third test rebuild of precise pangolin

Colin Watson cjwatson at
Tue Apr 3 14:00:56 UTC 2012

On Mon, Apr 02, 2012 at 10:15:38AM -0500, Kate Stewart wrote:
>    Given the rather large number of packages in main showing up on the
> list[1], and where we are in the release cycle,  if you could take a
> look a the list,  and please help resolve those rebuild ftbfs failures
> that you're able to assist with as soon as possible,  it would be
> appreciated. 

We're down to around 20 failures in main.  Here's a partial analysis of
the remaining failures:

aisleriot (i386, amd64, armhf):
  Fixed upstream
  I would have just backported it but the desktop team repository for
  this has some unreleased changes in it dating from months ago, and I
  don't know whether they'd like to just take 3.3.2 which I believe has
  this fixed.

bzr (armhf):
  Test failure on
  some smart server error that's beyond me to debug.  Could probably use
  the attention of a bzr hacker.

dbus-c++ (i386, amd64, armhf):
  Fixable with a merge from Debian, which I've just done.

evolution-exchange (i386, amd64, armhf):
evolution-indicator (i386, amd64, armhf):
  Bug indicates that this was fixed in evolution, so we should be able
  to retry shortly.

genshi (i386, amd64, armhf):
  Bug indicates that this is due to recent changes in Python 2.7 and
  that bugs have been filed with Genshi upstream, although I see no
  progress on those.

gnome-keyring (amd64):
  We've been here before (around beta-2, I think); its tests are racy,
  and IIRC the only thing to be done for now is to mash retry until it
  succeeds.  My previous investigations suggested that this should be
  better for 12.10 when we move to new upstream code, although I wasn't
  able to pin down a specific commit that fixed it since the code has
  been split in two so bisecting is painful.

heartbeat (i386, amd64, armhf):
  Needs to move to single <glib.h> includes rather than <glib/*.h>.

ibus-hangul (i386, amd64, armhf):
  Probably an ibus API change.  There are newer versions in Debian, so
  it's probably just a matter of selecting one that's a reasonable
  balance between building and not introducing too many new features.

libatomic-ops (i386):
  As recorded on the bug, this generally seems to fail on virtualised
  buildds but succeed on the regular ones, so we've been ignoring this
  for the purposes of test rebuilds.

libical (i386, armhf):
  There are lots of errors in here interleaved with real build output.
  The first two seem to be "No rule to make target
  `src/libical/icalderivedvalue.h', needed by `src/libicalss/icalss.h'"
  and "fatal error: libical/ical.h: No such file or directory", but it's
  kind of hard to unpick.  Perhaps this isn't parallel-make-safe?

m2crypto (armhf):
  Test failures in test_cookie_str_changed_mac and test_sslv23_no_v2.
  Possibly related to OpenSSL 1.0.1?  I've hit retry just in case, since
  the build that failed used libssl-dev 1.0.1-2ubuntu2 and there've been
  a couple of possibly-relevant changes since then.

memcached (i386):
  I couldn't reproduce this locally.  Perhaps the server team could take
  a look.

mpfr4 (amd64):
  ICE, as noted in the bug.  Somebody needs to extract the usual stuff
  from /usr/share/doc/gcc-4.6/README.Bugs for a compiler bug report.

pacemaker (i386, amd64, armhf):
  Needs to move to single <glib.h> includes rather than <glib/*.h>.

python-fixtures (i386):
  Looks like this might be a testtools regression.

python2.7 (amd64):
  Test suite hang.  I don't know what's going on here.

strigi (armhf):
  Symbols file mismatch.

subunit (i386, amd64, armhf):
  "ImportError: cannot import name Python26TestResult".  Robert Collins
  said something on IRC about a known testtools regression.

subversion (i386, amd64, armhf):
  Assorted test failures.  At first glance it looks as though they might
  be due to sort ordering changes in Python dictionaries; perhaps a
  sprinkling of sorted() or equivalent would help.

telepathy-qt4 (i386):
  "FAIL!  : TestStreamedMediaChan::testOutgoingCallBusy() Compared
  values are not the same".  Don't ask me why.

Colin Watson                                       [cjwatson at]

More information about the ubuntu-devel mailing list