[FFmpeg-devel] FATE review

[Måns Rullgård]

FATE is generally in good shape at the moment, so now is a good time
to look at the persistent failures.

- AVR32
  GCC bugs cause some tests to fail.

- Blackfin, gcc 4.5
  Linker admits being broken, presumably triggered by something gcc did:
    /usr/libexec/gcc/bfin-unknown-linux-uclibc/ld: BFD (Linux/GNU Binutils) 2.20.51.0.9.20100526 assertion fail bfd/elf32-bfin.c:4524
    /usr/libexec/gcc/bfin-unknown-linux-uclibc/ld: LINKER BUG: .rofixup section size mismatch
    /usr/libexec/gcc/bfin-unknown-linux-uclibc/ld: final link failed: Nonrepresentable section on output

- PA-RISC
  Some tests intermittently failing with "Received signal 15: terminating"
  Something is killing the ffmpeg process.  Test timing out?

- PPC, gcc 4.0.4
  Many VP8 tests fail.

- Sparc/Linux, gcc 4.2.3 and 4.2.4
  h264-conformance-mv1_brcm_d fails with output cut short and
  "Received signal 15: terminating."  Timeout?
  gcc 4.3 unaffected

- Sparc/Solaris, gcc 3.4.6
  mpc demuxer broken as previously discussed.

- Sparc64/Linux
  VP8 test files missing.

- Sparc64/OpenBSD
  Threaded encoding fails intermittently.

- IA64
  Shared libs fail to link with GPREL22 error.

- x86_32/Linux
  gcc 2.95.3
    seek_test fails to link with pthread error.  Unable to reproduce
    on my systems.
  gcc svn 161806 (2010-07-04)
    Unidentified error in regtest-mxf
  icc 10.1.022
    regtest-mp2 fails, both encoded and decoded differ
  suncc 5.11
    regtest fails flashsv, jpegls, mxf, pixfmt, rgb, wmav1, wmav2
    Most failures in decoding part of test.

- x86_64/FreeBSD, clang
  Segfault in regtest-nuv.  The second loop in rtjpeg.c:get_block()
  is miscompiled.  Easily reproduced.

- x86_64/Linux
  icc 11.1.059
    mpeg4-als and vc1 tests fail
  icc 11.0.083
    Fails regtest-error (mpeg4), all(?) h264, svq3, vc1
  icc 10.1.022
    Fails vc1
  suncc 5.11
    regtest fails ffm, jpegls, pcm, pixfmt, wmav1, wmav2
    MLP decoder disabled.

-- 
M?ns Rullg?rd
mans at mansr.com