3.14 pytest issues: suppress sqlite3 warnings, add context managers to mp calls - #3719

OK this should be working now. I think it may be an issue with PROJ-- have just suppressed the sqlite warnings specific to the cartopy tests in the image_path function in pytest

see #3718 (comment)

suspect the new FDSN bug is 3.14 being stricter about a missing pool.close() or pool.terminate() in mass downloader. may have to go through all these.

I've tried locally with 3.14 though and could not replicate any of these.. have you been able to get these fails locally?

I've tried locally with 3.14 though and could not replicate any of these.. have you been able to get these fails locally?

I have not.. but I think the issue could be here:

Traceback (most recent call last):
  File "<string>", line 1, in <module>
    from multiprocessing.resource_tracker import main;main(14)
  File "/home/runner/miniconda3/envs/test/lib/python3.14/multiprocessing/resource_tracker.py", line 34, in <module>
    _IGNORED_SIGNALS = (signal.SIGINT, signal.SIGTERM)
                        ^^^^^^^^^^^^^
AttributeError: module 'signal' has no attribute 'SIGINT' (consider renaming '/home/runner/miniconda3/envs/test/lib/python3.14/site-packages/obspy/signal/__init__.py' since it has the same name as the standard library module named 'signal' and prevents importing that standard library module)

Which leaves the mp stuff hanging specifically in pytest due to the overlapping "signal" imports. I already had to change something else regarding this for the 1.5 release using some creative importing. Fortunately by wrapping the 3-4 multiprocessing calls in a context manager (now done here) the problem doesn't seem to fail in pytest anymore and is probably a bit more robust anyway.

Have re-shuffled a bit and mostly am now just suppressing the specific sqlite3 pytest warnings in pytest.ini.

There was also a new issue with py3.10 ubuntu mindep, the 5 minute timeout didn't seem to be enough. Have bumped to 10 and made the libmamba solver explicitly used. On that tangent maybe we should turn off the 3.8 and 3.9 tests now?

Looking at this again, I think it should go into maintenance branch? I've merged maintenance into master, so the sqlalchemy fixes are in there now and I will retrigger tests here now to see what the status is.

so the thing with "signal" being shadowed by obspy.signal is solved in #3789 - leaving this open for the other issues, we need to check if they still come up.