Skip to content

Authored AP242 location to a pocket wall (75 +0.1/0) is neither drawn nor reported #2192

Closed

Description

@pzfreo

What

On main (db3b053, after #2191), the reproducer from #2189/#2190 builds cleanly in those respects: no out-of-bounds ink, and 35/85 placed. But one authored dimension still vanishes without any finding.

The STEP has a toleranced Location_LinearDistance from datum C (the x = −60 end face) to the pocket's side wall (the x = 15 face, normal −x, at z = 10): 75 +0.1/0. It does not appear on the drawing. Lint has no pmi_dropped, pmi_not_extracted or similar finding naming it. The other authored distance in the file, 120 ±0.05 from C to the far end face, is drawn correctly.

rc7 and rc8 behave the same way.

Reproduction

https://gist.github.com/pzfreo/7c6a0be95a567e2774924207133b2d40 (specify-plate-callout.step, repro.py). On main, repro.py prints only:

pmi_dropped | PMI '60' not placed (no room beside the front)
pmi_dropped | m_gdt0 not placed (no legible room in any side strip or sheet fallback)
pmi_dropped | m_gdt4 not placed (no legible room on its surface-normal left strip)

Nothing is reported for 75 +0.1/0.

Expected

The 75 +0.1/0 dimension is drawn, for example in the plan view, where the pocket wall is visible. If it can't be drawn, it is reported, so that an authored tolerance never disappears without a word.

Also seen

'60' (the bore's basic location from C) and two position frames are dropped for "no room" on an A1 sheet that is mostly empty.

Activity

  1. pzfreo commented on Oct 5, 2026

    @pzfreo
    OwnerAuthor

    PR #2193 addresses the missing authored 75 +0.1/0 dimension, with a common face-supported plan witness and explicit zero-tolerance preservation. Full lint on main actually reports this source as pmi_not_extracted and authored_dim_source_unresolved; the gist repro filters those codes, which made it appear silent. The bore 60 dimension and two GD&T frames are still reported drops and remain open under this issue.

  2. pzfreo commented on Oct 5, 2026

    @pzfreo
    OwnerAuthor

    Progress on #2192: #2193 restores the authored 75 +0.1/0; stacked #2194 proves the plane-to-bore-axis 60 and sends cylindrical position frames to their true face-on view. Neither PR closes this issue yet. On a normal A3 1:1 diagnostic, the bore frame lands, but 60 and datum B still drop in crowded plan/normal corridors; three authored 40 locations also drop. An A1 1:2 diagnostic placed the three original targets but lost other required dimensions, so it is not an acceptance result. The remaining work is requirement-aware placement and/or a coherent normal-size sheet split, related to #1866 and #1754. I have asked for the preferred sheet outcome before extending the document planner.

  3. pzfreo commented on Oct 6, 2026

    @pzfreo
    OwnerAuthor

    Checked on PR #2197 (88e2111): 75 +0.1/0 is now drawn, from C to the pocket wall in the plan, along with 120 ±0.05. That fixes this issue once #2197 is merged. What remains on the same file is filed as https://github.com/pzfreo/draftwright/issues/2198 (duplicate hole locations) and https://github.com/pzfreo/draftwright/issues/2199 (basic dimensions not boxed). Three basic "40" locations and m_gdt4 are still dropped for "no room", and are reported.

  4. pzfreo commented on Oct 6, 2026

    @pzfreo
    OwnerAuthor

    RC9 follow-up: PR #2200 preserves the eleven placed AP242 theoretical locations as boxed basic dimensions; PR #2201 removes planned duplicates for the 35, 85 and bore 60 locations. On their combined Specify plate diagnostic (default automatic layout), the three authored 40 locations and datum B still dropped. The placement trace showed datum B targeted a non-existent side/left strip, not a full page; PR #2202 gives that Y-normal datum a legal plan/below candidate through the shared solver. A combined diagnostic with #2202 places datum B, leaving exactly the three reported 40 drops. Those are still open work; no layout completeness claim yet. Section A–A also remains separately skipped because its cutting line conflicts with required leaders.

  5. pzfreo commented on Oct 6, 2026

    @pzfreo
    OwnerAuthor

    Further #2192 progress: PR #2203 retires a provisional section cutting-plane reservation before imported PMI dimension solve (after it protected early hole callouts). On the combined Specify plate diagnostic this recovers two of the three authored 40 locations; datum B is placed by #2202. The third 40 remains reported as pmi_dropped. Exact-ink diagnosis on an A1 1:1 diagnostic shows its left candidate conflicts with generated hole-pattern pitch dim_pitch_plan0_0; its right fallback conflicts with datum B m_gdt4. This is a priority/order issue, not simply page area. I tested shortest-to-longest PMI ordering and it did not recover the dimension, so that experimental change was discarded. Next work is to let required source PMI solve before generated pitch, then try to place the pitch in remaining room or report its drop. #2192 stays open.

  6. pzfreo commented on Oct 6, 2026

    @pzfreo
    OwnerAuthor

    PR #2204 (stacked after #2203) now defers generated pattern pitch until imported source-owned linear PMI has solved, then tries the normal pitch placer in the remaining room. The new Specify source-to-drawing regression verifies all three distinct basic 40 locations and both 60/96 grid-pitch dimensions on the diagnostic. A combined default automatic render with #2200/#2201/#2202/#2203/#2204 produced all three 40s, datum B, and both pitches with no PMI/pitch drops or ink-overlap lint. The optional section still reports its own conflict; pre-existing unsupported internal-thread PMI and hole-coverage warnings remain. This is tested code, not merged yet; #2192 stays open until the stack is green and merged.

  7. pzfreo commented on Oct 6, 2026

    @pzfreo
    OwnerAuthor

    Progress on the rc9/rc10 PMI completeness follow-ups: #2202 (alternate datum view) merged as 4054a656. #2204 (source AP242 dimensions before generated pattern pitch) is updated on top of that merge, with focused Specify and CTC01 checks passing locally; its new CI run is in progress. I am leaving this tracker open until #2204 is green and the combined default output is verified, rather than treating the earlier fixed-sheet diagnostic alone as closure.

  8. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    #2204 merged to main on 2026-10-07. It orders authored AP242 PMI before generated pattern pitch and includes a real Specify-plate regression fixture, addressing the unreported 75 +0.1/0 pocket-wall measurement. I am keeping #2192 open until the other noted PMI drops (the basic 60 and two position frames) are verified on current main or explicitly split into their own tracked outcome; a merged precedence fix alone is not proof that every annotation survives.

  9. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Verified on current main 7655ffc with the vendored Specify AP242 plate, pmi="annotate", A1 1:1 permissive: source dimension:0:1:4:18 renders as pmi_x_16 (75 +0.1/0); source dimension:0:1:4:3 renders as pmi_x_2 (60); all three source control frames geometric_tolerance:0:1:4:19/:23/:27 have m_gdt0/1/2 ink. There is no pmi_dropped or gdt_dropped finding. The original silent-loss complaint and the additional reported PMI drops are therefore resolved on main. This drawing still reports step_dim_withheld and section_dropped; those are separate, explicit outcomes, not grounds to keep this PMI-specific issue open. My previous comment associated the 75 fix specifically with #2204; the verification proves the current-main result, not which individual earlier PR supplied every part of it.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions