|
|
|
|
@@ -858,3 +858,587 @@ Final rx10b clean validation: gate.py ran without --quick on wad, sd_drv and pdm
|
|
|
|
|
[libs/RVL_SDK/src/fa/driver/sd_drv] objdiff: code 10068/11760 data 3592/3592 functions 23/26; instruction-exact 23/26, unchanged.
|
|
|
|
|
[libs/RVL_SDK/src/fa/pdm_partition] objdiff: code 3380/3716; no data; functions 19/20; instruction-exact 19/20, unchanged.
|
|
|
|
|
Regressions 0; net forbidden patterns 0; readability warnings 0. GATE PASS. Full log: build/rx10b/final-gate.log; fresh report: build/rx10b/final-report.json.
|
|
|
|
|
|
|
|
|
|
## rx10c: target-preserved dead operations and WAD link audit
|
|
|
|
|
|
|
|
|
|
Fresh branch agent/w1005/rx10c at a06cbaa8. PR1196 landed the two rx10b exact WAD functions; production starts clean, WAD 39/40. Scope is WADBackupEx then partition validation and the three SD functions. Reuse only the selected private function bodies, preserving landed code. Export targets with unreachable-block elimination disabled to inspect source operations hidden by the default Ghidra export. The Matching flip is authorized only after WAD reaches 40/40.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-replay`. Fresh current-unit replay of the 99.12429 private function body only. Verify the four source-shape deficits and preserve all 39 landed exact WAD functions. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-read`. Keep the real installed-content count read at a pointer helper boundary. The target stores the success status before loading/testing count; unlike prior by-value validation, this preserves the count-memory dependency through inline expansion. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-zero`. Keep the real installed-content count read at a pointer helper boundary. The target stores the success status before loading/testing count; unlike prior by-value validation, this preserves the count-memory dependency through inline expansion. Result: 99.100750%, instructions 1059/1062, positional differences 801.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-size-helper`. Keep the real installed-content count read at a pointer helper boundary. The target stores the success status before loading/testing count; unlike prior by-value validation, this preserves the count-memory dependency through inline expansion. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-array-block`. Probe real serialized header storage/view forms against the three separate base/member addi pairs. Same 128-byte extent, 64-byte alignment and fields. Integer-view variant is diagnostic only; prefer plain typed member access for any retained source. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-array-member`. Probe real serialized header storage/view forms against the three separate base/member addi pairs. Same 128-byte extent, 64-byte alignment and fields. Integer-view variant is diagnostic only; prefer plain typed member access for any retained source. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-integer-view`. Probe real serialized header storage/view forms against the three separate base/member addi pairs. Same 128-byte extent, 64-byte alignment and fields. Integer-view variant is diagnostic only; prefer plain typed member access for any retained source. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-void-view-helper`. Probe real serialized header storage/view forms against the three separate base/member addi pairs. Same 128-byte extent, 64-byte alignment and fields. Integer-view variant is diagnostic only; prefer plain typed member access for any retained source. Result: 99.114876%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-status-else`. Compare explicit status validation and dead success assignments with target 0x1e50-0x1e70. Each path computes the same status and preserves calls; do not infer exactness from total instruction count. Result: 96.709040%, instructions 1059/1062, positional differences 938.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-after-status`. Compare explicit status validation and dead success assignments with target 0x1e50-0x1e70. Each path computes the same status and preserves calls; do not infer exactness from total instruction count. Result: 98.742935%, instructions 1058/1062, positional differences 925.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-check-count-result`. Compare explicit status validation and dead success assignments with target 0x1e50-0x1e70. Each path computes the same status and preserves calls; do not infer exactness from total instruction count. Result: 99.030130%, instructions 1061/1062, positional differences 894.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-validation-comma`. Compare explicit status validation and dead success assignments with target 0x1e50-0x1e70. Each path computes the same status and preserves calls; do not infer exactness from total instruction count. Result: 98.742935%, instructions 1058/1062, positional differences 925.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-validation-conditional`. Compare explicit status validation and dead success assignments with target 0x1e50-0x1e70. Each path computes the same status and preserves calls; do not infer exactness from total instruction count. Result: 98.639360%, instructions 1059/1062, positional differences 805.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-nested-first-validation`. Compare explicit status validation and dead success assignments with target 0x1e50-0x1e70. Each path computes the same status and preserves calls; do not infer exactness from total instruction count. Result: 98.822040%, instructions 1059/1062, positional differences 805.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-diagnostic-opt_dead_assignments`. Lever 18e diagnostic only, never retain pragma. Identify whether the missing status store and folded member addresses arise in dead assignment, propagation or common-subexpression optimization. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-diagnostic-opt_propagation`. Lever 18e diagnostic only, never retain pragma. Identify whether the missing status store and folded member addresses arise in dead assignment, propagation or common-subexpression optimization. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-diagnostic-opt_common_subs`. Lever 18e diagnostic only, never retain pragma. Identify whether the missing status store and folded member addresses arise in dead assignment, propagation or common-subexpression optimization. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-null-init`. Test a reused real header view at initialization and use. Target member addresses retain two addis despite a stack-known base. Dead initializer tests are diagnostic unless assembly proves that source operation; integer offsetof form is diagnostic only. Result: 97.662900%, instructions 1059/1062, positional differences 940.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-file-view-reuse`. Test a reused real header view at initialization and use. Target member addresses retain two addis despite a stack-known base. Dead initializer tests are diagnostic unless assembly proves that source operation; integer offsetof form is diagnostic only. Result: 98.971750%, instructions 1059/1062, positional differences 910.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-buffer-reuse`. Test a reused real header view at initialization and use. Target member addresses retain two addis despite a stack-known base. Dead initializer tests are diagnostic unless assembly proves that source operation; integer offsetof form is diagnostic only. Result: 97.662900%, instructions 1059/1062, positional differences 940.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-scoped-null-init`. Test a reused real header view at initialization and use. Target member addresses retain two addis despite a stack-known base. Dead initializer tests are diagnostic unless assembly proves that source operation; integer offsetof form is diagnostic only. Result: 98.971750%, instructions 1059/1062, positional differences 910.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-offset-integer-diagnostic`. Test a reused real header view at initialization and use. Target member addresses retain two addis despite a stack-known base. Dead initializer tests are diagnostic unless assembly proves that source operation; integer offsetof form is diagnostic only. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-state-return-helper`. Factor the actual post-unmount state reset into a scalar-returning inline helper. Target flag load uses r3 before the returned zero occupies r3; test that source lifetime boundary without new state or sequencing effects. Result: 92.631580%, instructions 57/57, positional differences 4.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-state-helper-local-zero`. Factor the actual post-unmount state reset into a scalar-returning inline helper. Target flag load uses r3 before the returned zero occupies r3; test that source lifetime boundary without new state or sequencing effects. Result: Compile failed; excluded from attempt count.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-state-helper-global`. Factor the actual post-unmount state reset into a scalar-returning inline helper. Target flag load uses r3 before the returned zero occupies r3; test that source lifetime boundary without new state or sequencing effects. Result: 92.631580%, instructions 57/57, positional differences 4.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-disk-identity`. Inspect null-validation lowering against target li zero plus unsigned register compare. Pointer subtraction from zero is diagnostic only and will not be retained; the identity helper and unsigned-address forms test operand provenance. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-disk-offset-zero`. Inspect null-validation lowering against target li zero plus unsigned register compare. Pointer subtraction from zero is diagnostic only and will not be retained; the identity helper and unsigned-address forms test operand provenance. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-disk-unsigned-null`. Inspect null-validation lowering against target li zero plus unsigned register compare. Pointer subtraction from zero is diagnostic only and will not be retained; the identity helper and unsigned-address forms test operand provenance. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-media-getter`. Keep the actual media query at an inline helper boundary after installing the mounted device, matching the target store-before-load ordering. No extra calls or changed state are intended; audit before allocation. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-device-install`. Keep the actual media query at an inline helper boundary after installing the mounted device, matching the target store-before-load ordering. No extra calls or changed state are intended; audit before allocation. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-member-offset`. Diagnostic serialization-address helper boundary. The target materializes the buffer base then the member address at each call. Same named field offsets, extent and alignment; retain only a natural typed form if this identifies the missing phase. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-member-offset-int`. Diagnostic serialization-address helper boundary. The target materializes the buffer base then the member address at each call. Same named field offsets, extent and alignment; retain only a natural typed form if this identifies the missing phase. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-offset-zero`. Diagnostic serialization-address helper boundary. The target materializes the buffer base then the member address at each call. Same named field offsets, extent and alignment; retain only a natural typed form if this identifies the missing phase. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-required`. Diagnostic serialization-address helper boundary. The target materializes the buffer base then the member address at each call. Same named field offsets, extent and alignment; retain only a natural typed form if this identifies the missing phase. Result: 98.539550%, instructions 1067/1062, positional differences 920.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-end`. Diagnostic serialization-address helper boundary. The target materializes the buffer base then the member address at each call. Same named field offsets, extent and alignment; retain only a natural typed form if this identifies the missing phase. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-result-long`. Test actual 32-bit scalar status/count type provenance at the missing dead-store boundary. Keep object layout and bit patterns; unsigned-result variant is diagnostic pending full signed-return and out-parameter review. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-result-unsigned`. Test actual 32-bit scalar status/count type provenance at the missing dead-store boundary. Keep object layout and bit patterns; unsigned-result variant is diagnostic pending full signed-return and out-parameter review. Result: 98.973630%, instructions 1059/1062, positional differences 805.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-long`. Test actual 32-bit scalar status/count type provenance at the missing dead-store boundary. Keep object layout and bit patterns; unsigned-result variant is diagnostic pending full signed-return and out-parameter review. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-status-long`. Test actual 32-bit scalar status/count type provenance at the missing dead-store boundary. Keep object layout and bit patterns; unsigned-result variant is diagnostic pending full signed-return and out-parameter review. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-both-zero`. Target retains a dead start-output initialization after reading both partition fields. Test genuine output initialization/decoded-field lifetimes while preserving values, bounds and every input byte. Result: 89.845240%, instructions 84/84, positional differences 19.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-count-zero-after-start`. Target retains a dead start-output initialization after reading both partition fields. Test genuine output initialization/decoded-field lifetimes while preserving values, bounds and every input byte. Result: 89.845240%, instructions 84/84, positional differences 19.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-zero-after-decode`. Target retains a dead start-output initialization after reading both partition fields. Test genuine output initialization/decoded-field lifetimes while preserving values, bounds and every input byte. Result: 83.654760%, instructions 83/84, positional differences 72.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-combined-entry`. Decode the two real partition-entry fields through one output helper. Target prefetches count bytes before the dead start store; a combined helper makes that boundary explicit without inventing a carrier object. Result: 89.845240%, instructions 84/84, positional differences 19.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-combined-entry-temporaries`. Decode the two real partition-entry fields through one output helper. Target prefetches count bytes before the dead start store; a combined helper makes that boundary explicit without inventing a carrier object. Result: 83.654760%, instructions 83/84, positional differences 72.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-combined-entry-after-start`. Decode the two real partition-entry fields through one output helper. Target prefetches count bytes before the dead start store; a combined helper makes that boundary explicit without inventing a carrier object. Result: 83.654760%, instructions 83/84, positional differences 72.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-query-normalize`. Reconstruct the complete card-content query/validation boundary, including its success normalization. Earlier helpers isolated only the count check; target stores zero between the query-error path and count-error path. Result: 98.634650%, instructions 1061/1062, positional differences 903.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-and-query`. Reconstruct the complete card-content query/validation boundary, including its success normalization. Earlier helpers isolated only the count check; target stores zero between the query-error path and count-error path. Result: 98.352165%, instructions 1060/1062, positional differences 920.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-contents-operation`. Reconstruct the complete card-content query/validation boundary, including its success normalization. Earlier helpers isolated only the count check; target stores zero between the query-error path and count-error path. Result: 98.724106%, instructions 1054/1062, positional differences 933.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-reuse-transfer-buffer`. Reuse a truthful generic buffer local for disjoint serialization/transfer phases with real uses on both assignments. Target header field calls retain a base copy, which can survive when a named source local has multiple live definitions. Header field values and I/O remain unchanged. Result: 98.971750%, instructions 1059/1062, positional differences 910.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-reuse-file-header-buffer`. Reuse a truthful generic buffer local for disjoint serialization/transfer phases with real uses on both assignments. Target header field calls retain a base copy, which can survive when a named source local has multiple live definitions. Header field values and I/O remain unchanged. Result: 98.971750%, instructions 1059/1062, positional differences 910.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-reuse-certificate-buffer`. Reuse a truthful generic buffer local for disjoint serialization/transfer phases with real uses on both assignments. Target header field calls retain a base copy, which can survive when a named source local has multiple live definitions. Header field values and I/O remain unchanged. Result: 98.971750%, instructions 1059/1062, positional differences 910.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-ignored-status`. Target skips reserved-helper null/status checks yet retains dead error reports. Test the original-author omission: ignore the helper result and recheck the preceding successful block-write status. Its sole call has a nonnull global buffer; error behavior is unchanged for all actual calls. Prior-copy form is diagnostic only. Result: 95.653150%, instructions 228/222, positional differences 125.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-ignored-status`. Target skips reserved-helper null/status checks yet retains dead error reports. Test the original-author omission: ignore the helper result and recheck the preceding successful block-write status. Its sole call has a nonnull global buffer; error behavior is unchanged for all actual calls. Prior-copy form is diagnostic only. Result: 97.364870%, instructions 227/222, positional differences 117.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-void-nonnull-stale-status`. Target skips reserved-helper null/status checks yet retains dead error reports. Test the original-author omission: ignore the helper result and recheck the preceding successful block-write status. Its sole call has a nonnull global buffer; error behavior is unchanged for all actual calls. Prior-copy form is diagnostic only. Result: 97.364870%, instructions 227/222, positional differences 117.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-result-prior-copy`. Target skips reserved-helper null/status checks yet retains dead error reports. Test the original-author omission: ignore the helper result and recheck the preceding successful block-write status. Its sole call has a nonnull global buffer; error behavior is unchanged for all actual calls. Prior-copy form is diagnostic only. Result: 97.364870%, instructions 227/222, positional differences 117.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-clear-return-pointer`. Target skips reserved-helper null/status checks yet retains dead error reports. Test the original-author omission: ignore the helper result and recheck the preceding successful block-write status. Its sole call has a nonnull global buffer; error behavior is unchanged for all actual calls. Prior-copy form is diagnostic only. Result: 96.058556%, instructions 227/222, positional differences 117.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-output-nonnull`. Trace the real reserved-sector success value through inline expansion and the caller error path. The helper has one proven nonnull global caller. Parameter-passthrough variants are diagnostic, not landing candidates; output status and ordinary local status preserve actual behavior. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-result-parameter`. Trace the real reserved-sector success value through inline expansion and the caller error path. The helper has one proven nonnull global caller. Parameter-passthrough variants are diagnostic, not landing candidates; output status and ordinary local status preserve actual behavior. Result: 97.364870%, instructions 227/222, positional differences 117.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-const-result-parameter`. Trace the real reserved-sector success value through inline expansion and the caller error path. The helper has one proven nonnull global caller. Parameter-passthrough variants are diagnostic, not landing candidates; output status and ordinary local status preserve actual behavior. Result: 97.364870%, instructions 227/222, positional differences 117.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-initial-success`. Trace the real reserved-sector success value through inline expansion and the caller error path. The helper has one proven nonnull global caller. Parameter-passthrough variants are diagnostic, not landing candidates; output status and ordinary local status preserve actual behavior. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-duplicate-success`. Trace the real reserved-sector success value through inline expansion and the caller error path. The helper has one proven nonnull global caller. Parameter-passthrough variants are diagnostic, not landing candidates; output status and ordinary local status preserve actual behavior. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-result-short`. Trace the real reserved-sector success value through inline expansion and the caller error path. The helper has one proven nonnull global caller. Parameter-passthrough variants are diagnostic, not landing candidates; output status and ordinary local status preserve actual behavior. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-result-byte`. Trace the real reserved-sector success value through inline expansion and the caller error path. The helper has one proven nonnull global caller. Parameter-passthrough variants are diagnostic, not landing candidates; output status and ordinary local status preserve actual behavior. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-success-goto`. Combine the proven nonnull helper call with explicit success/error continuations. Target has unconditional success branches over an unreachable pair of report calls; compare when the frontend versus backend discards those blocks. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-error-label`. Combine the proven nonnull helper call with explicit success/error continuations. Target has unconditional success branches over an unreachable pair of report calls; compare when the frontend versus backend discards those blocks. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-early-helper-return`. Combine the proven nonnull helper call with explicit success/error continuations. Target has unconditional success branches over an unreachable pair of report calls; compare when the frontend versus backend discards those blocks. Result: 96.914410%, instructions 228/222, positional differences 110.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-switch-success-first`. Switch lowering defers case elimination and can retain the target dead report blocks after the reserved-sector success value is propagated. Preserve report order, return values and the single real buffer operation. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-switch-error-first`. Switch lowering defers case elimination and can retain the target dead report blocks after the reserved-sector success value is propagated. Preserve report order, return values and the single real buffer operation. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-switch-success-first`. Switch lowering defers case elimination and can retain the target dead report blocks after the reserved-sector success value is propagated. Preserve report order, return values and the single real buffer operation. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-switch-error-first`. Switch lowering defers case elimination and can retain the target dead report blocks after the reserved-sector success value is propagated. Preserve report order, return values and the single real buffer operation. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-bounded-signed-pairs`. Keep the bounded lower-byte arithmetic signed until it meets the unsigned high byte, or decode genuine 16-bit halves. Every signed intermediate is at most 0xffffff and cannot overflow. Test regrouping before allocation without changing decoded bits. Result: 87.476190%, instructions 86/84, positional differences 84.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-bounded-signed-low`. Keep the bounded lower-byte arithmetic signed until it meets the unsigned high byte, or decode genuine 16-bit halves. Every signed intermediate is at most 0xffffff and cannot overflow. Test regrouping before allocation without changing decoded bits. Result: 89.845240%, instructions 84/84, positional differences 19.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-low-half-helper`. Keep the bounded lower-byte arithmetic signed until it meets the unsigned high byte, or decode genuine 16-bit halves. Every signed intermediate is at most 0xffffff and cannot overflow. Test regrouping before allocation without changing decoded bits. Result: 87.154760%, instructions 88/84, positional differences 87.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-high-half-helper`. Keep the bounded lower-byte arithmetic signed until it meets the unsigned high byte, or decode genuine 16-bit halves. Every signed intermediate is at most 0xffffff and cannot overflow. Test regrouping before allocation without changing decoded bits. Result: 87.345240%, instructions 86/84, positional differences 67.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-named-status-output`. Test provenance of genuine escaped status/count storage. Target writes success before the count load and success join; named output pointers and pointer-taking validation change alias visibility without adding state, changing values or introducing volatile accesses. Result: 99.112990%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-status-helper-pointer-count`. Test provenance of genuine escaped status/count storage. Target writes success before the count load and success join; named output pointers and pointer-taking validation change alias visibility without adding state, changing values or introducing volatile accesses. Result: 96.709040%, instructions 1059/1062, positional differences 938.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-pointer`. Test provenance of genuine escaped status/count storage. Target writes success before the count load and success join; named output pointers and pointer-taking validation change alias visibility without adding state, changing values or introducing volatile accesses. Result: 98.870056%, instructions 1059/1062, positional differences 909.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-void-view`. Test provenance of genuine escaped status/count storage. Target writes success before the count load and success join; named output pointers and pointer-taking validation change alias visibility without adding state, changing values or introducing volatile accesses. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-status-zero-last`. Tie the success value to the actual final state reset and test its lifetime before allocation. All paths return the same status and reset identical fields; assignment-return forms are diagnostic unless their ordering and readability match the target. Result: 92.631580%, instructions 57/57, positional differences 4.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-media-assignment-return`. Tie the success value to the actual final state reset and test its lifetime before allocation. All paths return the same status and reset identical fields; assignment-return forms are diagnostic unless their ordering and readability match the target. Result: 89.192986%, instructions 57/57, positional differences 6.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-return-reset-value`. Tie the success value to the actual final state reset and test its lifetime before allocation. All paths return the same status and reset identical fields; assignment-return forms are diagnostic unless their ordering and readability match the target. Result: 92.631580%, instructions 57/57, positional differences 4.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-read-after-zero`. Test an actual status/count read whose temporary is dead or reused at the success join. The target retains both zero stores around count validation; copy propagation after dead-store analysis may preserve that source operation. No additional side effects or changed outcomes. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-dead-read-after-zero`. Test an actual status/count read whose temporary is dead or reused at the success join. The target retains both zero stores around count validation; copy propagation after dead-store analysis may preserve that source operation. No additional side effects or changed outcomes. Result: 98.742935%, instructions 1058/1062, positional differences 925.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-read-at-join`. Test an actual status/count read whose temporary is dead or reused at the success join. The target retains both zero stores around count validation; copy propagation after dead-store analysis may preserve that source operation. No additional side effects or changed outcomes. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-dead-read-at-join`. Test an actual status/count read whose temporary is dead or reused at the success join. The target retains both zero stores around count validation; copy propagation after dead-store analysis may preserve that source operation. No additional side effects or changed outcomes. Result: 98.742935%, instructions 1058/1062, positional differences 925.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-chained-status-zero`. Test an actual status/count read whose temporary is dead or reused at the success join. The target retains both zero stores around count validation; copy propagation after dead-store analysis may preserve that source operation. No additional side effects or changed outcomes. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-signed-count-read`. Test an actual status/count read whose temporary is dead or reused at the success join. The target retains both zero stores around count validation; copy propagation after dead-store analysis may preserve that source operation. No additional side effects or changed outcomes. Result: 98.742935%, instructions 1058/1062, positional differences 925.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-checked-header-const-pointer`. Header-view validation boundary diagnostic. A dynamic null guard recreates separate base/member adds but leaves extra tests. Check whether a genuine serialized-buffer length check or const pointer parameter lets the frontend prove the guard while preserving address formation. The constant-size guard is diagnostic only. Result: 98.540490%, instructions 1067/1062, positional differences 920.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-checked-header-stack-size`. Header-view validation boundary diagnostic. A dynamic null guard recreates separate base/member adds but leaves extra tests. Check whether a genuine serialized-buffer length check or const pointer parameter lets the frontend prove the guard while preserving address formation. The constant-size guard is diagnostic only. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-checked-header-member-range`. Header-view validation boundary diagnostic. A dynamic null guard recreates separate base/member adds but leaves extra tests. Check whether a genuine serialized-buffer length check or const pointer parameter lets the frontend prove the guard while preserving address formation. The constant-size guard is diagnostic only. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-checked-header-unsigned-null`. Header-view validation boundary diagnostic. A dynamic null guard recreates separate base/member adds but leaves extra tests. Check whether a genuine serialized-buffer length check or const pointer parameter lets the frontend prove the guard while preserving address formation. The constant-size guard is diagnostic only. Result: 98.540490%, instructions 1067/1062, positional differences 920.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-helper-after-caller`. Test when the private helper is optimized relative to call-site expansion. A later definition or switch-based null case may expose the known nonnull buffer only during backend CFG folding, preserving target dead diagnostics. Real helper semantics and caller results remain unchanged. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-helper-nonnull-after-caller`. Test when the private helper is optimized relative to call-site expansion. A later definition or switch-based null case may expose the known nonnull buffer only during backend CFG folding, preserving target dead diagnostics. Real helper semantics and caller results remain unchanged. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-helper-switch-null`. Test when the private helper is optimized relative to call-site expansion. A later definition or switch-based null case may expose the known nonnull buffer only during backend CFG folding, preserving target dead diagnostics. Real helper semantics and caller results remain unchanged. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-const-parameter`. Test the first null-validation value separately from branch layout. Target materializes a null register before cmplw; compare pointer-constant and real status-validation boundaries without changing validation outcomes. A static constant is diagnostic unless it emits no additional data. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-constant-null`. Test the first null-validation value separately from branch layout. Target materializes a null register before cmplw; compare pointer-constant and real status-validation boundaries without changing validation outcomes. A static constant is diagnostic unless it emits no additional data. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-static-constant-null`. Test the first null-validation value separately from branch layout. Target materializes a null register before cmplw; compare pointer-constant and real status-validation boundaries without changing validation outcomes. A static constant is diagnostic unless it emits no additional data. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-guard-status`. Test the first null-validation value separately from branch layout. Target materializes a null register before cmplw; compare pointer-constant and real status-validation boundaries without changing validation outcomes. A static constant is diagnostic unless it emits no additional data. Result: 93.159720%, instructions 146/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-unsigned-guard-status`. Test the first null-validation value separately from branch layout. Target materializes a null register before cmplw; compare pointer-constant and real status-validation boundaries without changing validation outcomes. A static constant is diagnostic unless it emits no additional data. Result: 93.159720%, instructions 146/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-initialize-operation`. Move the complete real backup-header initialization and device-id query into one inline operation. Unlike a field-address wrapper, the header parameter has all field stores as uses before its member-address call; compare address propagation in backend-00. Result: 99.035780%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-content-mask-operation`. Move the complete conditional content-mask operation across the header-parameter boundary. Same predicate and copy; target emits buffer-base addi only in the copy arm. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-device-mac-operation`. Move the transfer-id query and conditional MAC copy together. A real call separates pointer binding from member-address construction; target retains the split base/member adds after that call. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-separate-query-status`. Separate the call status from the escaped final result, reusing one truthful status local through query normalization. backend-00 contains the target first zero store, but backend-01 removes it because both successors overwrite result. A normalized status edge tests when that overwrite proof becomes available. Result: 99.124290%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-normalize-query-status`. Separate the call status from the escaped final result, reusing one truthful status local through query normalization. backend-00 contains the target first zero store, but backend-01 removes it because both successors overwrite result. A normalized status edge tests when that overwrite proof becomes available. Result: 98.441620%, instructions 1061/1062, positional differences 903.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-normalized-success-check`. Separate the call status from the escaped final result, reusing one truthful status local through query normalization. backend-00 contains the target first zero store, but backend-01 removes it because both successors overwrite result. A normalized status edge tests when that overwrite proof becomes available. Result: 99.280600%, instructions 1059/1062, positional differences 798.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-query-count-shared-error`. Separate the call status from the escaped final result, reusing one truthful status local through query normalization. backend-00 contains the target first zero store, but backend-01 removes it because both successors overwrite result. A normalized status edge tests when that overwrite proof becomes available. Result: 98.865350%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-existing-flags-query`. Use the existing flags accessor or a plain scalar flag-clear helper for the final transition. Target old flags use r3 before the zero return; inspect whether an inline returned value changes the load/mask lifetime and scheduling without additional state. Result: 92.631580%, instructions 57/57, positional differences 4.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-flags-status-helper`. Use the existing flags accessor or a plain scalar flag-clear helper for the final transition. Target old flags use r3 before the zero return; inspect whether an inline returned value changes the load/mask lifetime and scheduling without additional state. Result: 92.631580%, instructions 57/57, positional differences 4.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-const-flags-helper`. Use the existing flags accessor or a plain scalar flag-clear helper for the final transition. Target old flags use r3 before the zero return; inspect whether an inline returned value changes the load/mask lifetime and scheduling without additional state. Result: 92.631580%, instructions 57/57, positional differences 4.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-finalize-mask-parameter-helper`. Use the existing flags accessor or a plain scalar flag-clear helper for the final transition. Target old flags use r3 before the zero return; inspect whether an inline returned value changes the load/mask lifetime and scheduling without additional state. Result: 92.631580%, instructions 57/57, positional differences 4.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-array-address`. Test address-of-object versus array-decay provenance at the actual reserved-sector call. Both designate the same 512-byte buffer; target null/result paths disappear only after enough code exists to leave the dead diagnostics. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-element-address`. Test address-of-object versus array-decay provenance at the actual reserved-sector call. Both designate the same 512-byte buffer; target null/result paths disappear only after enough code exists to leave the dead diagnostics. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-void-array-address`. Test address-of-object versus array-decay provenance at the actual reserved-sector call. Both designate the same 512-byte buffer; target null/result paths disappear only after enough code exists to leave the dead diagnostics. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-whole-buffer-pointer`. Use the real fixed-size sector-buffer type at the private helper boundary. Pointer-to-array preserves the same address, extent and null semantics while changing frontend object-address knowledge. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-checked-negative`. The helper returns only zero or a negative error. Test equivalent signed error predicates so constant/range folding reaches the target unreachable-report layout at a different backend phase. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-checked-minus-one`. The helper returns only zero or a negative error. Test equivalent signed error predicates so constant/range folding reaches the target unreachable-report layout at a different backend phase. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-checked-success-nonnegative`. The helper returns only zero or a negative error. Test equivalent signed error predicates so constant/range folding reaches the target unreachable-report layout at a different backend phase. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-negative`. The helper returns only zero or a negative error. Test equivalent signed error predicates so constant/range folding reaches the target unreachable-report layout at a different backend phase. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-minus-one`. The helper returns only zero or a negative error. Test equivalent signed error predicates so constant/range folding reaches the target unreachable-report layout at a different backend phase. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nonnull-success-nonnegative`. The helper returns only zero or a negative error. Test equivalent signed error predicates so constant/range folding reaches the target unreachable-report layout at a different backend phase. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-split-file-remaining`. Target file phase uses r30 for name length and file bytes remaining, r28 for saved stream position, and r23 for transfer/padding size. The private source incorrectly extends the content-remaining local through the file phase and uses readSize for padding. Split or reuse only nonoverlapping real values, with truthful names, to recover target interference before structural gaps are addressed. Result: 99.166664%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-padding-size-reuse`. Target file phase uses r30 for name length and file bytes remaining, r28 for saved stream position, and r23 for transfer/padding size. The private source incorrectly extends the content-remaining local through the file phase and uses readSize for padding. Split or reuse only nonoverlapping real values, with truthful names, to recover target interference before structural gaps are addressed. Result: 99.161960%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-split-and-padding`. Target file phase uses r30 for name length and file bytes remaining, r28 for saved stream position, and r23 for transfer/padding size. The private source incorrectly extends the content-remaining local through the file phase and uses readSize for padding. Split or reuse only nonoverlapping real values, with truthful names, to recover target interference before structural gaps are addressed. Result: 99.180790%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-name-and-remaining`. Target file phase uses r30 for name length and file bytes remaining, r28 for saved stream position, and r23 for transfer/padding size. The private source incorrectly extends the content-remaining local through the file phase and uses readSize for padding. Split or reuse only nonoverlapping real values, with truthful names, to recover target interference before structural gaps are addressed. Result: 99.209040%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-aligned-count`. Storage diagnostic for the first dead success store. Count is an ES output at target stack offset 0x60. Check alignment and volatile-read barriers to locate the optimizer dependency; volatile variants will not be retained without independent target proof of that qualification. Result: 98.953860%, instructions 1058/1062, positional differences 943.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-volatile-count-diagnostic`. Storage diagnostic for the first dead success store. Count is an ES output at target stack offset 0x60. Check alignment and volatile-read barriers to locate the optimizer dependency; volatile variants will not be retained without independent target proof of that qualification. Result: Compile failed; excluded from attempt count.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-volatile-status-diagnostic`. Storage diagnostic for the first dead success store. Count is an ES output at target stack offset 0x60. Check alignment and volatile-read barriers to locate the optimizer dependency; volatile variants will not be retained without independent target proof of that qualification. Result: 98.838980%, instructions 1061/1062, positional differences 921.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-file-phase-combined`. Combine target-proven file-phase lifetime corrections: byteCount holds the path length then remaining file bytes, and size holds transfer then padding bytes. The content remaining lifetime now ends before files. Preserve all four known structural gaps for separate investigation. Result: 99.223170%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-enum-status`. Test ordinary error-status typing after the target file lifetimes are reconstructed. Enum and SDK status remain signed 32-bit values; no storage or API change. Result: 99.223170%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-sdk-status`. Test ordinary error-status typing after the target file lifetimes are reconstructed. Enum and SDK status remain signed 32-bit values; no storage or API change. Result: 99.223170%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-result-normalization-recheck`. Normalize the special ES failure separately from branching on that failure. This tests whether CFG simplification after dead-store analysis retains the otherwise dead success store. All outcomes and calls remain identical; duplicated predicate forms require readability review before retention. Result: 98.624290%, instructions 1062/1062, positional differences 251.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-result-default-recheck`. Normalize the special ES failure separately from branching on that failure. This tests whether CFG simplification after dead-store analysis retains the otherwise dead success store. All outcomes and calls remain identical; duplicated predicate forms require readability review before retention. Result: 98.634650%, instructions 1061/1062, positional differences 905.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-status-normalization-recheck`. Normalize the special ES failure separately from branching on that failure. This tests whether CFG simplification after dead-store analysis retains the otherwise dead success store. All outcomes and calls remain identical; duplicated predicate forms require readability review before retention. Result: 98.540490%, instructions 1061/1062, positional differences 903.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-pack-4`. Diagnostic only: model the real serialized header with natural 4-byte versus 8-byte record alignment. Offsets and 112-byte header/128-byte block extents are identical. Test whether field-address lowering depends on record alignment rather than invented register state; prefer the existing shared header definition in retained code. Result: 99.223170%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-pack-8`. Diagnostic only: model the real serialized header with natural 4-byte versus 8-byte record alignment. Offsets and 112-byte header/128-byte block extents are identical. Test whether field-address lowering depends on record alignment rather than invented register state; prefer the existing shared header definition in retained code. Result: 99.223170%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-null-after-signature`. Diagnostic null-check placement around the real signature stores. The sole call uses the global sector buffer; compare whether nonnull knowledge appears late enough to preserve dead reports. The post-dereference/member guard variants are not retained without target/source-quality proof. Result: 95.509010%, instructions 231/222, positional differences 123.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-null-member-address-guard`. Diagnostic null-check placement around the real signature stores. The sole call uses the global sector buffer; compare whether nonnull knowledge appears late enough to preserve dead reports. The post-dereference/member guard variants are not retained without target/source-quality proof. Result: 94.972980%, instructions 230/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-null-typed-view-guard`. Diagnostic null-check placement around the real signature stores. The sole call uses the global sector buffer; compare whether nonnull knowledge appears late enough to preserve dead reports. The post-dereference/member guard variants are not retained without target/source-quality proof. Result: 95.202705%, instructions 229/222, positional differences 122.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-buffer-metadata-header`. Reuse the actual metadata-view pointer for later disjoint serialization-buffer roles under the neutral name buffer. This may stop the earlier metadata view from replacing later owner-pointer uses, and tests whether genuine repeated pointer definitions preserve header base copies. All view extents and field accesses remain valid. Result: 99.084750%, instructions 1059/1062, positional differences 910.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-buffer-metadata-header-files`. Reuse the actual metadata-view pointer for later disjoint serialization-buffer roles under the neutral name buffer. This may stop the earlier metadata view from replacing later owner-pointer uses, and tests whether genuine repeated pointer definitions preserve header base copies. All view extents and field accesses remain valid. Result: 99.084750%, instructions 1059/1062, positional differences 910.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-buffer-metadata-certificates`. Reuse the actual metadata-view pointer for later disjoint serialization-buffer roles under the neutral name buffer. This may stop the earlier metadata view from replacing later owner-pointer uses, and tests whether genuine repeated pointer definitions preserve header base copies. All view extents and field accesses remain valid. Result: 99.223170%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
### rx10c backend and semantic audit
|
|
|
|
|
|
|
|
|
|
Fresh captures c-backup-phase-reuse, c-fat32-ignored-status, c-init-media-getter, c-finalize-status-zero and c-partition-combined-entry all passed independent ordinary compilation: all 40 WAD, 26 SD and 20 partition function instruction streams agree, and each complete ELF object is byte-identical. Validation JSON is beside each capture under build/rx10/mwdbg-c-*. The debugger transport and compiler flags were unchanged.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx: backend-00 B29 has `li r54,0; stw r54,result; lwz r103,installedContentCount`, and B31 writes r54 to result again. Backend-01 removes only the B29 store because the count-error and count-success successors overwrite it. The three header member pointers are already `addi ...,r1,8/32/104(headerBlock)` in backend-00; target emits a separate base add then member add. Thus the four missing opcodes have two different origins, backend dead-store elimination and earlier address folding. The repeated source-level status reads did not change either. Fresh pass 2 still chooses fd v68 as an optimistic spill candidate at degree 29/cost 42, simplifying metadata v67 afterward at degree 18. Reverse coloring retains metadata r25, files r24, size r23, imported bytes r22, output r21, index r20, encryption r19, buffer index r18, remaining r17, chunk r16, callback r15, fd r14. Successful copy merging remains the chunk temporary into its actual local; it does not explain the four absent opcodes.
|
|
|
|
|
- pfd_sddrv_init: the real media getter leaves backend-00's immediate null comparison and backend-01's hoisted loads unchanged. Target disk/report/state colors already agree with v32 r30, v40 r31 and v42/v60/v72 r29. Their simplify degrees are 13, 12 and 22/14/19 respectively. The success value v38 merges into physical r3; incoming disk r3->v32 is rejected for interference. These are correct colors on the wrong 143/144 instruction/control-flow shape, not an allocator tie.
|
|
|
|
|
- pfd_sddrv_finalize: backend-01 B13 places the media zero store and return-zero definition before masking flags. Old flags v56 and masked flags v57 simplify at degree 4 and both color r5; state base v55 has degree 7 and colors r6, zero v58 and field base v59 have degree 3 and color r0/r4. The old value therefore cannot take target r3 because the return-zero interval already occupies it. Only the two call-status copies merge into report argument r4. There is no spill, and the media-first scheduling survives plain reset helpers and status lifetime variants.
|
|
|
|
|
- pdm_part_is_master_boot_sector: combining the two real entry fields into one helper does not retain the target even/odd byte grouping or delayed zero store. Backend-01 still forms byte2+byte1 and byte3+byte0 and writes the dead start zero before prefetching count bytes. High-byte shifts v46/v56 color r31 after simplifying at degrees 16/15; target wants the first high-byte value in r30. Start sum v54 colors r30 at degree 19. The input-buffer and completed-start test copies are the only successful GPR merges. No spill accounts for the discrepancy.
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb: ignoring the private helper return and using its proven nonnull global argument retains the target error-call region, but backend-00 still tests the preceding write result v39 after the signature stores. It simplifies at degree 20 and colors r28, with both result/physical-r3 merge directions rejected for interference across pf_memset. That live interval creates the extra saved-result copy, status test, first report argument setup and final return copy. Report base v45 colors r31, buffer bases v49/v101 r30 and state bases v52/v93 r29, matching target. This 227/222 candidate is a useful source-shape diagnostic, not a match.
|
|
|
|
|
|
|
|
|
|
No diagnostic pragma, volatile qualification, pointer subtraction from null, integer-address field accessor, duplicated wire-header type, constant-size guard, status-passthrough helper or artificial prior-status copy is accepted for production. The pointer-subtraction trial has undefined C semantics and is excluded from semantic acceptance. Volatile count did not compile and lacks independent target evidence. The successful constant-null initialization trial emitted no production change. Search candidates remain subject to line-by-line semantic review.
|
|
|
|
|
|
|
|
|
|
Read-only prior-art searches for the exact three SDK/WAD function names returned no results. A locked origin fetch found only PR1197 in eZiText after our a06cbaa8 baseline; none of the owned source files changed. Branch remains agent/w1005/rx10c, with no rebase or other-worktree action.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-reuse-stream-position`. Target gives r28 to both saved file position and signature position in nonoverlapping stream phases. Reuse one signed stream position local under the neutral name position, retaining every seek offset and comparison. Result: Compile failed; excluded from attempt count.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-expected-stream-size-first`. Target stream-call comparisons put returned r3 before expected size/position. Mirror the equivalent equality operands at 9 complete call conditions; no evaluation order or value changes because the other operand is an unmodified scalar/sizeof expression. Result: 99.223170%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-position-and-comparisons`. Combine truthful stream-position reuse with the target call/expected-value comparison operand order. These are independent of the four previously mapped pre-allocation gaps. Result: Compile failed; excluded from attempt count.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-reuse-stream-position-c89`. Target gives r28 to both saved file position and signature position in nonoverlapping stream phases. Reuse one signed stream position local under the neutral name position, retaining every seek offset and comparison. Result: 99.270250%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-position-and-comparisons-c89`. Combine truthful stream-position reuse with the target call/expected-value comparison operand order. These are independent of the four previously mapped pre-allocation gaps. Result: 99.270250%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-null-address-switch`. Test explicit null-validation control-flow lowering after the separate initialized-success path is restored. Target needs a materialized zero and register comparison; preserve all pointer values and return outcomes. Result: 94.375000%, instructions 144/144, positional differences 15.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-null-success-else`. Test explicit null-validation control-flow lowering after the separate initialized-success path is restored. Target needs a materialized zero and register comparison; preserve all pointer values and return outcomes. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-null-status-guard`. Test explicit null-validation control-flow lowering after the separate initialized-success path is restored. Target needs a materialized zero and register comparison; preserve all pointer values and return outcomes. Result: 93.020836%, instructions 146/144, positional differences 139.
|
|
|
|
|
|
|
|
|
|
Further rx10c source-shape trials: the init address switch reaches 144/144 by adding a branch-over-branch, not by recovering the target materialized-null comparison. Its 15 positional differences and 94.375% are therefore a false instruction-count gain. Reusing the actual signed stream position across file and signature phases brings the private backup trial to 99.27025%, still 1058/1062. Equality-operand mirroring does not change its object. Production remains unchanged; no Matching flip has been attempted because WADBackupEx is not exact.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-position-before-header`. Confirmed file-phase allocator order is byteCount r30, position r29, header r28; target needs byteCount r30, header r29, position r28. Declare the actual per-loop position and header locals in the opposite order while leaving every initialization/call in place; no added state or changed lifetime overlap. Result: 99.260826%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-before-position`. Confirmed file-phase allocator order is byteCount r30, position r29, header r28; target needs byteCount r30, header r29, position r28. Declare the actual per-loop position and header locals in the opposite order while leaving every initialization/call in place; no added state or changed lifetime overlap. Result: 99.223170%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-write-return-local`. Name the 9 actual stream return values before comparison. Target compares returned r3 on the left, while direct-call expressions reverse it in the source. One neutral transferResult local has nonoverlapping call-result lifetimes; all casts, calls and branch outcomes are preserved. Result: 99.317330%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-seek-return-local`. Name the 3 actual stream return values before comparison. Target compares returned r3 on the left, while direct-call expressions reverse it in the source. One neutral transferResult local has nonoverlapping call-result lifetimes; all casts, calls and branch outcomes are preserved. Result: 99.298490%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-stream-return-local`. Name the 12 actual stream return values before comparison. Target compares returned r3 on the left, while direct-call expressions reverse it in the source. One neutral transferResult local has nonoverlapping call-result lifetimes; all casts, calls and branch outcomes are preserved. Result: 99.345570%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-metadata-void`. Use a raw allocated metadata buffer with its genuine typed TMD view, or the real serialized header byte member for I/O. Target keeps metadata owner r25 distinct from the early view r16 and evaluates header-buffer address before stream/context address. Storage sizes, ownership and field layouts are unchanged. Result: Compile failed; excluded from attempt count.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-metadata-bytes`. Use a raw allocated metadata buffer with its genuine typed TMD view, or the real serialized header byte member for I/O. Target keeps metadata owner r25 distinct from the early view r16 and evaluates header-buffer address before stream/context address. Storage sizes, ownership and field layouts are unchanged. Result: Compile failed; excluded from attempt count.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-byte-io`. Use a raw allocated metadata buffer with its genuine typed TMD view, or the real serialized header byte member for I/O. Target keeps metadata owner r25 distinct from the early view r16 and evaluates header-buffer address before stream/context address. Storage sizes, ownership and field layouts are unchanged. Result: 99.345570%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-metadata-void-header-byte-io`. Use a raw allocated metadata buffer with its genuine typed TMD view, or the real serialized header byte member for I/O. Target keeps metadata owner r25 distinct from the early view r16 and evaluates header-buffer address before stream/context address. Storage sizes, ownership and field layouts are unchanged. Result: Compile failed; excluded from attempt count.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-metadata-void-complete`. Complete raw metadata storage diagnostic using byte arithmetic for padding, or use the actual metadata owner for full-TMD export/hash/write. The early TMD-view lifetime ends before the buffer is overwritten with the exported full TMD. Target retains owner r25 at those three later calls; every address and allocated extent is unchanged. Result: 99.345570%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-metadata-bytes-complete`. Complete raw metadata storage diagnostic using byte arithmetic for padding, or use the actual metadata owner for full-TMD export/hash/write. The early TMD-view lifetime ends before the buffer is overwritten with the exported full TMD. Target retains owner r25 at those three later calls; every address and allocated extent is unchanged. Result: 99.345570%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-export-metadata-owner`. Complete raw metadata storage diagnostic using byte arithmetic for padding, or use the actual metadata owner for full-TMD export/hash/write. The early TMD-view lifetime ends before the buffer is overwritten with the exported full TMD. Target retains owner r25 at those three later calls; every address and allocated extent is unchanged. Result: 99.392654%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-raw-export-metadata-owner`. Complete raw metadata storage diagnostic using byte arithmetic for padding, or use the actual metadata owner for full-TMD export/hash/write. The early TMD-view lifetime ends before the buffer is overwritten with the exported full TMD. Target retains owner r25 at those three later calls; every address and allocated extent is unchanged. Result: 99.392654%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-padding-size-lifetime`. The remaining file-padding temporary is target r3 versus source r8. Test the real rounded-size lifetime through the size local, one expression, a scoped named value, or the disjoint stream-result scalar renamed neutrally. All arithmetic remains unsigned 32-bit and all file-size values are preserved. Result: 99.392654%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-padding-expression`. The remaining file-padding temporary is target r3 versus source r8. Test the real rounded-size lifetime through the size local, one expression, a scoped named value, or the disjoint stream-result scalar renamed neutrally. All arithmetic remains unsigned 32-bit and all file-size values are preserved. Result: 99.406780%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-padding-local`. The remaining file-padding temporary is target r3 versus source r8. Test the real rounded-size lifetime through the size local, one expression, a scoped named value, or the disjoint stream-result scalar renamed neutrally. All arithmetic remains unsigned 32-bit and all file-size values are preserved. Result: 99.406780%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-padding-transfer-value`. The remaining file-padding temporary is target r3 versus source r8. Test the real rounded-size lifetime through the size local, one expression, a scoped named value, or the disjoint stream-result scalar renamed neutrally. All arithmetic remains unsigned 32-bit and all file-size values are preserved. Result: 99.406780%, instructions 1058/1062, positional differences 924.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-buffer-local`. Bind the actual serialized-header buffer before I/O and use that view consistently. Target prepares the header argument before the stream/hash argument at three calls and separately forms member addresses. One genuine header pointer preserves extent and alignment; compare address lowering without an artificial barrier. Result: 99.891716%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-buffer-bound-at-use`. Bind the actual serialized-header buffer before I/O and use that view consistently. Target prepares the header argument before the stream/hash argument at three calls and separately forms member addresses. One genuine header pointer preserves extent and alignment; compare address lowering without an artificial barrier. Result: 99.778720%, instructions 1061/1062, positional differences 938.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-view-io`. Bind the actual serialized-header buffer before I/O and use that view consistently. Target prepares the header argument before the stream/hash argument at three calls and separately forms member addresses. One genuine header pointer preserves extent and alignment; compare address lowering without an artificial barrier. Result: 99.426550%, instructions 1058/1062, positional differences 925.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-header-and-padding`. Combine consistent serialized-header pointer use with the scoped file-padding value. Header pointer used for both fields and I/O recovers all three target base/member addi pairs and argument ordering; padding local gives the target r3. Only the pre-count success store should remain absent. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-status-pointer-all`. Isolate the single missing pre-count status store after recovering every other target opcode. Compare genuine shared-status/count pointer use throughout the function or a separate real ES query-result local. Self-assignment/round-trip status variants are diagnostic only, not proposed production source. Result: 99.894540%, instructions 1061/1062, positional differences 938.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-pointer-all`. Isolate the single missing pre-count status store after recovering every other target opcode. Compare genuine shared-status/count pointer use throughout the function or a separate real ES query-result local. Self-assignment/round-trip status variants are diagnostic only, not proposed production source. Result: 99.651600%, instructions 1062/1062, positional differences 120.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-card-status-transfer`. Isolate the single missing pre-count status store after recovering every other target opcode. Compare genuine shared-status/count pointer use throughout the function or a separate real ES query-result local. Self-assignment/round-trip status variants are diagnostic only, not proposed production source. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-card-status-content`. Isolate the single missing pre-count status store after recovering every other target opcode. Compare genuine shared-status/count pointer use throughout the function or a separate real ES query-result local. Self-assignment/round-trip status variants are diagnostic only, not proposed production source. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-status-current`. Isolate the single missing pre-count status store after recovering every other target opcode. Compare genuine shared-status/count pointer use throughout the function or a separate real ES query-result local. Self-assignment/round-trip status variants are diagnostic only, not proposed production source. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-status-transient`. Isolate the single missing pre-count status store after recovering every other target opcode. Compare genuine shared-status/count pointer use throughout the function or a separate real ES query-result local. Self-assignment/round-trip status variants are diagnostic only, not proposed production source. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-after-success`. Test whether conditional count validation dependent on the normalized status preserves its first store until later constant folding. Query outcomes and cleanup remain identical. The signed-nonzero decomposition is diagnostic only; retain only readable status-flow reconstruction with exact target instructions. Result: 99.661020%, instructions 1061/1062, positional differences 940.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-with-success`. Test whether conditional count validation dependent on the normalized status preserves its first store until later constant folding. Query outcomes and cleanup remain identical. The signed-nonzero decomposition is diagnostic only; retain only readable status-flow reconstruction with exact target instructions. Result: 99.566860%, instructions 1064/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-nested-count-success`. Test whether conditional count validation dependent on the normalized status preserves its first store until later constant folding. Query outcomes and cleanup remain identical. The signed-nonzero decomposition is diagnostic only; retain only readable status-flow reconstruction with exact target instructions. Result: 99.661020%, instructions 1061/1062, positional differences 940.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-switch`. Test whether conditional count validation dependent on the normalized status preserves its first store until later constant folding. Query outcomes and cleanup remain identical. The signed-nonzero decomposition is diagnostic only; retain only readable status-flow reconstruction with exact target instructions. Result: 99.557440%, instructions 1062/1062, positional differences 9.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-count-signed-nonzero`. Test whether conditional count validation dependent on the normalized status preserves its first store until later constant folding. Query outcomes and cleanup remain identical. The signed-nonzero decomposition is diagnostic only; retain only readable status-flow reconstruction with exact target instructions. Result: 99.509415%, instructions 1063/1062, positional differences 941.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-nonnull-count-helper`. Compare scalar versus output-pointer contracts at the actual ES query/count validation boundary. A real input-null check in the validator is folded for its stack argument. The target first status store remains the only missing instruction; require matching CFG before retaining any helper. Result: 99.175140%, instructions 1068/1062, positional differences 933.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-nonnull-count-output`. Compare scalar versus output-pointer contracts at the actual ES query/count validation boundary. A real input-null check in the validator is folded for its stack argument. The target first status store remains the only missing instruction; require matching CFG before retaining any helper. Result: 99.175140%, instructions 1068/1062, positional differences 933.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-normalized-query-output`. Compare scalar versus output-pointer contracts at the actual ES query/count validation boundary. A real input-null check in the validator is folded for its stack argument. The target first status store remains the only missing instruction; require matching CFG before retaining any helper. Result: 97.325800%, instructions 1064/1062, positional differences 923.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-normalize-query-else-block`. Place the normalized query-success store in an explicit predecessor block before count validation, keeping both actual cleanup branches. Compare whether block merging occurs after dead-store elimination; this specifically tests the sole remaining source-shape gap, not register ties. Result: 99.661020%, instructions 1061/1062, positional differences 940.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-normalize-query-positive-block`. Place the normalized query-success store in an explicit predecessor block before count validation, keeping both actual cleanup branches. Compare whether block merging occurs after dead-store elimination; this specifically tests the sole remaining source-shape gap, not register ties. Result: 99.557440%, instructions 1062/1062, positional differences 14.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-normalize-nested-count-block`. Place the normalized query-success store in an explicit predecessor block before count validation, keeping both actual cleanup branches. Compare whether block merging occurs after dead-store elimination; this specifically tests the sole remaining source-shape gap, not register ties. Result: 99.661020%, instructions 1061/1062, positional differences 940.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-normalize-loop-break`. Place the normalized query-success store in an explicit predecessor block before count validation, keeping both actual cleanup branches. Compare whether block merging occurs after dead-store elimination; this specifically tests the sole remaining source-shape gap, not register ties. Result: 99.557440%, instructions 1062/1062, positional differences 14.
|
|
|
|
|
|
|
|
|
|
All 15 original rx10c searches completed under the shared 24-slot limiter: WAD seeds 40/41/42 ran 1047/1089/1043 trials; partition 1880/1861/1879; init 1658/1689/1681; finalize 1464/1525/1515; FAT32 1430/1505/1497. Total 22,763 scored trials, no saved best or solution. Three additional 1200-second WAD searches (seeds 43/44/45) now start from the newly reconstructed 99.90584% source with only the pre-count status store absent. These test the substantially changed header and scalar lifetimes, preserving the limiter and the 39 exact functions.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-pfd_sddrv_finalize-consistent-state-pointer`. Follow the header-buffer finding: use one genuine driver-state pointer consistently in field accesses and existing accessors, or scope it to the final reset. Earlier state-pointer trials retained mixed direct-global helper arguments. Preserve structure, field types and external calls; inspect scheduling before allocation. Result: 67.666664%, instructions 52/57, positional differences 45.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-pfd_sddrv_finalize-direct-state-pointer`. Follow the header-buffer finding: use one genuine driver-state pointer consistently in field accesses and existing accessors, or scope it to the final reset. Earlier state-pointer trials retained mixed direct-global helper arguments. Preserve structure, field types and external calls; inspect scheduling before allocation. Result: 65.491230%, instructions 51/57, positional differences 52.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_finalize, `c-pfd_sddrv_finalize-scoped-reset-pointer`. Follow the header-buffer finding: use one genuine driver-state pointer consistently in field accesses and existing accessors, or scope it to the final reset. Earlier state-pointer trials retained mixed direct-global helper arguments. Preserve structure, field types and external calls; inspect scheduling before allocation. Result: 92.631580%, instructions 57/57, positional differences 4.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-pfd_sddrv_init-consistent-state-pointer`. Follow the header-buffer finding: use one genuine driver-state pointer consistently in field accesses and existing accessors, or scope it to the final reset. Earlier state-pointer trials retained mixed direct-global helper arguments. Preserve structure, field types and external calls; inspect scheduling before allocation. Result: 86.000000%, instructions 138/144, positional differences 136.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-pfd_sddrv_init-direct-state-pointer`. Follow the header-buffer finding: use one genuine driver-state pointer consistently in field accesses and existing accessors, or scope it to the final reset. Earlier state-pointer trials retained mixed direct-global helper arguments. Preserve structure, field types and external calls; inspect scheduling before allocation. Result: 84.534720%, instructions 137/144, positional differences 136.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-dead-initializers-device`. Target materializes zero before its unsigned disk-null comparison. Test ordinary initialization of the actual address-taken device/status outputs and local status, all overwritten before any live use. A removed initializer may leave the target zero value through constant propagation; no new state or changed caller behavior. Result: 95.138885%, instructions 145/144, positional differences 135.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-dead-initializers-status`. Target materializes zero before its unsigned disk-null comparison. Test ordinary initialization of the actual address-taken device/status outputs and local status, all overwritten before any live use. A removed initializer may leave the target zero value through constant propagation; no new state or changed caller behavior. Result: 95.138885%, instructions 145/144, positional differences 135.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-dead-initializers-result`. Target materializes zero before its unsigned disk-null comparison. Test ordinary initialization of the actual address-taken device/status outputs and local status, all overwritten before any live use. A removed initializer may leave the target zero value through constant propagation; no new state or changed caller behavior. Result: 95.138885%, instructions 143/144, positional differences 138.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-dead-initializers-device-status`. Target materializes zero before its unsigned disk-null comparison. Test ordinary initialization of the actual address-taken device/status outputs and local status, all overwritten before any live use. A removed initializer may leave the target zero value through constant propagation; no new state or changed caller behavior. Result: 94.444440%, instructions 146/144, positional differences 135.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-dead-initializers-device-result`. Target materializes zero before its unsigned disk-null comparison. Test ordinary initialization of the actual address-taken device/status outputs and local status, all overwritten before any live use. A removed initializer may leave the target zero value through constant propagation; no new state or changed caller behavior. Result: 95.138885%, instructions 145/144, positional differences 135.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-dead-initializers-status-result`. Target materializes zero before its unsigned disk-null comparison. Test ordinary initialization of the actual address-taken device/status outputs and local status, all overwritten before any live use. A removed initializer may leave the target zero value through constant propagation; no new state or changed caller behavior. Result: 95.138885%, instructions 145/144, positional differences 135.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_init, `c-init-dead-initializers-device-status-result`. Target materializes zero before its unsigned disk-null comparison. Test ordinary initialization of the actual address-taken device/status outputs and local status, all overwritten before any live use. A removed initializer may leave the target zero value through constant propagation; no new state or changed caller behavior. Result: 94.444440%, instructions 146/144, positional differences 135.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-initialized-result`. Test ordinary initialization of an existing handle, count or status local before its first live assignment. The author-retained dead code may originate in local initialization; every value is overwritten before use. Only a target-position instruction recovery can be accepted. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-initialized-content-status`. Test ordinary initialization of an existing handle, count or status local before its first live assignment. The author-retained dead code may originate in local initialization; every value is overwritten before use. Only a target-position instruction recovery can be accepted. Result: 98.664780%, instructions 1061/1062, positional differences 970.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-initialized-transfer-result`. Test ordinary initialization of an existing handle, count or status local before its first live assignment. The author-retained dead code may originate in local initialization; every value is overwritten before use. Only a target-position instruction recovery can be accepted. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-initialized-file-handle`. Test ordinary initialization of an existing handle, count or status local before its first live assignment. The author-retained dead code may originate in local initialization; every value is overwritten before use. Only a target-position instruction recovery can be accepted. Result: 99.758940%, instructions 1062/1062, positional differences 120.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-initialized-read-size`. Test ordinary initialization of an existing handle, count or status local before its first live assignment. The author-retained dead code may originate in local initialization; every value is overwritten before use. Only a target-position instruction recovery can be accepted. Result: 99.835220%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-initialized-size`. Test ordinary initialization of an existing handle, count or status local before its first live assignment. The author-retained dead code may originate in local initialization; every value is overwritten before use. Only a target-position instruction recovery can be accepted. Result: 98.674200%, instructions 1061/1062, positional differences 969.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-initialized-operation-status`. Test initialization and dominance of the actual prior successful block-write status when the nonnull reserved-sector helper return is ignored. Target folds both null and status checks yet retains the dead report region. Nested-success form keeps identical calls and errors while exposing the success edge directly. Result: 97.364870%, instructions 227/222, positional differences 117.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-success-before-reserved`. Test initialization and dominance of the actual prior successful block-write status when the nonnull reserved-sector helper return is ignored. Target folds both null and status checks yet retains the dead report region. Nested-success form keeps identical calls and errors while exposing the success edge directly. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-success-before-and-after-reserved`. Test initialization and dominance of the actual prior successful block-write status when the nonnull reserved-sector helper return is ignored. Target folds both null and status checks yet retains the dead report region. Nested-success form keeps identical calls and errors while exposing the success edge directly. Result: 97.207210%, instructions 216/222, positional differences 100.
|
|
|
|
|
|
|
|
|
|
- pfd_sddrv_build_fat32_mbr_bpb, `c-fat32-nested-prior-write-success`. Test initialization and dominance of the actual prior successful block-write status when the nonnull reserved-sector helper return is ignored. Target folds both null and status checks yet retains the dead report region. Nested-success form keeps identical calls and errors while exposing the success edge directly. Result: 92.815315%, instructions 227/222, positional differences 60.
|
|
|
|
|
|
|
|
|
|
### rx10c one-instruction WAD result
|
|
|
|
|
|
|
|
|
|
Private candidate `build/rx10c/trials/c-backup-header-and-padding/wad.c` is 99.90584%, 1061/1062 instructions. Its patch against this branch is `build/rx10c/backup-99.90584.patch`; production source is unchanged. `build/rx10c/prove_backup_gap.py` proves that deleting target instruction 132 (`stw r3,0x7c(r1)`) and adjusting branches across that location makes all 1061 remaining instructions identical. All 97 relocations also agree after accounting for that shift and resolving local literal labels by section offset and bytes. The first naive relocation comparison failed only because compiler literal labels differ; the byte/offset comparison passes. Private pool: all 18 strings identical; other WAD functions: 39/39 exact. This proof isolates the defect but is not ctxdiff zero or an exact match.
|
|
|
|
|
|
|
|
|
|
The earlier three missing header-address instructions were source shape, not register ties. A single `WADBackupHeaderBlock* headerBuffer = &headerBlock` used consistently for both fields and I/O survives backend-01 as a real header base v59. The backend then retains `addi r3,r59,8/32/104` at the member calls. It later rematerializes the stack base separately, exactly as target. Using a pointer only for fields or only for I/O did not recover these instructions. No new aggregate, field, alignment or carrier object was needed.
|
|
|
|
|
|
|
|
|
|
Separate true file-phase values recovered the remaining allocation: `byteCount` holds path length and remaining file bytes, `position` holds disjoint file/signature seek offsets, and `size` holds transfer/padding sizes. Using the metadata owner for full-TMD export/hash/write ends the early `titleMeta` view lifetime and permits the target r16 zero constants. A scoped `roundedSize` for the 32-byte file length receives r3 rather than the source read-result local's r8. Named `transferResult` values preserve target operand order at stream comparisons. Every name describes all its uses.
|
|
|
|
|
|
|
|
|
|
Fresh captures `c-backup-neutral-phases` and `c-backup-one-store` independently recompile byte-identically through the ordinary compiler: all 40 function streams and the complete ELF agree. Their validation JSON files are under `build/rx10/mwdbg-c-backup-*/ordinary-validation.json`. In the one-store capture backend-00 B29 still contains `li r55,0; stw r55,result; lwz r105,installedContentCount`; backend-01 removes that store because B30 and B31 overwrite result. B31 retains its zero store. Thus the sole remaining gap is before allocation. The target register and branch mapping is otherwise exact after the one-instruction adjustment; further declaration shuffles cannot be called a solution to that gap.
|
|
|
|
|
|
|
|
|
|
Whole-function status/count pointers, conditional status-dependent validation, explicit normalization predecessor blocks, normal dead initializers, and nullable validator/output contracts did not restore the target store. Count-switch and initialized-file-handle trials reach 1062 instructions with the wrong extra branch or entry initialization, so they are rejected. Explicit prior-success initialization in FAT32 removes the six target dead report instructions; nested success leaves the old result live across the buffer memset and still adds five instructions. Consistent SD state-pointer use removes target global-base loads, while scoped reset pointers leave the finalize scheduling discrepancy unchanged. No private helper, diagnostic type/qualifier, self-assignment or constant guard is retained.
|
|
|
|
|
|
|
|
|
|
One-store allocation confirmation: pass 2 selects fd v70 as an optimistic spill candidate at degree 29/cost 42; metadata v69 then simplifies at degree 18 and colors r25. The core colors remain files r24, size r23, imported bytes r22, output r21, index r20, encryption r19, buffer index r18, remaining r17, chunk r16, callback r15, fd r14. After v299's degree-29/cost-17 optimistic selection, byteCount v42, fileHeaderBuffer v43 and position v56 simplify at degrees 28/27/26 and reverse-color r30/r29/r28, matching target. The shortened titleMeta v58 simplifies at degree 19 and receives r16. roundedSize v41 simplifies early at degree 20; its legal-color mask is 0x40801ff8, whose first legal register is r3. contentStatus v55 has degree 12 and mask 0xfc941ff8, also selecting r3. Header base v59 has become dead/rematerialized by coloring (degree zero); its nominal r0 assignment is not a live header-pointer register. Only chunk temporary v238 coalesces into actual chunkSize v60 in pass 1; all 75 pass-2 candidates are rejected for interference. These traces explain why the remaining defect cannot be repaired by choosing a different physical color.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-indexed-zero`. Target delays the dead start-array zero until after all eight input bytes have been loaded. Test indexed access to the real local arrays alongside or instead of the advancing output cursors, so loop strength reduction may expose the common address after dead-store scheduling. Arrays, input bytes, store order and unsigned arithmetic stay identical. Result: 83.559525%, instructions 88/84, positional differences 87.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-indexed-start`. Target delays the dead start-array zero until after all eight input bytes have been loaded. Test indexed access to the real local arrays alongside or instead of the advancing output cursors, so loop strength reduction may expose the common address after dead-store scheduling. Arrays, input bytes, store order and unsigned arithmetic stay identical. Result: 77.107140%, instructions 90/84, positional differences 89.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-indexed-count`. Target delays the dead start-array zero until after all eight input bytes have been loaded. Test indexed access to the real local arrays alongside or instead of the advancing output cursors, so loop strength reduction may expose the common address after dead-store scheduling. Arrays, input bytes, store order and unsigned arithmetic stay identical. Result: 83.547620%, instructions 88/84, positional differences 87.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-indexed-stores`. Target delays the dead start-array zero until after all eight input bytes have been loaded. Test indexed access to the real local arrays alongside or instead of the advancing output cursors, so loop strength reduction may expose the common address after dead-store scheduling. Arrays, input bytes, store order and unsigned arithmetic stay identical. Result: 74.511900%, instructions 94/84, positional differences 93.
|
|
|
|
|
|
|
|
|
|
- pdm_part_is_master_boot_sector, `c-partition-indexed-whole-loop`. Target delays the dead start-array zero until after all eight input bytes have been loaded. Test indexed access to the real local arrays alongside or instead of the advancing output cursors, so loop strength reduction may expose the common address after dead-store scheduling. Arrays, input bytes, store order and unsigned arithmetic stay identical. Result: 89.464290%, instructions 84/84, positional differences 22.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-signed-count`. Check the genuine ES output count scalar type at the last missing-store boundary. Signed and unsigned corresponding 32-bit integer types may alias; the zero check and unsigned loop comparison keep identical bit semantics. No width, address or output extent changes. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-signed-long-count`. Check the genuine ES output count scalar type at the last missing-store boundary. Signed and unsigned corresponding 32-bit integer types may alias; the zero check and unsigned loop comparison keep identical bit semantics. No width, address or output extent changes. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-unsigned-int-count`. Check the genuine ES output count scalar type at the last missing-store boundary. Signed and unsigned corresponding 32-bit integer types may alias; the zero check and unsigned loop comparison keep identical bit semantics. No width, address or output extent changes. Result: Compile failed; excluded from attempt count.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-preserve-query-error`. Make the existing normalized status a real input on the count-error edge before cleanup. Preserving an existing error tests generic error-handling structure; subtraction is diagnostic only and will not be retained. This distinguishes dead-store removal before versus after status load forwarding, without changing the count-zero outcome. Result: 99.811676%, instructions 1064/1062, positional differences 937.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-preserve-query-error-conditional`. Make the existing normalized status a real input on the count-error edge before cleanup. Preserving an existing error tests generic error-handling structure; subtraction is diagnostic only and will not be retained. This distinguishes dead-store removal before versus after status load forwarding, without changing the count-zero outcome. Result: 99.529190%, instructions 1065/1062, positional differences 931.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-error-status-subtraction-diagnostic`. Make the existing normalized status a real input on the count-error edge before cleanup. Preserving an existing error tests generic error-handling structure; subtraction is diagnostic only and will not be retained. This distinguishes dead-store removal before versus after status load forwarding, without changing the count-zero outcome. Result: 99.905840%, instructions 1061/1062, positional differences 939.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-no-prior-error`. The prior-error guard retains the exact missing status store but adds an always-false cleanup test. Compare ordinary zero/nonnegative status predicates to determine whether late bit/range simplification removes that test after dead-store analysis. All actual incoming statuses at this boundary are zero; no change is retained without exact code and readability review. Result: 99.811676%, instructions 1064/1062, positional differences 937.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-nonnegative-status`. The prior-error guard retains the exact missing status store but adds an always-false cleanup test. Compare ordinary zero/nonnegative status predicates to determine whether late bit/range simplification removes that test after dead-store analysis. All actual incoming statuses at this boundary are zero; no change is retained without exact code and readability review. Result: 99.811676%, instructions 1064/1062, positional differences 937.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-nonnegative-status-bit`. The prior-error guard retains the exact missing status store but adds an always-false cleanup test. Compare ordinary zero/nonnegative status predicates to determine whether late bit/range simplification removes that test after dead-store analysis. All actual incoming statuses at this boundary are zero; no change is retained without exact code and readability review. Result: 100.000000%, instructions 1062/1062, positional differences 0.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-status-identity`. The error-preservation guard reads escaped result after storing the normalized content status. Compare it with that actual normalized status rather than immediate zero: load forwarding may make the two operands identical late enough to remove the guard while preserving the earlier store. No new variable or altered error outcome; difference spelling is diagnostic only. Result: 99.811676%, instructions 1064/1062, positional differences 937.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-status-identity-mirrored`. The error-preservation guard reads escaped result after storing the normalized content status. Compare it with that actual normalized status rather than immediate zero: load forwarding may make the two operands identical late enough to remove the guard while preserving the earlier store. No new variable or altered error outcome; difference spelling is diagnostic only. Result: 99.811676%, instructions 1064/1062, positional differences 937.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-status-difference`. The error-preservation guard reads escaped result after storing the normalized content status. Compare it with that actual normalized status rather than immediate zero: load forwarding may make the two operands identical late enough to remove the guard while preserving the earlier store. No new variable or altered error outcome; difference spelling is diagnostic only. Result: 100.000000%, instructions 1062/1062, positional differences 0.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-unsigned-error-bit`. Find the plainest negative-error test that preserves the target redundant status store. The high-bit test of the existing signed 32-bit status is equivalent to nonnegative status on this compiler; compare ordinary unsigned mask, signed shift, boolean comparison and unsigned range forms. No state or side effects are added. Result: 100.000000%, instructions 1062/1062, positional differences 0.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-signed-error-shift`. Find the plainest negative-error test that preserves the target redundant status store. The high-bit test of the existing signed 32-bit status is equivalent to nonnegative status on this compiler; compare ordinary unsigned mask, signed shift, boolean comparison and unsigned range forms. No state or side effects are added. Result: 99.811676%, instructions 1064/1062, positional differences 937.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-success-comparison-bool`. Find the plainest negative-error test that preserves the target redundant status store. The high-bit test of the existing signed 32-bit status is equivalent to nonnegative status on this compiler; compare ordinary unsigned mask, signed shift, boolean comparison and unsigned range forms. No state or side effects are added. Result: 99.622410%, instructions 1066/1062, positional differences 916.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-unsigned-status-range`. Find the plainest negative-error test that preserves the target redundant status store. The high-bit test of the existing signed 32-bit status is equivalent to nonnegative status on this compiler; compare ordinary unsigned mask, signed shift, boolean comparison and unsigned range forms. No state or side effects are added. Result: 99.717514%, instructions 1065/1062, positional differences 929.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-exact-neutral-names`. Use streamResult for both writes and seeks, and length for NAND read counts and rounded file lengths. Preserve declaration order and token position otherwise. The zero-content branch preserves an existing negative status by checking its error bit; after zero normalization this is behaviorally identical to assigning -3002, and late bit simplification retains exactly the target dead store. Result: 100.000000%, instructions 1062/1062, positional differences 0.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-negative-predicate`. Readability check: express the genuine signed-error predicate through a plain scalar helper, which is valid for every 32-bit status and has no carrier state. Prefer a normal negative comparison if the inline boundary retains the target store; compare to the exact direct sign-bit form. Result: 100.000000%, instructions 1062/1062, positional differences 0.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-nonnegative-predicate`. Readability check: express the genuine signed-error predicate through a plain scalar helper, which is valid for every 32-bit status and has no carrier state. Prefer a normal negative comparison if the inline boundary retains the target store; compare to the exact direct sign-bit form. Result: 99.622410%, instructions 1066/1062, positional differences 916.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-negative-bit-predicate`. Readability check: express the genuine signed-error predicate through a plain scalar helper, which is valid for every 32-bit status and has no carrier state. Prefer a normal negative comparison if the inline boundary retains the target store; compare to the exact direct sign-bit form. Result: 100.000000%, instructions 1062/1062, positional differences 0.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-exact-error-predicate`. Replace the low-level error-bit spelling with a plain, real signed-error predicate: _WADIsError(result) returns result < 0. The scalar inline boundary keeps the target status store and compiles to the same exact function. No carrier object, artificial arithmetic or qualifier is needed. Result: 100.000000%, instructions 1062/1062, positional differences 0.
|
|
|
|
|
|
|
|
|
|
- WADBackupEx, `c-backup-exact-predicate-no-status-temp`. With the plain signed-error helper in place, test removing the named normalized-content status and writing literal success directly. Prefer this smaller source if both target zero stores and physical colors remain exact. Result: 100.000000%, instructions 1062/1062, positional differences 0.
|
|
|
|
|
|
|
|
|
|
### rx10c exact WADBackupEx and link candidate
|
|
|
|
|
|
|
|
|
|
The missing-store hypothesis now has an exact plain-C form. On the zero-content cleanup edge, preserve an already-negative status before selecting -3002. A direct `result >= 0` condition keeps two unwanted test instructions, but a plain scalar `_WADIsError(s32 result)` helper returning `result < 0` moves the signed-error predicate across the inline boundary. That boundary allows the predicate to simplify after the escaped-status store is considered live. Both target zero stores survive and the test disappears. No error-bit mask or diagnostic subtraction is retained. The earlier normalized-content status temporary can also be removed: literal `result = 0` at the two target points is exact with the helper. This is a real status predicate, valid for every signed 32-bit result; it creates no carrier state. At this call site result has just been normalized to zero, so the error guard is redundant in reachable behavior, consistent with the target's dead status store. This redundancy is explicit for source review.
|
|
|
|
|
|
|
|
|
|
Final source trial: `c-backup-exact-predicate-no-status-temp`, objdiff 100.0%, 1062/1062, positional diffs 0; every one of the 40 WAD functions is exact. Rebuilt production object confirms `ctxdiff: diffs 0` and 18 identical strings. The final names are `streamResult` for seek/write results and `length` for NAND read counts/rounded file lengths; declaration order is unchanged by renaming. Only WADBackupEx and its adjacent plain error predicate changed in wad.c.
|
|
|
|
|
|
|
|
|
|
The additional WAD seeds 43/44/45 completed 1082/1075/1084 trials from the 99.90584% state without a saved candidate. Across all 18 searches: 26,004 trials. The exact result came from manual status-flow reconstruction, not a search mutation. All searches completed before link validation.
|
|
|
|
|
|
|
|
|
|
Per the explicit rx10c instruction, changed only wad/wad.c to Matching in configure.py. Both raw symorder logs are under build/rx10c/wad-symorder-{before,after}-link.log. They report ORDER MISMATCH for compiler-local literal names and the source-only zero-size ...data.0 base marker. All 40 text symbols have identical names, offsets and sizes. After excluding zero-size markers, every data/sdata symbol offset and size agrees. The source .data bytes exactly equal the first 458 target bytes; the target has six unnamed trailing zero bytes. The .sdata section is exactly 64 bytes on both sides. The initial strict data-size assertion exposed these six zeros; no source padding or artificial symbol was added. Evidence is in build/rx10c/wad-data-layout.json. The real linked DOL resolves the harmless differences: doldiff reports 0 differing bytes, equal 3,867,904-byte files, and SHA1 26116613f624061ba99c8d1a299aaa6efa85670d. This confirms the Matching flip, not merely an object score.
|
|
|
|
|
|
|
|
|
|
Final-helper backend confirmation is available in `build/rx10/mwdbg-c-backup-final`. Backend-00 B29 writes literal-zero v106 to result before loading/testing installedContentCount. B30 loads that escaped result into v71, and the inlined signed predicate lowers to `rlwinm v70,v71,1,31,31` followed by its conditional branch. Backend-01 forwards the zero and removes B30's predicate/test entirely, while B29 keeps `stw v106,result`; B32 reuses v106 for the second success store. This is exactly the target pre-allocation control flow. Direct signed comparisons kept the extra test; arbitrary status subtraction removed the store again. The final source retains only the ordinary signed predicate helper and the two literal success assignments.
|
|
|
|
|
|
|
|
|
|
Both target and source data sections have alignment 8. The target's .data size 464 is the source's 458-byte payload rounded to that alignment. Both .sdata sections are 64 bytes. The successful link and zero-byte doldiff confirm linker alignment supplies the six tail zeros without source padding or a section/symbol trick.
|
|
|
|
|
|
|
|
|
|
Final-helper allocator map: normalized success v106 simplifies at degree 12 and takes r3 from legal mask 0xfc941ff8; roundedSize v41 simplifies at degree 20 and takes r3 from mask 0x40801ff8 on its separate lifetime. FD v69 is the degree-29/cost-42 optimistic candidate, and metadata v68 simplifies afterward at degree 18, yielding target fd r14/metadata r25. File-phase byteCount v42, header view v43 and position v55 simplify at degrees 28/27/26 and color r30/r29/r28. The early metadata view v57 uses r16. Only chunk temporary v241 merges into chunkSize v59 in pass 1, and pass 2 has no successful coalesces. Final backend-04 agrees with all target physical operands, confirmed by the ordinary 1062/1062 zero-diff object.
|
|
|
|
|
|
|
|
|
|
### rx10c final validation and retained scope
|
|
|
|
|
|
|
|
|
|
- libs/RVL_SDK/src/wad/wad: exact functions 39/40 -> 40/40; code 20252 -> 24500/24500; data 528 -> 528/528.
|
|
|
|
|
WADBackupEx: 97.385124% -> 100.000000%; 141 compiled manual rx10c trials, plus the search runs above.
|
|
|
|
|
- libs/RVL_SDK/src/fa/driver/sd_drv: exact functions 23/26 -> 23/26; code 10068 -> 10068/11760; data 3592 -> 3592/3592.
|
|
|
|
|
pfd_sddrv_init: 93.125000% -> 93.125000%; 22 compiled manual rx10c trials, plus the search runs above.
|
|
|
|
|
pfd_sddrv_finalize: 92.631580% -> 92.631580%; 12 compiled manual rx10c trials, plus the search runs above.
|
|
|
|
|
pfd_sddrv_build_fat32_mbr_bpb: 95.202705% -> 95.202705%; 39 compiled manual rx10c trials, plus the search runs above.
|
|
|
|
|
- libs/RVL_SDK/src/fa/pdm_partition: exact functions 19/20 -> 19/20; code 3380 -> 3380/3716; data 0 -> 0/0.
|
|
|
|
|
pdm_part_is_master_boot_sector: 87.392860% -> 87.392860%; 15 compiled manual rx10c trials, plus the search runs above.
|
|
|
|
|
|
|
|
|
|
Final clean gate (without --quick) passed on wad, sd_drv and pdm_partition. Fresh objects, final report, and linked DOL agree. Zero regressions, zero added forbidden patterns and zero readability warnings. WAD is now Matching with 24,500 linked code bytes. Remaining SD/partition functions keep their baseline source and scores; their private trials are not retained. No function remains untried, and no unresolved source-shape difference is mislabeled as a proven allocator tie.
|
|
|
|
|
|
|
|
|
|
After the clean build, ctxdiff reports zero for WADBackupEx 1062/1062, WADImportEx 1146/1146, WAD_815C1288 61/61, WADImportGetBlocks 298/298, WADImportDVDExForBS 263/263 and pfd_sddrv_store_mbr_buf 202/202. All WAD functions are 100% in the fresh report. Final doldiff is 0 bytes with equal file lengths; DOL SHA1 is 26116613f624061ba99c8d1a299aaa6efa85670d. All debugger captures used for acceptance passed ordinary-compilation whole-ELF equality, including the final plain signed-error predicate.
|
|
|
|
|
|
|
|
|
|
Retained files are configure.py (only the authorized WAD Matching flip), libs/RVL_SDK/src/wad/wad.c (WADBackupEx plus _WADIsError), and this log. No shared header, SD/partition source, symbol configuration or other worktree was changed. The only symorder discrepancy is local-label naming/zero-size base marker; equal symbol extents, aligned data bytes and the linked DOL prove real order agrees. The error-preserving guard is deliberately redundant after success normalization; this is the source-review point behind the target's otherwise dead status store.
|
|
|
|
|
|
|
|
|
|
Final evidence: build/rx10c/final-gate.log, final-report.json, final-summary.json, final-doldiff.log, final-symorder.log, final-capture-audit.log, and build/rx10/mwdbg-c-backup-final/. All 18 source searches and debugger jobs completed. No push, PR, merge or rebase was performed.
|
|
|
|
|
|
|
|
|
|
```text
|
|
|
|
|
full build: ok
|
|
|
|
|
main.dol sha1: 26116613f624061ba99c8d1a299aaa6efa85670d
|
|
|
|
|
[libs/RVL_SDK/src/wad/wad] pool: IDENTICAL
|
|
|
|
|
[libs/RVL_SDK/src/wad/wad] objdiff: code 24500/24500 data 528/528 functions 40/40 fuzzy 100.0000 linked code 24500
|
|
|
|
|
[libs/RVL_SDK/src/wad/wad] instruction-exact functions: 40/40
|
|
|
|
|
[libs/RVL_SDK/src/wad/wad] section .data size 464 match 100.0
|
|
|
|
|
[libs/RVL_SDK/src/wad/wad] section .sdata size 64 match 100.0
|
|
|
|
|
[libs/RVL_SDK/src/wad/wad] section .text size 24500 match 100.0
|
|
|
|
|
[libs/RVL_SDK/src/wad/wad] baseline: code 20252/24500 data 528 functions 39 fuzzy 99.5466
|
|
|
|
|
[libs/RVL_SDK/src/fa/driver/sd_drv] pool: IDENTICAL
|
|
|
|
|
[libs/RVL_SDK/src/fa/driver/sd_drv] objdiff: code 10068/11760 data 3592/3592 functions 23/26 fuzzy 99.1582 linked code 0
|
|
|
|
|
[libs/RVL_SDK/src/fa/driver/sd_drv] instruction-exact functions: 23/26
|
|
|
|
|
[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 99.158165
|
|
|
|
|
[libs/RVL_SDK/src/fa/driver/sd_drv] below 100: pfd_sddrv_init 93.125
|
|
|
|
|
[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_build_fat32_mbr_bpb 95.202705
|
|
|
|
|
[libs/RVL_SDK/src/fa/driver/sd_drv] baseline: code 10068/11760 data 3592 functions 23 fuzzy 99.1582
|
|
|
|
|
[libs/RVL_SDK/src/fa/pdm_partition] pool: IDENTICAL
|
|
|
|
|
[libs/RVL_SDK/src/fa/pdm_partition] objdiff: code 3380/3716 data None/None functions 19/20 fuzzy 98.8601 linked code 0
|
|
|
|
|
[libs/RVL_SDK/src/fa/pdm_partition] instruction-exact functions: 19/20
|
|
|
|
|
[libs/RVL_SDK/src/fa/pdm_partition] section .text size 3716 match 98.86006
|
|
|
|
|
[libs/RVL_SDK/src/fa/pdm_partition] below 100: pdm_part_is_master_boot_sector 87.39286
|
|
|
|
|
[libs/RVL_SDK/src/fa/pdm_partition] baseline: code 3380/3716 data None functions 19 fuzzy 98.8601
|
|
|
|
|
regressions vs baseline: 0
|
|
|
|
|
global matched_code_percent: 93.65019 -> 93.79202
|
|
|
|
|
global fuzzy_match_percent: 99.79623 -> 99.79994
|
|
|
|
|
global complete_code_percent: 79.26653 -> 80.08451
|
|
|
|
|
global matched_data_percent: 99.99410 -> 99.99410
|
|
|
|
|
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
|
|
|
|
|
```
|
|
|
|
|
|