Currently if a page has a polygonal clip-path, we return false from SVGIntegrationUtils::UsingSimpleClipPathForFrame and bail out of creating a complex clip region for WR. Instead, we fall back to creating a mask image.

Then, when we try to do a compositor hit-test on that element, it uses the bounding rect of the mask and can therefore result in incorrect hit results.

With fission this means that the target layersId for a hit-test can be wrong, and the input events can end up getting sent to the wrong process. This can be seen using the test page here - load it in a fission-enabled browser, and mouse over the iframe at the top. Any time you're over the iframe, but outside of the green triangle, it should say the mouse cursor is over the iframe. But with fission enabled it says "parent document" on the entire bounding rect of the green triangle instead, which is incorrect.

Based on a discussion with :jrmuizel today - we should support sending the clip-path information to WR and use it to do precise hit-testing. This will involve:

1. changing this line to return true
2. Adding a case StyleBasicShape::Tag::Polygon: to this switch statement that encodes the StyleGenericPolygon contents into a new clip type and sending it to WR
3. Supporting that new clip type throughout WR wherever needed (intentionally vague because I don't know what all this entails)
4. Specifically updating hit_test.rs to do a precise hit-test against the polygon clip

As a result of step 1, we should never hit this line when WR is enabled. And APZ will get correct results when hit-testing against clip-paths, which will make the attached testcase pass.

Comment on attachment 9185822 [details] [diff] [review]
clippath+fisison+wr testcase

Replacing raw patch with an equivalent phabricator revision, rebased on top of bug 1675378

Tracking this for M7 (rather than M6c) as polygonal clip-paths seem like an edge case that nightly users are unlikely to run into often.

I'll work on this.

Depends on D105396

Depends on D105397

Brad, there have been patches on this from 18 days ago, but no reviewer is assigned yet. What are the next steps here?
Fission M7 has to be fixed in Fx88 (soft freeze on March 18).

(In reply to Neha Kochar [:neha] from comment #8)

Brad, there have been patches on this from 18 days ago, but no reviewer is assigned yet. What are the next steps here?
Fission M7 has to be fixed in Fx88 (soft freeze on March 18).

These are WIP patches that got de-prioritized as I focused on other work. I'll come back to the them in the coming days. They aren't ready for review yet.

Jeff Muizelaar messaged on slack about this bug and said that polygonal clip-paths work, it’s just hit testing them that’s broken and that the graphics team doesn’t know of any content that actually breaks because of this.
Based on this info, moving this out to M7a (Fx89). Brad, please prioritize this for Fission.

This leaves the actual clipping in draw_clip_batch_list still as a TODO.

Depends on D105548

This performs a winding number calculation on the polygon and matches the
result against the fill rule of the supplied polygon.

Depends on D105548

This test was originally written by Kats, but I can't figure out how to
preserve authorship.

Depends on D109946

https://treeherder.mozilla.org/#/jobs?repo=try&revision=d6b27fb4411a733de211bd6cc36986b2e89a2eb8

Some test failures I'll need to resolve: https://treeherder.mozilla.org/jobs?repo=try&revision=8ce0b27e667664b3f7a56adfb2351ab7de182b5e&selectedTaskRun=DfcGgc5aS0mdavobpFbCSQ.0

Updated try run: https://treeherder.mozilla.org/#/jobs?repo=try&revision=20053d1fd94791581a1e63c96cf56cf35ebb457f

This test fails in automation, but those failures are difficult to reproduce
in local builds. When attempting to reproduce locally, the whole test sometimes
times out. Isolating specific subtests seems to affect the test outcomes as
well. Splitting the test into pieces seems to be the best way to make failures
in this test more actionable.

Depends on D109947

Depends on D111167

Backed out for causing build bustages.

Backout link : https://hg.mozilla.org/integration/autoland/rev/4d65631b94f0695ab1f25f893803bea89accb3bc

Push with failures: https://treeherder.mozilla.org/jobs?repo=autoland&revision=4cf2b6a199509fb559eadf675ac3520f58f2565b&searchStr=linux%2Cx64%2Cwebrender%2Copt%2Cwebrender%2Cstandalone%2Cwebrender-linux64-release%2Cwrench

Failure log: https://treeherder.mozilla.org/logviewer?job_id=336252149&repo=autoland&lineNumber=13617

Backed out 7 changesets (Bug 1675375) for causing webrender lint tidy failures.
Backout link: https://hg.mozilla.org/integration/autoland/rev/d1db6e82c992be27bf96839225c5c6ca2e091d7f
Push with failures, failure log.

Backed out 7 changesets (Bug 1675375) for causing mochitest plain failures in test_group_hittest-2.html
Backout link: https://hg.mozilla.org/integration/autoland/rev/3734414bbb3d3c9e53ccc4515f65cced1cab17a0
Push with failures, failure log.

Hey, just a heads up that in bug 1704070 I added a new subtest to test_group_hittest.html, so the Part 6 patch will need to be rebased across that. Apologies for any inconvenience.

Backed out 7 changesets (bug 1675375) for Mochitest failures in gfx/layers/apz/test/mochitest/test_group_hittest-2.html. CLOSED TREE

Log:
https://treeherder.mozilla.org/jobs?repo=autoland&group_state=expanded&revision=1a2cb51e0573b839f7a2edcd09f9cd54a042e43b

Push with failures:
https://treeherder.mozilla.org/jobs?repo=autoland&group_state=expanded&revision=1a2cb51e0573b839f7a2edcd09f9cd54a042e43b

Backout:
https://hg.mozilla.org/integration/autoland/rev/b79bcdd75afa6f8739c162e2873c6f8cd0708092

https://hg.mozilla.org/mozilla-central/rev/dc7b605c2911
https://hg.mozilla.org/mozilla-central/rev/d4b0a140332c
https://hg.mozilla.org/mozilla-central/rev/06c404b94962
https://hg.mozilla.org/mozilla-central/rev/fa02c774226a
https://hg.mozilla.org/mozilla-central/rev/fb5f5a643fa6
https://hg.mozilla.org/mozilla-central/rev/82cf330340dc
https://hg.mozilla.org/mozilla-central/rev/3da14f29d4d7