Correct memory-card conversion constant ownership #697
Reference in New Issue
Block a user
Delete Branch "agent/bittle/memorycard-float-boundary"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Relocation-backed boundary correction
Move the MemoryCard/MemoryCardBase .sdata2 boundary from 0x81694A80 to 0x81694A78. The eight-byte double at 0x81694A78 (
4330000080000000, signed integer conversion bias) is used by MemoryBase::set_textbox at function +0x4C (SDA21 relocation). It belongs to the base unit, not MemoryCard.After correction:
Verification
Rebased and fully rebuilt on
a81b7c36fe. Only two adjacent boundaries in config/43U/splits.txt change; symbol addresses/sizes and C/C++ sources are untouched. All affected source .text bytes remain identical. Full 43U all_source/progress/report/ok passes; DOL SHA1 is 26116613f624061ba99c8d1a299aaa6efa85670d. Workflow guards 7/7, pool tests 8/8, jump-table tests 20/20; worktree checks pass. ctxdiff uses the disclosed local pyelftools/Capstone compatibility adapter. No new padding, assembly or weak-vtable changes. No code/link percentage gain claimed; affected units were already linked. Left for Cole's review and merge.Rebased the unchanged boundary correction onto
7a24a5c1after #698. Fresh all_source/progress/report/ok passes and the DOL SHA1 remains 26116613f624061ba99c8d1a299aaa6efa85670d. Tested local tree85911bc753equals the new remote tree. Heada371d4842f.Reviewed and merged: relocation-backed split boundary correction, splits.txt only. The gate's complete_data drop is the bytes moving to the adjacent unit (total unchanged, both units' matched data now 100%). Full build ok, DOL sha1 26116613f624061ba99c8d1a299aaa6efa85670d. Good evidence.