Commit 6fb2ca2
authored
infra: Scope incremental cascade per parameter set (#5086)
##### What this PR does / why we need it:
When an incremental class is parametrized at class level, a failure in
one parameter set (e.g. ipv4) should not cascade into another (e.g.
ipv6) since they are independent test cycles with fresh fixtures.
This was discovered while working on PR #5026 — adding
`@pytest.mark.jira("CNV-88755", run=False)` to a single test caused 6
xfails instead of just 2 skips, because `_previousfailed` is stored on
the class node which is shared across all parametrize combinations.
##### Which issue(s) this PR fixes:
##### Special notes for reviewer:
- Non-parametrized incremental classes are unaffected — all their tests
share the same key so behavior is identical to before.
- Currently two incremental classes use class-level parametrize:
-
`tests/network/l2_bridge/migration_stuntime/test_migration_stuntime.py`
- `tests/network/localnet/migration_stuntime/test_migration_stuntime.py`
##### jira-ticket:
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Tests**
* Improved incremental test failure tracking for parametrized tests, so
failures are tracked per-parameter set—leading to clearer xfail
attribution and more reliable test outcomes.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Signed-off-by: Anat Wax <awax@redhat.com>1 parent e42559f commit 6fb2ca2
1 file changed
Lines changed: 6 additions & 2 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
676 | 676 | | |
677 | 677 | | |
678 | 678 | | |
679 | | - | |
| 679 | + | |
| 680 | + | |
| 681 | + | |
| 682 | + | |
680 | 683 | | |
681 | 684 | | |
682 | 685 | | |
| |||
732 | 735 | | |
733 | 736 | | |
734 | 737 | | |
735 | | - | |
| 738 | + | |
| 739 | + | |
736 | 740 | | |
737 | 741 | | |
738 | 742 | | |
| |||
0 commit comments