Test List
https://wpt.fyi/results/css/css-view-transitions/inline-with-offset-from-containing-block.html
Rationale
This test has always had an extremely permissive fuzzy annotation -- it was added with up to 1400 pixels allowed to be completely different and that was later increased to 1633 pixels to get it passing in Safari.
This fuzziness is due to text getting rasterized at a low resolution and then getting scaled. That in-and-of-itself might be worth avoiding, but it seems to not be what this test is trying to test, given this history of fuzzy annotation.
Firefox currently is flagged as failing the test because we have more fuzzy pixels than the currently-annotated tolerance, but I think fundamentally we behave the same as Safari does (and as Chrome did when the test was checked in).
But rather than increase the pixel-mismatch-count to get Firefox passing (which would make the test even more permissive of arbitrary breakage), let's reduce the amount of text in the testcase so that we can lower the number of fuzzy pixels that we need to allow. I propose we change "Some text" to "A B", and then reduce the fuzzy annotation to match observed needed levels.
Test List
https://wpt.fyi/results/css/css-view-transitions/inline-with-offset-from-containing-block.html
Rationale
This test has always had an extremely permissive fuzzy annotation -- it was added with up to
1400pixels allowed to be completely different and that was later increased to1633pixels to get it passing in Safari.This fuzziness is due to text getting rasterized at a low resolution and then getting scaled. That in-and-of-itself might be worth avoiding, but it seems to not be what this test is trying to test, given this history of fuzzy annotation.
Firefox currently is flagged as failing the test because we have more fuzzy pixels than the currently-annotated tolerance, but I think fundamentally we behave the same as Safari does (and as Chrome did when the test was checked in).
But rather than increase the pixel-mismatch-count to get Firefox passing (which would make the test even more permissive of arbitrary breakage), let's reduce the amount of text in the testcase so that we can lower the number of fuzzy pixels that we need to allow. I propose we change "Some text" to "A B", and then reduce the fuzzy annotation to match observed needed levels.