08:02
<hsivonen>
GPHemsley: thanks. I posted http://lists.w3.org/Archives/Public/public-css-testsuite/2015Jan/0016.html
09:58
<annevk>
http://intertwingly.net/blog/2015/01/11/URL-Work-Status is somewhat hard to grok. The barrier to entry at the WHATWG is too high, yet everywhere else he hits a door.
10:13
<hemanth>
'http://f: 21 / b ? d # e' interesting
10:16
hemanth
tried to reflect on ES6 Reflect API http://h3manth.com/new/blog/2015/es6-reflect-api/
10:58
<MikeSmith>
annevk: https://twitter.com/ttepasse/status/554341595185946625
10:58
<annevk>
"fora⊙an" that's a long time ago
10:59
<MikeSmith>
2005 it seems
11:01
<annevk>
That was at Opera, around the time of Opera's 10 year anniversary and around the time of me getting hired for a longer period
11:04
<MikeSmith>
then before I started at Opera
11:05
<annevk>
I guess I switched to using @opera.com a little later, when everything was more permanent
13:34
<MikeSmith>
https://community.rapid7.com/community/metasploit/blog/2015/01/11/google-no-longer-provides-patches-for-webview-jelly-bean-and-prior
13:35
<MikeSmith>
seems pretty surprising if true
13:38
<caitp>
it's pretty sad for people who are contract-tied to their old phones and can't upgrade yet, or who are financially unable to upgrade
13:38
<caitp>
but surprising?
13:50
<annevk>
Hmm, I wanted to write about custom elements, but I think I should first explain web platform objects... Turtles, meh
17:17
<wanderview>
JakeA: annevk: if content runs window.caches.addAll(requests)... should those be intercepted by ServiceWorker? Does Cache add()/addAll() implicitly get the skip service worker flag?
17:18
<annevk>
wanderview: I feel like they should go through the worker
17:18
<wanderview>
annevk: unless initiated by the ServiceWorker?
17:18
<annevk>
wanderview: fetches from a service worker can never go through that service worker
17:19
<annevk>
(nor any other one at the moment, but it seems like that might change going forward)
17:20
<annevk>
wanderview: that should happen automatically in Fetch due to the associated client of the request
17:21
<wanderview>
annevk: yea... its a quirk of our cache add/addAll implementation that we don't get it automatically... Cache is using a lower level API
17:23
<JakeA>
wanderview: agree with annevk
17:23
<wanderview>
thanks
17:24
<wanderview>
so we can have window cache.addAll()... go to SW which then does more Cache operations
17:24
<wanderview>
nested within the cache.addAll()
17:27
<annevk>
turtles!
17:28
<annevk>
but yeah, that might get hairy quickly
17:28
<annevk>
need some kind of atomicity otherwise you get races
17:30
<wanderview>
I think this probably is mostly ok since the spec is written async with many operations in flight already
17:30
<wanderview>
its kind of tricky for the gecko implementation, though...
17:35
<JakeA>
annevk: if a call results in an error that isn't speced (perhaps implementation specific), should the browser throw the most appropriate error, or its this a signal that the implementation or the spec is wrong?
17:37
<JakeA>
annevk: I guess what I'm asking is: are there specs that loosely define throwing like "if an error occurs, throw an appropriate error"
17:37
<JakeA>
(feels like insufficient specing to me)
17:38
<Ms2ger>
There's lots of insufficient speccing
17:38
<Ms2ger>
Like HTML "if a network error occurred"
17:40
<jgraham>
The bit after the comma seems particularly poor though
17:40
<jgraham>
The nature of the error shouldn't be left to chance^Wdevelopers
17:47
<annevk>
JakeA: that would be a bug in the spec
17:48
<annevk>
JakeA: the only thing we're loose on is the error message, as that can be localized and such
17:55
<JakeA>
annevk: thought so, ta
18:33
<jgraham>
Does anyone know what happened to the testharness.js-based Chromium Wervice-Worker tests?
18:33
<jgraham>
Did they ever get submitted to wpt?
18:34
<jgraham>
Probably the ones at https://code.google.com/p/chromium/codesearch#chromium/src/third_party/WebKit/LayoutTests/http/tests/serviceworker/&q=service-worker&sq=package:chromium&type=cs
18:35
<Ms2ger>
Speaking of chromium tests
18:35
<Ms2ger>
jsbell, any news on TextEncoder/Decoder?
18:38
<jgraham>
Seems like jsbell might also be able to help with my question
18:38
Ms2ger
approaches jsbell from behind
18:39
<jsbell>
Ms2ger: thanks for the ping... I started to move some of them my local w-p-t repo before the holidays, need to get back to it.
18:39
<jsbell>
... into my local...
18:39
<Ms2ger>
Nice to hear that
18:39
<jsbell>
jgraham: ??
18:39
<Ms2ger>
<jgraham> Does anyone know what happened to the testharness.js-based Chromium Wervice-Worker tests?
18:39
<Ms2ger>
<jgraham> Did they ever get submitted to wpt?
18:39
<Ms2ger>
<jgraham> Probably the ones at https://code.google.com/p/chromium/codesearch#chromium/src/third_party/WebKit/LayoutTests/http/tests/serviceworker/&q=service-worker&sq=package:chromium&type=cs
18:40
<Ms2ger>
jsbell, fwiw, smaller PRs tend to get landed quicker, so you don't necessarily need to wait until you've got them all
18:40
<jsbell>
not submitted yet, we still plan to
18:41
<jgraham>
jsbell: If you can get them submitted we can get them reviewed (although technically that's not needed if you are confident they're good to go)
18:41
<jgraham>
jsbell: The Mozilla implementors would like to run them
18:42
<jsbell>
jgraham: I'll chat w/ the team tonight, see if anyone can take it on. (new quarter, new priorities, yadda yadda)
18:42
<jgraham>
Yup
18:42
<jgraham>
If you need any help let me know
18:42
<jgraham>
Like, if it's as simple as "copy the files from the Chromium tree into the wpt tree" I can just do that for you ;)
18:52
<jsbell>
jgraham: you're welcome to give it a shot. The big issue is likely to be that we don't run them w/ serve.py so many likely have assumptions about our test server config that'll need correcting
18:53
<Ms2ger>
jsbell, any news on moving to wptrunner?
18:54
<jsbell>
ms2ger: no news
18:55
<jgraham>
jsbell: OK, great. Like I say we have interest in running these which I assume (optimistically!) means we have resources to make it happen ;)
18:57
<wanderview>
yea... the tests were helpful when I ran them manually... it would be nice to get them imported before we ship
18:57
<Ms2ger>
wanderview, put it on your todo list :)
18:58
<wanderview>
Ms2ger: the last time I checked the blink folks were waiting on PRs in review with jgraham... once they uplift them to the wpt repo I don't think there is much for us to do besides pull them in
18:59
<wanderview>
it seems we're past that point now, though... so hopefully they can get uplifted
19:00
<jgraham>
Right, the testharness.js changes landed
19:04
<wanderview>
cool
20:03
<TabAtkins>
annevk: The mutation events crashers were about people using nodes exposed by mutation events (particularly removal events) and making more mutations in/with them - they were incompletely sterilized and still thought they had some sort of connection to the DOM, which causes inconsistent state when mutating, and eventually hits crashes.
20:03
<TabAtkins>
It's like the problems with the instance tree in <use> - nothing inherent to the tech, but persistent crashiness due to human frailty in coding, and that just being something which regularly exposed such problems.
20:08
<annevk>
TabAtkins: there are some concerns around synchronous mutation that would be relevant here
20:08
<TabAtkins>
annevk: Likely, yeah. People could be hanging onto references to the old version of the element.
20:09
<annevk>
well there's no old version in solution 1)
20:09
<TabAtkins>
Yeah, but there's all the ordering and raciness there. :(
20:10
<annevk>
why?
20:10
<TabAtkins>
I explained in the email, I thought. Is it still unclear?
20:11
<annevk>
yeah I guess, not sure it's insurmountable
20:12
<TabAtkins>
It's not insurmountable, it's just a persistent annoyance. Upgrading makes things more declarative, which matches the spirit of using HTML better.
20:12
<TabAtkins>
If you could only use custom elements by constructing them, it wouldn't matter - the ordering constraints of JS would make the right behavior fall out.
20:14
<annevk>
if you ensure your dependencies are loaded you can have constructors during parsing
20:14
<TabAtkins>
"if you ensure" is the hard part. It means, for example, that you need to load your element definitions in a sync script block before your markup.
20:55
<Hixie>
annevk: lol. if the barrier to entry is too high when the barrier is one's own ability to do the work...
20:56
<Ms2ger>
mailing list is a support forum
21:30
jgraham
wonders how long checking out blink is supposed to take
21:30
<miketaylr>
multiple hours, last i did that
21:34
<roc>
TabAtkins: FWIW the problems Blink had with <use> may have been Blink's approach, not inherent. We always had a simple cloning strategy and didn't seem to have these problems
21:35
<roc>
mutation events, on the other hand .... blergh
21:35
<jgraham>
miketaylr: I'm up to 1:45 and no end in sight
21:36
<jgraham>
But my internal monolouge of the process has started to sound like the narration of a castaway
21:36
<miketaylr>
heh
21:37
<miketaylr>
be sure to find a volley ball to keep you company
22:31
<jgraham>
Hmm, so 2.5 hours in chromium checkout failed due to low disk space
22:33
<Ms2ger>
It's what, 50GB?
22:38
<jgraham>
I hope not
22:39
<jgraham>
If it is then even deleting B2G isn't going to help
22:40
<Ms2ger>
"Either way, fetch checks out more than 10GB"
22:40
<Ms2ger>
Sounds like I overestimated a bit
22:40
<jgraham>
I think I had 16 when it stopped
22:41
<jgraham>
B2G is 26, but I think that's almost all android
23:44
<jamesr__>
my chromium checkout is roughly 20GB, excluding build artifacts
23:45
<jamesr__>
debug build is 27GBish, depending on configuration
23:45
<jamesr__>
and what targets you wanna build
23:48
<caitp->
that's my favourite thing about v8
23:48
<caitp->
tiny