[FFmpeg-trac] #6403(undetermined:new): Use of both posix_memalign() and realloc() on same memory block not supported
FFmpeg
trac at avcodec.org
Wed May 31 02:57:30 EEST 2017
#6403: Use of both posix_memalign() and realloc() on same memory block not
supported
-------------------------------------+-------------------------------------
Reporter: jrummell | Owner:
Type: defect | Status: new
Priority: normal | Component:
Version: git-master | undetermined
Keywords: | Resolution:
Blocking: | Blocked By:
Analyzed by developer: 0 | Reproduced by developer: 0
-------------------------------------+-------------------------------------
Comment (by jrummell):
Since I'm not sure what Chromium / Chrome does, I asked
[https://groups.google.com/a/chromium.org/forum/#!msg/chromium-
dev/QaFjUlGD6mI/AQPatcW4BAAJ others].
It appears that Chromium / Chrome uses
[https://cs.chromium.org/chromium/src/third_party/tcmalloc/chromium/src/debugallocation.cc?l=1202
tcmalloc]. I was pointed to several internal bugs from 2013 that discussed
this issue, and the resolution back then was that since the documentation
for realloc() does not indicate that it definitely maintains memory
alignment, it should not be allowed (and thus the warning in debug
builds). I checked realloc() documentation, and did not find anything that
says that alignment is preserved (although it would if the memory could be
realloc'd in place). So I'm unlikely to make a strong case that tcmalloc
needs changing.
Since there is data in FFmpeg that needs to be aligned, having Chromium
move from using posix_memalign() in general seems like it would result in
slower performance. Unfortunately there's no av_aligned_alloc() to help
distinguish between the places where aligned memory is required, and where
it's optional.
--
Ticket URL: <https://trac.ffmpeg.org/ticket/6403#comment:6>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker
More information about the FFmpeg-trac
mailing list