Debian Patches
Status for python-yamlfix/1.19.1-2
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| click-8.2.patch | Fix the test suite with click 8.2 =================================================================== |
Peter Pentchev <roam@ringlet.net> | no | 2025-02-21 | ||
| pytest-9-compat.patch | Pass a list to @pytest.mark.parametrize pytest 9 raises PytestRemovedIn10Warning (an error under the project's filterwarnings = ["error"]) when parametrize is given a non-Collection iterable such as itertools.product(). Materialise the product into a list. |
Dale Richards <dale@dalerichards.net> | no | 2026-09-08 | ||
| skip-tests-broken-by-ruyaml-0.92.patch | Skip tests that assume ruyaml < 0.92 line-folding behaviour ruyaml 0.92 changed how long scalar lines are folded and how anchors are emitted alongside duplicate merge keys, so five tests that pin the exact wrapped output now fail. yamlfix itself only passes the data through ruyaml, and upstream has not yet adapted to the new behaviour. Skip the affected tests until upstream resolves the incompatibility. . Closes: #1115776, #1136143 |
Dale Richards <dale@dalerichards.net> | yes | 2026-09-08 | ||
| glob-cache-test-count-traversals.patch | Assert the glob cache's invariant instead of timing the run test_glob_cache_is_fast created 2000 files, ran the cli over them and required the whole run to finish within a hardcoded 2 seconds. Almost none of that time is spent globbing: on amd64 the directory scanning accounts for about 0.01s of a 0.33s run, the rest being the unrelated work of parsing and re-emitting 1000 yaml files through ruyaml. The threshold therefore measured the speed of the machine rather than the cache, and the test failed on riscv64, where the same run takes 7.5s. . What the cache actually guarantees is that the number of directory traversals depends on the number of distinct glob patterns, not on the number of files found: without it each of the 2000 files re-scans the directory once per exclude pattern. So count the traversals and assert that no pattern is ever read from disk twice. That is deterministic, holds on every architecture, and is a stricter check than the timeout -- with the cache defeated it reports "3004 traversals were made for 6 distinct patterns" rather than relying on the run happening to overshoot a stopwatch. . The companion test_glob_cache_excludes_slow is left alone: it compares a cached run against an uncached one, so it is a ratio rather than an absolute time and is not architecture-sensitive in the same way. |
Dale Richards <dale@dalerichards.net> | no | 2026-09-16 |