[FFmpeg-trac] #5048(undetermined:closed): -map 0:v causes segfaults on 2pass libvpx-vp9 on multi video-track file
FFmpeg
trac at avcodec.org
Sun Dec 6 01:53:49 CET 2015
#5048: -map 0:v causes segfaults on 2pass libvpx-vp9 on multi video-track file
-------------------------------------+-------------------------------------
Reporter: kagami | Owner:
Type: defect | Status: closed
Priority: important | Component:
Version: git-master | undetermined
Keywords: libvpx | Resolution: invalid
crash SIGSEGV | Blocked By:
Blocking: | Reproduced by developer: 0
Analyzed by developer: 0 |
-------------------------------------+-------------------------------------
Comment (by kagami):
>you argue that it is perfectly ok for libvpx to crash if the logfile that
is passed does not match the source file?
No, segfault is just the symptom here.
>Generally, libraries should never crash on invalid input
Completely agree. This crash should be fixed in libvpx and this issue
should have a bit different name and description. I just don't have enough
knowledge to analyze whether ffmpeg correctly encodes two VP9 tracks with
2pass at the same time (corrupted logfile is just an assumption), so I
demonstated the most obvious consequence from the user point of view in
this ticket. It still would be bad if you get output but in bad quality if
e.g. libvpx can tolerate wrong input.
I think it might be related to the logfile processing because:
* I encode two tracks but get only single log file
* If I encode only 1 track on a second pass, I still get segfault
I may be complete wrong here though. So ideally I should file crash bug to
the libvpx first, wait for the fix, make sure library explicitly forbids
incorrect logs/input, try again. I just thought it would be easier if
someone at ffmpeg bugtracker could tell whether it's related to ffmpeg or
just libvpx issue.
--
Ticket URL: <https://trac.ffmpeg.org/ticket/5048#comment:4>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker
More information about the FFmpeg-trac
mailing list