port 2D step 5a: EterPack/EterPackManager + EterBase file units against the real 40250 packs
- Mirror copies: EterPack (EterPack, EterPackManager, EterPackCursor, CSHybridCrypt policy, Inline.h)
and EterBase logic units tea/lzo/Timer/Stl/Random, with PORT-tagged width fixes (LONG index
fields, static_assert(sizeof(TEterPackIndex) == 192), tea_word = Win32 unsigned long).
- TEA pack keys are extracted from the reference EterPack.cpp at configure time into
build/.../EterPackKeys.generated.h; nothing key-bearing is tracked.
- Platform EterBase: FileBase (stdio, case-insensitive fallback), MappedFile (mmap /
MapViewOfFile), CRC32, Debug, Utils StringPath/StringLowers.
- Win32Crt: CreateDirectory/DeleteFile/_access/MAKEFOURCC; Win32MinMax.h min/max (force-included
on MinGW, whose windows.h omits them in C++).
- tests/port_eterpack_test: registers pack/Index like PackInitialize and checks 119
registrations / 103 packs + root / 54891 entries / 52609 paths / 2282 overrides with
first-registered-wins, and that exactly the 4 short-layout SECURITY files fail to load.
- 2R count corrected (136 -> 119 registrations, 103 distinct packs + root).
Not runtime-reachable yet (asset_io pack backend is step 5b), so port-map statuses stay TODO.
port_gate: macos/android/ios/windows PASS.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>