Repository navigation
Commit c76ed80
committed
Don't hold the hatch uv var in a parallel test
The first version of uv_path_env_counts_only_when_a_guard_holds_it held
HATCH_ENV_TYPE_VIRTUAL_UV_PATH with an EnvGuard. Other utils::hatch
tests are not #[serial], so while the guard was live they saw uv
switched on: uv_sources_refuse_project_rewrites_under_the_uv_installer
failed 2-3 times in 100 loops of `utils::hatch` on a clean env. Even an
empty-value guard leaked, because EnvGuard's drop restores the previous
value before it releases the key.
The test now only checks that an unguarded host value stays hidden; the
uv branch itself is covered by the direct uv_source_names(.., true)
asserts. utils::hatch now passes 300/300 loops clean and 100/100 with
the host var set.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011f9aeR2MHrmJ9Fh9wmm3X11 parent 52db0e1 commit c76ed80
1 file changed
Lines changed: 7 additions & 16 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1739 | 1739 | | |
1740 | 1740 | | |
1741 | 1741 | | |
1742 | | - | |
1743 | | - | |
1744 | | - | |
| 1742 | + | |
| 1743 | + | |
| 1744 | + | |
| 1745 | + | |
| 1746 | + | |
| 1747 | + | |
1745 | 1748 | | |
1746 | 1749 | | |
1747 | | - | |
1748 | | - | |
| 1750 | + | |
1749 | 1751 | | |
1750 | 1752 | | |
1751 | 1753 | | |
| |||
1755 | 1757 | | |
1756 | 1758 | | |
1757 | 1759 | | |
1758 | | - | |
1759 | | - | |
1760 | | - | |
1761 | | - | |
1762 | | - | |
1763 | | - | |
1764 | | - | |
1765 | | - | |
1766 | | - | |
1767 | | - | |
1768 | | - | |
1769 | 1760 | | |
1770 | 1761 | | |
1771 | 1762 | | |
| |||
0 commit comments