Afterburner Black Falcon - Freezes after Dev Logo/Copyright Screen

[4XPG]

I recall this game used to be working back on v0.9.1 (according to the forum entry, it doesn't work since 0.9.5-0.9.6 and still never been fixed since then), but it freezes after the dev logo FMV/copyright screen on the later version (I'm using v0.9.9-80 in this case).

The sound thread appears to be stuck at endless loop, here's the log:
https://gist.github.com/4XPG/1a529bc99d1959f6c4fc

[unknownbrackets]

Does it actually work in v0.9.1?
https://github.com/hrydgard/ppsspp/releases/tag/v0.9.1

Does it work in JPCSP or something else? Just want to make sure you have a good rip.

-[Unknown]

[daniel229]

The last working build is v0.9.6-453-gda97a67,It hanged up since #5044 ,

[4XPG]

It worked on both old v0.9.1 (bad sounds, but the rest are fine) and JPCSP (100% working).

[sum2012]

This is v0.9.9.1-144-g8edf255 scempeg log:
https://gist.github.com/sum2012/e5fb7cb7dc1e4a0a60fe
It freeze in 47:31:509 decodeThread D[ME]: HLE\sceMpeg.cpp:1607 80618001=sceMpegGetAtracAu(0bfff380, 00000004, 0bfff44c, 0bfc8e40)

[daniel229]

revert unknownbrackets@af411e6 the video won't hang,but frameskipping.

[unknownbrackets]

That implies that the pts is either not getting set correctly or getting stuck. When it's hung, what is m_videopts? And what are m_readSize and m_streamSize at that point?

-[Unknown]

[daniel229]

How to find these value?

[daniel229]

commemt out this in sceMpegGetAtracAu() the video won't hang also,and no frameskipping in intro video.

/*  if (ctx->atracRegistered && ctx->mediaengine->IsNoAudioData() && !ctx->endOfAudioReached) {
        WARN_LOG(ME, "Audio end reach. pts: %i dts: %i", (int)atracAu.pts, (int)ctx->mediaengine->getLastTimeStamp());
        ctx->endOfAudioReached = true;
        result = ERROR_MPEG_NO_DATA;
    }*/

[unknownbrackets]

If you only comment out ctx->endOfAudioReached = true;, it doesn't work, right? What if you only comment out result = ERROR_MPEG_NO_DATA;?

-[Unknown]

[daniel229]

Only works with comment out result = ERROR_MPEG_NO_DATA;

[unknownbrackets]

What's strange is, it seems to be adding extra packets, one would think they might include audio.

Maybe it's something wrong with how it's starting up, like not doing the warm up (which I'm pretty sure mpeg is supposed to, but eventually I'll need to take the plunge on sceMpeg tests, unless someone else wants to write some and help...)

-[Unknown]

[sum2012]

Can we have a work around ?

[sum2012]

Update the log to v1.1.1-43-g6c31612
(I have switched to USA version to debug)
https://gist.github.com/sum2012/d062601b36159dde64d6

[sum2012]

I change the log to
WARN_LOG(ME, "Audio end reach. pts: %i dts: %i m_videopts %i m_audiopts %i", (int)atracAu.pts, (int)ctx->mediaengine->getLastTimeStamp(), ctx->mediaengine->m_videopts, ctx->mediaengine->m_audiopts);

output:
01:53:818 decodeThread W[ME]: HLE\sceMpeg.cpp:1609 Audio end reach. pts: 866654 dts: 780780 m_videopts 774774 m_audiopts 780834

[unknownbrackets]

So then getAudioTimeStamp() is ahead of the video indeed? Hmm. That implies to me that it's decoding audio frames faster than video...

-[Unknown]

[sum2012]

So just return 0 would solve the problem

[LunaMoo]

More like a hack than solution, since 0 just means no error when it clearly got one it was designed to detect. You can do that via CWCheat if you want to, just find the syscall in ppsspp disassembly and force v0 or whatever v0 is passed to after that syscall to 0 - this can be done by setting it's value to 0(li v0,0) or by moving value of zero register(move v0,zero).

BTW is that code related -

ppsspp/Core/HLE/sceMpeg.cpp

Line 1503 in c2f4fad

if (ctx->atracRegistered && (ctx->mediaengine->getVideoTimeStamp() > ctx->mediaengine->getAudioTimeStamp() + getMaxAheadTimestamp(mpegRingbuffer)))

It seems like it was stopping this from happening, disabled because it does't happen on psp?

[sum2012]

In the real PSP,this game of video don't return ERROR_MPEG_NO_DATA
JPCSPTrace log:
https://gist.github.com/sum2012/43a7a73f9a0dac1175a0

[unknownbrackets]

Yes, but maybe something else is causing it to call that too frequently. Does it call it the same number of times?

The real firmware returns no data when there's no audio remaining. It doesn't even mean end; it means there's no audio data in the buffer, afaict. So the issue here can be any of:

1. It's reading audio data out faster than video data, and shouldn't be.
2. It's not filling the buffer as fast as it should be.
3. There's some other bug somewhere (e.g. it's ignoring data that was fed to it.)

I'm pretty sure it should NOT return 0 when there's no audio data, and doing so will DEFINITELY break other games. I'm also pretty sure it should not delay to manually sync audio with video, since that is the game's job, and iirc, doing that again will ALSO break other games.

-[Unknown]

[sum2012]

If I make change,at least the following:
if ((int)ctx->mediaengine->getLastTimeStamp() > ctx->mediaengine->m_audiopts)
result = ERROR_MPEG_NO_DATA;

Don't break Obscure the Aftermath,Pacman land 3 ,it still hit ERROR_MPEG_NO_DATA;

[sum2012]

@unknownbrackets jpcsp fix this game by jpcsp/jpcsp@f4a49b8
//It seems that sceMpegGetAtracAu returns ERROR_MPEG_NO_DATA only when both
// the audio and the video have reached the end of the stream.
// No error is returned when only the audio has reached the end of the stream.

[unknownbrackets]

I'm really not sure that's correct, since that sounds like it contradicts firmware tests I did previously, although it's been a while. I'd like to see actual firmware tests confirming this.

-[Unknown]

[sum2012]

(no wrong)
This game's firmware is 2.82

[slowfellow]

Can someone please adress this issue?, I very much like to play After Burner on the latest version.
Regards

[sum2012]

@slowfellow what device are you using ? Windows 32 bit,Windows 64 bit,Android

[slowfellow]

I am using both Windows32 and Android 5.1.1 (Xperia Z5 compact)

[sum2012]

The is test build
Windows 32 build
https://drive.google.com/file/d/0B3OaSdeV0L8kVTI1MG1GRUlaaWc/view?usp=sharing
Android build
https://drive.google.com/file/d/0B3OaSdeV0L8kaHlCbmRkWWctcW8/view?usp=sharing
If it break other game of video ,I would appreciate that you tell me.Thanks

[sum2012]

oop ,my change break Obscure The Aftermath

[slowfellow]

Thanx, I will try them out, windows version now supports Afterburner. Thanx, will try android later.

[slowfellow]

Why not make it an optional switch if it breaks some games, but supports others?

[sum2012]

Becuase I didn't test when I compiled

[slowfellow]

Well I have tested some 20 games so far everything works, maybe this Obscure, The Aftermath is a combination of things

[unknownbrackets]

If you un-revert e866d83, does this work? Just wondering, since it's related to the jpcsp change.

-[Unknown]

[daniel229]

No,it doesn't.

[daniel229]

Fixed in #8867

[unknownbrackets]

Oh, that makes sense. So it was likely related to how it was reading the header.

Thanks for testing.

-[Unknown]

[sum2012]

Video flicking in beginning.
Old version (eg: v1.1.1-577-gefc1768) no problem.
I post here or open new issue ?

[unknownbrackets]

Hmm, probably a new issue is better since it started at a different time - better to avoid getting confused with the old issue.

-[Unknown]