02:43
<sideshowbarker>
jarhar: The code in the PR branch of https://github.com/WebKit/WebKit/pull/18785 has a working implementation of auto-expanding details. If you have time to build from that and try it, let me know if you find any problems. I put it together about a month ago but so far it’s otherwise been unreviewed (maybe due to waiting to get the content-visibility implementation further along).
12:06
<Noam Rosenthal>
Is there an existing term in the HTML spec for “topmost same-agent ancestor”? Something equivalent to “local root” in chromium? (Perhaps Domenic knows?) or do I need to get that manually?
12:06
<Domenic>
I'm not familiar with that Chromium term. Ancestor of what?
12:07
<Noam Rosenthal>
A document’s ancestor. The most top level ancestor that shares an event loop with this doc
12:08
<Domenic>
You'd have to crawl for that manually. Sounds kinda sketchy though to expose agent allocation in that way...
12:09
<Noam Rosenthal>
It’s exposed anyway since this is about timing/who’s blocking the main thread
12:09
<annevk>
It seems okay, except that it can be cross-origin today which is probably not a property we want to build on
12:12
<Noam Rosenthal>
The idea is that each of these docs keeps the tab of how long an animation frame takes. It’s observable today because you can see you’ve been blocked… very careful about not exposing more than that. It went through chrome security and will be happy for spec scrutiny once the draft is up
12:18
<annevk>
Hmm, but in theory that could block across top-level browsing contexts as well, no? If they end up in the same process. There are quite a few things that can be observable which are nevertheless not properties we want to enshrine forever.
12:28
<Noam Rosenthal>
Hmm, but in theory that could block across top-level browsing contexts as well, no? If they end up in the same process. There are quite a few things that can be observable which are nevertheless not properties we want to enshrine forever.
Absolutely. Let's talk about the specifics when the draft is up
13:05
<Noam Rosenthal>
You'd have to crawl for that manually. Sounds kinda sketchy though to expose agent allocation in that way...
Is that so also for getting the nearest ancestor thats top level in terms of same origin? (Nearest ancestor frame who has no parent or a cross origin parent)
13:11
<annevk>
Noam Rosenthal: yeah, we don't have any agent traversal of that kind
21:22
<jarhar>
do you also need to enable a flag after building?
21:34
<sideshowbarker>
Yeah, the “HTML <details> autoexpand” feature flag, in the Develop > Feature Flags… menu (what used to be called the Preferences Experimental Features menu)