Feature or enhancement

Currently you can use--match and--ignore options for positive and negative filtering of test cases by name. But negative patterns always win. So you can select a class of tests and then exclude some tests from it, but you cannot exclude a class of tests except some tests.

Many programs that have similar options (can also be named--include/--exclude) use different algorithm. They apply rules in the order, and the last rule wins. I propose to implement this in Python regrtests too. Examples:

RunFileTests tests intest_os, excludingtest_write:

./python -m test test_os -m FileTests -i test_write

Run all tests intest_os, excludingFileTests tests, but includingtest_write:

./python -m test test_os -i FileTests -m test_write

And, of course, any combinations are valid. This applies also to options that read patterns from files:--matchfile and--ignorefile. The implementation is actually somewhat simpler than the current one.

Linked PRs

• gh-110918: regrtest: allow to intermix --match and --ignore options #110919
• [3.12] gh-110918: regrtest: allow to intermix --match and --ignore options (GH-110919) #111167
• [3.11] gh-110918: regrtest: allow to intermix --match and --ignore options (GH-110919) #111168
• gh-110918: Fix side effects of regrtest test_match_tests() #116718
• [3.12] gh-110918: Fix side effects of regrtest test_match_tests() (GH-116718) #116726
• [3.11] gh-110918: Fix side effects of regrtest test_match_tests() (GH-116718) #116727