Summary: | REGRESSION(r264280) [GTK][WPE] fast/gradients/conic-gradient-alpha.html is failing | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Lauro Moura <lmoura> | ||||
Component: | WebKitGTK | Assignee: | Nobody <webkit-unassigned> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | bugs-noreply, cgarcia, clopez, darin | ||||
Priority: | P2 | ||||||
Version: | WebKit Nightly Build | ||||||
Hardware: | Unspecified | ||||||
OS: | Unspecified | ||||||
See Also: | https://bugs.webkit.org/show_bug.cgi?id=214221 | ||||||
Attachments: |
|
Description
Lauro Moura
2020-07-13 09:56:07 PDT
Oops, looks like I broke this. Happy to help fix it. Any indication what’s going wrong yet? It seems the generated gradient is now black&white: https://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20(Tests)/r264280%20(14625)/fast/gradients/conic-gradient-alpha-diffs.html Maybe a bug in the blend function? There are two problems: 1. Now the blend is done using premultiplied alpha, since the first color stop is #00FF0000, the premultiplied is just 0, so we always end up with black color. So, just changing to use blendWithoutPremultiply() would fix the bug but... 2. There's a bug in blendWithoutPremultiply(), from is used in both fromSRGB and toSRGB. Fixing 1 and 2 I get the right output, but the test doesn't pass yet, there's a one pixel difference for some reason. Created attachment 404338 [details]
Patch
Comment on attachment 404338 [details]
Patch
Nice! Thanks, this seems like a great fix.
Committed r264449: <https://trac.webkit.org/changeset/264449> |