Mark/pc2cad

Fork 0

test_benchmark_sample_runs_on_committed_corpus: cone tier shifted after fitter fix #64

New issue

Closed

opened 2026-09-15 02:55:44 +00:00 by Mark · 2 comments

Mark commented

2026-09-15 02:55:44 +00:00

Owner

Symptom

tests/test_methodology.py::test_benchmark_sample_runs_on_committed_corpus asserts:

assert int(res["verdict_tier"]) == 1

For the methodology_cone.ply benchmark. After commit 852ad5c (fitter fix: cone axis hint + angle floor + sphere plausibility), this returns 4 instead of 1.

Cause

The fitter now correctly emits a cone primitive for methodology_cone.ply (was: 0 cones, just planes + extrudes). The benchmark's tier ranking was tuned against the broken output, so its tier verdict shifted from 1 to 4.

What's expected

The cone fixture was always supposed to fit as a cone. The new behavior is more correct. The test's expected verdict_tier needs to be updated to reflect that a correctly fitted cone is still tier-1 territory, or the tier-ranking logic needs revisiting.

Reproduction

pytest tests/test_methodology.py::test_benchmark_sample_runs_on_committed_corpus

After commit 852ad5c: fails with assert 4 == 1.
Before commit 852ad5c: passes.

Scope

This is a test-expectation follow-up, not a fitter bug. The fitter change is correct and should stay. The benchmark expectation needs to be re-evaluated against the corrected fitter output.

## Symptom `tests/test_methodology.py::test_benchmark_sample_runs_on_committed_corpus` asserts: ``` assert int(res["verdict_tier"]) == 1 ``` For the `methodology_cone.ply` benchmark. After commit `852ad5c` (fitter fix: cone axis hint + angle floor + sphere plausibility), this returns 4 instead of 1. ## Cause The fitter now correctly emits a cone primitive for `methodology_cone.ply` (was: 0 cones, just planes + extrudes). The benchmark's tier ranking was tuned against the *broken* output, so its tier verdict shifted from 1 to 4. ## What's expected The cone fixture was always supposed to fit as a cone. The new behavior is more correct. The test's expected `verdict_tier` needs to be updated to reflect that a *correctly* fitted cone is still tier-1 territory, or the tier-ranking logic needs revisiting. ## Reproduction ``` pytest tests/test_methodology.py::test_benchmark_sample_runs_on_committed_corpus ``` After commit `852ad5c`: fails with `assert 4 == 1`. Before commit `852ad5c`: passes. ## Scope This is a test-expectation follow-up, not a fitter bug. The fitter change is correct and should stay. The benchmark expectation needs to be re-evaluated against the corrected fitter output.

Mark added the

bug

triage-approved

labels

2026-09-15 02:55:44 +00:00

Mark commented

2026-09-15 02:57:57 +00:00

Author

Owner

kilo-gsd launcher attempted to dispatch this ticket but preflight failed: ruff reports 21 lint errors on origin/main (commit 852ad5c). kilo-gsd.py exited at preflight without claiming the issue; no in-progress label was applied. A lint-cleanup ticket creation was attempted but failed (HTTP 422 — likely missing label definition). Until the lint debt is cleared, fixit work on pc2cad is gated by a separate cleanup ticket that does not yet exist.

kilo-gsd launcher attempted to dispatch this ticket but preflight failed: ruff reports 21 lint errors on origin/main (commit 852ad5c). kilo-gsd.py exited at preflight without claiming the issue; no in-progress label was applied. A lint-cleanup ticket creation was attempted but failed (HTTP 422 — likely missing label definition). Until the lint debt is cleared, fixit work on pc2cad is gated by a separate cleanup ticket that does not yet exist.

Mark added the

blocked:preflight

label

2026-09-15 17:23:57 +00:00

Mark commented

2026-09-15 17:26:56 +00:00

Author

Owner

kilo-gsd-runner (launcher pass on kanban t_de18e5eb): preflight blocked this ticket because ruff check . against Mark/pc2cad exits non-zero (2 E402 errors in tests/test_reconstruction.py plus 19 lines of context, 23 stdout lines total).

Lane-fix applied: filed Fix the lint on pc2cad at #67 with lint-cleanup + blocked:preflight labels. Released the stuck in-progress claim on this issue (kilo-gsd.py's auto-cleanup release_claim failed silently after a 422 on cleanup-ticket creation due to Forgejo's int64 label-id requirement on the issue-creation endpoint vs. name-acceptance on the label-attachment endpoint).

This issue now carries blocked:preflight and will be skipped by the picker until #67 merges and ruff check . exits 0.

Log: /home/b/.hermes/logs/kilo-monitor/pc2cad-64.log

kilo-gsd-runner (launcher pass on kanban t_de18e5eb): preflight blocked this ticket because `ruff check .` against Mark/pc2cad exits non-zero (2 E402 errors in tests/test_reconstruction.py plus 19 lines of context, 23 stdout lines total). Lane-fix applied: filed `Fix the lint on pc2cad` at #67 with `lint-cleanup` + `blocked:preflight` labels. Released the stuck `in-progress` claim on this issue (kilo-gsd.py's auto-cleanup release_claim failed silently after a 422 on cleanup-ticket creation due to Forgejo's int64 label-id requirement on the issue-creation endpoint vs. name-acceptance on the label-attachment endpoint). This issue now carries `blocked:preflight` and will be skipped by the picker until #67 merges and `ruff check .` exits 0. Log: /home/b/.hermes/logs/kilo-monitor/pc2cad-64.log

Mark referenced this issue

2026-09-16 01:56:04 +00:00

Fix the lint on pc2cad #67

Mark referenced this issue from a commit

2026-09-16 19:30:35 +00:00

fix(test): align methodology_cone verdict_tier expectation with post-852ad5c fitter

Mark referenced this issue from a pull request that will close it,

2026-09-16 19:32:17 +00:00

fix(test): align methodology_cone verdict_tier expectation with post-852ad5c fitter (Mark/pc2cad#64) #71

Mark closed this issue

2026-09-16 20:48:56 +00:00

Mark referenced this issue