[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#968503: nmu: mutter, muffin on 32-bit architectures (and gtk+3.0 on m68k)



Package: release.debian.org
Severity: normal
User: release.debian.org@packages.debian.org
Usertags: binnmu
X-Debbugs-Cc: m68k@buildd.debian.org, ftpmaster@ports-master.debian.org

pango 1.46.0-1 introduced an ABI break on 32-bit architectures, affecting
packages that subclass PangoRenderer (which is rare, but is done by some
important packages like GTK). I've just uploaded 1.46.0-2 to revert the
ABI break, with a patch that was already accepted upstream.

According to codesearch there only seem to be 8 affected packages, most
of which haven't been rebuilt for a long time according to their buildd
logs, so their binaries already expect the 1.44.x and 1.46.0-2 ABI and
do not need to be changed.

However, packages that were unlucky enough to already be rebuilt against
1.46.0-1 might crash with 1.46.0-2 on 32-bit architectures. On release
architectures, that means mutter and its Cinnamon fork muffin:

nmu mutter_3.36.5-1 . armel armhf i386 mipsel . unstable . -m "rebuild against pango1.0 with #968337 fixed"
dw mutter_3.36.5-1 . armel armhf i386 mipsel . unstable . -m "libpango1.0-dev (>= 1.46.0-2)"
nmu muffin_4.6.3-1 . armel armhf i386 mipsel . unstable . -m "rebuild against pango1.0 with #968337 fixed"
dw muffin_4.6.3-1 . armel armhf i386 mipsel . unstable . -m "libpango1.0-dev (>= 1.46.0-2)"

To -ports people (cc'd): for the -ports architectures, mutter and
muffin should be rebuilt as above on all the 32-bit ports, with the
same dep-wait. Additionally, some differences in build order mean that
gtk+3.0 needs a rebuild on m68k (but not on release architectures or
other 32-bit ports).

Thanks,
    smcv


Reply to: