Drop sage-cleaner interface

[antonio-rojas]

This doesn't work with meson, since sage-cleaner is not available, and breaks parallel doctesting

It is only used in interfaces/quit.py, where we can replace it by the existing temporary dir mechanism

Fixes #39873

[tobiasdiez]

Is it much work to migrate this to the built-in python tempdir?

(Triggered by the fact the last PR that touched the cleaner interface #33213 notes "Afterward, the custom functions tmp_dir() and tmp_filename() can be deprecated in favor of tempfile.TemporaryDirectory() and tempfile.NamedTemporaryFile().")

[antonio-rojas]

I haven't been able to make it work. If I use TemporaryDirectory then the directory seems to be deleted immediately, unless I use delete=False, in which case it is not removed on sage exit.

This seems to be similar to the parallel doctesting issues we found on Python 3.13 upgrade, see #39201 (cc @tornaria)

[tobiasdiez]

Okay, thanks for checking!

[antonio-rojas]

cc @orlitzky for input since I just noticed his previous sage-cleaner removal attempt at #33213 (which didn't go through at the end)

Note that this doesn't actually remove the sage-cleaner script, and in particular the sage script installed by setuptools still uses it. OTOH the meson sage script never called sage-cleaner, and I haven't seen any issue in a few weeks of testing: everything (temporary dirs and spawned subprocesses) is properly cleaned up on exit. I tried crashing a sage process which had gap and gp subprocesses running and they were killed too (temporary files weren't deleted, but I don't think anybody should expect a crashing program to clean up its temporary files)

[tobiasdiez]

Looks good from my side.

[orlitzky]

temporary files weren't deleted, but I don't think anybody should expect a crashing program to clean up its temporary files)

A lot of these can be "fixed" by #36322. The sage tmp_filename() and tmp_dir() functions leave the files/directories until sage exits cleanly. Then if sage does not exit cleanly, stuff gets left behind. But often, a temporary file/directory is much more temporary than that. And in that case, using the python tempfile module directly lets us clean up earlier (when we are done with the file/directory), making it irrelevant if sage eventually crashes. In any case, this is a minor issue IMO. I've been in no hurry to work on issue #36322 because the replacement code will be a lot nicer when we can assume python-3.12.

I also do not think there are any major problems with leaving processes behind. It depends on the program, but e.g. GAP and maxima are well-behaved, and it isn't really a sage problem if some optional package doesn't handle SIGPIPE correctly. Sage (the library) can still be used to launch other processes... but again, if the user wants to launch a daemon using popen or something like that... okay? I think the least surprising behavior is if the sage shell acts like the python shell in this regard (i.e. no sage-cleaner).

[github-actions]

Documentation preview for this PR (built with commit 18c2dfd; changes) is ready! 🎉
This preview will update shortly after each push to this PR.