Code Lyoko audio softlock (now: bad playback)

[Xibancry]

Game or games this happens in

ULES-01095 - Code Lyoko: Quest for Infinity

What area of the game / PPSSPP

Many areas affected. Only when some voice audio tracks play. The text box showing dialogue appears but it fails to play the voice audio preventing it from closing and effectively softlocking the game everytime it happens (this is apparently consistent with some dialogues). Only solution around it is to reset the game.

Adding a savefile previous to softlock. Softlock occurs on the third dialogue just after defeating the first enemy in the level.

ULES01095DATA00.zip

What should happen

The game should play the dialogue normally without preventing the dialogue box from closing and not softlocking the game.

Logs

ppsspplog.txt

Platform

Windows

Mobile device model or graphics card (GPU)

RTX 2080 Super

PPSSPP version affected

v1.18.1 - 0f50225

Last working version

v1.16.6

Graphics backend (3D API)

Direct3D 11

Checklist

• Test in the latest git build in case it's already fixed.
• Search for other reports of the same issue.
• Try resetting settings or older versions and include if the issue is related.
• Try without any cheats and without loading any save states.
• Include logs or screenshots of issue.

[sum2012]

@Xibancry
"Sector 5" or "Deseat sector" lock up ?
edit: should be "Sector 5"

[Xibancry]

Yes, it's on sector 5 for the savefile I sent. It's not the only place it happens

[sum2012]

Media log easier to check
https://gist.githubusercontent.com/sum2012/de3af54e161e8eacb48086b34315397f/raw/17207eac434bdb4fee0ab0cfb2b688f2543a19fb/gistfile1.txt

18:26:659 ATRAC3 Load D[ME]: HLE\AtracCtx.cpp:235 ATRAC analyzed: AT3Plus channels: 2 filesize: 70164 bitrate: 94 kbps jointStereo: 0
18:26:659 ATRAC3 Load D[ME]: HLE\AtracCtx.cpp:236 dataoff: 164 firstSampleOffset: 4011 endSample: 251620
18:26:659 ATRAC3 Load D[ME]: HLE\AtracCtx.cpp:237 loopStartSample: 4379 loopEndSample: 255999
18:26:659 ATRAC3 Load I[ME]: HLE\AtracCtx.cpp:699 0=sceAtracSetDataAndGetID(08a5c340, 00000800): atrac3+ stereo audio
18:26:659 ATRAC3 Load D[ME]: HLE\sceAtrac.cpp:702 00000000=sceAtracSetLoopNum(0, -1)
18:26:659 ATRAC3 Load D[ME]: HLE\sceAtrac.cpp:463 00000000=sceAtracGetRemainFrame(0, 08a5c1dc[00000002])
18:26:677 ATRAC3 play D[ME]: HLE\sceAtrac.cpp:281 00000000=sceAtracDecodeData(0, 08a7c380, 09fb7aa8[000006e5], 09fb7aac[00000000], 09fb7abc[1])
18:26:693 ATRAC3 play D[ME]: HLE\sceAtrac.cpp:532 00000000=sceAtracGetStreamDataInfo(0, 09fb7ab8[08a5c40c], 09fb7ab4[00000438], 09fb7ab0[00000800])
18:26:693 ATRAC3 play E[ME]: at3_standalone\compat.cpp:19 Atrac3/3+: Frame data doesn't match channel configuration!
18:26:693 ATRAC3 play D[ME]: HLE\sceAtrac.cpp:281 80630024=sceAtracDecodeData(0, 08a7df14, 09fb7aa8[00000000], 09fb7aac[00000001], 09fb7abc[0])
18:26:693 ATRAC3 play E[ME]: HLE\sceAtrac.cpp:382 UNIMPL sceAtracGetInternalErrorInfo(0, 09fb7aa4)
18:26:694 ATRAC3 play D[ME]: HLE\sceAtrac.cpp:253 00000000=sceAtracAddStreamData(0, 00000438)
18:26:694 ATRAC3 play D[ME]: HLE\sceAtrac.cpp:463 00000000=sceAtracGetRemainFrame(0, 09fb7abc[00000003])

ATRAC3 related

[sum2012]

Even v1.16.6 don't softlock, the voice has some problem

[sum2012]

v1.17.1-264 good
v1.17.1-268 bad
It mean #19053

[sum2012]

git bisect result: a1ac79c

[hrydgard]

Thank you for the bisect, not surprising, it's when we switched the decoder to not use ffmpeg.

I think this is another case of not detecting mono properly, so the decoder tries to parse it as stereo and fails. Will need to figure out how it worked before..

[sum2012]

@hrydgard Before it cannot decode mono correctly , but it do not have error .

[sum2012]

@hrydgard I have a idea. For streamed ATRAC3+ , I think that we do not change track_.channels , so it assume stereo.
edit: No, 8:26:659 ATRAC3 Load I[ME]: HLE\AtracCtx.cpp:699 0=sceAtracSetDataAndGetID(08a5c340, 00000800): atrac3+ stereo audio So it detect wrong the atrac as stereo not mono early ?

[sum2012]

Force to change to mono still fail.

		if (!decoder_->Decode(indata, track_.bytesPerFrame, &bytesConsumed, outputChannels_, (int16_t*)outbuf, &outSamples)) {
			track_.channels = 1;
			if (!decoder_->Decode(indata, track_.bytesPerFrame, &bytesConsumed, outputChannels_, (int16_t*)outbuf, &outSamples)) {			
			// Decode failed.
			*SamplesNum = 0;
			*finish = 1;
			return ATRAC_ERROR_ALL_DATA_DECODED;
			}
		}

[sum2012]

Ignore decode error can work around....Before v1.19 release make the hack ?

[hrydgard]

If we can't fix it, yeah.

[hrydgard]

#20005 will probably work as a workaround, but I don't think it'll fix broken audio.. Something is really going wrong in decoding here.

[sum2012]

#20005 work as a workaround
Jpcsp emulator play the voice correctly
PPSSPP never play the voice correctly

[sum2012]

Jpcsp emulator log
https://gist.githubusercontent.com/sum2012/77328a1337e7dd96b2252bba4e7c4dd2/raw/52b997c6216503e05bbb0dbb31bb5d469a90d80d/gistfile1.txt

edit:
#20005 Warn log this two times , then it can play the bad voice
53:30:276 ATRAC3 play W[ME]: at3_standalone\compat.cpp:23 Atrac3/3+: Frame data doesn't match channel configuration! ch_block 1 >= num_channel_blocks 1,

From JPCSP emulator log, channels=2 all the times

[hrydgard]

Hm. Might have to dig into how JPCSP handles this then... Thanks for letting me know.

[sum2012]

Can you code (logging) the data of the "data chunk" in PPSPPP so that I can compare it in JPCSP emulator ?

5:05:40 DEBUG hle.sceAtrac3plus - ATRAC3 Load thread - sceAtracSetDataAndGetID buffer=0x08A5C340, bufferSize=0x800
05:05:40 DEBUG hle.sceAtrac3plus - ATRAC3 Load thread - FileSize 0xA5BE4
05:05:40 DEBUG hle.sceAtrac3plus - ATRAC3 Load thread - WAVE format: magic=0x20746D66('fmt '), chunkSize=52, compressionCode=0xFFFE, channels=2, sampleRate=44100, bitrate=12058, bytesPerFrame=0x230, hiBytesPerSample=0, codingMode=0
05:05:40 DEBUG hle.sceAtrac3plus - ATRAC3 Load thread - Additional chunk data: 22 00 00 08 03 00 00 00 BF AA 23 E9 58 CB 71 44 A1 19 FF FA 01 E4 CE 62 01 00 28 45 00 00 00 00 00 00 00 00
05:05:40 DEBUG hle.sceAtrac3plus - ATRAC3 Load thread - FACT Chunk: chunkSize=8, endSample=0x25D391, sampleOffset=0xAFE
05:05:40 DEBUG hle.sceAtrac3plus - ATRAC3 Load thread - Loop #0: LoopInfo[cuePointID 0, type 0, startSample 0x0, endSample 0x25D391, fraction 0, playCount 0]
05:05:40 DEBUG hle.sceAtrac3plus - ATRAC3 Load thread - DATA Chunk: data offset=0xA4, data size=0xA5B40
05:05:40 DEBUG hle.sceAtrac3plus - ATRAC3 Load thread - hleSetHalfwayBufferAndGetID returning atID=0x0

[sum2012]

I don't know this code
https://github.com/hrydgard/ppsspp/blob/master/Core/HLE/AtracCtx.cpp#L361
Why break after "}" ?
Same with https://github.com/hrydgard/ppsspp/blob/master/Core/HLE/AtracCtx.cpp#L373
https://github.com/hrydgard/ppsspp/blob/master/Core/HLE/AtracCtx.cpp#L404

[anr2me]

I don't know this code https://github.com/hrydgard/ppsspp/blob/master/Core/HLE/AtracCtx.cpp#L361 Why break after "}" ? Same with https://github.com/hrydgard/ppsspp/blob/master/Core/HLE/AtracCtx.cpp#L373 https://github.com/hrydgard/ppsspp/blob/master/Core/HLE/AtracCtx.cpp#L404

Probably because you can get compilation error when declaring and initializing a variable inside case without {}

[hrydgard]

It's nicer to have the break inside the {} of the case, but it doesn't really matter.

I'll look into this more in a little while.

[hrydgard]

Hm, jpcsp seems to handle the channel blocks exactly the same:

https://github.com/jpcsp/jpcsp/blob/cd20cf312b358b4260f26f6754f9c62926c70ba6/src/jpcsp/media/codec/atrac3plus/Atrac3plusDecoder.java#L96

Wonder what is different..

[sum2012]

edit last: I know what is the problem

[hrydgard]

So what is it? :)

I merged your logging change.

[sum2012]

Updated PPSSPP log:
https://gist.github.com/sum2012/355e85c0347ccdacadf8a249018796e0
Our code of "Additional chunk data" is match JPCSP emulator.

[sum2012]

After update source, debug build (VS2022):
34:41:289 ATRAC3 play E[HLE]: HLE\HLE.cpp:992 HLE function stack mismatch (D:\project\9\ppsspp\Core\HLE\sceAtrac.cpp:253)! stackSize = 0