This goes beyond bug 225156 by marking everything that takes > 0.5s to run on my MacBook Pro (16-inch, 2019) as slow. This makes a larger number of tests get skipped by default than with ./Tools/Scripts/test-webkitpy, but halves the execution time. Having looked at a graph of test execution time, I think the > 0.5s cutoff is the right one to balance running as much as possible as quickly as possible.
Created attachment 441380 [details] Patch
Comment on attachment 441380 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=441380&action=review > Tools/Scripts/webkitpy/style/main_unittest.py:36 > +from webkitpy.test.markers import slow Can we maybe not do this file? I only call it out because the style checker is the thing engineers not familiar with webkitpy are most likely to change. Would be good if we kept full coverage for folks making those changes, since they're unlikely to know to change the flags they're passing to test-webkitpy.
(In reply to Jonathan Bedard from comment #2) > Comment on attachment 441380 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=441380&action=review > > > Tools/Scripts/webkitpy/style/main_unittest.py:36 > > +from webkitpy.test.markers import slow > > Can we maybe not do this file? I only call it out because the style checker > is the thing engineers not familiar with webkitpy are most likely to change. > Would be good if we kept full coverage for folks making those changes, since > they're unlikely to know to change the flags they're passing to > test-webkitpy. Note that this currently has no impact on `test-webkitpy`; there's probably an argument that we want to default to running all the tests and then allowing developers to opt-in to running the "not slow" subset (which once we're using pytest everywhere is just `pytest -m 'not slow'`).
<rdar://problem/84548211>