Hi Alexandre
But 3.8 does not even build most of the days....
Trying rerun the Salsa CI does not help because log is too long and truncated.
The full log from salsaci is available in the artifacts section of the
project on salsa:
https://salsa.debian.org/python-team/packages/matplotlib/-/artifacts
which shows some tests failing.
They are almost all in matplotlib/tests/test_backends_interactive.py
failing with either SIGABRT or dying with a «could not connect to
display :1» error (despite the entire test suite being with xvfb, so
there should be a display available). The tests involve running code in
subprocesses which may or may not be relevant.
I don't know if that gives us a clue as to why they fail... the tests
pass when run locally, so it's hard to see what the next steps would be.
I suspect uploading the current matplotlib 3.8 that is in git to
experimental is a reasonable next step. If this failure isn't seen on
the buildd then that's a useful data point for future work - and if it
is seen on the buildd then it's more urgent future work.
Getting that uploaded helps with one level of disentanglement of
build-dep-cycles and also gets some actual tests being run for
matplotlib since they are currently not, and fixes a problem with numpy2
and matplotlib (only seen on arch:all builds which have not been done
since the numpy2 transition started).
Let's get that through experimental, working ok, and then uploaded to
sid prior to thinking about 3.10... but yes, let's look at 3.10 in
experimental immediately after.
regards
Stuart
--
Stuart Prescott http://www.nanonanonano.net/ stu...@nanonanonano.net
Debian Developer http://www.debian.org/ stu...@debian.org
GPG fingerprint 90E2 D2C1 AD14 6A1B 7EBB 891D BBC1 7EBB 1396 F2F7
