interop
interop copied to clipboard
Remove all tests using SVG filters with backdrop-filter from Interop 2025
Test List
backdrop-filter-reference-filter-mutated.html backdrop-filter-reference-filter.html backdrop-filter-svg-background-image-blur.html backdrop-filter-svg-foreignObject.html backdrop-filter-svg.html
Rationale
Similar to https://github.com/web-platform-tests/interop/issues/938
cc @progers @jrmuizel
Sounds good to me.
@chrishtr @progers Friendly ping :) I think SVG is really its own focus area in general and the implementation burden is particularly high (requires LBSE in WebKit at least to fix this, which is a multi-year project).
Other than backdrop-filter-svg-foreignObject.html, these tests are testing url() references for backdrop-filter, which work in the regular filter property in all engines. Do these tests have the same implementation burden and spec issues as those raised in https://github.com/web-platform-tests/interop/issues/938? It seems like fixing these just involves bringing reference support to backdrop-filter.
Other than backdrop-filter-svg-foreignObject.html, these tests are testing url() references for
backdrop-filter, which work in the regularfilterproperty in all engines. Do these tests have the same implementation burden and spec issues as those raised in #938? It seems like fixing these just involves bringing reference support to backdrop-filter.
@mattwoodrow looked into this, and this is what he mentioned:
Our UI side rendering of filters (CoreAnimation) doesn't support SVG filter references (or any kind of SVG filters), and our WebProcess side doesn't support backdrop-filter (we have no code to capture a backdrop). We'd have to write an implementation of backdrop-filter from scratch, and it'd suffer from the same extremely slow rendering issues as we already have for filter + SVG reference
The reason they work for filter is because there is non-CoreAnimation based implementation of filter, but that is not the case for backdrop-filter.
We think our time would be better spent focusing on the other backdrop-filter tests, rather than writing a brand new implementation of backdrop-filter just for SVG filter references.
@nt1m - backdrop-filter-svg-background-image-blur.html passes: backdrop-filter-svg-background-image-blur.html.
Should it be excluded from the list?
https://wpt.fyi/results/css/filter-effects?label=master&label=experimental&aligned&view=interop&q=backdrop-filter-reference-filter-mutated.html%20or%20backdrop-filter-reference-filter.html%20or%20backdrop-filter-svg-background-image-blur.html%20or%20backdrop-filter-svg-foreignObject.html%20or%20backdrop-filter-svg.html is the link for all the mentioned tests.
@progers @chrishtr Gentle ping on this issue :)
@nt1m - backdrop-filter-svg-background-image-blur.html passes: backdrop-filter-svg-background-image-blur.html.
Should it be excluded from the list?
It passes because it's a mismatch test comparing against blank, so passes regardless of whether the filter is applied or not.
I think it still belongs in this set, but I guess there's no harm in leaving it in.
I'm so sorry about delaying this unnecessarily. I am sympathetic to wanting to spend our time on the most impactful interop areas and am okay with this test change proposal. One of the most important backdrop-root interop issues for our users, in terms of bugs filed, involves the backdrop-root definition and associated tests, and I am so happy to see the progress that's been made there (which had a similar CoreAnimation/CoreGraphics dependency).
Thanks!