freenode :: #whatwg

9 Jan 2017
08:40annevkJakeA: in case you forgot about here's a friendly reminder (I'm in no hurry though, just noticed it's likely after your break)
09:35JakeAannevk: cheers. Hadn't forgotten, but yeah, will start this today. Last time I looked Chrome was crashing on the fetch tests, so might end up down a rabbit hole
09:36annevkWell, as I said, I don't think we're in a particular rush with that refactoring, so up to you
09:57annevkjgraham: is there any H/2 talk with regards to WPT?
09:58annevkjgraham: and also, is there a way to enforce H/1 for certain tests?
10:19jgrahamannevk: No and therefore no. I mean everything enforce http/1.1 at the moment; if you aren't running through wptserve it's an unsupported configuration
10:19annevkjgraham: wptserve is used for WebSocket too?
10:20jgrahamOh, no
10:20annevkjgraham: or to ask differently, tracking issues should be filed there?
10:20jgrahampywebsocket is used then
10:20annevkI see
10:20annevkI guess when I end up writing those tests I can say they assume H/1
10:20jgrahamIf http/2 is required I guess I should add that to my list of "reasons to spend more time on web-platform-tests this year"
10:21jgrahamThat is, if there are things that can't be tested or might work differently with http/2
10:21annevkjgraham: it would be good for Fetch at least to test H/2 framing and such
10:21jgrahamOK. There is a bug somewhere, right?
10:22jgrahamCan you add a comment if there isn't one already
10:22annevkjgraham: I don't know, I just got curious
10:22annevkjgraham: there is a bug about sending responses earlier
10:22annevkjgraham: should I file a bug against wptserve?
10:23jgrahamannevk: Sure
10:24Ms2gerannevk, hey, you know a few things about incumbent globals, right?
10:25annevkMs2ger: don't use m?
10:26annevkMs2ger: the main thing I know about incumbent globals is that they're not a shared concept across all browsers
10:26annevkMs2ger: main important thing, that is
10:28annevkjgraham: filed
10:29annevkjgraham: is the other issue, with early responses
10:29annevkjgraham: that would be more interesting to have I think since it enables completely new scenarios and bugs
10:30Ms2gerIt seems like uses `callback` as a WebIDL callback function (which has an incumbent global stored as the "callback context"), but it's created as a JS function, which doesn't have that
10:30annevkjgraham: basically being able to send a response from the server the moment the server has all the request headers, not necessarily the request body
10:32annevkMs2ger: because of step 1?
10:32Ms2gerYeah, just does FunctionCreate
10:33annevkMs2ger: new issue
10:34annevkMs2ger: that should indeed some some kind of IDLFunctionCreate
10:34Ms2gerOn its way
10:34Ms2gerThe annoying thing is that it's called without JS on the stack
11:53JakeAannevk: I'm getting "updating and loading test manifest; this may take several minutes" on web-platform-tests. It doesn't seem to ever finish
11:54JakeAI'm also getting a ManifestVersionMismatch error in the console
11:54JakeA(server console that is)
11:56annevkJakeA: what are you trying to do?
11:56JakeAannevk: run the fetch tests
11:57JakeAannevk: I visit http://web-platform.test:8000/tools/runner/index.html and get that error
11:57annevkJakeA: I see, you can also go to e.g. http://web-platform.test:8000/fetch/api/headers/headers-errors.html directly
11:57annevkJakeA: that's what I usually do
11:57jgrahamJakeA: Delete your MANIFEST.json
11:57annevkJakeA: jgraham might be able to help with questions regarding the runner
11:57annevkThere you go
11:57jgrahamWe changed the manifest format and it probably isn't intelligent enough to just regenerate it in the new format
11:58* annevk hasn't used the runner
11:58jgrahamYeah, the runner isn't great
11:59JakeAjgraham: where's MANIFEST.json?
11:59JakeAoh top dir
12:00JakeAjgraham: thanks, that's working now
12:03jgrahamApologies for the inconvenience
12:05MikeSmithah yeah
12:05MikeSmithCindy ran into the same problem and pinged me about it
12:05MikeSmithtold her the same thing
12:05MikeSmithrm MANIFEST.json && ./manifest
12:06jgrahamWe can probably make it happen by magic, although if the rebuild takes too long the request might time out
12:41JakeAjgraham: is there a process for running the tests over https?
13:00JakeAjgraham: figured it out. Filed as a reminder to document it
13:11annevkWhy does wpt-stability-bot report results in seemingly random order?
13:11gsneddersannevk: Python hash order, maybe?
13:11gsneddersannevk: feel free to file a bug, tho
13:12annevkgsnedders: can you point me to the code perhaps?
13:12annevkgsnedders: maybe I can propose a change
13:12gsnedders /, IIRC
13:12annevk should probably mention where you can patch it
13:13gsneddersI mean it could also be somewhere in wptrunner
13:13gsneddersannevk: idk how has access to that, maybe only jgraham
13:14gsneddersclass LogHandler is where it all goes wrong
13:14gsneddersthough replacing those with ordered dicts is harder than I'd like
13:15annevkYeah, was just looking at that
13:15annevkI'm going to file an issue for now and get back to IDNA
13:36jochen__any suggestions what name to use to refer to an "inline style declaration"
13:36jochen__as opposed to a stylesheet
13:37jochen__i want to say something "if an url() is part of an inline style, then do this"
13:46annevkjochen__: it sounds like that requires a hook in the CSS parser or some such
13:47annevkjochen__: so probably best to ask the CSS WG for such a hook
13:47annevkjochen__: unless you're happy with your text not being the long term solution, in which case you can do whatever
13:47annevk(within reason)
14:37jgrahamMikeSmith, JakeA: I think the manifest version mismatch problem is fixed now
14:41MikeSmithjgraham: super, thanks
14:41MikeSmithwill let Cindy know too
14:41MikeSmithannevk: spaces around em dash?
14:43annevkMikeSmith: seems to be the norm
14:43annevkI personally think it reads better too, but I know it's sometimes frowned upon too
14:44MikeSmithits not the norm in any style guide I have ever seen, nor in pretty much any print newspaper or magazine or book
14:45MikeSmithbut that said I can live it as long as we do it consistently
14:46MikeSmithanything is better than continuing to instead do -- forever
14:47gsnedderstwo hyphens is normally en-dash
14:47gsneddersbut yeah, I agree with annevk that it reads better IMO, but AFAIK pretty much every style guide ever says no spaces
14:48jgrahamI really will never understandand believe me I have triedwhy style guides think no space is a good idea
14:49annevkI don't think any WHATWG Standard avoids spaces
14:50MikeSmithyall sound like you probably think two spaces after a period reads better too
14:51jgrahamI hope not
14:51annevkNooo, those are bad
14:52annevkAnd trimmed by CSS unless you use   so rather pointless anyway
14:53annevkIf you have a string with that reads "\u1234" how do you convert the \u1234 to a code point in the easiest way with Python?
14:53annevkDo I need my own tokenizer of sorts or is there a simple utility?
14:58MikeSmithnow that is something that jgraham and gsnedders definitely know more about than I do
14:59jgrahamYou have a literal python escape in a string?
14:59jgrahamIn [2]: "\\u1234".decode("unicode-escape")
14:59jgrahamOut[2]: u'\u1234
15:04annevkjgraham: cool
15:18* wanderview considers staying an extra day or two after SW meeting in tokyo...
15:21JakeAjgraham: cheers!
15:24wanderviewoooh... JakeA is writing WPT tests?
15:35annevkjgraham: I guess they're not literal Python escapes after all, since they can be up to six digits
15:43jgrahamannevk: Ah, python ones are like "\\U00101234".decode("unicode-escape") for non-BMP
15:44jgrahamSeems like they got scared that someone would sneak in some more codepoints
16:08MikeSmithgsnedders: the tree at is not correct, right?
16:16MikeSmiththat is, it parses <!DOCTYPE html><noscript><img src=foo></noscript> into <html><head><noscript></noscript></head><body><img src=foo>...
16:16MikeSmithbut browsers do not seem to do that that
16:19zcorpanMikeSmith: scripting disabled?
16:19MikeSmithyeah I guess thats the difference
16:20MikeSmithvnu parser runs in the HTML checker with scripting disabled
17:07annevkjgraham: turned out they were of the \uXXXX type after all, but that doesn&#39;t really work with Python if the string contains those and non-ASCII code points
17:08jgrahamWhat doesn&#39;t work?
17:10annevkjgraham: it turned a legitimate \u00df into \u00XX\u00XX where those XX were the utf-8 byte sequence for U+00DF
17:11annevkjgraham: I then tried to run decode(&quot;utf-8&quot;) first and then run decode(&quot;escape-string&quot;), but yielded similar bad results
17:11annevkA custom method did the trick however...
17:12annevkI&#39;m guessing I should really be using Python 3.0 to avoid this, but then that doesn&#39;t come by default on the OS
17:12jgrahamIn [8]: &quot;\\u00df&quot;.decode(&quot;unicode-escape&quot;)
17:12jgrahamOut[8]: u&#39;\xdf&#39;
17:12jgrahamIn [9]: &quot;\xdf&quot;.decode(&quot;unicode-escape&quot;)
17:12jgrahamOut[9]: u&#39;\xdf&#39;
17:14annevkjgraham: but the input is something like &quot;\\u1234 \u00DF&quot; (so escapes and actual code points)
17:14annevkAnyway, I wrote a custom thing
17:15annevkjgraham: if you want to review or improve upon my way, it&#39;s in as
17:17jgrahamannevk: Oh I see. If you already have a unicode string, running .decode() on it will probably cause badness (attempting to encode to ascii first or something)
17:24DomenicWait when in the Tokyo service workers meeting?
17:26annevkDomenic: it&#39;s 3-4 April, JakeA can give you the details
17:26DomenicHmm OK
17:26Domenicfoolip is out there all of March and I am scheduling my trip to probably overlap with his
17:26DomenicMaybe I should go for end of March / beginning of April for maximum overlap
17:27annevkWe&#39;re likely moving house, so no trips
18:47JakeAwanderview: I know right! Dangerous times. Hoping to have something to PR tomorrow. Just getting up to speed today & got sidetracked a little
18:47JakeABut seems like something I should be doing with all spec things
18:47JakeADomenic: you&#39;re more than welcome to the SW meeting. Happy for streams to be part of that too.
10 Jan 2017
No messages
Last message: 226 days and 14 hours ago