Bug#1038827: libwebm: Failed tests in big-endian architectures

Boyuan Yang <[email protected]> Wed, 21 Jun 2023 16:51:46 -0400
Newsgroups	gmane.linux.debian.ports.hppa
Organization	Debian Project
Message-ID	<c093472d5c07324fb415a90e114e5398569c55b2.camel__19405.2746215434$1687380890$gmane$org@debian.org>
Source: libwebm
Version: 1.0.0.29-1
Tags: sid
Severity: important
X-Debbugs-CC: [email protected] [email protected] [email protected] [email protected] [email protected]

Hi all,

Package libwebm was introduced into Debian in 2021 for .webm support
(and to avoid library vendor copies). However, its unit test has been broken
on big-endian architectures since very beginning as shown in buildd page at
https://buildd.debian.org/status/package.php?p=libwebm (s390x/hppa/powerpc/
ppc64/sparc64 for 1.0.0.30-2 and earlier), causing FTBFS.

Currently an convenient copy of libwebm is bundled and embedded in src:aom,
and embedded libwebm lacks unit tests. When I am working on src:aom and
trying to replace bundled libwebm with the packaged one, current failed unit
tests in libwebm will need some handling, or at least in the release archtiecture
(s390x, such as
https://buildd.debian.org/status/fetch.php?pkg=libwebm&arch=s390x&ver=1.0.0.30-2&stamp=1687373358&raw=0
):


==================================================
./testing/mkvmuxer_tests.cc:863: Failure
Expected equality of these values:
  muxer_mm.luminance_min()
    Which is: 30
  parser_mm->luminance_min
    Which is: 0
./testing/mkvmuxer_tests.cc:864: Failure
Expected equality of these values:
  muxer_mm.luminance_max()
    Which is: 40
  parser_mm->luminance_max
    Which is: 0
./testing/mkvmuxer_tests.cc:865: Failure
Expected equality of these values:
  muxer_mm.r()->x()
    Which is: 0.1
  parser_mm->r->x
    Which is: 0
./testing/mkvmuxer_tests.cc:866: Failure
Expected equality of these values:
  muxer_mm.r()->y()
    Which is: 0.2
  parser_mm->r->y
    Which is: 0
./testing/mkvmuxer_tests.cc:867: Failure
Expected equality of these values:
  muxer_mm.g()->x()
    Which is: 0.1
  parser_mm->g->x
    Which is: 0
./testing/mkvmuxer_tests.cc:868: Failure
Expected equality of these values:
  muxer_mm.g()->y()
    Which is: 0.2
  parser_mm->g->y
    Which is: 0
./testing/mkvmuxer_tests.cc:869: Failure
Expected equality of these values:
  muxer_mm.b()->x()
    Which is: 0.1
  parser_mm->b->x
    Which is: 0
./testing/mkvmuxer_tests.cc:870: Failure
Expected equality of these values:
  muxer_mm.b()->y()
    Which is: 0.2
  parser_mm->b->y
    Which is: 0
./testing/mkvmuxer_tests.cc:871: Failure
Expected equality of these values:
  muxer_mm.white_point()->x()
    Which is: 0.1
  parser_mm->white_point->x
    Which is: 0
./testing/mkvmuxer_tests.cc:872: Failure
Expected equality of these values:
  muxer_mm.white_point()->y()
    Which is: 0.2
  parser_mm->white_point->y
    Which is: 0
[  FAILED  ] MuxerTest.Colour (0 ms)
[ RUN      ] MuxerTest.ColourPartial
[       OK ] MuxerTest.ColourPartial (0 ms)
[ RUN      ] MuxerTest.Projection
./testing/mkvmuxer_tests.cc:930: Failure
Expected equality of these values:
  muxer_proj.pose_yaw()
    Which is: 1
  parser_proj->pose_yaw
    Which is: 0
./testing/mkvmuxer_tests.cc:931: Failure
Expected equality of these values:
  muxer_proj.pose_pitch()
    Which is: 2
  parser_proj->pose_pitch
    Which is: 0
./testing/mkvmuxer_tests.cc:932: Failure
Expected equality of these values:
  muxer_proj.pose_roll()
    Which is: 3
  parser_proj->pose_roll
    Which is: 0
[  FAILED  ] MuxerTest.Projection (0 ms)

===================================================


I am looking for help from porters with experience in big-endian architectures
on debugging of the test failures. Given the high popcon of reverse dependencies
(src:aom, 30000+), I may allow unittest failures for now but still keep this bug
report open against src:libwebm. Any help would be appreciated.

Best,
Boyuan Yang
signature.asc (application/pgp-signature, 833 B)