| 11:13 | <Noam Rosenthal> | To give people time to prepare for tomorrow's WHATNOT, I've posted the topics currently with Agenda+ here: https://github.com/whatwg/html/issues/12275#issuecomment-4081641430 It's ok to add more issues but people might not have time to review before the meeting. |
| 12:55 | <Shannon Booth> | I looked into it some more and ran into a difference between browser behaviour here as well by extending the test a bit. All browsers agree child removal does not trigger script execution, but they don't agree on whether a mutation of character data for a child should or not. Chromium and Webkit dont execute but Firefox does. Raised an issue for it on HTML tracker |
| 13:26 | <annevk> | That sounds like an odd difference. |
| 13:42 | <annevk> | Shannon Booth: I had a quick look at the issue you filed and I actually can't find why https://github.com/web-platform-tests/wpt/blob/master/dom/nodes/insertion-removing-steps/script-does-not-run-on-child-removal.window.js should not run. It's not "parser-inserted" and "already started" also does not get set (there's an early return for non-script types). So why do we early return when the children changed steps run? |
| 14:03 | <Noam Rosenthal> | Dominic Farolino: do you remember this stuff from 2 years ago? ^^^ |
| 14:15 | <annevk> | keithamus: do you know if there's anything standardized around the scroll position of popovers? I don't recall this being discussed and I can't immediately find anything. |
| 14:19 | <keithamus> | I do not recall anything around that, no. |
| 15:13 | <Shannon Booth> | Right, that's why I reported the issue, I couldn't find where it is specified. But when I looked deeper it turns out that browsers also differ in behaviour that child CharacterData nodes are being mutated. Chromium/Webkit seem to early return in that case, Gecko seems to not. |
| 15:30 | <annevk> | keithamus: if you have a lot of options in base select it's definitely a thing that comes up. And it does seem that Chrome has made some decisions there that made sense to them. (I haven't looked at the source though.) Not sure if we want this fully nailed down as it's user interaction, but it might well be something people want control over. |
| 15:35 | <Luke Warlow> | Can you explain more what you're seeing? Because I also don't recall any discussions here. But personally I lean more towards speccing behaviours as much as possible. I thought this was just using popover + anchor positioning for any positioning behaviours |
| 15:41 | <annevk> | Luke Warlow: well say you have multiple "pages" worth of options and you select an option somewhere in the second page. As long as the selected option remains visible Chrome will remember the scroll position and show it in the same position when you close and reopen. Otherwise it'll be top-aligned or the prior remembered position (not entirely sure about the exact logic here). |
| 15:42 | <annevk> | Not sure yet what we want as we probably have to deviate from macOS behavior here. |
| 16:25 | <Noam Rosenthal> | Shannon Booth, annevk : chromium code seems to have a comment pointing to an outdated spec text: // <spec>When a script element el that is not parser-inserted experiences one of So I think this was standard behavior at some point and some spec PR along the line regressed it |
| 16:34 | <annevk> | Thanks! That would almost certainly be Dominic Farolino's changes so his input would be useful. |
| 16:34 | <Noam Rosenthal> | I think it was actually much much earlier, around 2018. Looking for the exact PR. |
| 16:38 | <Noam Rosenthal> | Indeed, It's in https://github.com/whatwg/html/commit/ddd2d0dd8ddee1d56e87e6e4bc149bd6e5a93d11, line 62388. |
| 17:10 | <Noam Rosenthal> | https://github.com/whatwg/html/pull/12282 and https://github.com/whatwg/dom/pull/1460 should fix this (need some work) |
| 18:12 | <annevk> | Oh, are you also going to fix the issue we're having with Trusted Types? It needs to know who changed things. |
| 18:27 | <Noam Rosenthal> | I wasn't in this PR.... I think that involves a bit more as we need to thread that info from the parser |
| 18:29 | <annevk> | Yeah makes sense, just curious about plans. |
| 18:32 | <Noam Rosenthal> | I actually had zero plans about this, but was waiting for something else so decided to fix that script mutation regression thingy :) |
| 21:52 | <zcorpan> | @noamr:matrix.org: did you write down somewhere your idea about how to implement idscope efficiently? |