03:32
<annevk>
Domenic_: because if you want to transport some JSON you need to know whether you can use a BOM in front of it or not
03:32
<annevk>
Domenic_: you can't anticipate clients on the web
04:01
<Hixie_>
cabanier: i haven't looked at canvas things for a while, sorry. been trying to get security and scripting figured out.
04:08
<cabanier>
Hixie_: no problem! As long as it's on your to do list.
04:09
<Hixie_>
yup, definitely there
04:09
<Hixie_>
you can see what's on the todo list here btw: http://www.whatwg.org/issues/
04:20
<jgraham>
Do we have any generally agreed on standards for web-platform-tests code? Basically what should I mention when I do code review with webapps?
04:21
<jgraham>
I guess only darobin is here at the moment :|
04:21
<darobin>
jgraham: standards? you mean like coding conventions?
04:26
<jgraham>
darobin: Yeah. Also what to look for other than just "does the test match the spec"
04:26
<jgraham>
So for example "use of external resources is bad"
04:26
<jgraham>
s/bad/forbidden/ I guess
04:26
<MikeSmith>
jgraham: should mention about, use idlharness for interface tests rather than writing them manually
04:26
<darobin>
ah
04:27
<darobin>
no, we don't have a checklist
04:27
<jgraham>
MikeSmith: Ah, good point
04:27
<darobin>
it would be good to have a checklist
04:27
<jgraham>
Yup, that's what I'm trying to write
04:28
<jgraham>
Well actually at the moment I have got as far as "here are the review tools". But the next thing will be what to check for when you do the review
04:28
<jgraham>
Which should actually go in the documentation, so I should write it up properly
04:29
<MikeSmith>
jgraham: I think another thing to point out is that people should get familiar with the set of available assert_* methods and use the most specific one available when there is one
04:32
<jgraham>
MikeSmith: Added
05:58
<jgraham>
MikeSmith, zcorpan_: https://etherpad.mozilla.org/MeHiumCufk
06:01
<MikeSmith>
jgraham: thanks will try to think if I can add anything useful there
06:03
<zcorpan_>
jgraham: why " - Best to avoid the pattern where you have a test that asserts that the feature is supported and bails if it isn't" ?
06:06
<jgraham>
zcorpan_: I mean "bails the whole page". When people do this they often depend on some assumptions about feature detection and then assume that none of their other tests will work if the feature detection failed. It seems much better to just run all the tests and let them fail unless that will take a rally long time
06:06
<zcorpan_>
oh, ok
06:06
<jgraham>
Getting a consistent number of test results is helpful for some kinds of tracking
06:07
<zcorpan_>
sometimes there's a separate test that just checks whether the feature is there, independently of the rest of the tests, which seems OK
06:07
<zcorpan_>
since if it fails it's very easy to see why
06:07
<jgraham>
Yep, that is fine
06:08
<zcorpan_>
maybe reword that point to be a bit clearer
06:09
<jgraham>
Sure
06:09
<jgraham>
These are just notes at the moment
06:26
<MikeSmith>
darobin_: https://etherpad.mozilla.org/MeHiumCufk
06:26
<darobin_>
MikeSmith: ta
07:19
<annevk>
http://lists.w3.org/Archives/Public/www-style/2013Nov/0085.html
07:19
<annevk>
"You should have attended the META conference in september..."
07:19
<annevk>
o_O
08:17
<jgraham>
Ms2ger: https://etherpad.mozilla.org/MeHiumCufk
08:18
<Ms2ger>
Nice
08:21
<jgraham>
Ms2ger: Contributions welcome
08:38
<annevk>
http://xkcd.com/1289/ brilliant
08:39
<Ms2ger>
jgraham, when is the session?
09:00
<philipj_>
jgraham, what's the best way to run web-platform-tests locally?
09:00
<philipj_>
it looks like it needs a web server because of the paths starting with /
09:00
<philipj_>
also, the ability to run all the tests in a directory would be nice
09:02
<jgraham>
philipj_: The best way is to check out the jgraham/python_review_2 branch and follow the instructions in README.md
09:02
<jgraham>
There isn't any way to run a whole directory automatically because that's assumed to be browser-specific
09:02
<jgraham>
I guess you want to talk to Dirk Pranke
09:03
<philipj_>
jgraham, on t we had these runner.php in some places, which isn't browser-specific
09:03
<philipj_>
but it doesn't exist, I get it :)
09:04
<jgraham>
Right, but some tests depend on not being in an iframe and so on
09:04
<jgraham>
Could probably make a simple cross-browser harness that would be OK for local usage
09:04
<jgraham>
In fact Ms2ger might alrady have one
09:04
<philipj_>
oh, never encountered that for media tests, but of course that's sounds like an insurmountable problem
09:05
<jgraham>
Well window.open helps
09:05
<jgraham>
I think the main thing that's hard to do is recover from crashes or hangs or whatever :)
09:05
<philipj_>
recovering from a crash is indeed hard
09:06
<jgraham>
(so doing something simple and entirely in-browser might make sense for non-production use cases)
09:07
<philipj_>
jgraham, does serve.py need to be run as root?
09:07
<jgraham>
philipj_: No
09:08
<jgraham>
It doesn't use port 80, for better or worse
09:08
<philipj_>
I get socket.error: [Errno 110] Connection timed out
09:08
<jgraham>
On startup?
09:08
<jgraham>
Or when trying to access a TC?
09:08
<philipj_>
on startup
09:08
<jgraham>
Do you already have something running on port 8000?
09:09
<philipj_>
Not to my knowledge
09:09
<jgraham>
By default it tries to use 8000 and a couple of random free ports
09:09
<jgraham>
(you can change that in config.json)
09:10
<philipj_>
lsof | grep -w 8000 does find a few things, but it looks like it's port 8000 on the other side of the TCP connection
09:10
<jgraham>
(Is that different to lsof -i :8000 ?)
09:10
<philipj_>
I have no idea, I've hardly ever used lsof
09:11
<jgraham>
Right, that's the only usage I know :)
09:11
<philipj_>
lsof -i :8000 finds one of the four things I saw before
09:11
<philipj_>
anyway, it's not because of the port, no random port I tested works
09:11
<jgraham>
Oh
09:12
<jgraham>
Well can I see the error?
09:12
<philipj_>
jgraham, http://pastebin.com/C00FKaZM
09:12
<philipj_>
Python is 2.7.4
09:13
<philipj_>
wow, pastebin has really horrible ads...
09:15
<jgraham>
Doesn't work from China :) (but I can read it anyway)
09:16
<Ms2ger>
philipj_, pastebin.mozilla.org doesn't have ads ;)
09:16
<jgraham>
It isn't very obvious. Did you edit /etc/hosts?
09:16
<philipj_>
Ms2ger, thanks!
09:16
<Ms2ger>
And my cross-browser test runner does use iframes (https://bitbucket.org/ms2ger/test-runner)
09:17
<sangwhan__>
philipj_: if all fails "python -m SimpleHTTPServer" works as a quick hack
09:18
<philipj_>
jgraham, I don't have anything weird in my /etc/hosts
09:19
<philipj_>
jgraham, maybe I should mentioned that I merged your branch to master instead of actually checking it out, will try doing as you said instead
09:19
<jgraham>
philipj_: You are supposed to add web-platform.test and some other subdomains
09:19
<philipj_>
jgraham, oh look, I started reading at the submodule stuff
09:20
<jgraham>
I should fix to not throw though
09:20
<jgraham>
There was an older model where it used "localhost"
09:20
<jgraham>
But localhost turns out to be a magic name
09:20
<philipj_>
you could try resolving all of the required hosts and telling me to RTFM if it fails I guess
09:21
<jgraham>
That is what it's trying to do, almost
09:21
<jgraham>
It just doesn't work if web-platform.test itself doesn't resolve
09:21
<jgraham>
Which is a bug
09:24
<philipj_>
jgraham, glad to me of service by not reading the readme :)
09:24
<philipj_>
well now I have it running, so thanks!
09:24
<jgraham>
Great :)
09:34
<jgraham>
~.
09:38
<Ms2ger>
annevk, yt?
09:39
<annevk>
Ms2ger: sure
09:39
<annevk>
Ms2ger: sitting next to jgraham, plotting
09:40
<Ms2ger>
Well, that's an excellent combination :)
09:40
<Ms2ger>
So the load event fired on window apparently needs to have target==document
09:41
<Ms2ger>
Want to check if https://github.com/mozilla/servo/pull/1213 is a reasonable way to implement that?
09:44
<annevk>
Ms2ger: http://www.whatwg.org/specs/web-apps/current-work/#events-and-the-window-object
09:45
<annevk>
Ms2ger: not sure what is going on there
09:45
<annevk>
Ms2ger: or why it is non-compliant
09:46
<Ms2ger>
Because a fully compliant implementation of page loading is hard
09:47
<annevk>
AbstractEventTarget::from_document(document) looks icky
09:47
<Ms2ger>
Yeah, that's working around the lack of inheritance
09:49
<annevk>
some sad code
09:49
<Ms2ger>
So the issue is that step 7.2 of http://www.whatwg.org/specs/web-apps/current-work/multipage/the-end.html#the-end can't really be done if you call into http://dom.spec.whatwg.org/#concept-event-dispatch
09:50
<annevk>
remind me
09:51
<Ms2ger>
Come again?
09:52
<annevk>
How can it not be done?
09:54
<Ms2ger>
Because http://dom.spec.whatwg.org/#concept-event-dispatch step 3 sets target to the wrong value, and you'd have to change it to the document somewhere in the middle of the algorithm
10:00
<annevk>
Beautiful
10:00
<annevk>
Ms2ger: file a bug
10:01
<annevk>
Add this to the have Hixie_ merge HTML and DOM
10:01
<annevk>
bucket
10:01
<hober>
we're gonna need a bigger bucket
10:04
<Ms2ger>
annevk, not sure how that'd help, except that then Hixie_ would need to fix it :)
10:05
<annevk>
Ms2ger: for starters we'd have the whole object chain in one spec
10:06
<annevk>
Ms2ger: anyway, file a bug, disregard bucket for now
10:06
<annevk>
kthxbai
10:06
<Ms2ger>
https://www.w3.org/Bugs/Public/show_bug.cgi?id=23792
13:18
<philipj>
zcorpan, what do you think about https://github.com/w3c/web-platform-tests/pull/429 ?
13:19
<philipj>
VTTCue is the only path for testing TextTrackCue, but it looks a bit odd to test TextTrackCue using a new VTTCue(...)
13:19
<zcorpan>
i'll have a look
13:25
<zcorpan_>
annoying how critic handles moved and edited files
13:31
<zcorpan_>
philipj: do the tests pass?
13:33
<zcorpan_>
philipj: the only things changed are s/TextTrackCue/VTTCue/ basically, right?
13:33
<philipj>
zcorpan_, only s/new TextTrackCue/new VTTCue/ actually
13:34
<philipj>
I didn't change any tests to check for the interface object, if there is any
13:37
<zcorpan_>
so it seems there are some spec changes that affect these tests. do you want me to comment on that in this review?
13:37
<zcorpan_>
unrelated to the VTTCue change that is
13:41
<zcorpan_>
philipj: ^
13:41
<philipj>
zcorpan_, the fact that not all of these tests are still correct?
13:42
<philipj>
I'm aware of that, and think it's better to review that when moving out of old-tests
13:42
<zcorpan_>
ok, wfm
13:44
<philipj>
zcorpan_, how are those two reviews to import more of our tests coming along?
13:48
<zcorpan_>
philipj: you mean https://critic.hoppipolla.co.uk/r/307 and https://critic.hoppipolla.co.uk/r/74 ?
13:48
<philipj>
zcorpan_, I think those are the ones you showed me last time, yes
13:48
<zcorpan_>
not much has happened
13:50
<zcorpan_>
i think the wpt-serve thing makes changes to the files in 74 which makes it more interesting
13:51
<jgraham>
zcorpan_: You should forward critic feature requests to jl :)
13:51
<zcorpan_>
yeah i can file a bug on critic
14:33
<philipj>
zcorpan_, https://github.com/w3c/web-platform-tests/pull/430 is also for you
14:34
<philipj>
(I'm preparing to re-import some of the track tests and want them to not be so ugly first)
14:35
<zcorpan_>
hmmm. i'm a bit worried about conflicts with existing PRs there
14:37
<zcorpan_>
like wpt-serve touches dnd tests iirc
14:38
<zcorpan_>
i'll hand it over to jgraham
14:41
<zcorpan_>
now time to sleep
14:41
<philipj>
oh right, you're at TPAC :)
14:41
<zcorpan_>
yep
14:42
<zcorpan_>
i think i'll do some squatting in the morning
14:43
<jgraham>
zcorpan_: I am hoping that the merge conflicts won't be too hard to sort out
14:43
<philipj>
if it's an issue, I can revert the dnd changes, it's just the media stuff I really care about
14:43
<jgraham>
Since the demantics of the test don't change it should be obvious what the right side of the diff to take is
14:44
<jgraham>
s/diff/merge/
14:44
<zcorpan_>
yeah, ok
14:44
<zcorpan_>
*poof*
20:27
<jensnockert>
Domenic_: Good work on the Streams API btw.
20:27
<Domenic_>
jensnockert: thank you! Great comment BTW.
20:28
<jensnockert>
Thanks.
21:03
<jensnockert>
Domenic_: What I find most unclear is, if I write something to an 'identity' pipe, can it return the same object on the other end? Or is it always a copy?
21:05
<jensnockert>
For example, a tee pipe, is the same object returned on all endpoints?
21:12
<Domenic_>
jensnockert: nothing would stop you from returning the same object.
21:12
<jensnockert>
Cool.
21:13
<Domenic_>
I think the built-in tee pipe would not do any sort of clone, although you could imagine someone building a StructuredCloneTeeStream or something
21:13
<jensnockert>
Yeah, but anything that is up to the implementer is cool.
21:14
<jensnockert>
Err… programmer.
21:15
<jensnockert>
Whatever the word is for 'The person implementing the custom stream type'
23:14
<WeirdAl>
streams API?
23:15
<WeirdAl>
can someone give me a link?
23:36
<smaug____>
WeirdAl: https://dvcs.w3.org/hg/streams-api/raw-file/tip/preview.html
23:37
<smaug____>
(looks like that API could use events, not promises, and be simpler for the user that way)
23:45
<WeirdAl>
ok, I need to read this spec :)
23:49
<Hixie_>
anyone got IE10+ ?
23:50
<Hixie_>
what's the log say for http://software.hixie.ch/utilities/js/live-dom-viewer/?saved=2632 ?
23:57
<smaug____>
10 says BackCompat, document has no title
23:57
<smaug____>
(using http://www.browserstack.com)