+1 Maintenance Report

Graham Inggs ginggs at ubuntu.com
Mon Jul 26 13:30:13 UTC 2021

Last week was my first shift on +1 maintenance.  Below are the
packages I looked at, grouped by their current status.


I found that the autopkgtests had already regressed in Debian since
January 2021 (Debian bug #988836).  I confirmed that version
0.5.32+dfsg1-1 in -testing had also regressed, then opened a MP
against hints-ubuntu.  Once this was merged,
moment-timezone-js/0.5.32+dfsg1-2+2021a migrated.

There were autopkgtest failures on arm64 and ppc64el.  I found a sleep
delay in mailman3-api test to wait for the mailman3 service to
restart.  I confirmed that increasing the delay from 10 to 30 in a PPA
allowevagrant (2.2.9+dfsg-1ubuntu1d the tests to passes reliably on
all architectures.  I uploaded mailman3/3.3.3-1ubuntu1 and it

There was a missing build on arm64 which was last retried on
2021-05-10.  I retried it, but it failed OOM.  I also tried reducing
parallelism in a PPA, which passed on arm64, but failed at first on
ppc64el.  I then retried the arm64 build in the archive, and this time
it succeeded and polymake migrated.

There were missing builds on amd64, arm64 ppc64el and s390x.  I found
these were due to 1000+ missing symbols because of building with LTO
enabled.  In a PPA, I tested adding libsass to lto-disabled-list and
found the builds on am64 and arm64 were successful, but ppc64el and
s390x were still missing a handful of symbols.  Instead of trying
maintaining a delta of 1000+ C++ symbols in Ubuntu, I decided to
upload lto-disabled-list/14 and libsass/3.6.4+20210501-1ubuntu1, and
both migrated.  While investigating, I had noticed Debian had
libsass/3.6.5-1 in experimental, which only contained some additional
bugfixes compared to the previous snapshot.  I merged and uploaded
libsass/3.6.5-1ubuntu1 which also migrated.

vagrant / vagrant-bindfs
I found that the autopkgtests of vagrant-bindfs had already regressed
in -release.  I filed LP: #1937076 and opened a MP against
hints-ubuntu.  Once this was merged, vagrant/2.2.9+dfsg-1ubuntu1

There were missing builds on arm64 and riscv64 where micropython had
built in the past.  I found LP: #1931955 where William Wilson had
attached a patch.  I confirmed in a PPA that this at least fixed the
builds on arm64 and riscv64, but did not help armhf, ppc64el and s390x
where micropython had never built.  I sponsored the upload of
micropython/1.14+ds-1ubuntu1 and it migrated.

varnish / varnish-modules
A new upload of varnish caused the autopkgtests of varnish-modules to
fail.  I uploaded a no-change rebuild of varnish-modules, then
triggered the autopkgtests of varnish/6.5.2-1 and
varnish-modules/0.16.0-2.1build1 together, and they both migrated.
This is being addressed in Debian bug #991348.

The autopkgtest on arm64 regressed.  I found this was due to the arm64
build being disable in 2.35.0+ds-1.1 and opened a MP against
hints-ubuntu.  Once this was merged, angelscript/2.35.0+ds-1.1

openmsx / openmsx-catapult
openmsx-catapult and openmsx were blocked on each other due to a
missing build on riscv64.  After investigating, I found upstream had
missed a small change when they enabled the build on riscv64.  After
testing in a PPA that the build was successful, I submitted the change
upstream, uploaded openmsx/17.0-1ubuntu1 and it migrated.

glibc/ xmltooling
I found autopkgtests of xmltooling had regressed in release sometime
around mid-July 2021.  I filed LP: #1937348 then and a MP against
hints-ubuntu.  Once this was merged, glibc/2.33-0ubuntu9 was no longer
blocked by xmltooling.


There were missing builds on arm64, armhf, riscv64 and 390x.  This was
caused by stylish-haskell/ adding a build-dependencies on
libghc-ghc-lib-parser-dev and libghc-ghc-lib-parser-prof which has not
built on these architectures.  I filed LP: #1937070 requesting removal
of the old stylish-haskell binaries.

There were missing builds on armhf, riscv64 and s390x.  This was due
to qemu-web-desktop/21.05.03-2 restricting its architectures to those
where qemu-system-* is available.  I filed LP: #1937275 requesting
removal of the old qemu-web-desktop binaries.


opa-basic-tools has an unsatisfiable dependency on linux-cpupower
which is not built in Ubuntu.  I found LP: #1215411 and marked it also
affecting opa-ff.

I was able to confirm the segmentation fault in DBstats locally.  I
eventually found LP: #1895685, re-opened it, and tagged it

Missing builds on all architectures.  I filed LP: #1937225 and tested
the patch from the Debian bug in a PPA.  While the builds were
successful, the autopkgtests only passed on amd64 and armhf.

Missing builds on amd64, arm64, armhf, ppc64el and s390x.  I found LP:
#1934534 and tagged it 'update-excuse'.

Missing builds on amd64, arm64, armhf, ppc64el and s390x.  I found LP:
#1934829 and tagged it 'update-excuse'.

Missing builds on ppc64el and s390x.  I found LP: #1931225 and marked
it also affecting xorgxrdp.

Autopkgtest regressions on arm64 and armhf.  I filed LP: #1937253 and
tagged it 'update-excuse'.

Autopkgtest regression on armhf:  I filed lp: #1937256 and tagged it

I was unable to reproduce the ctt and kleborate autopkgtests locally,
even in a VM with only 512MB RAM allocated.  I did find that cct's
autopktests in Debian started to take around double the time with the
new version of ncbi-blast, but no bugs have been filed yet.

More information about the ubuntu-devel mailing list