Skip to content

Restore pocket-wall 75 +0.1/0 authored PMI - #2193

Merged
pzfreo merged 5 commits into
mainfrom
fix/pmi-pocket-wall-witness
Oct 6, 2026
Merged

pzfreo merged 5 commits into
mainfrom
fix/pmi-pocket-wall-witness

Conversation

@pzfreo

@pzfreo pzfreo commented Oct 5, 2026 •

Copy link
Copy Markdown
Owner

Addresses the missing authored 75 +0.1/0 dimension in #2192. Does not close the issue: the bore 60 dimension and two GD&T frames remain dropped.

The STEP source references datum C and a pocket wall with offset face centres. Their centre-to-centre vector is oblique, but the faces share a true X-normal witness at Y=26, Z=10. The extractor now proves that shared witness and selects a plan-view X dimension, then the placement path accepts that view. It also preserves the authored zero lower deviation, rendering +0.1/0.

Evidence: the source fixture originally failed the unprojected station proof; a regression test now checks the source record, label, witness, and queued plan candidates. On an explicit A1 1:2 build the 75 dimension placed in plan and its source no longer produced lint. scripts/pr-check --static passed; 151 focused tests passed (1 slow deselected). A temporary removal of the new witness fallback and of zero-tolerance preservation each made the regression fail.

Full lint on main already reports this missing source as pmi_not_extracted and authored_dim_source_unresolved. The public repro filters those codes, making it appear silent. Remaining #2192 drops are reported and are not claimed fixed here.

@pzfreo

pzfreo commented Oct 5, 2026

Copy link
Copy Markdown
Owner Author

CI failures in this run are reproducible on current main (db3b053) without this PR: tests/test_issue_1516_dimension_text.py::test_style_preserves_complete_plate_measurements[inline-aligned-first-None] loses a Y hole-location dimension on A3, and tests/test_hole_table.py::TestHoleTable::test_table_and_balloons_keep_lint_clean[first] selects scale 2 where the test expects 1. Both targeted tests fail locally on main. They do not exercise PMI extraction or this PR path. I am keeping this PR open, not treating the red matrix as evidence that #2192 is fixed.

@pzfreo
pzfreo enabled auto-merge (squash) October 6, 2026 06:43
pzfreo pushed a commit that referenced this pull request Oct 6, 2026
* Prove bore-axis location PMI and orient cylindrical frames

* Test geometric tolerance axis extraction

* Speed up PMI layout retries for #2192 (#2195)

* perf: probe GD&T corridor retries before CAD construction

* perf: skip impossible joint principal-view scales

* perf: avoid inadmissible no-ISO page proposal

* refactor: keep PMI probe within source-shape guards

* fix: retain AP242 plan witness stations for Y dimensions

* fix: reserve routes for unhinted AP242 linear PMI

* refactor: keep PMI reservation within review guards

* fix: preserve page choice and authored PMI intent in probe stack

---------

Co-authored-by: Paul Fremantle <pzfreo@github.com>

---------

Co-authored-by: Paul Fremantle <pzfreo@github.com>
@pzfreo
pzfreo disabled auto-merge October 6, 2026 07:41
@pzfreo
pzfreo enabled auto-merge (squash) October 6, 2026 08:08
@pzfreo
pzfreo merged commit 05a1bb5 into main Oct 6, 2026
19 checks passed
@pzfreo
pzfreo deleted the fix/pmi-pocket-wall-witness branch October 6, 2026 09:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant