Repository navigation
difftest: surface.surface_p9 analytic diagonal regressions - #1041
Conversation
Co-Authored-By: GPT-6.1 Sol Fast <noreply@openai.com>
Add named cylinder and cone controls that exercise the existing diagonal continuation and compare complete native split arrays. Preserve old fixtures, generator inputs, guards, numerical formulas and tolerances. Co-Authored-By: GPT-6.1 Sol Fast <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol Fast <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol Fast <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol Fast <noreply@openai.com>
|
@bherbruck — reviewed draft head |
bherbruck
left a comment
There was a problem hiding this comment.
Approved by the automated first-pass review: CI green, mergeable, no CI/build/dependency file changes, no red-flag patterns in added code (process/network/env/unsafe outside FFI/blobs/hidden unicode), 0 known-mismatch entries added. Logic is gated by CI and the difftest merge queue.
Scoped completion evidence
Units:
surface.surface_p9.Four named real-cylinder/canonical-cone regression fixtures exercise the existing
FindTessellationSplitsdiagonal continuation in direction U. No productiongeometry or representation was added. Each kind has identical enabled/disabled
controls: fractional domain
[0.125,0.2,0.375,0.8], controls[0.05,15,0].Expected edits (exact):
fixtures/surface.surface_p9/analytic-diagonals.jsoncrates/difftest/src/units/surface_surface_p9.rsPublished head and verification
Head
d8d3972f7939a733135ed630c6d1536a7fea929b, safely synchronized with authoritative mainb52ba92814f3d72d1008270df84b19cbd2e6c896.Latest upstream observed at publication:
b52ba92814f3d72d1008270df84b19cbd2e6c896.Independent final source review: no findings, fingerprint
4d240e3afa68b30a4250399e765f4122bf01da30b5f44dd06e7e29c950691c5d.Reviewed binary PR diff unchanged across both safe upstream merges; all 221 old
fixtures, minimum 600 generated cases, generator inputs/RNG/seeds, native guards,
formulas, tolerances and floors are unchanged. Total stored fixtures:225.
Local macOS Rust1.98/CPP checks on this exact head:
Shared reference libraries were reused read-only, with hashes/mtimes unchanged;
pinned upstream is c9c5979. This local from-source reference has no BUILDINFO/key
marker and is not canonical Linux CI proof.
analytic_diagonal_cylinder_u_enabled,analytic_diagonal_cone_u_enabled_u_disabledcontrolsThe focused CPP test also perturbs an interior split while preserving status;
the unchanged comparator detects exactly one mismatch. Existing generated
cylinder/cone suppression remains executable-byte-identical; only its stale
comment was corrected. No shim/FFI/plan/fixture-removal or guard changes.
Related work and status
#1031/#1042 curve/cache interfaces remain untouched. #1047 p6 Sphere, #1048 p4
Ellipse equations and #1049 p7 Ellipse classification are independent unit scopes.
Available follow-ups: other kinds/directions and broader generator coverage are not
reserved or certified by these four cases.
Ready for review by explicit user instruction; canonical CI is still pending., including
native sweep, perturbation, value-share and exact-base retention. Source success,
synthetic queue checks and actual merging are separate. Maintainers retain workflow
authorization, approval, queue admission and merging.
Metadata-only upstream status correction synchronized before canonical validation; all five local checks reran successfully on the resulting exact head.
User-authorized review status
Marked ready for review at the user’s request after local native validation and independent source review. Canonical CI was authorization-blocked at the latest audit; this transition does not claim green CI, approval, queue admission or merge readiness. Maintainer workflow authorization and all required checks remain mandatory. No source commit or workflow configuration changed.