You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Check the specific Windows GIX_TEST_IGNORE_ARCHIVES=1 failures
This modifies the `test-fixtures-windows` job that tests on Windows
with `GIX_TEST_IGNORE_ARCHIVES=1` so that, instead of checking that
no more than 14 failures occur, it checks that the failing tests
are exactly those that are documented in GitoxideLabs#1358 as expected to fail.
The initial check that no tests have *error* status is preserved,
with only stylistic changes, and kept separate from the subsequent
logic so that the output is clearer.
The new steps are no longer conditional on `nextest` having exited
with a failure status, since (a) that was probably unnecessary
before and definitely unnecessary now, (b) at last for now, the
comparison is precise, so it would be strange to pass if the diff
were to have changes on *all* lines, and (c) this makes it slightly
less likely that GitoxideLabs#1358 will accidentally stay open even once fixed.
The current approach is to actually retrieve the list of tests
expected to fail on Windows with `GIX_TEST_IGNORE_ARCHIVES=1` from
the GitoxideLabs#1358 issue body. This has the advantage that it automatically
keeps up to date with changes made to that issue description, but
this is of course not the only possible approach for populating the
expected value.
Two changes should be made before this is ready:
- As noted in the "FIXME" comment, the job should currently fail
becuase the performance test reported to fail in GitoxideLabs#1358 is not
being filtered out from the expected failures list. It's left in
as of this commit, to verify that the job is capable of failing.
(After that, the performance test should either be filtered out
or removed from the list in GitoxideLabs#1358, but the former approach is
currently preferable because I have not done diverse enough
testing to check if the failure on my main Windows system is due
to that system being too slow rather than a performance bug.)
- The scratchwork file should be removed once no longer needed.
0 commit comments