Karim Jerbi eabc7e216b fix(underlay): clamp hostile arc sweep before sizing pieces ...

Bug:
arc_cubics turned a raw sweep angle from the DGN file straight into a
piece count - ceil(sweep ...) as usize fed with_capacity, and an extreme
sweep (huge, or +/-inf) produced a count that overflows usize
allocation, causing a capacity overflow panic while attaching the file.
Legal DGN arcs top out near 104 radians.

Reproduce:
1. Craft a DGN with an arc element whose sweep angle is enormous (1e18+,
   +/-inf).
2. ATTACH it.
3. Before: "capacity overflow" at raw_vec during model build.
4. After: the arc clamps to +/-4096 quarter-turns (sign preserved, NaN
   passes through, +/-inf handled) and renders as many cubics as fit.

Fix:
The clamp happens in the one funnel (arc_cubics) before ceil: sweep is
limited to +/-4096 x 90 degrees, far above the legal 104 rad but orders
below what overflows the count. The sign of the sweep is preserved so
arcs keep their direction, and NaN passes through to the existing
empty-arc handling. Tests cover three sweep magnitudes (including
+/-inf) building without overflow.

2026-09-30 19:15:49 +01:00

..

dgn7.rs

DGN underlays: global origin, stored extents, colour table display, 3D cells

2026-09-29 11:58:47 +03:00

dgn8.rs

fix(dgn): a short 2D text element no longer panics the reader

2026-09-29 23:11:33 +03:00

dwf.rs

fix(dwf): bound XML element depth before roxmltree and Canvas walk

2026-09-30 19:15:49 +01:00

mod.rs

DWF/DGN underlays: draw content as vector wires

2026-09-29 12:43:20 +03:00

model.rs

fix(underlay): clamp hostile arc sweep before sizing pieces

2026-09-30 19:15:49 +01:00

wires.rs

DWF/DGN underlays: draw content as vector wires

2026-09-29 12:43:20 +03:00