Repository navigation
Suggestion: Using capturemanager.global_and_fixture_disabled #172
Description
I came across some odd behavior when I was debugging something related to terminal logging in pytest.
And noticed pytest-timeout was disabling capturemanager and doing it's own printing
Ref: https://github.com/pytest-dev/pytest-timeout/blob/2.3.1/pytest_timeout.py#L511
I think the better approach here may be to use global_and_fixture_disabled which is a context manager to temporarily disable capturemanager and write your own terminal outputs while it is disabled
https://github.com/pytest-dev/pytest/blob/main/src/_pytest/capture.py#L819
Any thoughts on whether this would make the logging logic simpler here?
Activity
[-]Suggestion: Using caplog.global_and_fixture_disabled[/-][+]Suggestion: Using capturemanager.global_and_fixture_disabled[/+]flub commentedon Apr 15, 2024
Sounds reasonable, it's likely pytest-timemout predates that context manager. I suppose your problem was that the capturing was never re-enabled but the test run continued?
Want to do a PR to switch to it and and make sure test coverage catches the edge cases you found?
ci: switch pytest-timeout to signal method; restore Phase 14 tests
Arham-Begani commentedon Oct 19, 2026
I reproduced a related diagnostic gap on current main (1bbaf76) with pytest 9.1.1: a thread timeout prints its stack dump but loses stdout/stderr buffered by the test's
capsysfixture. A subprocess test that prints distinct stdout/stderr markers and then times out shows neither marker in the output.I'll prepare a focused change using
global_and_fixture_disabled, preserving fixture-buffered output before dumping diagnostics, and regression coverage for capture fixtures and disabled global capture. This work is being performed by an AI coding assistant with the account owner's authorization; no human review is claimed.