freenode :: #whatwg

17 May 2017
01:41knownanoni need a friend(s)
06:54XhmikosRMikeSmith: https://github.com/w3c/markup-validator/compare/master...XhmikosR:master?w=1
06:55* MikeSmith looks
06:55MikeSmithdang thats a long changelog
06:56XhmikosRit's the whitespace changes hence ?w=1
06:57MikeSmithOK
06:57XhmikosRshould I keep tabs for indentation?
06:57XhmikosRit seems this repo is using tabs
06:57MikeSmithno dont bother
06:57MikeSmithwell if you have time
06:57MikeSmithI guess it is better to be consistent
06:58MikeSmithbut that said, all this code it legacy stuff anyway
06:58MikeSmithso why do the PNGs show up in the CL?
06:58MikeSmith> htdocs/images/markup_validation_service.png
06:58MikeSmith> Diff not rendered.
06:59XhmikosRyeah the web interface doesn't load everything
06:59XhmikosRMikeSmith: only dupe gif/png is v15445
06:59XhmikosRhow does the badge work?
06:59XhmikosR URI = http://validator.w3.org/images/v15445
06:59XhmikosR Local URI = images/v15445
06:59MikeSmithno clue
06:59XhmikosRdoesn't specify the extension
06:59* MikeSmith looks
07:00MikeSmithoh
07:00MikeSmithwell W3C server use conneg
07:00MikeSmith*servers
07:00XhmikosRcorrection: there's also images/valid_icons which have both gif and png
07:01MikeSmithah Losslessly compress PNG files. is why the PNGs changed
07:01XhmikosRI guess it should be left like that otherwise we'd need to redirect
07:01MikeSmithyeah I suppose so
07:03MikeSmithOK LGTM
07:04MikeSmithuntested but not sure how to test it anyway
07:04XhmikosRI'll revert to tabs for indentation
07:04XhmikosRextra noise and it's not consistent for the repo
07:08MikeSmithOK
07:08MikeSmithyeah that would be good
07:09MikeSmithanyway if you raise an actual PR for it I will review it and merge it and push it out to production
07:09MikeSmithwith help from somebody on the W3C systems team to kick teh W3C Varnish caches
07:10XhmikosRand to verify we don't break anything with the gif images removal :)
07:10MikeSmithyeah
07:10annevkzcorpan: how much of the ?wss variant tests of websockets/ actually work?
07:12zcorpanannevk: i think secure websocket doesn't work at all with wptserve (yet). But we used to run ?wss in Spartan
07:16annevkzcorpan: but constants.js doesn't use the secure port because secure sockets doesn't work, but then the tests still don't actually work
07:16annevkI dunno, seems annoying
07:18annevkzcorpan: are there any cross-origin tests?
07:22zcorpanannevk: no
07:23annevkzcorpan: would they be hard to make?
07:23annevkzcorpan: I haven't really figured out the setup
07:23annevkzcorpan: I need something like opening-handshake/003.html but then to an IDNA subdomain
07:23annevkzcorpan: so I can test the origin field of the events (which doesn't appear to be tested at all)
07:27zcorpanannevk: seems like it shouldn't be hard. Changing constants.js to SCHEME_DOMAIN_PORT = 'ws://{{domains[www1]}}:' + PORT; seems like it successfully opens a websocket
07:28annevkzcorpan: it failed for IDNA
07:29XhmikosRMikeSmith: I split the patches for easier reviewing. Feel free to squash and merge those after proper reviewing, especially for the removed files
07:29annevkzcorpan: I did new WebSocket(url.replace("://", "://IDNA."))
07:33zcorpanannevk: SCHEME_DOMAIN_PORT = 'ws://xn--n8j6ds53lwwkrqhv28a.web-platform.test:' + PORT; seems to work for me
07:33annevkzcorpan: does it also work with non-ASCII input?
07:34annevkI wonder what's going on
07:35zcorpanannevk: Failed to construct 'WebSocket': The URL 'ws://.web-platform.test:50591/origin' is invalid.
07:35zcorpanannevk: I suppose the test needs to declare encoding
07:36annevkzcorpan: yeah, add that and it'll still fail (at least in my experience)
07:36annevkzcorpan: but in a different way
07:38zcorpanannevk: with <meta charset=utf-8> in opening-handshake/003.html it passes in chrome and firefox, if I comment out `assert_equals(e.wasClean, true);` (which is not relevant to whether open handshake was successful)
07:38zcorpanannevk: what fails for you?
07:39annevkzcorpan: okay if I get back to you in a bit?
07:39zcorpansure
07:47annevkFor BroadcastChannel Fx already did ASCII weirdly enough
07:48annevkOkay, back to WebSocket
07:50annevkzcorpan: it works for me too now
07:51annevkzcorpan: I suspect I had ?wss there still or some such
07:51zcorpanannevk: ok, great!
07:52MikeSmithXhmikosR: thanks will take a look
08:21MikeSmithXhmikosR: thanksdeployed to production
08:21XhmikosRMikeSmith: did you regenerate the minified files?
08:22MikeSmithno
08:22MikeSmithdunno how
08:22XhmikosRthere&#39;s a makefile
08:22MikeSmithis there are target for that in the makefile?
08:22MikeSmithah
08:22XhmikosRI skipped this because I thought you guys do it
08:22MikeSmithyeah
08:30MikeSmithXhmikosR: OK did that too
08:31XhmikosRMikeSmith: all seems to be ok
08:31XhmikosRone thing I couldn&#39;t change due to the way things are done is this
08:31XhmikosR <style type=&quot;text/css&quot; media=&quot;all&quot;>
08:31XhmikosR @import &quot;./style/base&quot;;
08:31XhmikosR </style>
08:31XhmikosRthis should be a link
08:31MikeSmithhmm yeah dunno why it is that way
08:32XhmikosRand perhaps add the favicon to the repo so that it&#39;s not served from w3.org
08:32MikeSmithoh
08:32MikeSmiththat would be good yeah
08:34XhmikosRI&#39;m not familiar with the src in order to make these changes myself
08:35MikeSmithIm not either
08:35MikeSmithnobody else around now is either
08:35XhmikosRalso are you sure the assets are being flushed?
08:36XhmikosRfor js combined I see last modified Tue, 14 Jul 2015 10:45:55 GMT
08:36XhmikosRand for the css
08:36MikeSmithoh
08:36XhmikosRbasically all assets
08:36MikeSmithyeah I need systems team to do the Varnish thing
08:37XhmikosRoh ok then what we see live is the old stuff
08:38MikeSmithyeah
08:39MikeSmiththere is some mechanism that I can supposedly use myself to purge teh caches
08:39MikeSmithbut every time Ive tried it it seems to have no effect
08:40MikeSmithso I just wait til somebody who has perms can do it however they do
08:45MikeSmithXhmikosR: check again later and if they havent been updated by tomorrow lemme know
08:46MikeSmithduring May some people are off or only partially around
08:48XhmikosRcool, I&#39;ll see if I can drop @import
08:51MikeSmithk
08:52XhmikosRmaybe some of the branches could be removed if merged?
08:53XhmikosRMikeSmith: https://github.com/w3c/markup-validator/compare/master...XhmikosR:master but it&#39;s totally untested
08:58* MikeSmith looks
09:00XhmikosRare there any tests we could run to verify we don&#39;t break anything?
09:01XhmikosRin theory my changes shouldn&#39;t break anything but I&#39;d feel more confident if we could verify it
09:02MikeSmithif tehre were tests I guess there would be targets in the makefile
09:04XhmikosRtest:
09:04XhmikosR misc/testsuite/harness.py --validator_uri=$(VALIDATOR_URI) run
09:05MikeSmith> AttributeError: &#39;ValidatorTestCase&#39; object has no attribute &#39;_cleanups&#39;
09:05MikeSmith> Makefile:43: recipe for target &#39;test&#39; failed
09:06XhmikosRthe src is a mess :/
09:06MikeSmithyup
09:07MikeSmiththis is the reason I never look at it
09:07MikeSmithanyway https://github.com/w3c/markup-validator/compare/master...XhmikosR:master LGTM
09:07MikeSmithI have no reason to expect it would not work as expected ni production
09:09XhmikosRme neither
09:09XhmikosRI have one more patch: switched to https for all http://validator.w3.org/ refs
09:09MikeSmithah good yes please
09:10XhmikosRverifying we don&#39;t break anything
09:10XhmikosRtotal mess
09:10XhmikosRfor example
09:10XhmikosR<xsl:param name=&quot;host&quot; select=&quot;&#39;validator.w3.org&#39;&quot;/>
09:10XhmikosR<xsl:param name=&quot;self&quot; select=&quot;concat(&#39;http://&#39;, $host, &#39;/whatsnew.atom&#39;)&quot;/>
09:10XhmikosRthis would break
09:11MikeSmithoh man xsl
09:14XhmikosRand would probably break some tests
09:14XhmikosRwith referer and stuff
09:15XhmikosRon the other hand it should fix some things like the favelet over https
09:16XhmikosRhmm scratch that since there is a redirect anyway
09:17XhmikosRMikeSmith: https://github.com/w3c/markup-validator/compare/master...XhmikosR:master
09:17XhmikosRif you still feel confident and take the resposibility, I can make a PR :)
09:18* MikeSmith looks
09:19XhmikosRI also changed any URI encoded instances to https
09:21XhmikosRhmm why is htdocs\dev\tests\trailing-nul.html
09:21XhmikosR marked as binary in the diff?
09:22XhmikosRsomething isn&#39;t right
09:22XhmikosRhmm unless that&#39;s what it tests :)
09:25MikeSmithdunno but LGTM overall
09:26MikeSmithso please do PR it
09:30tobieWhat&#39;s the difference between the ECMAScript global object and the Primary Global Object if any?
09:31annevktobie: Primary is the Window object, abstractly defined
09:32annevktobie: I think we should get rid of the PrimaryGlobal extended attribute personally and require Exposed to always be used
09:33annevktobie: that might be relatively easy cleanup if you want to do it, though will require lots of downstream changes, but those won&#39;t be hard
09:33annevk(and I&#39;m happy to help out with downstream)
09:33tobieannevk: is there an open issue for this?
09:34tobieannevk: sorry: do you know of the top of your head if there&#39;s an issue open for this?
09:34annevktobie: no
09:34annevktobie: to the latter
09:35annevktobie: I might have filed one though
09:37tobiehttps://github.com/heycam/webidl/issues/365
09:40annevktobie: https://www.w3.org/Bugs/Public/show_bug.cgi?id=26425
11:37XhmikosRon a side note MikeSmith, the CSS and JS files don&#39;t have cache busting of any kind
11:38XhmikosRalso, it doesn&#39;t seem you guys redirect to https
11:38XhmikosRC:\Users\XhmikosR\Desktop>curl -ILl http://validator.w3.org/
11:38XhmikosRHTTP/1.1 200 OK
12:02MikeSmithXhmikosR: I think that might be intentional but I dunno
12:04MikeSmithXhmikosR: if anybody uses http://validator.w3.org/ to check a document with a non-obsolete doctype I have long ago made it redirect to https://validator.w3.org/nu/ anyway
12:05XhmikosRMikeSmith: so those changes are not in httpd.conf in the repo
12:06XhmikosRI&#39;m using for years https://github.com/h5bp/html5-boilerplate/blob/master/dist/.htaccess with some minor adaptations. I believe we should use this here too
12:06XhmikosRbut since I don&#39;t know how you maintain this thing, it&#39;s just a suggestion
12:21smaugI wonder, given that Promises are an easy way to write jank-y web pages
12:21smaugshould there be some way to schedule Promises in a different way
12:22smaugsomething like IdlePromise
12:22smaugor AsyncPromise
12:23annevksmaug: how did you arrive at that premise?
12:24smaugprofiling Firefox devtools
12:24smaugand it shows up quite often
12:24smauglong promise chains
12:25smaugall resolved at the end of a microtask
12:25smaugwhich looks like async operation to JS
12:25smaugbut from UA point of view is sync
12:28annevksmaug: are microtasks costly then?
12:28smaugannevk: that is not the point
12:28smaugpoint is running too much JS during a task
12:29smaugwhich means no user input can be processed, nor do painting etc
12:29annevkIf you didn&#39;t want that presumably you&#39;d use setTimeout(..., 0)
12:30smaugannevk: well, that complicates the code a bit
12:30smaugsince one can&#39;t use the then() stuff
12:33annevksmaug: as I understand it a lot of libraries wanted to microtask-timing for their stuff so it would happen before paint, but I guess it&#39;s getting abused then
12:33smaugI can totally see reasons for the current scheduling
12:33annevksmaug: we could add more promise-based APIs though, for timers or run in next task, but it&#39;s unclear if those would see adoption
12:33smaugbut I can also see reasons to have different kind of scheduling
12:35ondrasDomenic: is it possible that jsdom&#39;s XHR would fail when sending ArrayBufferView instances? (works with .buffer though)
12:35annevksmaug: filing a bug against HTML seems reasonable if you want to further explore it
12:36annevksmaug: it seems you keep running into this, https://github.com/whatwg/html/issues/512 has similar suggestions (something that doesn&#39;t block paint, but runs asap)
13:50MikeSmithXhmikosR: caches purged
14:20annevkhsivonen: pretty good thread you got going there: http://unicode.org/pipermail/unicode/2017-May/thread.html#5389
14:27annevkEven includes a couple of asides on file systems
14:28noxannevk: Is the tone always like that on this ml?
14:29annevknox: I don&#39;t know, I don&#39;t read it often
14:29noxannevk: Good life choice I assume.
14:30annevknox: I kinda would like to be more involved with Unicode at times, but the current process is just so closed it&#39;s not really worth investing time in
14:31noxannevk: If I ever get involved, it will only be to make Fitzpatrick modifiers apply on the baguette emoji to indicate how baked it is.
14:31annevknox: Unicode also raises the &quot;cannot publish as HTML flag&quot;, which makes me less interested
14:31noxannevk: You mean the dozens of PDFs?
14:32annevknox: yes
14:32noxYeah that&#39;s sad.
14:32annevknox: with terrible xref etc.
14:32noxannevk: If only we had a plain text format with a precise definition of hyperlinks. :(
14:32annevkJavaScript switched recently and it&#39;s been so much better for collaboration
14:33annevkBut JavaScript has the advance-by-in-person-meeting flag raised, which is also sad
14:39gsneddersfoolip: how did you manage to repro the Edge flakiness on Edge 15? I can&#39;t get it to reproduce there at all?
14:39foolipgsnedders: I just reloaded in BrowserStack over and over
14:40gsneddershmm, ok
14:40foolipIt only took 5 or so tries to see it
14:40gsneddersfoolip: I&#39;m 99% sure this is a JIT bug with typeof HTMLAllCollection, FWIW
14:42gsneddersfoolip: I couldn&#39;t get it to repro on BrowserStack
14:42foolipgsnedders: do you mean that there&#39;s some fast path where they forgot about the special behavior?
14:42gsneddersfoolip: but it being a JIT bug dependent on previous executions make it unsurprising it went away trying to reduce it
14:42gsneddersfoolip: well, in some sense, yes
14:43gsneddersfoolip: though obviously the JIT compiler may not have such a fast-path explicitly there
14:44gsneddersfoolip: I expect it type-specialises and then fails if it then sees some new document object it hadn&#39;t seen before?
14:45foolipgsnedders: your expectations are better than mine I think :)
14:45foolipgsnedders: did you reduce many of those for farre perhaps?
14:45gsneddersand jl, etc. :)
14:45fooliphej farre!
14:46gsneddersit&#39;s reliably failing on Edge 15 on browserstack for me
14:46farrehej foolip!
14:46foolipfarre: waddayu working on?
14:46gsnedders(hej farre!)
14:47gsneddersfoolip: completely consistent on 15, completely random on 14 unless devtools are open (which apparently disables JIT)
14:48farrefoolip: currently timeout and/or requestIdleCallback stuff
14:49farrefoolip: you?
14:49farre(hej gsnedders!)
14:49foolipfarre: mmm, I guess requestIdleCallback is one where basically no shared tests can be written
14:50farrefoolip: we have some in wpt
14:50foolipfarre: Fullscreen, and generally web-platform-tests infrastructure, like the 2-way sync we have now.
14:50farrefoolip: https://github.com/w3c/web-platform-tests/tree/master/html/webappapis/idle-callbacks
14:51foolipfarre: oh, much more than I would have guessed
14:53foolipfarre: looks like callback-suspended.html is failing for us, I suppose that&#39;s not the most important of the lot?
14:53farrefoolip: could be that I&#39;ve written a crappy test
14:53foolipfarre: unpossible
14:53farrefoolip: I&#39;ve had a hard time getting wpt running with chrome.
14:55farrefoolip: for all browsers without some kind of bfcache, navigation should reload and have the test pass, but that might not be working
14:56foolipfarre: I was looking at our imported tests that run in content_shell, not a full Chrome. differences are possible
14:57jgrahamfarre: I&#39;m working on making that easier fwiw
14:57jgrahamSHould be able to expose it through mach soon
14:57farrehej jgraham!
14:57farrejgraham: and that would be awesome
14:58gsneddersoh wait, Edge does its compilation off-thread, IIRC, so a JIT bug might appear even more randomly depending on whether it JITs the code before the broken call happens
14:58foolipgsnedders: trying to minimize?
14:58gsneddersfoolip: trying to understand why I can&#39;t observe any randomness in 15
14:58gsneddersfoolip: I can&#39;t really minimize until I can repro it :\
15:00gsneddersalso we should really be using assert_in_array instead of assert_true(foo || bar)
15:02gsneddersI&#39;m seeing it about 50/50 in Edge 14, and never in 15.
15:05gsneddersoh, no, finally seen it repro
15:22XhmikosRMikeSmith: something isn&#39;t right
15:25XhmikosRMikeSmith: scripts/combined is broken
15:25XhmikosRstyles/base is fine
16:23annevkjgraham: you should email dev-platform once done, if not already planned
16:31DomenicAnyone want to tl;dr the Unicode thread for those of us who are only slightly morbidly curious?
16:34annevkDomenic: folks seem okay with changing UTF-8 error handling, some don&#39;t think it matters how you error, and some are upset like I am
17:21jgrahamannevk: I will
18:14ondrasDomenic: after looking into jsdom&#39;s xhr::send, I conclude that it indeed fails when sending ArrayBufferViews. I would send a PR, but am not sure how to actually *detect* an ArrayBufferView subtype... what would you suggest?
18:14Domenicondras: maybe a better conversation for the jsdom issue tracker or the #jsdom channel
18:14DomenicBut anyway, ArrayBuffer.isView
18:14ondrasah, thanks a lot.
19:00jugglinmikeJakeA: Do you have a moment to confirm my reading of the service workers spec?
19:21wanderviewjugglinmike: he&#39;s at google I/O I think
19:24jugglinmikewanderview: ah, got it. I found help from Marijn over in #chromium
19:24wanderviewI should probably lurk in that channel
19:24jugglinmikeso many channels
19:26wanderviewjugglinmike: there is a #blink too...
19:31jugglinmikeoh boy
19:40gsnedders#blink is dead now, AFAIK
19:40gsneddersI mean it probably still has people in it
19:40gsneddersBut I think officially it&#39;s dead
21:35AutomatedTesterTabAtkins: was /deep/ dropped from CSS selectors ?
21:36TabAtkinsKicked over to https://drafts.csswg.org/css-scoping/#deep-combinator and spelled differently
21:37TabAtkinsNote that it&#39;s still only in the static profile, which means usable in querySelector() in related APIs, not in CSS stylesheets.
21:37AutomatedTester TabAtkins I am interested in querySelector cases
21:37* AutomatedTester goes to read
21:38AutomatedTesterTabAtkins: thank you!
22:07DomenicFirefox got a name option for workers already; that was fast! :D
22:51wanderviewDomenic: yea, baku jumped right on that
22:52DomenicIt makes me happy when a vendor can file an issue, go through a relatively-fast standardization process, and have multiple vendors start implementing their idea within a month or so
22:53wanderviewDomenic: I think having the tests helps a lot... with something like this writing the test takes longer than the impl
22:53Domenic:D
22:54wanderviewanyway, hooray for the process working!
18 May 2017
No messages
   
Last message: 5 days and 4 hours ago