Add a test to ensure APZ+WR hit-testing robustness
Categories
(Core :: Panning and Zooming, task, P3)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox75 | --- | fixed |
People
(Reporter: kats, Assigned: kats)
References
Details
Attachments
(1 file)
The APZ hit-testing codepath when using WR is a little finicky in that we have to maintain an invariant that's hard to check for in the code. Specifically we need to ensure that the scene data that WR does a hit-test against "matches" the state that APZ has when it receives the result of the hit-test. By "matches" I mean that the two pieces of data were generated from the same layers transaction.
In bug 1580178 nical is adding a new nonblocking hit-test mechanism, and I wrote a test locally to try and hammer those codepaths and make sure they didn't break the above invariant. I think it would be useful to land this test into the tree so it runs in automation and can catch regressions. This bug is to do that.
Current WIP is in https://treeherder.mozilla.org/#/jobs?repo=try&group_state=expanded&revision=6e654374b684c9a29f1a17aa6681f5d7b32f7dfc but is showing some unrelated errors that I need to investigate.
|
Assignee | |
Comment 1•5 years ago
|
||
https://treeherder.mozilla.org/#/jobs?repo=try&group_state=expanded&selectedJob=290432650&revision=92c6120fecfc8ed3ddb5c9d84c12640dbaf4ead1 is looking better, after I made it a subtest of test_group_hittest. I think it was running into that weird setup where it puts the test inside an iframe in the main runner and blowing up.
|
|
Assignee | |
Comment 2•5 years ago
|
||
This test is primarily to ensure that the APZ+WR hit-testing codepath
maintains the proper invariants (see
https://bugzilla.mozilla.org/show_bug.cgi?id=1617998#c0). However we can run
it in non-WR as well just to catch any random regressions.
|
||
Comment 4•5 years ago
|
||
| bugherder | ||
Description
•