Kingdom Hearts: BBS, game stays on white screen after completing events.

[Kenny4343]

From the forum member VanitasZD:
"After completing an event (e.g defeat all the Unversed in Olympus Colloseum), it freezes on the white screen. Why does it do this?"

It happened on maleficent's castle for me, and it doesn't really freeze, it keeps running but does nothing, here's a log: http://db.tt/2v94qP9l

[unknownbrackets]

Is it reproducible every time? Did this always happen, or is it new (and if so, since when)?

-[Unknown]

[Kenny4343]

Every single time that I tried (three, gave up after that), it didn't happen on previous events of this kind that I first did, and no one had said anything either, but it certainly started somewhere after video support... It'll be a pain to reproduce it to test on other builds as I'm on the hardest difficulty LOL, but if someone else doesn't, I'll see if I can find it.

[solarmystic]

@Kenny4343

Could be a little bit more specific in your problem description?

Stating the platform you experienced the freeze on would be a nice start. (Windows? Android? IOS? Symbian?)

Settings used in the emulator? (Fast mem etc)

Hardware utilized?

Like @unknownbrackets mentioned, problem repro (the exact steps undertaken to reproduce the problem) would also be great.

[Kenny4343]

I don't know what did Vanitas use, as always from my side it's on android, SGS3, settings are default ones, with and without fast mem it still happens, and as explained, as already said it happens fter in-game events (Not sure if all of them) for me it was in maleficent's castle where you have to beat a great group of unversed, there the screen slowly becomes white as everytime you finish one of these events and stays white.

Edit: Now that I realize it, people were saying how they were stuck after the tutorial, so I'm guessing it must be after completing the event of fighting the light balls which is similar, I'm just guessing though.

[Kenny4343]

Surprisingly, as told by amj530 on the forums:
"This happened 2 times to me, but I waited a few minutes and then continued normally."
It does continue after you wait for a few minutes on the whitescreen, although that's not normal either..

[solarmystic]

@Kenny4343

If waiting makes it continue normally, could activating turbo help to speed up the process?

[Kenny4343]

No, that's why I had given up before. I turbo'ed it for at least a minute or half a minute all the three times I tried, since the game didn't freeze as I said, but just stayed on the blank screen.

[Squall-Leonhart]

does an FFMPEG occur at this point?

[Kenny4343]

No, it wouldn've been simply skipable if so.

[Squall-Leonhart]

some ffmpegs are definitely not skippable in this game.

[Kenny4343]

That doesn't change the fact that there isn't any in there.
And you might be right but each cutscene and FMV has always been skippable for me, so.

[Squall-Leonhart]

which character are you playing

[Kenny4343]

Ventus, I assume this is the issue because of how people also kept saying they'd be stuck after completing the tutotial (which is pretty much the same, after completing the light balls event).

[Squall-Leonhart]

can you attach an affected save. i can't find anything relating to a light balls event on google

and the tutorial all goes fine for me(unless its a different tutorial than the one at the games start?)

[Kenny4343]

Hmm, thing is I already got past that part, but I'll try to get there again to give you the save file, moreover to check if the issue still persists.

[Squall-Leonhart]

I can see it now, i found where the orbs of light even occurs xD, i was a little further in and forgot that battles name.

@oioitff's last build using the Filter does not have the issue so that might be a place to start investigating.

[Kenny4343]

Okay here you go: https://www.dropbox.com/sh/7dq4128t4lv6ss6/UDtCHUssZC
It's a bit far from where it actually happens sorry, but it shouldn't take you long, it's the closest savefile I could find, all you have to do is to keep going until you reach the cutscene where you got to fight a big group of maleficent's goons (that's the event) once you complete it, the issue happens.
Edit: oh lol sorry, just saw your post, so did you get to reproduce it?

[Squall-Leonhart]

03:32:240 Sd::UpdateVS I[HLE]: HLE\sceAtrac.cpp:833 sceAtracReleaseAtracID(8)
03:33:845 idle0 I[HLE]: GLES\Framebuffer.cpp:610 Decimating FBO for 04174000 (480 x 272 x 1), age 6

Pauses on white screen for almost 3 minutes and then continues with

06:15:492 user_main I[HLE]: HLE\scePower.cpp:260 222=scePowerGetPllClockFrequencyInt()
06:15:493 user_main I[HLE]: HLE\scePower.cpp:226 scePowerSetClockFrequency(333,333,166)
06:15:508 idle0 I[HLE]: GLES\Framebuffer.cpp:610 Decimating FBO for 00154000 (128 x 128 x 0), age 6
06:15:508 idle0 I[HLE]: GLES\Framebuffer.cpp:610 Decimating FBO for 00154000 (128 x 128 x 2), age 6
06:15:786 user_main I[HLE]: GLES\Framebuffer.cpp:415 Creating FBO for 00154000 : 128 x 128 x 0
06:15:786 user_main I[HLE]: GLES\Framebuffer.cpp:415 Creating FBO for 00154000 : 128 x 128 x 2
06:16:873 user_main I[HLE]: HLE\sceAtrac.cpp:1060 sceAtracSetDataAndGetID(0900d4a0, 0001e110)
06:16:874 user_main W[HLE]: HLE\sceAtrac.cpp:984 This is an atrac3 audio

[unknownbrackets]

If you can compile, it'd be very useful to know if there's any function active in that time. For example, if you use VerySleepy to profile it (starting when it starts the white screen, stopping afterward, and turn off the throttle), it should say what function is spending the most time.

I just wonder if something is doing a huge loop or something, since it eventually recovers...

-[Unknown]

[Squall-Leonhart]

good idea unknownbrackets

[Squall-Leonhart]

Sleep 10.828945 10.828945 16.638676 16.638676 KERNELBASE [unknown] 0
DrvPresentBuffers 6.637321 14.249403 10.198245 21.894211 nvoglv32 [unknown] 0
GLES_GPU::FastRunLoop 2.452177 31.664965 3.76777 48.653225 PPSSPPWindows
GLES_GPU::ExecuteOp 2.01636 16.184216 3.098138 24.867051 PPSSPPWindows
ShaderManager::ApplyShader 1.503011 6.233249 2.309377 9.577388 PPSSPPWindows
CityHash32 1.13189 1.13189 1.739149 1.739149 PPSSPPWindows
RtlTryEnterCriticalSection 0.856798 0.856798 1.31647 1.31647 ntdll [unknown] 0

[unknownbrackets]

Hmm, that looks mostly normal. It seems like it's rendering something.

And an actual PSP shows exactly the same thing, just without the delay? There's no extra scene that is being rendered wrong or something?

By the way, Sleep means you had throttle on...

-[Unknown]

[Squall-Leonhart]

PSP, JPCSP and @oioitff media builds all behave like this video

https://www.youtube.com/watch?feature=player_detailpage&v=0NiPpdYMV50#t=233s

master pauses for 3ish minutes between the level up/slowmo and the information screen at 4:00

Just took another profile
WaitForSingleObject 1705.134117 1705.134117 59.042949 59.042949 kernel32 [unknown] 0
WaitForMultipleObjects 392.268939 392.268939 13.582929 13.582929 kernel32 [unknown] 0
RtlInitializeExceptionChain 262.541412 2887.08559 9.090909 99.969877 ntdll [unknown] 0
MsgWaitForMultipleObjects 131.270706 131.270706 4.545455 4.545455 USER32 [unknown] 0
PeekMessageW 131.270706 131.270706 4.545455 4.545455 USER32 [unknown] 0
DispatchMessageW 130.398656 130.446623 4.515258 4.516919 USER32 [unknown] 0
Sleep 98.868343 98.868343 3.423472 3.423472 KERNELBASE [unknown] 0
DrvPresentBuffers 3.253231 6.95261 0.112648 0.240745 nvoglv32 [unknown] 0
GLES_GPU::FastRunLoop 0.973009 13.651314 0.033692 0.472698

NativeRender 0 128.877043 0 4.46257 PPSSPPWindows

[Squall-Leonhart]

Sorted by inclusive instead

RtlInitializeExceptionChain 262.541412 2887.08559 9.090909 99.969877 ntdll [unknown] 0
BaseThreadInitThunk 0 2624.544178 0 90.878968 kernel32 [unknown] 0
WaitForSingleObject 1705.134117 1705.134117 59.042949 59.042949 kernel32 [unknown] 0
endthreadex 0 1311.837125 0 45.424423 MSVCR100 [unknown] 0
WorkerThread::WorkFunc 0 1050.165646 0 36.363636 PPSSPPWindows
std::thread::RunAndDelete<std::thread::Func<std::tr1::_Bind<void void std::tr1::_Bind1<std::tr1::_Callable_pmf<void (__thiscall http::Download::*const)(void) http::Download 0> http::Download *> > > > 0 1050.165646 0 36.363636 PPSSPPWindows
DrvCopyContext 0.120022 396.016339 0.004156 13.712688 nvoglv32 [unknown] 0
WaitForMultipleObjects 392.268939 392.268939 13.582929 13.582929 kernel32 [unknown] 0
[691A98CC] 0 262.54141 0 9.090909 nvoglv32 0
[746412DA] 0 262.537384 0 9.09077 DSOUND 0
[691A7EC7] 0 262.47841 0 9.088728 nvoglv32 0
[74642E13] 0 262.467377 0 9.088346 DSOUND 0

[Squall-Leonhart]

this comment by solarmystic might relate to this issue

#2039 (comment)

[unknownbrackets]

Unfortunately, probably not. In that case, it's hanging trying to play a video.

-[Unknown]

[Squall-Leonhart]

last working 40877cf
first build able to confirm as broken since then 89b2fdf

builds between these 2 change sets won't load the mickey riding a book video and the pause menu won't appear (FPS counter stops dead) so i can't get in game to test them.

there are also alot of No Builds in that gap.

[unknownbrackets]

Oh, I wonder if it's the same as the loop in Twin Brave, which seems ADSR related. That's the most likely thing in that range.

Does it help to replace this in Core/HW/SasAudio.cpp:

    case STATE_SUSTAIN:
        WalkCurve(sustainRate, sustainType);
        break;
    case STATE_RELEASE:
        WalkCurve(releaseRate, releaseType);
        if (height_ <= 0) {
            height_ = 0;
            SetState(STATE_OFF);
        }
        break;

With:

    case STATE_SUSTAIN:
        WalkCurve(sustainRate, sustainType);
        if (height_ <= 0) {
            height_ = 0;
            SetState(STATE_RELEASE);
        }
        break;
    case STATE_RELEASE:
        WalkCurve(releaseRate, releaseType);
        if (height_ <= 0) {
            height_ = 0;
            SetState(STATE_OFF);
        }
        break;

-[Unknown]

[Squall-Leonhart]

unfortunately not :(

or, wait, let me try from a full rebuild

no, it did not help.

[Squall-Leonhart]

though the issue does seem to be in the SasAudio source.

[unknownbrackets]

You could try a hack (if you have a savestate right before it, for example) by replacing this in Core/HLE/sceSas.cpp:

u32 sceSasGetEndFlag(u32 core) {
    return 0xFFFFFFFF;
}

If it doesn't hang this way, that means we're keeping a sound effect on too long.

-[Unknown]

[Squall-Leonhart]

Your code would not compile,

However, entered as

u32 sceSasGetEndFlag(u32 core) {
    u32 endFlag = 0;
    for (int i = 0; i < sas->maxVoices; i++) {
        if (!sas->voices[i].playing)
            endFlag |= (1 << i);
        return 0xFFFFFFFF;
    }

and the problem is gone, so yes, it seems a sound effect is being held onto too long.

[Squall-Leonhart]

u32 sceSasGetEndFlag(u32 core) {
    u32 endFlag = -1;
    for (int i = 0; i < sas->maxVoices; i++) {
        if (!sas->voices[i].playing)
            endFlag |= (1 << i);
    }

also works

[Squall-Leonhart]

not sure if this breaks something, but it works

u32 sceSasGetEndFlag(u32 core) {
    u32 endFlag = 0;
    for (int i = 0; i < sas->maxVoices; i++) {
         if (sas->voices[i].playing ? true : false);
            endFlag |= (1 << i);
    }

    DEBUG_LOG(HLE,"sceSasGetEndFlag(%08x)", endFlag);
    return endFlag;
}

[unknownbrackets]

if (sas->voices[i].playing ? true : false); // <--- notice the stray semicolon

Those semicolons, they'll get you every time.

-[Unknown]

[Squall-Leonhart]

not as stray as you would think,

with the ;. the following line is considered an empty controlled statement and the game transitions past to the next screen immediately
without the ; the next line is not considered an empty controlled statement and hangs on the white screen

so would that imply that

            endFlag |= (1 << i);

is not returning the desired result?

[unknownbrackets]

With the ;, your code is equivalent to:

u32 sceSasGetEndFlag(u32 core) {
    u32 endFlag = 0;
    for (int i = 0; i < sas->maxVoices; i++) {
        endFlag |= (1 << i);
    }

    DEBUG_LOG(HLE,"sceSasGetEndFlag(%08x)", endFlag);
    return endFlag;
}

In other words, the problem is the voice is still playing. You could try (SasAudio.cpp again):

void ADSREnvelope::Step() {
    // Let's end the envelope right away.
    height_ = 0;
    SetState(STATE_OFF);
    return;

    // This unreachable code here mainly to look pretty.
    switch (state_) {
...

To verify if it's the ADSR envelope at all. If that doesn't help, you can try disabling looping:

    else if (flags == 3) {
        if (loopEnabled_) {
            // Nah, looping is for suckers.
            //loopAtNextBlock_ = true;
        }
    }

If neither helps, then maybe we're not handling the voice right at all...

-[Unknown]

[Squall-Leonhart]

Verified to be the ADSR Envelop with the first code change.

Disabling Looping also works. but i assume the second block of code disables the envelope anyway.

[hrydgard]

Somebody mentioned a while ago that the functionality of the Sas library appears to be very similar to the PSX SPU. So maybe some of this is relevant: http://psx.rules.org/spu.txt

[Squall-Leonhart]

@unknownbrackets
i utilised your first suggestion with a slight alteration

    case STATE_SUSTAIN:
        WalkCurve(sustainRate, sustainType);
        if (height_ < sustainLevel) {
            height_ = 0;
            SetState(STATE_RELEASE);
        }
        break;
    case STATE_RELEASE:
        WalkCurve(releaseRate, releaseType);
        if (height_ <= 0) {
            height_ = 0;
            SetState(STATE_OFF);
        }
        break;

this works, but im probably breaking things elsewhere

[unknownbrackets]

@hrydgard according to this it doesn't stop on sustain:
https://github.com/smokku/pcsx-rearmed/blob/master/plugins/dfsound/adsr.c#L146

PEOPS has some interesting comments in its code, which seem to disagree on the matter.

Hmm. Maybe I can find a way to test it.

-[Unknown]

[Squall-Leonhart]

dfsound is the evolution of PEOPS Dsound after pete released the source

PEOPS has a number of bugs not present in dfsound, so the comments in peops cannot be completely trusted to be correct (not that I don't trust Pete, what i don't trust is what was and wasn't known at the time he created the plugin)

[Squall-Leonhart]

Fixed by #2186