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
Currently, the leak checking encompasses the whole setup/call/teardown cycle. This is maybe not optimal, since some pytest fixtures appear to leak, and those leaks are usually not of interest.
It would be possible to split the refleak run cycle todo
setup()
cleanup()
before = get_counts()
test()
cleanup()
after = get_counts()
deltas = after - before
teardown()
instead of the current
setup()
test()
teardown()
cleanup()
after = get_counts()
deltas = after - before
before = after
but it comes at the price of two cleanups, which I guess are required before "get_counts" for stable numbers.
The text was updated successfully, but these errors were encountered:
Not sure how feasible this is though: Calling via the hook pytest_runpytest_call enables e.g. the logging fixture (on by default) which hold references from call to teardown and would seem to leak if done like this. OTOH, calling directly without the hook disables output capture, so also not ok.
Currently, the leak checking encompasses the whole setup/call/teardown cycle. This is maybe not optimal, since some pytest fixtures appear to leak, and those leaks are usually not of interest.
It would be possible to split the refleak run cycle todo
instead of the current
but it comes at the price of two cleanups, which I guess are required before "get_counts" for stable numbers.
The text was updated successfully, but these errors were encountered: