test_profile_operations_in_result_and_report: pre-existing flake in pipeline test #66

New issue

Closed

opened 2026-09-15 02:56:22 +00:00 by Mark · 1 comment

Mark commented

2026-09-15 02:56:22 +00:00

Owner

Symptom

tests/test_pipeline.py::test_profile_operations_in_result_and_report flakes in full-suite runs. Pre-existing; not caused by recent fitter fixes.

Repro

Passes alone:

pytest tests/test_pipeline.py::test_profile_operations_in_result_and_report

Flakes in full suite at random.

Status

Not introduced by commit 852ad5c (verified by running the test against pre-fix f9627bb: also flakes).

This test is one of several in tests/test_pipeline.py that exhibit run-order-dependent behavior (same family as the test_sample_plane_cyl_detects_cylinder flake we've seen for months). The root cause is the same: open3d's internal RANSAC RNG isn't fully exposed to Python seeding, so accumulated state across prior tests affects the fixture under test.

Workaround

Run the test alone when investigating, or accept the flake in CI as a known issue.

Filed at user's direction

User asked for triage-approved Forgejo issues for everything that has moved beyond a single fix-and-verify cycle. This is one of three such issues from the most recent session.

## Symptom `tests/test_pipeline.py::test_profile_operations_in_result_and_report` flakes in full-suite runs. Pre-existing; not caused by recent fitter fixes. ## Repro Passes alone: ``` pytest tests/test_pipeline.py::test_profile_operations_in_result_and_report ``` Flakes in full suite at random. ## Status Not introduced by commit `852ad5c` (verified by running the test against pre-fix `f9627bb`: also flakes). This test is one of several in `tests/test_pipeline.py` that exhibit run-order-dependent behavior (same family as the `test_sample_plane_cyl_detects_cylinder` flake we've seen for months). The root cause is the same: open3d's internal RANSAC RNG isn't fully exposed to Python seeding, so accumulated state across prior tests affects the fixture under test. ## Workaround Run the test alone when investigating, or accept the flake in CI as a known issue. ## Filed at user's direction User asked for triage-approved Forgejo issues for everything that has moved beyond a single fix-and-verify cycle. This is one of three such issues from the most recent session.

Mark added the

bug

triage-approved

labels

2026-09-15 02:56:22 +00:00

Mark commented

2026-09-15 03:11:10 +00:00

Author

Owner

kilo-gsd preflight-failed on this issue (Mark/pc2cad#66): ruff check . returned 2 E402 errors in tests/test_reconstruction.py:21-22 (module-level imports not at top of file). The runner's preflight is hard-gated on clean lint per the 2026-07-08 protocol (see kilo-gsd-lint-gate-whole-repo-2026-06-23 skill). The runner also failed to file its own lint-cleanup ticket (POST got HTTP 422) and to release its in-progress claim (DELETE 204 but the issue-side endpoint requires the label ID, not the name — known Forgejo quirk).

Lane cleanup performed by launcher: deleted the stuck in-progress label via DELETE /repos/Mark/pc2cad/issues/66/labels/1244. Issue is now clean for re-dispatch.

Resolution path:

Fix the 2-line E402 in tests/test_reconstruction.py (move the imports above sys.path.insert), commit on a branch, PR.
After merge, re-dispatch #66 — preflight will pass.

The original issue body itself prescribes the workaround for the test flake (accept the flake in CI as a known issue), but the runner can't even start to apply that workaround until lint is clean.

kilo-gsd preflight-failed on this issue (Mark/pc2cad#66): `ruff check .` returned 2 E402 errors in `tests/test_reconstruction.py:21-22` (module-level imports not at top of file). The runner's preflight is hard-gated on clean lint per the 2026-07-08 protocol (see kilo-gsd-lint-gate-whole-repo-2026-06-23 skill). The runner also failed to file its own `lint-cleanup` ticket (POST got HTTP 422) and to release its `in-progress` claim (DELETE 204 but the issue-side endpoint requires the label ID, not the name — known Forgejo quirk). Lane cleanup performed by launcher: deleted the stuck `in-progress` label via `DELETE /repos/Mark/pc2cad/issues/66/labels/1244`. Issue is now clean for re-dispatch. Resolution path: 1. Fix the 2-line E402 in `tests/test_reconstruction.py` (move the imports above `sys.path.insert`), commit on a branch, PR. 2. After merge, re-dispatch #66 — preflight will pass. The original issue body itself prescribes the workaround for the test flake (`accept the flake in CI as a known issue`), but the runner can't even start to apply that workaround until lint is clean.

Mark added the

blocked:preflight

in-progress

labels

2026-09-15 17:29:46 +00:00

Mark removed the

labels

2026-09-16 01:41:30 +00:00