[FFmpeg-devel] [PATCH 2/3] Acquire lock when initializing parsers.

Michael Niedermayer michaelni at gmx.at
Thu Dec 6 13:44:03 CET 2012


On Thu, Dec 06, 2012 at 08:16:26AM +0100, Reimar Döffinger wrote:
> 
> 
> On 6 Dec 2012, at 03:55, Michael Niedermayer <michaelni at gmx.at> wrote:
> 
> > On Thu, Dec 06, 2012 at 03:07:21AM +0100, Michael Niedermayer wrote:
> >> On Tue, Nov 27, 2012 at 09:59:31PM +0100, Reimar Döffinger wrote:
> >>> This is necessary since they might be initializing or
> >>> even using static VLC tables.
> >> 
> >> this breaks ffmpeg.c with multiple inputs
> >> the parsers get initialized from read packet and ffmpeg runs the
> >> 2 demuxers from seperate threads if iam guessing right
> >> i wasnt able to make it happen in gdb though ...
> >> 
> >> also this patchset breaks the frame thread encoder, i think ive fixed
> >> that one but iam not sure what to do about the parser issue quickly
> >> now so Ill revert the patch so we have time to think and work on
> >> fixing this without pressure ...
> > 
> > To reproduce: (with git prior to reverts)
> > ./ffmpeg  -y -i matrixbench_mpeg2.mpg -i matrixbench_mpeg2.mpg -map 0:0 -map 1:0 out.mp4
> > 
> > Fails >50% of the time but not in gdb
> 
> Don't know when I will be able to run myself, so could someone explain what "fails" means?

sorry, fails == aborts with assert failures, i belive i saw both
asserts from the lock functions fail but maybe it was just one


> Also I expect that enabling core dumps should allow getting a backtrace.

> Does ffmpeg.c provide a lock manager?

no

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

I have often repented speaking, but never of holding my tongue.
-- Xenocrates
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20121206/9830d1e5/attachment.asc>


More information about the ffmpeg-devel mailing list