Closed adamchainz closed 5 months ago
@gaborbernat I haven’t got a Windows setup for verifying this on - would you mind trying this on your test suite?
I also do not have one :D used the Windows GHA to catch this :D
Okay, I tested on Windows in CI too, in #458. The test suite is not ready to run on Windows without more work, so I’m going to merge and release this without any test for now.
Please give 2.14.2 a try!
Fixes #456.
The relevant crash from the test log actually seems to be the initial failure:
PyObject_GetAttrString
fails to getclock_gettime
andclock_gettime_ns
on Windows because they only exist on Unix. It returnsNULL
and each time sets theAttributeError
as the current error.patch_if_needed
did not clear this, leading to Python's call check to raise aSystemError
caused by theAttributeError
, making time-machine fail to start travelling.I'm not sure which exact change in Python made this failure occur, so far it seems the
SystemError
should have been raised on older versions too. Ah well.@gaborbernat I haven’t got a Windows setup for verifying this on - would you mind trying this on your test suite?