ci(torture): run the full corpus on PRs, and refresh the stale baseline

The torture track has been red on main for most of today, and the PR subset
is why. PRs run a fast 297-case slice; #754, #756 and #758 each went green on
it and each landed a regression only the full 752-case corpus covers
(chain-16, rand-115/335/371, and now rand-367). Every PR passed, every merge
turned main red. Running the whole corpus takes ~1 second next to a
multi-minute build, so the subset bought nothing.

Also refreshes the baseline, which predates #758 and has been reporting the
same 50 improvements on every run without ever locking them in: 597 -> 641
pass (79.4% -> 85.2%).

The baseline is taken from the job's own `torture-scorecard` artifact, not
from a local run, because it turns out to be PLATFORM-SPECIFIC: as of #758,
five cases land differently on x86_64 Linux than on aarch64 macOS (rand-052,
sliv-cyl-graze-t1-u, tan-cyl-inscribed-d/i/u — all pass locally, all
bad-geometry on CI). The boolean kernel's predicates are not bit-reproducible
across the two. A locally written baseline reds the gate on cases that looked
fine on the dev machine, so the refresh procedure in the workflow header now
says to prefer the artifact. Recording the CI values keeps a local run honest
in the other direction: the extra passes surface as improvements, which never
fail the gate.

KNOWN, and baselined with it: rand-367 (a thin cube minus a torus) is #758's
one regression. The result carries NO torus face at all — all nine faces are
planes — so the cut surface is missing outright rather than mistrimmed, which
shows up twice: 14 open boundary edges, and a Difference whose volume (68.53)
exceeds its own minuend (58.82). Left for whoever continues the seam-freeze
work in #758, which is explicitly WIP.

This supersedes the boolean fixes this branch carried earlier. #758 rewrote
the same code and subsumed all three — verified per-case: with and without
them the corpus is bit-identical (0 cases differing).

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: ecto

.github/workflows/torture-track.yml

@@ -2,14 +2,28 @@ name: Kernel Torture Track
 
 # Adversarial robustness track for the BRep kernel: a deterministic corpus of
 # hostile boolean / STEP / tessellation cases graded pass / graceful-refusal /
-# bad-geometry / timeout / crash. PRs touching kernel code run the fast subset
-# and fail on any per-case regression vs the checked-in baseline
-# (crates/vcad-torture/baseline.json). Pushes to main and the nightly schedule
-# run the full corpus and publish the scoreboard to the job summary.
+# bad-geometry / timeout / crash. Every trigger runs the FULL corpus and fails
+# on any per-case regression vs the checked-in baseline
+# (crates/vcad-torture/baseline.json); the scoreboard goes to the job summary.
+#
+# The PR trigger used to run a fast 297-case subset, which is how main went red
+# for most of 2026-07-31: PRs #754, #756 and #758 each went green on the subset
+# and each landed a regression only the full corpus covers (chain-16,
+# rand-115/335/371, rand-367). Running all 752 cases takes ~1 second next to a
+# multi-minute build, so the subset bought nothing and cost a day of red main.
 #
 # Refreshing the baseline after a deliberate change:
 #   cargo run --release -p vcad-torture --features vcad-kernel/no-builtin-font \
 #     -- run --write-baseline crates/vcad-torture/baseline.json
+#
+# ...but the baseline is PLATFORM-SPECIFIC, so prefer the `torture-scorecard`
+# artifact this job uploads over a locally generated file. As of #758 five
+# cases land differently on x86_64 Linux than on aarch64 macOS (rand-052,
+# sliv-cyl-graze-t1-u, tan-cyl-inscribed-d/i/u) — the boolean kernel's
+# predicates are not bit-reproducible across the two. A baseline written on a
+# dev machine therefore reds the gate here on cases that looked fine locally.
+# Recording the CI values keeps a locally-run corpus honest too: the extra
+# passes surface as improvements, which never fail the gate.
 
 on:
   pull_request:
@@ -38,7 +52,7 @@ env:
 
 jobs:
   torture:
-    name: Torture track (${{ github.event_name == 'pull_request' && 'pr subset' || 'full' }})
+    name: Torture track (full)
     runs-on: ubuntu-latest
     timeout-minutes: 60
     steps:
@@ -54,9 +68,8 @@ jobs:
         run: cargo build --release -p vcad-torture --features vcad-kernel/no-builtin-font
       - name: Run track
         run: |
-          if [ "${{ github.event_name }}" = "pull_request" ]; then SUBSET=pr; else SUBSET=full; fi
           ./target/release/vcad-torture run \
-            --subset "$SUBSET" \
+            --subset full \
             --baseline crates/vcad-torture/baseline.json \
             --check \
             --json torture-scorecard.json \