| 11:23 | <smaug____> | I wonder how much webcomponents implementation slows down normal DOM operations and whether to have a fast path when page doesn't use webcomponents |
| 12:08 | <annevk> | smaug____: custom elements, shadow trees, or both? |
| 12:09 | <smaug____> | both |
| 12:09 | <smaug____> | CE adds CEReactions overhead, and element creation etc |
| 12:09 | <smaug____> | shadow dom affects to event dispatch and such |
| 12:31 | <smaug____> | hmm, https://w3c.github.io/touch-events/#dom-touchevent-targettouches may need some tweaks in shadow DOM case. |
| 12:31 | <smaug____> | depends on what 'target' refers to |
| 13:38 | <annevk> | smaug____: there's an issue on that |
| 13:39 | <smaug____> | there is? |
| 13:39 | <smaug____> | there is that other touch event issue |
| 13:40 | <smaug____> | perhaps they are the same issue |
| 13:40 | <smaug____> | but I thought the other one was about retargeting touch.target in general |
| 13:41 | <smaug____> | but targetTouches needs to also change because of touch.target == event.target check |
| 13:41 | <annevk> | smaug____: it's supposed to be for touches/targetTouches/changedTouches |
| 13:42 | <annevk> | smaug____: the idea is to store all of those in a flat list (event's touch target list), which then gets modified as appropriate by the dispatch algorithm |
| 13:42 | <annevk> | smaug____: and then make sure touch.target grabs it from the correct location |
| 13:43 | <smaug____> | sure. It is just that other lists don't need to be changed, only their touch objects target |
| 13:43 | <smaug____> | targetTouches list itself needs to be modified |
| 13:44 | <annevk> | oh sorry |
| 13:44 | <annevk> | touch events :/ |
| 13:46 | <smaug____> | yeah, not exactly simple API, and designed behind closed doors |
| 14:13 | <wanderview> | falken: for the chrome-extension:// service workers... do you intercept cross-origin subresource requests to http/https? and do you enforce CORS there? |
| 14:53 | <smaug____> | annevk: do you think touch.targetTouches list object should be modified or replaced with a new one? |
| 14:54 | <smaug____> | er, touchEvent.targetTouches |
| 14:55 | <annevk> | smaug____: I guess you need multiple lists otherwise you can hold onto one and then if you peek at the appropriate time it'd be similar to the window.event leak |
| 14:56 | <smaug____> | right |
| 14:56 | <smaug____> | yeah, I was thinking too to create a new list |
| 14:56 | <smaug____> | just looking what blink does |
| 14:56 | <annevk> | smaug____: I doubt they do anything at the moment |
| 14:56 | <annevk> | smaug____: I think both Chrome and Safari haven't fixed these issues |
| 15:02 | <smaug____> | blink does adjust touch objects |
| 15:02 | <smaug____> | somehow |
| 15:53 | <smaug____> | window.event leak is different |
| 15:53 | <smaug____> | it is an object hanging on the global |
| 16:49 | <Domenic> | annevk: I don't quite understand what's going on in https://github.com/w3c/web-platform-tests/pull/10712 . You're using createElementNS, but browsers still end up with Element and not HTMLAnchorElement? |
| 16:50 | <Domenic> | What does that have to do with node document vs. global? |
| 16:50 | <annevk> | smaug____: sure, but conceptually it’s similar; you hold onto some object; some code runs while the event is in the shadow tree, can you find a way in? |
| 16:50 | <smaug____> | annevk: how is that different to event itself? |
| 16:50 | <annevk> | Domenic: they do not end up with Element? |
| 16:50 | <annevk> | smaug____: good point |
| 16:50 | <Domenic> | annevk: check http://w3c-test.org/submissions/10712/dom/nodes/Document-constructor.html |
| 16:51 | <Domenic> | If you use your script editor to log a.constructor.name it's Element |
| 16:52 | <annevk> | Domenic: ah, html -> xhtml |
| 16:52 | <annevk> | Namespaces are hard, can fix later |
| 16:52 | <Domenic> | Oh, OK. Then what is your comment about, node document vs. global? |
| 16:53 | <annevk> | Domenic: I draw the wrong conclusions thus far; I was thinking the URL stuff failed because it used the global to find a base URL and encoding |
| 16:54 | <Domenic> | I see, OK |
| 16:54 | <annevk> | Domenic: the output not being ASCII did not tip me off… |
| 16:54 | <Domenic> | Yeah, thus the comment I just added about including a sanity check. The failure mode being ".href just gives you back what you set" makes this test hard to reason about. |
| 16:55 | <annevk> | Will do |
| 17:31 | <annevk> | Domenic: I think you missed https://github.com/w3c/web-platform-tests/pull/10686 since it wasn't linked from a GitHub comment |
| 17:32 | <annevk> | Oh, and I had not removed the "needs tests" label |
| 20:09 | <howdoi> | What's am I missing on TransformStream here? -> https://jsfiddle.net/gnumanth/yve4froq/ |
| 20:10 | <howdoi> | 'pipeThrough' on 'ReadableStream': parameter 1's 'writable' property is undefined. |
| 20:13 | <howdoi> | shouldn't TransformStream return both write and read streams? |
| 20:13 | <howdoi> | /cc @Domenic ^^ |
| 20:14 | <Domenic> | howdoi: you are passing a function, not a TransformStream. |
| 20:15 | <howdoi> | Domenic: duh! my bad, thanks, so it's rather `await resp.body.pipeThrough(markedTransform())` is there a way to `.pipeTo` console? |
| 20:16 | <Domenic> | No, neat idea though. |
| 20:16 | <Domenic> | pipeTo(new WritableStream({ write: console.log })) should probably work though |
| 20:18 | <howdoi> | agree, just for debugging purpose |
| 20:22 | <domfarolino> | Domenic: could you tell me what `console.countReset()`produces in Edge (any output at all?, warning that no previous count exists?) |
| 20:23 | <Domenic> | domfarolino: it outputs " null: 0" |
| 20:23 | <Domenic> | If I call it again it outputs nothing |
| 20:23 | <Domenic> | This seems surprising |
| 20:23 | <domfarolino> | Whoah really |
| 20:24 | <Domenic> | Uhhh now I've broken the console by reloading |
| 20:24 | <Domenic> | I guess reloading on about:blank does not make the Edge console happy |
| 20:24 | <Domenic> | Anyway same results on another random page |
| 20:24 | <howdoi> | https://jsfiddle.net/gnumanth/yve4froq/3/ done :-) |
| 20:25 | <howdoi> | we can't pipeTO dom yet, right? (unless we do that iframe trick thinge) |
| 20:26 | <domfarolino> | Oh boy lol |
| 20:26 | <howdoi> | :D |
| 20:26 | <domfarolino> | Ok, I am leaning towards what all browsers are doing now for timeEnd with no timer (a warning) |
| 20:28 | <howdoi> | is something wrong with `.pipeThrough(new TextDecoder())` ? |
| 20:28 | <domfarolino> | Will see if we can get edge and node on board since they're the impls that have countReset as of now |
| 20:32 | <domfarolino> | Domenic: would you mind calling it again with a label like "a" or something? Want to make sure the behavior you're seeing is not a default conversion issue, but an actual anomaly with the API |
| 20:55 | <Domenic> | domfarolino: uhhh it appears that it's similar. The first time I call it it prints " a: 0". Future times it prints nothing. |
| 20:55 | <domfarolino> | Gotcha, thanks. That's messy |
| 21:11 | <domfarolino> | Domenic: RE you're console comment: ahh that's right, whoops I misunderstood. Hopefully we can get this specc'd that'll be nice :) |
| 21:11 | <Domenic> | Yeah current Edge behavior seems pretty bonkers |