Last Ranker crashes when battle in the Coliseum

[daniel229]

disable atreac3+ plugin can "fix" this issue.

savedata
https://docs.google.com/file/d/0BzGZGDfFE68zNUZvdDIxMi1VbEU/edit?usp=sharing

[daniel229]

It seems like one commit that between 0.76-1140-gbe483a1 and 0.76-1157-g7d566b9 causes the issue.Before 1141,it random crashed.after 1156,It definitely crashes.Disable atreac3+ plugin,no crash in any builds.

[daniel229]

in __sceAudio.cpp
revert
if (chan.sampleQueue.size() > 0 || chan.sampleAddress == 0)
to
if (chan.sampleQueue.size() > chan.sampleCount * 2 * chanQueueMaxSizeFactor || chan.sampleAddress == 0) {

then the definitely crash return to random crash.

[sum2012]

VS2010 crash screen

This time I do not know correctly fix

[sum2012]

@ daniel229
Please try work or not
https://docs.google.com/file/d/0B3OaSdeV0L8kS1p1cXJaVmcwZTA/edit?usp=sharing

@unknownbrackets
I hope this is not hack...

u32 _AtracDecodeData(int atracID, u8* outbuf, u32 *SamplesNum, u32* finish, int *remains)
{
    Atrac *atrac = getAtrac(atracID);

    u32 ret = 0;
    if(SamplesNum > finish) //added
        return 0x80630024;//ERROR_ATRAC_ALL_DATA_DECODED added

[unknownbrackets]

@sum2012 hmm, that fixes it really? I don't understand.

In sceAtracDecodeData(), both values are initialized to 0. That check should never have effect there, because 0 > 0 is always false.

In SasAtrac3::getNextSamples(), it's used in a loop, so it seems more likely. finish is either 0 or 1, but SamplesNum is again initialized to 0. 0 > 1 and 0 > 0 are both always false. Either way, it can't be greater than finish, can it?

@daniel229 and @sum2012, it seems like this is a crash inside the atrac decoder. Either bad params are being passed to it (e.g. a bad pointer, or a size that goes beyond the pointer), or the decoder is simply buggy. Or both.

-[Unknown]

[sum2012]

@unknownbrackets so that I need @daniel229 confirm work
Add 1 more picture

[unknownbrackets]

As you can see from the stack, it's actually crashing in that other dll. I can't really help you with that dll, but I will say from a general standpoint, if you want to debug a crash inside a dll, you probably need to compile that dll in debug mode if possible (and most importantly have its pdb.)

-[Unknown]

[daniel229]

@sum2012 no crash,but missing BGM.

[daniel229]

run with a debug dll.

[sum2012]

My error screen is differult to you

Again no idea to fix

[daniel229]

Test with the maxim-at3plus,It stops BGM instead of crash at that crash point.

[sum2012]

stops BGM log:
47:02:476 ATRAC3 play I[ME]: HW\MediaEngine.cpp:101 FF: Frame data doesn't match channel configuration!
47:02:476 ATRAC3 play E[ME]: HLE\sceAtrac.cpp:599 avcodec_decode_audio4: Error decoding audio -1094995529
47:02:476 ATRAC3 play E[ME]: HLE\sceAtrac.cpp:787 UNIMPL sceAtracGetInternalErrorInfo(0, 09fdf66c)
47:02:504 ATRAC3 play W[KERNEL]: C:\BuildAgent\work\acf56f986e98e7c9\Core/HLE/sceKernel.h:454 Kernel: Bad object handle 0 (00000000)
47:02:504 ATRAC3 play E[IO]: HLE\sceIo.cpp:1033 sceIoLseek(4, 10d308, 0) - ERROR: invalid file

ppsspp-v0.9.1-2038-g90eed17 full log
https://gist.github.com/sum2012/a478046e21ed545d3b2b

[daniel229]

Still cut off with FFMPEG updated
40:41:618 ATRAC3 play I[ME]: HW\MediaEngine.cpp:84 FF: Frame data doesn't match channel configuration!
40:41:618 ATRAC3 play E[ME]: HLE\sceAtrac.cpp:608 avcodec_decode_audio4: Error decoding audio -1094995529
40:41:618 ATRAC3 play E[ME]: HLE\sceAtrac.cpp:818 UNIMPL sceAtracGetInternalErrorInfo(0, 09fdf66c)
40:41:651 ATRAC3 play E[IO]: HLE\sceIo.cpp:1055 sceIoLseek(4, 10d308, 0) - ERROR: invalid file
40:41:651 ATRAC3 play W[IO]: HLE\sceIo.cpp:796 sceIoRead(4, 099b7c80, 0): error 80020323

[hrydgard]

We are obviously decoding bad data for some reason, but at least we aren't crashing now...

[daniel229]

That BGM is playback well on ffmpeg player until the end,It does not stop,and looping the last frame.The other BGM files that from the iso have the same issue on ffmpeg player.

[unknownbrackets]

Does this still show the same "FF: Frame data doesn't match channel configuration!" error?

-[Unknown]

[daniel229]

Still,but I get it works one time.
16:29:064 ATRAC3 play E[ME]: HLE\sceAtrac.cpp:1445 sceAtracSetLoopNum(1, 0):no loop information
16:29:127 ATRAC3 play I[ME]: HLE\sceAtrac.cpp:1084 sceAtracReleaseAtracID(1)
16:29:205 ATRAC3 play I[ME]: HW\MediaEngine.cpp:85 FF: Frame data doesn't match channel configuration!
16:29:205 ATRAC3 play E[ME]: HLE\sceAtrac.cpp:655 avcodec_decode_audio4: Error decoding audio -1094995529
16:29:205 ATRAC3 play E[ME]: HLE\sceAtrac.cpp:911 UNIMPL sceAtracGetInternalErrorInfo(0, 09fdf66c)
16:29:208 ATRAC3 play E[IO]: HLE\sceIo.cpp:1084 sceIoLseek(4, 10d308, 0) - ERROR: invalid file
16:29:208 ATRAC3 play W[IO]: HLE\sceIo.cpp:804 sceIoRead(4, 099b7c80, 0): error 80020323
16:29:208 ATRAC3 play I[ME]: HLE\sceAtrac.cpp:1084 sceAtracReleaseAtracID(0)
16:29:629 ATRAC3 play I[ME]: HLE\sceAtrac.cpp:1628 0=sceAtracSetMOutHalfwayBufferAndGetID(08ccdb80, 00000c64, 00000c64)
16:29:629 ATRAC3 play W[ME]: HLE\sceAtrac.cpp:1273 This is an atrac3+ mono audio
16:29:629 ATRAC3 play E[ME]: HLE\sceAtrac.cpp:1445 sceAtracSetLoopNum(0, 0):no loop information

[daniel229]

Look like it usually works,and occasionally not work

[hrydgard]

Still happens?

[daniel229]

Yes,it still happens

[unknownbrackets]

So, just to be clear, does it still crash, or does it just not loop properly now?

-[Unknown]

[daniel229]

Not crash,just stop suddently in certain point of the bgm,not a loop,and sometime not happen.

[daniel229]

Still got "Error decoding audio" sometime in v1.0.1-224-g0978500 with default setting,but it won't happen when CPU clock change to 1000.

[unknownbrackets]

Is it affected by IO timing method?

-[Unknown]

[daniel229]

Yes,it is affected IO timing method.I usually try cpu clock 0 or 1000 with IO timing method fast.1000 always work.Just combined other IO timing method and got this result.

cpu clock 0

Fast not work
Host not work
Simulate UMD delays not work

cpu clock 1000

Fast work
Host work
Simulate UMD delays not work