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
{{ message }}
Repository navigation
test_repeat is skipped across all -compat wrapped libraries #413
Not sure if it's worth inventing a whole new framework for xfailing parts of a test. OTOH, this is relatively common need to be able to express "this fails for uint64 and is fine otherwise".
To close the loop: gh-424 restricted drawing repeats to signed integers only, with a #TODO comment and a matching note in #301. I thought about adding a second test to draw all dtypes, so that we have both a "correct" test (xfailed all around) and some actual testing or repeats, but in the end it feels like too much hassle for too little gain.
Activity
ev-br commentedon Feb 8, 2026
Ran a small check to see if there are other tests of this kind:
So after #418, and outside of
test_special_cases, the only two offenders aretest_finfowith float32/complex64 : this is a known wontfix, test_finfo fails on numpy #317test_repeatNot sure if it's worth inventing a whole new framework for xfailing parts of a test. OTOH, this is relatively common need to be able to express "this fails for uint64 and is fine otherwise".
ev-br commentedon Feb 17, 2026
To close the loop: gh-424 restricted drawing
repeatsto signed integers only, with a#TODOcomment and a matching note in #301. I thought about adding a second test to draw all dtypes, so that we have both a "correct" test (xfailed all around) and some actual testing orrepeats, but in the end it feels like too much hassle for too little gain.