Since it's just been a month since my last +1 rotation, I tried to focus
on following up on past stuff and getting more update-excuse bugs filed.
### emscripten ###
Tests for 2.0.12~dfsg-2ubuntu1 failed due to llvm compatibility issues.
Debian has incremented this package a few times in experimental, and I
found their first increment, 2.0.13-1, resolved the compatibility
issue. I sync'd that version in and emscripten migrated successfully.
### libreoffice vs. openjdk-12 ###
Test was failing on armhf but passed on retrigger. Both
libreoffice/1:7.1.5~rc2-0ubuntu1 and openjdk-12 are migrated in impish.
### wpewebkit ###
A simple rebuild succeeded, and the package migrated successfully.
### websockify ###
A simple rebuild succeeded, and the package migrated successfully.
### ovn / 21.06.0-0ubuntu1 build failure on armhf ###
A simple rebuild succeeded, and the package migrated successfully.
(I found a comment in upstream's bug tracker suggesting the testsuite is
known to be a bit flaky, but nothing tangible enough worth filing a bug
about...)
### node-rollup-plugin-node-polyfills and ts-node ###
Both of these packages seem to just have flaky tests, and migrated
successfully on retriggering. I suspect adjustment to timeouts might
resolve the flakiness if it resurfaces, and opened bug LP: #1938948 on
that note.
### aodh / 1:12.0.0+git2021072116.10383f4d-0ubuntu2 ###
The upstream requires.txt had a depends on sqlalchemy 1.4 which is not
available yet. I proposed a MP to remove the versioned depends, which
was accepted and landed (thanks Corey!) aodh has migrated successfully
in impish.
### node-websocket-driver ###
The issue here is just that there's a warning emitted during testing
that trips autopkgtest's "any stderr is fail" rule. I opened LP:
#1938949 with a modification of the test to filter the warning, but
before uploading would like to get a sanity check from someone else that
papering this over is a good enough workaround. (Feel free to upload
this yourself if you agree.)
### crash / 7.3.0-1ubuntu1 test failure on ppc64el ###
Guessing this was just a race condition with the archive; it wanted a
version of linux-image that wasn't available when the test executed.
Whatever the cause, a simple rebuild succeeded and allowed crash to
migrate.
### breezy ###
Since several things are backed up on breezy (including silver-platter,
breezy-debian, etc.), I followed on from work by Dan and Christian and
spent some time studying the package internals. I expanded on Dan's
partial-fix to address a couple more test cases. Essentially, python
has recently changed where '.' in sys.path causes issues with module
loading; Dan's fix is to substitute CWD here, and that fixed a number of
problems; my addition was to update a few test cases that were expecting
'.' in error messages to instead accept '.*' so they'll match the CWD.
The remaining test failures appear to be caused by something in some
code that manually creates a module-space, 'breezy.testingplugins' and
then inserts synthetic modules into it. Maybe whatever changed for
handling '.' affected artificial module loading functionality as well?
Dan suggested extracting a test case to send upstream might be useful.
breezy/plugin.py:_load_plugin_module() might be the place to start for
that.
### python-meshio / 4.4.6-1 test failure on s390x ###
Guessing a tolerance value used in a test is triggering some data type
oddity on s390x. I filed update-excuse bug LP: #1939057 for tracking
this.
### mutter on armhf ###
Successfully migrated on a simple retrigger.
The stacking/override-redirect.metatest test seems to pass or fail about
50/50, so seems to have something non-deterministic going on. vanvugt
notes it occurs upstream as well. I filed LP: #1939058 for an
update-excuse next time it happens. If it does, maybe the testcase
could be disabled until there's an upstream fix?
### mesa vs. mutter on ppc64el ###
After mutter itself migrated, it hit a test failure with mesa on
ppc64el. Something to do with missing drivers for the Clutter backend.
I didn't have time to investigate, but retriggered it in case it's just
something else being flaky.
### thin / 1.8.0-1 test failure on armhf ###
The issue here has to do with the squid proxy on armhf. I'm not sure
how to debug this one, but someone familiar with squid's configuration
on armhf probably could. Opened update-excuse bug LP #1939086.
### gauche-gtk ###
Heather's analysis suggested this project may be abandoned and perhaps
should be dropped. This seems sensible to me as well. There are some
rdepends that would need dealt with as well, notably wiliki, scmail, and
skktools. wiliki hasn't been updated since precise, and scmail since
natty. skktools is just a suggests. There's several other gauche-*
source and binary packages but I verified none have rdepends beyond this
set.
I think a good next action would be to file removal requests for wiliki,
scmail, gauche, libgauche-*, and gauche-*.
### glance ###
This just needs another dependency, python3-oslo.limit. glance is
maintained by Canonical folks, so I just filed LP: #1938965 to track the
update-excuse for future +1'ers.
### netplan.io ###
The test log showed it's failing on `netplan apply` but didn't show that
command's stderr. I noticed slyon was already working on this, and
opened update-excuse LP: #1875677 with him subscribed to coordinate.
Indeed, he debugged the issue locally and found it to be a race
condition. He uploaded 0.103-0ubuntu3 with a revert, allowing
netplan.io to successfully migrate.
### open-ath9k-htc-firmware / 1.4.0-106-gc583009+dfsg1-1 ###
This FTBFS may be straightforward to fix:
Essentially, from what I can tell, this package grabs a copy of binutils
and applies some patches. One of these patches,
binutils-2.34_fixup.patch, fails to apply with the latest binutils.
However I *think* the patch simply needs to be refreshed. In any case
there's only four hunks failing so probably not hard to update.
It looks like upstream routinely refreshes these patches, but just
hasn't done so for binutils 2.37 yet.
### delve / 1.6.1-1 test failure on arm64 ###
From Dan's analysis I wonder if we could just force-reset-test this on
arm64, but Brian rightly points out there's no urgency so for now can
wait for a response and/or fix from upstream or the kernel team.
### limnoria ###
Not entirely sure of the cause of this test failure, but Debian is
hitting it too. I opened LP: #1938957 to track on our end.
### linux ###
The page feels like it has a larger than usual number of linux-*
packages. Are these migration items already on someone's todo list?
### Week's Stats ###
2021-07-30 223 update excuse records found
2021-08-02 431 update excuse records found
253 update excuse records found
2021-08-03 225 update excuse records found
2021-08-04 211 update excuse records found
2021-08-05 209 update excuse records found
2021-08-06 210 update excuse records found
--
ubuntu-devel mailing list
ubuntu-devel@lists.ubuntu.com
Modify settings or unsubscribe at:
https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
