SemianiakaVY/openpencil
1
0
Fork 0
Code Issues Pull requests Actions 1 Packages Projects Releases Wiki Activity

test(canvas): budget isolated seam pixels in the pan-cache diff

Browse source 
Windows' scalar rounding lands one glyph/tile edge a hair differently
(exactly 1 px on the CI runner), which the zero-diff rule read as
misregistration. Real misregistration swaps whole fills — hundreds of
contiguous pixels — so a 4-pixel isolated budget keeps the guarantee
while absorbing the platform rounding the comment already describes.
This commit is contained in:
Kayshen-X 2026-08-09 11:54:06 +08:00
parent ba2f55d4b5
commit b01551b7fe
1 changed files with 8 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
crates/op-host-native/src/widget_host/pan_cache_scroll_tests.rs
View file

@ -94,6 +94,13 @@ fn reference_frame(zoom: f32, dpi: f32, pan_x: f32, pan_y: f32) -> Vec<u8> {
/// misregistration produces (a shifted band swaps whole fills).
fn diff_report(actual: &[u8], expected: &[u8], dpi: f32) -> Option<String> {
const TOLERANCE: i32 = 8;
// A handful of isolated pixels is the same seam-rounding class as the
// per-channel tolerance above, just past its threshold — Windows'
// scalar math rounds one glyph/tile edge differently (observed: exactly
// 1 px on the CI runner). Misregistration cannot hide here: a shifted
// band swaps whole fills, which is hundreds of pixels in contiguous
// columns, far past this budget.
const ISOLATED_PIXEL_BUDGET: usize = 4;
let (pw, ph) = ((W as f32 * dpi) as i32, (H as f32 * dpi) as i32);
let stride = (pw * 4) as usize;
let (mut count, mut x0, mut y0, mut x1, mut y1) = (0usize, pw, ph, -1, -1);
@ -115,7 +122,7 @@ fn diff_report(actual: &[u8], expected: &[u8], dpi: f32) -> Option<String> {
}
}
}
if count == 0 {
if count <= ISOLATED_PIXEL_BUDGET {
return None;
}
let profile: Vec<String> = (0..pw as usize)