Repository navigation
test: ask the platform ignore hook without the ci shard targets list - #32209
Conversation
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: this is a test-only fixture change. It turns off the shard plugin's [DESIGN-REVIEWED] 188cc77 Description read[DESCRIPTION-READ] 5d97b0845922f56147144d08f19c23e838fa930e217e6d7bdc62bf050c15c925 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
|
[operator: bolichen97] Heads-up: #32183 and #32209 each stay red on the other's main bug, so neither can go green alone.
Suggest folding one change into the other PR so a single PR carries both fixes and can pass CI. Happy to help if useful. |
Under --file-shard-targets the ci_file_shards plugin skips every unlisted path by design, so the hook chain answered for the Windows list and the nested-path assertion failed on every leaf-test run. The fixture lifts that list for these questions only. Closes #32199.
edb0249 to
188cc77
Compare
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Tier 1 auto-approve: test (1 files). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, security checklist all-NO, no reviewer-lane findings, readiness signal fresh for this head SHA, declared category matches the diff. Category: test-only change: the platform ignore-hook test lifts the CI file-shard target list for its own questions.
bolichen97
left a comment
There was a problem hiding this comment.
Approving head 188cc77. The gate showed 96 checks passed. Sensitive Change Review failed on the GitHub API rate limit, and its rerun passed.
This is a test-only change in test/test_platform_collect_ignore.py. Under CI's --file-shard-targets, the ci-file-shards plugin skips every unlisted path, which would answer the platform-ignore question for these tests. The fixture now sets that plugin's keep to None through monkeypatch, only when the plugin is loaded, so it is restored after each test. No product code changes.
Problem / Motivation
Goal: A PR that changes only test files no longer fails Backend Tests on
test_platform_collect_ignore.pywhen its shard draws that file.test/test_platform_collect_ignore.py::test_the_windows_list_ignores_exactly_the_listed_test_filesfails on every leaf-test run that puts it in a shard's corpus gates (#32199). Those gates run with--file-shard-targets=(#31516), and theci_file_shardsplugin then skips every unlisted path, by design. The test asks the whole hook chain whethertest/fixtures/<listed name>is ignored, so the shard plugin's answer stands in for the Windows list's and the assertion fails.Closes #32199
Why it matters
#32183 (main's line-budget fix) is red on Backend Tests (3.12, 7) for this alone, and any other test-only PR whose shard draws this file is too.
Not a goal
--file-shard-targetsskips. ci: collect reduced-scope backend tests from a targets file #31516 skips unlisted paths on purpose, so collection stays fast.test/conftest.pyapplies it.What changed (motivation → approach → change)
The test's
windows_listfixture sets the registeredci-file-shardsplugin's targets list toNonefor the test, when the plugin is loaded. The test still asks the real hook chain, now without the CI targets list answering for it. Test-only.Backwards compatibility
Compatible: test-only; no runtime or CI code changes.
Removes nothing: every assertion stays; only the CI targets list stops answering for them.
Tests
-p scripts.ci_file_shards --file-shards 1 --file-shard 1 --file-shard-targets=<file listing this test>): 1 failed onorigin/main0795ebc479, 3 passed with this change, 5 runs in a row. With--file-shards 8 --file-shard 7 -n 4: 3 passed, twice.test/test_platform_collect_ignore.pyandtest/test_ci_file_shards.pypass (29 tests).Proof: not provable -- test-only change
Manual verification
N/A — unit coverage sufficient: the change is a fixture line in one test.
Related Issues
Follow-up to #31516. Unblocks #32183.
Sibling sweep
Swept: git grep -n "pytest_ignore_collect(" -- test/
Siblings: none --
test_platform_collect_ignore.pyis the only test that asks the live session's hook chain about paths outside its own filePattern harvest
Not generalizable: one test asked the live session's hook chain, which CI's file-shard plugin also answers when a targets list is set.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)