UninitCond in vp3_dequant

[Matt Wolenetz]

Summary of the bug:
==7732== Conditional jump or move depends on uninitialised value(s)
==7732== at 0xAA54C9: vp3_dequant (vp3.c:1365)
==7732== by 0xAA608E: render_slice (vp3.c:1582)
==7732== by 0xAA816B: vp3_decode_frame (vp3.c:2104)
==7732== by 0x966AA4: frame_worker_thread (pthread.c:394)
==7732== by 0x5A57E99: start_thread (pthread_create.c:308)

I hit this in current master (fd68371b072824ee6f63e38e1b2f6921a05bedcc) as well as when synced to much older commits (I checked as far back as ~Oct 1 2012).

How to reproduce:

./configure --valgrind=<<path to valgrind>> --disable-optimizations --extra-cflags='-fno-inline -fno-omit-frame-pointer -fno-builtin -fno-optimize-sibling-calls' --enable-debug --enable-static --disable-shared && make clean && make -j32 && <<path to valgrind>> --tool=memcheck --num-callers=30 --gen-suppressions=all --demangle=no --leak-check=yes --show-possibly-lost=no ./ffmpeg_g -i <<media file>> -f null - 2>&1

See attached media file. I found this with Valgrind-3.9.0.SVN from ​http://src.chromium.org/chrome/trunk/deps/third_party/valgrind/binaries

Patches should be submitted to the ffmpeg-devel mailing list and not this bug tracker.

[Carl Eugen Hoyos]

To make this a valid ticket, please provide the valgrind / ffmpeg command line together with the complete, uncut console output.

[Matt Wolenetz]

Uncut console output showing issue from git-master

[Matt Wolenetz]

Synced to fd68371b072824ee6f63e38e1b2f6921a05bedcc, the attached repro.log was produced by:

../../trunk/src/third_party/valgrind/linux_x64/bin/valgrind --tool=memcheck --num-callers=30 --gen-suppressions=all --demangle=no --leak-check=yes --show-possibly-lost=no ./ffmpeg_g -i big_dims.ogv -f null - 2>&1 | tee repro.log

[Carl Eugen Hoyos]

Attached patch fixes the messages here, I don't know if there is an issue at all.

[Elon Musk]

Let me repeat once more what is already mentioned:

Patches should be submitted to the ffmpeg-devel mailing list and not this bug tracker.

[Matt Wolenetz]

I haven't tested upstream ffmpeg, but that attached patch fixes the issue downstream. Please get it submitted to ffmpeg-devel. Thanks!

[Carl Eugen Hoyos]

Replying to wolenetz:

I haven't tested upstream ffmpeg, but that attached patch fixes the issue downstream.

If you believe the patch fixes an issue (I don't know), please consider submitting it or set up a git clone and ask Michael to merge the patch.

[Carl Eugen Hoyos]

While I have seen no indication that there was a bug, my patch was committed under a pseudonym.