Replace shouldYield with hasPendingUserInput |
||||||
Issue descriptionAs per TPAC 2018, we should replace performance.shouldYield with performance.hasPendingUserInput, a function that allows checking for the presence of pending input events (in Chrome, those that are queued by the compositor) while the main thread is blocked. The event types that should be checked are to be provided as a list of event type strings supplied to the function.
,
Nov 30
Oh, something that came up at TPAC that we haven't discussed. Folks from Apple expressed that they'll want some wiggle room here, and were pushing for a higher level API. Folks from MS expressed that they want something precisely specified. I think resolving this is unlikely to change Chrome's implementation much, but it could change naming. I opened up an issue on the explainer here: https://github.com/tdresser/should-yield/issues/1.
,
Nov 30
,
Dec 15
Regarding receiving input information from cross-origin iframes- would it make sense to gate the experimental implementation on OOPIF support/strict site isolation? This would allow the MTEQ to satisfy the P&S requirements discussed in https://github.com/tdresser/should-yield/issues/2 without implementing off-main-thread hit testing in the renderer.
,
Dec 17
Ouch, good point. The work here (https://chromium-review.googlesource.com/c/chromium/src/+/1318054) and followup patches will enable a fairly easy solution, right? I'm not sure whether it would be worth shipping an origin trial without support on mobile. +Jason to see whether he thinks that would potentially be reasonable.
,
Dec 17
Just chatted with Jason - it sounds like there's precedent for shipping an origin trial desktop only. I think it makes sense to push for this.
,
Dec 17
Discussed offline with Tim. In short, it seems reasonable to run an origin trial that is not available on mobile, if you think you can get useful feedback from desktop only. There is precedent for such trials, and the OT framework allows configuring a trial to be available only on specific platforms.
,
Dec 17
Great! > The work here (https://chromium-review.googlesource.com/c/chromium/src/+/1318054) and followup patches will enable a fairly easy solution, right? It depends on whether or not it becomes possible to perform all steps up to (and including) frame attribution on the compositor thread. +Dave, does this sound reasonable to you? Our goal with the current implementation is to be able to tell if the compositor has received a WebInputEvent in a frame without yielding the event loop from client JS.
,
Jan 2
Hey all, tackling and demonstrating wins on mobile is critical for the web platform scheduling effort. Let's discuss what we need to do to make this work, rather than punting on mobile.
,
Jan 3
Shubhie, why do you think having support for mobile in the origin trial is required? We plan to ship on all platforms, this just lets us get the OT out the door sooner. Are there specific things you're hoping to learn about this API via OT that are mobile specific?
,
Jan 4
|
||||||
►
Sign in to add a comment |
||||||
Comment 1 by tdres...@chromium.org
, Nov 30