Regression:Bakemonogatari corrupted frames in video.

[daniel229]

Since #5212

Reverted that change

[sum2012]

Do #5245 fix it ?

[unknownbrackets]

I guess this is somehow affecting the speed of decoding or whether it waits or something. Is this near the front of the video or in the middle?

It may be that these values need to come from feedback from the decoder.

-[Unknown]

[daniel229]

@sum2012 that does not help.
@unknownbrackets It is the biginning of video,the first frame is correct,then it is being corrupted for a second.
here is a video
http://www.youtube.com/watch?v=Vr1zDG0IE54

[dbz400]

@daniel229 , if just simply set the following , it did help , right ?

ctx->avc.avcDecodeResult = 0;

[daniel229]

That does not help,and the first frame start to be corrupted,the master build the first frame is fine.Revert either one would not help,should revert both.

[dbz400]

I see.

[unknownbrackets]

My theory at least in the past was that initAddr should contain a "1" when a frame was fully decoded, and "0" otherwise, or something.

So possibly, for the first few calls this game expects a 0, but another game might expect a 1.

Best thing would be to look at the code around this call in one of the affected games and see what logic it uses based on the resulting value. Clearly if this value changes whether there are corrupted frames or not, it must affect the game's behavior in some way.

-[Unknown]

[dbz400]

Yep, Project Diva 2 same as well .It start at 1st frame