[MPlayer-users] Major regression in r36452, sig 11 demux_open

Reimar Döffinger Reimar.Doeffinger at gmx.de
Tue Sep 24 09:49:33 CEST 2013


Josh <joshf87 at live.com> wrote:
>On 9/22/2013 03:34, Reimar Döffinger wrote:
>> On Sun, Sep 22, 2013 at 12:29:05AM -0400, Josh wrote:
>>> I thought this was an FFmpeg regression but reverting to a known
>>> good FFmpeg revision didn't fix the regression. Basically, in every
>>> single h264/webm file I am getting a "Signal 11: demux_open" error.
>>> I have traced this to r36452. There might be other formats affected.
>>
>> Should be fixed.
>> It only happened "randomly", depending on your system, compilation
>> options, configuration etc.
>> The backtrace was completely misleading though.
>> I don't know if you care about these things, but
>> http://www.drmemory.org/ is a great tool for Windows that like
>> valgrind on Linux/OSX makes tracking down these things much simpler.
>> _______________________________________________
>> MPlayer-users mailing list
>> MPlayer-users at mplayerhq.hu
>> https://lists.mplayerhq.hu/mailman/listinfo/mplayer-users
>>
>>
>
>Is this easier to decipher? Do you care about the 'leak' errors?

Not easier to decipher, but at least contains the correct information :-)
But just to avoid misunderstandings: you can confirm it is fixed?

>Error #1: UNINITIALIZED READ: reading register eax
># 0 av_dict_get                       [libavutil/dict.c:43]

This one points directly to the issue: an uninitialized dict pointer.
Some of the leaks sound worrying, but I suspect they only appear because MPlayer crashed and this could not free things. So unless they appear when it does not crash they are not relevant.



More information about the MPlayer-users mailing list