I've found exact place of deadlock.
it is
http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/pthread.c;hb=HEAD#l859

patch for debug tracing: http://pastie.org/2988933
debug log: http://pastie.org/2988951
call stack: http://pastie.org/2988941

2011/12/9 Sergey Radionov <[email protected]>

>
>
> 2011/12/8 Ronald S. Bultje <[email protected]>
>
> Hi,
>>
>>
>> On Thu, Dec 8, 2011 at 5:42 AM, Jean-Baptiste Kempf <[email protected]>wrote:
>>
>>> Hello *<tab>,
>>>
>>> On Wed, Dec 07, 2011 at 07:32:00PM +0100, Ronald S. Bultje  wrote :
>>> > w32thread: port fixes to pthread_cond_broadcast() from x264.
>>>
>>> This improves the situation quite a bit, but isn't perfect.
>>>
>>> Noticeable on 7, but very visible on XP, frame-based decoding on h264,
>>> can deadlock quite a bit, notably with many cores.
>>
>>
>> Note that the bug report I saw involved slice threading, not frame
>> threading, IIRC.
>>
> Bug reported by myself was really with slice threading as long as I found
> it when playing mpeg file. And subj patch successfully fix it. But now
> problem is with .avi files, and therefor with frame threading.
>
>
>> Anyway, please show me a backtrace of all threads on a full debug build
>> of Libav, ideally compiled without -fomit-frame-pointer and with -O0.
>> Whether it's VLC or avconv isn't totally relevant.
>>
>> Ronald
>>
>> _______________________________________________
>> libav-devel mailing list
>> [email protected]
>> https://lists.libav.org/mailman/listinfo/libav-devel
>>
>>
>
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to