v bugs

KNOWN BUGS / ISSUES

To be fixed, probably.

Known issues are separate as aren't always possible to fix, not really a bug, too hard/time consuming, or low interest.

general

general issues

average bitrate is borked for some metas with subsongs
- codec/bitrate breaks when loading segmented stream as a part of another segmented stream
- segments + layers have incorrect bitrate (EAXMA loop)
- complex .txtp mixes
plugins: seek code could be faster
extensionless files don't work inside folders with extension in foobar ("folder.etc/file", "folder.etc\file")
mpg123 forces endianess/aligment in 8-bit buffer, problematic with some systems?
log: can't write long lines (unknown bug)
cleanup: crlfs + tabs

known general issues

codecs: ADPCM variants (ex. DSP, IMA, PS-ADPCM) may need to preserve history when looping (hard to test)
coding_utils: ffmpeg_make_riff don't always create proper RIFF headers (ok for now but may break if FFmpeg changes stuff)
code: performance could be improved, generally unsafe/brittle
- not very interesting or useful, improved little by little
code: warning with disabled external libs could be improved
- wrapping more stuff into would result in uglier code just to catch
- preferably should not disable in meta, but in decoder (init_x fails if not implemented)
  - should set a init_x callback to "noop" (same for other decoder calls)
common: dual extensions (ex. ".cdk.wav") not supported
- not many known, not that easy to detect
common: Gamecube games (unlike Wii) sample rate should be slightly higher due to a hardware bug
- instead of 48000/32000 it's closer to ~48043/32029
  - https://dolphin-emu.org/blog/2017/11/03/dolphin-progress-report-october-2017/
  - https://github.com/dolphin-emu/dolphin/pull/6040
- can't detect in many cases Wii or GC, and not all games are coded expecting this
  - ex. Killer7 sounds faster on GC vs PS2, but on files sounds the same
- some formats already have sample rate 0xBBAA/48042, acknowledging this (h4m videos)
- should just override with TXTP
os: Linux can't open companion with different case
- hard and messy to fix, rename/symlink as needed
- could try opening companion file with same caps? (.ABK > .AST)
32/64 diffs due to float rounding (+- samples in play settings/etc)
- ao\Cloudphobia (PC)[OGG]\FallAp.ao
- ao\GEO ~~The Sword Millennia~~ Kasumi no Tani no Kaibutsu (PC)[OGG]\town.ao
- ogg-opus\Monster Boy
- opusx\Touhou Genso Wanderer -Reloaded- (Switch)[OPUS]\BGM_8.opusx
- ubi-dat\Rayman 2 (DC)[AICA]

components

components issues

make: ZIP target could include licenses
foobar: print loop times as xx:xxx (standard)
winamp: unicode build problems with UNICODE Windows flag (MSVC default)
winamp: improve conversions wpath to path, etc
winamp: for $AUTOTRACK doesn't like TRACKNUMBER (more common) and wants TRACK
audacious: some compiler warnings could be improved
audacious: missing C flags? "-O3 -ffast-math" etc
audacious: should be named audacious-vgmstream
CLI: files with unicode names don't work in Windows CMD (messy to fix and preserve multiplatform code)
- wmain/WinMain: get wchar args, convert to utf8 (MultiByteToWideChar), pass to normal main
- internally use utf8
- when opening new files (WSTREAMFILE?), convert and use _wfopen
- printf convert to widechar before print too?
- https://github.com/thpatch/win32_utf8
cmake: missing flags in external libs?
- some calls seem to overwrite them

known component issues

foobar: can't write to wave virtual txtp
- not fixable?
audacious: can't support virtual txtp
- blocked by probe.cc
xmplay: no config dialog
- no requests to fix it...
xmplay: can't use subsongs
- use TXTP
- xmplay expects that channels/sample rate doesn't change, which isn't true for vgmstream's subsongs
- see xmp-gme for a possible example
- accepts ":// (file).txtp" if flag XMPIN_FLAG_OWNFILE is set
- workarounds:
xmplay: extension list bigger than ~1000 crashes the options panel (~v3.8.2.17)
xmplay: GCC linking can be improved (GCC complaints)
- not important (check openmpt for tips)
xmplay: add tags
- not very used
AIMP plugin:
- https://github.com/ArtemIzmaylov/aimp_vgmstream
- https://github.com/lltcggie/vgmstream/tree/aimp
Deafbeef plugin
- https://github.com/jchv/deadbeef-vgmstream
- https://bitbucket.org/losnoco/deadbeef_plugins/src/master/plugins/vgm/

codecs

bugs

PS-ADPCM/XA: not fully accurate (implementations vary, see xa_decoder.c)
interleave_short_frame won't work in many cases of variable frame size (rarely used in practice)
Guitar Hero 2 last stream of half frame rate ignored in some files
reset_mpeg may need to work like flush_mpeg (to be tested)
Juiced 1/2 (PC): use giant .dsb WMA and some kind of offset of unknown format to locate sections
Metal Gear Solid Twin Snakes strange voice ogg
FFmpeg bugs:
- simulated looping, potentially buggy with unknown formats
- simulated Opus encoder delay
- simulated ATRAC3 encoder delay
- wrong encoder delay: ALAC, MP4, etc
- xwma 1ch 32000hz wrong (wma looks ok)
- "-Wcast-align" not working (can't do aligned/unaligned memory)
- inverted signal in ATRAC3 decoding
- internal OPUS
- can't seek at all in some formats like .smk (even on video players)
- broken XMA1 seeking leading to inf-loops in 5.1.2
  - "XMA1 doesn't need seeking", silently fixed after pointing out internals need flushing
- mistakes in ADPCM codecs like DIV vs SHR in MSADPCM (not used, just warning)
- seeking to zero breaks encoder delay and other stuff in various codecs (fixed in some cases)
fsb multichannel at9/celt faulty in some configs (unused?)
Wwise DSP ADPCM aligment in ParseAdpcmHeader
Ubi SB/BAO? weird XMA1 with wrong counter in XMA1 block header, has issues in FFmpeg
dsp: could try to reject .dsp with too much data (accounting some padding) to be mono
- problem: some headers lie/set nibble counts as if they were stereo about [ESPN 2002]
```
if ((header.nibble_count / 2) + 0x4000 < get_streamfile_size(streamFile) - 0x60)
    goto fail;
```
EALayer3: free bitrate misdetected by mpg123 at times
- force free bitrate for ealayer3 by sending fake 2 free silent BR packets?
KTAC: buggy mode 1
ealayer3 causes lots of restarts
- complex to fix ATM, would need to save granules
ADX: figure out frame size value what does exactly
some decoders should be implemented in float math (MTA2, EA-XA, etc)
ADPCM (ex. psx) use hist_16/32 more consistently
FFmpeg less frames to detect? use less IO buffer?
PSX optimize: maybe upper filters just set 0s in table
westwood clean decoder/meta
clean psx decoding old, doc about clamping
ima/xbima/msima/msadpcm optimize frame read
.ogg in Undertale have malformed granulepos in several of the files
- would need detection to set discard seeks
awc: try if block have more frames than usual (rather than next block having extra frames?)
- rdr2 music desyncs

new

.H4M + AFC: used? [Picmin (GC)]
ESS: unsupported Eugen Systems' games [RUSE, Wargame, Act of Aggression]
- decoder: http://forum.xentax.com/viewtopic.php?f=17&t=15423
Codemasters ADPCM: unsupported [DIRT, GRID]
- decoder: http://forum.xentax.com/viewtopic.php?f=10&t=13681
Evolution engine sound [Warframe, Darkness II, Star Trek, Dark Sector]
- http://forum.xentax.com/viewtopic.php?f=17&t=13701
.MED: Missing IMA 3-bit ADPCM
.SNEO: mutant Ogg with external? Vorbis codebooks (S-Neo Dreamcast games)
- Kimi ga Nozomu Eien
  - OggVorbis Decoder v1.00 for Dreamcast Programmed by S.Nakamoto(S-NEO)
  - vorbissneo / Xiphophorus libVorbis I 20011231 / BCV_huff_book_line_NNNxNN_classN
- Aikagi, Majo no Ocha Kai, Tamakyuu
  - OggVorbis Decoder for Dreamcast v1.20 Programmed by S.Nakamoto(S-NEO)
  - vorbissneo / Xiph.Org libVorbis I custom 20021109 / vorbisS3 [V1.0 RC3]
UEBA: mutant bink audio [Fortnite new, Robocop]
XVAG MP3 Surround: unknown actual 2ch > 6ch mapping
adventure games in scummvm
- Tinsel adpcm: Discworld, Discworld 2: Missing Presumed ...!?
  - https://github.com/scummvm/scummvm/blob/master/engines/tinsel/adpcm.cpp
- Last Express:
  - https://github.com/scummvm/scummvm/blob/master/engines/scumm/he/sound_he.cpp
- etc
.xwb: strange DSP (2-bit?) [Blossom Tales (Switch)]
EALayer3: free bitrate misdetected by mpg123 at times
- force free bitrate for ealayer3 by sending fake 2 free silent BR packets?
ea sx.exe new features (loop in sps)
ATRAC9: new decoder modes
Ogg: include internal decoder?
- https://github.com/nothings/stb/blob/master/stb_vorbis.c
mp2.bin custom in various shovelware games

encryption

Rome 2 Total War: XORed WEMs
Tales of Asteria
.rpgmvm: RPG Maker MV files M4A (iOS/Android?), XOR'ed with 0x10 key in System.json
Rock band (PSP) encrypted .ATM
Guitar Hero Legends encrypted *.wem
.sec SITERSKAIN: xor encrypted?
Zone4 .sod strange character swap table
Darkwind (PC)
Guitar Hero 6 Warriors of Rock custom? FSB encryption
hca phantasy star 2 bad key detection

silent files but uses a bad key with high score forcing min test frames seems to work, but also makes key detection slower

sample issues

some formats loop_end may be 1 sample off, depending if loop end means "loop after playing this sample" or "loop when reaching this sample"
- ex. some files suspiciously use "loop_end=123455, num_samples=123456")
encoder delay to be researched/fixed:
- XMA: most formats should be fine work but there may be bugs
- ATRAC3: GSB+GSP, RXWS, tri-Ace AAC, VAWX, XWB
- MPEG: SCD, XVAG, MSF (probably not used)
- ATRAC9: various
loop end +1 in XMA and AT3 in psb, maybe others
fix: dsp_bytes_to_samples, DSP frames can end before 0x08?
mp3 counter: ignore frames with only pre-data
wma v2 sample counts
xma encoder delay bugs (not reading skips ok?): xwb xma1, eaxma, gtd (glitch at the beginning)
#t trim command has issues with interleave last b/c total samples change and calcs depend on those
OGL Vorbis: num_samples in non-looping files is a bit off (approx calcs)
wma_get_samples wrong for 6ch 00000162 xwma konami
BINK: bink total samples is slightly off vs binkconv
opus number of samples maybe fixed for all frames

known improvements

For fun but not really useful

TXTP inline segments/layers (for mini/virtual txth)
- file > file: segments
- file : file: layers
- virtual txtp are limited so not really useful (maybe for CLI)
- which chars to use for all OSs?
TXTP: some kind of remap ea that would change FL FC FR BL BR LFE to standard FL FR FC LFE BL BR
- better as a mixing related macro?
TXTH/TXTP: force internal name
- ever useful?
TXTH: add xor mode/key? (IQ Remix)
- too limited, better pre-decrypted
builds: use DLL delayed loading configurable during compile (substitute functions)
- not very compatible and kinda strange
- better build single lib
calculate codec_bitrate for CBR codecs:
- let codecs set frame_size (ex. NXOPUS)
- get_frame_size can't be used since some VBR codecs report frame_size (ex. FFmpeg)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

v bugs

KNOWN BUGS / ISSUES

general

general issues

known general issues

components

components issues

known component issues

meta

meta issues

known meta issues

codecs

bugs

new

encryption

sample issues

known improvements

Clone this wiki locally