flintlib
msolve hangs CI after its tests are completed
Check this one for instance: https://github.com/flintlib/flint/actions/runs/18970263945/job/54176148389
It seems to do all tests, but does not manage to exit. Not sure why, but feels like some Github Action specific issue.
@vneiger have you seen anything similar?
No, I don't remember seeing something similar in msolve CI runs. I have just looked at the CI workflow files of both flint and msolve and could not identify a meaningful difference, possibly responsible for this (but I am far from expert on this stuff). Also I've just re-run make check on my side with "-j5 --output-sync=target" and did not get anything unusual.
You write "for instance": has this been observed several times already?
Here are all the instances I can find: https://github.com/flintlib/flint/actions/runs/17006782978/job/48217495792 https://github.com/flintlib/flint/actions/runs/16720460021/job/47323521714 https://github.com/flintlib/flint/actions/runs/17264574164/job/48993599495 https://github.com/flintlib/flint/actions/runs/18882313196/job/53888512505 https://github.com/flintlib/flint/actions/runs/18984838940/job/54226011237 https://github.com/flintlib/flint/actions/runs/18970263945/job/54176148389 https://github.com/flintlib/flint/actions/runs/18968335803/job/54169886357
In a recent CI run I got a check failure for diff_realroot1:
https://github.com/vneiger/flint/actions/runs/19030225200/job/54342150611#step:13:180
When inspecting the content of the checks in the above runs that were hanging indefinitely, exactly one "PASS: ..." is missing compared to a full successful run of checks. And this happens to be the PASS concerning diff_realroot1.
So, I suppose it was not hanging after completing all the tests, but was hanging due to some bug in the check for diff_realroot1. I've opened https://github.com/algebraic-solving/msolve/issues/241 .
We've not managed to reproduce the failed (or rather hanging) realroot1 test on our machines, although it continued to appear sporadically in FLINT CI jobs. A possible culprit piece of code was identified and fixed a few days ago (https://github.com/algebraic-solving/msolve/pull/244), so let's keep an eye on whether this issue persists in FLINT CI jobs.