perf(anatomy): V1-2.2 skip arc spread sampling where unused

Third security review LOW: the f32 spread of an arc is sampled only
when an outline is drawn and under its vertex cap; the variants'
extremes still join the bounds.
This commit is contained in:
bchanot committed 2026-11-01 13:31:39 +01:00
1 parent 86dce16b09
commit c73415abef
2 files changed
+15 -1

No files matched your search

+9
View File
@@ -4429,6 +4429,15 @@ arithmetic drift raised to `4 × 2⁻²³` of the radius. Bounds of every arc gr
edge counts them). LOW kept, pre-existing, outside the diff: relative commands accumulate in the browser's `f32` current
point; the worst aligned case (4 096 segments at the 256-span offset limit) drifts 6 % of the span, contrived.
Third security review, on `86dce16`: PASS, no MEDIUM. Both siblings refuse or match (an f32 browser model on ~300 000
random arcs: no counterexample at 0.5 CSS px; worst under-cover 0.13 units on generic arcs, about 0.1 px of disc radius).
LOWs: the spread's sampling ran for bounds alone and past the vertex cap, now skipped there (`86dce16` → the next commit;
the variants' extremes still join the bounds); open, declared: a semicircle exactly tangent to a frame edge counts the
spread of its variants (a few thousandths of a unit), so a detail `view_box` cropped exactly to an arc's extreme is
refused as out of frame; the spread is an estimate (one ulp, one axis at a time, rotation not moved); a zero-width needle of
a refused flat arc may run past the frame as a hairline nothing fills; arcs cost about 6 µs each natively, bounded by
the 2 MiB document only (≈ 1.8 s for a document of nothing but arcs, in the author's own preview).
Mutations (`.claude/audits/2026-11-01-v1-2-2-mutations-interior.py`): 29 killed, 2 equivalent: Z22 (the floor prune
removed; a cell that cannot reach the floor never holds an accepted disc) and Z25 (a path without an outline accepted;
unreachable). S1, S8 and F16 of the previous scripts, whose anchors moved, re-anchored and killed. Module 580 438 B gzip -9