Repository navigation
Open
Labels
in progresssomeone is working on it (see comments)someone is working on it (see comments)needs-triageno priority yet; a maintainer sorts it with /triageno priority yet; a maintainer sorts it with /triage
Description
solidcore/AGENTS.md §4b says the existing UpstreamFixes flags are retired. This issue tracks retiring the remaining ones in small batches, one PR per batch. Each PR deletes the field, the code path the flag turned off, its preset line in crates/smlib/src/core.rs, its fixes.json entry and the tests' "off" halves.
Builds on tdamsma's reachability audit (#13843), which used an earlier preset and test set.
Survey (2026-11-09, origin/main b653486)
Full report and one CSV row per flag: https://gist.github.com/brett19/18a1b63a4f7e7a8419669769003b79bd
Method:
- Every library read of a flag was instrumented (1137 reads).
- Each flag was measured as the default kernel plus that one flag, over the goldens cases that reach it (3502 capped unit runs).
- The kernel rig was run with
fixes.jsonand with[]. For each of the 130 cases that fail without flags, a ddmin found the minimal set of flags it needs.
| group | flags | meaning |
|---|---|---|
| A | 86 | no golden case and no rig case changes (5 of them are read nowhere) |
| B | 170 | values move, nothing regresses; each moved row is blessed with a reason |
| C | 30 | some rows regress; each needs judgement |
| D | 1 | HELD_BACK (fillet_round_ends_on_outline) |
| E | 18 | already retired or in an open PR |
Things to know before retiring:
fillet_gn_section_generator: on its own, this flag makesblend.fillet_geom_p2/gen0014allocate without bound. Kernel: variable rounds are the union of their balls by default; retire fillet_gn_section_generator (moved from solidcore#2936) #13945, which retires it, adds a test for that case.ssi_tangent_curve_tracer: on its own, this takesoffset.offset_geometry_creationfrom 23 s to more than 600 s.blend.fillet_executive_p3under the preset: goes from 35 s to 946 s, through a combination of flags.- solidcore: Offset Face: mitred rounds refused when the moved top goes below their depth (the C++ builds them) #13841's trio (
fillet_refuse_rails_leaving_faces,fillet_cliff_overflow,fillet_rail_end_on_continued_surface): the rig only ever needs them together. Retire them as one bundle (B17), after Kernel: variable rounds are the union of their balls by default; retire fillet_gn_section_generator (moved from solidcore#2936) #13945. - The
*_statuses_pass_onfamily: retire each with its unit's typed-error work (solidcore: Errors: report solidcore-specific, descriptive errors instead of bare SMLib statuses and SMLib names #13790), not as a bulk flip. - Stale
fixes.json: the rig'sfixes.jsonis missing 9 flags that SolidCraft's preset turns on. The report lists them.
Batches
Order: A (by area), then B17/B18, then B by area, then C. A batch never touches a flag claimed by an open PR. Each PR runs the full goldens and the kernel rig.
- dead flags, retired by solidcore: retire five flags that no code reads (fillet corner/domain statuses, offset gap, ssi alignment, closed-curve trim arc) #14091 (
fillet_corner_statuses,fillet_edge_curve_domain_statuses,offset_tangent_gap_statuses,ssi_alignment_statuses,offset_closed_curve_trim_arc) -
fillet_trace_rejects_short_step(solidcore: retire fillet_trace_rejects_short_step (WR-503) #13982, merged) -
fillet_refuse_crossing_rails(solidcore: retire the crossing-rails refusal (WR-460); the band test (WR-542) refuses its cases with a fit #14047) - A4 curve (solidcore: retire the curve flags that change no result (A4) #14288), A7 intersect (solidcore: retire the intersect flags that change no result (A7) #14347), A8 mass (solidcore: retire the mass flags that change no result (A8) #14429), A11 primitive (solidcore: retire the primitive flags that change no result (A11) #14448), A14 sweep (solidcore: retire the sweep flags that change no result (A14) #14478)
- A6 geom
- Status 2026-11-14: 187 flags remain on main (A 31, B 124, C 28, D 1, E 3). 14 are in open retire PRs and 10 are held. Experiment: retire every UpstreamFixes flag SolidCraft runs on, at once #15956 (
status:decision) would retire nearly all of them at once, so new batches wait on that decision. - A1–A3, A9, A13, A15, A16 (after the open PRs on the same files)
- B17–B29 bundles and area batches, then the remaining B batches
- C batches
Changes a default: no
Metadata
Metadata
Assignees
Labels
in progresssomeone is working on it (see comments)someone is working on it (see comments)needs-triageno priority yet; a maintainer sorts it with /triageno priority yet; a maintainer sorts it with /triage
Activity
solidcore: retire the intersect flags that change no result (A7)
solidcore: retire the curve flags that change no result (A4) (tangenc…
solidcore: retire the intersect flags that change no result (A7) (tan…
solidcore: retire the mass flags that change no result (A8) (tangency…
solidcore: retire the primitive flags that change no result (A11) (ta…
solidcore: retire the sweep flags that change no result (A14)
solidcore: retire the sweep flags that change no result (A14)
solidcore: retire the sweep flags that change no result (A14) (tangen…
github-actions commentedon Nov 10, 2026
Claimed by @austin62jake through #14679. The claim lapses after 14 days without pushes.
austin62jake commentedon Nov 10, 2026
A16 (topology) is in draft #14679:
loop_orientation_floor_from_loop_box,trim_boundary_seam_wrap,imprint_extension_keeps_analytic_curve,shell_inner_shell_of_statusesandis_iso_curve_at_propagates_canonical_failure.Already done on main, from checking that each batch's flags still exist at
87400c80e: A4, A5, A6, A7, A8, A11, A12 and A14 have no flags left.Written by Claude Code (Anthropic) on behalf of @austin62jake.
plan: the five A16 flags' records note that the flags are removed (ta…
solidcore: retire five topology UpstreamFixes flags (tangency-labs#14246
solidcore: retire remove_slit_loops_where_solids_touch too (A3)
301 remaining items
plan: update the topology-2 addenda to the merged-main row counts
solidcore: a face with no edges boxes to its domain, also with a vert…
solidcore: the fold witness ignores degenerate points of the original…
solidcore: a cell past its surface's domain by rounding is decided on…
solidcore: keep upstream's result where a passed-on status refused a …
solidcore: expectations: rows the fallbacks keep return to main
solidcore: retire brep_edit_statuses_pass_on (tangency-labs#16956)
solidcore: a tight-box cell is decided on its intersection with the s…
solidcore: tight-box cell domain intersection only where evaluation c…
solidcore: a self-loop proto edge on a vertex is degenerate; retire p…
solidcore: retire four topology-2 UpstreamFixes flags (edge splits, t…
solidcore: goldens for the mass rows that now read the trimmed-face t…
solidcore: retire four offset and local-operation flags (tangency-lab…
solidcore: retire imprint_extends_edge_to_proto_vertex (tangency-labs…
solidcore: always split SSI curves at singular points (retire WR-77 f…