Log file Viewer

[10:50:28] <ronny> xafer: sup

[10:57:06] <xafer> ronny: hello

[10:58:42] <ronny> xafer: i have a very vauge hunch on why your issue happened in combiation with parallel testing and the custom tmpdir fixture

[10:59:12] <ronny> (and why it was unlikely to occur with lest parallel runner)

[11:04:16] <ronny> xafer: main problem is interaction with getbasetemp

[11:04:19] <xafer> Anyway we could clarify this hunch ?

[11:05:14] <ronny> i have an idea on fixing the underlyin problem, but that is a rewrite of the pytest config object and plugin manager initialization

[11:05:24] <xafer> ^^

[11:06:37] <ronny> my current hunch is that due to a race condition the tmpdir manager is initialized twice

[11:07:04] <xafer> To be honest, I don't have that much time, so the proposed fix works for me. It's just that I'd like to be sure the real issue has been identified.

[11:07:06] <ronny> and py.test removes tempdirs after a few calls

[11:07:37] <ronny> xafer: as far as i can tell your original code just made it more likely to trigger the behaviour

[11:07:56] <ronny> so bruno fixed it by using the normal code path that doesnt trigger it

[11:08:09] <xafer> Maybe we could add a few trace calls for the 'basetemp.remove()' case ?

[11:10:12] <ronny> probably - its really not a pleasure to debug that one ^^

[11:28:06] <xafer> I'm testing with https://github.com/xavfernandez/pytest/commit/22584011e6ca101d6853cb142d49c1b857181e76

[11:28:51] <xafer> This place was apparently not called

[11:47:00] <xafer> ronny: it seems we are only hitting https://github.com/pytest-dev/pytest/blob/fde0890d584cedb43bfb7c5abecdc61af74a9784/_pytest/tmpdir.py#L56 once

[12:25:09] <ronny> xafer: once per test run?

[12:25:34] <xafer> no once for the whole run

[12:25:35] <ronny> intersting, completely not what i was thinkingof

[12:26:11] <xafer> cf https://travis-ci.org/pypa/pip/jobs/146619665

[12:31:00] <xafer> Note that I also set keep=0 to avoid the pruning

[12:31:26] <xafer> Which does not change the result

[12:31:43] <xafer> So it is unclear where the cleanup happens :-/