib source reconstruction, 2026-10-02
Base `b26e248b`, branch `agent/bittle/nhttp-source-reconstruction`, 43U only.
Owned source: `libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL.c`. No headers, metadata,
configuration, linking flags, binary data, or other source files changed.
The earlier `four-leaves-r9.attempts.md` and other worker logs are untouched.
## Target-grounded changes
### NHTTPi_compareToken
The target enters through its case-folded equality test, then checks the raw left
byte for NUL/space before advancing the two pointers. It retains the unsigned
byte separately and sign-extends it for both the case fold and termination check.
The baseline unconditional loop placed that control-flow boundary differently
and compiled to 44 instructions rather than the target's 45.
The retained while condition captures the raw u8 load and explicitly sign-extends
it into LowerCase. The body reuses that captured byte for termination. LowerCase
returns its conditional value directly, preserving the target's separate folded
value lifetime. This recovers the complete 45-instruction control-flow shape.
The common helper's other callers remain unchanged at the instruction level.
A separate token-only helper gave the same result; it was removed to avoid
unnecessary duplication.
Remaining differences: two initial constant loads are reversed, and eight
instruction operands use different scratch/callee-saved registers. This is not
an exact match.
### NHTTPi_Base64Encode
The target retains signed length/padding comparisons. The length local is now
s32, matching those comparisons without the previous mixed-sign comparisons.
Each quartet snapshots its three signed input bytes before writing the output,
then combines each later byte's high fragment with the preceding byte's masked
low fragment. This expresses the target's byte-load and index dependency
structure more closely, preserving the 119-instruction body size.
The target and existing implementation use signed byte shifts, a NUL-terminated
input, and full three-byte reads even for a short final group. This candidate does
not broaden the API to arbitrary binary data or repair those original constraints.
The repository's caller supplies a separate temporary source and authorization
output buffer (`d_nhttp.c`, NHTTP_CreateRequest proxy authorization path).
## Experiments
- Recovered the previously logged loop-test/raw-byte/ternary-fold approach rather
than repeat register/declaration searches. Local conditional-loop alternatives
were measured against the target before retaining the compact while condition.
- One exploratory assignment/read condition was found to be unsequenced. It was
withdrawn, never behavior-tested, and is not part of the retained source.
- Signed/unsigned/raw byte representations and explicit sextet fragments were
checked against target load/sign-extension evidence. Signed byte snapshots
with the retained index grouping gave the best compact source.
- Stream-cursor forms generated identical code and were not retained.
- A separate quartet helper improved Base64 only 67.63025 -> 67.79832%; it was
discarded as unnecessary source expansion. There is no new helper in the diff.
- No declaration/register sweep, inline assembly, use-site volatile, dummy value,
padding, compiler-option change, or linking experiment was performed.
## Verification
Before -> after:
- NHTTPi_compareToken: 54.622223 -> 98.844444% fuzzy; 44 -> 45 instructions against
target 45. Final ctxdiff: 10 positional differences.
- NHTTPi_Base64Encode: 60.285713 -> 67.63025% fuzzy; 119/119 instructions unchanged.
Final ctxdiff: 67 positional differences (baseline 58); aligned objdiff improves
despite that positional metric. Exact dependency scheduling remains open.
- Unit fuzzy: 87.93594 -> 93.03203%.
- Exact functions 11/14, exact code 1388/2248, data 112/112 (100%) and linked status
are unchanged. All 11 previously exact functions have ctxdiff diffs 0.
- The unchanged strnicmp remains 99.76471%, 51/51 instructions, two constant-order
differences. LowerCase's other exact users remain exact.
- Pool: the one alphabet string is identical. .data's 65 bytes and .rodata's
36 bytes equal the prefixes of their 72-byte/40-byte target sections. Existing
trailing target alignment extents are unchanged; no padding was introduced.
- Literal-reference audit: 14 functions analyzed, no candidates, skips or errors;
zero statically addressed call arguments, so this is not a separate proof of
complete data matching.
- Full 1027-unit report: zero exact-function/code/data/fuzzy regressions.
- Full 43U build and build/43U/ok passed. DOL SHA1 remains
`26116613f624061ba99c8d1a299aaa6efa85670d`.
Host behavior validation compiles both unchanged production sources (baseline
and candidate) with -fsigned-char and fixed-width SDK type shims:
- 75,536 compareToken cases: all 65,536 first-byte pairs plus 10,000 seeded
multi-byte/case-fold/mismatch cases, against an independent token reference.
- 3,033 Base64 cases: known answers, every length 0..1024, and 2,000 seeded ASCII
strings up to 4096 bytes, against Python base64. Return length, NUL terminator
and a 16-byte output canary are checked. Inputs have readable zero tail bytes
for the original full-triplet reads. High-bit/binary inputs are not claimed.
Private local evidence: `/tmp/nhttp-source-reconstruction/` contains baseline and
final reports, all function ctxdiffs, pool/literal audits, full-build.log,
validate_nhttp.py, vectors.log, and the rejected local experiments. No retail
assembly or binary is included in this commit.
GATE: validated partial control/data-flow/type reconstruction; no regressions.
Exact matching and linking remain open. Frozen for independent parent validation.

View File

@@ -0,0 +1,199 @@
# System remainder decompilation, 2026-10-02
Baseline: `449529d92405f8f7fda610660c7794914a81dfe6`.
Branch: `agent/bittle/allhands-system-decompile`.
Owned units: `src/utility/iplESMisc.cpp`, `src/scene/setting/AOSS.c`,
`src/scene/setting/ATERM.c`. Parent additionally approved the single ESMisc
return-type declaration change in `include/utility/iplESMisc.h`.
The current user priority is complete source reconstruction first, exact code
matching second, linking last. This candidate is decompilation progress, not an
exact match or a linking change. No configuration, symbol, split, or link flag
was changed. AOSS and ATERM are unchanged.
## Initial audit
- Read AGENTS.md, CONTRIBUTING.md, local skill inventory, AOSS/ATERM attempt
histories, and ESMisc rounds 9/11. The sole local skill is unrelated TypeSafe
integration guidance.
- Configured and built only 43U with the approved Wibo wrapper.
- Before source work, pool_diff found ATERM 0/0 strings and AOSS 1/1 identical.
ESMisc first diverged at string 83; both contained 114 strings.
- ESMisc's 30 exact functions were preserved in earlier rounds, but a meaningful
helper reconstruction had been discarded solely because it was not exact.
Reconstructed it from the current C++ and target object, without importing an
old experimental source or altering linker behavior.
## ESMisc reconstruction
Restored three ordinary inline boundaries named by the target's diagnostic
literals: `verifySavedataZD`, `DeleteTicketsForce`, and `InitSavedata`.
`ESMisc::DeleteUnauthorizedData` returns the initializer status.
Concrete correctness corrections:
1. Use a real NAND_MAX_PATH buffer and a real caller-owned, 32-byte-aligned
NANDFileInfo. Removed the out-of-object `path - 8` and `fileInfo - 0x20`
accesses. The caller also owns the aligned, rounded ticket-view scratch.
2. Traverse `ESTitleId` values by typed array index, with one unconditional loop
increment for every title. The old verifier branch continued without
advancing its separate byte offset and could repeat the same title.
3. Replace the handwritten nested 64-bit switch-lowering expression with the
five actual title cases and the independently masked save-title predicate.
4. Preserve the two ES_ListTitlesOnCard results independently of inlined file
and ticket operation status. The public declaration now returns s32.
5. A negative NANDRead result jumps to cleanup, which frees the save buffer
before closing the still-open file. The old source took the success-path
close before freeing it. Successful or invalid-size reads retain their
close-before-delete/free path.
6. Use sizeof(ESTicketView) for the 0xD8-byte copy and OSRoundUp32B for the 0xE0
scratch/allocation extent, preserving the target's distinct sizes.
### ABI and control-flow evidence
Target function: address 0x813673E0, size 0x704 (449 instructions).
At original object offsets 0x1D20 and 0x1D9C, the two enumeration return values
are saved in r27. Calls through both verifier/ticket paths leave that saved
value independent. At 0x23C0 it is copied to return register r3 after the
optional title-list free. The prior void declaration discarded this ABI.
The only declaration is in include/utility/iplESMisc.h. The only call is
src/system/iplSystem.cpp:680; it intentionally ignores the result. The symbol
name is unchanged because the return type is not encoded in this member name.
All 1027 compiled source objects were compared after a forced full rebuild;
iplSystem and every unit except iplESMisc were byte/relocation unchanged.
The target's verifier branch joins the common increment at 0x238C (index +1,
byte offset +8). Its negative-read path joins cleanup at 0x2104, whereas its
successful close begins at 0x20E0. The reconstruction restores these distinct
paths instead of compensating with stack offsets or register tricks.
### Measurements
| Unit / measure | Before | After |
| --- | ---: | ---: |
| ESMisc exact functions | 30/31 | 30/31 |
| ESMisc exact code | 9404/11200 | 9404/11200 |
| ESMisc fuzzy | 96.76036% | 99.33143% |
| DeleteUnauthorizedData fuzzy | 79.797325% | 95.830734% |
| Function instructions | 434/449 | 446/449 |
| ESMisc matched data | 4416/4416 | 4416/4416 |
| ATERM exact / fuzzy / data | 17/26; 96.624664%; 18584/18864 | unchanged |
| AOSS exact / fuzzy / data | 16/21; 96.777916%; 3928/3928 | unchanged |
| Whole-report fuzzy | 99.588615% | 99.598236% |
pool_diff now reports all 114 strings identical. The source .data is 4378
bytes and exactly equals the original section's first 4378 bytes. The target
section is 4384 bytes and ends in six zero alignment bytes. No synthetic
padding was added. The objdiff data score remains 100%, and the literal audit
checks 204 arguments in all 30 exact functions with zero candidates or errors.
The candidate remains non-exact. ctxdiff shows 446/449 instructions and
remaining register allocation, -1 mask materialization, and scratch-pointer
lifetime differences. No brute-force declaration/register search was run.
## Validation
Commands:
- python3 configure.py --version 43U --wrapper ../toolchain/wibo-build/wibo
- Delete generated native `build/43U/src/**/*.o`, retaining a private baseline
object snapshot for comparison.
- WIBO_SJIS_MISSING_IMPORTS=1 ../.venv/bin/ninja all_source build/43U/report.json build/43U/ok
- PYTHONPATH=../local-tools ../.venv/bin/python tools/decomp-assist/pool_diff.py
- PYTHONPATH=../local-tools ../.venv/bin/python tools/decomp-assist/ctxdiff.py src/utility/iplESMisc DeleteUnauthorizedData__Q33ipl7utility6ESMiscFPQ23EGG4Heap
- PYTHONPATH=../local-tools ../.venv/bin/python tools/decomp-assist/literal_reference_diff.py src/utility/iplESMisc
- Compare every unit in the fresh report and all 1027 source objects' allocated
bytes and normalized relocations against the baseline.
- git diff --check
Full rebuild passed (1031 Ninja steps). DOL SHA1:
`26116613f624061ba99c8d1a299aaa6efa85670d`.
Only the ESMisc report changed: its fuzzy and one function's fuzzy improved.
No report metric regressed. Only ESMisc's allocated .text and .data changed:
.text 11360 -> 11408 bytes (390 relocations on both sides), .data 4378 -> 4378
bytes (zero relocations). Every other allocated section/object is unchanged.
A temporary host harness extracted the actual reconstructed helper bodies,
substituted fake Wii/heap calls, and exercised 12 control-flow scenarios:
first and second enumeration errors, title-list allocation failure, mixed
Zelda/blocked/ordinary title traversal, negative read, short read, second-save
verification failure, absent file, save-buffer allocation failure, ChangeUid
failure, ticket enumeration failure, and zero-ticket handling. Tests ran with
AddressSanitizer and UndefinedBehaviorSanitizer. LeakSanitizer is unavailable
under the execution environment's ptrace, so only that component was disabled.
The harness is supplemental behavioral validation, not a retail runtime test.
GATE: focused decompilation improvement; full 43U/DOL and no-regression checks
pass; exact-matching gate remains incomplete at 30/31. No linking claimed.
## Follow-on: recover the last ESMisc structural differences
Baseline for this follow-on is the accepted reconstruction `1f9db89d`:
30/31 exact functions, function 95.830734%, unit 99.33143%, 446/449 instructions,
data 4416/4416. The source/header correction remains intact.
Natural source attempts, in order:
| Attempt | Instructions | Function fuzzy | Disposition |
| --- | ---: | ---: | --- |
| Ticket helper uses error-first cleanup and returns early with no tickets | 446/449 | 95.830734% | Same generated code |
| Caller gives aligned ticket scratch an explicit pointer lifetime | 448/449 | 98.004456% | Improved |
| Move aligned ticket scratch into the ticket helper | 446/449 | 95.817375% | Rejected |
| Scope caller's scratch pointer to successful title enumeration | 448/449 | 98.89755% | Retained |
| Restore nested ticket helper with scoped scratch | 448/449 | 98.89755% | Same generated code |
| Ticket helper returns its actual ES operation status | 449/449 | 99.14254% | Retained |
The scratch pointer is initialized only in the block reached after successful
second enumeration, and remains live across the title loop. This restores the
saved scratch cursor and per-iteration 64-bit mask materialization. No pointer
is used before initialization and no artificial use extends its lifetime.
The private ticket helper now returns the status already obtained from
ES_GetTicketViews or ES_DeleteTicket, including its zero-ticket early return
and normal cleanup. Its caller intentionally discards that status. The return
contract of an independently emitted original helper cannot be established
because this helper was fully inlined; this is a natural equivalent source
form whose zero-ticket branch structure is confirmed by the target. It does
not overwrite the outer enumeration status or change the public ABI further.
Final ctxdiff: 449/449 instructions, 74 differences. Every mnemonic,
non-register operand, stack offset, resolved call target and branch target
agrees. The remaining differences are register substitutions, including the
title list, scratch cursor, title index, verifier flags and title halves. No
register/declaration permutation search was performed after reaching this
structural agreement.
Measurements after a fresh full report:
- ESMisc exact functions 30/31 -> 30/31
- ESMisc exact code 9404/11200 -> 9404/11200
- ESMisc fuzzy 99.33143% -> 99.8625%
- DeleteUnauthorizedData fuzzy 95.830734% -> 99.14254%
- Matched data 4416/4416 -> 4416/4416
- Pool: all 114 strings identical
- Literal audit: 204 arguments in 30 exact functions, no candidates/errors
- Every other unit's full report is unchanged
- Full 43U build/check and DOL SHA1 pass:
26116613f624061ba99c8d1a299aaa6efa85670d
- The same 12 ASan/UBSan host control-flow scenarios pass on helper bodies
freshly extracted from this final candidate (LeakSanitizer disabled only
for the environment limitation already noted)
- git diff --check passes
This remains a partial matching improvement, not an exact function or unit.
### Rejected ATERM ownership experiment
Read-only target evidence identified that the old typed reply view spans the
last eight bytes of one BSS object and the following option-buffer allocation.
With parent approval, tried a real two-object protocol-workspace aggregate and
corresponding narrow metadata consolidation. It preserved storage, addresses,
all 17 exact functions and matched data, and reduced the main protocol from
972 to 959 instructions (target 951). However, its unit fuzzy regressed from
96.624664% to 96.49781%; wider grouping and a decoder-helper experiment also
failed the no-regression condition. All ATERM source and symbol metadata were
restored. Fresh report after restoration was identical to the accepted ESMisc
baseline, and the DOL hash passed. Detailed rejected source/metadata and
address evidence were kept privately for future work; none is in this commit.

View File

@@ -0,0 +1,23 @@
# iplSetting worker3 — Setting::draw (Oct 2 2026)
Worktree: `/workspace/worktrees/iplSetting-draw`
Branch: `agent/grok/decomp-worker-3/draw`
Unit: `src/scene/setting/iplSetting.cpp` `Setting::draw` only.
`scanAP` / `convertRevIP` source not touched. File stays `NonMatching` (not linked).
## Result
**632/632 byte-identical** to retail (`draw__Q33ipl5scene7SettingFv`, 2528 bytes, 0 diffs).
A HEAD compile of the same TU differs from this object only in `draw`. No sibling codegen change (`-ipa file`).
## Levers that closed it
1. `wideTopL` / `wideTopR` as separate locals (no CSE of the int-to-float). Gets `xoris r6,r6` and `fsubs f2`.
2. One `void* held` is the browser thread, then overwritten by the standard-buffer fetch, so browser and `standardBuffer` share r31.
3. Tex0 material pointer reused as the scroll `FrameController*` (`wideMaterial`), with `initFrame`/`restart` duplicated in each arm.
4. Field copy through `volatile GXRenderModeObj*` so loads and stores interleave on r4/r0.
5. Sample-pattern loop is direct `dst[1]=src[1]; dst[2]=src[2]` with `src` declared before `dst` is assigned (`u8* src; u8* dst = ...; src = ...`).
## Not linked
`configure.py` still has `Object(NonMatching, "scene/setting/iplSetting.cpp")`. DOL SHA1 is unchanged because this TU is not linked. Do not flip the object to Matching until `scanAP` / `convertRevIP` (and the rest of the TU) match.

View File

@@ -0,0 +1,75 @@
# iqdec_b65_frv32 / TMCJPEGDEC_decode_iquant (worker1)
Worktree: `/workspace/worktrees/iqdec-b65-worker1`
Branch: `agent/grok/decomp-worker-1/iqdec_b65`
## Scores
| State | Insns | Diffs | Notes |
|---|---|---|---|
| main baseline | 276/276 | **30** reg-only | entry/DC extract + AC coloring |
| this PR (sibling semantics) | 276/276 | **13** reg-only | structural 0 |
| target wall | 276/276 | 13 | AC huff/zztbl/zz permutation |
## Sibling semantics ported (from matching `decode_iquant_rc` C reference under `#else`)
- `u16*` fast indexing in `decodeLongHuffman` / `decodeACHuffman` (`entry = (u16*)(huff_tbl+9)`, `entry += 2`, `limit->bitLength = entry[0]`)
- comma-op `load_buff`: `if (load_ret = TMCJPEGDEC_load_buff(work), load_ret < 0)`
- early `s32* conv_row = (s32*)conv_row_ptr`
- `1UL << r/t`, `extra -= (tmp - 1)` paren form
- bitCount update before bitBuf read on DC/AC extract
These fixed the 17 non-AC diffs (30→13).
## Remaining wall (13 reg-only)
Permutation of 4 callee-saved across AC setup:
| Value | Ours | Target |
|---|---|---|
| `huff_tbl` (pACHuffTbl) | r29 | **r25** |
| `huff_sym` (pACHuffSym) | r30 | **r24** |
| `zztbl` | r24 | **r29** |
| `zz` | r25 | **r30** |
Cascaded: `tmp-1` in r7 vs r6; `zz<<2` index in r6 vs r7.
Insns 0–114 (through `ac_fast` load) are byte-identical. Divergence starts at AC huff loads (115–116).
## Coloring levers tried (≥12) — none beat 13
1. Separate AC-block `ac_huff_tbl`/`ac_huff_sym` → 32 (DC shifted)
2. DC-only block-local huff → 46 (prologue shifted)
3. Fully scoped DC/AC huff (no top-level) → 32
4. Early `zztbl` at function start → 138 (schedule)
5. `zztbl` before huff (before memset) → 32 (schedule)
6. Swap AC huff_tbl/huff_sym assign order → 13 (swapped wrong load order)
7. Swap DC huff assign order → build fail / unrestored
8. Decl reorder huff near zztbl → 17 (DC shifted)
9. `u8 zz` → 36
10. `u32 zz` → 13 (no change)
11. Move `zz = zztbl[idx]` after `load_buff` → 100
12. Eliminate `zz` temp (double `zztbl[idx]`) → 100
13. Huff before `ac_fast` → 14
14. Split `r` into `dc_r` + `r` → 13
15. Null-kill `huff_*=0` before AC reassign → 13
16. DC passes `work->pDCHuff*` directly → 30
17. Sibling decl order + `ac_vl`/`ac_sym` → 96
18. Scope DC HuffmanEntry temps → 33
19. Dummy pad locals → 13
20. Sibling `*(s32*)((u8*)block + zz*4)` store form → 13
21. `zztbl` first in decl list → 36
22. Coalesce huff_tbl via reassigning `dc_fast` → 36
23. `u32 idx` → 33
24. `s16 idx` → build fail
25. `unsigned long tmp` → 30
26. AC huff via `work->` at call only (no AC locals) → 120
27. Nested temp copies into huff_* → 32
28. `u8* zztbl` non-const → 13
29. **declsearch** 80 evals on decls 118–139 → best still (0, 13)
Root cause: `huff_sym`/`huff_tbl` keep DC-path preferred colors (r29/…) when reassigned for AC; target recycles freed `r24` (dc_predict) / `r25` (dc `r`) for AC huff and pushes `zztbl`/`zz` high. Any split of DC/AC names reshuffles the already-matching DC body.
## Recommendation
Need a shape that keeps DC 0–114 identical **and** births AC huff into r25/r24 (e.g. true live-range split / coalescing onto freed dc_predict/r without extra locals that raise pressure). Sibling `#else` C uses separate `ac_vl`/`ac_sym` and never shares DC huff locals — porting that cleanly without DC regen is the open problem.

View File

@@ -0,0 +1,26 @@
# odh `cdj_d_decompressLoop` (worker-2) 2026-10-02
Claim was only `cdj_d_decompressLoop`. colorConv, huffmanDecoder, and LineConv11 were not modified in the committed diff.
## Result
`cdj_d_decompressLoop` **899/899, 0 diffs, byte-identical** (3596 bytes) to retail. Was 43, then 7.
## What matched
- Inlined `(blockRowSize * 2)` and `((u32)blockWidth << 4)` (init `slwi r18` / `slwi r17`).
- `planeOffset += (u32)master->blocksWide << 6` coalesces the 1x2 update into `add r15, r15, r0`.
- `int nextBlockX = master->blockX + 1` with `(u16)` compare (same shape as compressLoop).
- Second 1x1 chroma sum written blockX term first.
- idct destinations that need plane-first `add` use `(u8*)((u32)planeOffset + (u32)master->workBuffer) + 8`.
- 2x2 tail stores the updated plane in `requiredBlockEnd` and the end pointer in `planeOffset`. That is what puts the plane in r15 and the end in r16 (`addi r15, r3, -8` / `add r16, r15, r18`). Assigning the plane back to `planeOffset` left those two callee-saves swapped (7 diffs). Splitting the `-8` let MWCC cancel it against the later `+8` and dropped to 897.
colorConv left at the 18 pairwise wall. Not in this diff.
## colorConv (2026-10-02)
`cdj_c_colorConv` **77/77, 0 diffs, byte-identical** (308 bytes). Was 18, then 13, then 8.
- Drop `paddedWidth`. The row loop adds `paddedDimensions[0]`, which hoists `lhz r24` and `lis r23` / `addi r10, r23, 0`.
- Pad loop: `u16 remainder = dimensions[i] & 7`, then `(dimensions[i] + 8) - remainder`. Width stays r8, the indexed load stays r7, the mask stays r9. Naming both a dimension and a remainder swaps width off r8.
- Declare `sourceStride` after `rowIndex` (cb/cr stay above it). That assigns stride r29 and `cbPlane` r22. Computing stride before the planes shortens the function to 76 and shifts the callee-saves. huffmanDecoder and LineConv11 were not edited. `cdj_d_decompressLoop` stays 899/899.

View File

@@ -0,0 +1,67 @@
# Data d16 attempts
HEAD and fetched origin/main are 6f49df32. Both owned units are unchanged on origin/main. Scope is data symbols only inside the two assigned split ranges; no source/code/configure/split changes.
Baseline: MyTiManager instruction-exact 91/91, code 20196/20196, data 148/1132; fa/pf_path instruction-exact 27/27, code 11872/11872, data 8/56. Pools identical, 7 and 0 strings. Initial pool commands with unit paths failed because this version of pool_diff accepts object paths; corrected commands passed before any mutation.
Fresh ELF audit: MyTiManager source .data now contains a genuine 36-byte weak Animation vtable at +0x3b0. Target +0x3b0..0x3d8 remains zero without outgoing relocations. Canonical retained Animation vtable is 0x8165D258, with seven real virtual slots. This improves the earlier evidence in data-d4, whose source ended at +0x3b0. The current source ends at +0x3d4; the last four target bytes are zero section alignment. Tail ownership will be tested only with this type/relocation evidence.
fa/pf_path: both .sdata byte arrays are identical for all 48 bytes. The target null literal at +4 is recorded as four bytes, compiler char[2] is two bytes. The target initializer at +40 is split into two single bytes, compiler emits one two-byte const array, used by lbz with addends 0 and 1. The final target space literal is recorded as two bytes, compiler emits a four-byte narrow char string with explicit trailing zero bytes. .sdata2 has two identical two-byte arrays similarly split into individual byte symbols; trailing two bytes are zero alignment. No initializer difference or non-exact function exists.
## pf_path literal extent proof
lbl_81698394 at unchanged .sdata:0x81698394 is char[4] literal `"\x20\0\0"`, not char[2]: PFPATH_CheckExtShortName loads it at .text+0x2b78 and +0x2bcc and passes it to PFSTR_StrNCmp; the compiled @2271 has size 4 and bytes 20000000 at the identical +0x2c offset. Extent 2 -> 4 recovers the complete real string including its explicit two zeros and implicit terminator. No address, section size, source object or total_data changes.
Minimal correction selected: other pf_path aliases and initializer sublabels already compare correctly through exact section bytes. Leave all names and other extents alone.
## MyTiManager tail extent proof
lbl_816681E8 at unchanged .data:0x816681E8 covers the discarded weak Animation vtable footprint, size 0x24 rather than 0x28: the compiler emits __vt__Q39textinput4util9Animation at identical .data+0x3b0 with two ABI header words and seven virtual slots, ending +0x3d4; include/keyboard/tiUtil.h declares exactly those seven methods, and the retained retail Animation vtable at 0x8165D258 also has size 0x24 and precisely the same seven relocation targets. Target createMemoInputForm .text+0x9de/+0x9fe and createLetterInputForm +0xb06/+0xb26 refer to the retained canonical table, while source equivalents refer to this weak copy; this demonstrates linker deduplication. The four unrelocated zero bytes at 0x8166820c..0x81668210 align the next unit to eight bytes and are not an eighth virtual method. Correct only the extent, keep the existing label, all addresses, all 984 section bytes and total_data 1132. No source vtable suppressed and no weak-pointer bytes fabricated.
The old log could not infer the tail when its source emitted no Animation table. The current source now supplies independent table-layout evidence. Every source tail relocation is an extra linker-discarded weak entry; no target relocation is removed or altered. The goal depends on correcting true object versus section-alignment ownership, not inventing a name for zero bytes.
## Iteration measurements
After pf_path extent correction, fresh objdiff report: matched_data 8 -> 56/56, .sdata and .sdata2 both 100%, code 11872/11872, functions 27/27. After MyTiManager correction, fresh objdiff report: matched_data 148 -> 1132/1132, all six non-text sections 100%, code 20196/20196, functions 91/91. Correct source-only Ninja targets subsequently report no work to do. Extracted object paths are outputs of SPLIT, not explicit Ninja targets; the two commands naming them reported unknown targets after the required split succeeded, and were corrected to source-only targets before final verification.
Objdiff v3.4.5 source inspection confirms diff_data_section uses the greatest owned-symbol endpoint, excluding physical section-end alignment, and accepts exact raw section bytes when every target relocation matches; extra source weak relocations do not veto this route. Both corrections describe actual object extents and preserve every physical section byte. Names need no correction for these units.
No open function exists in either unit: all 118 functions were exact before this data task. No compiler experiments, code changes, dummy objects, fake virtual slots, forced placement, symbol renames, configure changes, split changes, or remote mutations. The earlier broad requirement for new exact functions does not apply to these explicitly assigned already-code-100% units; preserve exact counts instead.
## Final clean gate and audit
Full non-quick two-unit gate rebuilt 43U from scratch and passed. Fresh post-build pool comparisons remain identical. All 91 MyTiManager and 27 pf_path functions were checked individually with ctxdiff.py: every one reports diffs 0 and identical instruction counts. Output is /tmp/data-d16.ctxdiff-all.txt.
Live progress and build/43U/report.json regenerated after the gate. Both units matched_data == unchanged total_data, code == unchanged total_code, and every section reports 100%. Source bodies and every symbol name/address/section are unchanged; an automated line-by-line config audit confirms exactly the two extent corrections proved above. No remaining non-exact functions or unresolved non-text sections.
```text
full build: ok
main.dol sha1: 26116613f624061ba99c8d1a299aaa6efa85670d
[src/keyboard/MyTiManager] pool: IDENTICAL
[src/keyboard/MyTiManager] objdiff: code 20196/20196 data 1132/1132 functions 91/91 fuzzy 100.0000 linked code 20196
[src/keyboard/MyTiManager] instruction-exact functions: 91/91
[src/keyboard/MyTiManager] section .bss size 72 match 100.0
[src/keyboard/MyTiManager] section .ctors size 4 match 100.0
[src/keyboard/MyTiManager] section .data size 984 match 100.0
[src/keyboard/MyTiManager] section .sbss size 16 match 100.0
[src/keyboard/MyTiManager] section .sdata size 24 match 100.0
[src/keyboard/MyTiManager] section .sdata2 size 32 match 100.0
[src/keyboard/MyTiManager] section .text size 20196 match 100.0
[src/keyboard/MyTiManager] baseline: code 20196/20196 data 148 functions 91 fuzzy 100.0000
[libs/RVL_SDK/src/fa/pf_path] pool: IDENTICAL
[libs/RVL_SDK/src/fa/pf_path] objdiff: code 11872/11872 data 56/56 functions 27/27 fuzzy 100.0000 linked code 11872
[libs/RVL_SDK/src/fa/pf_path] instruction-exact functions: 27/27
[libs/RVL_SDK/src/fa/pf_path] section .sdata size 48 match 100.0
[libs/RVL_SDK/src/fa/pf_path] section .sdata2 size 8 match 100.0
[libs/RVL_SDK/src/fa/pf_path] section .text size 11872 match 100.0
[libs/RVL_SDK/src/fa/pf_path] baseline: code 11872/11872 data 8 functions 27 fuzzy 100.0000
regressions vs baseline: 0
global matched_code_percent: 90.94304 -> 90.94304
global fuzzy_match_percent: 99.58742 -> 99.58742
global complete_code_percent: 73.73083 -> 73.73083
global matched_data_percent: 99.36639 -> 99.42271
forbidden patterns added (net, per file): 0
readability warnings (net, per file; must be 0 in the final result): 0
note: config touched: config/43U/symbols.txt (orchestrator reviews every config/symbols change)
GATE PASS
```

View File

@@ -1746,3 +1746,557 @@ Final authority lines:
[libs/RVL_SDK/src/kpad/KPAD] instruction-exact functions: 21/29
GATE PASS
```
# Round 9: MAX fresh branchf7b1b9b0
Initial owned objects rebuilt and fresh objdiff baseline generated directly; pools identical101/57/0. Fresh counts SD122/129,Address90/94,KPAD21/29, data100% in all3. New near-miss hypotheses use accessor references to stored pointer cells and postincrement-return inline boundaries; prior rounds did not change the original objects.
## src/scene/sdChannelSelect/iplSDChannelSelect flushSaveDataAndMountSD__Q33ipl5scene15SDChannelSelectFv
Fetch origin/main 6cb5ac9dedfbf99dc591e2cabc9d62d2c849a69d. Initial assembly diagnosis: (2, 2) structural/positional differences; instructions 35/35
Remote source equals freshly built baseline; still open 99.65714%. Structural diagnosis: 35/35,frame0x10; only heap/manager load order reversed. Target loads heap after flag store, then reloads manager. New accessor reference-return types describe read-only views of stored pointer cells; actual calls/field layout remain identical. Prior trials bound pointer values after loading them.
- save-manager getter returns a const reference to the stored pointer: (2, 2) structural/positional; instructions 35/35; restored.
- heap getter returns a const reference to the stored pointer: (2, 2) structural/positional; instructions 35/35; restored.
- both argument getters return read-only references to their pointer cells: (2, 2) structural/positional; instructions 35/35; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesByUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main c7ceb3efe077dca7dd49508cf4e2bd1faab56b54. Initial assembly diagnosis: (2, 4) structural/positional differences; instructions 101/101
Remote source equals freshly built baseline; still open 97.81188%. Structural diagnosis: 101/101 or102/102,frame0x40; target preserves old count inr3 and computes increment inr0, blocking early threshold loading. New inline postincrement-return boundary and copy-plus-increment helper test this real result lifetime; no artificial uses or dummy objects.
- pointer postincrement-result inline boundary after collecting a title: (2, 4) structural/positional; instructions 101/101; restored.
- reference postincrement-result inline boundary after collecting a title: (2, 4) structural/positional; instructions 101/101; restored.
- append postincrement-result inline boundary after collecting a title: (2, 4) structural/positional; instructions 101/101; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesFromNandUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main c7ceb3efe077dca7dd49508cf4e2bd1faab56b54. Initial assembly diagnosis: (2, 4) structural/positional differences; instructions 102/102
Remote source equals freshly built baseline; still open 97.833336%. Structural diagnosis: 101/101 or102/102,frame0x40; target preserves old count inr3 and computes increment inr0, blocking early threshold loading. New inline postincrement-return boundary and copy-plus-increment helper test this real result lifetime; no artificial uses or dummy objects.
- pointer postincrement-result inline boundary after collecting a title: (2, 4) structural/positional; instructions 102/102; restored.
- reference postincrement-result inline boundary after collecting a title: (2, 4) structural/positional; instructions 102/102; restored.
- append postincrement-result inline boundary after collecting a title: (2, 4) structural/positional; instructions 102/102; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect flushSaveDataAndMountSD__Q33ipl5scene15SDChannelSelectFv
Fetch origin/main 7216fd527fee4db057af192c416f9d2fbe0e547a. Initial assembly diagnosis: (2, 2) structural/positional differences; instructions 35/35
Remote source equals freshly built baseline; still open 99.65714%. Structural diagnosis: Further structural lever: scalarized aggregate of the real heap/manager arguments, with read-only request/helper boundaries; target frame0x10 and35 instructions, only two argument loads inverted.
- read-only real flush argument aggregate: (2, 2) structural/positional; instructions 35/35; restored.
- mutable request fields assigned heap then manager: (2, 2) structural/positional; instructions 35/35; restored.
- mutable request with manager initialized last via constructor: (2, 2) structural/positional; instructions 35/35; restored.
## src/scene/address/iplAddressEdit update_friendinfo__Q33ipl5scene11AddressEditFv
Fetch origin/main 7216fd527fee4db057af192c416f9d2fbe0e547a. Initial assembly diagnosis: (2, 2) structural/positional differences; instructions 38/38
Remote source equals freshly built baseline; still open 99.42105%. Structural diagnosis: Structural diagnosis:38 instructions/frame0x10 equal; resolved wcsncpy source and destination LEAs inverted. No immediate store/reload pair or extent overlap; test scalarized copy request with source-first real fields before register-only work.
- read-only name-copy argument aggregate: (11, 32) structural/positional; instructions 40/38; restored.
- mutable name-copy source and destination field assignments: (2, 2) structural/positional; instructions 38/38; restored.
- read-only name-copy request constructor: (2, 2) structural/positional; instructions 38/38; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect create__Q33ipl5scene15SDChannelSelectFv
Fetch origin/main 2da1c58f2c1b14de0754929ddbc23394a011d7a5. Initial assembly diagnosis: (4, 39) structural/positional differences; instructions 145/146
Remote source equals freshly built baseline; still open 97.73972%. Structural diagnosis: 145/146 frame0x20; missing BS2-manager body reload, global/string base colors reversed. New lexical scopes for manager acquisition and an actual abort-state for-loop; no added data or volatile without proof.
- poll using a switch and separate completed-state exit label: build failed i-ipl-workers\data- # d5\include\scene\iplSceneManager.h:14 # Z:\mnt\drive2\projects\wii-ipl-workers\data-d5\include\iplSceneUI.h:11 # Z:\mnt\drive2\projects\wii-ipl-workers\data- # d5\src\scene\sdChannelSelect\iplSDChannelSelect.cpp:6) ### mwcceppc.exe Compiler: # File: src\scene\sdChannelSelect\iplSDChannelSelect.cpp # --------------------------------------------------------- # 258: } # Error: ^ # (10159) undefined label 'abortFinished' # Too many errors printed, aborting program User break, cancelled... ninja: build stopped: subcommand failed. ; restored.
- bind the update receiver as a read-only reference in the loop body: (4, 39) structural/positional; instructions 145/146; restored.
- hold the observed abort state as a leading initialized for-loop variable: (7, 42) structural/positional; instructions 149/146; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesByChannelOrder__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main 02dbcc53bdf195d3b2aa22f56460ad2c2f913dd2. Initial assembly diagnosis: (2, 22) structural/positional differences; instructions 126/126
Remote source equals freshly built baseline; still open 97.29365%. Structural diagnosis: 126/126 frame0x60; title-ID inner scope changes index/add operand allocation; new pointer-to-row and explicit signed page offset views after hoisting the real ID, retaining safe typed entry access.
- hoist the title and index through a pointer to the selected channel page: (2, 7) structural/positional; instructions 126/126; restored.
- hoist the title and apply the row index before dereferencing its array: (2, 7) structural/positional; instructions 126/126; restored.
- hoist the title and use an explicitly unsigned native channel index: (2, 7) structural/positional; instructions 126/126; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesBySpecialChannels__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main ed2d5b30b46940f63b16670dd1f7b44b2ac06984. Initial assembly diagnosis: (8, 16) structural/positional differences; instructions 361/361
Remote source equals freshly built baseline; still open 97.55125%. Structural diagnosis: 361/361 frame0x80; four repeated counter/threshold schedules differ16. New count read-modify-write expression semantics and inline returning the original count; source paths and threshold signedness preserved.
- use a named previous count in each completed title increment: (8, 16) structural/positional; instructions 361/361; restored.
- sequence the postincrement result as a discarded comma-expression: (8, 16) structural/positional; instructions 361/361; restored.
- put postincrement behind a helper returning the original count: (8, 16) structural/positional; instructions 361/361; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect setChannelScissor__Q33ipl5scene15SDChannelSelectCFPCQ33ipl5scene12SDChannelObj
Fetch origin/main 5a332eceba97b348917577a016edf49c8b2bd519. Initial assembly diagnosis: (23, 68) structural/positional differences; instructions 233/233
Remote source equals freshly built baseline; still open 93.4721%. Structural diagnosis: 233/233 frame0xe0; framebuffer-width/height conversions and projection/scissor lifetimes dominate68 differences; new real framebuffer-dimension and scissor-rectangle aggregate views before register-only tuning.
- declare framebuffer width before height then initialize both from named dimension references: (23, 68) structural/positional; instructions 233/233; restored.
- retain one named projection-width local shared by both scissor branches: (51, 195) structural/positional; instructions 230/233; restored.
- represent the four actual computed scissor bounds as fields of a local rectangle: (192, 247) structural/positional; instructions 251/233; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect create__Q33ipl5scene15SDChannelSelectFv
Fetch origin/main a8ab2099cdf66a22205c6faec73834eb7f771168. Initial assembly diagnosis: (4, 39) structural/positional differences; instructions 145/146
Remote source equals freshly built baseline; still open 97.73972%. Structural diagnosis: Correct failed exit-label trial; elapsedTicks is the real statement after the poll. Failed compiler trial did not count toward the three successful variants.
- switch-based poll with a defined abort-finished label before the elapsed tick computation: (9, 40) structural/positional; instructions 145/146; restored.
## src/scene/address/iplAddressEdit get_friendinfo__Q33ipl5scene11AddressEditFv
Fetch origin/main a8ab2099cdf66a22205c6faec73834eb7f771168. Initial assembly diagnosis: (10, 68) structural/positional differences; instructions 82/84
Remote source equals freshly built baseline; still open 88.03571%. Structural diagnosis: 82/84 frame0x420; target keeps name/display addresses in one saved register across virtual pane lookups. New mutable input pointer/array/String views test MWCC rematerialization and alias assumptions without changing shared signatures.
- keep both text inputs in mutable character pointers before the virtual lookups: (10, 68) structural/positional; instructions 82/84; restored.
- bind actual mutable name and display arrays by their real extents: (10, 68) structural/positional; instructions 82/84; restored.
- read both text arrays through a mutable String reference shared across pane calls: (10, 68) structural/positional; instructions 82/84; restored.
## src/scene/address/iplAddressEdit create__Q33ipl5scene11AddressEditFv
Fetch origin/main a8ab2099cdf66a22205c6faec73834eb7f771168. Initial assembly diagnosis: (19, 323) structural/positional differences; instructions 820/820
Remote source equals freshly built baseline; still open 97.44634%. Structural diagnosis: 820/820 frame0x60; ordinary construction removes the doubled allocation guard but leaves two missing name/display LEAs. New mutable text input views paired with ordinary allocation preserve literal source order and avoid manual constructor calls.
- ordinary balloon allocation with a mutable name pointer before animator restart: (23, 331) structural/positional; instructions 818/820; restored.
- ordinary balloon allocation with mutable real name/display array references: (23, 331) structural/positional; instructions 818/820; restored.
- ordinary balloon allocation and a mutable String receiver for both mode-zero text inputs: (23, 331) structural/positional; instructions 818/820; restored.
## src/scene/address/iplAddressEdit set_err_msg__Q33ipl5scene11AddressEditFPwUl8NWC24Err
Fetch origin/main a8ab2099cdf66a22205c6faec73834eb7f771168. Initial assembly diagnosis: (1, 30) structural/positional differences; instructions 76/75
Remote source equals freshly built baseline; still open 98.666664%. Structural diagnosis: 76/75 frame0x60; target has undefined message ID for other errors, safe default cannot be removed. Try safe grouped valid errors with an explicit default and a truly initialized index object; unsupported input behavior remains defined.
- assign the safe fallback explicitly inside every switch case including default: (6, 28) structural/positional; instructions 73/75; restored.
- use an initialized message-selection aggregate holding the actual index: (1, 30) structural/positional; instructions 76/75; restored.
- use a safe if/else selection testing the original supported error groups: (18, 47) structural/positional; instructions 65/75; restored.
## libs/RVL_SDK/src/kpad/KPAD KPADGetProjectionPos
Fetch origin/main fd775c34dcac2c7110d7729ebd0bd6e3ee7bba62. Initial assembly diagnosis: (0, 9) structural/positional differences; instructions 19/19
Remote source equals freshly built baseline; still open 96.57895%. Structural diagnosis: Decoded19/19 leaf structure exact; target double scale usesf0 and half/aspectf4/f3. New actual scalar helper boundaries for correction, half-height, and complete coordinate projection; registers last.
- put the double calibration product behind a scalar-return inline helper: (0, 10) structural/positional; instructions 19/19; restored.
- put the half-height product behind a scalar-return inline helper: (0, 10) structural/positional; instructions 19/19; restored.
- use a complete projection helper with a const coordinate input and precomputed height: (0, 10) structural/positional; instructions 19/19; restored.
## libs/RVL_SDK/src/kpad/KPAD KPADSetSensorHeight
Fetch origin/main fd775c34dcac2c7110d7729ebd0bd6e3ee7bba62. Initial assembly diagnosis: (0, 7) structural/positional differences; instructions 52/52
Remote source equals freshly built baseline; still open 99.23077%. Structural diagnosis: Decoded52/52 frame0x30,7 colors; named-zero and squared-height order yield3 base-register colors. New sensor-center subobject view and channel-base lifetime forms keep actual indexed controller fields, without volatile or dummy state.
- name the actual sensor-center subobject pointer after controller indexing: (0, 3) structural/positional; instructions 52/52; restored.
- form the indexed controller pointer only after computing both real squared dimensions: (0, 3) structural/positional; instructions 52/52; restored.
- assign the squared-height contribution into the diagonal temporary before adding width: (0, 8) structural/positional; instructions 52/52; restored.
## libs/RVL_SDK/src/kpad/KPAD calc_acc_horizon
Fetch origin/main fd775c34dcac2c7110d7729ebd0bd6e3ee7bba62. Initial assembly diagnosis: (2, 35) structural/positional differences; instructions 101/101
Remote source equals freshly built baseline; still open 97.27723%. Structural diagnosis: Decoded101/101 frame0x30,35 pure floating-register colors. New in-place magnitude/normalization lifetimes and product subscopes before using declsearch on any surviving register-only candidate.
- reuse magnitude as the blend weight instead of retaining a second smoothing scalar: (2, 35) structural/positional; instructions 101/101; restored.
- normalize the interpolated horizon coordinates in place before updating the circle: (2, 37) structural/positional; instructions 101/101; restored.
- scope the smoothing weight inside its actual interpolation block: (2, 35) structural/positional; instructions 101/101; restored.
## libs/RVL_SDK/src/kpad/KPAD read_kpad_acc
Fetch origin/main fd775c34dcac2c7110d7729ebd0bd6e3ee7bba62. Initial assembly diagnosis: (16, 66) structural/positional differences; instructions 400/400
Remote source equals freshly built baseline; still open 98.15%. Structural diagnosis: Decoded400/400 frame0x50,5 CFG differences; real smoothing/guard reconstruction narrows to cmpwi5 instead ofcmplwi5. New boolean switch and nested primary/fallback switch boundaries preserve the complete unsupported-format path.
- switch on the unsigned equality predicate of the valid fallback format: (12, 197) structural/positional; instructions 403/400; restored.
- nest the fallback single-case switch inside the primary format switch default: (2, 2) structural/positional; instructions 400/400; restored.
- subtract the real fallback format before testing the zero switch case: (1, 1) structural/positional; instructions 400/400; restored.
## libs/RVL_SDK/src/kpad/KPAD select_1obj_first
Fetch origin/main 842c850c10099d9937f1e02f8625a2f0a985abf9. Initial assembly diagnosis: (2, 26) structural/positional differences; instructions 109/109
Remote source equals freshly built baseline; still open 98.348625%. Structural diagnosis: Decoded109/109 frame0x20; 26 float/base colors, direct candidate-end comparison reaches22. New real geometry operand/lifetime forms and shared output-candidate pointer before declaration search.
- use scale-first products for the two actual horizon offsets after direct-end comparison: (2, 22) structural/positional; instructions 109/109; restored.
- evaluate horizon dot-product axis operands first with direct candidate-end comparison: (4, 22) structural/positional; instructions 109/109; restored.
- bind the two actual output candidate records through one local pointer: (2, 22) structural/positional; instructions 109/109; restored.
## libs/RVL_SDK/src/kpad/KPAD calc_dpd_variable
Fetch origin/main 842c850c10099d9937f1e02f8625a2f0a985abf9. Initial assembly diagnosis: (14, 45) structural/positional differences; instructions 250/250
Remote source equals freshly built baseline; still open 94.82%. Structural diagnosis: Decoded250/250 frame0x20; filter magnitude register reuse and late geometry schedules differ. New explicit absolute-value branch, direction scalar components, and in-place distance filter weight; preserve FP precision and all writes.
- assign filter magnitude using explicit sign branches rather than a conditional expression: (14, 45) structural/positional; instructions 250/250; restored.
- reuse the filtered distance magnitude through its gain multiplication: (14, 43) structural/positional; instructions 250/250; restored.
- name both actual direction components within the final geometric projection scope: (12, 40) structural/positional; instructions 250/250; restored.
## libs/RVL_SDK/src/kpad/KPAD KPADRead
Fetch origin/main 842c850c10099d9937f1e02f8625a2f0a985abf9. Initial assembly diagnosis: (11, 89) structural/positional differences; instructions 459/459
Remote source equals freshly built baseline; still open 97.52723%. Structural diagnosis: Decoded459/459 frame0xc0; raw-copy endpoint grouping and gravity x/y signed halfword loads differ. New complete raw-copy inline boundary, typed three-factor gravity locals, and decrement loop form; no field type/width changes.
- move the complete raw-ring copy into an inline operation returning its final sample index: (46, 250) structural/positional; instructions 459/459; restored.
- use signed-int product temporaries for both complete gravity validation products: (7, 83) structural/positional; instructions 459/459; restored.
- express the raw-ring copy as a for-loop with the decrement in its update clause: (19, 377) structural/positional; instructions 458/459; restored.
## libs/RVL_SDK/src/kpad/KPAD KPADInit
Fetch origin/main 842c850c10099d9937f1e02f8625a2f0a985abf9. Initial assembly diagnosis: (9, 81) structural/positional differences; instructions 185/185
Remote source equals freshly built baseline; still open 96.51351%. Structural diagnosis: Decoded185/185 frame0x90; repeat-delay constant and saved float/matrix colors. New repeat initializer inline boundary, zero/one calibration scope, and real matrix reference pointer scope; arrays/stores unchanged.
- initialize all six real repeat state fields through a complete inline operation: (9, 81) structural/positional; instructions 185/185; restored.
- initialize zero before one while retaining the actual saved scalar locals: (9, 81) structural/positional; instructions 185/185; restored.
- bind the initialized rotation matrix pointer inside the controller loop instead of retaining it outside: (9, 65) structural/positional; instructions 185/185; restored.
## libs/RVL_SDK/src/kpad/KPAD calc_acc_horizon
Fetch origin/main 6f6f63cb81bdb8298e0270364f6b94dae2b36a6c. Initial assembly diagnosis: (2, 35) structural/positional differences; instructions 101/101
Remote source equals freshly built baseline; still open 97.27723%. Structural diagnosis: Register-only final search after this round three distinct successfully built lifetime/scope attempts. Reconstruct the best19-color seed from previous complete assignment order, then search the remaining swaps/moves with a substantially larger budget; decode paired loads and FP compares.
- Reproduced best declaration seed, decoded (0, 19); every assignment and computation order unchanged.
- Extended declsearch: declaration block: | f32 targetY; | f32 circleY; | f32 deltaX; | f32 oldX; | f32 projectedX; | f32 productX; | f32 targetX; | f32 oldY; | f32 smoothing; | f32 magnitude; | f32 normalizedX; | f32 normalizedY; | f32 nextX; | f32 nextY; | f32 normalized; | f32 circleX; | f32 deltaCircleX; | f32 deltaCircleY; | f32 oldCircleY; | f32 oldCircleX; | f32 unitY; | f32 unitX; | f32 blend; | start (0, 19) | improved (0, 17) | best (0, 17) after 852 builds; kept in source: | f32 targetY; | f32 circleY; | f32 deltaX; | f32 oldX; | f32 projectedX; | f32 productX; | f32 targetX; | f32 normalizedY; | f32 smoothing; | f32 magnitude; | f32 normalizedX; | f32 oldY; | f32 nextX; | f32 nextY; | f32 normalized; | f32 circleX; | f32 deltaCircleX; | f32 deltaCircleY; | f32 oldCircleY; | f32 oldCircleX; | f32 unitY; | f32 unitX; | f32 blend; |
Nonexact extended-search candidate restored; raw authority (0, 17).
## libs/RVL_SDK/src/kpad/KPAD KPADGetProjectionPos
Fetch origin/main a9a3f36cf03be45e012ddbb0b86cc689755c89c0. Initial assembly diagnosis: (0, 9) structural/positional differences; instructions 19/19
Remote source equals freshly built baseline; still open 96.57895%. Structural diagnosis: Register-only last after three new scalar/full projection helper attempts; no frame/CFG differences. Remove the independent named double result so its anonymous calibration temporary can coalesce, while naming only actual half/aspect factors. Search six real local declarations, preserving every assignment.
- Anonymous double-calibration seed: decoded (0, 9).
- Anonymous-calibration declsearch: declaration block: | f32 halfHeight; | f32 aspect; | f32 half; | f32 x; | f32 y; | f32 height; | start (0, 9) | improved (0, 5) | best (0, 5) after 36 builds; kept in source: | f32 aspect; | f32 halfHeight; | f32 half; | f32 x; | f32 y; | f32 height; |
Nonexact projection search candidate restored; raw authority (0, 5).
## libs/RVL_SDK/src/kpad/KPAD calc_dpd_variable
Fetch origin/main a9a3f36cf03be45e012ddbb0b86cc689755c89c0. Initial assembly diagnosis: (14, 45) structural/positional differences; instructions 250/250
Remote source equals freshly built baseline; still open 94.82%. Structural diagnosis: Further structural reconstruction of late geometry: target loads directionY/tangentY, negates directionY, then loads directionX/tangentX and multiplies in that order. Earlier direct const/scalar views retained12 differences; name the real dot/cross terms and test declaration/expression order.
- declare the cross-product rotation before the dot-product rotation: (22, 47) structural/positional; instructions 250/250; restored.
- name all four geometric input components in target operand-load order: (12, 41) structural/positional; instructions 250/250; restored.
- name the real dot/cross product component values in target multiplication order: (14, 41) structural/positional; instructions 250/250; restored.
## libs/RVL_SDK/src/kpad/KPAD KPADSetSensorHeight
Fetch origin/main a9a3f36cf03be45e012ddbb0b86cc689755c89c0. Initial assembly diagnosis: (0, 7) structural/positional differences; instructions 52/52
Remote source equals freshly built baseline; still open 99.23077%. Structural diagnosis: Remaining3 base-register colors after named-zero geometry reconstruction. New actual scalar-return scale helper takes only a read-only sensor-center input instead of the prior entire controller/height helper; this exposes a real inline/alias boundary.
- scalar sensor-scale helper with const center input: (1, 13) structural/positional; instructions 52/52; restored.
- scalar sensor-scale helper with mutable center input: (1, 13) structural/positional; instructions 52/52; restored.
- scalar sensor-scale helper with separated offset inputs: (32, 58) structural/positional; instructions 59/52; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesByUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main a9a3f36cf03be45e012ddbb0b86cc689755c89c0. Initial assembly diagnosis: (2, 4) structural/positional differences; instructions 101/101
Remote source equals freshly built baseline; still open 97.81188%. Structural diagnosis: Four schedule-only differences remain after the new old-count helper attempts. Test immutable formal pointer variables without altering pointee types or mangled public signature, then copy the real incremented count through memcpy as a memory-write boundary.
- const only the count pointer formal: (2, 4) structural/positional; instructions 101/101; restored.
- const every real pointer formal: (2, 4) structural/positional; instructions 101/101; restored.
- write the actual incremented output count through the ordinary memory-copy primitive: (29, 53) structural/positional; instructions 105/101; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesFromNandUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main a9a3f36cf03be45e012ddbb0b86cc689755c89c0. Initial assembly diagnosis: (2, 4) structural/positional differences; instructions 102/102
Remote source equals freshly built baseline; still open 97.833336%. Structural diagnosis: Four schedule-only differences remain after the new old-count helper attempts. Test immutable formal pointer variables without altering pointee types or mangled public signature, then copy the real incremented count through memcpy as a memory-write boundary.
- const only the count pointer formal: (2, 4) structural/positional; instructions 102/102; restored.
- const every real pointer formal: (2, 4) structural/positional; instructions 102/102; restored.
- write the actual incremented output count through the ordinary memory-copy primitive: (29, 52) structural/positional; instructions 106/102; restored.
## src/scene/address/iplAddressEdit update_friendinfo__Q33ipl5scene11AddressEditFv
Fetch origin/main a9a3f36cf03be45e012ddbb0b86cc689755c89c0. Initial assembly diagnosis: (2, 2) structural/positional differences; instructions 38/38
Remote source equals freshly built baseline; still open 99.42105%. Structural diagnosis: Two source/destination LEA schedule differences; test genuine array-element/array-object addresses and a read-only getter returning the actual name array reference, no raw offsets or pinned objects.
- pass explicit addresses of the real source and destination first elements: (2, 2) structural/positional; instructions 38/38; restored.
- cast the address of the actual destination name-array object: (2, 2) structural/positional; instructions 38/38; restored.
- read the actual source array through a const member getter returning its full array reference: (2, 2) structural/positional; instructions 38/38; restored.
- Anonymous double projection best5 diagnosis: [(11, ('fmuls', 'f4, f1, f4'), ('fmuls', 'f1, f1, f4')), (12, ('fmuls', 'f1, f3, f2'), ('fmuls', 'f2, f3, f2')), (13, ('fmuls', 'f2, f3, f4'), ('fmuls', 'f1, f3, f1')), (14, ('fmul', 'f0, f1, f0'), ('fmul', 'f0, f2, f0')), (15, ('stfs', 'f2, 4(r3)'), ('stfs', 'f1, 4(r3)'))]. Temporary diagnosis source restored.
## libs/RVL_SDK/src/kpad/KPAD KPADGetProjectionPos
Fetch origin/main a9a3f36cf03be45e012ddbb0b86cc689755c89c0. Initial assembly diagnosis: (0, 9) structural/positional differences; instructions 19/19
Remote source equals freshly built baseline; still open 96.57895%. Structural diagnosis: Anonymous-double calibration and named coefficient search reaches5 differences, all in y-half-height/X/Y-aspect colors. New actual coordinate-result lifetime boundaries; target19 instructions with self-coalesced Xf2/Yf1 and no stack.
- inline only the vertical coordinate product into its output assignment: (0, 0) structural/positional; instructions 19/19; candidate kept.
## Round9 exact KPAD projection candidate and gate
KPADGetProjectionPos: removing the named double scale result lets the correction product coalesce in f0; named real half/aspect locals put those coefficients in f4/f3; inlining only the vertical coordinate expression into dest->y removes its competing local lifetime and fixes the last5 colors. Final19/19 instructions, ctxdiff0, exact-name objdiff100.0%; pool identical0 strings. The existing horizontal local preserves source.x before either output write, including dest/src aliasing. No new volatile, asm, symbol/extent change or data object.
All-unit quick authority gate: full43U build passed; DOL SHA1 26116613f624061ba99c8d1a299aaa6efa85670d; all3 pools identical;0 regressions/forbidden/readability warnings. SD122/129 unchanged, Address90/94 unchanged, KPAD21->22/29; code6756->6832/13056, data8032/8032 unchanged.
## libs/RVL_SDK/src/kpad/KPAD KPADSetSensorHeight
Fetch origin/main 8864436cde34e234c6853e021fe428b802033f4a. Initial assembly diagnosis: (0, 7) structural/positional differences; instructions 52/52
Remote source equals freshly built baseline; still open 99.23077%. Structural diagnosis: Register-only last after >=3 distinct successfully built new structural/alias/scope variants. Use declsearch on real subobject/scan locals with all assignments preserved; decode PPC paired operations and FP comparisons before scoring.
- New register-only seed decoded (0, 3).
- Register-only last declsearch: declaration block: | Vec2* center; | KPADInside* kpad; | f32 halfHeightSquared; | f32 halfHeight; | f32 barOffsetX; | f32 barDiagonal; | f32 halfWidthSquared; | f32 halfWidth; | f32 negativeSensorHeight; | f32 barDiagonalSquared; | f32 zero; | start (0, 3) | best (0, 3) after 146 builds; source restored; best order was: | Vec2* center; | KPADInside* kpad; | f32 halfHeightSquared; | f32 halfHeight; | f32 barOffsetX; | f32 barDiagonal; | f32 halfWidthSquared; | f32 halfWidth; | f32 negativeSensorHeight; | f32 barDiagonalSquared; | f32 zero; |
Nonexact declaration candidate restored; raw authority (0, 3).
## libs/RVL_SDK/src/kpad/KPAD select_1obj_first
Fetch origin/main 8864436cde34e234c6853e021fe428b802033f4a. Initial assembly diagnosis: (2, 26) structural/positional differences; instructions 109/109
Remote source equals freshly built baseline; still open 98.348625%. Structural diagnosis: Register-only last after >=3 distinct successfully built new structural/alias/scope variants. Use declsearch on real subobject/scan locals with all assignments preserved; decode PPC paired operations and FP comparisons before scoring.
- New register-only seed decoded (0, 22).
- Register-only last declsearch: declaration block: | KPADDPDObject* object; | f32 scale; | f32 productX; | f32 productY; | f32 offsetX; | f32 offsetY; | start (0, 22) | best (0, 22) after 36 builds; source restored; best order was: | KPADDPDObject* object; | f32 scale; | f32 productX; | f32 productY; | f32 offsetX; | f32 offsetY; |
Nonexact declaration candidate restored; raw authority (2, 22).
## libs/RVL_SDK/src/kpad/KPAD select_1obj_first
Fetch origin/main 6f49df32b3d1be83a41250c810670894ebe27efe. Initial assembly diagnosis: (2, 26) structural/positional differences; instructions 109/109
Remote source equals freshly built baseline; still open 98.348625%. Structural diagnosis: Apply the newly proven projection lifetime lever to the scan: target scale is an anonymousf0 load and each input coordinate coalesces through its use. Test removal of named real scale/product/component temporaries with exact typed field accesses, no arithmetic reassociation.
- inline the scale field into both real offset products: (2, 17) structural/positional; instructions 109/109; restored.
- inline the dot/cross component products into the two real offset initializers: (4, 22) structural/positional; instructions 109/109; restored.
- inline candidate input fields directly into both projected positions: (0, 13) structural/positional; instructions 109/109; restored.
- combine anonymous scale and candidate coordinate inputs: (0, 8) structural/positional; instructions 109/109; restored.
- combine anonymous product and candidate coordinate inputs: (2, 13) structural/positional; instructions 109/109; restored.
- inline all scale/product/coordinate inputs: (0, 0) structural/positional; instructions 109/109; candidate kept.
## Round9 exact single-object scan candidate and gate
select_1obj_first: after structural const/operand trials and declsearch, the projection lifetime finding led to removing named scale, dot/cross component and candidate x/y temporaries. Direct typed field expressions and direct candidate-array loop endpoint recover every target color. Final109/109 instructions/ctxdiff0, exact-name objdiff100.0%; projection remains19/19/ctxdiff0. No extra memory objects/volatile/asm or raw offset casts.
All-unit quick authority gate GATE PASS: correct DOL hash; pools identical;0 regressions/forbidden/readability warnings. SD122/129 and Address90/94 unchanged. KPAD22->23/29; round baseline21->23, code6756->7268/13056 (+512), all data8032/8032 unchanged.
Read-only origin dependency audit: KPAD dependencies and all owned function symbol definitions are unchanged from f7b1b9b0. Scene dependency changes only rename unrelated fields/getters at unchanged offsets or add IPL_CHANNEL_SELECT_CPP blocks (inactive in both owned scene TUs); owned source is unchanged remotely. No duplicated remote function match found. No symbols.txt changes made locally.
## libs/RVL_SDK/src/kpad/KPAD calc_acc_horizon
Fetch origin/main 6f49df32b3d1be83a41250c810670894ebe27efe. Initial assembly diagnosis: (2, 35) structural/positional differences; instructions 101/101
Remote source equals freshly built baseline; still open 97.27723%. Structural diagnosis: Apply two newly exact SDK lifetime reconstructions after the extended declaration search plateau17: remove real named field/product/interpolation temporaries so anonymous expressions can coalesce. Preserve the original floating multiplication/addition grouping, calls, guards and each object-field write.
- read normal components directly in their real products: (2, 35) structural/positional; instructions 101/101; restored.
- read the old horizon axis directly in its interpolation expressions: (2, 28) structural/positional; instructions 101/101; restored.
- inline the actual first dot-product component: (2, 35) structural/positional; instructions 101/101; restored.
- inline the blend and X-delta subexpressions: (2, 35) structural/positional; instructions 101/101; restored.
- read the old circle components directly in their interpolated results: (2, 33) structural/positional; instructions 101/101; restored.
- combine normal, axis and dot-product anonymous inputs: (0, 13) structural/positional; instructions 101/101; restored.
- combine normal, axis, dot, blend and delta anonymous inputs: (0, 21) structural/positional; instructions 101/101; restored.
- combine every actual field/product/interpolation input: (0, 19) structural/positional; instructions 101/101; restored.
- write circle interpolation results directly before deriving their true residuals: (2, 35) structural/positional; instructions 101/101; restored.
- combine all anonymous inputs with direct circle result stores: (0, 19) structural/positional; instructions 101/101; restored.
## libs/RVL_SDK/src/kpad/KPAD calc_acc_horizon
Fetch origin/main b33a493cec82cb6dbb068d07e1ad17c6d0204e32. Initial assembly diagnosis: (2, 35) structural/positional differences; instructions 101/101
Remote source equals freshly built baseline; still open 97.27723%. Structural diagnosis: New anonymous normal/axis/product inputs lower35 to13 exact positional differences with zero decoded structural difference; revisit register-only last declsearch only on this improved readable source, preserving all initialized assignments and rounding.
- Anonymous field/product declsearch: declaration block: | f32 magnitude; | f32 normalizedX; | f32 normalizedY; | f32 blend; | f32 projectedX; | f32 smoothing; | f32 deltaX; | f32 nextX; | f32 nextY; | f32 normalized; | f32 unitX; | f32 oldCircleX; | f32 oldCircleY; | f32 unitY; | f32 circleX; | f32 deltaCircleX; | f32 circleY; | f32 deltaCircleY; | start (0, 13) | improved (0, 11) | improved (0, 9) | improved (0, 3) | best (0, 3) after 650 builds; kept in source: | f32 magnitude; | f32 normalizedX; | f32 normalizedY; | f32 blend; | f32 projectedX; | f32 smoothing; | f32 deltaX; | f32 nextX; | f32 nextY; | f32 normalized; | f32 oldCircleY; | f32 circleX; | f32 oldCircleX; | f32 unitY; | f32 unitX; | f32 deltaCircleX; | f32 circleY; | f32 deltaCircleY; |
Nonexact anonymous-field/product declaration candidate restored; raw authority (0, 3).
## libs/RVL_SDK/src/kpad/KPAD KPADRead
Fetch origin/main b33a493cec82cb6dbb068d07e1ad17c6d0204e32. Initial assembly diagnosis: (11, 89) structural/positional differences; instructions 459/459
Remote source equals freshly built baseline; still open 97.52723%. Structural diagnosis: Raw-copy target computes the output cursor before loading the ring index, whereas prior pointer-first grouping retained index computation outside the cursor scope and enlarged the frame. New actual cursor/index scope ordering and fewer named count/index lifetimes preserve the complete copy and calibration logic.
- compute the copy cursor first and the actual ring index inside its scope: (40, 282) structural/positional; instructions 459/459; restored.
- combine cursor/index scope ordering with anonymous ring count loading: build failed libs/RVL_SDK/src/kpad/KPAD.c -o build/43U/src/libs/RVL_SDK/src/kpad && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d ### mwcceppc.exe Compiler: # File: libs\RVL_SDK\src\kpad\KPAD.c # ------------------------------------- # 1454: calc_button_repeat(kpad, device, ringCount); # Error: ^^^^^^^^^ # (10140) undefined identifier 'ringCount' # Too many errors printed, aborting program User break, cancelled... ninja: build stopped: subcommand failed. ; restored.
- reuse the actual later sample-count variable in the raw-copy loop: (40, 282) structural/positional; instructions 459/459; restored.
- combine cursor/index scope with both anonymous and reused count values: build failed libs/RVL_SDK/src/kpad/KPAD.c -o build/43U/src/libs/RVL_SDK/src/kpad && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d ### mwcceppc.exe Compiler: # File: libs\RVL_SDK\src\kpad\KPAD.c # ------------------------------------- # 1453: calc_button_repeat(kpad, device, ringCount); # Error: ^^^^^^^^^ # (10140) undefined identifier 'ringCount' # Too many errors printed, aborting program User break, cancelled... ninja: build stopped: subcommand failed. ; restored.
## libs/RVL_SDK/src/kpad/KPAD calc_dpd_variable
Fetch origin/main b33a493cec82cb6dbb068d07e1ad17c6d0204e32. Initial assembly diagnosis: (14, 45) structural/positional differences; instructions 250/250
Remote source equals freshly built baseline; still open 94.82%. Structural diagnosis: Distance filter target multiplies its weight in place in f2 then produces distance delta in f1. Remove the named next-value local and use the stored true dist_vec field before its speed/position updates; test direct geometric expressions without changing operation grouping.
- derive distance speed from the actual stored delta after an in-place gain product: (14, 43) structural/positional; instructions 250/250; restored.
- combine anonymous stored distance delta with the explicit magnitude sign branch: (14, 43) structural/positional; instructions 250/250; restored.
- combine anonymous distance delta with direction-first rotation products: (12, 38) structural/positional; instructions 250/250; restored.
- Anonymous-field/product horizon best3 diagnosis: [(44, ('fmuls', 'f5, f8, f6'), ('fmuls', 'f8, f8, f6')), (47, ('fmuls', 'f2, f5, f2'), ('fmuls', 'f2, f8, f2')), (48, ('fmuls', 'f0, f5, f0'), ('fmuls', 'f0, f8, f0'))]. Temporary diagnosis source restored.
## libs/RVL_SDK/src/kpad/KPAD calc_acc_horizon
Fetch origin/main e3457d11a22a9f0d8300daed615914fb230bea7d. Initial assembly diagnosis: (2, 35) structural/positional differences; instructions 101/101
Remote source equals freshly built baseline; still open 97.27723%. Structural diagnosis: Remaining3 instructions in the anonymous-field/product best declaration candidate all identify the same structural temporary boundary: target overwrites magnitudef8 with magnitude*blend, source creates independent smoothingf5. Reuse the real magnitude as the interpolation weight after its normalization work is finished.
- reuse magnitude in place for the true interpolation weight in the best3 real-local context: (0, 0) structural/positional; instructions 101/101; candidate kept.
## Round9 exact acceleration horizon candidate and gate
calc_acc_horizon: anonymous normal/old-axis/first-product inputs reduce35 colors to13; declaration search reduces13->3. All three remaining instructions prove the same temporary difference: target overwrites magnitude f8 with magnitude*blend, while source creates smoothing f5. Reusing magnitude as its interpolation weight recovers101/101 instructions and ctxdiff0/exact-name objdiff100.0%. No arithmetic reassociation or changed object writes/calls/guards. Projection and single-object scan remain exact.
All-unit quick gate GATE PASS, correct DOL hash, identical pools,0 regressions/forbidden/readability warnings. SDK23->24/29; round21->24, code6756->7672/13056 (+916), data8032/8032. SD122/129, Address90/94 and their code/data totals unchanged.
KPADRead failed anonymous-ring-count trials diagnosed: ringCount also supplies the full batch elapsed count to calc_button_repeat after available is capped by caller count. It must retain the original batch count; replacing it with available would change behavior and was not done. Failed compilations do not count toward the three successful source attempts.
## Round9 final full clean gate and remaining-function audit
Full non-quick gate across all3 units GATE PASS; DOL SHA1 26116613f624061ba99c8d1a299aaa6efa85670d; all pools identical;0 regressions/forbidden/readability warnings; every owned data section100%. No data symbol rename or extent correction needed/made. All open target function extents end before/equal the next symbol; no adjacent same-slot word/single/double stack store/reload pair justified new volatile. No byte16/byte8 mask mismatch justified widening.
src/scene/sdChannelSelect/iplSDChannelSelect: exact 122 ->122/129; code 29412 ->29412/33828; data 2960 ->2960/2960.
src/scene/address/iplAddressEdit: exact 90 ->90/94; code 23044 ->23044/27112; data 2560 ->2560/2560.
libs/RVL_SDK/src/kpad/KPAD: exact 21 ->24/29; code 6756 ->7672/13056; data 8032 ->8032/8032.
MAX round produced 3 new exact functions; 98 logged successfully built distinct source variants plus1720 declaration-search builds (failed builds excluded). All16 remaining functions have >=3 distinct successfully built source variants in this round. Every nonexact source/header trial was restored; only the newly exact SDK function bodies and this log differ from f7b1b9b0. Unsupported NWC24-error input contract remains unproved; initialized fallback retained.
- src/scene/sdChannelSelect/iplSDChannelSelect create__Q33ipl5scene15SDChannelSelectFv 97.73972%; 3 successful distinct source trials; 145/146 frame0x20; missing BS2-manager body reload; switch/for-loop polling and update-reference scopes leave gap.
- src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesByUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl 97.81188%; 6 successful distinct source trials; 101/101 frame0x40; counter add/store and first threshold load schedule differ4; postincrement-result helpers and const formal pointer variables unchanged; copied count adds4 instructions.
- src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesFromNandUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl 97.833336%; 6 successful distinct source trials; 102/102 frame0x40; counter add/store and first threshold load schedule differ4; old-count helpers/const formals unchanged; copied count adds4 instructions.
- src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesByChannelOrder__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl 97.29365%; 3 successful distinct source trials; 126/126 frame0x60; entry indexing and counter/threshold lifetimes; title-ID hoist plus pointer-row/unsigned channel index views leave7; nonexact restored.
- src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesBySpecialChannels__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl 97.55125%; 3 successful distinct source trials; 361/361 frame0x80; four repeated counter/threshold schedules differ16; named previous counts, comma postincrement and old-count-return helper unchanged.
- src/scene/sdChannelSelect/iplSDChannelSelect flushSaveDataAndMountSD__Q33ipl5scene15SDChannelSelectFv 99.65714%; 6 successful distinct source trials; 35/35 frame0x10; heap/manager argument loads reversed; pointer-cell getter const references and real flush-request aggregates all retain2.
- src/scene/sdChannelSelect/iplSDChannelSelect setChannelScissor__Q33ipl5scene15SDChannelSelectCFPCQ33ipl5scene12SDChannelObj 93.4721%; 3 successful distinct source trials; 233/233 frame0xe0; framebuffer conversion and scissor/projection coordinate lifetimes; dimension references retain68, shared width230 instructions, bounds aggregate251.
- src/scene/address/iplAddressEdit create__Q33ipl5scene11AddressEditFv 97.44634%; 3 successful distinct source trials; 820/820 frame0x60; target keeps two text addresses across lookups and has one allocation guard; ordinary new with mutable pointer/array/String views remains818/820; restored.
- src/scene/address/iplAddressEdit get_friendinfo__Q33ipl5scene11AddressEditFv 88.03571%; 3 successful distinct source trials; 82/84 frame0x420; target saves both text pointers before virtual pane lookups; mutable pointer/array/String views still82/84; rematerialization gap remains.
- src/scene/address/iplAddressEdit update_friendinfo__Q33ipl5scene11AddressEditFv 99.42105%; 6 successful distinct source trials; 38/38 frame0x10; resolved wcsncpy source/destination LEAs reversed; mutable request fields/constructor, actual array addresses and const array-reference getter retain2.
- src/scene/address/iplAddressEdit set_err_msg__Q33ipl5scene11AddressEditFPwUl8NWC24Err 98.666664%; 3 successful distinct source trials; 76/75 frame0x60; target leaves message ID undefined on unsupported errors; initialized aggregate76, exhaustive safe switch73, grouped safe if65; safe fallback retained.
- libs/RVL_SDK/src/kpad/KPAD KPADSetSensorHeight 99.23077%; 6 successful distinct source trials; 52/52 frame0x30; named-zero and center-subobject view fixes floating colors but leaves3 array-base colors;146-build declsearch plateau3; const/mutable center scalar helpers13 differences.
- libs/RVL_SDK/src/kpad/KPAD read_kpad_acc 98.15%; 3 successful distinct source trials; 400/400 frame0x50; smoothing/guard reconstruction reaches one cmpwi5 instead of targetcmplwi5; unsigned predicate switch403, nested switch2 differences, format-relative switch1; restored.
- libs/RVL_SDK/src/kpad/KPAD calc_dpd_variable 94.82%; 9 successful distinct source trials; 250/250 frame0x20; distance-weight and late rotation geometry load/product schedules; anonymous stored distance delta/direction-first products lower to38; no exact candidate.
- libs/RVL_SDK/src/kpad/KPAD KPADRead 97.52723%; 5 successful distinct source trials; 459/459 frame0xc0; raw-copy cursor endpoint and ring-index/load schedules plus gravity x/y order; complete copy helper459 with250 differences, gravity product locals83, for-loop458; cursor/index scope459 instructions/282 differences; removing full batch count fails because repeat timing still uses it; nonexact restored.
- libs/RVL_SDK/src/kpad/KPAD KPADInit 96.51351%; 3 successful distinct source trials; 185/185 frame0x90; delay constant/saved-float/matrix colors; repeat initializer and zero/one order unchanged81, matrix pointer loop scope improves65; restored.
New exact names: KPADGetProjectionPos, calc_acc_horizon, select_1obj_first.
Final full-gate authority lines:
```
[src/scene/sdChannelSelect/iplSDChannelSelect] instruction-exact functions: 122/129
[src/scene/address/iplAddressEdit] instruction-exact functions: 90/94
[libs/RVL_SDK/src/kpad/KPAD] instruction-exact functions: 24/29
GATE PASS
```
Gate-passing source commits: 459cc944 (projection), 021a0466 (single-object scan), b7d4457c (acceleration horizon). Final clean build retains all3 exact matches and every previous exact function. Only libs/RVL_SDK/src/kpad/KPAD.c and this attempts log changed.
## Round10 MAX continuation baseline and acceptance gates
Fresh source baseline d78210528f882407ff363d4a112114670a216ce3. Read AGENTS.md; no subagents or remote writes. Pools first: SDK identical0 strings, SD identical101, Address identical57. Source trials must preserve pool and prior exact functions; only instruction-exact plus exact-name objdiff100 candidates may be retained/committed after all-unit quick gate. Final clean all-unit gate and three successfully built distinct source attempts per remaining function required. Failed compilations excluded.
libs/RVL_SDK/src/kpad/KPAD baseline {"fuzzy_match_percent": 98.81893, "total_code": "13056", "matched_code": "7672", "matched_code_percent": 58.762253, "total_data": "8032", "matched_data": "8032", "matched_data_percent": 100.0, "total_functions": 29, "matched_functions": 24, "matched_functions_percent": 82.75862, "total_units": 1}
src/scene/sdChannelSelect/iplSDChannelSelect baseline {"fuzzy_match_percent": 99.582596, "total_code": "33828", "matched_code": "29412", "matched_code_percent": 86.945724, "total_data": "2960", "matched_data": "2960", "matched_data_percent": 100.0, "total_functions": 129, "matched_functions": 122, "matched_functions_percent": 94.57364, "total_units": 1}
src/scene/address/iplAddressEdit baseline {"fuzzy_match_percent": 99.52479, "total_code": "27112", "matched_code": "23044", "matched_code_percent": 84.995575, "total_data": "2560", "matched_data": "2560", "matched_data_percent": 100.0, "total_functions": 94, "matched_functions": 90, "matched_functions_percent": 95.74468, "total_units": 1}
## libs/RVL_SDK/src/kpad/KPAD KPADSetSensorHeight
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (0, 7) structural/positional differences; instructions 52/52
Remote source equals freshly built baseline; still open 99.23077%. Structural diagnosis: Target/ours 52 instructions and frame0x30; width/height squared and zero local lifetime differ in four floating colors and array-base uses r4 instead of target r3 in three instructions. No branch or arithmetic difference. Build genuine shorter squared/negative-height lifetimes and readonly pointer definitions before register-only search; original target PSQ decoded and ctxdiff recorded.
- compute the two real squared dimensions directly in the diagonal sum: (0, 3) structural/positional; instructions 52/52; restored.
- negate sensor height directly in its real center-field store: (1, 14) structural/positional; instructions 52/52; restored.
Trial harness diagonal-sum transform omitted its Python return; aborted before writing source, corrected, not counted as a source attempt. Remote dependency audit d7821052->26d7e1e0: only configure.py switches unrelated already-matched wprintf to Matching; owned source, headers and symbol definitions unchanged.
## libs/RVL_SDK/src/kpad/KPAD KPADSetSensorHeight
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (0, 7) structural/positional differences; instructions 52/52
Remote source equals freshly built baseline; still open 99.23077%. Structural diagnosis: Target/ours 52 instructions and frame0x30; width/height squared and zero local lifetime differ in four floating colors and array-base uses r4 instead of target r3 in three instructions. No branch or arithmetic difference. Build genuine shorter squared/negative-height lifetimes and readonly pointer definitions before register-only search; original target PSQ decoded and ctxdiff recorded.
- inline only the real diagonal sum into sqrt: (0, 0) structural/positional; instructions 52/52; candidate kept.
## Round10 exact sensor-height candidate and gate
KPADSetSensorHeight: target has the diagonal sum as an anonymous sqrt input; named-zero/squared-height order plus removal of the real diagonal-sum local makes the array-base register reuse match as well. 52/52 instructions, frame0x30, ctxdiff0, exact-name objdiff100.0%; pool identical. All-unit quick gate GATE PASS, correct DOL SHA1,0 regressions/forbidden/readability warnings. SDK24->25/29; code7672->7880/13056; data8032/8032 unchanged. SD122/129 and Address90/94 unchanged. Only the exact SDK function and attempts log changed; no headers or data-symbol edits.
## libs/RVL_SDK/src/kpad/KPAD read_kpad_acc
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (16, 66) structural/positional differences; instructions 400/400
Remote source equals freshly built baseline; still open 98.15%. Structural diagnosis: 400/400 frame0x50; smoothing magnitude/delta boundaries and split error/device guards reduce the old mismatch to a single final switch cmpwi5 versus target cmplwi5. Try unsigned relational guards and branch-to-accepted loop forms on that complete readable reconstruction, preserving all unsupported-format returns.
- bound the valid fallback format with unsigned relational guards: (10, 195) structural/positional; instructions 401/400; restored.
- dispatch each accepted format through a single-pass loop with two break arms: (10, 194) structural/positional; instructions 399/400; restored.
- use a looping fallback switch whose accepted arm exits the loop: (15, 198) structural/positional; instructions 401/400; restored.
- write the primary and fallback cases as exclusive unsigned bounds: (11, 200) structural/positional; instructions 402/400; restored.
- handle the fallback format in a one-case switch followed by a primary-format guard: (13, 196) structural/positional; instructions 399/400; restored.
## libs/RVL_SDK/src/kpad/KPAD calc_dpd_variable
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (14, 45) structural/positional differences; instructions 250/250
Remote source equals freshly built baseline; still open 94.82%. Structural diagnosis: 250/250 frame0x20; target distance weight is updated in place and late geometry coalesces direction products through their sums. Test real Vec2 rotation/midpoint boundaries, anonymous rotation or midpoint values, and direct normal-component input lifetimes before declaration search. Retain original operation grouping and every object-field write.
- hold the true dot/cross rotation in a Vec2 local: build failed lang=c -MMD -c libs/RVL_SDK/src/kpad/KPAD.c -o build/43U/src/libs/RVL_SDK/src/kpad && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d ### mwcceppc.exe Compiler: # File: libs\RVL_SDK\src\kpad\KPAD.c # ------------------------------------- # 963: f32 scaleX = 0.5f * (kpad->dpdState.candidates[0].x + kpad->dpdState.candidates[1].x); # Error: ^^^ # (10141) expression syntax error # Too many errors printed, aborting program User break, cancelled... ninja: build stopped: subcommand failed. ; restored.
- evaluate the actual Vec2 cross coordinate before its dot coordinate: build failed lang=c -MMD -c libs/RVL_SDK/src/kpad/KPAD.c -o build/43U/src/libs/RVL_SDK/src/kpad && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d ### mwcceppc.exe Compiler: # File: libs\RVL_SDK\src\kpad\KPAD.c # ------------------------------------- # 963: f32 scaleX = 0.5f * (kpad->dpdState.candidates[0].x + kpad->dpdState.candidates[1].x); # Error: ^^^ # (10141) expression syntax error # Too many errors printed, aborting program User break, cancelled... ninja: build stopped: subcommand failed. ; restored.
- inline midpoint x and y into both actual geometric products: (16, 39) structural/positional; instructions 250/250; restored.
- inline the dot rotation expression into its actual geometric products: (19, 45) structural/positional; instructions 250/250; restored.
- inline both dot and cross rotation expressions into their geometric products: (10, 23) structural/positional; instructions 250/250; restored.
- inline all rotation and midpoint expressions: (12, 35) structural/positional; instructions 250/250; restored.
- read normal components through a const subobject view: (12, 38) structural/positional; instructions 250/250; restored.
- combine a readonly normal view with anonymous midpoint products: (16, 39) structural/positional; instructions 250/250; restored.
## libs/RVL_SDK/src/kpad/KPAD KPADRead
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (11, 89) structural/positional differences; instructions 459/459
Remote source equals freshly built baseline; still open 97.52723%. Structural diagnosis: 459/459 frame0xc0; copy cursor uses target statuses+available then decrement, source (available-1) scale before pointer-add. Ring count is read from an actual u8 field, never changed and still supplies repeat elapsed count; test safe byte/halfword local widths, real cursor lifetime sharing, and the full copy cursor endpoint before declaration search.
- keep the actual full batch count in its byte-width local: (11, 89) structural/positional; instructions 459/459; restored.
- keep the actual full batch count widened to a halfword local: (11, 89) structural/positional; instructions 459/459; restored.
- preserve the nonnegative wrapping ring index as an unsigned local: (12, 89) structural/positional; instructions 459/459; restored.
- use an end cursor and copy into the preceding real status slot: (75, 284) structural/positional; instructions 459/459; restored.
- reuse the actual later output cursor for the initial raw-copy loop: (46, 276) structural/positional; instructions 459/459; restored.
- combine a byte batch count with the real end-cursor copy: (75, 284) structural/positional; instructions 459/459; restored.
- combine a halfword batch count and shared output cursor: (46, 276) structural/positional; instructions 459/459; restored.
## libs/RVL_SDK/src/kpad/KPAD calc_dpd_variable
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (14, 45) structural/positional differences; instructions 250/250
Remote source equals freshly built baseline; still open 94.82%. Structural diagnosis: Anonymous rotation inputs improve45 to23 raw differences with unchanged250 instructions. Distance target coalesces the original quotient f2 into its absolute interpolation weight: reuse the real value local after deriving dx, preserving its original zero-validity store. Test combined anonymous-rotation and actual weight reuse before registers.
- reuse the distance quotient local for its actual absolute filtering weight: (14, 43) structural/positional; instructions 250/250; restored.
- combine distance-value reuse with anonymous dot/cross rotation inputs: (10, 23) structural/positional; instructions 250/250; restored.
- fix C89 leading declarations for a real Vec2 rotation local: build failed c -MMD -c libs/RVL_SDK/src/kpad/KPAD.c -o build/43U/src/libs/RVL_SDK/src/kpad && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d ### mwcceppc.exe Compiler: # File: libs\RVL_SDK\src\kpad\KPAD.c # ------------------------------------- # 961: scaleX = 0.5f * (kpad->dpdState.candidates[0].x + kpad->dpdState.candidates[1].x); # Error: ^^^^^^ # (10140) undefined identifier 'scaleX' # Too many errors printed, aborting program User break, cancelled... ninja: build stopped: subcommand failed. ; restored.
- combine the real Vec2 rotation and distance weight reuse: build failed c -MMD -c libs/RVL_SDK/src/kpad/KPAD.c -o build/43U/src/libs/RVL_SDK/src/kpad && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d ### mwcceppc.exe Compiler: # File: libs\RVL_SDK\src\kpad\KPAD.c # ------------------------------------- # 960: scaleX = 0.5f * (kpad->dpdState.candidates[0].x + kpad->dpdState.candidates[1].x); # Error: ^^^^^^ # (10140) undefined identifier 'scaleX' # Too many errors printed, aborting program User break, cancelled... ninja: build stopped: subcommand failed. ; restored.
- reuse distance weight and inline midpoint inputs as well: (12, 35) structural/positional; instructions 250/250; restored.
## libs/RVL_SDK/src/kpad/KPAD KPADInit
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (9, 81) structural/positional differences; instructions 185/185
Remote source equals freshly built baseline; still open 96.51351%. Structural diagnosis: 185/185 frame0x90; target keeps zero/one and controller/matrix values in different saved registers and computes the repeat-delay constant earlier. Negative-sine narrowing coalesces to target f0 versus named rotation result f2. Test anonymous true trig result stores, matrix lifetime/representation, and loop-local calibration temporaries before register search.
- store each actual trig result directly into its matrix element: (13, 86) structural/positional; instructions 185/185; restored.
- combine direct trig stores with a controller-loop matrix pointer scope: (13, 70) structural/positional; instructions 185/185; restored.
- keep the actual width height and diagonal calibration locals inside their loop: (9, 81) structural/positional; instructions 185/185; restored.
- view the real rotation matrix through the SDK Mtx element layout: build failed contract off -lang=c -MMD -c libs/RVL_SDK/src/kpad/KPAD.c -o build/43U/src/libs/RVL_SDK/src/kpad && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d ### mwcceppc.exe Compiler: # File: libs\RVL_SDK\src\kpad\KPAD.c # ------------------------------------- # 1530: initial_rotation_(*matrix)[0][0] = one; # Error: ^ # (10377) illegal operands 'int' [ 'int' # Too many errors printed, aborting program User break, cancelled... ninja: build stopped: subcommand failed. ; restored.
- combine local calibration and anonymous trig results: (13, 86) structural/positional; instructions 185/185; restored.
- combine typed matrix elements and anonymous trig results: build failed contract off -lang=c -MMD -c libs/RVL_SDK/src/kpad/KPAD.c -o build/43U/src/libs/RVL_SDK/src/kpad && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d build/43U/src/libs/RVL_SDK/src/kpad/KPAD.d ### mwcceppc.exe Compiler: # File: libs\RVL_SDK\src\kpad\KPAD.c # ------------------------------------- # 1529: initial_rotation_(*matrix)[0][0] = one; # Error: ^ # (10377) illegal operands 'int' [ 'int' # Too many errors printed, aborting program User break, cancelled... ninja: build stopped: subcommand failed. ; restored.
## libs/RVL_SDK/src/kpad/KPAD calc_dpd_variable
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (14, 45) structural/positional differences; instructions 250/250
Remote source equals freshly built baseline; still open 94.82%. Structural diagnosis: Anonymous rotation inputs improve45 to23 raw differences with unchanged250 instructions. Distance target coalesces the original quotient f2 into its absolute interpolation weight: reuse the real value local after deriving dx, preserving its original zero-validity store. Test combined anonymous-rotation and actual weight reuse before registers.
- fix C89 leading declarations for a real Vec2 rotation local: (62, 136) structural/positional; instructions 252/250; restored.
- combine the real Vec2 rotation and distance weight reuse: (62, 136) structural/positional; instructions 252/250; restored.
## libs/RVL_SDK/src/kpad/KPAD KPADInit
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (9, 81) structural/positional differences; instructions 185/185
Remote source equals freshly built baseline; still open 96.51351%. Structural diagnosis: Trig and geometry structural trials preserve185 instructions but leave temporary rounding/register schedules. Target negative sine remains double until its actual field store. Test that real precision/lifetime boundary and the corrected exact SDK matrix layout before final declaration permutations. Failed C89/matrix harness transforms are excluded from successful trial counts.
- retain negative sine at math-library precision until its real float matrix store: (9, 80) structural/positional; instructions 185/185; restored.
- retain every true trig result at library precision until the float store: (9, 80) structural/positional; instructions 185/185; restored.
- combine a double negative-sine result and controller-loop matrix scope: (9, 64) structural/positional; instructions 185/185; restored.
- access actual matrix rows and columns through Mtx: (9, 81) structural/positional; instructions 185/185; restored.
- combine library-precision sine and SDK typed matrix elements: (9, 80) structural/positional; instructions 185/185; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect create__Q33ipl5scene15SDChannelSelectFv
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (4, 39) structural/positional differences; instructions 145/146
Remote source equals freshly built baseline; still open 97.73972%. Structural diagnosis: Target146 instructions/frame0x20 versus ours145; target reloads the BS2 manager in the polling body while ours reuses the condition pointer. Calls and literal pool already agree. Test actual polling pointer scope and loop exit boundaries before any register search; no unjustified volatile.
- retain the actual polling manager as a for-loop condition value: (6, 21) structural/positional; instructions 148/146; restored.
- use explicit polling and ready labels for the same loop condition: (9, 40) structural/positional; instructions 145/146; restored.
- bind the actual polling state reference in the loop condition: (9, 40) structural/positional; instructions 145/146; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesByUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (2, 4) structural/positional differences; instructions 101/101
Remote source equals freshly built baseline; still open 97.81188%. Structural diagnosis: Target and source have identical101/102 instruction counts and frame0x40; the only four differences are counter add/store and first threshold load scheduling. New source lifetime tests qualify the actual title-name input, give its output address a real cursor local, and group the escaped page/channel outputs in their true stack order. No widening, volatile or alias-type violations.
- use a const title-name input at the ordinary name-copy boundary: (2, 4) structural/positional; instructions 101/101; restored.
- give the real title-name output address its own local lifetime: (2, 4) structural/positional; instructions 101/101; restored.
- group the actual page and channel lookup outputs as a ChannelLocation: (16, 17) structural/positional; instructions 101/101; restored.
- combine readonly name input and named output cursor: (2, 4) structural/positional; instructions 101/101; restored.
- combine actual location outputs and readonly title-name input: (16, 17) structural/positional; instructions 101/101; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesFromNandUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (2, 4) structural/positional differences; instructions 102/102
Remote source equals freshly built baseline; still open 97.833336%. Structural diagnosis: Target and source have identical101/102 instruction counts and frame0x40; the only four differences are counter add/store and first threshold load scheduling. New source lifetime tests qualify the actual title-name input, give its output address a real cursor local, and group the escaped page/channel outputs in their true stack order. No widening, volatile or alias-type violations.
- use a const title-name input at the ordinary name-copy boundary: (2, 4) structural/positional; instructions 102/102; restored.
- give the real title-name output address its own local lifetime: (2, 4) structural/positional; instructions 102/102; restored.
- group the actual page and channel lookup outputs as a ChannelLocation: (16, 17) structural/positional; instructions 102/102; restored.
- combine readonly name input and named output cursor: (2, 4) structural/positional; instructions 102/102; restored.
- combine actual location outputs and readonly title-name input: (16, 17) structural/positional; instructions 102/102; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesByChannelOrder__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (2, 22) structural/positional differences; instructions 126/126
Remote source equals freshly built baseline; still open 97.29365%. Structural diagnosis: 126/126 frame0x60; a real titleId hoist previously reduced22 to7 differences, with three channel-address operand differences and four postcopy counter/threshold schedules. New contiguous channel-row indexing order, real name input/output lifetimes, and title-state scope preserve the complete ordered scan.
- use index-first subscripting of the actual typed channel row: (2, 7) structural/positional; instructions 126/126; restored.
- use a readonly pointer to the complete channel row array: (2, 7) structural/positional; instructions 126/126; restored.
- combine the hoisted title value with readonly collected-name input: (2, 7) structural/positional; instructions 126/126; restored.
- combine index-first row addressing and a real name output cursor: (2, 7) structural/positional; instructions 126/126; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesBySpecialChannels__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl
Fetch origin/main 26d7e1e052a80ea55d0591a82653370965ad5bb2. Initial assembly diagnosis: (8, 16) structural/positional differences; instructions 361/361
Remote source equals freshly built baseline; still open 97.55125%. Structural diagnosis: 361/361 frame0x80; four identical counter/threshold schedules differ16 total. Test the same real name-input/output lifetime boundaries, and anonymous cache/lookup title values preserving every special-channel ordering branch.
- make each actual special-channel collected-name input readonly: (8, 16) structural/positional; instructions 361/361; restored.
- name each real special-channel name output cursor before memcpy: (8, 16) structural/positional; instructions 361/361; restored.
- combine readonly collected-name inputs and named output cursors: (8, 16) structural/positional; instructions 361/361; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect flushSaveDataAndMountSD__Q33ipl5scene15SDChannelSelectFv
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (2, 2) structural/positional differences; instructions 35/35
Remote source equals freshly built baseline; still open 99.65714%. Structural diagnosis: 35/35 frame0x10; only heap r4 and save-manager r3 argument-load order differs. Test actual returned file/output-field lifetimes and readonly pointee reference getters scoped to this TU. Keep the flag setter before heap/manager loads, every following store/call unchanged, and no volatile.
- name the actual returned asynchronous save file before its field store: (2, 2) structural/positional; instructions 35/35; restored.
- bind the real asynchronous file output field by reference at the call: (2, 2) structural/positional; instructions 35/35; restored.
- call through actual manager and heap object references at the flush boundary: (2, 2) structural/positional; instructions 35/35; restored.
## src/scene/sdChannelSelect/iplSDChannelSelect setChannelScissor__Q33ipl5scene15SDChannelSelectCFPCQ33ipl5scene12SDChannelObj
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (23, 68) structural/positional differences; instructions 233/233
Remote source equals freshly built baseline; still open 93.4721%. Structural diagnosis: 233/233 frame0xe0; target keeps projection width in saved f29, source retains widthScale there; framebuffer conversion scheduling and matrix temporary colors follow from those lifetimes. Test a projection-width local scoped only to the transformed branch, anonymous matrix bound quotients, and direct quotient use before register-only last.
- keep a branch-local projection width distinct from final scissor width: (23, 68) structural/positional; instructions 233/233; restored.
- combine branch-local projection width with direct horizontal quotient use: (36, 71) structural/positional; instructions 233/233; restored.
- inline the actual top projection quotient into its matrix bound: (23, 67) structural/positional; instructions 233/233; restored.
- inline all actual matrix bound quotients while retaining each final bound: (38, 71) structural/positional; instructions 233/233; restored.
- combine anonymous matrix quotients and branch-local projection width: (38, 71) structural/positional; instructions 233/233; restored.
## src/scene/address/iplAddressEdit create__Q33ipl5scene11AddressEditFv
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (19, 323) structural/positional differences; instructions 820/820
Remote source equals freshly built baseline; still open 97.44634%. Structural diagnosis: 820/820 frame0x60; target has one balloon allocation guard and keeps both text input addresses across virtual pane lookups. Ordinary new recovers the real allocation boundary but leaves818/820 instructions. Test pointer-definition constness, a String pointer receiver reused across the real text reads, and an actual name/display input aggregate; keep literal token order.
- combine ordinary balloon allocation and a const name-pointer definition: (23, 331) structural/positional; instructions 818/820; restored.
- combine ordinary allocation and an actual String pointer used for both text inputs: (23, 331) structural/positional; instructions 818/820; restored.
- combine ordinary allocation and a real pair of readonly name/display input addresses: (23, 331) structural/positional; instructions 818/820; restored.
## src/scene/address/iplAddressEdit get_friendinfo__Q33ipl5scene11AddressEditFv
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (10, 68) structural/positional differences; instructions 82/84
Remote source equals freshly built baseline; still open 88.03571%. Structural diagnosis: 82/84 frame0x420; target computes text LEAs before virtual lookups and retains them in r29, source rematerializes them afterwards. Test const pointer definitions and the real copied friend-info/String object view lifetimes; no extra buffers or dummy objects.
- give both existing readonly text pointers const definitions: (10, 68) structural/positional; instructions 82/84; restored.
- reuse a String pointer receiver for every real string operation: (10, 39) structural/positional; instructions 83/84; restored.
- read the actual copied friend-info object through a named const reference: (14, 71) structural/positional; instructions 80/84; restored.
## src/scene/address/iplAddressEdit update_friendinfo__Q33ipl5scene11AddressEditFv
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (2, 2) structural/positional differences; instructions 38/38
Remote source equals freshly built baseline; still open 99.42105%. Structural diagnosis: 38/38 frame0x10; target builds readonly name-source r4 before global destination r3, source inverts those two LEAs. Source/destination fields and wcsncpy relocation are proven. Test a const String pointer receiver, actual typed name-array pointer, and a source-array reference with copy length derived from the true ten-character limit; no raw offsets or pinned data.
- read the copy source through a const String pointer definition: (2, 2) structural/positional; instructions 38/38; restored.
- read the source through a pointer to the actual complete name-array type: (2, 2) structural/positional; instructions 38/38; restored.
- bind the actual source array and derive the existing ten-character copy limit: (2, 2) structural/positional; instructions 38/38; restored.
## src/scene/address/iplAddressEdit set_err_msg__Q33ipl5scene11AddressEditFPwUl8NWC24Err
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (1, 30) structural/positional differences; instructions 76/75
Remote source equals freshly built baseline; still open 98.666664%. Structural diagnosis: 76/75 frame0x60; the target message-ID register is undefined for unsupported errors. Retain the initialized server fallback. New readonly cache-object access, anonymous actual error-code formatting input, and a typed supported-message enum test legitimate source boundaries without uninitialized assumptions.
- read the actual cache error through a typed readonly FriendListCache pointer: (1, 30) structural/positional; instructions 76/75; restored.
- pass the actual cache error directly into its wide-format call: (1, 30) structural/positional; instructions 76/75; restored.
- select initialized supported message values through a real enum local: (1, 30) structural/positional; instructions 76/75; restored.
## libs/RVL_SDK/src/kpad/KPAD KPADInit
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (9, 81) structural/positional differences; instructions 185/185
Remote source equals freshly built baseline; still open 96.51351%. Structural diagnosis: Register-only last after the logged structural/local-lifetime/operand trials. Search declaration permutations on the best newly tested readable context; keep every real initialized assignment, existing exact function and every field/call unchanged. Decode PSQ and FP comparisons; raw ctxdiff and objdiff remain authority.
- Declaration-search context /tmp/fz16r10.init.more.2.c: (9, 64) structural/positional differences; instructions 185/185; initialized real repeated-delay value is40000, with all field writes retained.
- KPADInit register-only last declsearch: declaration block: | KPADInside* kpad; | f32 distanceValue; | f32 sensorDistance; | f32 rotationElement; | f64 negativeSine; | f32 referenceWidth; | f32 referenceHeight; | f32 objectInterval; | f32 zero; | f32 degreesToRadians; | f32 one; | u32 i; | s32 chan; | BOOL enabled; | u16 repeatDelay; | start (6, 64) | improved (6, 53) | improved (6, 50) | best (6, 50) after 312 builds; kept in source: | f32 referenceHeight; | f32 objectInterval; | f32 sensorDistance; | f32 rotationElement; | f64 negativeSine; | f32 referenceWidth; | KPADInside* kpad; | f32 distanceValue; | f32 zero; | f32 degreesToRadians; | f32 one; | u32 i; | s32 chan; | BOOL enabled; | u16 repeatDelay; |
Nonexact declaration candidate restored; (8, 50) structural/positional differences; instructions 185/185.
## libs/RVL_SDK/src/kpad/KPAD KPADRead
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (11, 89) structural/positional differences; instructions 459/459
Remote source equals freshly built baseline; still open 97.52723%. Structural diagnosis: New target instruction84 moves available into the real decrement counter before the ring-count clear; prior endpoint tests assigned it after that clear. Extend the real counter lifetime first, then combine statuses+available endpoint and ring-index scope. Preserve the full repeat batch count; test gravity y/x operand order only after the copy structure.
- initialize the actual raw-copy counter before clearing the batch field: (9, 86) structural/positional; instructions 459/459; restored.
- combine early decrement-count initialization and a statuses-plus-available endpoint: (44, 283) structural/positional; instructions 459/459; restored.
Read cursor-scope harness omitted the extracted constant defining the old index block; aborted before source write, corrected, not counted. Real early counter assignment improved89->86/459 before the field clear.
## libs/RVL_SDK/src/kpad/KPAD KPADRead
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (11, 89) structural/positional differences; instructions 459/459
Remote source equals freshly built baseline; still open 97.52723%. Structural diagnosis: New target instruction84 moves available into the real decrement counter before the ring-count clear; prior endpoint tests assigned it after that clear. Extend the real counter lifetime first, then combine statuses+available endpoint and ring-index scope. Preserve the full repeat batch count; test gravity y/x operand order only after the copy structure.
- combine early counter initialization and pointer-before-ring-index scope: (40, 282) structural/positional; instructions 459/459; restored.
- give the real copy pointer a leading declaration with the early counter scope: (40, 282) structural/positional; instructions 459/459; restored.
- combine early counter and ring-index scope with target gravity operand order: (36, 282) structural/positional; instructions 459/459; restored.
## libs/RVL_SDK/src/kpad/KPAD read_kpad_acc
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (16, 66) structural/positional differences; instructions 400/400
Remote source equals freshly built baseline; still open 98.15%. Structural diagnosis: Further actual inline boundary: the 400-instruction target accepts either FS format through two unsigned equality checks and a shared negative exit. A boolean support predicate may retain that positive-branch graph; compare real unsigned scalar, const sample-pointer and byte-pointer helpers after the smoothing/guard reconstruction, before registers.
- unsigned scalar format with explicit accepted returns: (15, 204) structural/positional; instructions 404/400; restored.
- readonly sample pointer with accepted positive returns: (15, 204) structural/positional; instructions 404/400; restored.
- readonly format-byte pointer with accepted positive returns: (15, 204) structural/positional; instructions 404/400; restored.
- unsigned scalar short-circuit accepted predicate: (13, 197) structural/positional; instructions 403/400; restored.
- unsigned scalar switch returning real format support: (20, 207) structural/positional; instructions 408/400; restored.
## libs/RVL_SDK/src/kpad/KPAD calc_dpd_variable
Fetch origin/main 805e84045c82629c645e9ef1787c30d2bc33b5e7. Initial assembly diagnosis: (14, 45) structural/positional differences; instructions 250/250
Remote source equals freshly built baseline; still open 94.82%. Structural diagnosis: Best anonymous rotation inputs lower45->23/250 while preserving0x20 frame. Target negates directionY before later rotation products and coalesces midpoint-X into the projected-Y expression. Test those actual sign/input/intermediate lifetimes, plus readonly normal inputs, on the anonymous rotation context before declaration search.
- give the real negative direction-Y input an independent lifetime before anonymous products: (13, 34) structural/positional; instructions 250/250; restored.
- name normal-Y then normal-X only at the late positional projection: (24, 45) structural/positional; instructions 250/250; restored.
- reuse the midpoint-X scalar for its actual projected-Y value after projected-X is computed: (12, 35) structural/positional; instructions 250/250; restored.
- use a readonly normal subobject view with anonymous rotation inputs: (10, 23) structural/positional; instructions 250/250; restored.
## Round10 final full clean gate and remaining-function audit
Full non-quick gate across all3 units GATE PASS; DOL SHA1 26116613f624061ba99c8d1a299aaa6efa85670d; all pools identical;0 regressions/forbidden/readability warnings; every owned data section100%. No data symbol rename or extent correction needed/made. All open target function extents end before/equal the next symbol; no adjacent same-slot word/single/double stack store/reload pair justified new volatile. No byte16/byte8 mask mismatch justified widening.
libs/RVL_SDK/src/kpad/KPAD: exact 24 ->25/29; code 7672 ->7880/13056; data 8032 ->8032/8032.
src/scene/sdChannelSelect/iplSDChannelSelect: exact 122 ->122/129; code 29412 ->29412/33828; data 2960 ->2960/2960.
src/scene/address/iplAddressEdit: exact 90 ->90/94; code 23044 ->23044/27112; data 2560 ->2560/2560.
MAX round produced 1 new exact function; 89 logged successfully built distinct source variants plus312 declaration-search builds (failed builds excluded). All15 remaining functions have >=3 distinct successfully built source variants in this round. Every nonexact source/header trial was restored; only the newly exact SDK function bodies and this log differ from d7821052. Unsupported NWC24-error input contract remains unproved; initialized fallback retained.
- libs/RVL_SDK/src/kpad/KPAD read_kpad_acc 98.15%; 10 successful distinct source trials; 400/400 frame0x50; smoothing/guard reconstruction has only cmpwi5 versus cmplwi5; relational401, positive single-pass399, actual support-predicate helpers403-408; restored.
- libs/RVL_SDK/src/kpad/KPAD calc_dpd_variable 94.82%; 15 successful distinct source trials; 250/250 frame0x20; distance weight and rotation/product schedules; anonymous rotations reduce45->23, true rotation aggregate252, normal/sign/midpoint reuse do not close gap; restored.
- libs/RVL_SDK/src/kpad/KPAD KPADRead 97.52723%; 12 successful distinct source trials; 459/459 frame0xc0; cursor endpoint/index/counter scope and gravity operand schedules; safe u8/u16 full batch counts unchanged89, early real counter reduces86, endpoint scopes frame0xd0/282; restored.
- libs/RVL_SDK/src/kpad/KPAD KPADInit 96.51351%; 9 successful distinct source trials; 185/185 frame0x90; repeat constant/global base and saved FP/matrix colors; actual double sine and matrix scope64, 312-build declaration search50 with6 decoded structural differences; restored.
- src/scene/sdChannelSelect/iplSDChannelSelect create__Q33ipl5scene15SDChannelSelectFv 97.73972%; 3 successful distinct source trials; 145/146 frame0x20; polling condition pointer is reused instead of target body reload; pointer-for-loop148, labels/reference145; restored.
- src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesByUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl 97.81188%; 5 successful distinct source trials; 101/101 frame0x40; four counter/threshold schedules; const name/output cursor unchanged4, true location aggregate17; restored.
- src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesFromNandUsage__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl 97.833336%; 5 successful distinct source trials; 102/102 frame0x40; four counter/threshold schedules; const name/output cursor unchanged4, location aggregate17; restored.
- src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesByChannelOrder__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl 97.29365%; 4 successful distinct source trials; 126/126 frame0x60; entry-address operand order plus counter/threshold schedule; hoisted ID and new row-array/name views remain7; restored.
- src/scene/sdChannelSelect/iplSDChannelSelect collectTitlesBySpecialChannels__Q33ipl5scene15SDChannelSelectFPClPClPUxPcPUl 97.55125%; 3 successful distinct source trials; 361/361 frame0x80; four repeated count/threshold schedules differ16; readonly name and real output-cursor lifetimes unchanged; restored.
- src/scene/sdChannelSelect/iplSDChannelSelect flushSaveDataAndMountSD__Q33ipl5scene15SDChannelSelectFv 99.65714%; 3 successful distinct source trials; 35/35 frame0x10; heap/manager argument load order differs2; real returned-file/output-field and manager/heap reference lifetimes unchanged; restored.
- src/scene/sdChannelSelect/iplSDChannelSelect setChannelScissor__Q33ipl5scene15SDChannelSelectCFPCQ33ipl5scene12SDChannelObj 93.4721%; 5 successful distinct source trials; 233/233 frame0xe0; framebuffer conversion/projection-width/output lifetimes; branch-local projection width68, anonymous top quotient67, anonymous four quotients71; restored.
- src/scene/address/iplAddressEdit create__Q33ipl5scene11AddressEditFv 97.44634%; 3 successful distinct source trials; 820/820 frame0x60; target one allocation guard and retained two text addresses; ordinary new plus const-pointer/String-pointer/actual text-input pair still818/820; restored.
- src/scene/address/iplAddressEdit get_friendinfo__Q33ipl5scene11AddressEditFv 88.03571%; 3 successful distinct source trials; 82/84 frame0x420; text pointers rematerialize after virtual lookups; const definitions82, complete String pointer83/84, copied friend reference80; restored.
- src/scene/address/iplAddressEdit update_friendinfo__Q33ipl5scene11AddressEditFv 99.42105%; 3 successful distinct source trials; 38/38 frame0x10; wcsncpy source/destination LEAs reversed; const String pointer, full source-array pointer/reference keep2; restored.
- src/scene/address/iplAddressEdit set_err_msg__Q33ipl5scene11AddressEditFPwUl8NWC24Err 98.666664%; 3 successful distinct source trials; 76/75 frame0x60; target leaves message ID undefined for unsupported errors; const cache, anonymous error format input and supported enum still76; initialized fallback retained.
New exact names: KPADSetSensorHeight.
Final full-gate authority lines:
```
[libs/RVL_SDK/src/kpad/KPAD] instruction-exact functions: 25/29
[src/scene/sdChannelSelect/iplSDChannelSelect] instruction-exact functions: 122/129
[src/scene/address/iplAddressEdit] instruction-exact functions: 90/94
GATE PASS
```
Final source-isolation/remote audit: origin/main 8a07484e8aea9bdb38db258f5a6eb16347197c86 has no owned source/header/symbol changes from d7821052. SDK source differs only inside newly exact KPADSetSensorHeight; scene sources, headers, symbol metadata and configure.py unchanged locally. Nonexact trials restored. Commit64bbfbba contains the exact height match and its quick all-unit gate; this final audit records the clean full gate and89 successfully built distinct source variants plus312 declaration builds.

View File

@@ -0,0 +1,244 @@
# EXIF parser reconstruction and aliasing audit
Base: `2a33684c` on `agent/bittle/allhands-jpeg-decompile`; EXIF source itself is
unchanged from baseline `449529d9`. This is a separate candidate from the prior
JPEG converter/IDCT commit. Those files and their logs were left unchanged.
## Retained reconstruction
1. The IFD0 entry parameter is `u8*`, with corresponding call-site casts from the
parser's read-only directory cursor. With the reconstructed `const u8*`
parameter, MWCC hoists all four version-byte loads before their stores.
Removing this unsupported qualifier reproduces the original interleaved
load/store sequence for tags 0x9000, 0x9101 and 0xA000.
2. IFD0's decoded field type is widened to `u32` and both entry fields use the
existing `readExifU16` reader. This reproduces the target's eager zero
extension and initial instruction order; it does not change TIFF field size.
3. IFD1 explicitly includes ignored 0x011C and 0xA000..0xA002 cases alongside
existing ignored tags. Target dispatch compares against 0x011C and 0xA004;
these case labels improve the recovered decision tree. The precise original
no-op case list cannot be proved uniquely, since unknown tags also return
unchanged. No new tag interpretation or payload write is introduced.
4. IFD1 rational denominators use `thumbnailData + (offset + 4)` without a
redundant same-type cast.
No new exact functions. Exact stays 3/6, matched code 1348/5088 bytes, allocated
data 0 bytes / 100%, and linking is unchanged. Unit fuzzy 96.50157 -> 98.94104.
| Function | Before | After | Source/target instructions |
| --- | ---: | ---: | ---: |
| TMCCJPEGDecGetOffsetEXIF | 100 | 100 | 155/155, ctxdiff 0 |
| TMCCJPEGDecGetInfoEXIF | 100 | 100 | 132/132, ctxdiff 0 |
| TMCJPEGDEC_exif_parse | 99.43396 | 99.43396 | 212/212 |
| TMCJPEGDEC_IFD0_tag_parse | 94.46986 | 99.386696 | 480/481 |
| TMCJPEGDEC_IFD1_tag_parse | 93.099174 | 96.14876 | 239/242 |
| TMCJPEGDEC_ThumbnailCheck | 100 | 100 | 50/50, ctxdiff 0 |
IFD0 still lacks the original low-tag `bltlr` leaf. Aligning that instruction and
its branch displacements leaves 32 GPR-operand differences in the two rational
field cases. IFD1 still has decision-tree and register differences. The parser
retains 21 GPR-operand differences. Nothing here claims exact completion.
## Concrete behavioral distinction
An internal IFD0 entry can be arranged with its four version bytes overlapping
the destination by one byte. Initial source bytes 01 02 03 04 produce:
- Baseline compiled code: destination 01 02 03 04
- Original object: destination 01 01 01 01
- Corrected compiled code: destination 01 01 01 01
This is evidence about the internal parser's byte-copy/aliasing behavior. It is
not a claim that ordinary nonoverlapping EXIF input was decoded incorrectly, or
that a public API vulnerability has been established. Both byte orders and all
three version fields were checked at offsets -3 through +3, giving 42 overlap
cases. The unchanged baseline object fails this regression test; the final
object passes all 42. Local destination casts, explicit loops and removal of
other const qualifiers did not fix the discrepancy and were restored.
## Validation
- Pool checked before tuning and after the final build: identical, zero strings
- No allocated non-code sections in either EXIF object
- Fresh official report: all six function scores compared with baseline,
zero regressions; all three exact functions retain ctxdiff zero differences
- Full 4.3U build: PASS
- DOL SHA1: `26116613f624061ba99c8d1a299aaa6efa85670d`
- `git diff --check 2a33684c`: PASS, including the committed candidate diff
- No header/configuration/linking edits, no changes to the separately owned
entropy decoder, no remote writes or original binary data committed
Build/report commands:
```sh
python3 configure.py --version 43U --wrapper ../toolchain/wibo-build/wibo
WIBO_SJIS_MISSING_IMPORTS=1 ../.venv/bin/ninja
build/tools/objdiff-cli report generate -p . -o build/43U/report.json -f json
PYTHONPATH=../local-tools ../.venv/bin/python tools/decomp-assist/pool_diff.py build/43U/src/libs/RVLMiddleware/TMC_JPEG/src/exif/exif_parse.o build/43U/obj/libs/RVLMiddleware/TMC_JPEG/src/exif/exif_parse.o
PYTHONPATH=../local-tools ../.venv/bin/python tools/decomp-assist/ctxdiff.py libs/RVLMiddleware/TMC_JPEG/src/exif/exif_parse <symbol>
../.venv/bin/python tools/decomp-assist/jpeg_exif_differential_audit.py
../.venv/bin/python tools/decomp-assist/jpeg_exif_differential_audit.py --overlap-only
```
## Differential test scope
`jpeg_exif_differential_audit.py` is a read-only bounded integer PPC interpreter
and deterministic test generator. It reads local ELF objects and never writes
original executable bytes to the repository. It is a separate test tool; the
previous converter/IDCT audit remains unchanged.
- 267,282 paired calls, seed 449529
- Every 16-bit tag selector in both byte orders for both IFD handlers
- Field types 0, 3, 4 and 0xFFFF; valid, boundary, out-of-range and wrapping
offsets for supported scalar/rational/byte-array tags
- Header sizes and offsets, valid/invalid byte order and TIFF magic, truncated
entry arrays, IFD0/IFD1/Exif-directory chains and directory-boundary lengths
- 42 overlapping version-byte cases that distinguish baseline from original
- Every non-stack byte compared, including guard regions; parser return values,
stack pointer, LR and GPR14..GPR31 preservation checked
- Complete instruction-address coverage of IFD0 and the central parser in both
objects; IFD1 covers every source instruction and 241/242 target instructions
(one unreachable return in the target dispatch is not visited)
Limitations: this is not formal equivalence or a Wii runtime test. Instruction
coverage is not exhaustive path/input coverage. The custom interpreter models
only the observed integer subset, condition comparisons and internal EXIF calls;
unsupported instructions/relocations/callees fail closed. FPU, XER overflow,
exceptions, timing, cache/MMIO and a full system ABI are not modeled. Relocations
use collision-checked symbol tokens; actual linked addresses are not tested.
Public JPEG I/O, thumbnail decoding, arbitrary pointer values and all possible
payload alias layouts are outside this test. The three exact wrappers are
checked by objdiff/ctxdiff rather than executing their external dependencies.
Run from the repository root with pyelftools and Capstone available, and without
Python's `-O` option. `--source-object` and `--target-object` can select separately
compiled objects, allowing the overlap regression to be reproduced on baseline
without changing the working tree. Final output is in
`jpeg-exif-allhands-differential-results.txt`.
## Existing external branch overlap
Read-only comparison used the already fetched ref
`origin/agent/w1009/tmc` at `cd90be696cb86258a1bc9fb9dc2b304f2bed318d`.
No fetch, author communication, branch checkout or merge was performed.
Its `1c7cf40c` independently removes const from both tag-entry parameters, so the
central IFD0 qualifier correction overlaps. Its `81440c32` then uses a u16 tag,
u16 type and explicit tag remask. Its no-op case ranges differ from this
candidate's explicit ignored tags. It also changes the shared 16-bit reader and
parser declaration order. This candidate does not contain those changes.
An isolated `/tmp` compile of that branch's EXIF source, using the current
checkout's original MWCC flags, produced:
| Measure/function | This candidate | cd90be69 EXIF object |
| --- | ---: | ---: |
| Unit fuzzy | 98.94104 | 99.01337 |
| Central parser | 99.43396 | 98.679245 |
| IFD0 | 99.386696 | 98.37838 |
| IFD1 | 96.14876 | 99.194214 |
| Exact functions | 3/6 | 3/6 |
Neither candidate wholly supersedes the other. This candidate has the better
IFD0 and preserves the current parser score; the external candidate has the
better IFD1. Integration must preserve the parser score and independently
revalidate any combined source. The isolated comparison is not a full gate of
the external branch or its other files.
Relative to baseline 449529d9, the external RGB565 edits concern its setter and
do not overlap the prior candidate's RGB444 body. Its RGBA420 edge edits overlap
that prior candidate's changed expression context. It does not change IDCT.
The previously committed JPEG candidate remains separate and unchanged.
## Attempts
All rejected/equal variants were restored. The lower-tag 0x0100 experiment had a
fuzzy improvement but was explicitly discarded after review: the original
branch shape does not establish that particular no-op label, and it changed
other lower dispatch comparisons. It is absent from the final source.
### trials
```text
TMCJPEGDEC_IFD0_tag_parse word-width-field-type: 94.469860 -> 95.122660; RETAIN
TMCJPEGDEC_IFD0_tag_parse explicit-ignored-compression-return: 95.122660 -> 94.914764; restore
TMCJPEGDEC_IFD0_tag_parse explicit-ignored-compression-break: 95.122660 -> 95.122660; restore
TMCJPEGDEC_IFD1_tag_parse recognized-ignored-planar-and-pixel-tags: 93.099174 -> 95.942150; RETAIN
TMCJPEGDEC_IFD1_tag_parse ignored-low-strip-offset-own-return: 95.942150 -> 95.342970; restore
TMCJPEGDEC_IFD1_tag_parse ignored-orientation-own-return: 95.942150 -> 95.533060; restore
TMCJPEGDEC_IFD1_tag_parse ignored-datetime-own-return: 95.942150 -> 95.528920; restore
```
### control-trials
```text
TMCJPEGDEC_IFD0_tag_parse explicit-unknown-tag-return: 95.122660 -> 94.498960; restore
TMCJPEGDEC_IFD0_tag_parse explicit-unknown-tag-break: 95.122660 -> 95.122660; restore
TMCJPEGDEC_IFD1_tag_parse explicit-unknown-tag-return: 95.942150 -> 95.876030; restore
TMCJPEGDEC_IFD1_tag_parse explicit-unknown-tag-break: 95.942150 -> 95.942150; restore
TMCJPEGDEC_IFD1_tag_parse ifd1-orientation-uses-unknown-exit: 95.942150 -> 95.756195; restore
TMCJPEGDEC_IFD1_tag_parse ifd1-planar-uses-unknown-exit: 95.942150 -> 91.694214; restore
```
### ignored-tag-trials
```text
TMCJPEGDEC_IFD0_tag_parse ignored-image-fields-before-compression: 95.122660 -> 95.110180; restore
TMCJPEGDEC_IFD1_tag_parse ignored-image-fields: 95.942150 -> 95.735535; restore
TMCJPEGDEC_IFD1_tag_parse ifd1-planar-logical-case-order: 95.942150 -> 95.942150; restore
```
### reader-trials
```text
TMCJPEGDEC_IFD0_tag_parse shared-exif-tag-reader: 95.122660 -> 95.195430; RETAIN
TMCJPEGDEC_IFD0_tag_parse shared-exif-type-reader: 95.195430 -> 95.517670; RETAIN
TMCJPEGDEC_IFD0_tag_parse shared-exif-short-reader: 95.517670 -> 95.517670; restore
TMCJPEGDEC_IFD1_tag_parse shared-exif-tag-reader: 95.942150 -> 95.942150; restore
TMCJPEGDEC_IFD1_tag_parse shared-exif-short-reader: 95.942150 -> 95.942150; restore
```
### lower-tag-trials
```text
TMCJPEGDEC_IFD0_tag_parse ignored-lower-tags-(256,): 95.517670 -> 93.004160; restore
TMCJPEGDEC_IFD0_tag_parse ignored-lower-tags-(257,): 95.517670 -> 93.004160; restore
TMCJPEGDEC_IFD0_tag_parse ignored-lower-tags-(258,): 95.517670 -> 95.505196; restore
TMCJPEGDEC_IFD0_tag_parse ignored-lower-tags-(256, 257): 95.517670 -> 93.004160; restore
TMCJPEGDEC_IFD0_tag_parse ignored-lower-tags-(256, 258): 95.517670 -> 93.004160; restore
TMCJPEGDEC_IFD0_tag_parse ignored-lower-tags-(257, 258): 95.517670 -> 95.505196; restore
TMCJPEGDEC_IFD0_tag_parse ignored-lower-tags-(256, 257, 258): 95.517670 -> 95.505196; restore
TMCJPEGDEC_IFD1_tag_parse ignored-lower-tags-(256,): 95.942150 -> 96.665290; RETAIN
TMCJPEGDEC_IFD1_tag_parse ignored-lower-tags-(257,): 96.665290 -> 96.665290; restore
TMCJPEGDEC_IFD1_tag_parse ignored-lower-tags-(258,): 96.665290 -> 95.735535; restore
TMCJPEGDEC_IFD1_tag_parse ignored-lower-tags-(256, 257): 96.665290 -> 96.644630; restore
TMCJPEGDEC_IFD1_tag_parse ignored-lower-tags-(256, 258): 96.665290 -> 96.256195; restore
TMCJPEGDEC_IFD1_tag_parse ignored-lower-tags-(257, 258): 96.665290 -> 95.735535; restore
TMCJPEGDEC_IFD1_tag_parse ignored-lower-tags-(256, 257, 258): 96.665290 -> 95.735535; restore
```
### byte-copy-trials
```text
TMCJPEGDEC_IFD0_tag_parse alias-preserving-version-byte-copy: 95.517670 -> 95.517670; restore
```
### version-loop-trials
```text
TMCJPEGDEC_IFD0_tag_parse forward-version-byte-loops: 95.517670 -> 95.517670; restore
```
### qualifier-trials
```text
TMCJPEGDEC_IFD0_tag_parse mutable-entry-view: 95.517670 -> 99.386696; RETAIN
TMCJPEGDEC_IFD1_tag_parse mutable-entry-view: 95.942150 -> 95.942150; restore
```
### pointer-trials
```text
TMCJPEGDEC_IFD1_tag_parse mutable-rational-byte-view: 95.942150 -> 95.942150; restore
TMCJPEGDEC_IFD1_tag_parse direct-denominator-byte-address: 95.942150 -> 96.148760; RETAIN
```

View File

@@ -0,0 +1,6 @@
TMCJPEGDEC_IFD0_tag_parse 131072 selector cases PASS
TMCJPEGDEC_IFD1_tag_parse 131072 selector cases PASS
TMCJPEGDEC_IFD0_tag_parse 132746 cases PASS; own-body instruction coverage 480 / 480 source; 481 / 481 original
TMCJPEGDEC_IFD1_tag_parse 132704 cases PASS; own-body instruction coverage 239 / 239 source; 241 / 242 original
TMCJPEGDEC_exif_parse 1832 cases PASS; own-body instruction coverage 212 / 212 source; 212 / 212 original
TOTAL 267282 paired EXIF calls PASS

View File

@@ -0,0 +1,82 @@
# JPEG follow-on reconstruction
Base: `813660b9`, branch `agent/bittle/jpeg-reconstruction-next`.
The already-reviewed EXIF candidate `b9305280` was preserved unchanged as local
cherry-pick `164dbbc5`. This log concerns a separate follow-on to that candidate.
## Retained change
IFD0 rational denominator addresses now use the same explicit offset grouping as
IFD1: add four to the TIFF offset, then add the thumbnail-data base. The redundant
same-type pointer cast is removed. This is a two-line source change; it introduces
no new tag interpretation, new memory access, or function.
- `TMCJPEGDEC_IFD0_tag_parse`: 99.386696 -> 99.49065 fuzzy
- EXIF unit: 98.94104 -> 98.98035 fuzzy
- EXIF exact functions: 3/6 -> 3/6; matched code remains 1348/5088 bytes
- IFD0 remains 480/481 instructions; this is not an exact match or linking result
- All other 33 owned functions retain their baseline scores
No changes are retained in the color converters or IDCT. No declaration-order
search, forced registers, volatile, assembly, padding, headers, configuration,
linking changes, remote writes, or edits to the entropy decoder were made.
## Rejected reconstruction trials
Every equal or regressing trial was restored before validation.
- Separate rational numerator/denominator pointer lifetimes in IFD0 and IFD1:
unchanged from the retained expression; restored
- RGB444's common edge-sibling channel evaluation and clipping order with a
separate luminance value: 98.71429 -> 98.62637; restored
- Explicit odd-coefficient differences in the luminance second-pass butterfly:
83.05058 -> 81.42802; restored
- Extract the luminance first-pass row transform to a static inline helper:
83.05058 -> 81.12062; restored. The trial script initially omitted helper
local declarations, causing a compiler error; that experiment was corrected,
measured, and then fully restored
- Extract RGBA pixel packing to a static inline helper:
- 411: 82.6033 -> 81.72314; restored
- 422 and edge: unchanged at 87.72298 and 90.08849; restored
- 420 and edge: unchanged at 89.40523 and 87.458336; restored
- 211 and edge: unchanged at 91.15385 and 92.0; restored
- 444: unchanged at 90.670105; restored
- 444 edge: 88.90909 -> 87.70909; restored
These were source-structure trials, not permutations of register declarations.
The source-neutral scheduling differences that remain were not chased through
artificial source changes.
## Verification
The baseline report was generated after building the unchanged cherry-picked
candidate. The final report was freshly generated after restoring every rejected
trial and rebuilding all four owned objects.
- Full 4.3U build: PASS
- All 34 owned function scores compared against the baseline: zero regressions
- All 11 previously exact functions retain exact-name objdiff 100 and ctxdiff 0
- Pool checks before trials and after the final build: all four identical,
zero strings
- Neither source nor original object has allocated non-code sections in any of
the four owned units
- DOL SHA1: `26116613f624061ba99c8d1a299aaa6efa85670d`
- The existing read-only bounded PPC audit tools were inspected before use;
their scripts are unchanged
- Converter/IDCT audit: all 9,456 paired calls pass
- EXIF audit: all 267,282 paired calls pass, including the 42 version-byte
overlap cases preserved by the preceding candidate
- `git diff --check 813660b9`: PASS, including the base-relative committed diff
- The working tree is paused for parent review after the focused local commit
These interpreter tests are bounded samples, not formal equivalence or Wii
runtime tests. Their existing coverage and modeling limitations still apply.
## Existing external overlap
The prior EXIF log documents the already fetched
`origin/agent/w1009/tmc` branch's overlap with EXIF and color reconstruction.
That overlap remains relevant to parent integration. This follow-on was derived
from the current source, the prior owned reconstruction, and local object
comparisons. No external branch was merged, copied, fetched, or contacted.
The parent should preserve the previously queued EXIF candidate separately.

View File

@@ -0,0 +1,63 @@
main/libs/RVLMiddleware/TMC_JPEG/src/buffer/idct_block_var 89.06892 -> 89.06892
POOL IDENTICAL up to 0 (mine=0 base=0)
No allocated non-code sections in either object
main/libs/RVLMiddleware/TMC_JPEG/src/exif/exif_parse 98.94104 -> 98.98035
POOL IDENTICAL up to 0 (mine=0 base=0)
TMCCJPEGDecGetOffsetEXIF: src 0x26c base 0x26c insns 155/155; diffs 0: []
TMCCJPEGDecGetInfoEXIF: src 0x210 base 0x210 insns 132/132; diffs 0: []
TMCJPEGDEC_IFD0_tag_parse 99.386696 -> 99.49065
TMCJPEGDEC_ThumbnailCheck: src 0xc8 base 0xc8 insns 50/50; diffs 0: []
No allocated non-code sections in either object
main/libs/RVLMiddleware/TMC_JPEG/src/texturecvtr/Texture_MCUtoRGB565 96.44502 -> 96.44502
POOL IDENTICAL up to 0 (mine=0 base=0)
TMCJPEGDEC_converterYUV422toRGB565edge: src 0x1ac base 0x1ac insns 107/107; diffs 0: []
TMCJPEGDEC_converterYUV211toRGB565: src 0x188 base 0x188 insns 98/98; diffs 0: []
TMCJPEGDEC_converterYUV211toRGB565edge: src 0x1b4 base 0x1b4 insns 109/109; diffs 0: []
TMCJPEGDEC_converterYUV444toRGB565edge: src 0x1a0 base 0x1a0 insns 104/104; diffs 0: []
TMCJPEGDEC_converterYUV400toRGB565: src 0x118 base 0x118 insns 70/70; diffs 0: []
TMCJPEGDEC_converterYUV400toRGB565edge: src 0x14c base 0x14c insns 83/83; diffs 0: []
No allocated non-code sections in either object
main/libs/RVLMiddleware/TMC_JPEG/src/texturecvtr/Texture_MCUtoRGBA8 90.32323 -> 90.32323
POOL IDENTICAL up to 0 (mine=0 base=0)
TMCJPEGDEC_converterYUV400toRGBA8: src 0x130 base 0x130 insns 76/76; diffs 0: []
TMCJPEGDEC_converterYUV400toRGBA8edge: src 0x164 base 0x164 insns 89/89; diffs 0: []
No allocated non-code sections in either object
34 owned function scores checked; 11 exact functions preserved; zero regressions
26116613f624061ba99c8d1a299aaa6efa85670d build/43U/main.dol
TMCJPEGDEC_IFD0_tag_parse 131072 selector cases PASS
TMCJPEGDEC_IFD1_tag_parse 131072 selector cases PASS
TMCJPEGDEC_IFD0_tag_parse 132746 cases PASS; own-body instruction coverage 480 / 480 source; 481 / 481 original
TMCJPEGDEC_IFD1_tag_parse 132704 cases PASS; own-body instruction coverage 239 / 239 source; 241 / 242 original
TMCJPEGDEC_exif_parse 1832 cases PASS; own-body instruction coverage 212 / 212 source; 212 / 212 original
TOTAL 267282 paired EXIF calls PASS
TMCJPEGDEC_converterYUV411toRGBA8: 100 cases PASS; instruction coverage source 242/242, original 242/242
TMCJPEGDEC_converterYUV411toRGBA8edge: 400 cases PASS; instruction coverage source 112/112, original 112/112
TMCJPEGDEC_converterYUV422toRGBA8: 100 cases PASS; instruction coverage source 148/148, original 148/148
TMCJPEGDEC_converterYUV422toRGBA8edge: 400 cases PASS; instruction coverage source 113/113, original 113/113
TMCJPEGDEC_converterYUV420toRGBA8: 100 cases PASS; instruction coverage source 153/153, original 153/153
TMCJPEGDEC_converterYUV420toRGBA8edge: 400 cases PASS; instruction coverage source 120/120, original 120/120
TMCJPEGDEC_converterYUV211toRGBA8: 100 cases PASS; instruction coverage source 104/104, original 104/104
TMCJPEGDEC_converterYUV211toRGBA8edge: 400 cases PASS; instruction coverage source 115/115, original 115/115
TMCJPEGDEC_converterYUV444toRGBA8: 100 cases PASS; instruction coverage source 97/97, original 97/97
TMCJPEGDEC_converterYUV444toRGBA8edge: 400 cases PASS; instruction coverage source 110/110, original 110/110
TMCJPEGDEC_converterYUV400toRGBA8: 100 cases PASS; instruction coverage source 57/76, original 57/76
TMCJPEGDEC_converterYUV400toRGBA8edge: 400 cases PASS; instruction coverage source 70/89, original 70/89
TMCJPEGDEC_set_converterRGBA8: 1024 cases PASS; instruction coverage source 168/169, original 169/170
TMCJPEGDEC_converterYUV411toRGB565: 100 cases PASS; instruction coverage source 217/217, original 217/217
TMCJPEGDEC_converterYUV411toRGB565edge: 400 cases PASS; instruction coverage source 106/106, original 106/106
TMCJPEGDEC_converterYUV422toRGB565: 100 cases PASS; instruction coverage source 136/136, original 136/136
TMCJPEGDEC_converterYUV422toRGB565edge: 400 cases PASS; instruction coverage source 107/107, original 107/107
TMCJPEGDEC_converterYUV420toRGB565: 100 cases PASS; instruction coverage source 141/141, original 141/141
TMCJPEGDEC_converterYUV420toRGB565edge: 400 cases PASS; instruction coverage source 114/114, original 114/114
TMCJPEGDEC_converterYUV211toRGB565: 100 cases PASS; instruction coverage source 98/98, original 98/98
TMCJPEGDEC_converterYUV211toRGB565edge: 400 cases PASS; instruction coverage source 109/109, original 109/109
TMCJPEGDEC_converterYUV444toRGB565: 100 cases PASS; instruction coverage source 91/91, original 91/91
TMCJPEGDEC_converterYUV444toRGB565edge: 400 cases PASS; instruction coverage source 104/104, original 104/104
TMCJPEGDEC_converterYUV400toRGB565: 100 cases PASS; instruction coverage source 51/70, original 51/70
TMCJPEGDEC_converterYUV400toRGB565edge: 400 cases PASS; instruction coverage source 64/83, original 64/83
TMCJPEGDEC_set_converterRGB565: 1024 cases PASS; instruction coverage source 168/169, original 169/170
TMCJPEGDEC_IdctBlock_Lumi: 704 cases PASS; instruction coverage source 257/257, original 257/257
TMCJPEGDEC_IdctBlock_Col: 704 cases PASS; instruction coverage source 454/454, original 454/454
TOTAL: 9456 differential cases PASS; GPR14-GPR31, stack pointer and LR preserved

View File

@@ -0,0 +1,276 @@
# JPEG remainder: semantic audit and source-shape reconstruction
Baseline: `449529d9`, branch `agent/bittle/allhands-jpeg-decompile`, 4.3U only.
No remote writes, header changes, configuration changes, linking changes, assembly,
volatile forcing, padding, or edits to the separately owned entropy decoder.
## Result
All 28 owned functions were already implemented. No semantic mismatch was found
in a bounded differential comparison of their current compiled PPC against their
original ELF function bodies. This is additional test evidence, not a formal
proof of equivalence and not a Wii runtime test.
Three natural source-shape changes improved the report without a function-score
regression:
- Luminance IDCT now uses explicit odd-coefficient sum/difference temporaries,
matching the existing chroma first-pass butterfly. Function fuzzy
82.31518 -> 83.05058; 257/257 instructions, 187 ctxdiff differences remain
- RGB444 consumes the luminance temporary as the red channel and uses the
existing exact edge sibling's red/green packing association. Function fuzzy
97.96703 -> 98.71429; 91/91 instructions, 13 ctxdiff differences remain
- RGBA420 edge consumes the luminance temporary as red after calculating the
other channels. Function fuzzy 87.4 -> 87.458336; 120/120 instructions,
77 ctxdiff differences remain
No new exact matches. This candidate is not an exact-matching or linking result.
| Unit | Exact before -> after | Fuzzy before -> after | Allocated data |
| --- | --- | --- | --- |
| Texture_MCUtoRGBA8 | 2/13 -> 2/13 | 90.318985 -> 90.32323 | 0 bytes; 100% |
| Texture_MCUtoRGB565 | 6/13 -> 6/13 | 96.40103 -> 96.44502 | 0 bytes; 100% |
| idct_block_var | 0/2 -> 0/2 | 88.80309 -> 89.06892 | 0 bytes; 100% |
All eight previously exact functions still have exact-name objdiff 100% and
ctxdiff zero differences. Each setter remains 169/170 instructions, with one
folded buffer-base address operation; the other 26 functions retain their
original instruction counts. Every owned function and both input objects were
checked for allocated non-code sections: none. Pool checks were run before tuning and after the final build; all three were
identical, with zero strings.
The baseline comparison was regenerated by rebuilding saved unmodified versions
of the three owned sources, reporting them, then restoring/rebuilding the final
sources. All 28 function scores were compared; zero regressions.
Full build: PASS. `git diff --check`: PASS.
DOL SHA1: `26116613f624061ba99c8d1a299aaa6efa85670d`.
Commands used:
```sh
python3 configure.py --version 43U --wrapper ../toolchain/wibo-build/wibo
WIBO_SJIS_MISSING_IMPORTS=1 ../.venv/bin/ninja
build/tools/objdiff-cli report generate -p . -o build/43U/report.json -f json
PYTHONPATH=../local-tools ../.venv/bin/python tools/decomp-assist/pool_diff.py <source-object> <original-object>
PYTHONPATH=../local-tools ../.venv/bin/python tools/decomp-assist/ctxdiff.py <unit> <symbol>
../.venv/bin/python tools/decomp-assist/jpeg_differential_audit.py
sha1sum build/43U/main.dol
```
## Differential test scope and limitations
The committed `jpeg_differential_audit.py` is a bounded integer PPC interpreter
and deterministic test generator. It reads the locally available source/original
ELF objects and does not distribute original executable bytes. Capstone and
pyelftools are required. Run it from the repository root after building the
three objects; do not run Python with assertions disabled.
- 9,456 paired calls across all 28 functions, seed 449529
- Normal and edge converters: scales 1, 2, 4, 8; multiple MCU origins; both
edge-dimension branches; zero, partial and full extents; constant and random
luminance/chroma bytes, including negative signed chroma and clipping
- Setters: all 256 component selectors and four supported IDCT sizes;
randomized 16-bit image dimensions; return values, converter identities and
non-stack side effects compared
- IDCT: all row extents 1..8 and column-mode representatives 1, 2, 3, 8;
DC-only, sparse and dense coefficients; positive/negative coefficients and
clipping-boundary DC values; pitches 8, 16, 32
- Every non-stack byte is compared, including guard regions, along with setter
return values. Stack pointer, LR and GPR14..GPR31 restoration are checked
- Full instruction-address coverage was reached for all 20 non-grayscale
converter bodies and both IDCT functions. This does not mean exhaustive
branch/path/input coverage. Four grayscale bodies retain 19 unvisited
instructions each in saturation paths that follow an unsigned byte load;
each setter has one unvisited negative-selector branch instruction
- Function address relocations are normalized to deterministic symbol tokens,
with collision checks. Real linked addresses and integration are not tested
- Only the observed integer instruction subset is implemented. Unsupported
instructions/relocations/callees fail closed. No FPU, exceptions, cache/MMIO,
timing, alignment faults, XER behavior or complete condition-register model
- `memset` is modeled by its memory effects and ABI-permitted volatile-register
clobbers; its actual library machine code is not executed
- Inputs are bounded samples. Invalid pointers, zero scale, arbitrary invalid
IDCT descriptors, all 32-bit coefficient patterns, concurrency, real JPEG
streams and other units are not covered
- This custom interpreter is itself test tooling rather than a formally verified
CPU implementation. Official objdiff, ctxdiff and full-DOL checks remain the
matching/integration authority
Exact run output is preserved in `jpeg-remainder-differential-results.txt`.
## Attempts
Only strictly improved candidates without other function regressions were kept.
No declaration-order permutation search was performed. Failed/equal candidates
were restored. The one temporary compile error in the luminance-consumption
script was a duplicate local declaration; it was corrected and no broken
candidate was retained.
### type-trials
```text
TMCJPEGDEC_converterYUV411toRGBA8 signed-chroma-pointers: 82.603300 -> 82.500000; restore
TMCJPEGDEC_converterYUV411toRGBA8 signed-chroma-coefficients: 82.603300 -> 82.603300; restore
TMCJPEGDEC_converterYUV411toRGBA8 signed-chroma-domain: 82.603300 -> 81.570250; restore
TMCJPEGDEC_converterYUV411toRGBA8edge signed-chroma-pointers: 93.883930 -> 91.964290; restore
TMCJPEGDEC_converterYUV411toRGBA8edge signed-chroma-coefficients: 93.883930 -> 93.750000; restore
TMCJPEGDEC_converterYUV411toRGBA8edge signed-chroma-domain: 93.883930 -> 92.187500; restore
TMCJPEGDEC_converterYUV422toRGBA8 signed-chroma-pointers: 87.722980 -> 87.722980; restore
TMCJPEGDEC_converterYUV422toRGBA8 signed-chroma-coefficients: 87.722980 -> 87.722980; restore
TMCJPEGDEC_converterYUV422toRGBA8 signed-chroma-domain: 87.722980 -> 87.722980; restore
TMCJPEGDEC_converterYUV422toRGBA8 byte-luminance: 87.722980 -> 87.689186; restore
TMCJPEGDEC_converterYUV422toRGBA8edge signed-chroma-pointers: 90.088490 -> 90.088490; restore
TMCJPEGDEC_converterYUV422toRGBA8edge signed-chroma-coefficients: 90.088490 -> 90.088490; restore
TMCJPEGDEC_converterYUV422toRGBA8edge signed-chroma-domain: 90.088490 -> 88.628320; restore
TMCJPEGDEC_converterYUV422toRGBA8edge byte-luminance: 90.088490 -> 90.088490; restore
TMCJPEGDEC_converterYUV420toRGBA8 signed-chroma-pointers: 89.405230 -> 89.176470; restore
TMCJPEGDEC_converterYUV420toRGBA8 signed-chroma-coefficients: 89.405230 -> 89.307190; restore
TMCJPEGDEC_converterYUV420toRGBA8 signed-chroma-domain: 89.405230 -> 86.464050; restore
TMCJPEGDEC_converterYUV420toRGBA8 byte-luminance: 89.405230 -> 87.379080; restore
TMCJPEGDEC_converterYUV420toRGBA8edge signed-chroma-pointers: 87.400000 -> 80.666664; restore
TMCJPEGDEC_converterYUV420toRGBA8edge signed-chroma-domain: 87.400000 -> 80.666664; restore
TMCJPEGDEC_converterYUV211toRGBA8 signed-chroma-pointers: 91.153850 -> 91.153850; restore
TMCJPEGDEC_converterYUV211toRGBA8 signed-chroma-coefficients: 91.153850 -> 91.153850; restore
TMCJPEGDEC_converterYUV211toRGBA8 signed-chroma-domain: 91.153850 -> 91.153850; restore
TMCJPEGDEC_converterYUV211toRGBA8 byte-luminance: 91.153850 -> 91.153850; restore
TMCJPEGDEC_converterYUV211toRGBA8edge signed-chroma-pointers: 92.000000 -> 91.869570; restore
TMCJPEGDEC_converterYUV211toRGBA8edge signed-chroma-coefficients: 92.000000 -> 91.869570; restore
TMCJPEGDEC_converterYUV211toRGBA8edge signed-chroma-domain: 92.000000 -> 91.869570; restore
TMCJPEGDEC_converterYUV444toRGBA8 signed-chroma-pointers: 90.670105 -> 90.515465; restore
TMCJPEGDEC_converterYUV444toRGBA8 signed-chroma-coefficients: 90.670105 -> 90.515465; restore
TMCJPEGDEC_converterYUV444toRGBA8 signed-chroma-domain: 90.670105 -> 90.618550; restore
TMCJPEGDEC_converterYUV444toRGBA8 byte-luminance: 90.670105 -> 90.670105; restore
TMCJPEGDEC_converterYUV444toRGBA8edge signed-chroma-pointers: 88.909090 -> 88.909090; restore
TMCJPEGDEC_converterYUV444toRGBA8edge signed-chroma-coefficients: 88.909090 -> 88.909090; restore
TMCJPEGDEC_converterYUV444toRGBA8edge signed-chroma-domain: 88.909090 -> 87.645454; restore
TMCJPEGDEC_converterYUV444toRGBA8edge byte-luminance: 88.909090 -> 88.909090; restore
TMCJPEGDEC_converterYUV411toRGB565 signed-chroma-pointers: 90.599075 -> 90.599075; restore
TMCJPEGDEC_converterYUV411toRGB565 signed-chroma-coefficients: 90.599075 -> 90.599075; restore
TMCJPEGDEC_converterYUV411toRGB565 signed-chroma-domain: 90.599075 -> 90.599075; restore
TMCJPEGDEC_converterYUV422toRGB565 signed-chroma-pointers: 96.691180 -> 96.691180; restore
TMCJPEGDEC_converterYUV422toRGB565 signed-chroma-coefficients: 96.691180 -> 96.691180; restore
TMCJPEGDEC_converterYUV422toRGB565 signed-chroma-domain: 96.691180 -> 96.691180; restore
TMCJPEGDEC_converterYUV420toRGB565 signed-chroma-pointers: 94.609930 -> 94.609930; restore
TMCJPEGDEC_converterYUV420toRGB565 signed-chroma-coefficients: 94.609930 -> 94.609930; restore
TMCJPEGDEC_converterYUV420toRGB565 signed-chroma-domain: 94.609930 -> 94.609930; restore
TMCJPEGDEC_converterYUV420toRGB565 byte-luminance: 94.609930 -> 92.907800; restore
TMCJPEGDEC_converterYUV444toRGB565 signed-chroma-pointers: 97.967030 -> 97.362640; restore
TMCJPEGDEC_converterYUV444toRGB565 signed-chroma-coefficients: 97.967030 -> 97.802200; restore
TMCJPEGDEC_converterYUV444toRGB565 signed-chroma-domain: 97.967030 -> 97.340660; restore
TMCJPEGDEC_converterYUV444toRGB565 byte-luminance: 97.967030 -> 91.373630; restore
```
### idct-trials
```text
TMCJPEGDEC_IdctBlock_Lumi nested-odd-output-sum: 82.315180 -> 79.708170; restore
TMCJPEGDEC_IdctBlock_Lumi explicit-ac-or-tree: 82.315180 -> 74.984436; restore
TMCJPEGDEC_IdctBlock_Lumi ac-coefficients-direct-condition: 82.315180 -> 74.984436; restore
TMCJPEGDEC_IdctBlock_Lumi nested-even-output-sum: 82.315180 -> 82.315180; restore
TMCJPEGDEC_IdctBlock_Lumi direct-product-shifts: 82.315180 -> 82.315180; restore
TMCJPEGDEC_IdctBlock_Col nested-odd-output-sum: 92.475770 -> 87.907486; restore
TMCJPEGDEC_IdctBlock_Col explicit-ac-or-tree: 92.475770 -> 90.651985; restore
TMCJPEGDEC_IdctBlock_Col ac-coefficients-direct-condition: 92.475770 -> 90.651985; restore
TMCJPEGDEC_IdctBlock_Col nested-even-output-sum: 92.475770 -> 92.475770; restore
TMCJPEGDEC_IdctBlock_Col direct-product-shifts: 92.475770 -> 92.475770; restore
```
### packing-trials
```text
TMCJPEGDEC_converterYUV411toRGBA8 explicit-channel-clamp-branches: 82.603300 -> 61.962810; restore
TMCJPEGDEC_converterYUV411toRGBA8 structured-nonnegative-clamp: 82.603300 -> 68.442150; restore
TMCJPEGDEC_converterYUV411toRGBA8edge explicit-channel-clamp-branches: 93.883930 -> 80.008930; restore
TMCJPEGDEC_converterYUV411toRGBA8edge structured-nonnegative-clamp: 93.883930 -> 85.714290; restore
TMCJPEGDEC_converterYUV422toRGBA8 explicit-channel-clamp-branches: 87.722980 -> 71.222980; restore
TMCJPEGDEC_converterYUV422toRGBA8 structured-nonnegative-clamp: 87.722980 -> 75.533780; restore
TMCJPEGDEC_converterYUV422toRGBA8edge explicit-channel-clamp-branches: 90.088490 -> 76.097340; restore
TMCJPEGDEC_converterYUV422toRGBA8edge structured-nonnegative-clamp: 90.088490 -> 82.123890; restore
TMCJPEGDEC_converterYUV420toRGBA8 explicit-channel-clamp-branches: 89.405230 -> 73.372550; restore
TMCJPEGDEC_converterYUV420toRGBA8 structured-nonnegative-clamp: 89.405230 -> 76.797386; restore
TMCJPEGDEC_converterYUV420toRGBA8edge explicit-channel-clamp-branches: 87.400000 -> 77.841670; restore
TMCJPEGDEC_converterYUV420toRGBA8edge structured-nonnegative-clamp: 87.400000 -> 82.400000; restore
TMCJPEGDEC_converterYUV211toRGBA8 explicit-channel-clamp-branches: 91.153850 -> 74.701920; restore
TMCJPEGDEC_converterYUV211toRGBA8 structured-nonnegative-clamp: 91.153850 -> 82.211540; restore
TMCJPEGDEC_converterYUV211toRGBA8edge explicit-channel-clamp-branches: 92.000000 -> 77.730440; restore
TMCJPEGDEC_converterYUV211toRGBA8edge structured-nonnegative-clamp: 92.000000 -> 84.043480; restore
TMCJPEGDEC_converterYUV444toRGBA8 explicit-channel-clamp-branches: 90.670105 -> 73.030930; restore
TMCJPEGDEC_converterYUV444toRGBA8 structured-nonnegative-clamp: 90.670105 -> 81.082470; restore
TMCJPEGDEC_converterYUV444toRGBA8edge explicit-channel-clamp-branches: 88.909090 -> 75.463640; restore
TMCJPEGDEC_converterYUV444toRGBA8edge structured-nonnegative-clamp: 88.909090 -> 81.000000; restore
TMCJPEGDEC_converterYUV411toRGB565 normalize-packed-channels: 90.599075 -> 87.004610; restore
TMCJPEGDEC_converterYUV411toRGB565 group-red-green-packing: 90.599075 -> 88.755760; restore
TMCJPEGDEC_converterYUV411toRGB565 pack-before-output-address: 90.599075 -> 90.599075; restore
TMCJPEGDEC_converterYUV411toRGB565 explicit-channel-clamp-branches: 90.599075 -> 59.511520; restore
TMCJPEGDEC_converterYUV411toRGB565 structured-nonnegative-clamp: 90.599075 -> 71.640560; restore
TMCJPEGDEC_converterYUV411toRGB565edge normalize-packed-channels: 95.377360 -> 93.820755; restore
TMCJPEGDEC_converterYUV411toRGB565edge group-red-green-packing: 95.377360 -> 93.613205; restore
TMCJPEGDEC_converterYUV411toRGB565edge pack-before-output-address: 95.377360 -> 95.377360; restore
TMCJPEGDEC_converterYUV411toRGB565edge explicit-channel-clamp-branches: 95.377360 -> 79.801890; restore
TMCJPEGDEC_converterYUV411toRGB565edge structured-nonnegative-clamp: 95.377360 -> 86.179245; restore
TMCJPEGDEC_converterYUV422toRGB565 explicit-channel-clamp-branches: 96.691180 -> 78.507355; restore
TMCJPEGDEC_converterYUV422toRGB565 structured-nonnegative-clamp: 96.691180 -> 79.044120; restore
TMCJPEGDEC_converterYUV420toRGB565 normalize-packed-channels: 94.609930 -> 90.248230; restore
TMCJPEGDEC_converterYUV420toRGB565 group-red-green-packing: 94.609930 -> 91.560290; restore
TMCJPEGDEC_converterYUV420toRGB565 pack-before-output-address: 94.609930 -> 94.609930; restore
TMCJPEGDEC_converterYUV420toRGB565 explicit-channel-clamp-branches: 94.609930 -> 73.127660; restore
TMCJPEGDEC_converterYUV420toRGB565 structured-nonnegative-clamp: 94.609930 -> 76.751770; restore
TMCJPEGDEC_converterYUV420toRGB565edge normalize-packed-channels: 94.692986 -> 89.912280; restore
TMCJPEGDEC_converterYUV420toRGB565edge group-red-green-packing: 94.692986 -> 91.798250; restore
TMCJPEGDEC_converterYUV420toRGB565edge pack-before-output-address: 94.692986 -> 94.692986; restore
TMCJPEGDEC_converterYUV420toRGB565edge explicit-channel-clamp-branches: 94.692986 -> 81.894740; restore
TMCJPEGDEC_converterYUV420toRGB565edge structured-nonnegative-clamp: 94.692986 -> 86.315790; restore
TMCJPEGDEC_converterYUV444toRGB565 normalize-packed-channels: 97.967030 -> 96.648350; restore
TMCJPEGDEC_converterYUV444toRGB565 group-red-green-packing: 97.967030 -> 98.681320; RETAIN
TMCJPEGDEC_converterYUV444toRGB565 explicit-channel-clamp-branches: 98.681320 -> 85.120880; restore
TMCJPEGDEC_converterYUV444toRGB565 structured-nonnegative-clamp: 98.681320 -> 82.802200; restore
```
### control-trials
```text
structured-col-low-mode-first: {'TMCJPEGDEC_IdctBlock_Lumi': 82.31518, 'TMCJPEGDEC_IdctBlock_Col': 92.47577} -> {'TMCJPEGDEC_IdctBlock_Lumi': 82.31518, 'TMCJPEGDEC_IdctBlock_Col': 92.47577}; restore
structured-col-high-mode-first: {'TMCJPEGDEC_IdctBlock_Lumi': 82.31518, 'TMCJPEGDEC_IdctBlock_Col': 92.47577} -> {'TMCJPEGDEC_IdctBlock_Lumi': 82.31518, 'TMCJPEGDEC_IdctBlock_Col': 50.55066}; restore
shared-full-row-butterfly: {'TMCJPEGDEC_IdctBlock_Lumi': 82.31518, 'TMCJPEGDEC_IdctBlock_Col': 92.47577} -> {'TMCJPEGDEC_IdctBlock_Lumi': 83.05058, 'TMCJPEGDEC_IdctBlock_Col': 92.47577}; RETAIN
```
### luma-reuse-trials
```text
TMCJPEGDEC_converterYUV422toRGBA8edge consume-luminance-as-blue: 90.088490 -> 90.088490; restore
TMCJPEGDEC_converterYUV422toRGBA8edge consume-luminance-as-red: 90.088490 -> 89.955750; restore
TMCJPEGDEC_converterYUV422toRGBA8edge consume-luminance-as-green: 90.088490 -> 90.088490; restore
TMCJPEGDEC_converterYUV420toRGBA8edge consume-luminance-as-blue: 87.400000 -> 87.400000; restore
TMCJPEGDEC_converterYUV420toRGBA8edge consume-luminance-as-red: 87.400000 -> 87.458336; RETAIN
TMCJPEGDEC_converterYUV211toRGBA8 consume-luminance-as-blue: 91.153850 -> 91.153850; restore
TMCJPEGDEC_converterYUV211toRGBA8 consume-luminance-as-red: 91.153850 -> 86.067310; restore
TMCJPEGDEC_converterYUV211toRGBA8 consume-luminance-as-green: 91.153850 -> 78.307690; restore
TMCJPEGDEC_converterYUV211toRGBA8edge consume-luminance-as-blue: 92.000000 -> 91.565216; restore
TMCJPEGDEC_converterYUV211toRGBA8edge consume-luminance-as-red: 92.000000 -> 87.617390; restore
TMCJPEGDEC_converterYUV211toRGBA8edge consume-luminance-as-green: 92.000000 -> 79.547830; restore
TMCJPEGDEC_converterYUV444toRGBA8 consume-luminance-as-blue: 90.670105 -> 90.670105; restore
TMCJPEGDEC_converterYUV444toRGBA8 consume-luminance-as-red: 90.670105 -> 85.216490; restore
TMCJPEGDEC_converterYUV444toRGBA8 consume-luminance-as-green: 90.670105 -> 74.577320; restore
TMCJPEGDEC_converterYUV444toRGBA8edge consume-luminance-as-blue: 88.909090 -> 88.909090; restore
TMCJPEGDEC_converterYUV444toRGBA8edge consume-luminance-as-red: 88.909090 -> 85.645454; restore
TMCJPEGDEC_converterYUV444toRGBA8edge consume-luminance-as-green: 88.909090 -> 78.772730; restore
TMCJPEGDEC_converterYUV411toRGB565edge consume-luminance-as-blue: 95.377360 -> 91.396225; restore
TMCJPEGDEC_converterYUV411toRGB565edge consume-luminance-as-red: 95.377360 -> 93.707550; restore
TMCJPEGDEC_converterYUV411toRGB565edge consume-luminance-as-green: 95.377360 -> 95.330185; restore
TMCJPEGDEC_converterYUV444toRGB565 consume-luminance-as-blue: 98.681320 -> 98.681320; restore
TMCJPEGDEC_converterYUV444toRGB565 consume-luminance-as-red: 98.681320 -> 98.714290; RETAIN
```
### rgb444-trials
```text
TMCJPEGDEC_converterYUV444toRGB565 reused-luminance-green-red-group: 98.714290 -> 89.923080; restore
TMCJPEGDEC_converterYUV444toRGB565 reused-luminance-blue-green-group: 98.714290 -> 98.549450; restore
TMCJPEGDEC_converterYUV444toRGB565 reused-luminance-red-blue-group: 98.714290 -> 90.945050; restore
TMCJPEGDEC_converterYUV444toRGB565 reused-luminance-red-green-group: 98.714290 -> 98.714290; restore
TMCJPEGDEC_converterYUV444toRGB565 separate-luminance-green-red-group: 98.714290 -> 98.131870; restore
TMCJPEGDEC_converterYUV444toRGB565 separate-luminance-blue-green-group: 98.714290 -> 98.516490; restore
TMCJPEGDEC_converterYUV444toRGB565 separate-luminance-red-blue-group: 98.714290 -> 91.406590; restore
TMCJPEGDEC_converterYUV444toRGB565 separate-luminance-red-green-group: 98.714290 -> 98.681320; restore
```

View File

@@ -0,0 +1,29 @@
TMCJPEGDEC_converterYUV411toRGBA8: 100 cases PASS; instruction coverage source 242/242, original 242/242
TMCJPEGDEC_converterYUV411toRGBA8edge: 400 cases PASS; instruction coverage source 112/112, original 112/112
TMCJPEGDEC_converterYUV422toRGBA8: 100 cases PASS; instruction coverage source 148/148, original 148/148
TMCJPEGDEC_converterYUV422toRGBA8edge: 400 cases PASS; instruction coverage source 113/113, original 113/113
TMCJPEGDEC_converterYUV420toRGBA8: 100 cases PASS; instruction coverage source 153/153, original 153/153
TMCJPEGDEC_converterYUV420toRGBA8edge: 400 cases PASS; instruction coverage source 120/120, original 120/120
TMCJPEGDEC_converterYUV211toRGBA8: 100 cases PASS; instruction coverage source 104/104, original 104/104
TMCJPEGDEC_converterYUV211toRGBA8edge: 400 cases PASS; instruction coverage source 115/115, original 115/115
TMCJPEGDEC_converterYUV444toRGBA8: 100 cases PASS; instruction coverage source 97/97, original 97/97
TMCJPEGDEC_converterYUV444toRGBA8edge: 400 cases PASS; instruction coverage source 110/110, original 110/110
TMCJPEGDEC_converterYUV400toRGBA8: 100 cases PASS; instruction coverage source 57/76, original 57/76
TMCJPEGDEC_converterYUV400toRGBA8edge: 400 cases PASS; instruction coverage source 70/89, original 70/89
TMCJPEGDEC_set_converterRGBA8: 1024 cases PASS; instruction coverage source 168/169, original 169/170
TMCJPEGDEC_converterYUV411toRGB565: 100 cases PASS; instruction coverage source 217/217, original 217/217
TMCJPEGDEC_converterYUV411toRGB565edge: 400 cases PASS; instruction coverage source 106/106, original 106/106
TMCJPEGDEC_converterYUV422toRGB565: 100 cases PASS; instruction coverage source 136/136, original 136/136
TMCJPEGDEC_converterYUV422toRGB565edge: 400 cases PASS; instruction coverage source 107/107, original 107/107
TMCJPEGDEC_converterYUV420toRGB565: 100 cases PASS; instruction coverage source 141/141, original 141/141
TMCJPEGDEC_converterYUV420toRGB565edge: 400 cases PASS; instruction coverage source 114/114, original 114/114
TMCJPEGDEC_converterYUV211toRGB565: 100 cases PASS; instruction coverage source 98/98, original 98/98
TMCJPEGDEC_converterYUV211toRGB565edge: 400 cases PASS; instruction coverage source 109/109, original 109/109
TMCJPEGDEC_converterYUV444toRGB565: 100 cases PASS; instruction coverage source 91/91, original 91/91
TMCJPEGDEC_converterYUV444toRGB565edge: 400 cases PASS; instruction coverage source 104/104, original 104/104
TMCJPEGDEC_converterYUV400toRGB565: 100 cases PASS; instruction coverage source 51/70, original 51/70
TMCJPEGDEC_converterYUV400toRGB565edge: 400 cases PASS; instruction coverage source 64/83, original 64/83
TMCJPEGDEC_set_converterRGB565: 1024 cases PASS; instruction coverage source 168/169, original 169/170
TMCJPEGDEC_IdctBlock_Lumi: 704 cases PASS; instruction coverage source 257/257, original 257/257
TMCJPEGDEC_IdctBlock_Col: 704 cases PASS; instruction coverage source 454/454, original 454/454
TOTAL: 9456 differential cases PASS; GPR14-GPR31, stack pointer and LR preserved

View File

@@ -0,0 +1,319 @@
#!/usr/bin/env python3
"""Read-only, bounded PPC integer interpreter for differential JPEG checks.
Loads source and original ELF .text; never copies original code into sources.
Only memset is external. No FPU/system instructions are supported.
"""
import random
import re
import struct
import zlib
import capstone
from elftools.elf.elffile import ELFFile
U32 = 0xFFFFFFFF
def signed(x):
x &= U32
return x - 0x100000000 if x & 0x80000000 else x
def rotate(x, n):
return (x << n | x >> (32 - n if n else 32)) & U32
def mask(b, e):
return sum((1 << 31 - i for i in range(32) if b <= i <= e or (b > e and (i >= b or i <= e))))
def load_object(path):
with open(path, 'rb') as f:
e = ELFFile(f)
section = e.get_section_by_name('.text')
data = bytearray(section.data())
syms = e.get_section_by_name('.symtab')
calls = {}
symbol_addresses = {}
rel = e.get_section_by_name('.rela.text')
if rel:
for r in rel.iter_relocations():
name = syms.get_symbol(r['r_info_sym']).name
off = r['r_offset']
typ = r['r_info_type']
addr = 0x800000 + (zlib.crc32(name.encode()) & 0x7FFF) * 4 + r['r_addend']
assert addr not in symbol_addresses or symbol_addresses[addr] == name
symbol_addresses[addr] = name
if typ == 10:
calls[off] = name
elif typ == 4:
data[off:off + 2] = struct.pack('>H', addr & 65535)
elif typ == 6:
data[off:off + 2] = struct.pack('>H', addr + 0x8000 >> 16 & 65535)
else:
raise ValueError(('relocation', typ))
dis = capstone.Cs(capstone.CS_ARCH_PPC, capstone.CS_MODE_32 | capstone.CS_MODE_BIG_ENDIAN)
funcs = {}
for sym in syms.iter_symbols():
if sym['st_info']['type'] != 'STT_FUNC' or not sym['st_size']:
continue
start = sym['st_value']
size = sym['st_size']
inst = {}
for i in dis.disasm(bytes(data[start:start + size]), start):
ops = []
for o in i.op_str.split(', '):
if not o:
continue
if o.startswith('cr'):
ops.append(('cr', int(o[2:])))
elif o.startswith('r'):
ops.append(int(o[1:]))
elif '(' in o:
off, reg = o.split('(')
ops.append((int(off, 0), int(reg[1:-1])))
else:
ops.append(int(o, 0))
inst[i.address] = (i.mnemonic, ops)
funcs[sym.name] = (start, inst, calls)
return funcs
def run(fn, initial, args):
pc, ins, calls = fn
mem = bytearray(initial)
r = [i * 1122867 & U32 for i in range(32)]
r[1] = 0x3F000
r[3:3 + len(args)] = args
saved_registers = r[14:].copy()
cr = [0] * 8
ctr = 0
lr = 0
steps = 0
coverage = set()
def ld(a, n):
assert 0 <= a <= len(mem) - n, hex(a)
return int.from_bytes(mem[a:a + n], 'big')
def st(a, v, n):
assert 0 <= a <= len(mem) - n, hex(a)
mem[a:a + n] = (v & (1 << 8 * n) - 1).to_bytes(n, 'big')
while True:
steps += 1
if steps > 100000:
raise RuntimeError('step bound')
coverage.add(pc)
op, a = ins[pc]
pc += 4
dot = op.endswith('.')
op = op.rstrip('.')
val = None
if op == 'blr':
break
elif op == 'bl':
assert calls.get(pc - 4) == 'memset', calls.get(pc - 4)
assert r[3] + r[5] <= len(mem)
mem[r[3]:r[3] + r[5]] = bytes([r[4] & 255]) * r[5]
# Model ABI-permitted call clobbers rather than preserving temporaries.
for reg in [0] + list(range(4, 13)):
r[reg] = (0xBAD00000 + reg) & U32
for field in [0, 1, 5, 6, 7]:
cr[field] = -1
lr = pc
elif op == 'b':
pc = a[0]
elif op == 'bdnz':
ctr = ctr - 1 & U32
if ctr:
pc = a[0]
elif op in ['beq', 'bne', 'ble', 'bge', 'blt', 'bgt']:
c = cr[a[0][1]] if len(a) > 1 else cr[0]
if {'beq': c == 0, 'bne': c != 0, 'ble': c <= 0, 'bge': c >= 0, 'blt': c < 0, 'bgt': c > 0}[op]:
pc = a[-1]
elif op in ['cmpwi', 'cmpw', 'cmplw']:
c = a[0][1] if isinstance(a[0], tuple) else 0
a = ins[pc - 4][1]
if isinstance(a[0], tuple):
a = a[1:]
x = r[a[0]]
y = a[1] if op == 'cmpwi' else r[a[1]]
if op != 'cmplw':
x = signed(x)
y = signed(y)
cr[c] = (x > y) - (x < y)
elif op in ['lbz', 'lhz', 'lwz', 'lmw']:
off, reg = a[1]
addr = (r[reg] if reg else 0) + off & U32
if op == 'lmw':
for j in range(a[0], 32):
r[j] = ld(addr + (j - a[0]) * 4, 4)
else:
r[a[0]] = ld(addr, {'lbz': 1, 'lhz': 2, 'lwz': 4}[op])
elif op in ['stb', 'sth', 'stw', 'stwu', 'stmw']:
off, reg = a[1]
addr = (r[reg] if reg else 0) + off & U32
if op == 'stmw':
for j in range(a[0], 32):
st(addr + (j - a[0]) * 4, r[j], 4)
else:
st(addr, r[a[0]], {'stb': 1, 'sth': 2, 'stw': 4, 'stwu': 4}[op])
if op == 'stwu':
r[reg] = addr
elif op in ['stbx', 'sthx']:
st((r[a[1]] if a[1] else 0) + r[a[2]] & U32, r[a[0]], 1 if op == 'stbx' else 2)
elif op == 'mflr':
r[a[0]] = lr
elif op == 'mtlr':
lr = r[a[0]]
elif op == 'mtctr':
ctr = r[a[0]]
else:
d = a[0]
if op == 'li':
val = a[1]
elif op == 'lis':
val = a[1] << 16
elif op == 'mr':
val = r[a[1]]
elif op == 'add':
val = r[a[1]] + r[a[2]]
elif op == 'addi':
val = (r[a[1]] if a[1] else 0) + a[2]
elif op == 'addis':
val = (r[a[1]] if a[1] else 0) + (a[2] << 16)
elif op == 'subf':
val = r[a[2]] - r[a[1]]
elif op == 'subfic':
val = a[2] - r[a[1]]
elif op == 'mulli':
val = r[a[1]] * a[2]
elif op == 'mullw':
val = r[a[1]] * r[a[2]]
elif op == 'divw':
x = signed(r[a[1]])
y = signed(r[a[2]])
val = abs(x) // abs(y) * (-1 if (x < 0) != (y < 0) else 1)
elif op == 'or':
val = r[a[1]] | r[a[2]]
elif op == 'andc':
val = r[a[1]] & ~r[a[2]]
elif op == 'andi':
val = r[a[1]] & a[2]
elif op == 'neg':
val = -r[a[1]]
elif op == 'extsb':
val = (r[a[1]] & 255) - (256 if r[a[1]] & 128 else 0)
elif op == 'srawi':
val = signed(r[a[1]]) >> a[2]
elif op == 'srwi':
val = r[a[1]] >> a[2]
elif op == 'slwi':
val = r[a[1]] << a[2]
elif op == 'clrlwi':
val = r[a[1]] & (1 << 32 - a[2]) - 1
elif op == 'rotlwi':
val = rotate(r[a[1]], a[2])
elif op == 'rlwinm':
val = rotate(r[a[1]], a[2]) & mask(a[3], a[4])
else:
raise ValueError((hex(pc - 4), op, a))
r[d] = val & U32
if dot:
cr[0] = (signed(val) > 0) - (signed(val) < 0)
assert r[1] == 0x3F000, 'Stack pointer not restored'
assert lr == 0, 'Link register not restored'
assert r[14:] == saved_registers, 'Nonvolatile registers not restored'
return (mem, r[3], coverage)
def put(m, a, v, n=4):
m[a:a + n] = (v & (1 << 8 * n) - 1).to_bytes(n, 'big')
def compare(src, orig, name, m, args, case, returns=False):
a, ra, ca = run(src[name], m, args)
b, rb, cb = run(orig[name], m, args)
# The stack can differ because each implementation owns its frame.
if a[:0x3E000] != b[:0x3E000]:
diff = [hex(i) for i in range(0x3E000) if a[i] != b[i]][:12]
raise AssertionError((name, case, 'memory mismatch', diff))
if returns and ra != rb:
raise AssertionError((name, case, 'return', ra, rb))
return (ca, cb)
def main():
total_cases = 0
rng = random.Random(449529)
root = 'build/43U/{}/libs/RVLMiddleware/TMC_JPEG/src/{}'
for unit in ['texturecvtr/Texture_MCUtoRGBA8', 'texturecvtr/Texture_MCUtoRGB565', 'buffer/idct_block_var']:
src = load_object(root.format('src', unit) + '.o')
orig = load_object(root.format('obj', unit) + '.o')
assert src.keys() == orig.keys(), (unit, 'Function sets differ')
assert len(src) == (2 if unit.startswith('buffer/') else 13)
for name in src:
n = 0
sa = set()
sb = set()
if 'IdctBlock' in name:
for rows in range(1, 9):
for cols in [1, 2, 3, 8]:
for pattern in range(22):
m = bytearray([0xA5]) * 0x40000
block = [0] * 64
for y in range(rows):
for x in range(cols):
if pattern < 6:
block[y * 8 + x] = [0, 1, -1, 0x3FFFF, 0x40000, -0x40000][pattern] if x == 0 and y == 0 else 0
elif pattern < 14:
block[y * 8 + x] = rng.randrange(-10000, 10001) if rng.randrange(4) == 0 else 0
else:
block[y * 8 + x] = rng.randrange(-150000, 150001)
for i, v in enumerate(block):
put(m, 0x1000 + i * 4, v)
pitch = [8, 16, 32][pattern % 3]
args = [0x1000, 0x2000, pitch, rows << 4 | cols]
ca, cb = compare(src, orig, name, m, args, (rows, cols, pattern))
sa |= ca
sb |= cb
n += 1
elif 'set_converter' in name:
for component in range(256):
for mode in [1, 2, 4, 8]:
m = bytearray([0xA5]) * 0x40000
work = 0x1000
state = 0x4000
put(m, work + 0x17FC, component, 1)
put(m, work + 0x19DC, mode, 1)
put(m, work + 0x19E4, state)
put(m, state + 0x24, rng.randrange(65536), 2)
put(m, state + 0x26, rng.randrange(65536), 2)
ca, cb = compare(src, orig, name, m, [work], (component, mode), True)
sa |= ca
sb |= cb
n += 1
else:
sampling = re.search('YUV(\\d+)to', name)[1]
w, h = {'411': (32, 8), '422': (16, 8), '420': (16, 16), '211': (8, 16), '444': (8, 8), '400': (8, 8)}[sampling]
for scale in [1, 2, 4, 8]:
for pattern in range(25):
for border in range(4 if name.endswith('edge') else 1):
m = bytearray([0xA5]) * 0x40000
work = 0x1000
state = 0x4000
tex = 0x8000
x = 32 * (pattern % 2)
y = 16 * (pattern % 3)
ww = w // scale
hh = h // scale
put(m, work + 0x19E4, state)
put(m, state + 0x20, scale, 1)
put(m, state + 0x2C, 96)
put(m, state + 0x48, tex)
put(m, state + 0x18, x if border & 1 else x + 1)
put(m, state + 0x1C, y if border & 2 else y + 1)
put(m, state + 0x16, pattern % (ww + 1), 1)
put(m, state + 0x17, pattern % (hh + 1), 1)
for i in range(0x184):
m[work + 0x1858 + i] = [0, 127, 128, 255][pattern % 4] if pattern < 4 else rng.randrange(256)
ca, cb = compare(src, orig, name, m, [work, x, y], (scale, pattern, border))
sa |= ca
sb |= cb
n += 1
total_cases += n
print(f'{name}: {n} cases PASS; instruction coverage source {len(sa)}/{len(src[name][1])}, original {len(sb)}/{len(orig[name][1])}', flush=True)
print(f'TOTAL: {total_cases} differential cases PASS; GPR14-GPR31, stack pointer and LR preserved')
if __name__ == '__main__':
main()

View File

@@ -0,0 +1,368 @@
#!/usr/bin/env python3
"""Read-only, bounded PPC integer interpreter for differential EXIF checks.
Loads source and original ELF .text; never copies original code into sources.
Only internal EXIF calls are supported. No FPU/system instructions are supported.
"""
import random
import re
import struct
import zlib
import capstone
from elftools.elf.elffile import ELFFile
U32 = 0xFFFFFFFF
def signed(x):
x &= U32
return x - 0x100000000 if x & 0x80000000 else x
def rotate(x, n):
return (x << n | x >> (32 - n if n else 32)) & U32
def mask(b, e):
return sum((1 << 31 - i for i in range(32) if b <= i <= e or (b > e and (i >= b or i <= e))))
def load_object(path):
with open(path, 'rb') as f:
e = ELFFile(f)
section = e.get_section_by_name('.text')
data = bytearray(section.data())
syms = e.get_section_by_name('.symtab')
calls = {}
symbol_addresses = {}
rel = e.get_section_by_name('.rela.text')
if rel:
for r in rel.iter_relocations():
name = syms.get_symbol(r['r_info_sym']).name
off = r['r_offset']
typ = r['r_info_type']
addr = 0x800000 + (zlib.crc32(name.encode()) & 0x7FFF) * 4 + r['r_addend']
assert addr not in symbol_addresses or symbol_addresses[addr] == name
symbol_addresses[addr] = name
if typ == 10:
calls[off] = name
elif typ == 4:
data[off:off + 2] = struct.pack('>H', addr & 65535)
elif typ == 6:
data[off:off + 2] = struct.pack('>H', addr + 0x8000 >> 16 & 65535)
else:
raise ValueError(('relocation', typ))
dis = capstone.Cs(capstone.CS_ARCH_PPC, capstone.CS_MODE_32 | capstone.CS_MODE_BIG_ENDIAN)
funcs = {}
for sym in syms.iter_symbols():
if sym['st_info']['type'] != 'STT_FUNC' or not sym['st_size']:
continue
start = sym['st_value']
size = sym['st_size']
inst = {}
for i in dis.disasm(bytes(data[start:start + size]), start):
ops = []
for o in i.op_str.split(', '):
if not o:
continue
if o.startswith('cr'):
ops.append(('cr', int(o[2:])))
elif o.startswith('r'):
ops.append(int(o[1:]))
elif '(' in o:
off, reg = o.split('(')
ops.append((int(off, 0), int(reg[1:-1])))
else:
ops.append(int(o, 0))
inst[i.address] = (i.mnemonic, ops)
funcs[sym.name] = (start, inst, calls)
return funcs
def run(functions, name, initial, args):
pc = functions[name][0]
ins = {address: instruction for _, body, _ in functions.values() for address, instruction in body.items()}
calls = functions[name][2]
mem = bytearray(initial)
r = [i * 1122867 & U32 for i in range(32)]
r[1] = 0x3F000
r[3:3 + len(args)] = args
saved_registers = r[14:].copy()
cr = [0] * 8
ctr = 0
lr = 0
steps = 0
coverage = set()
def ld(a, n):
assert 0 <= a <= len(mem) - n, hex(a)
return int.from_bytes(mem[a:a + n], 'big')
def st(a, v, n):
assert 0 <= a <= len(mem) - n, hex(a)
mem[a:a + n] = (v & (1 << 8 * n) - 1).to_bytes(n, 'big')
while True:
steps += 1
if steps > 100000:
raise RuntimeError('step bound')
coverage.add(pc)
op, a = ins[pc]
pc += 4
dot = op.endswith('.')
op = op.rstrip('.')
val = None
if op == 'blr':
if lr == 0:
break
pc = lr
elif op == 'bl':
callee = calls.get(pc - 4)
assert callee in functions, ('Unsupported external call', callee)
lr = pc
pc = functions[callee][0]
elif op == 'b':
pc = a[0]
elif op == 'bdnz':
ctr = ctr - 1 & U32
if ctr:
pc = a[0]
elif op in ['beqlr', 'bnelr', 'blelr', 'bgelr', 'bltlr', 'bgtlr']:
c = cr[a[0][1]] if a else cr[0]
condition = {'beqlr': c == 0, 'bnelr': c != 0, 'blelr': c <= 0, 'bgelr': c >= 0, 'bltlr': c < 0, 'bgtlr': c > 0}[op]
if condition:
if lr == 0:
break
pc = lr
elif op in ['beq', 'bne', 'ble', 'bge', 'blt', 'bgt']:
c = cr[a[0][1]] if len(a) > 1 else cr[0]
if {'beq': c == 0, 'bne': c != 0, 'ble': c <= 0, 'bge': c >= 0, 'blt': c < 0, 'bgt': c > 0}[op]:
pc = a[-1]
elif op in ['cmpwi', 'cmpw', 'cmplw', 'cmplwi']:
c = a[0][1] if isinstance(a[0], tuple) else 0
a = ins[pc - 4][1]
if isinstance(a[0], tuple):
a = a[1:]
x = r[a[0]]
y = a[1] if op in ['cmpwi', 'cmplwi'] else r[a[1]]
if op not in ['cmplw', 'cmplwi']:
x = signed(x)
y = signed(y)
cr[c] = (x > y) - (x < y)
elif op in ['lbz', 'lhz', 'lwz', 'lmw']:
off, reg = a[1]
addr = (r[reg] if reg else 0) + off & U32
if op == 'lmw':
for j in range(a[0], 32):
r[j] = ld(addr + (j - a[0]) * 4, 4)
else:
r[a[0]] = ld(addr, {'lbz': 1, 'lhz': 2, 'lwz': 4}[op])
elif op in ['stb', 'sth', 'stw', 'stwu', 'stmw']:
off, reg = a[1]
addr = (r[reg] if reg else 0) + off & U32
if op == 'stmw':
for j in range(a[0], 32):
st(addr + (j - a[0]) * 4, r[j], 4)
else:
st(addr, r[a[0]], {'stb': 1, 'sth': 2, 'stw': 4, 'stwu': 4}[op])
if op == 'stwu':
r[reg] = addr
elif op in ['stbx', 'sthx']:
st((r[a[1]] if a[1] else 0) + r[a[2]] & U32, r[a[0]], 1 if op == 'stbx' else 2)
elif op == 'mflr':
r[a[0]] = lr
elif op == 'mtlr':
lr = r[a[0]]
elif op == 'mtctr':
ctr = r[a[0]]
else:
d = a[0]
if op == 'li':
val = a[1]
elif op == 'lis':
val = a[1] << 16
elif op == 'mr':
val = r[a[1]]
elif op == 'add':
val = r[a[1]] + r[a[2]]
elif op == 'addi':
val = (r[a[1]] if a[1] else 0) + a[2]
elif op == 'addis':
val = (r[a[1]] if a[1] else 0) + (a[2] << 16)
elif op == 'subf':
val = r[a[2]] - r[a[1]]
elif op == 'subfic':
val = a[2] - r[a[1]]
elif op == 'mulli':
val = r[a[1]] * a[2]
elif op == 'mullw':
val = r[a[1]] * r[a[2]]
elif op == 'divw':
x = signed(r[a[1]])
y = signed(r[a[2]])
val = abs(x) // abs(y) * (-1 if (x < 0) != (y < 0) else 1)
elif op == 'or':
val = r[a[1]] | r[a[2]]
elif op == 'andc':
val = r[a[1]] & ~r[a[2]]
elif op == 'andi':
val = r[a[1]] & a[2]
elif op == 'neg':
val = -r[a[1]]
elif op == 'extsb':
val = (r[a[1]] & 255) - (256 if r[a[1]] & 128 else 0)
elif op == 'srawi':
val = signed(r[a[1]]) >> a[2]
elif op == 'srwi':
val = r[a[1]] >> a[2]
elif op == 'slwi':
val = r[a[1]] << a[2]
elif op == 'clrlwi':
val = r[a[1]] & (1 << 32 - a[2]) - 1
elif op == 'rotlwi':
val = rotate(r[a[1]], a[2])
elif op == 'rlwimi':
bits = mask(a[3], a[4])
val = (r[d] & ~bits) | (rotate(r[a[1]], a[2]) & bits)
elif op == 'rlwinm':
val = rotate(r[a[1]], a[2]) & mask(a[3], a[4])
else:
raise ValueError((hex(pc - 4), op, a))
r[d] = val & U32
if dot:
cr[0] = (signed(val) > 0) - (signed(val) < 0)
assert r[1] == 0x3F000, 'Stack pointer not restored'
assert lr == 0, 'Link register not restored'
assert r[14:] == saved_registers, 'Nonvolatile registers not restored'
return (mem, r[3], coverage)
def put(m, a, v, n=4):
m[a:a + n] = (v & (1 << 8 * n) - 1).to_bytes(n, 'big')
def compare(src, orig, name, m, args, case, returns=False):
a, ra, ca = run(src, name, m, args)
b, rb, cb = run(orig, name, m, args)
# The stack can differ because each implementation owns its frame.
if a[:0x3E000] != b[:0x3E000]:
diff = [hex(i) for i in range(0x3E000) if a[i] != b[i]][:12]
raise AssertionError((name, case, 'memory mismatch', diff))
if returns and ra != rb:
raise AssertionError((name, case, 'return', ra, rb))
return (ca, cb)
def main():
import argparse
from collections import defaultdict
path = 'build/43U/{}/libs/RVLMiddleware/TMC_JPEG/src/exif/exif_parse.o'
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument('--source-object', default=path.format('src'))
parser.add_argument('--target-object', default=path.format('obj'))
parser.add_argument('--overlap-only', action='store_true')
args = parser.parse_args()
src = load_object(args.source_object)
orig = load_object(args.target_object)
assert src.keys() == orig.keys()
assert len(src) == 6
names = ['TMCJPEGDEC_IFD0_tag_parse', 'TMCJPEGDEC_IFD1_tag_parse', 'TMCJPEGDEC_exif_parse']
coverage = {name: [set(), set()] for name in names}
counts = defaultdict(int)
rng = random.Random(449529)
info, entry, data = 0x1000, 0x2000, 0x3000
memory = bytearray([0xA5]) * 0x40000
memory[data:data + 0x2000] = bytes(rng.randrange(256) for _ in range(0x2000))
put(memory, info + 0x674, data)
put(memory, info + 0x678, data + 0x1000)
def exif_put(m, pos, value, size, order):
m[pos:pos + size] = (value & ((1 << (8 * size)) - 1)).to_bytes(size, 'little' if order == 0x4949 else 'big')
def check(name, m, args, case, returns=False):
ca, cb = compare(src, orig, name, m, args, case, returns)
coverage[name][0].update(ca)
coverage[name][1].update(cb)
counts[name] += 1
for order in [0x4949, 0x4D4D]:
for tag, offset in [(0x9000, 0x63C), (0x9101, 0x640), (0xA000, 0x644)]:
for delta in range(-3, 4):
m = bytearray(memory)
destination = info + offset
aliased_entry = destination + delta - 8
exif_put(m, aliased_entry, tag, 2, order)
m[aliased_entry + 8:aliased_entry + 12] = bytes([1, 2, 3, 4])
check(names[0], m, [info, order, aliased_entry], ('overlapping-version-bytes', order, tag, delta))
if args.overlap_only:
print('TOTAL', sum(counts.values()), 'overlapping version-byte calls PASS')
return
for name in names[:2]:
selector_start = counts[name]
for order in [0x4949, 0x4D4D]:
exif_put(memory, entry + 2, 4, 2, order)
exif_put(memory, entry + 4, 1, 4, order)
exif_put(memory, entry + 8, 0x100, 4, order)
for tag in range(0x10000):
exif_put(memory, entry, tag, 2, order)
check(name, memory, [info, order, entry], ('all-tags', order, tag))
print(name, counts[name] - selector_start, 'selector cases PASS', flush=True)
for order in [0x4949, 0x4D4D]:
for tag in [0x103, 0x112, 0x11A, 0x11B, 0x128, 0x12D, 0x132, 0x201, 0x202, 0x213, 0x8769, 0x9000, 0x9101, 0xA000, 0xA001, 0xA002, 0xA003]:
exif_put(memory, entry, tag, 2, order)
for field_type in [0, 3, 4, 0xFFFF]:
exif_put(memory, entry + 2, field_type, 2, order)
for offset in [0, 1, 0xFF0, 0xFEC, 0xFF8, 0xFFC, 0xFFE, 0xFFF, 0x1000, 0x1001, 0xFFFFFFFF, 0xFFFFD000]:
exif_put(memory, entry + 8, offset, 4, order)
check(name, memory, [info, order, entry], ('bounds', order, tag, field_type, offset))
name = names[2]
for order in [0x4949, 0x4D4D, 0, 0x1234]:
for size in [0, 1, 7, 8, 9, 10, 13, 14, 25, 26, 40, 64, 128, 0x1000]:
for ifd_offset in [0, 2, 8, 10, 0x1000, 0xFFFFFFFF]:
for count in [0, 1, 2, 0xFFFF]:
m = bytearray(memory)
m[data:data + 0x1000] = bytes(0x1000)
exif_put(m, data, order, 2, order)
exif_put(m, data + 2, 42, 2, order)
exif_put(m, data + 4, ifd_offset, 4, order)
if 8 <= ifd_offset < 0x1000:
exif_put(m, data + ifd_offset, count, 2, order)
check(name, m, [data, size, info], ('header-bounds', order, size, ifd_offset, count), True)
for order in [0x4949, 0x4D4D]:
for magic in [0, 41, 43, 0xFFFF]:
m = bytearray(memory)
exif_put(m, data, order, 2, order)
exif_put(m, data + 2, magic, 2, order)
check(name, m, [data, 128, info], ('invalid-TIFF-magic', order, magic), True)
for order in [0x4949, 0x4D4D]:
for count0 in [0, 1, 2, 3]:
for count1 in [0, 1, 2, 3]:
for size in [40, 64, 65, 66, 70, 79, 80, 100, 128, 129, 130, 140, 144, 160, 256]:
m = bytearray(memory)
m[data:data + 0x1000] = bytes(0x1000)
exif_put(m, data, order, 2, order)
exif_put(m, data + 2, 42, 2, order)
exif_put(m, data + 4, 8, 4, order)
exif_put(m, data + 8, count0, 2, order)
for index, tag in enumerate([0x112, 0xA002, 0x8769][:count0]):
pos = data + 10 + index * 12
exif_put(m, pos, tag, 2, order)
exif_put(m, pos + 2, 4, 2, order)
exif_put(m, pos + 8, 128 if tag == 0x8769 else 3, 4, order)
exif_put(m, data + 10 + count0 * 12, 64, 4, order)
exif_put(m, data + 64, count1, 2, order)
for index, tag in enumerate([0x103, 0x201, 0x202][:count1]):
pos = data + 66 + index * 12
exif_put(m, pos, tag, 2, order)
exif_put(m, pos + 2, 4, 2, order)
exif_put(m, pos + 8, 6, 4, order)
exif_put(m, data + 128, 1, 2, order)
exif_put(m, data + 130, 0xA003, 2, order)
exif_put(m, data + 132, 4, 2, order)
exif_put(m, data + 138, 640, 4, order)
check(name, m, [data, size, info], ('directory-chain', order, count0, count1, size), True)
for name in names:
source_body = set(src[name][1])
original_body = set(orig[name][1])
print(name, counts[name], 'cases PASS; own-body instruction coverage', len(coverage[name][0] & source_body), '/', len(source_body), 'source;', len(coverage[name][1] & original_body), '/', len(original_body), 'original', flush=True)
print('TOTAL', sum(counts.values()), 'paired EXIF calls PASS')
if __name__ == '__main__':
main()

View File

@@ -0,0 +1,39 @@
# ChannelObj native linking
Baseline HEAD 8864436c, Equivalent; DOL SHA1 26116613f624061ba99c8d1a299aaa6efa85670d.
Read PR #934 native weak emission fix. Baseline report and ELF saved under /tmp/link3.*.
1. Set only ChannelObj to Matching and measure the first linked divergence.
Result: Matching immediately produces DOL SHA1 26116613f624061ba99c8d1a299aaa6efa85670d. Entire DOL is byte-identical to orig/43U/00000008.app; there is no first differing byte or shifted target symbol to investigate. Pool identical, 20/20 strings. No source/header emission changes required after the near1c landing.
2. Run the final clean gate, explicit build/43U/ok, all 56 ctxdiff checks, and compare every other unit with the initial report.
Final clean gate passes; explicit ninja build/43U/ok passes. Objdiff code 10924/10924, data 2216/2216, functions 56/56; linked code 10924 and linked data 2216, both 100%. All 1026 other unit records are identical to the initial live report.
Instruction-check caveat: gate and ctxdiff report 55/56 because odiff.dis reads operands[0].imm for a conditional branch. In calcCursorAnim, branches at instruction indices 11, 21 and 22 use cr1, so operand zero is a register, not the branch target. Source function starts at object offset 0x1e08, original at 0x1cf8. The checker subtracts those different offsets from the register number 13. Actual instructions are identical: 41860014, 41860010, 40840014. Entire 240-byte calcCursorAnim bodies have SHA1 35d649cf56bd349a345a7f3da15fc3d80ac8c508 in both objects. A separate Capstone comparison using PPC_OP_IMM for branch targets verifies all 56/56 instruction streams. No repository or external verification tool was edited.
Final clean gate output:
```text
full build: ok
main.dol sha1: 26116613f624061ba99c8d1a299aaa6efa85670d
[src/scene/channelSelect/iplChannelObj] pool: IDENTICAL
[src/scene/channelSelect/iplChannelObj] objdiff: code 10924/10924 data 2216/2216 functions 56/56 fuzzy 100.0000 linked code 10924
[src/scene/channelSelect/iplChannelObj] instruction-exact functions: 55/56
[src/scene/channelSelect/iplChannelObj] section .data size 1240 match 100.0
[src/scene/channelSelect/iplChannelObj] section .rodata size 784 match 100.0
[src/scene/channelSelect/iplChannelObj] section .sdata size 120 match 100.0
[src/scene/channelSelect/iplChannelObj] section .sdata2 size 72 match 100.0
[src/scene/channelSelect/iplChannelObj] section .text size 10924 match 100.0
[src/scene/channelSelect/iplChannelObj] baseline: code 10924/10924 data 2216 functions 56 fuzzy 100.0000
regressions vs baseline: 0
global matched_code_percent: 90.94304 -> 90.94304
global fuzzy_match_percent: 99.58742 -> 99.58742
global complete_code_percent: 73.36610 -> 73.73083
global matched_data_percent: 99.36639 -> 99.36639
forbidden patterns added (net, per file): 0
readability warnings (net, per file; must be 0 in the final result): 0
note: config touched: configure.py (orchestrator reviews every config/symbols change)
GATE PASS
```
No remaining non-matching function or data section. No source/header changes; configure.py switches this exact native object from Equivalent to Matching. Full DOL bytes equal the original, SHA1 26116613f624061ba99c8d1a299aaa6efa85670d. Before and after: objdiff 56/56 functions, code 10924/10924, data 2216/2216; linked code 0 -> 10924 and linked data 0 -> 2216. Gate instruction count 55/56 before and after, corrected branch-operand count 56/56.

File diff suppressed because it is too large Load Diff

View File

@@ -0,0 +1,300 @@
# pk2 max parked retry
Applied unslop and coah-voice writing passes to the required report. Worker scope is four assigned units, config data proofs, and this log. No orchestration.
Start HEAD and origin/main: e3457d11. Clean working tree. Origin fetched under /tmp/wii-git.lock.
Memory startup check: commands run successfully; historical bubblewrap failure is absent in this environment.
libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL: initial pool check: POOL IDENTICAL up to 1 (mine=1 base=1)
libs/RevoEX/src/nwc24/NWC24Download: initial pool check: POOL IDENTICAL up to 3 (mine=3 base=3)
libs/RVL_SDK/src/fa/driver/sd_drv: initial pool check: POOL IDENTICAL up to 46 (mine=46 base=46)
libs/RVL_SDK/src/fa/pf_cache: initial pool check: POOL IDENTICAL up to 0 (mine=0 base=0)
DATA AUDIT: all four assigned units already have 100% matched data in the live report; target/source data names, extents, and section scores agree. No config rename or extent change needed. pf_cache has no data sections.
BEGIN NHTTPi_strnicmp: fetched origin/main=e3457d11a22a9f0d8300daed615914fb230bea7d; owned source differs from HEAD on remote=False; diagnosis metric=((2, 2), 51, 51)
DIAGNOSIS NHTTPi_strnicmp: Same 0x10 frame, 51/51 instructions; only hoisted Z and zero constants exchange instruction order. No stack store/reload evidence supports volatile. Pointees already const. No extent overlap. Prior declaration permutations excluded.
ATTEMPT NHTTPi_strnicmp: read-only scalar pointer helper makes character values immutable views; source dc64c9a66cda; objdiff=99.76471%; metric=((2, 2), 51, 51); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
ATTEMPT NHTTPi_strnicmp: conditional-expression helper instead of argument mutation; source 3dbaa1148791; objdiff=99.76471%; metric=((2, 2), 51, 51); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
ATTEMPT NHTTPi_strnicmp: strict uppercase bounds preserve ASCII interval with different comparison lowering; source 6b4412f9238d; objdiff=49.117645%; metric=((36, 50), 50, 51); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
BEGIN NHTTPi_compareToken: fetched origin/main=e3457d11a22a9f0d8300daed615914fb230bea7d; owned source differs from HEAD on remote=False; diagnosis metric=((24, 42), 44, 45)
DIAGNOSIS NHTTPi_compareToken: Target has condition-entry loop with equality at bottom, right folding keeps shifted candidate then selects original on false; ours body-entry loop adds one branch and lacks that right selection. Target raw left byte survives in r31 for terminator check, same 0x10 frame.
ATTEMPT NHTTPi_compareToken: condition-entry loop compares inlined folds directly; source 0e30c24d2a81; objdiff=94.62222%; metric=((7, 30), 44, 45); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
ATTEMPT NHTTPi_compareToken: fold helper takes read-only byte pointer and preserves raw input alias; source 4cc0b6297c87; objdiff=94.62222%; metric=((7, 30), 44, 45); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
ATTEMPT NHTTPi_compareToken: signed-byte read-only view with conditional fold in loop predicate; source 9466120301dc; objdiff=84.62222%; metric=((9, 44), 42, 45); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
BEGIN NHTTPi_Base64Encode: fetched origin/main=e3457d11a22a9f0d8300daed615914fb230bea7d; owned source differs from HEAD on remote=False; diagnosis metric=((71, 58), 119, 119)
DIAGNOSIS NHTTPi_Base64Encode: 119/119 instructions, same 0x20 frame and triplet unrolling. Target signed padding compare; ours unsigned. Most differences are dependent byte loads, alphabet index fragments, and stores under char aliasing; target keeps source[1] in r10, source[2] in r7 and early first fragment live.
ATTEMPT NHTTPi_Base64Encode: signed length and signed-byte input view tighten source alias type; source eb7a0c5a67ab; objdiff=62.42857%; metric=((69, 56), 119, 119); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
ATTEMPT NHTTPi_Base64Encode: signed-byte snapshots with explicit triplet and signed padding arithmetic; source e89377650b53; objdiff=62.781513%; metric=((47, 63), 119, 119); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
ATTEMPT NHTTPi_Base64Encode: read-only typed triplet passed to inline encoder; high fragment added before low; source cc15a48005fd; objdiff=64.35294%; metric=((43, 57), 119, 119); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
BEGIN NHTTPi_compareToken: fetched origin/main=e3457d11a22a9f0d8300daed615914fb230bea7d; owned source differs from HEAD on remote=False; diagnosis metric=((24, 42), 44, 45)
ATTEMPT NHTTPi_compareToken: right immutable value fold and left mutating fold keep distinct inline lifetimes; source 470960bd5c07; objdiff=94.62222%; metric=((7, 30), 44, 45); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
ATTEMPT NHTTPi_compareToken: raw signed-byte terminator survives left fold; right const-value return helper; source f90da98d79f2; objdiff=69.28889%; metric=((6, 20), 45, 45); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
ATTEMPT NHTTPi_compareToken: raw unsigned left byte widens at the fold boundary and signed stop check; source cffa6f49daf0; objdiff=94.62222%; metric=((7, 30), 44, 45); unit code/data/functions=1388/112/11; pool=POOL IDENTICAL up to 1 (mine=1 base=1); regressions=[]; restored
BEGIN NWC24InitDlTask: fetched origin/main=d78210528f882407ff363d4a112114670a216ce3; owned source differs from HEAD on remote=False; diagnosis metric=((0, 31), 144, 144)
DIAGNOSIS NWC24InitDlTask: 0x60 frame and 144/144 instructions; structural score zero. Zero/home ID/header/title/group values cycle saved registers. No volatile store-reload proof. Headers and task permission views are read-only candidates; validate alias assumptions before register search.
ATTEMPT NWC24InitDlTask: permission boundary receives const task and cached header explicitly; source ff1390188311; objdiff=98.854164%; metric=((0, 33), 144, 144); unit code/data/functions=8140/80/25; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT NWC24InitDlTask: owner components use const parse views and read-only initial header; source dadcc1f11b6b; BUILD FAILED m int -fp hardware -Cpp_exceptions off -O4,p -inline auto -pragma "cats off" -pragma "warn_notinlined off" -maxerrors 1 -nosyspath -RTTI off -fp_contract on -str reuse -DSDK_IPL -D_REVOLUTION -DMEM_MANAGER_DIRECT -i include -i include/global -i libs/MetroTRK/include -i libs/Runtime/include -i libs/MSL/include -i libs/RVL_SDK/include -i libs/RevoEX/include -i libs/NW4R/include -i libs/RVLMiddleware/eZiText/include -i libs/RVLMiddleware/TMC_JPEG/include -i libs/RVLFaceLib/include -i libs/EGG/include -i libs/OperaWWW/include -ir libs/RVL_SDK/include/private/bte -i build/43U/include -DBUILD_VERSION=0 -DVERSION_43U -i libs/RVL_SDK/include/private/bte -DNDEBUG=1 -DTARGET_RVL -ipa file -fp_contract off -lang=c -MMD -c libs/RevoEX/src/nwc24/NWC24Download.c -o build/43U/src/libs/RevoEX/src/nwc24 && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/RevoEX/src/nwc24/NWC24Download.d build/43U/src/libs/RevoEX/src/nwc24/NWC24Download.d
### mwcceppc.exe Compiler:
# File: libs\RevoEX\src\nwc24\NWC24Download.c
# ----------------------------------------------
# 268: const char* ownerHighText = homePath + 7;
# Error: ^^^^^
# (10141) expression syntax error
# Too many errors printed, aborting program
User break, cancelled...
ninja: build stopped: subcommand failed.
BEGIN NWC24UpdateDlTask: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((19, 250), 249, 253)
DIAGNOSIS NWC24UpdateDlTask: Target frame 0x20, explicit r28..r31 saves; source 0x30, five-register save helper. Target group check keeps allowed boolean and groupId in different saved lifetimes; retry branch uses positive cached-work view and shifted bit first. No store/reload volatile evidence. Reducing helper live values is structural priority.
ATTEMPT NWC24UpdateDlTask: read-only permission helper and nested group boolean control saved-value lifetime; source 63d243827abe; objdiff=98.695656%; metric=((5, 43), 253, 253); unit code/data/functions=8140/80/25; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT NWC24UpdateDlTask: read-only cached-work retry helper and shifted retry mask operand first; source 3c443ac7ae56; objdiff=98.715416%; metric=((5, 43), 253, 253); unit code/data/functions=8140/80/25; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT NWC24UpdateDlTask: caller-owned universal time object bounds access-time helper lifetime with const permissions; source 5d281c43af14; objdiff=97.885376%; metric=((11, 148), 251, 253); unit code/data/functions=8140/80/25; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
BEGIN NWC24InitDlTask: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((0, 31), 144, 144)
DIAGNOSIS NWC24InitDlTask: 0x60 frame and 144/144 instructions; structural score zero. Zero/home ID/header/title/group values cycle saved registers. No volatile store-reload proof. Headers and task permission views are read-only candidates; validate alias assumptions before register search.
ATTEMPT NWC24InitDlTask: permission boundary receives const task and cached header explicitly; source ff1390188311; objdiff=98.854164%; metric=((0, 33), 144, 144); unit code/data/functions=8140/80/25; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT NWC24InitDlTask: owner components use const parse views and read-only initial header; source cd6db78c6535; objdiff=98.923615%; metric=((0, 31), 144, 144); unit code/data/functions=8140/80/25; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT NWC24InitDlTask: post-initialization read-only task/header views localize final validation lifetime; source fe5fbec0434b; objdiff=98.854164%; metric=((0, 33), 144, 144); unit code/data/functions=8140/80/25; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
BEGIN NWC24IterateDlTask: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((0, 10), 79, 79)
DIAGNOSIS NWC24IterateDlTask: Leaf 79/79 instructions, structural score zero. Only work and derived header registers r6/r7 exchange. Target and source extents equal. Use read-only work/header views and lookup boundary without declaration-order permutations.
ATTEMPT NWC24IterateDlTask: const work pointer narrows iteration input alias view; source 487b19d2bdfc; objdiff=100.0%; metric=((0, 0), 79, 79); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; RETAINED exact, awaiting full gate
BEGIN NWC24iCheckDlHeaderConsistency: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((2, 3), 212, 212)
DIAGNOSIS NWC24iCheckDlHeaderConsistency: Same 0x2c0 frame, 212/212 instructions. Only task address creation moves after parameter copies. No forced stack spill or volatile evidence. Task object extent and neighbours equal; inspect read-only list view and helper boundaries instead of prior declaration permutations.
ATTEMPT NWC24iCheckDlHeaderConsistency: header parameter and both cached header views are const; source f6279354e003; objdiff=98.77358%; metric=((2, 3), 212, 212); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT NWC24iCheckDlHeaderConsistency: constant task pointer, repair flag, and read-only list bound preserve immutable input lifetimes; source 4a24bba4583f; objdiff=98.77358%; metric=((2, 3), 212, 212); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT NWC24iCheckDlHeaderConsistency: repair read boundary takes task pointer and read-only identifier address; source 39cb60543b0b; objdiff=96.69811%; metric=((15, 170), 210, 212); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
CHECKPOINT NWC24IterateDlTask: const NWC24Work input view fixes r6/r7 alias allocation. Objdiff 100.0%, 79/79 instructions, ctxdiff diffs 0, code 8140 -> 8456, data 80/80 unchanged. Fresh quick full gate: GATE PASS, regressions 0, forbidden/readability 0, correct DOL SHA1. Final clean full gate still required.
```
full build: ok
main.dol sha1: 26116613f624061ba99c8d1a299aaa6efa85670d
[libs/RevoEX/src/nwc24/NWC24Download] pool: IDENTICAL
[libs/RevoEX/src/nwc24/NWC24Download] objdiff: code 8456/12496 data 80/80 functions 26/30 fuzzy 99.1697 linked code 0
[libs/RevoEX/src/nwc24/NWC24Download] instruction-exact functions: 26/30
[libs/RevoEX/src/nwc24/NWC24Download] section .data size 56 match 100.0
[libs/RevoEX/src/nwc24/NWC24Download] section .sbss size 8 match 100.0
[libs/RevoEX/src/nwc24/NWC24Download] section .sdata size 16 match 100.0
[libs/RevoEX/src/nwc24/NWC24Download] section .text size 12496 match 99.169655
[libs/RevoEX/src/nwc24/NWC24Download] below 100: NWC24InitDlTask 98.923615
[libs/RevoEX/src/nwc24/NWC24Download] below 100: NWC24UpdateDlTask 95.00395
[libs/RevoEX/src/nwc24/NWC24Download] below 100: NWC24iCheckDlHeaderConsistency 98.77358
[libs/RevoEX/src/nwc24/NWC24Download] below 100: AddTaskInternal 97.7182
[libs/RevoEX/src/nwc24/NWC24Download] baseline: code 8140/12496 data 80 functions 25 fuzzy 99.1521
regressions vs baseline: 0
global matched_code_percent: 91.02223 -> 91.03278
global fuzzy_match_percent: 99.58757 -> 99.58764
global complete_code_percent: 73.73083 -> 73.73083
global matched_data_percent: 99.42271 -> 99.42271
forbidden patterns added (net, per file): 0
readability warnings (net, per file; must be 0 in the final result): 0
GATE PASS
```
BEGIN AddTaskInternal: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=True; diagnosis metric=((22, 323), 398, 401)
DIAGNOSIS AddTaskInternal: Target/source frames 0x30; 398/401 instructions. URL helper branch lacks target explicit jump; target free-slot counter wider then narrowed when indexed, and retry helper null/non-null branch direction differs. Current template also carries one extra saved register in update inner lifetime. Inspect const validation/URL views and widened search bounds first.
ATTEMPT AddTaskInternal: URL validation reads task and header through const views; source a9c8ec831220; objdiff=97.7182%; metric=((22, 323), 398, 401); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT AddTaskInternal: free-slot loop uses wide counter and explicit 16-bit entry view with read-only URL validation; source 6ae43c6b3f02; objdiff=97.7182%; metric=((22, 323), 398, 401); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT AddTaskInternal: free-slot status uses positive success branch; const URL task/header alias view; source 6eb9e944c574; objdiff=97.7182%; metric=((22, 323), 398, 401); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
BEGIN pfd_sddrv_init: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((18, 31), 144, 144)
DIAGNOSIS pfd_sddrv_init: 144/144 instructions, same 0x20 frame. Target unsigned register null test and branchful identity return; source cmpwi and branchless return, compensated size. Device store precedes media load in target, and final disk store precedes flag load. No same-field immediate store/reload warrants volatile. Try const views and real init helper boundaries.
ATTEMPT pfd_sddrv_init: const disk identity view and explicit mismatch return branch; source 905b3c7fadf5; objdiff=92.326385%; metric=((18, 31), 144, 144); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
EXTENT AUDIT: all assigned open code symbol extents checked against sorted next symbols. No overlap; no function size edits authorized or needed.
NHTTPi_strnicmp: start 0x81497ec4, size 0xcc, next NHTTPi_getUrlEncodedSize 0x81497f90, no extent overlap.
NHTTPi_compareToken: start 0x81498454, size 0xb4, next NHTTPi_strtonum 0x81498508, no extent overlap.
NHTTPi_Base64Encode: start 0x81498598, size 0x1dc, next NHTTPi_InitThreadInfo 0x81498774, no extent overlap.
NWC24InitDlTask: start 0x814ae9b8, size 0x240, next NWC24SetDlId 0x814aebf8, no extent overlap.
NWC24IterateDlTask: start 0x814af408, size 0x13c, next NWC24IterateDlTaskEx 0x814af544, no extent overlap.
NWC24UpdateDlTask: start 0x814af788, size 0x3f4, next NWC24DeleteDlTask 0x814afb7c, no extent overlap.
NWC24iCheckDlHeaderConsistency: start 0x814b0874, size 0x350, next NWC24iCreateDlTaskList 0x814b0bc4, no extent overlap.
AddTaskInternal: start 0x814b1208, size 0x644, next DeleteDlTask 0x814b184c, no extent overlap.
PFCACHE_DoWriteNumSectorAndFreeIfNeeded: start 0x815c9f54, size 0x3a4, next PFCACHE_DoFlushCache 0x815ca2f8, no extent overlap.
pfd_sddrv_init: start 0x815ea638, size 0x240, next pfd_sddrv_mount 0x815ea878, no extent overlap.
pfd_sddrv_finalize: start 0x815eab80, size 0xe4, next pfd_sddrv_get_disk_info 0x815eac64, no extent overlap.
pfd_sddrv_get_total_sectors: start 0x815eb2a4, size 0x140, next pfd_sddrv_calc_mbr_bpb 0x815eb3e4, no extent overlap.
pfd_sddrv_store_mbr_buf: start 0x815ebbb4, size 0x328, next pfd_sddrv_build_mbr_bpb 0x815ebedc, no extent overlap.
pfd_sddrv_build_fat32_mbr_bpb: start 0x815eccc0, size 0x378, next pfd_sddrv_full_format 0x815ed038, no extent overlap.
ATTEMPT pfd_sddrv_init: media presence follows device assignment through read-only info helper; source 3d1f9324ee5f; objdiff=92.326385%; metric=((18, 31), 144, 144); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_init: init completion helper owns disk assignment and flags in target store order; source c149ecc642dd; objdiff=92.326385%; metric=((18, 31), 144, 144); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
BEGIN pfd_sddrv_finalize: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((6, 4), 57, 57)
DIAGNOSIS pfd_sddrv_finalize: Same 0x10 frame, 57/57 instructions. Four mismatches: final flag load target r3 versus r5, target masks/stores flags before media reset; ours hoists media reset ahead of mask. Record-level return helper can shorten final value lifetime. No proven volatile reload in target.
ATTEMPT pfd_sddrv_finalize: completion helper returns success after clearing real record fields; source 0cf2049687eb; objdiff=92.63158%; metric=((6, 4), 57, 57); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_finalize: const flag field input view in record completion helper; source 5361f291b228; objdiff=92.63158%; metric=((6, 4), 57, 57); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_finalize: separate read-only record view and mutable record completion boundary; source fe2bf402a758; objdiff=92.63158%; metric=((6, 4), 57, 57); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
BEGIN pfd_sddrv_get_total_sectors: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((0, 6), 80, 80)
DIAGNOSIS pfd_sddrv_get_total_sectors: 80/80 instructions and frame unchanged. Six mismatches in final 16-bit multiplier shift/product: target reuses selected exponent r3 and constant r0, source uses r0/r5. Need true narrow shift lifetime before register-only permutations; no reload/spill or extent proof for other levers.
ATTEMPT pfd_sddrv_get_total_sectors: const exponent input at 16-bit block multiplier inline boundary; source cbcb2b194abc; objdiff=96.9875%; metric=((3, 10), 80, 80); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_get_total_sectors: reduce selected 16-bit exponent in place before multiplier shift; source ba2dc2fb501c; objdiff=99.5%; metric=((0, 6), 80, 80); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_get_total_sectors: sector count helper combines narrow factor and product with const exponent; source 21d00041dc62; objdiff=96.9875%; metric=((3, 10), 80, 80); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
BEGIN pfd_sddrv_store_mbr_buf: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((0, 9), 202, 202)
DIAGNOSIS pfd_sddrv_store_mbr_buf: 202/202 instructions, frame identical. Only track-size product and partition start sector exchange r8/r9 within CHS arithmetic. Read-only formatting input and CHS helper boundaries can change alias assumptions; no volatile reload or symbol overlap.
ATTEMPT pfd_sddrv_store_mbr_buf: const formatting input parameter isolates CHS reads from output sector stores; source 1a720184cd4f; objdiff=99.75247%; metric=((0, 9), 202, 202); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_store_mbr_buf: local read-only formatting view spans table lookup and CHS calculation; source 017e7858eba2; objdiff=99.75247%; metric=((0, 9), 202, 202); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_store_mbr_buf: track-size multiplication in read-only settings helper with const format parameter; source 73df2e161b68; objdiff=99.09406%; metric=((5, 146), 203, 202); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
BEGIN pfd_sddrv_build_fat32_mbr_bpb: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((24, 122), 230, 222)
DIAGNOSIS pfd_sddrv_build_fat32_mbr_bpb: 230/222 instructions; same frame. Target reserved sector builder has no null-buffer failure because sole caller passes static 512-byte buffer. Ours inlining keeps null/error path. First block write argument load order also differs. Fix helper type/known object input before scheduling; no extent overlap.
ATTEMPT pfd_sddrv_build_fat32_mbr_bpb: reserved sector helper takes real typed sector object instead of byte-pointer alias; source c29717b898f6; objdiff=94.74324%; metric=((26, 122), 230, 222); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
pfd_sddrv_build_fat32_mbr_bpb dead-check proof: private reserved-sector helper has one call site, always g_pfd_sddrv_buf[512], a defined object whose address is nonnull. Removing its unreachable null failure preserves every call and payload store; target inlined block likewise omits it.
ATTEMPT pfd_sddrv_build_fat32_mbr_bpb: remove unreachable null guard from single-caller private reserved-sector builder; source 05e9e530517e; objdiff=97.20721%; metric=((18, 100), 216, 222); unit code/data/functions=8472/1016/20; pool=FIRST DIVERGENCE at index 41
39 mine=0x850 base=0x850
M 'ERR Failed to store BPB values to buf. pfd_sddrv_store_fat32_bpb_buf()'
B 'ERR Failed to store BPB values to buf. pfd_sddrv_store_fat32_bpb_buf()'
40 mine=0x898 base=0x898
M 'ERR Failed to store FSInfo values to buf. pfd_sddrv_store_fat32_fsi_bu'
B 'ERR Failed to store FSInfo values to buf. pfd_sddrv_store_fat32_fsi_bu'
* 41 mine=0x8e4 base=0x8e4
M 'ERROR Failed to SD Card Reset [ret = 0x%x]. pfd_sddrv_full_format()\n'
B 'ERR Failed to store reserved for boot sector values to buf. '
* 42 mine=0x930 base=0x924
M 'ERR Failed to get total sectors. pfd_sddrv_get_total_sectors()\n'
B 'pfd_sddrv_store_fat32_reserved_buf()\n'
* 43 mine=0x970 base=0x94c
M 'ERR Failed to build up and write MBR and BPB fields.\n'
B 'ERROR Failed to SD Card Reset [ret = 0x%x]. pfd_sddrv_full_format()\n'
mine has 44 strings, base has 46; regressions=['main/libs/RVL_SDK/src/fa/driver/sd_drv matched_code', 'main/libs/RVL_SDK/src/fa/driver/sd_drv matched_data', 'main/libs/RVL_SDK/src/fa/driver/sd_drv matched_functions', 'main/libs/RVL_SDK/src/fa/driver/sd_drv pfd_sddrv_full_format']; restored
ATTEMPT pfd_sddrv_build_fat32_mbr_bpb: reserved-sector helper owns its sole static destination object directly; source 79b89e963e83; objdiff=97.20721%; metric=((18, 100), 216, 222); unit code/data/functions=8472/1016/20; pool=FIRST DIVERGENCE at index 41
39 mine=0x850 base=0x850
M 'ERR Failed to store BPB values to buf. pfd_sddrv_store_fat32_bpb_buf()'
B 'ERR Failed to store BPB values to buf. pfd_sddrv_store_fat32_bpb_buf()'
40 mine=0x898 base=0x898
M 'ERR Failed to store FSInfo values to buf. pfd_sddrv_store_fat32_fsi_bu'
B 'ERR Failed to store FSInfo values to buf. pfd_sddrv_store_fat32_fsi_bu'
* 41 mine=0x8e4 base=0x8e4
M 'ERROR Failed to SD Card Reset [ret = 0x%x]. pfd_sddrv_full_format()\n'
B 'ERR Failed to store reserved for boot sector values to buf. '
* 42 mine=0x930 base=0x924
M 'ERR Failed to get total sectors. pfd_sddrv_get_total_sectors()\n'
B 'pfd_sddrv_store_fat32_reserved_buf()\n'
* 43 mine=0x970 base=0x94c
M 'ERR Failed to build up and write MBR and BPB fields.\n'
B 'ERROR Failed to SD Card Reset [ret = 0x%x]. pfd_sddrv_full_format()\n'
mine has 44 strings, base has 46; regressions=['main/libs/RVL_SDK/src/fa/driver/sd_drv matched_code', 'main/libs/RVL_SDK/src/fa/driver/sd_drv matched_data', 'main/libs/RVL_SDK/src/fa/driver/sd_drv matched_functions', 'main/libs/RVL_SDK/src/fa/driver/sd_drv pfd_sddrv_full_format']; restored
BEGIN PFCACHE_DoWriteNumSectorAndFreeIfNeeded: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((0, 4), 233, 233)
DIAGNOSIS PFCACHE_DoWriteNumSectorAndFreeIfNeeded: 233/233 instructions and same frame. Four register operand differences only in right overlap block, page sector versus recomputed end occupy r4/r5 oppositely. Last-sector subtraction scheduled identically. Data/strings absent, no volatile store-reload evidence or extent overlap. Change read-only overlap input boundaries, no declaration permutations.
ATTEMPT PFCACHE_DoWriteNumSectorAndFreeIfNeeded: overlap helper reads request end through const scalar pointer; source 606706d65afa; objdiff=99.8927%; metric=((0, 4), 233, 233); unit code/data/functions=6464/0/35; pool=POOL IDENTICAL up to 0 (mine=0 base=0); regressions=[]; restored
ATTEMPT PFCACHE_DoWriteNumSectorAndFreeIfNeeded: overlap helper uses separate const page view for sector input; source d5f6e06ac285; objdiff=99.8927%; metric=((0, 4), 233, 233); unit code/data/functions=6464/0/35; pool=POOL IDENTICAL up to 0 (mine=0 base=0); regressions=[]; restored
ATTEMPT PFCACHE_DoWriteNumSectorAndFreeIfNeeded: typed read-only request-end record across overlap update boundary; source 9a7956d662f5; objdiff=99.8927%; metric=((0, 4), 233, 233); unit code/data/functions=6464/0/35; pool=POOL IDENTICAL up to 0 (mine=0 base=0); regressions=[]; restored
BEGIN NWC24iCheckDlHeaderConsistency: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=True; diagnosis metric=((2, 3), 212, 212)
NWC24iCheckDlHeaderConsistency local type proof: DlTaskData ends with url[0xec], fileName[0x5c], optOut byte and 3 reserved bytes, total 0x200; matches opaque NWC24DlTask 0x200. Stack object replacement represents the same task, with no padding added or extent change.
ATTEMPT NWC24iCheckDlHeaderConsistency: real structured task stack object cast at the opaque API boundary; source fcd54af0851f; objdiff=98.77358%; metric=((2, 3), 212, 212); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT NWC24iCheckDlHeaderConsistency: structured task receives opaque view through typed inline boundary; source 301ca3eb1edc; objdiff=98.77358%; metric=((2, 3), 212, 212); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
ATTEMPT NWC24iCheckDlHeaderConsistency: typed task pointer stays structured until each opaque read/delete boundary; source 1bc6e69d54e9; objdiff=98.77358%; metric=((2, 3), 212, 212); unit code/data/functions=8456/80/26; pool=POOL IDENTICAL up to 3 (mine=3 base=3); regressions=[]; restored
BEGIN PFCACHE_DoWriteNumSectorAndFreeIfNeeded: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((0, 4), 233, 233)
ATTEMPT PFCACHE_DoWriteNumSectorAndFreeIfNeeded: request-end formal is reduced in place to overlap count after deriving last sector; source b10773207f13; objdiff=99.8927%; metric=((0, 4), 233, 233); unit code/data/functions=6464/0/35; pool=POOL IDENTICAL up to 0 (mine=0 base=0); regressions=[]; restored
ATTEMPT PFCACHE_DoWriteNumSectorAndFreeIfNeeded: overlap count starts from end and subtracts page before computing last sector; source 85f5d01a2a49; objdiff=99.8927%; metric=((0, 4), 233, 233); unit code/data/functions=6464/0/35; pool=POOL IDENTICAL up to 0 (mine=0 base=0); regressions=[]; restored
ATTEMPT PFCACHE_DoWriteNumSectorAndFreeIfNeeded: const source-sector lifetime retained for both overlap updates without separate overlap object; source f643ee71364a; objdiff=99.8927%; metric=((0, 4), 233, 233); unit code/data/functions=6464/0/35; pool=POOL IDENTICAL up to 0 (mine=0 base=0); regressions=[]; restored
BEGIN pfd_sddrv_finalize: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((6, 4), 57, 57)
PFD_SDDRV_INFO type trial proof: media_inserted/media_ejected are 32-bit truth values. Target init and callbacks write 0/1; lifecycle code tests against zero, and offsets remain 0x10/0x14. BOOL preserves width and layout and adds no volatile.
ATTEMPT pfd_sddrv_finalize: media insertion field models target truth value as SDK BOOL; source b5214114691d; objdiff=92.63158%; metric=((6, 4), 57, 57); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_finalize: both lifecycle truth fields model SDK BOOL with unchanged offsets; source f4736e296197; objdiff=92.63158%; metric=((6, 4), 57, 57); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_finalize: signed word mask expression preserves only the initialized-bit clearing; source e483acc64c10; objdiff=92.63158%; metric=((6, 4), 57, 57); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
BEGIN pfd_sddrv_get_total_sectors: fetched origin/main=26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source differs from HEAD on remote=False; diagnosis metric=((0, 6), 80, 80)
ATTEMPT pfd_sddrv_get_total_sectors: wide unsigned shift factor is narrowed at final sector multiplication; source eef542901c40; objdiff=99.5%; metric=((0, 5), 80, 80); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_get_total_sectors: signed arithmetic shift factor narrowed at product, exponent constrained to zero through two; source 7322e282bd0f; objdiff=99.5%; metric=((0, 5), 80, 80); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
ATTEMPT pfd_sddrv_get_total_sectors: factor uses explicit unsigned shift expression and single 16-bit conversion; source 8666142025c4; objdiff=96.9875%; metric=((3, 10), 80, 80); unit code/data/functions=8940/3592/21; pool=POOL IDENTICAL up to 46 (mine=46 base=46); regressions=[]; restored
## Final open-function audit before clean gate
Every open symbol was checked after fresh origin fetch. Data and strings stayed identical; all failed variants restored. No declaration permutations used. All 13 open functions have at least three distinct source-level attempts with successful object builds.
OPEN NHTTPi_strnicmp: 99.76471%; 51/51 instructions; Z and zero initialization order exchanged, every later instruction identical; 3 distinct successful source attempts; all reverted.
OPEN NHTTPi_compareToken: 54.622223%; 44/45 instructions; loop entry and right-fold conditional value lifetime differ; 6 distinct successful source attempts; all reverted.
OPEN NHTTPi_Base64Encode: 60.285713%; 119/119 instructions; source byte/table lookup scheduling and signed padding comparisons differ; 3 distinct successful source attempts; all reverted.
OPEN NWC24InitDlTask: 98.923615%; 144/144 instructions; saved registers for zeros, owner ID, header and permission lifetimes differ; 3 distinct successful source attempts; all reverted.
OPEN NWC24UpdateDlTask: 95.00395%; 249/253 instructions; additional saved register/frame and permission/access/retry helper boundaries differ; 3 distinct successful source attempts; all reverted.
OPEN NWC24iCheckDlHeaderConsistency: 98.77358%; 212/212 instructions; stack task address materialization scheduled after saved argument copies; 6 distinct successful source attempts; all reverted.
OPEN AddTaskInternal: 97.7182%; 398/401 instructions; URL/free-slot/retry branch boundaries and saved register lifetimes differ; 3 distinct successful source attempts; all reverted.
OPEN pfd_sddrv_init: 92.326385%; 144/144 instructions; null/identity branch lowering and device/media/flag store-load scheduling differ; 3 distinct successful source attempts; all reverted.
OPEN pfd_sddrv_finalize: 92.63158%; 57/57 instructions; final flag-mask register and media-reset scheduling differ; 6 distinct successful source attempts; all reverted.
OPEN pfd_sddrv_get_total_sectors: 99.5%; 80/80 instructions; six register differences in narrow multiplier shift and product; 6 distinct successful source attempts; all reverted.
OPEN pfd_sddrv_store_mbr_buf: 99.75247%; 202/202 instructions; nine register differences for CHS track product and partition start; 3 distinct successful source attempts; all reverted.
OPEN pfd_sddrv_build_fat32_mbr_bpb: 94.75225%; 230/222 instructions; reserved-sector inline null/error path and block-write arguments differ; 3 distinct successful source attempts; all reverted.
OPEN PFCACHE_DoWriteNumSectorAndFreeIfNeeded: 99.8927%; 233/233 instructions; request end and page sector exchange r4/r5 in right-overlap arithmetic; 6 distinct successful source attempts; all reverted.
Source-quality audit: only accepted source change is const NWC24Work* in NWC24IterateDlTask. No comments, new asm, volatile, register keyword, label pins, dummy objects, function renames, symbol/config changes, shared headers, other units, link flags, or remote writes. Data totals remain 112/80/3592/0, all 100%. Local accepted source commit f9c79662.
## Final full clean gate over all four units
Command: python3 /mnt/drive2/projects/wii-ipl-workers/_restore0928-tools/gate.py libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL libs/RevoEX/src/nwc24/NWC24Download libs/RVL_SDK/src/fa/driver/sd_drv libs/RVL_SDK/src/fa/pf_cache
```text
full build: ok
main.dol sha1: 26116613f624061ba99c8d1a299aaa6efa85670d
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] pool: IDENTICAL
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] objdiff: code 1388/2248 data 112/112 functions 11/14 fuzzy 87.9359 linked code 0
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] instruction-exact functions: 11/14
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] section .data size 72 match 100.0
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] section .rodata size 40 match 100.0
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] section .text size 2248 match 87.93594
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] below 100: NHTTPi_strnicmp 99.76471
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] below 100: NHTTPi_compareToken 54.622223
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] below 100: NHTTPi_Base64Encode 60.285713
[libs/RevoEX/src/nhttp/NHTTP_stdlib_RVL] baseline: code 1388/2248 data 112 functions 11 fuzzy 87.9359
[libs/RevoEX/src/nwc24/NWC24Download] pool: IDENTICAL
[libs/RevoEX/src/nwc24/NWC24Download] objdiff: code 8456/12496 data 80/80 functions 26/30 fuzzy 99.1697 linked code 0
[libs/RevoEX/src/nwc24/NWC24Download] instruction-exact functions: 26/30
[libs/RevoEX/src/nwc24/NWC24Download] section .data size 56 match 100.0
[libs/RevoEX/src/nwc24/NWC24Download] section .sbss size 8 match 100.0
[libs/RevoEX/src/nwc24/NWC24Download] section .sdata size 16 match 100.0
[libs/RevoEX/src/nwc24/NWC24Download] section .text size 12496 match 99.169655
[libs/RevoEX/src/nwc24/NWC24Download] below 100: NWC24InitDlTask 98.923615
[libs/RevoEX/src/nwc24/NWC24Download] below 100: NWC24UpdateDlTask 95.00395
[libs/RevoEX/src/nwc24/NWC24Download] below 100: NWC24iCheckDlHeaderConsistency 98.77358
[libs/RevoEX/src/nwc24/NWC24Download] below 100: AddTaskInternal 97.7182
[libs/RevoEX/src/nwc24/NWC24Download] baseline: code 8140/12496 data 80 functions 25 fuzzy 99.1521
[libs/RVL_SDK/src/fa/driver/sd_drv] pool: IDENTICAL
[libs/RVL_SDK/src/fa/driver/sd_drv] objdiff: code 8940/11760 data 3592/3592 functions 21/26 fuzzy 99.0544 linked code 0
[libs/RVL_SDK/src/fa/driver/sd_drv] instruction-exact functions: 21/26
[libs/RVL_SDK/src/fa/driver/sd_drv] section .bss size 608 match 100.0
[libs/RVL_SDK/src/fa/driver/sd_drv] section .data size 2576 match 100.0
[libs/RVL_SDK/src/fa/driver/sd_drv] section .rodata size 368 match 100.0
[libs/RVL_SDK/src/fa/driver/sd_drv] section .sbss size 40 match 100.0
[libs/RVL_SDK/src/fa/driver/sd_drv] section .text size 11760 match 99.05442
[libs/RVL_SDK/src/fa/driver/sd_drv] below 100: pfd_sddrv_init 92.326385
[libs/RVL_SDK/src/fa/driver/sd_drv] below 100: pfd_sddrv_finalize 92.63158
[libs/RVL_SDK/src/fa/driver/sd_drv] below 100: pfd_sddrv_get_total_sectors 99.5
[libs/RVL_SDK/src/fa/driver/sd_drv] below 100: pfd_sddrv_store_mbr_buf 99.75247
[libs/RVL_SDK/src/fa/driver/sd_drv] below 100: pfd_sddrv_build_fat32_mbr_bpb 94.75225
[libs/RVL_SDK/src/fa/driver/sd_drv] baseline: code 8940/11760 data 3592 functions 21 fuzzy 99.0544
[libs/RVL_SDK/src/fa/pf_cache] pool: IDENTICAL
[libs/RVL_SDK/src/fa/pf_cache] objdiff: code 6464/7396 data None/None functions 35/36 fuzzy 99.9865 linked code 0
[libs/RVL_SDK/src/fa/pf_cache] instruction-exact functions: 35/36
[libs/RVL_SDK/src/fa/pf_cache] section .text size 7396 match 99.98648
[libs/RVL_SDK/src/fa/pf_cache] below 100: PFCACHE_DoWriteNumSectorAndFreeIfNeeded 99.8927
[libs/RVL_SDK/src/fa/pf_cache] baseline: code 6464/7396 data None functions 35 fuzzy 99.9865
regressions vs baseline: 0
global matched_code_percent: 91.02223 -> 91.03278
global fuzzy_match_percent: 99.58757 -> 99.58764
global complete_code_percent: 73.73083 -> 73.73083
global matched_data_percent: 99.42271 -> 99.42271
forbidden patterns added (net, per file): 0
readability warnings (net, per file; must be 0 in the final result): 0
GATE PASS
```
Before -> after instruction-exact/code/data:
NHTTP_stdlib_RVL: 11/14 -> 11/14; 1388/2248 -> 1388/2248; 112/112 -> 112/112.
NWC24Download: 25/30 -> 26/30; 8140/12496 -> 8456/12496; 80/80 -> 80/80.
sd_drv: 21/26 -> 21/26; 8940/11760 -> 8940/11760; 3592/3592 -> 3592/3592.
pf_cache: 35/36 -> 35/36; 6464/7396 -> 6464/7396; 0/0 -> 0/0.
Only NWC24IterateDlTask gained exact bytes. Const work view is proven by objdiff 100%, 79/79 identical instructions, and the final clean gate. All other trials are reverted. Remaining functions and reasons are in the OPEN audit above. No uncertainty about measured gains; source-level causes for the parked compiler differences remain unresolved.
Official declsearch scoring only after structural/source trials. No permutations and no source mutations.
NHTTPi_strnicmp: structural 2 exact 2.
NHTTPi_compareToken: structural 24 exact 42.
NHTTPi_Base64Encode: structural 71 exact 58.
NWC24InitDlTask: structural 0 exact 31.
NWC24UpdateDlTask: structural 19 exact 250.
NWC24iCheckDlHeaderConsistency: structural 2 exact 3.
AddTaskInternal: structural 20 exact 323.
pfd_sddrv_init: structural 18 exact 31.
pfd_sddrv_finalize: structural 6 exact 4.
pfd_sddrv_get_total_sectors: structural 0 exact 6.
pfd_sddrv_store_mbr_buf: structural 0 exact 9.
pfd_sddrv_build_fat32_mbr_bpb: structural 24 exact 122.
PFCACHE_DoWriteNumSectorAndFreeIfNeeded: structural 0 exact 4.
Accepted symbol final fresh ctxdiff:
```text
src 0x13c base 0x13c insns 79/79
diffs 0: []
```
Final progress/report.json and build/43U/ok targets pass. The full clean gate remains GATE PASS. Source change and all non-text scores are unchanged after scoring.

View File

@@ -0,0 +1,460 @@
# prior1 MAX reference round
Scope: VIInit, Window::DrawFrame, SOGetSockName and owned data. No wprintf edits.
Origin fetch succeeded; 26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source and symbols match HEAD/origin. Shell execution works; historical bubblewrap blocker does not apply.
BASELINE main/libs/RevoEX/src/so/SOBasic {'fuzzy_match_percent': 99.79452, 'total_code': '4088', 'matched_code': '3836', 'matched_code_percent': 93.83562, 'total_data': '144', 'matched_data': '144', 'matched_data_percent': 100.0, 'total_functions': 22, 'matched_functions': 21, 'matched_functions_percent': 95.454544, 'total_units': 1}
Initial string pool IDENTICAL; data already 100% for every owned section. No symbol renames or extent changes needed.
BASELINE main/libs/NW4R/src/lyt/lyt_window {'fuzzy_match_percent': 99.75864, 'total_code': '11352', 'matched_code': '9848', 'matched_code_percent': 86.751236, 'total_data': '316', 'matched_data': '316', 'matched_data_percent': 100.0, 'total_functions': 21, 'matched_functions': 20, 'matched_functions_percent': 95.2381, 'total_units': 1}
Initial string pool IDENTICAL; data already 100% for every owned section. No symbol renames or extent changes needed.
BASELINE main/libs/RVL_SDK/src/vi/vi {'fuzzy_match_percent': 99.97521, 'total_code': '11296', 'matched_code': '9944', 'matched_code_percent': 88.03116, 'total_data': '1944', 'matched_data': '1944', 'matched_data_percent': 100.0, 'total_functions': 30, 'matched_functions': 29, 'matched_functions_percent': 96.666664, 'total_units': 1}
Initial string pool IDENTICAL; data already 100% for every owned section. No symbol renames or extent changes needed.
VIInit structural diagnosis: both frames 0x30; 338/338 instructions. Only 14 differences: clip active-height/top locals r29/r30/r31 rotate with shared framebuffer/PreCB/PostCB zero. No target spill/reload proof for new volatile. Reference changes declarations, explicit narrowing in AdjustPosition, and DVD-stop inline boundary; data 1944/1944.
VIInit | V1 reference removes unused format, DVD previous state block scoped | insns 338/338 structural/exact (0, 14) | objdiff 99.7929 | lost [] | data 1944/1944 | POOL IDENTICAL up to 7 (mine=7 base=7) | 63b67f44ab8a
199 M ('slwi', 'r29, r7, 1') B ('slwi', 'r30, r7, 1')
201 M ('subf', 'r26, r6, r29') B ('subf', 'r26, r6, r30')
222 M ('and', 'r30, r25, r10') B ('and', 'r31, r25, r10')
225 M ('and', 'r29, r26, r12') B ('and', 'r30, r26, r12')
227 M ('add', 'r7, r7, r30') B ('add', 'r7, r7, r31')
235 M ('li', 'r31, 0') B ('li', 'r29, 0')
238 M ('divw', 'r8, r29, r5') B ('divw', 'r8, r30, r5')
248 M ('sth', 'r31, 0x16(r4)') B ('sth', 'r29, 0x16(r4)')
250 M ('sth', 'r31, 0x18(r4)') B ('sth', 'r29, 0x18(r4)')
255 M ('stw', 'r31, 0x20(r4)') B ('stw', 'r29, 0x20(r4)')
260 M ('stb', 'r31, 0x3c(r4)') B ('stb', 'r29, 0x3c(r4)')
262 M ('stw', 'r31, 0x44(r4)') B ('stw', 'r29, 0x44(r4)')
274 M ('stw', 'r31, 0(0)') B ('stw', 'r29, 0(0)')
275 M ('stw', 'r31, 0(0)') B ('stw', 'r29, 0(0)')
VIInit | V2 reference scalar declaration group and required volatile DVD read | insns 338/338 structural/exact (0, 14) | objdiff 99.7929 | lost [] | data 1944/1944 | POOL IDENTICAL up to 7 (mine=7 base=7) | b4b22213ff0c
199 M ('slwi', 'r29, r7, 1') B ('slwi', 'r30, r7, 1')
201 M ('subf', 'r26, r6, r29') B ('subf', 'r26, r6, r30')
222 M ('and', 'r30, r25, r10') B ('and', 'r31, r25, r10')
225 M ('and', 'r29, r26, r12') B ('and', 'r30, r26, r12')
227 M ('add', 'r7, r7, r30') B ('add', 'r7, r7, r31')
235 M ('li', 'r31, 0') B ('li', 'r29, 0')
238 M ('divw', 'r8, r29, r5') B ('divw', 'r8, r30, r5')
248 M ('sth', 'r31, 0x16(r4)') B ('sth', 'r29, 0x16(r4)')
250 M ('sth', 'r31, 0x18(r4)') B ('sth', 'r29, 0x18(r4)')
255 M ('stw', 'r31, 0x20(r4)') B ('stw', 'r29, 0x20(r4)')
260 M ('stb', 'r31, 0x3c(r4)') B ('stb', 'r29, 0x3c(r4)')
262 M ('stw', 'r31, 0x44(r4)') B ('stw', 'r29, 0x44(r4)')
274 M ('stw', 'r31, 0(0)') B ('stw', 'r29, 0(0)')
275 M ('stw', 'r31, 0(0)') B ('stw', 'r29, 0(0)')
VIInit | V3 reference interrupt register value assignment order | insns 338/338 structural/exact (0, 0) | objdiff 100.0 | lost [] | data 1944/1944 | POOL IDENTICAL up to 7 (mine=7 base=7) | 4cf8217f30f5
VIInit | V4 reference clipping explicit u16 stores and grouped locals | insns 338/338 structural/exact (0, 14) | objdiff 99.7929 | lost [] | data 1944/1944 | POOL IDENTICAL up to 7 (mine=7 base=7) | 5e17b0aaa91d
199 M ('slwi', 'r29, r7, 1') B ('slwi', 'r30, r7, 1')
201 M ('subf', 'r26, r6, r29') B ('subf', 'r26, r6, r30')
222 M ('and', 'r30, r25, r10') B ('and', 'r31, r25, r10')
225 M ('and', 'r29, r26, r12') B ('and', 'r30, r26, r12')
227 M ('add', 'r7, r7, r30') B ('add', 'r7, r7, r31')
235 M ('li', 'r31, 0') B ('li', 'r29, 0')
238 M ('divw', 'r8, r29, r5') B ('divw', 'r8, r30, r5')
248 M ('sth', 'r31, 0x16(r4)') B ('sth', 'r29, 0x16(r4)')
250 M ('sth', 'r31, 0x18(r4)') B ('sth', 'r29, 0x18(r4)')
255 M ('stw', 'r31, 0x20(r4)') B ('stw', 'r29, 0x20(r4)')
260 M ('stb', 'r31, 0x3c(r4)') B ('stb', 'r29, 0x3c(r4)')
262 M ('stw', 'r31, 0x44(r4)') B ('stw', 'r29, 0x44(r4)')
274 M ('stw', 'r31, 0(0)') B ('stw', 'r29, 0(0)')
275 M ('stw', 'r31, 0(0)') B ('stw', 'r29, 0(0)')
VIInit | V5 hardware register value uses its actual u16 width | insns 338/338 structural/exact (0, 0) | objdiff 100.0 | lost [] | data 1944/1944 | POOL IDENTICAL up to 7 (mine=7 base=7) | 91d73ae4ffad
V5 clean full gate before commit
```
full build: ok
main.dol sha1: 26116613f624061ba99c8d1a299aaa6efa85670d
[libs/RVL_SDK/src/vi/vi] pool: IDENTICAL
[libs/RVL_SDK/src/vi/vi] objdiff: code 11296/11296 data 1944/1944 functions 30/30 fuzzy 100.0000 linked code 0
[libs/RVL_SDK/src/vi/vi] instruction-exact functions: 30/30
[libs/RVL_SDK/src/vi/vi] section .bss size 368 match 100.0
[libs/RVL_SDK/src/vi/vi] section .data size 1368 match 100.0
[libs/RVL_SDK/src/vi/vi] section .sbss size 176 match 100.0
[libs/RVL_SDK/src/vi/vi] section .sdata size 32 match 100.0
[libs/RVL_SDK/src/vi/vi] section .text size 11296 match 100.0
[libs/RVL_SDK/src/vi/vi] baseline: code 9944/11296 data 1944 functions 29 fuzzy 99.9752
regressions vs baseline: 0
global matched_code_percent: 91.05281 -> 91.09795
global fuzzy_match_percent: 99.58826 -> 99.58836
global complete_code_percent: 73.73083 -> 74.01916
global matched_data_percent: 99.42271 -> 99.42271
forbidden patterns added (net, per file): 0
readability warnings (net, per file; must be 0 in the final result): 0
note: no baseline for merge-base 26d7e1e0; compared against nearest snapshotted ancestor d7821052 (1 commits back)
GATE PASS
```
VIInit accepted locally: value now u16, matching the VI hardware register width. Both masks preserve 16 bits. This changes only register coloring, resolves every 14 instruction difference, and retains all sections at 100%. Clean full gate PASS; ctxdiff 338/338 diffs 0; pool 7/7 identical; zero regressions and forbidden/readability patterns. No configure.py or symbols change.
DrawFrame origin pre-function fetch: 26d7e1e052a80ea55d0591a82653370965ad5bb2; owned source identical to origin. Target frame 0xe0, 376/376 instructions. All 119 differences are registers: persistent flip table r21 vs r31 shifts geometry and alpha. Static table, branches, helper operations, floating-point order, and stack offsets identical; reference lacks the target-proven texture-count guard and has mutable bool. No volatile spill/reload proof; all 316 data bytes matched. Extent 0x5e0 ends exactly at DrawFrame4.
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W1 reference mutable bUseVtxCol | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | e4d4a68bdadd
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W2 material setup result separate declaration and assignment | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 645a40fa68ea
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W3 texture dimensions separate declaration and assignment | insns 381/376 structural/exact (161, 358) | objdiff 80.97607 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 356336b28e11
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W4 coordinates passed via reference to the quad array | insns 381/376 structural/exact (96, 326) | objdiff 76.89096 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | b75bb1a0e722
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W5 alpha parameter const-qualified at definition | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | c852083af7e9
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W6 explicitly narrow texture unit argument in vertex format and dimension query | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 19a85be12cc2
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W7 texture index ix uses u8 | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 5cef145ed39c
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W7 texture index iy uses u8 | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 962c307d665a
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W7 texture index ix,iy uses u8 | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | fba15f864c06
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W7 texture index ix,iy uses u16 | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 9230751b3b92
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W7 texture index ix,iy uses u32 | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 01617eecae4e
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W7 texture index ix,iy uses unsigned long | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | aefd446c1d1d
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W8 reference plain struct declaration for TextureFlipInfo | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 127501278333
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W9 reference texture-flip type declared outside anonymous namespace | BUILD FAILED | a29d9ea4e268
2/mwcceppc.exe -nodefaults -proc gekko -align powerpc -enum int -fp hardware -Cpp_exceptions off -O4,p -inline auto -pragma "cats off" -pragma "warn_notinlined off" -maxerrors 1 -nosyspath -RTTI off -fp_contract on -str reuse -DSDK_IPL -D_REVOLUTION -DMEM_MANAGER_DIRECT -i include -i include/global -i libs/MetroTRK/include -i libs/Runtime/include -i libs/MSL/include -i libs/RVL_SDK/include -i libs/RevoEX/include -i libs/NW4R/include -i libs/RVLMiddleware/eZiText/include -i libs/RVLMiddleware/TMC_JPEG/include -i libs/RVLFaceLib/include -i libs/EGG/include -i libs/OperaWWW/include -ir libs/RVL_SDK/include/private/bte -i build/43U/include -DBUILD_VERSION=0 -DVERSION_43U -i libs/RVL_SDK/include/private/bte -DNDEBUG=1 -DTARGET_RVL -ipa file -fp_contract off -Cpp_exceptions off -lang=c++ -MMD -c libs/NW4R/src/lyt/lyt_window.cpp -o build/43U/src/libs/NW4R/src/lyt && "/usr/bin/python3" tools/transform_dep.py build/43U/src/libs/NW4R/src/lyt/lyt_window.d build/43U/src/libs/NW4R/src/lyt/lyt_window.d
### mwcceppc.exe Compiler:
# File: libs\NW4R\src\lyt\lyt_window.cpp
# -----------------------------------------
# 48: u8 coords[VERTEXCOLOR_MAX][FLIPINDEX_MAX]; // 0x00
# Error: ^^^^^^^^^^^^^^^
# (10140) undefined identifier 'VERTEXCOLOR_MAX'
# Too many errors printed, aborting program
User break, cancelled...
ninja: build stopped: subcommand failed.
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W10 reference helper declarations before definitions | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | a4897278131e
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W9b reference global texture-flip struct with real array dimensions | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | c3ee228d2a7f
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W11 explicit inline boundary for GetTexutreFlipInfo | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 60/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | e833af53f19d
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W11 explicit inline boundary for GetLTTexCoord | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 383a08cfd390
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W11 explicit inline boundary for GetRTTexCoord | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 3d074b950c89
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W11 explicit inline boundary for GetLBTexCoord | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 7a08a15fb0f3
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W11 explicit inline boundary for GetRBTexCoord | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 0e6faa8f9403
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W12 geometry and texture-coordinate helpers explicit inline | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 60/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 1ccea4bdb91d
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W13 polygon size constructor explicitly initializes its dimensions | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | dadaf033b0f2
DrawFrame W14 final declaration-order search after structural/reference/type trials
declaration block:
const Size texSize = detail::GetTextureSize(frame.pMaterial, GX_TEXMAP0);
const ut::Color vtxColors[VERTEXCOLOR_MAX];
detail::TexCoords texCds[1];
math::VEC2 polPt;
Size polSize;
start (0, 119)
best (0, 119) after 52 builds; source restored; best order was:
const Size texSize = detail::GetTextureSize(frame.pMaterial, GX_TEXMAP0);
const ut::Color vtxColors[VERTEXCOLOR_MAX];
detail::TexCoords texCds[1];
math::VEC2 polPt;
Size polSize;
DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | W14 final declaration-order result | insns 376/376 structural/exact (0, 119) | objdiff 98.17819 | lost [] | data 316/316 | POOL IDENTICAL up to 1 (mine=1 base=1) | 3cc7e4df0b05
SOGetSockName origin pre-function fetch: 805e84045c82629c645e9ef1787c30d2bc33b5e7; owned source identical to origin. Pool 1/1 identical. Both frames 0x30,63/63 instructions; only first memcpy setup differs (ours destination/source/length, target source/length/destination). Neighbor extent starts at 0x814B34CC, exactly 0x814B33D0+0xFC; no overlap. All 144 target data bytes matched. Reference uses early goto conclude, address initializer, sizeof comparison, separate prepare assignment; wire request raw-offset expressions replaced by existing NameRequest field.
SOGetSockName | S1 reference statement order and conclude goto, with typed wire fields | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 33cc57840375
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S2 reference address initialization and separate prepare assignment | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | b5a1a7e874aa
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S3 first address copy uses an explicit const input view | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 53088c5d144f
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S4 const socket address view for validation and input, mutable output parameter | insns 63/63 structural/exact (2, 5) | objdiff 96.349205 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | c08e5df4c7de
5 M ('mr', 'r27, r4') B ('mr', 'r28, r3')
6 M ('mr', 'r28, r3') B ('mr', 'r27, r4')
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S5 first copy views its header as const bytes | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 9e0bb74dd11a
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S6 typed copy helper with const source boundary | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 6a49cab5c214
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S7 allocation size declared u32 | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 68140082eff9
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S7 allocation size declared u16 | insns 63/63 structural/exact (3, 4) | objdiff 95.71429 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 404166a4490e
22 M ('rlwinm', 'r31, r0, 0, 0x10, 0x1a') B ('rlwinm', 'r31, r0, 0, 0, 0x1a')
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S7 allocation size declared size_t | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 73d720d3670d
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S8 first copy separately reads length as u8 | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | c847ce10914f
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S8 first copy separately reads length as u16 | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | becaa347c984
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S8 first copy separately reads length as u32 | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | a1d015de3910
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S8 first copy separately reads length as int | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | d235318a13b5
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S8 first copy separately reads length as size_t | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 43864ee20f80
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S9 reference command allocation is int pointer with typed address-field view | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | f42be4f4821b
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S10 reply buffer held as opaque IOCTL buffer with typed length view | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 39d08eea8df3
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S11 promote first copy length through unsigned halfword | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 8dae136f16d1
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S12 syscall result declared s16 | insns 63/63 structural/exact (8, 20) | objdiff 91.5873 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 2e50fd103218
10 M ('extsh.', 'r3, r3') B ('cmpwi', 'r3, 0')
19 M ('b', '136') B ('b', '140')
29 M ('b', '96') B ('b', '100')
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
43 M ('extsh.', 'r30, r3') B ('cmpwi', 'r3, 0')
44 M ('blt', '20') B ('mr', 'r30, r3')
45 M ('lbz', 'r5, 0(r28)') B ('blt', '20')
46 M ('mr', 'r3, r27') B ('lbz', 'r5, 0(r28)')
47 M ('mr', 'r4, r28') B ('mr', 'r3, r27')
48 M ('bl', '0') B ('mr', 'r4, r28')
49 M ('mr', 'r4, r29') B ('bl', '0')
50 M ('mr', 'r5, r31') B ('mr', 'r4, r29')
51 M ('li', 'r3, 0xc') B ('mr', 'r5, r31')
52 M ('bl', '0') B ('li', 'r3, 0xc')
53 M ('mr', 'r4, r30') B ('bl', '0')
54 M ('li', 'r3, 0') B ('mr', 'r4, r30')
55 M ('bl', '0') B ('li', 'r3, 0')
56 M ('extsh', 'r3, r3') B ('bl', '0')
SOGetSockName | S12 syscall result declared long | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | b72c9cdddb5e
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S12 syscall result declared s32 | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 1ab005e08c72
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S13 reply initialized through memcpy destination return | insns 63/63 structural/exact (10, 20) | objdiff 89.12698 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 7693f4b6b84c
5 M ('mr', 'r29, r3') B ('mr', 'r28, r3')
18 M ('li', 'r29, -0x1c') B ('li', 'r30, -0x1c')
22 M ('rlwinm', 'r30, r0, 0, 0, 0x1a') B ('rlwinm', 'r31, r0, 0, 0, 0x1a')
23 M ('mr', 'r4, r30') B ('mr', 'r4, r31')
26 M ('mr', 'r28, r3') B ('mr', 'r29, r3')
28 M ('li', 'r29, -0x31') B ('li', 'r30, -0x31')
30 M ('stw', 'r29, 0(r3)') B ('stw', 'r28, 0(r3)')
31 M ('mr', 'r4, r27') B ('addi', 'r28, r3, 0x20')
32 M ('addi', 'r3, r3, 0x20') B ('mr', 'r4, r27')
34 M ('bl', '0') B ('mr', 'r3, r28')
35 M ('mr', 'r31, r3') B ('bl', '0')
37 M ('lbz', 'r8, 0(r27)') B ('mr', 'r5, r29')
38 M ('mr', 'r5, r28') B ('lbz', 'r8, 0(r27)')
39 M ('mr', 'r7, r31') B ('mr', 'r7, r28')
44 M ('mr', 'r29, r3') B ('mr', 'r30, r3')
46 M ('lbz', 'r5, 0(r31)') B ('lbz', 'r5, 0(r28)')
48 M ('mr', 'r4, r31') B ('mr', 'r4, r28')
50 M ('mr', 'r4, r28') B ('mr', 'r4, r29')
51 M ('mr', 'r5, r30') B ('mr', 'r5, r31')
54 M ('mr', 'r4, r29') B ('mr', 'r4, r30')
SOGetSockName | S14 reply pointer obtained from the initial address-field copy | insns 63/63 structural/exact (10, 6) | objdiff 90.07937 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 59b28a410bde
31 M ('mr', 'r4, r27') B ('addi', 'r28, r3, 0x20')
32 M ('addi', 'r3, r3, 0x20') B ('mr', 'r4, r27')
34 M ('bl', '0') B ('mr', 'r3, r28')
35 M ('mr', 'r28, r3') B ('bl', '0')
37 M ('lbz', 'r8, 0(r27)') B ('mr', 'r5, r29')
38 M ('mr', 'r5, r29') B ('lbz', 'r8, 0(r27)')
SOGetSockName | S15 socket command and read-only address initialized in one typed helper | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 53a9d11f1936
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
SOGetSockName | S16 source-first read-only request initializer returns the address buffer | insns 63/63 structural/exact (2, 19) | objdiff 95.39683 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 7a7c904069d2
5 M ('mr', 'r29, r3') B ('mr', 'r28, r3')
18 M ('li', 'r29, -0x1c') B ('li', 'r30, -0x1c')
22 M ('rlwinm', 'r30, r0, 0, 0, 0x1a') B ('rlwinm', 'r31, r0, 0, 0, 0x1a')
23 M ('mr', 'r4, r30') B ('mr', 'r4, r31')
26 M ('mr', 'r28, r3') B ('mr', 'r29, r3')
28 M ('li', 'r29, -0x31') B ('li', 'r30, -0x31')
30 M ('stw', 'r29, 0(r3)') B ('stw', 'r28, 0(r3)')
31 M ('addi', 'r31, r3, 0x20') B ('addi', 'r28, r3, 0x20')
32 M ('mr', 'r3, r31') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
37 M ('mr', 'r5, r28') B ('mr', 'r5, r29')
39 M ('mr', 'r7, r31') B ('mr', 'r7, r28')
44 M ('mr', 'r29, r3') B ('mr', 'r30, r3')
46 M ('lbz', 'r5, 0(r31)') B ('lbz', 'r5, 0(r28)')
48 M ('mr', 'r4, r31') B ('mr', 'r4, r28')
50 M ('mr', 'r4, r28') B ('mr', 'r4, r29')
51 M ('mr', 'r5, r30') B ('mr', 'r5, r31')
54 M ('mr', 'r4, r29') B ('mr', 'r4, r30')
SOGetSockName S17 final declaration-order search after reference/copy-type/helper trials
declaration block:
int size;
int result;
s32 rm;
NameRequest* request;
SOSockAddr* reply;
SOSockAddr* addr;
start (2, 3)
best (2, 3) after 36 builds; source restored; best order was:
int size;
int result;
s32 rm;
NameRequest* request;
SOSockAddr* reply;
SOSockAddr* addr;
SOGetSockName | S17 final declaration-order result | insns 63/63 structural/exact (2, 3) | objdiff 96.666664 | lost [] | data 144/144 | POOL IDENTICAL up to 1 (mine=1 base=1) | 73d59f1449b9
32 M ('mr', 'r3, r28') B ('mr', 'r4, r27')
33 M ('mr', 'r4, r27') B ('lbz', 'r5, 0(r27)')
34 M ('lbz', 'r5, 0(r27)') B ('mr', 'r3, r28')
FINAL OPEN AUDIT | SOGetSockName | objdiff 96.666664% | 25 distinct successful source trials | all unsuccessful source restored
FINAL OPEN AUDIT | DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc | objdiff 98.17819% | 23 distinct successful source trials | all unsuccessful source restored
Retained source: VIInit hardware register value u16 only, local commit bfe67f5a. No other source/config/header changes. All owned data sections 100%; data symbol/extent corrections unnecessary. DrawFrame remains register-only at 376/376, 119 differences; SOGetSockName remains first-copy setup scheduling at 63/63, 3 differences.
Final full clean gate over all owned units
```
full build: ok
main.dol sha1: 26116613f624061ba99c8d1a299aaa6efa85670d
[libs/RVL_SDK/src/vi/vi] pool: IDENTICAL
[libs/RVL_SDK/src/vi/vi] objdiff: code 11296/11296 data 1944/1944 functions 30/30 fuzzy 100.0000 linked code 0
[libs/RVL_SDK/src/vi/vi] instruction-exact functions: 30/30
[libs/RVL_SDK/src/vi/vi] section .bss size 368 match 100.0
[libs/RVL_SDK/src/vi/vi] section .data size 1368 match 100.0
[libs/RVL_SDK/src/vi/vi] section .sbss size 176 match 100.0
[libs/RVL_SDK/src/vi/vi] section .sdata size 32 match 100.0
[libs/RVL_SDK/src/vi/vi] section .text size 11296 match 100.0
[libs/RVL_SDK/src/vi/vi] baseline: code 9944/11296 data 1944 functions 29 fuzzy 99.9752
[libs/NW4R/src/lyt/lyt_window] pool: IDENTICAL
[libs/NW4R/src/lyt/lyt_window] objdiff: code 9848/11352 data 316/316 functions 20/21 fuzzy 99.7586 linked code 0
[libs/NW4R/src/lyt/lyt_window] instruction-exact functions: 20/21
[libs/NW4R/src/lyt/lyt_window] section .ctors size 4 match 100.0
[libs/NW4R/src/lyt/lyt_window] section .data size 256 match 100.0
[libs/NW4R/src/lyt/lyt_window] section .sbss size 8 match 100.0
[libs/NW4R/src/lyt/lyt_window] section .sdata2 size 48 match 100.0
[libs/NW4R/src/lyt/lyt_window] section .text size 11352 match 99.75864
[libs/NW4R/src/lyt/lyt_window] below 100: DrawFrame__Q34nw4r3lyt6WindowFRCQ34nw4r4math4VEC2RCQ44nw4r3lyt6Window5FrameRCQ34nw4r3lyt15WindowFrameSizeUc 98.17819
[libs/NW4R/src/lyt/lyt_window] baseline: code 9848/11352 data 316 functions 20 fuzzy 99.7586
[libs/RevoEX/src/so/SOBasic] pool: IDENTICAL
[libs/RevoEX/src/so/SOBasic] objdiff: code 3836/4088 data 144/144 functions 21/22 fuzzy 99.7945 linked code 0
[libs/RevoEX/src/so/SOBasic] instruction-exact functions: 21/22
[libs/RevoEX/src/so/SOBasic] section .bss size 40 match 100.0
[libs/RevoEX/src/so/SOBasic] section .data size 88 match 100.0
[libs/RevoEX/src/so/SOBasic] section .sbss size 8 match 100.0
[libs/RevoEX/src/so/SOBasic] section .sdata size 8 match 100.0
[libs/RevoEX/src/so/SOBasic] section .text size 4088 match 99.79452
[libs/RevoEX/src/so/SOBasic] below 100: SOGetSockName 96.666664
[libs/RevoEX/src/so/SOBasic] baseline: code 3836/4088 data 144 functions 21 fuzzy 99.7945
regressions vs baseline: 0
global matched_code_percent: 91.05281 -> 91.09795
global fuzzy_match_percent: 99.58826 -> 99.58836
global complete_code_percent: 74.01916 -> 74.01916
global matched_data_percent: 99.42271 -> 99.42271
forbidden patterns added (net, per file): 0
readability warnings (net, per file; must be 0 in the final result): 0
GATE PASS
```
Post-gate checks: ninja -C . build/43U/ok progress build/43U/report.json PASS. DOL SHA1 26116613f624061ba99c8d1a299aaa6efa85670d. VIInit ctxdiff 338/338 diffs 0; SOGetSockName 63/63 diffs 3. Final pools identical: vi 7/7, window 1/1, SOBasic 1/1. Git diff --check PASS.
Final owned metrics (exact functions, matched code bytes, matched data bytes): vi 29/30,9944/11296,1944/1944 -> 30/30,11296/11296,1944/1944; lyt_window 20/21,9848/11352,316/316 unchanged; SOBasic 21/22,3836/4088,144/144 unchanged.
Final origin/main advanced to 805e84045c82629c645e9ef1787c30d2bc33b5e7 during other workers: the local scope diff against merge-base 26d7e1e0 contains only vi.c and this attempts log; no out-of-scope edits. Worker did not rebase, link, push or contact upstream.

View File

@@ -0,0 +1,130 @@
# revisit leaf — second-pass evidence (agent/w1007/revisit @ origin/main 29fe1efa)
Scope: iplSDChannelTitle (99.98883), ut_ArchiveFontBase (98.52), lyt_window (99.76),
iplMemoryCardManager (97.88). Post-#834 (linkgap emit gates + .data object names).
## iplSDChannelTitle — all gates landed
Per-TU `IPL_SD_CHANNEL_TITLE_CPP` declared-only gates applied to:
RuntimeTypeInfo.h DynamicCast, iplMathTypes.h VEC3 3-arg ctor, GUIManager.h
(EventHandler setManager/setLatestEventCtrlNo/getLatestEventCtrlNo,
Component setTriggerTarget, PaneComponent setPane/getPane), iplGraphics.h
setOrthoTrans/setOrthoScale, iplSceneBase.h isReady/isResetProcessDone/getParent/
getChild/getNext/getPrev, iplFaderSceneBase.h initCalcNormal/calcCommonAfter,
iplNand.h File::isFinished/checkData, iplInterporation.h HermiteIntp<f32>::get
(nested under IPL_CHANNEL_TITLE_NOVTABLE spec), Rect.h 0-arg ctor
(IPL_SOUND_RECT_OUT_OF_LINE gate), iplGuiManager.h ~PaneManager.
Deleted 2 unreferenced extern "C" trampolines (onTitleButtonEvent/onTitlePaneEvent)
orig lacks. Result: all data sections 100%; .text 99.98883.
### VERIFIED WALL — dtor emission policy (decl-immune)
Residual .text gap = 2 unreferenced weak deleting-dtor bodies
(~FaderSceneBase 88B + ~FrameController 64B) + ~5 dangling no-reloc UNDs
(__vt__LangFile, __vt__SDButtonEventHandlerBase, __vt__FrameController,
mArg/typeInfo syms — symtab noise, no relocs). Orig's TU had these dtors
trivial/implicit: ~SDChannelTitle calls ~Base directly (28 insns), Rect
temporaries carry no dtor calls, so correct forms are `virtual ~X() {}`
inline or NO decl at all. Declaring them out-of-line adds real `bl` calls →
fuzzy dropped 99.98883→99.55799. Implicit/non-virtual variants still emit the
weak bodies AND slightly regress — front-end-level MWCC instantiation drag,
no decl form eliminates it. Remaining fn wall: flushSaveBeforeExit 98.59 —
&smArg materialization register choice (orig r3→scratch r4, mine r4/r5→recv-first);
8+ source forms all shift the home, never to r3.
## ut_ArchiveFontBase — ConstructOpAnalyzeGLGR 92.09 (956B)
20 normalized diffs, insns 116-147 of 239 — issue-order inside the
ready-set around the CalcOffsetSheetFlags/CalcSizeFlagSet scratch math.
Orig hoists lhz 0x1c (sheetGlyphCount) / lhz 0x24 (sheetCount) / lhz 0xe
(dataBlockCount) + lwz 0x4c(r20) into the marshal block.
FAILED variants: decl-order swap of the three u16s (identical 20 diffs —
scheduler ignores decl order); moving ctx->m* tail stores adjacent to the
u16 loads (regressed 20→108, whole-web rotation). Conclusion: pure list-
scheduling tie, same family as before.
## lyt_window — DrawFrame 98.18 (1504B)
376=376 insns. One callee-window permutation: orig numbers the static
flipInfos[] array base early (r21, first materialized at insn 36) pushing
arg webs to r27-r31; mine puts it deepest (r31), args r26-r30. Same web
count (12 callee webs r20-r31 both), identical prologue/epilogue; only the
allocator's numbering differs. Plus a repeated stfsx/fadds store-weave in
the 4 DRAW_QUAD_FOR_FRAME_1 corner blocks.
FAILED variants: scoping polPt/polSize inside the macro (385 insns,
worse); moving texSize decl after polPt/polSize (372 insns, worse).
NOTE: paired-single ops — capstone can't decode psq_*/ps_* insns, so
ndiff.py/disasm_fn.py truncate at the first one; use raw-word diffs.
## iplMemoryCardManager — 8 sub-100 fns all decomposed
- isMoveEnable/isCopyEnable (99.83), update_file_array (99.72),
getComment (98.94): 0 normalized diffs — pure regname webs.
- isBannerEnable (92.31, 26 insns, 7 diffs) + getBlocks (92.0, 25 insns,
6 diffs): single diff = `addi r11,r1,0x20` restgpr-setup placement —
orig emits it at tail (after the full member-chain load), mine hoists
to slot 11. Folded-return variant (`return icons[slot][mFile..]` no
locals) emits identical code — scheduler-internal, decl-form insensitive.
Folded form kept (same output, terser).
- _create_icon (87.55) + create_banner (90.42): `(base+const)+var`
member-chain regroup — orig computes (&mFileCell + memberOffs) +
(slot*row + file*0x15C combined index) with `this+0x10EC`/`this+0x110C`
bases cached in callee regs reused across both GX_TF branches, and
rematerializes &cell.icon per call (`add r3,...` then `add r24,...`
post-bl) while mine pins it; marshal order `li r4,0` before `mr r3`
in orig. FAILED variants: `MCFileCell* cell = &mFileCell[slot][file]`
(90 insns — MWCC folded the pointer, worse); `mFileCell[0] +
slot*0x7F + file` flat-index (89 insns — aggressive CSE, worse).
Same documented mcard2 wall.
All units stay NonMatching — no shims, no flips. Every residual is a
verified allocator/scheduler/emission-policy tie with source-form evidence.
## Round 2 (post-#834 rebase @11c75ddb) — sdChannelTitle emit-parity probe
Investigated per orchestrator hint whether the 152B weak-dtor + dangling-UND
residual was the same dedup-loss family as channelTitle's symbols.txt fix.
VERDICT: NOT a naming issue and NOT a flip blocker.
- orig iplSDChannelTitle.o: ZERO weak FUNC syms. Mine emits 2 unreferenced
weak deleting-dtor bodies: `__dt__FaderSceneBase` (88B @0x1f4) +
`__dt__FrameController` (64B @0xb30), plus 3 zero-reloc UND vtable refs
(`__vt__FrameController`, `__vt__SDButtonEventHandlerBase`, `__vt__LangFile`).
- orig iplFaderSceneBase.o's own view: `~FaderSceneBase` is DECLARED-ONLY
(its `__dt__` is UND there, defined in another TU).
- Tried `__declspec(novtable)` on all 4 classes under
IPL_SD_CHANNEL_TITLE_CPP: removed the 3 dangling `__vt__` UNDs (correct
mechanism — orig stores only most-derived vt) but weak bodies persisted.
- Tried implicit-trivial dtor (removing decl): `__dt__FrameController` emit
STOPPED (64B gone), but `calc()` vtable slot shifted 0x0c→0x08 →
calcFadein/calcFadeout single-word diffs (`lwz r12,0x0c(r12)` vs 0x08)
→ REVERTED everything.
- CRITICAL PROOF these bodies are harmless: main's iplChannelTitle.o
(unit fuzzy 100.0 post-#834) emits the IDENTICAL `__dt__FaderSceneBase`
88B + `__dt__FrameController` 64B weak bodies that orig's .o lacks.
They are linker-stripped dead emission; objdiff fn-pairing ignores them.
=> The 152B was never the blocker. Real residual = flushSaveBeforeExit
98.59 only.
## flushSaveBeforeExit — 6-diff → 2-diff minimum
Window is the `sm` (System globals) base web: orig binds &sm to r3 and
loads +0x94→r4 (store), +0x28→r4 (heap arg), +0x94→r3 (recv); mine binds
r4 and marshals recv before arg. Variants:
- saveHeap decl hoisted above getSDPrevPage store: 2 diffs (base→r5,
+0x28 load moved early) — closer but not orig's order.
- `flushAsync(System::getMem2App())` inline: same 2 diffs, base→r4.
Residual = base-web home + marshal eval order — allocator tie.
## GLGR — ready-set issue-order re-probes
- def-move (sheetGlyphCount/dataBlockCount to tail): 20→99 diffs, REGRESSION.
- `u32 remWorkSpace = ctx->remWorkSpace()` hoist: 20→25, REGRESSION.
- stepSheetFlags after flagsSheetsOff (dependency-chain reorder): identical
20 diffs — MWCC issue-order insensitive to stmt order here. Wall stands.
## DrawFrame / MCManager re-verified
DrawFrame 376=376: flipInfos base pinned r21(orig)/r31(mine) + stfsx weave
— callee-numbering tie unchanged. create_banner/_create_icon: `(base+const)
+var` regroup + recompute-vs-CSE (`add r3,r29,r25` remat per call in orig)
— same documented wall.

File diff suppressed because it is too large Load Diff

View File

@@ -0,0 +1,51 @@
# Run fixed-stack relocation correction
This is a narrow follow-up to #973, not another Run implementation or a new
match. The existing assembly, cache loop, register clearing, labels and handoff
remain intact. No linking flag changes.
## Evidence
The hash-verified original DOL and linked ELF contain Run at 0x8137B3EC,
size 0xAC (43 instructions). Its instructions at 0x8137B468/0x8137B46C establish
SP as 0x81600000. This architectural stack value should not depend on placement
of an unrelated formatted-output routine.
Automatic extraction had attached two relocations to __pformatter + 0x280:
R_PPC_ADDR16_HI at Run+0x7E and R_PPC_ADDR16_LO at Run+0x82. The linked formatter
address happens to make this expression equal 0x81600000 in this target, but
that is a false semantic dependency in the reconstruction.
DTK block_relocations uses instruction addresses, not ELF halfword offsets.
The two exclusions are therefore .text:0x8137B468 and .text:0x8137B46C.
The source removes the unrelated extern and uses the fixed high/low immediate.
No other source instructions or original bytes are replaced.
## Independent validation
- Full 43U all_source, progress, report and DOL check pass.
- Original app and rebuilt DOL SHA1 both remain
26116613f624061ba99c8d1a299aaa6efa85670d.
- Run remains 100%, 43/43 instructions, ctxdiff zero. No new match attribution.
- Whole report is unchanged: no code, data, function, linking or fuzzy delta.
- Pool: all 91 strings identical.
- All 1027 native source objects compared: only BS2Mach .text changes; size
16960 is unchanged. Exactly the two false text relocations are removed.
All other allocated objects and data sections are unchanged.
- All 1027 active extracted objects compared, plus the same obsolete cached
tiManager.o in both trees. Only BS2Mach .text changes: size 16980 unchanged;
only Run's high immediate placeholder bytes at +0x7E/+0x7F become 0x81,0x60.
Exactly the two formatter relocations disappear, with no added relocations.
Defined target symbols and non-text sections stay unchanged.
- Supplemental integer-PPC contract model checks the compiled source for 1,
4 and 131072 cache lines, preserved LR entry, SP 0x81600000, cleared GPRs and
alternating dcbz/dcbf events. The zero-count prefix retains the original CTR
wrap to 0xFFFFFFFF; the full 2^32-iteration path is not executed. This is an
abstract contract test, not actual cache/MMIO or console execution.
- Workflow/literal/jump-table/pool tests: 8/36/27/8 passed.
- Diff check passes.
Assembly was already introduced by #973. The original isolated full-Run
experiment is deliberately not included here. This correction adds no new
assembly instructions, forced register uses, retail .s objects or padding.
Please review the metadata/source constant attribution explicitly.

View File

@@ -0,0 +1,86 @@
# scene4 leaf — decode attempts & evidence
Branch `agent/w1008/scene4` off origin/main@e3457d11.
Scope: iplAddress, iplAddressEdit, iplSDChannelSelect are the only sub-100 units in
the scene/address + sdChannelSelect remainder (iplAddressAddSel, iplAddressInputName,
iplAddressSelect, addressHelpers, addressInputHelpers, agreeSettings, agreeHttp,
iplFaceSelect, iplNCDSetting all verified already 100% on main; no open PR claims).
## Decodes that landed (this leaf)
### Single-case `switch` produces the `b<eq> caseblock; b after` 2-branch guard
`switch (x) { case C:` emits `cmp; beq caseblock; b exit` — MWCC keeps the
dispatch's second `b` because the case-block is a separate labelled region.
Equivalent `if`, `if||`, `if&&`, `goto`-goto, `if/else`-goto all get normalized
to a single fused `beq`/`bne`. Used in:
- `Address::start_drag_event` (892B): outer `switch (mState) { case STATE_NORMAL:`
+ inner `switch (mMode) { case 0:` reproduces orig's two nested 2-branch guards.
Residual: 7 word diffs, pure regname — nigaoe web pinned r26 (orig) vs r29
(mine), miiPane web r29 vs r26. Allocator home-order tie; decl-hoist of
miiPane above the `if` was a no-op (web homes unchanged).
- `Address::onEventDerived` (1080B): TRIG case falls through to ON_POINT
(orig has NO `break`); the `mState == COVER_NORMAL || mState == NORMAL`
guard folds to `beq`+`bne` inside the TRIG case. Residual: 81 word diffs,
all regname — insn sequence identical.
### `addic.` fusion for `ptr = base + const` NULL-tested members
`record->attr.name` tested `!= 0` fuses into `addic.` only when the pointer is
materialized lazily at the compare — decl the local inside the `if`. Hoisted
decls emit `addi`+`cmpwi`. (movePane_onDrag `name` web.)
## Walls re-verified (evidence)
### `Address::movePane_onDrag` 92.45 (620 orig vs 612 mine)
- Orig emits `bne next; b exit` 2-branch guards for both `textBox != NULL` and
`name != NULL` (the `b` skips to the shared `width += 0.01f` block).
- This is switch-dispatch-shaped, but every legal polarity fails:
`switch(ptr) { case NULL: break; default: body }` bool-materializes
(`addic/subfe/cmpwi 1` — +3 insns); `switch (p != NULL) { case true:` also
bool-materializes; `do { if (!x) break; } while (0)` folds to single `beq`.
- Orig computes the record math (`lwz 0x70 / mulli 5 / lwz 0x274 / mulli 0x140`)
before the textBox branch — scheduler interleave around the guard.
- Residual also regname r27↔r29 (textBox web) + loop `name` web.
### `FriendListCache::update` 95.5 (160 orig vs 156 mine)
Orig remats `mInfos + index*0x140` per use site (memset, wcsncpy,
updateFriendInfo = 3 `mulli`+`add` pairs). Mine CSEs into callee r31.
`getInfo(index)` vs `mInfos[index]` vs `(int)index` cast — all VN-identical,
still CSE. Remat-vs-pin in the unusual remat-is-orig direction; no lever.
### `AddressEdit::create` 97.45 (3280B)
Two distinct residuals:
- Wholesale r30↔r31 web-order swap: orig pins string-table const web in r31
and `boardFile` (getScene+0xd20) in r30; mine the reverse. Webs born in the
same source order — allocator processing order tie.
- `friendText` (this+0x2b4 / this+0x2cc) pin: orig materializes
`addi r25, r29, 0x2b4` at decl point between `initFrame()` and `restart()`,
reassigns `addi r25, r29, 0x2cc` mid-block — the reg+const web is pinned.
Mine remats `addi r5, r29, 0x2b4` at each marshal. Variants tried: non-const
`wchar_t*`, decl moved before `animator =` — both no-op. Same this+const
pin family as get_friendinfo.
### `AddressEdit::get_friendinfo` 88.04 (336 orig vs 328 mine)
Orig: this→r28, `this+0x2b4` (mName) pinned in r29 materialized before the
FindPaneByName marshal — 2 extra insns vs mine's marshal-site remats, and one
more callee web (savegpr_28 vs my _29 window differs). Same `this+const`
pin-vs-remat wall as create's friendText.
### `AddressEdit::update_friendinfo` 99.42 (152B)
Single 2-insn scheduling swap: `addi r4,r30,0x2b4` vs `addi r3,r31,8` order
in the wcsncpy marshal.
### iplSDChannelSelect 99.58 — 7 fns
- `create` 97.74: regname r29↔r30 swap + marshal interleaves around `bl`s
(scheduler family); insn count equal.
- `collectTitles*` ×4 (97.29–97.83): documented ties — callee-web rotations
around the same ES_GetTitleCount/ES_GetTitles marshal blocks.
- `flushSaveDataAndMountSD` 99.66 (140B): marshal order tie.
- `setChannelScissor` 93.47 (932B): fp/scheduling family (DrawFrame-class).
## Notes
- ndiff.py NORMALIZES register names — "diffs 0" can hide regname-only
diffs. Always confirm with raw byte-compare of `.text` (readelf -sW for
addr/size, -SW for file offset).
- `objdiff-cli diff` is interactive-only (fails headless); use disasm_fn.py
+ byte-compare.