[anchor-position] Support testing of polyfill #47156
Conversation
|
There are no reviewers for this pull request. Please reach out on the chat room to get help with this. Thank you! |
|
Seems odd to do any kind of WPT change for this. This should probably done at the level of your test runner (perhaps finding a way to run the polyfill really early (WebKit has bootstrap scripts that run before everything else in the Web Inspector for instance). |
Adding this ( |
|
Yeah, I agree that changing the tests for the nuances of a single implementation is not ideal. The key complication is that currently the polyfill only affects elements in the DOM when the polyfill is loaded, which we are tracking here- oddbird/css-anchor-positioning#91. The WPT scripts mix the arrange and assertion steps, so there isn't a way to cleanly insert the polyfill between those steps, and running earlier before all WPT scripts wouldn't work. I don't think it's worth rewriting the runner before the polyfill supports dynamic elements, as that will change/lessen the requirements there. At that point, there will be a difference between "the polyfill is loaded" and "the polyfill has been applied". The polyfill will still likely take several animation frames to get elements positioned when the polyfill has been applied, so the change in this PR will still be useful. |
I'm contributing to the polyfill for anchor position, which we test against the WPT. When testing the polyfill, we need to add a bit of delay to ensure the polyfill is applied before testing.
In #38442, a
checkLayoutForAnchorPoswrapper forcheckLayoutwas added to support a way to add a delay only for the polyfill while not impacting other tests. We don't have a similar method fortest, so we have some false negatives when testing the polyfill.testForAnchorPosis intended to replace all instances oftestandpromise_testin the Anchor Position tests. It likely would work forasync_testas well, but I'm not seeing any instances of that in the Anchor Position tests. LikecheckLayoutForAnchorPos, this only adds a delay if a globalCHECK_LAYOUT_DELAYis true AND we are testing with--injected-script. This should have no impact on non-polyfill tests.I've applied this to
css/css-anchor-position/pseudo-element-anchor.html, and if this strategy is acceptable, I will transition the other tests as well.