Friday 23 April 2021

+1 maintenance report

Here's my report for the week of April 19-23.
Unfortunately, I got a bit distracted by a late netplan.io regression, which I
worked on a bit on the side.


### aws-shell ###

aws-shell / 0.2.1-1 ftbfs on amd64 (all)

This package is pretty outdated and unmaintained upstream. It depends on
prompt-toolkit 1.0, last released in Oct 2019 and was not updated to support
the new major versions of prompt-toolkit 2.x or 3.x
It was removed from Debian testing, so I opened a removal bug  on LP as well:
https://pad.lv/1925034


### auctex ###

auctex / 11.91-2ubuntu2 ftbfs on amd64 (all)

It took me some time to get into the emacs lisp language here. Turned out I
wasn't able to fix the problem, but Debian has a newer version. So I merged the
new package (incl. fix) from Debian.
I staged the upload in a PPA, because of final freeze:
https://launchpad.net/~slyon/+archive/ubuntu/plusone-hh-freeze/+packages


### baycomepp ###

baycomepp / 0.10-15 ftbfs on amd64

Ugh.. this package is very outdated. Upstream has vanished and I couldn't find
a Debian packaging repository. It uses source format 1.0 and is very hard to
work with. After getting into this non-quilt build I was able to revert a change
which was apparently made at some time in the past years, to fix the build with
a newer GCC. By re-defining the 'only_inline' macro to 'extern inline', i.e. as
it is currently in the orig sources.
I've sent the patch to Debian and Launchpad:
https://bugs.debian.org/987250
https://pad.lv/1925212


### badger ###

badger / 2.2007.2-2 ftbfs on armhf

This is a known and expected issue, I've created a LP bug, tagged 'ftbfs' for
reference, so others don't investigate too much about this issue:
https://pad.lv/1925214


### beets ###

beets / 1.4.9-7 ftbfs on amd64 (all)

This build PASSed locally (multiple times) in a Hirsute sbuild environment:
"Ran 1885 tests in 240.622s, OK (skipped=45, expected failures=9)"
So it was most probably an intermittent failure.


### sshuttle ###

I tried running the sshuttle test on a "m1.large" instance, to verify Bryce's
assumption about adding it to 'big_packages'. Unfortunately, canonistack seems
to be broken and does not spawn instances correctly (many requests fail with
503 errors).
This Canonistack problem is most probably related to RT#130493.


### atlas ###

atlas / 3.10.3-8ubuntu7 ftbfs on any arch

The build failure is fixed in Debian, but it required a complex merge, including
special handling of s390x instruction set and cross-compilation of Z14 on Z13.
Also, there were some incompatibilities between the new debhelper-compat 13 and
the previous dh-exec logic. I staged the upload in a PPA:
https://launchpad.net/~slyon/+archive/ubuntu/plusone-hh-freeze/+packages
https://pad.lv/1910830
https://bugs.debian.org/957015

### alltray ###

alltray / 0.71b-1.2 ftbfs on any arch

This package is unmaintained upstream, orphaned & removed in Debian. It depends
on the deprecated 'gdk-pixbuf-xlib-2.0' module. So I opened a removal bug:
https://pad.lv/1925459


### ap-utils ###

ap-utils / 1.5-4 ftbfs on any arch

GNU gettext was upgraded from 0.19 -> 0.21, which now requires the 'external'
parameter to the AM_GNU_GETTEXT macro and also does not allow the 'intl'
directory to be listed in the Makefile's SUBDIRS variable.
I staged the upload in a PPA and forwarded the patch to Debian:
https://launchpad.net/~slyon/+archive/ubuntu/plusone-hh-freeze/+packages
https://bugs.debian.org/978372


### gsmlib ###

gsmlib / 1.10+20120414.gita5e5ae9a-0.3ubuntu1 ftbfs on any arch

This is the same GNU_GETTEXT issue as described above in "ap-utils", fixed in
the same way.
I staged the upload in a PPA and forwarded the patch to Debian:
https://launchpad.net/~slyon/+archive/ubuntu/plusone-hh-freeze/+packages
https://bugs.debian.org/978339


### cgminer ###

cgminer / 4.9.2-1build2 ftbfs on any arch

This package showed a linking error ("multiple definition of ...") when compiled
with GCC >= 10, due to the new -fno-common default. Setting this back to
-fcommon fixes the build failure.
I staged the upload in a PPA:
https://launchpad.net/~slyon/+archive/ubuntu/plusone-hh-freeze/+packages


### clamfs ###

clamfs / 1.2.0-2build2 ftbfs on any arch

The upgrade to glibc 2.33 deprecated the mallinfo() function in favor of the new
mallinfo2() call. Special care needs to be taken for the printing of 'size_t',
having a different size on 32bit vs 64bit systems.
I staged the upload in a PPA and forwarded the patch to upstream:
https://launchpad.net/~slyon/+archive/ubuntu/plusone-hh-freeze/+packages
https://github.com/burghardt/clamfs/pull/64


### csh ###

csh / 20110502-6 ftbfs on any arch

The upgrade to glibc 2.32 deprecated the  sys_siglist, _sys_siglist and
sys_sigabbrev arrays. All programs should use strsignal() instead.
I staged the upload in a PPA:
https://launchpad.net/~slyon/+archive/ubuntu/plusone-hh-freeze/+packages


### gramophone2 ###

gramophone2 / 0.8.13a-3.1 ftbfs on any arch

GCC >= v10 uses -fno-common by default, which leads to "multiple
definitions of ..." errors during the build.
By defining the -fcommon flag via DEB_CFLAGS_MAINT_APPEND (before the
cdbs includes) the build can be fixed. The attached patch does exactly
this.
I staged the upload in a PPA and forwarded the fix to Debian:
https://launchpad.net/~slyon/+archive/ubuntu/plusone-hh-freeze/+packages
https://bugs.debian.org/966862


### ISO testing ###

As we are in release week, I've also helped with some ISO testing of the RC
images a bit:

Ubuntu Server Subiquity/amd64 (SUCCESS) Install BIOS
Ubuntu Base/amd64 (SUCCESS) Smoke test
Ubuntu Base/arm64 (SUCCESS) Smoke test
Ubuntu Base/armhf (SUCCESS) Smoke test


### TODO ###
Remove from the archive:
- aws-shell
- alltray, DONE by @doko

Upload to the archive (after the final freeze):
- auctex
- baycomepp
- atlas
- ap-utils
- gsmlib
- cgminer
- clamfs
- csh, UNAPPROVED pending
- gramophone2

Verify autopkgtest:
- sshuttle on "m1.small" vs "m1.large" instance (big_packages)


Cheers,
  Lukas