Bug 299189 – REGRESSION(300173@main): Pixel tolerance for WPT tests under /css/filter-effects/filter-function was increased to make them useless

• Status: RESOLVED
• Resolution: DUPLICATE
• Component: Layout and Rendering

[Said Abou-Hallawa]

The problem was the change in these tests:

imported/w3c/web-platform-tests/css/filter-effects/filter-function/filter-function-repeating-radial-gradient.html
imported/w3c/web-platform-tests/css/filter-effects/filter-function/filter-function-repeating-conic-gradient.html

After 300173@main, most of the pixels were changed in the actual image by just one in only one channel. For example white pixel (0xFF, 0XFF, 0XFF) became (0xFF, 0XFF, 0XFE). That is okay but the problem is at the edges of the conic and radial gradients, some pixels are changed by 255 in one channel. The number of these pixels is not large so the pixel difference was something like 0.29% . But the pixel tolerance had to be something like this:

<meta name="fuzzy" content="maxDifference=0-255; totalPixels=0-107939">

which made the test almost useless. Any regression will not be discoverable with large pixel tolerance.

[Radar WebKit Bug Importer]

<rdar://problem/160949031>

[Said Abou-Hallawa]

Pull request: https://github.com/WebKit/WebKit/pull/51016

[Said Abou-Hallawa]

Bug 299228 addresses this issue better than this bug. It does not change the WPT tests and it removes the pixel tolerance completely.

*** This bug has been marked as a duplicate of bug 299228 ***