Jeremy Kloth <jeremy.kloth+python-trac...@gmail.com> added the comment:
Also of note, a largish temporary directory (16K+ entries) seemed to be causing a slowdown in the cleanup of the tests, thus triggering the failures. A quick purge later and the tests seem to run to completion. Although the tests are currently passing, they are still broken from a Windows directory removal standpoint. It is good to know that the failures can be recreated if needed by refilling the tempdir (at least on my buildbot). This issue brings to light a few things: 1) tests MUST use the support.rmtree to prevent spurious failures 2) tests need to cleanup better; each run leaves ~10 entries in TEMP 3) a per-builder tempdir would mitigate the excessive tempdir issue Note that addressing 2 wouldn't eliminate 3 as non-buildbot usage can still fill the tempdir, just not as fast as our buildbot seems to :) I would really like to see #3 solved (maybe a `tmp` directory next to the `build` directory on the builders themselves?) If it is in a fixed location, I would then be able to use a tmpfs-like solution for it, reducing test times 20+% (in my hacked up local testing). ---------- _______________________________________ Python tracker <rep...@bugs.python.org> <https://bugs.python.org/issue35795> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
