+1 maintenance shift, Oct 3-7.
We're well into the freeze for kinetic release, and -proposed is in good
shape in terms of both number of packages (between 169 and 222 packages in
-proposed over the past week) and age of what's there. This means there's
really fairly little +1 maintenance work to be done (at least, that we can
reasonably expect to make progress on). I focused specifically on getting
NBS cleared out for the release.
Last week I got notification of blasr being fixed in Debian unstable, fixing
the last blocker for pbbam NBS. This brought the number of NBS packages in
unstable down to 2, with 3 reverse-dependencies.
ruby-vips
This package fails its autopkgtests against vips 8.1.0 in -proposed. It is
also a build failure (https://bugs.debian.org/1017186), but more important
is that it's a runtime regression. Somehow this regression was allowed into
Debian testing (temporary removal of ruby-vips from testing that allowed
vips to migrate and reset the baseline, maybe?) I checked upstream for a fix
and they don't have one, and I don't speak ruby, so I've removed ruby-vips
and its 1 reverse-dependency (ruby-image-processing) from kinetic. This
will let vips migrate, removing one of the reverse-dependencies of NBS
libopenexr25.
openvdb
New upstream version in -proposed was FTBFS, running out of memory.
Removing build parallelism was enough to reduce memory consumption and let
the package build on all archs. This migrated, and let us remove
libopenexr25.
prusa-slicer
The new upstream version of openvdb in -proposed itself introduces a library
transition. No-change rebuild of prusa-slicer uploaded.
insighttoolkit4
Infuriatingly, there are two insighttoolkit source packages: the new one
which builds fine, and the old one which fails to build and has all the
actual reverse-dependencies. And being old makes it particularly heinous to
git bisect the upstream to find where insighttoolkit5 fixed this, because of
all the other ways the intermediate commits will FTBFS with a recent
toolchain. Did a bisect and then only at the end did I notice there was a
Debian bug about the build failure that was marked 'pending' and had a
commit from the Debian maintainer that works around this. Anyway, uploaded
now with the upstream fix but this ended up failing to build with test
failures, so more investigation still needed.
libevent
And just as we were getting to the end, we were just in time to have another
library transition with libevent (transition needed because the library ABI
changed on rebuild due to dropped symbol with current glibc). So I:
- reviewed/accepted the package as a freeze exception
- did some no-change rebuilds of packages that depend on libevent-core
instead of libevent
- retriggered autopkgtests with --all-proposed so we were coherently testing
against the new binary packages
- dealt with libreswan, a reverse-dependency that's FTBFS in the release
pocket and in -proposed. It's also an unseeded universe package that's
two years behind Debian due to a delta and not having been merged, so
rather than having to drop it from the release I went ahead and merged the
new upstream version from Debian after confirming that it builds.
- noticed that the armhf autopkgtest queues had actually not been moving
since Wednesday. Poked at the infrastructure (the lxd autopkgtest
dispatcher missed a juju config change?) and got the queues moving again.
- calling it quits there for the week!
--
Steve Langasek Give me a lever long enough and a Free OS
Debian Developer to set it on, and I can move the world.
Ubuntu Developer https://www.debian.org/
slangasek@ubuntu.com vorlon@debian.org
