Resistance Retribution / Houfuku noToki crash in-game

[sum2012]

info log https://gist.github.com/sum2012/d3803a954fc24270d340

debug log https://drive.google.com/file/d/0B3OaSdeV0L8kVmszTHo4OWU5S3c/edit?usp=sharing

iso index https://gist.github.com/sum2012/adf600b324d3b58ebd96

first notice 37:01:136 ModuleLoader E[MODULE]: HLE\sceKernelModule.cpp:1612 UNIMPL sceKernelStopUnloadSelfModuleWithStatus(00000001, 00000000, 00000000, 00000000, 00000000): game has likely crashed

second notice E[IO]: HLE\sceIo.cpp:1634 ERROR_ERRNO_FILE_NOT_FOUND=sceIoOpenAsync(disc0:/PSP_GAME/USRDIR/GAMEDATA/COMMON/FONTS.HGZ, 00000001, 00000000) - file not found

The file exist in the iso index but end with .HOG \PSP_GAME\USRDIR\GAMEDATA\COMMON\FONTS.HOG

[sum2012]

JPCSP no probrem log: https://gist.github.com/sum2012/71ba4779bb48935d5985

[unknownbrackets]

It seems to still open the HOG files if the HGZ files aren't there, so that may be fine.

What does the code look like at 08807520? I assume this still happens. A JpcspTrace of sceKernelDeleteCallback might help:

sceKernelDeleteCallback 0xEDBA5844 1 x

Just to see if it does call it.

-[Unknown]

[sum2012]

First Thanks you tell me test in Sunday

I give you newest log first: (v0.9.6-406-g62eb701 ,(now crash due to on fast memory?)) info: https://gist.github.com/sum2012/9eb8c23fcbcd51a01128 debug log: https://drive.google.com/file/d/0B3OaSdeV0L8kSElXSW1YcEVtVjg/edit?usp=sharing

[sum2012]

JPCSP trace log: https://gist.github.com/sum2012/4e91cbd3bd05db670252

nothing special 08:24:26 ModuleLoader - sceKernelDeleteCallback 0x3E1433B = 0x0

What does the code look like at 08807520? Hope I do right.

[unknownbrackets]

I wonder if handling sceKernelStopUnloadSelfModuleWithStatus would actually fix this game...

-[Unknown]

[sum2012]

Add sceKernelCreateCallback in JPCSP trace log https://gist.github.com/sum2012/087dd982ba945751386b

PPSSPP debug log: 287=sceKernelCreateCallback(name=MEDIAPOWER, entry=08bf46c0, callbackArg=00000000) But in JPCSP trace log cannot find 08bf46c0

[sum2012]

change config to sceKernelCreateCallback 0xE81CAF8F 3 sxx Hmm more easy to compare https://gist.github.com/sum2012/d588c00039e5e69062b7

09:16:04 user_main - sceKernelCreateCallback 0x08DCF370('MEDIAPOWER'), 0x8BF46C0, 0x0 = 0x3E0C035 seem no problem in sceKernelCreateCallback

[unknownbrackets]

Is this improved at all by #5092?

-[Unknown]

[sum2012]

v0.9.6-420-g108d592 crash after in-game. I need go out now. Debug Log will proudce later https://gist.github.com/sum2012/e09eec24dcfa6e423319

[sum2012]

@unknownbrackets I forget to ask a Important question. Do make log need turn off fast memory ? edit: turn off fast memory info log (a bit better) https://gist.github.com/sum2012/39b5aa5bb5d1e8905cd0

[unknownbrackets]

Generally yes, I usually use fast memory off. It helps to know the address that failed, sometimes, and especially what thread it was on.

Hmm. I guess it could be io related...

-[Unknown]

[sum2012]

v0.9.6-429 debug log fast memory off,stop on first error https://drive.google.com/file/d/0B3OaSdeV0L8kRXlHaU5uNnpNTWM/edit?usp=sharing

[sum2012]

Maybe need fix invalid address before more test

[sum2012]

v0.9.8-500-gb0e97a0 Now debug log do not show OSK(freeze) spam of D[UTIL]: HLE\sceUtility.cpp:376 00000001=sceUtilityOskGetStatus() (Could not get time to draw a OSK ?) debug log: https://drive.google.com/file/d/0B3OaSdeV0L8kakI4ZmZWb2tETTA/edit?usp=sharing

Fast test osk good v0.9.6-813-g9798af3 debug log:https://drive.google.com/file/d/0B3OaSdeV0L8kSzNQbnYwNk9kZjQ/edit?usp=sharing

bad v0.9.6-837-g19a3ede

Compare https://github.com/hrydgard/ppsspp/compare/v0.9.6-813-g9798af3...v0.9.6-837-g19a3ede

[sum2012]

Hmm, just a sceUtility debug log also proudce the problem

[sum2012]

git bisect result This commit https://github.com/unknownbrackets/ppsspp/commit/1f5639a489dfe68129cf2ae2d2833cd03acd6c81

[sum2012]

int PSPOskDialog::Init(u32 oskPtr) {

ifdef DEBUG

 OSK_INIT_DELAY_US = 10000;

else

OSK_INIT_DELAY_US = 300000;

endif

@unknownbrackets @hrydgard @thedax How to get a log setting from sceUtility programmably?

[thedax]

@sum2012: You mean how can you tell what the log level for a certain module is set to while PPSSPP is running?

Probably something like this:

int logLevel = LogManager::GetInstance()->GetLogLevel(SCEUTILITY);

if (logLevel == LDEBUG) {
 // Do stuff if it's debug level
} else if (logLevel == LERROR) {
// Do something else if it's an error
}

Or use a switch statement:

int logLevel = LogManager::GetInstance()->GetLogLevel(SCEUTILITY);

switch (logLevel) {
 case LDEBUG:
      // Do stuff
      break;
 case LERROR:
      // Do other stuff
      break;
 default:
      // Do some other behaviour when it's not any of the above
      break;
}

[unknownbrackets]

Emulation behavior should definitely NOT change based on log level. Talk about a debugging nightmare.

-[Unknown]

[thedax]

Oh, I agree completely. I was just answering his basic question on how to obtain the logging level at runtime.

[daniel229]

The DEMO also gets a lot Invalid address. https://drive.google.com/file/d/0BzGZGDfFE68zSGNWcHVKX1pLWW8/edit?usp=sharing

[daniel229]

For the demo,after you killed the 3 robots in the first room and entry to another room,there is cutscene,if you don't skip,at the end it reportsa lot of invalid address,then after you killed one of the two big robots,it reports a lot of E[G3D]: GLES\SoftwareTransform.cpp:332 Normal projection mapping without normal? ,sometime it crashes.

[unknownbrackets]

Interesting. The invalid addresses happen even with jit off, so it doesn't look like a jit bug.

-[Unknown]

[sum2012]

Flush icache only related to jit ?

[unknownbrackets]

Correct... well, I guess there are also the replacement opcodes. Did it used to work without jit on?

-[Unknown]

[sum2012]

Doesn't

[daniel229]

Fixed one kind of the invalid addresses I mentioned in the https://github.com/hrydgard/ppsspp/issues/3110 ,that cutsene invalid addresses don't hang the emulator sometime.

[daniel229]

Those invalid addresses would make the game missing some graphics

non invalid addresses

invalid addresses

[sum2012]

Update the log to v0.9.9-84-g37b8f62 Info log:https://gist.github.com/sum2012/918e3536fe0c65485786 Debug log: https://drive.google.com/file/d/0B3OaSdeV0L8kTlU1WUtoVWN6NTg/edit?usp=sharing I find sceUsb staff

[daniel229]

Don't Register sceAtrac3plus() can avoid the invalid addresses and these graphical issue.https://github.com/hrydgard/ppsspp/issues/5631

[sum2012]

Good finding

[sum2012]

I test from start. Still freeze

[daniel229]

Above errors happen randomly.I don't get those errors this time.

[daniel229]

change this if (Memory::IsValidAddress(numSamplesAddr))Memory::Write_U32(numSamples, numSamplesAddr);

to if (Memory::IsValidAddress(numSamplesAddr))Memory::Write_U32(numSamples/2, numSamplesAddr); it would not get invalid addresses and these graphical issue.

[unknownbrackets]

It could be that this file decodes less samples each frame, whether due to mono or who knows what. But also, the first frame of audio actually is supposed to be a lower number of samples. Could just be that.

-[Unknown]

[daniel229]

It works with this.with atrac->currentSample <= 2048 still got error.

        if (Memory::IsValidAddress(numSamplesAddr))
            (atrac->atracChannels == 1 && atrac->currentSample <= 4096) ? Memory::Write_U32(numSamples / 2, numSamplesAddr) : Memory::Write_U32(numSamples, numSamplesAddr);

[daniel229]

It works with this

if (Memory::IsValidAddress(numSamplesAddr))
            (atrac->codecType == PSP_MODE_AT_3_PLUS && atrac->atracChannels == 1 && atrac->currentSample == 2048) ? Memory::Write_U32(numSamples / 32 + 1, numSamplesAddr) : Memory::Write_U32(numSamples, numSamplesAddr);

[unknownbrackets]

Interesting. Well, at least that supports the theory that it could have to do with the first time. Thanks.

Guess we'll have to test a bunch of permutations, long/short, at3/at3+, stereo/mono, full buffer/sliding buffer...

-[Unknown]

[daniel229]

Demo crashing after the big robots explode in the second room,same as the retail version. savestate for the demo https://drive.google.com/file/d/0BzGZGDfFE68zaUJjYkZNSTljVXM/edit?usp=sharing

[unknownbrackets]

199e6bcd3bd57e173d9704c4016a6759203e3b7b avoids that crash, but it's still doing something wrong. Also, Google doesn't seem to want me to download that file.

-[Unknown]

[daniel229]

Still crash with that change,I can't download from google either.Try this savestate https://onedrive.live.com/redir?resid=46301D80532F136C%21107

[unknownbrackets]

Hmm, that's jumping to 0 but I'm not sure from where. It's not in jit or vertexjit...

Do the graphical glitches still go away entirely without sceAtrac? I guess it's probably still bugs in that.

-[Unknown]

[daniel229]

without Register_sceAtrac3plus(); still get graphical glitches and crashing.

[ppsspp-gamer]

Tested resistance retribution (UCUS-98668) with PPSSPP build v0.9.9.1-537-gabb294f x64, OpenGL backend, after merge pull request #6935 from unknownbrackets/atrac-minor, this game can almost finish first mission without any graphical glich, but still ran into infinite loop after open the door as picture below. Real time save state is attached just before ppsspp hang, download save state from here: http://www.mediafire.com/download/d844cd9i1er4d7e/UCUS98668_1.00_0.zip.

[daniel229]

@ppsspp-gamer that area did not hang for me with JPN version. Same area crash point as the demo in retail version ,if it get lucky,it did not crash,it would get this error

snd_stream_d E[ME]: HLE\sceAtrac.cpp:1638 sceAtracSetAA3DataAndGetID(09007140, 51200, 14870528, 09fcc9b8): ATRAC UNKNOWN FORMAT

[sum2012]

@ppsspp-gamer save (USA version) media log: https://gist.github.com/sum2012/f39217a5c38b713fc71b

debug log: https://drive.google.com/file/d/0B3OaSdeV0L8kbi1aQzY5QUNwaFU/edit?usp=sharing

[ppsspp-gamer]

I believe the resistance retribution crash (UCUS-98668) is related to atrac functions. If I disable audio just before opening the door, then enable audio after that, I can finish the first mission without problems. Actually I can continue several missions without crashes.

Another real time save state before crash (this crash can be avoid by disabling audio). http://www.mediafire.com/download/yy2ldh49rajwm51/UCUS98668_1.00_0-02.zip

[unknownbrackets]

The emulator's option to disable audio only disables output; everything else related to audio (including all the atrac3+ decoding and etc.) is still done. That is, the option doesn't affect game behavior at all; it only affects whether we send the resulting data to DirectSound or ignore it during the last step in the process.

If turning off audio really avoids the crash, it may be related to the amount of audio data being generated or something (like an overflow in our DirectSound management), rather than decoding specifically. But that seems unlikely based on the other crash.

-[Unknown]

[sum2012]

@ppsspp-gamer I still get invalid address when I disable audio in option

[ppsspp-gamer]

@unknownbrackets but if audio is disable, then the StreamCallback in soundThread() function won't be called, why should the atrac3+ decding still be done?

@sum2012, then I guess I am only lucky, I play almost half of this game by using this trick.