fixed 1031056 1:22.0.0-1 tags 1031056 + bookworm thanks
El 9/5/24 a las 14:05, Thomas Goirand escribió:
I tried rebuilding 3 times ceilometer with 1 CPU (with nr_cpu=1 in grub) in a virtual machine, and I couldn't reproduce. Note that it has run unit tests twice on each run, with python 3.11 and 3.12, so effectively, I tried 6 times...
Thanks for checking. I see that you skipped bookworm in your tests, which is what I reported at the time. So I went ahead and checked the build in bullseye, bookworm and trixie again, using single-cpu systems of several flavors. This is what I got so far (numbers show failures/tries). bullseye 0/18 bookworm 17/32 trixie 0/32 I could do more tests than that, but with the above data I think it's fair to say that the problem I reported (FTBFS randomly in bookworm) is not fixed yet.
It's possible that the latest eventlet fixed it. Please try to reproduce this bug again, and re-open the bug if you feel like it's still not fixed.
Unfortunately, I can't do that because it's against the rules of version tracking. What I can do is to adjust the metadata to reflect reality, at least that way we document clearly and faithfully what's going on. I'm afraid that this might remain unfixed in bookworm forever, even if it happens the currently supported stable distribution. I understand that you are busy and have to prioritize. Considering that the bug happens randomly and it may be avoided using nocheck, I could agree that there are other FTBFS bugs which require more attention. For example, I would be interested to know your plans regarding python-pycdlib. I still think that one should be fixed in bookworm and trixie, because build-essential does not imply that /tmp is in tmpfs. (Please reply in #1002789). Thanks.
