mirror of
https://github.com/boykopovar/AnyPS5.git
synced 2026-10-29 12:15:39 +00:00
Dean
2a16d79fb9
fix(libSceCesCs): report the length of a rejected MBCS character as the console does
26 lines
15 KiB
Markdown
26 lines
15 KiB
Markdown
# libSceCesCs technical debt
|
|
|
|
### Silent stubs
|
|
|
|
- [sceCesMbcsUcsContextInit](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) stores the profile pointer in the first 8 bytes of the caller's context; the console's context layout is not known. It returns the invalid-parameter error for a null context and, as a payload process on a PS5 (firmware 13.60) did, for a null profile. The signature `int sceCesMbcsUcsContextInit(union* context, const union* mbcsUcsPrf)` follows the [Orbital IDA type database](https://github.com/AlexAltea/orbital/blob/master/tools/ida/db_types.json). `sceCesRefersUcsProfileCp1251` returns a profile accepted by `sceCesSbcToUtf8`, `sceCesUtf8ToSbc` and the MBCS context conversions, with the Windows-1251 table of Microsoft's [bestfit1251.txt](https://unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WindowsBestFit/bestfit1251.txt), except byte 0x98, which is unassigned
|
|
- [EUC-JP (CP51932) conversions](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs): `sceCesUcsProfileInitEucJpCp51932` marks the caller's sheet with an 8-byte tag and returns it as the profile. Decoding follows the console (see the `sceCesEucJpToUtf8` entry), with the Microsoft code page 932 table read through the JIS row/cell arithmetic ([JisX0208.hpp](../../../core/libs/prx/libSceCesCs/JisX0208.hpp)); the console does not decode or encode JIS X 0212 either way (U+02D8 answered `0x805C0020`). Encoding takes the lowest cell of a character. Signatures follow [orbital](https://github.com/AlexAltea/orbital/blob/master/tools/ida/db_types.json)
|
|
- [Shift-JIS (CP932) conversions](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs): decoding matches a payload process on a PS5 (firmware 13.60) for every two-byte input through `sceCesMbcsStrToUcs2Str`, in result, lengths and output. `sceCesUcsProfileInitSJis1997Cp932` tags the caller's sheet and returns it as the profile. The string conversions decode ASCII, half-width katakana (`0xA1`-`0xDF`) and two-byte sequences (`0x81`-`0x9F` and `0xE0`-`0xFC` lead, `0x40`-`0x7E` and `0x80`-`0xFC` trail) through the CP932 table ([JisX0208.hpp](../../../core/libs/prx/libSceCesCs/JisX0208.hpp)), including the IBM extensions `0xFA40`-`0xFC4B`. The user-defined area `0xF040`-`0xF9FC` and unassigned cells answer `UNASSIGNED_CODE`, an invalid lead or trail `INVALID_ENCODE`, a cut sequence `SRC_BUFFER_END`. Signatures follow [orbital](https://github.com/AlexAltea/orbital/blob/master/tools/ida/db_types.json)
|
|
|
|
### Unknown function info
|
|
|
|
- [Unicode scalar conversions](../../../core/libs/prx/libSceCesCs/Unicode.cpp) (libSceCesCs) - console error precedence, optional length pointers and length units are unverified. Signatures follow the [Orbital type database](https://github.com/AlexAltea/orbital/blob/b321fcdbd700cc02da1cb1ad25412781d5ec73c6/tools/ida/db_types.json); names are in the aerolib table pinned by `tools/nid_names.py`. Encoding follows [Unicode section 3.9](https://www.unicode.org/versions/Unicode17.0.0/core-spec/chapter-3/). Each call converts one scalar, including zero and U+FEFF, without a terminator or byte-order detection; lengths count bytes for UTF-8 and 16-bit units for UTF-16. Native forms use host byte order (supported hosts are little-endian). Errors follow the existing CES converters, not console measurements, and failure leaves outputs unchanged. No console or title comparison has been run
|
|
- [sceCesEucJpToUtf8 and sceCesEucJpToUcs2](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) with the CP51932 profile follow a payload process on a PS5 (firmware 13.60): every two-byte input and every `0x8F` three-byte input gives the console's result, length and output (the test pins a hash of both sweeps). Bytes below 0xA0 other than 0x0E, 0x8E and 0x8F convert to themselves; JIS X 0212 is not decoded, so `0x8F` forms return `0x805C0020` in rows X 0212 leaves empty and `0x805C0014` elsewhere; a null profile returns `0x805C0004`, and the lengths are cleared first and then set as the console sets them. Their signatures follow the Orbital type database linked above. The console reads any other profile pointer as a sheet, where these return the invalid-parameter error.
|
|
- [Shift-JIS and EUC-JP scalar conversions](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) - errors, their precedence and the optional length pointers of `sceCesSJisTo*`, `sceCesEucJpTo{Utf16,Utf32}*` and `sceCes{Ucs2,Utf8,Utf16,Utf32}*To{SJis,EucJp}` are unverified against the console. Signatures follow the Orbital type database linked above; the CP932 and CP51932 table and profile assumptions above apply. A profile of another encoding answers `INVALID_PARAMETER`, a surrogate or a value above U+10FFFF `INVALID_ENCODE`, a code point outside the BMP `UNASSIGNED_CODE`. UTF-8 input is decoded as `sceCesUtf8ToSbc` decodes it, so overlong forms answer `0x805C0015`. `sceCesSJisCodeToJisX0208` and `sceCesJisX0208ToSJisCode` do not check that the cell is assigned; their `INVALID_ENCODE` for rows or cells outside 1-94 and `INVALID_DST_BUFFER` for a null output are not console measurements
|
|
- [Strict JIS X 0208 profiles](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) - cell `0x2140` and U+FF3C throw: JIS0208.TXT gives U+005C, glibc 2.39 and Python's `euc_jp` U+FF3C. In `sceCesUcsProfileInitSJis1997X0208`, bytes `0x5C` and `0x7E` and U+005C, U+007E, U+00A5 and U+203E throw: SHIFTJIS.TXT gives JIS X 0201 Roman, CP932 and Python's `shift_jis` ASCII; the console's is unknown. The table, [JIS0208.TXT](https://unicode.org/Public/MAPPINGS/OBSOLETE/EASTASIA/JIS/JIS0208.TXT), is CP932's without rows 13 and 89-92 and with six cells changed, not measured. Per [SharpProspero](https://github.com/SvenGDK/SharpProspero/blob/5dfdb4b0b166e83dc9ae275001afb695f12c7174/src/SharpProspero/Interop/Text/Ces.cs#L57-L67), the `X0208Ss2` profile adds SS2 katakana and `X0208Ss2Ss3` SS3 JIS X 0212. Signatures follow Orbital; the conversions above take these profiles with the same errors, a missing SS2 or SS3 answering `INVALID_ENCODE` and Shift-JIS leads `0xF0`-`0xFC` `UNASSIGNED_CODE`
|
|
- [UTF-16 and UTF-32 forms of the single-byte and MBCS string conversions](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) - errors, their precedence and the optional length pointers of `sceCesSbcTo{Ucs2,Utf16,Utf32}*`, `sceCes{Ucs2,Utf16,Utf32}*ToSbc`, `sceCesMbcsStr{To,Get}{Utf16,Utf32}*` and `sceCes{Utf16,Utf32}Str{To,Get}Mbcs*` are unverified against the console. Signatures follow the Orbital type database linked above. They follow `sceCesSbcToUtf8`, `sceCesUtf8ToSbc` and the UTF-8 MBCS string forms with the profiles and tables listed above. A surrogate or a value above U+10FFFF answers `INVALID_ENCODE`, a valid code point no profile maps `UNASSIGNED_CODE`. The string forms stop at a zero code unit, write no terminator and count code units
|
|
- [GBK (CP936) and EUC-CN conversions](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) - `sceCesUcsProfileInitGbkCp936` and `sceCesUcsProfileInitEucCnGb2312` mark the caller's sheet with an 8-byte tag; the console's profile layout is not known. Both use one table ([Cp936.hpp](../../../core/libs/prx/libSceCesCs/Cp936.hpp)) read from Microsoft's [CP936.TXT](https://unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WINDOWS/CP936.TXT); the EUC-CN profile accepts only its `0xA1`-`0xFE` lead and trail bytes, so it includes the GBK cells inside that range and the CP936 forms of `0xA1A4` (U+00B7) and `0xA1AA` (U+2014) instead of GB 2312's. That the console's GB 2312 and CP936 tables equal it is not measured. User-defined areas and `sceCesGbUcsProfileSet*` are not handled. Errors, their precedence and the length pointers follow the Shift-JIS forms above and are unverified
|
|
- [GB 18030 profiles](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) - the console's tables are not measured, and the sources disagree on 26 sequences. The 2000 profile follows ICU's `gb-18030-2000.ucm` at icu-data d9d3a6ed (before ICU-8427), which matches Python's codec: `0xA8BC` is U+E7C7 and `0x8135F437` U+1E3F. The 2005 profile follows `gb-18030-2005.ucm` (91850aec), which swaps those two and maps 24 PUA cells to U+FE10-U+FE19, U+9FB4-U+9FBB and six supplementary ideographs, keeping the PUA code points as encode fallbacks. Two-byte cells are CP936 plus [Gb18030.hpp](../../../core/libs/prx/libSceCesCs/Gb18030.hpp); four-byte runs include ICU's `gb18030Ranges`. The inits tag the sheet like GBK's; the `sceCesGb*`, `sceCesMbc*` and MBCS string forms accept them. The full mapping is assumed as the default of `sceCesGbUcsProfileSetUdaMapping` and `Set4ByteCharRange`, which throw. Supplementary characters to UCS-2 answer `UNASSIGNED_CODE`; other errors follow the GBK forms
|
|
- [sceCesRefersUcsProfileCp1250](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) - returns a profile accepted by `sceCesSbcToUtf8`, `sceCesUtf8ToSbc` and the MBCS context conversions, with the Windows-1250 table of Microsoft's [bestfit1250.txt](https://unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WindowsBestFit/bestfit1250.txt), except bytes 0x81, 0x83, 0x88, 0x90 and 0x98, which are unassigned. PPSA34501 fails to start without this export (`qYu6y1E0Qbg`)
|
|
- [sceCesSbcToUtf8 and sceCesUtf8ToSbc](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) follow a payload process on a PS5 (firmware 13.60): the three tables match the console byte for byte, with Cp1252 bytes 0x81, 0x8D, 0x8F, 0x90 and 0x9D unassigned; a null profile returns `0x805C0004`; and the length and output are cleared first, as the console does. `sceCesUtf8ToSbc` decodes leads up to 0xFD (5- and 6-byte forms), returns `0x805C0015` for overlong forms and surrogates and reports the measured lengths. The console reads any other profile pointer as a table, where these return the invalid-parameter error; which of a null profile and a scalar above U+10FFFF it reports first is not measured
|
|
- [MBCS string conversions](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) - `sceCesUtf8StrToMbcsStr`, `sceCesUcs2StrToMbcsStr`, `sceCesMbcsStrToUtf8Str`, `sceCesMbcsStrToUcs2Str` and their `Get*Len` forms follow a payload process on a PS5 (firmware 13.60) with the Cp1252 and EUC-JP profiles: the lengths are cleared and a terminator is written first, then the context, the source and the output are checked in that order; a source limit of 0 reads up to the terminating zero; the output keeps room for a terminator; and on an error the characters before it stay converted, terminated and counted. They decode UTF-8 as `sceCesUtf8ToSbc` does, and a UTF-16 surrogate answers `0x805C0015`. The UCS-2 forms and the other profiles were not measured beyond these cases. A zero-filled context converted nothing and returned 0 on the console, where this returns the invalid-parameter error
|
|
- [EUC-JP and Shift-JIS encoding](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) with the CP51932 and CP932 profiles gives the result, lengths and bytes a payload process read on a PS5 (firmware 13.60) for every UCS-2 value through `sceCesUcs2StrToMbcsStr` (the test pins a hash of both sweeps): the JIS forms U+00A2, U+00A3, U+00A5, U+00AC, U+2014, U+2016, U+203E, U+2212 and U+301C encode to the cells of their Microsoft forms, U+FEFF is consumed and writes nothing, EUC-JP writes U+0080-U+009F other than U+008E and U+008F as single bytes and answers `UNASSIGNED_CODE` for U+000E, and Shift-JIS answers `0x805C0024` for U+0080-U+009F. Decoding those bytes back is unchanged
|
|
- [Generic MBCS character and Shift-JIS code helpers](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) - UTF-8/UTF-16 buffer errors and size pointers are unverified. Measured on a PS5: the single-character encoders (`sceCesUtf32ToMbc` over U+0000-U+110001 per profile, surrogates, a one-byte target) give 0x805C0015 for a surrogate, 0x805C0014 past U+10FFFF, the needed length with `DST_BUFFER_END` and 0 otherwise; a rejected decoded character (`sceCesMbcTo*` and each encoding's `To{Ucs2,Utf8,Utf16,Utf32}`, not `sceCesEucJpToUcs2`/`ToUtf8`) stores its length (the bytes before a NUL after the lead, the needed length when truncated, 1 for a bad lead, the sequence for a bad or unassigned trail), a zero UCS-2/UTF-32 character and no UTF-8/UTF-16 units. `sceCesMbcTo*` and `sceCes*ToMbc` take any profile the MBCS context accepts. `sceCesSJisGetCode`/`PutCode` are unchecked; `sceCesSJisCodeToJisX0213` and back use JIS X 0213 annex 1. Signatures follow Orbital
|
|
- [Single-byte profiles](../../../core/libs/prx/libSceCesCs/SingleByte.hpp) (libSceCesCs) - the console's tables for these 27 profiles are not verified. Each returns a profile accepted by `sceCesSbcToUtf8`, `sceCesUtf8ToSbc` and the MBCS context conversions, with bytes 0x00-0x7F unchanged and the high half of a public table: ISO 8859-1, -2, -4, -5, -9, -10, -13, -14, -15 and -16 from [unicode.org](https://unicode.org/Public/MAPPINGS/ISO8859/), code pages 437, 737, 775, 850, 852, 855, 860-863, 865 and 866 from [Microsoft's PC tables](https://unicode.org/Public/MAPPINGS/VENDORS/MICSFT/PC/), KOI8-R and KOI8-U from [unicode.org](https://unicode.org/Public/MAPPINGS/VENDORS/MISC/), and 1254, 1256 and 1258 from Microsoft's [best-fit tables](https://unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WindowsBestFit/), whose unassigned bytes map to C1 controls as in Cp1250. Profiles whose public table leaves bytes undefined or maps them to private use still throw
|
|
- [libSceCesCs firmware 3.20 pending exports](../../../core/libs/prx/libSceCesCs/PendingExports.cpp) - 62 symbol names are present in [the PS5 3.20 libSceCesCs export list](https://github.com/DNNDHH/PS5-3.20_Libs/blob/b2a77c09cd71c01f34c790cfbe98c751a13262ad/libSceCesCs.c). Their signatures and console behaviour are unverified. The declarations use the existing no-parameter unknown-signature convention solely to resolve imports, and every call throws through `NotImplemented_nid_no_patch`.
|
|
- [EUC-KR, UHC and Big5 conversions](../../../core/libs/prx/libSceCesCs/Export.cpp) (libSceCesCs) - the errors and Big5 duplicate choices are assumed; the decode tables were measured on a PS5 (`sceCesMbcToUtf32` and each `ToUtf8` over every one- and two-byte sequence): UHC and Big5 match, EUC-KR apart from its single bytes. The tables are Microsoft's [CP949.TXT](https://unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WINDOWS/CP949.TXT), plus `0xA2E8` U+327E from KS X 1001:2002, and [CP950.TXT](https://unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WINDOWS/CP950.TXT); EUC-KR is CP949 limited to bytes `0xA1`-`0xFE`, below `0x80` ASCII; the KS X 1003 profile is not implemented. User-defined areas answer `UNASSIGNED_CODE`. Big5 encoding takes the lower of two cells except for U+2550, U+255E, U+2561, U+256A, U+5341 and U+5345, as code page 950 does. The profiles tag the caller's sheet like the EUC-JP profile
|