Debian Patches
Status for picosvg/0.22.3+git20260303-3
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| 0001-Add-minimal-pyproject.toml.patch | Add minimal pyproject.toml | Boyuan Yang <byang@debian.org> | no | 2026-03-31 | ||
| 0002-Fix-arch-dependent-test-failures-from-skia-pathops.patch | Fix architecture-dependent test failures caused by skia-pathops skia-pathops (libskia) may produce geometrically identical path output but with subpaths in a different order depending on the CPU architecture. Additionally, floating-point intermediate values near rounding boundaries can produce slightly different results across architectures. Fix 1: Normalize subpath order before string comparison in _test() by sorting the subpaths within each <path d="..."> attribute. This makes the test architecture-independent for the subpath ordering issue, affecting test_topicosvg[global-fill-none...] and similar cases. Fix 2: Move clip-from-brazil-flag.svg from test_topicosvg to test_topicosvg_tricky with round_ndigits=1. On arm64 and ppc64el, the skia-pathops output differs from the x86-generated expected file by up to ~0.01 in coordinate values (e.g. 249.649 vs 249.65, 149.083 vs 149.086). Rounding to 1 decimal place makes both architectures produce the same values (e.g. both → 249.6, both → 149.1), so the comparison succeeds without changing the geometric meaning of the test. Fix 3: Lower round_ndigits from 2 to 1 for the stroke-circle-dasharray tricky test, so values near the rounding boundary (e.g. 93.49 vs 93.5) both round to the same value. The 0.01-unit precision difference is geometrically negligible in a 128-unit viewBox. |
Boyuan Yang <byang@debian.org> | no | 2026-04-06 |
All known versions for source package 'picosvg'
- 0.22.3+git20260303-3 (sid)
