mirror of
https://github.com/hrydgard/ppsspp.git
synced 2026-04-24 21:56:10 +03:00
[GH-ISSUE #4248] PPSSPP: Atrac3+ related issues after the merge of maxim-at3plus branch to master #1726
Labels
No labels
Atrac3+
Audio
CPU emulation
D3D11
D3D9 (removed)
Depth / Z
Feature Request
Font Atlas
GE emulation
Guardband / Range Culling
HLE/Kernel
I/O
Input/Controller
MP3
Multithreading
Needs hardware testing
Networking/adhoc/infrastructure
No Feedback / Outdated?
OpenGL
PGF / sceFont
PSMF / MPEG
Platform-specific (Android)
Platform-specific (Windows)
Platform-specific (iOS)
PowerVR GPU
SDL2
Saving issue
User Interface
Vulkan
arm64jit
armjit
armv6
x86jit
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
starred/ppsspp#1726
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
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?
Originally created by @solarmystic on GitHub (Oct 19, 2013).
Original GitHub issue: https://github.com/hrydgard/ppsspp/issues/4248
The recent merge of maxim's atrac3+ branch into master by @hrydgard is massive one, and brought along with it major improvements, but there could still be bugs related to it that needs flushing out.
I've decided to open this issue report as a catch-all for such bugs in games, which may not present themselves until wider testing is done by the PPSSPP community now that the master has been updated to include the new atrac3+ patch.
I'll start the ball rolling by presenting my findings. Other users are also welcome to report their at3+ related issues evident since the merge for other games here.
@hrydgard commented on GitHub (Oct 19, 2013):
Thank you for your initial testing!
All considered, I think it's gone very well so far, of course much thanks to Maxim's amazing work properly figuring out Atrac3+ and implementing it in ffmpeg!
@solarmystic commented on GitHub (Oct 19, 2013):
Further details about the GTA games here https://github.com/hrydgard/ppsspp/issues/3805#issuecomment-26659765 https://github.com/hrydgard/ppsspp/issues/3805#issuecomment-26660662 and for Trails in The Sky https://github.com/hrydgard/ppsspp/issues/4003
@daniel229 commented on GitHub (Oct 19, 2013):
Digimon Adventure crashes before the title.Before build perfect.
Super Robot Taisen OE crashes in the title,if disable fastmemory,the game can goes a little farther,but the BGM speed is too fast,and still crashes.Before build perfect run with or without fastmemory.
Last Ranker crashed https://github.com/hrydgard/ppsspp/issues/3122 is fixed,but the BGM still cut off.
Fairy Tail Zelef Kakusei JPN crashed is fixed,seems to be perfect now.
I have tested over 30 games,others seem to be fine.
Test on Win7 64bit with v0.9.1-2005-gdea4668 both X64 and Win32 build that download from orphis's site.
@unknownbrackets commented on GitHub (Oct 19, 2013):
Do you have a trace of either of the crashes?
-[Unknown]
@daniel229 commented on GitHub (Oct 20, 2013):
Is a stack trace in VS 2013 useful.
@unknownbrackets commented on GitHub (Oct 20, 2013):
Sure, although I wonder if the compile of ffmpeg will match a 2013 build...
-[Unknown]
@brujo5 commented on GitHub (Oct 20, 2013):
maxim's atrac3+ working great in all my games.
perfomance increased
sound/fx very clean
custum layout control is a great feature for mobiles.
people will love 0.9.5 in playstore.
@daniel229 commented on GitHub (Oct 20, 2013):
I tested the 32bit build.
Digimon Adventure show nothing.
Super Robot Taisen OE with fastmemory
Super Robot Taisen OE without fastmemory
@unknownbrackets commented on GitHub (Oct 20, 2013):
Hmm, workarea is null there? Hmm, it doesn't properly check for that. But still, it's a problem if this is happening now and wasn't before.
One of those (Digimon Adventure) looks similar to an error I got in Shadow of Destiny.
-[Unknown]
@sum2012 commented on GitHub (Oct 20, 2013):
My freind report One Piece ROMANCE DAWN (NPJH50679) will crash after opening video
@daniel229 commented on GitHub (Oct 20, 2013):
One Piece
@unknownbrackets commented on GitHub (Oct 20, 2013):
If it's not too slow, a stack trace with jit disabled would be interesting (unless it's the same.)
-[Unknown]
@daniel229 commented on GitHub (Oct 20, 2013):
stack trace for One piece with jit disabled
@daniel229 commented on GitHub (Oct 20, 2013):
stack trace for Digimon Adventure with jit disabled.
@unknownbrackets commented on GitHub (Oct 20, 2013):
Well, you can put at the top of
__KernelLockLwMutexsomething like:But, really, it seems like a buffer overflow (?) or something.
-[Unknown]
@daniel229 commented on GitHub (Oct 20, 2013):
with that changed.
One Piece and Digimon Adventure ,the debug console continue showing this a lot.
CRI Server M W[MM]: MemmapFunctions.cpp:95 ReadFromHardware: Invalid address 13f706bb PC 08abb9c0 LR 08abb9c0
Digimon Adventure still crash with release build, debug build is too slow to wait.
One Piece is working fine,but missing BGM. (maybe all sound,I do not hear anything)
@unknownbrackets commented on GitHub (Oct 20, 2013):
Okay, so there's somewhere probably where it's writing outside the expected buffer (maybe writing stereo instead of mono? or writing too large a packet?)
-[Unknown]
@sum2012 commented on GitHub (Oct 20, 2013):
How to make log now ?
Log.bat don't more work
@unknownbrackets commented on GitHub (Oct 20, 2013):
See #4239.
-[Unknown]
@internetakias commented on GitHub (Oct 20, 2013):
https://github.com/hrydgard/ppsspp/issues/3650#issuecomment-26663842
@sum2012 commented on GitHub (Oct 20, 2013):
@unknownbrackets
This is with your change debug log (Digimon Adventure)
https://docs.google.com/file/d/0B3OaSdeV0L8kT1BEYnZGRTZXdFE/edit?usp=sharing
@unknownbrackets commented on GitHub (Oct 20, 2013):
What are the values of
atrac->pFrame->nb_samples,atrac->pFrame->channels, anddecodedinsceAtracLowLevelDecode(insideif (got_frame) {)?Also, does it help (I don't expect it to...) to change this?
To:
-[Unknown]
@papel commented on GitHub (Oct 20, 2013):
I see that https://github.com/hrydgard/ppsspp/pull/3772 introduced a bug that crashes many games. The Maxim's decoder avoids the crash, but the music is not played.
Therefore, most of these bugs are duplicate
Puyo Pop Fever: 3650
Grand Theft Auto Chinatown Wars: 3805
Last Ranker: 3122
Trails in the Sky: 4003
A Japanese game: 3796
Class of Heroes 1&2: 4250
Ys Seven (https://github.com/hrydgard/ppsspp/issues/2391) was the only known game fixed by that commit. It crashed in some areas on Win64, but did not crash on Win32. I do not know about Android.
To revert that commit, just change the line:
bufferInfo->first.filePos = Sampleoffset;
to
bufferInfo->first.filePos = atrac->first.fileoffset;
I wonder how Ys Seven would be now if this change were reverted. I cannot test, because I would have a bad time downloading the new ffmpeg. I tested with older revision (that uses the plugin) and it fixes Trails in the Sky and breaks Ys Seven.
@daniel229 commented on GitHub (Oct 20, 2013):
@unknownbrackets
how to find out the value?
That change fixes all these crashes
@daniel229 commented on GitHub (Oct 20, 2013):
break point shows these values
@unknownbrackets commented on GitHub (Oct 20, 2013):
@papel rather, I think that change introduced correct behavior that code elsewhere in sceAtrac.cpp (but possibly in the same function) is not handling properly, revealing incorrect behavior. Before, it was always "resetting" to the end of the file. I don't think I've even been able to find a way to get an actual PSP to give such an invalid value.
@daniel229 which change, the
numSamplesone? Or making it so thatsceAtracGetBufferInfoForResetting()always incorrectly returns the end of the file?Hmm,
decodedthere is 0xcccccccc, I think. But anyway, those values don't seem wrong assuming it is stereo.-[Unknown]
@sum2012 commented on GitHub (Oct 20, 2013):
That change still need improve (Digimon Adventure)
48:48:531 CRI Atom Atr E[ME]: HLE\sceAtrac.cpp:1666 atracID: 1, avcodec_decode_audio4: Error decoding audio -22
full log:
https://gist.github.com/sum2012/176a887e5245cc85aa9c
@unknownbrackets commented on GitHub (Oct 20, 2013):
Hmm. So that log says:
So for some reason it's misdetecting this as stereo it seems like? Or we are using it incorrectly.
-[Unknown]
@daniel229 commented on GitHub (Oct 20, 2013):
Yes,the numSamples change.
@daniel229 commented on GitHub (Oct 20, 2013):
Digimon Adventure missing voice.
@sum2012 commented on GitHub (Oct 20, 2013):
@unknownbrackets This is the second change debug log
https://docs.google.com/file/d/0B3OaSdeV0L8kR3JlRzJnZHpydnc/edit?usp=sharing
Edit:I might need test in real PSP and old version
@daniel229 commented on GitHub (Oct 20, 2013):
With that change Super robot Taisen OE missing voice too,and infinite looping in battle.Before is the same looping if disable Atrac3+ plugin.
@unknownbrackets commented on GitHub (Oct 20, 2013):
I've noticed there are some subtle differences in the non-ffmpeg and ffmpeg code (and the non-ffmpeg code was being used here before.) Hmm.
Also, volume handling is now gone.
-[Unknown]
@sum2012 commented on GitHub (Oct 20, 2013):
I have tested old version no voice (Digimon Adventure),real PSP have voice
@daniel229 commented on GitHub (Oct 20, 2013):
Digimon Adventure with the Atrac3+ plugin,the voice is working fine here on 0.91-1949.
@sum2012 commented on GitHub (Oct 20, 2013):
@daniel229 yes,maybe I test too old version
@hrydgard commented on GitHub (Oct 20, 2013):
"FF: Channel layout 'stereo' with 2 channels does not match specified number of channels 1: ignoring specified channel layout"
This could just as well mean that we're playing garbage data. However would not be surprised if we are ignoring some bit, either.
@Ritori commented on GitHub (Oct 20, 2013):
SD Gundam G Genaration OverWorld crash with new atrac3+ at Opening video (can't skip it) build without this new code work fine
I used v0.9.1-2010-g023dbbc test on window.
because log don't work anymore i post screenshot then :)
@sum2012 commented on GitHub (Oct 20, 2013):
@Ritori Log.bat has been fixed since v0.9.1-2018-gfa2846b
@unknownbrackets change fix this game too,
edit:sound do not have probrem
changed log:
https://gist.github.com/sum2012/ae7937fcf74ef9bd2fdd
@Ritori commented on GitHub (Oct 20, 2013):
@sum2012 it still crash for me at Opening video.
here log
http://www.mediafire.com/download/j14ldodb2s5uwud/Ar3plus.7z
@sum2012 commented on GitHub (Oct 20, 2013):
@Ritori
Try this build
https://docs.google.com/file/d/0B3OaSdeV0L8kS3BEX0lKdnVISVk/edit?usp=sharing
maybe my setting is wrong.Now the sound do not have problem.
@daniel229 commented on GitHub (Oct 20, 2013):
seventh dragon 2020 2 missing voice,looping after battle.
18:04:461 at3 bgm thre W[ME]: HLE\sceAtrac.cpp:1112 This is an atrac3+ mono audio
18:04:461 at3 bgm thre E[ME]: HLE\sceAtrac.cpp:1041 avformat_open_input: Cannot open input -1094995529
18:08:199 at3 bgm thre E[ME]: HLE\sceAtrac.cpp:1203 sceAtracSetDataAndGetID(0923cc40, 000014c0): no free ID
18:10:601 at3 bgm thre E[ME]: HLE\sceAtrac.cpp:1203 sceAtracSetDataAndGetID(0923ff00, 00002630): no free ID
18:12:266 at3 bgm thre E[ME]: HLE\sceAtrac.cpp:1203 sceAtracSetDataAndGetID(092425b0, 00002350): no free ID
18:17:000 at3 bgm thre E[ME]: HLE\sceAtrac.cpp:1203 sceAtracSetDataAndGetID(0923e180, 00001d80): no free ID
18:18:665 at3 bgm thre E[ME]: HLE\sceAtrac.cpp:1203 sceAtracSetDataAndGetID(0923a8f0, 00002350): no free ID
@sum2012 commented on GitHub (Oct 20, 2013):
This issue cannot close until fix voice probrem
@solarmystic commented on GitHub (Oct 20, 2013):
@sum2012
I don't intend to close it until all the BGM cutoff (audio missing instead of game crashing) bugs remaining in the various games reported are fixed. Not to mention other bugs and issues that have yet to be reported.
Wider release of maxim's patch with 0.9.5 will probably bring in even more reports, and this issue page will useful to unify them.
@sum2012 commented on GitHub (Oct 20, 2013):
@solarmystic ok
@unknownbrackets commented on GitHub (Oct 20, 2013):
True, but it seems like a lot of the things with issues are mono. Also, at least a few of the games with issues do not import many atrac functions (according to logs, they don't import any of the resetting functions, etc., and they apparently worked before the merge.)
-[Unknown]
@internetakias commented on GitHub (Oct 20, 2013):
@papel
The PPF (or at least PPF2) bug was happening way before Unknown's commit was merged into the master branch.
EDIT: Disabling Atomic Audio Lock seems to have fixed the various audio glitches in PPF2. The only remaining glitch is that the music plays twice at the beginning
@CPkmn commented on GitHub (Oct 21, 2013):
I just merged the ffmpeg atrac3+ code to my build since it simplifies a lot with atrac playing and I noticed this happens in Super Danganronpa 2. I came up with a quick fix for it too...should I pull request? (by now I wouldn't be surprised if you figured it out though)
@sum2012 commented on GitHub (Oct 21, 2013):
UCAS40023 : 天地之門 back screen (instead of Shaking video) after merge of maxim-at3plus branch to master
https://github.com/hrydgard/ppsspp/issues/3374
ppsspp-v0.9.1-2057-g0dd0948 log:
https://gist.github.com/sum2012/2bb7cb47544418b2bfcb
debug log:
https://docs.google.com/file/d/0B3OaSdeV0L8kSjd6T1FDaTZpX3c/edit?usp=sharing
Need fill header information by code ?
25:05:194 read thread I[ME]: HW\MediaEngine.cpp:101 FF: Header missing
@Ritori commented on GitHub (Oct 21, 2013):
@sum2012 game work fine now but it only me or This game audio loop every I used load state*
can you confirm it? btw I used v0.9.1-2056-ga39974b
Edit:- I mean SD Gundam G Genaration OverWorld :)
@sum2012 commented on GitHub (Oct 21, 2013):
@Ritori
Yes I confirm the sound probrem .Test on ppsspp-v0.9.1-2057-g0dd0948 windows 32 bit
@CPkmn commented on GitHub (Oct 21, 2013):
@daniel229 / @sum2012 can you try Digimon Adventure with my patch? This fixed voices in Super Danganronpa 2, which had similar error messages to the one in your log and used low level atrac3+ mono for its voices (same as Digimon Adventure, from looking at your log).
https://github.com/CPkmn/ppsspp/compare/patch-5
@sum2012 commented on GitHub (Oct 21, 2013):
@CPkmn Yes .Digimon Adventure 's voice fixed.
Go ahead pull request
@hrydgard commented on GitHub (Oct 21, 2013):
We should probably only check and set that when "low level" sceatrac3 is used, otherwise we should trust the autodetect ... i think.
@CPkmn commented on GitHub (Oct 21, 2013):
@hrydgard it shouldn't matter since when atrac->atracChannels is 1, we always want mono (look at _AtracSetData and sceAtracLowLevelInitDecoder)...
@vnctdj commented on GitHub (Oct 22, 2013):
What about this red line in the log ?
I was trying Monster Hunter Freedom Unite (ULES01213) with ppsspp-v0.9.1-2078-g325f76a-windows-amd64 and this happened when my character appeared in the village.
I don't know if my post is useful... If not, forget it :)
@daniel229 commented on GitHub (Oct 22, 2013):
@CPkmn
Yes,fixed Digimon Adventure and Super Robot Taisen OE,
seventh dragon 2020-2 still missing.
@hrydgard commented on GitHub (Oct 22, 2013):
@vnctdj that's useful :) I wanted to know if it happened, and it appears it does, so I'll have to implement support for that. Shouldn't be too hard.
@CPkmn , I'll just do your mono fix as I don't see a pull request..
@vnctdj commented on GitHub (Oct 22, 2013):
@hrydgard : Thanks ! :)
@papel commented on GitHub (Oct 23, 2013):
I do not know if this code makes sense, but it fixes both Trails in the Sky and Ys Seven.
The uncommon test if-blocks were always false. I do not have many games to test.
github.com/papel/ppsspp@1ed58984a1@hrydgard commented on GitHub (Oct 23, 2013):
@papel, doesn't look unreasonable, but it's really hard to say what's right. @unknownbrackets went by test results, don't know if your code would pass it, but neither would the old one ...
Send a pull request, we can discuss it more there. Also those warn logs need to have the name of the function in the log message, otherwise it's hard to tell where it's from just looking at the log.
@daniel229 commented on GitHub (Oct 24, 2013):
Zero no Kiseki and Ao no Kiseki missing BGM after battle.
SOUND_THREAD E[ME]: HLE\sceAtrac.cpp:604 avcodec_decode_audio4: Error decoding audio -1094995529
@majesticR3 commented on GitHub (Oct 24, 2013):
Issue with GTA Liberty City Stories:
http://www.youtube.com/watch?v=sleBLKZRAsg
Look at FPS slowdowns when I switch radio stations
@papel commented on GitHub (Oct 25, 2013):
It would be good to have results of tests in real PSP with another at3+ file.
https://github.com/hrydgard/pspautotests/tree/master/tests/audio/atrac
I tried it with another at3+ file and the result was very different on emulator.
@unknownbrackets commented on GitHub (Oct 25, 2013):
Actually, I had tested against ~12 different atrac3 and atrac3+ files (mostly +), pulled from different games.
The test is completely repeatable, though. If you have a PSP you can run it yourself. We can't include commercial games' files in pspautotests, so I didn't want to include their results (which wouldn't be repeatable) either.
The results do differ, mainly based on the length of the file. More tests are needed for sure. Of other atrac funcs too.
-[Unknown]
@unknownbrackets commented on GitHub (Oct 26, 2013):
When I call sceAtracGetStreamDataInfo(), I get a value 140 bytes after the buffer start (PPSSPP returns the buffer start.) This is exactly how much bigger the "writable" value is than it should be.
Unfortunately, it's not 140 in every file, it varies. It might be the header size? It doesn't change after a decode (although the other value it returns does.)
Another interesting thing. Every time I decode a frame, that other value from sceAtracGetStreamDataInfo() increases (writableBytes.) The increase depends on the file (I suppose ATRAC3 vs. ATRAC3+), but it's exactly half the value that sceAtracGetBufferInfoForResetting() returns for "minWritableBytes". In other words, I guess "minWritableBytes" is generally 2 frames?
I'm sure there are more cases, though.
-[Unknown]
@RebelliousX commented on GitHub (Oct 27, 2013):
Tenchu Shadow Assassin freezes PPSSPP (not crashing, but nothing happens) when hit new game :(
I think it is related to AT3Plus. The game used to work good except minor graphical issues.
@unknownbrackets commented on GitHub (Oct 28, 2013):
So, one complication seems to be that we buffer in a different way than the PSP does. Unless we change that, matching the PSP's return values might be difficult at best.
It seems like it uses a ringbuffer, which excludes the header of the file. A game can just load the entire file into this buffer, or even create a buffer large enough but only load it partially at first ("halfway".)
Trails in the Sky, for example, creates a buffer that is not large enough for the entire file, and loads data into it using sceAtracAddStreamData() and sceAtracGetStreamDataInfo().
It seems like it can also use a "second buffer." I'm not clear on this at all, but perhaps it's needed when the primary buffer is not large enough or something.
Currently, our atrac code attempts to "reconstruct the file" into a large buffer that's not exposed to the game. It's possible that resetting the buffer or play position is causing us to write to the wrong places in this large buffer, or something.
-[Unknown]
@hrydgard commented on GitHub (Oct 28, 2013):
Yeah, we're really pretty far off from how it's supposed to be working. There's nothing stopping us from doing exactly the same thing as the real hardware and ring buffering the same way, but we might have to do the demuxing ourselves like we do in video playback (that is, not use libavformat to split into frames, but splitting ourselves and sending frames directly into libavcodec like we do in video playback).
It's just about figuring out all the little details of the buffering which seems to be quite complicated, yeah. Offsets, pointers, wrapping, and yes, that mysterious second buffer.
@daniel229 commented on GitHub (Nov 3, 2013):
Earth Defense Forces 2 missing BGM in the logo screen.tesed on ppsspp-v0.9.5-42-g5490ed4.
Game Main I[ME]: HW\MediaEngine.cpp:102 FF: Frame data doesn't match channel configuration!
Game Main E[ME]: HW\SimpleAT3Dec.cpp:128 Error decoding Atrac3+ frame
Game Main E[ME]: HW\MediaEngine.cpp:637 AT3 decode failed during video playback
@papel commented on GitHub (Nov 3, 2013):
Since the number of bug reports is growing, I think it would be good to use labels more often.
These are some issues related to Atrac3+ that could have the label "Atrac3+".
2554
3650
4423
3122
4250
2391
Do you have a list of games that call
sceAtracGetBufferInfoForResetting? The report server may have a list.Currently I know:
Ys Seven
Trails in the Sky First Chapter and Second Chapter (Maybe Third Chapter)
GTA: Chinatown Wars
GTA: Vice City
Hexyz Force (for sound effects)
GTA: Liberty City (Maybe?)
Class of Heroes (Maybe?)
@unknownbrackets commented on GitHub (Nov 3, 2013):
To be sure, there are games which do not call sceAtracGetBufferInfoForResetting which are having problems with ATRAC3+. Many of my games use that function actually, and many of them play fine. Here's an incomplete list of other games that use that function (not that I think they are broken):
Tales of Phantasia X / Narikiri Dungeon
Tales of Rebirth
LittleBigPlanet
Fat Princess
Persona 1
Persona 2
Zettai Hero Project
Disgaea 1
Disgaea 2
Valkyria Chronicles 2
Senjou no Valkyria 3
Patapon 3
ClaDun
ClaDun X2
Canabalt
Phantom Brave
Monster Hunter 3rd (demo)
Toukinden (demo)
Prinny 1
Prinny 2
Metal Slug XX
Hexyz Force does not use sceAtracGetBufferInfoForResetting. It is not capable of calling that function since it is not imported.
-[Unknown]
@papel commented on GitHub (Nov 3, 2013):
It is a large list. :(
I see sceAtracGetBufferInfoForResetting in the log of Hexyz Force. This is the log in the title screen (v0.9.5-44 win32).
@unknownbrackets commented on GitHub (Nov 3, 2013):
Hmm, something's wrong with my index file, then. Probably even more of my games call that function in that case.
-[Unknown]
@unknownbrackets commented on GitHub (Nov 3, 2013):
Naruto Ultimate Ninja Hero 3 (demo, Japanese) - which does not even seem to use sceAtrac at all? - is crashing now while decoding atrac3+ from an mpeg.
I assumed it was a bad pointer, so I tried giving it a very huge regular new'd array, which it also crashed on. So I suppose a bug in the atrac3+ decoder somewhere.... or just in our handling/demuxing.
It seems to be at least somewhat into the video, possibly at/near the end.
The nearest ffmpeg func to the crash is ff_atrac3p_generate_tones.
-[Unknown]
@vit9696 commented on GitHub (Dec 13, 2013):
Hope, it won't do harm to inform that currently Maxim's decoder has an issue when reading at3 files with ATRAC3+ encoded stream. I contacted him and a temporary fix (hack), according to his words, is to change the line 373 in libavcodec/atrac3plusdec.c from
return avpkt->size;
to
return avctx->block_align;
Perhaps, it will resolve some issues here. But, like I said, this is, perhaps, a temporary solution and once Maxim releases his decoder it may change.
@internetakias commented on GitHub (Dec 13, 2013):
Does the error code in this log mean anything? It's from when Puyo Pop Fever's opening music stops playing abruptly.
@unknownbrackets commented on GitHub (Jan 22, 2014):
ppsspp-ffmpeg has had its source updated, but only the Mac builds so far. Should we update Windows and Android too, etc?
Naruto Ultimate Ninja Heroes 3 (demo, Japanese) is still crashing in the same place (a few seconds into the intro video.) The demo is available here:
http://www.pspdemocenter.com/page.php?id=3194
-[Unknown]
@hrydgard commented on GitHub (Jan 22, 2014):
Yeah, I've just been too lazy to set up a working Windows build setup for FFMPEG again after I reinstalled my computer. I'll do it soon.
@daniel229 commented on GitHub (Jan 26, 2014):
Naruto Ultimate Ninja Accel 3 crashing problem has been fixed.
Ao no kiseki & Zero no kiseki BGM cut off problem has been fixed.
Last ranker BGM still cut off in coliseum battle.
@hrydgard commented on GitHub (Jan 26, 2014):
Yes, I finally got around to updating ffmpeg properly.
It might be a good idea to close this thread and open specific issues for any remaining atrac problems as this thread has just grown too large and complex, what do you think?
@solarmystic commented on GitHub (Jan 26, 2014):
@hrydgard
Agreed. This thread has served its purpose.
Closed.