[FFmpeg-devel] [PATCH] libavfilter:vf_thumbnail.c: Fix bug in buffer handling for RGB width

Chris Kennedy ckennedy at groovy.org
Thu Feb 26 15:18:41 CET 2015

On Sun, Feb 22, 2015 at 7:26 AM, Clément Bœsch <u at pkh.me> wrote:

> On Wed, Feb 18, 2015 at 06:20:54PM -0800, Chris Kennedy wrote:
> > More details attached, debug level and gdb backtrace.  Hopefully this
> > helps, and I will work on getting a file I can share showing the issue.
> >
> > Thanks
> >
> [...]
> > #14 0x000000000042790c in process_input (file_index=0) at ffmpeg.c:3418
> > #15 0x0000000000429397 in transcode_step () at ffmpeg.c:3719
> > #16 0x00000000004294a5 in transcode () at ffmpeg.c:3771
> > #17 0x00000000004299b3 in main (argc=21, argv=0x7fffffffdcf8) at
> ffmpeg.c:3951
> This doesn't really look like an overread, more like you hit a memory
> limit and your malloc implementation doesn't like it, or something along
> these lines (178 frames is a lot). Keep in mind that thumbnail might not
> be the best solution to getting thumb. Using scene detection in select
> filter and pick the Nth scene change might pick a better one, with less
> memory usage.
> You might want to try valgrind to make sure it's not an overread/write.

Yeah I am now going to just do the normal frame stuff with scene change or
similar, as it seems I explored it deeper and hit yet another oddity in an
email I sent last night about it where it crashed in a different way using
hardly any memory and when the program exits (yet only when it is parsing
the exact second that has some oddity, other video and other parts of the
exact same video do not trigger, larger frame counts don't either like 300).

Thanks for the help, yeah I wish I could send the video that triggers it,
but it is one of our DVD rips of Rurouni Kenshin we did with handbreak or
something years ago and must itself have some bad mpeg4 frame information
causing this I guess.  I am doubting this would be easy/humanly possible to
recreate and find a non-copyrighted video doing it without spending day and
night hunting something down (which the bug is so odd I can't really
understand what it is doing, it seems to only crash now at the end IF the
frame is extracted it always crashes on, odd).


> --
> Clément B.
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

More information about the ffmpeg-devel mailing list