forked from Ports/goldeneye-pc-port
James Dansereau
bbc32957e1
Docs-only. Records the mod-compatibility investigation so it is not re-derived when mod support is picked up. No code changed. Measured two shipped community mods against a vanilla NTSC-U ROM: - Goldfinger 64 v1.0 (12.8 MB xdelta): ROM 12 -> 24 MB, 797/803 vanilla filelist.u.csv slices differ. - Random Eye v1.1 (286 KB xdelta): ROM size byte-identical, yet 96.5% of bytes and 803/803 slices differ; content preserved but shifted (+7216/+7392, one file -229888). Both die at tools_pc/d43_emit.py:273 with a zlib error on a slice taken at a hardcoded vanilla offset. Cause: the file table is compiled in (ob.c -> file_resource_table.inc.c -> romassets_<region>.s) and the converters reuse the same vanilla offsets. ROM size is not the constraint; ROM code is never executed, so ASM hacks are inert regardless. D325 - Goldfinger 64 probe + mod taxonomy. D326 - Random Eye repacks too, so D325's hypothesised in-place tier is empty. D327 - plan settled: reading the table from the ROM is ruled out by measurement (hw_address absent in all four encodings tested, 0 hits for the segment-relative form; two structural fileentry scans found nothing); content-addressed offset remapping with gap inference is the viable route, since obInit differencing proves files are contiguous. Also recorded: GE maps its 0x7F000000 game segment directly from ROM, uncompressed, uniquely signature-findable at ROM 0x34b30. The 1964GEPD fork reviewed in D327 is GPL-2.0 against this MIT tree; no code was copied or vendored, facts re-derived from our own ROM and decomp.