07:22
<boblet>
Hey all, anyone know what’s up with Flexbox, Template Layout and Grid Positioning CSS3 modules? Are all being actively developed? Are all expected to continue?
08:34
<annevk>
jgraham, you around?
08:37
<annevk>
jgraham, function assert throws a new AssertionError, but what is catching that?
09:34
<annevk>
jgraham, your script is fucking complicated
09:34
<annevk>
jgraham, I put an hour into debugging this assert_throws madness but no luck
09:34
<annevk>
jgraham, too much functions invoking each other and all kinds of crazyness
09:36
<annevk>
jgraham, I did found some smallish bugs
09:40
<annevk>
btw, rather than <?xbl?> maybe we should have <html binding=""> instead?
09:41
<annevk>
(or maybe name it bindings as it presumably takes a space-separated list)
09:42
<annevk>
jgraham, we should have some more discussion on the framework
10:06
<annevk>
jgraham, http://bitbucket.org/ms2ger/web-dom-core/changeset/d5e2a858a519 has the minor fixes
10:22
<annevk>
prolly away most of the day -- trying to get back to Oslo
10:22
<jgraham>
Wow I missed anne by || that much
11:08
<jgraham>
Oh it was a simple bug
12:18
<jgraham>
I pushed the fix to W3C repository BTW
14:21
<espadrine>
I'm confused about "new Image(undefined,undefined);"
14:21
<espadrine>
Firefox, Chrome, Opera all do something different
14:21
<espadrine>
Firefox fires an error, Chrome turns undefined into the number 0, Opera treats this as new Image().
14:21
<espadrine>
Which one is right?
15:14
<Damark>
Hey
15:14
<Damark>
I want help
15:15
<gsnedders>
espadrine: I think, if I'm understanding WebIDL correctly, Chrome is correct
15:15
<gsnedders>
Damark: Then ask.
15:16
<Damark>
i need shells please
15:18
<jgraham>
Damark: I assume you are not talking about Mario Kart, but can't really work out what you might be talking about
15:19
<jgraham>
annevk: see logs
15:19
<annevk>
sure
15:19
<annevk>
might grab some food first
15:19
<jgraham>
annevk: np
15:19
<annevk>
just arrived at heathrow
15:19
<jgraham>
annevk: Fix pushed to W3C repository
15:19
<jgraham>
ah
15:19
<annevk>
k
15:19
<jgraham>
Simple bug that I should find a generic solution for
15:20
<annevk>
did you see my fixes?
15:20
<jgraham>
The template stuff requires explicit string conversions everywhere at the moment
15:20
<jgraham>
Or it just dies in confusing ways
15:20
<annevk>
i figured something went wrong with that
15:20
<annevk>
but I couldn't find it
15:20
<annevk>
it's a maze of functions
15:20
<jgraham>
I fixed this one instance but it whould probably be fixed in more places
15:21
<annevk>
i wonder if we can simplify it somehow
15:21
<annevk>
less elegant but more clear or some such :)
15:23
<jgraham>
Maybe :)
15:24
<annevk>
i'm mostly worried about you being the only person that can fix "simple bugs" :)
15:28
<jgraham>
So am I :)
15:39
<annevk>
jgraham, the error message is still incorrect btw
15:39
<annevk>
oh maybe not
15:39
<annevk>
never mind
15:46
<boblet>
annevk: yt?
15:46
<annevk>
yup
15:46
<boblet>
re: CSS3 layout modules, I’m wondering how they’re going — are all being actively developed? Are all expected to continue?
15:47
<annevk>
flexbox seems like it will happen in some form
15:47
<annevk>
not sure about any of the others
15:48
<boblet>
I guess having implementation is a big help
15:49
<annevk>
sure
15:50
<boblet>
I perceive flexbox as being good for web app interfaces, would that be right? also would grid positioning & template layout be better (in certain cases) for more web page content layout
15:50
<boblet>
?
15:54
<annevk>
flexbox can be used for web page layout as well
15:55
<annevk>
I haven't studied layout models enough to give a really good answer, though according to Tab (iirc) most should be doable with flexbox
15:57
<boblet>
annevk: ok. yeah I know flexbox should work for everything, but was wondering if setting up a page layout for content would be easier with the others, or what the different strengths of each one were. Maybe I should ping Tab :)
15:57
<boblet>
annevk: thanks for your help yo
16:15
<annevk>
jgraham, ms2ger had this feature request: http://bitbucket.org/ms2ger/web-dom-core/issue/1/make-assert_throws-do-an-instanceof-domexception
16:15
<annevk>
jgraham, of course that does not work for ECMAScript exceptions
16:18
<jgraham>
Sure but one could assume that anything with a code is a DOMException
16:19
<jgraham>
So it should be rather easy to implement I think
16:21
<annevk>
what might also be nice is adding a check for e.name
16:21
<annevk>
e.name == arg basically
16:32
<annevk>
wait, why does getElementsByTagName match on local name for HTML elements? I doubt that's true
18:24
gsnedders
wonders if you could write a generic runner for JS tests (for HTML5) using Watir
18:24
gsnedders
wonders if Watir works with IE9/Fx4/WebKit ToT
18:34
<jarib>
gsnedders: basically just opening a list of URLs?
18:35
<jarib>
i'd recommend looking at watir-webdriver instead, but i'm biased :)
18:35
<jgraham>
gsnedders: Why would you bother?
18:35
<jgraham>
To avoid the problems of <iframe>s
18:35
<jgraham>
?
19:17
<gsnedders>
jgraham: Process everything and come up with something listing tests that fail in all browsers too, for example
19:17
<gsnedders>
jarib: And read the DOM
19:20
<jarib>
gsnedders: watir doesn't support any webkit browsers, and its firefox support is pretty shaky (especially now when Jssh is being EOLed)
19:21
<gsnedders>
jarib: Ah, I thought the status of both of those was better than that
19:25
<jarib>
webdriver supports IE (though I haven't tested in IE9), chrome (though a rewritten driver just landed in the chromium tree), and firefox (Fx4 support just landed in trunk). a safari driver is under development
19:25
<jarib>
you might want to look at the selenium-webdriver gem to use that from ruby directly
19:25
<gsnedders>
webdriver can't actually access the DOM, though, right?
19:26
<jarib>
watir-webdriver is a watir wrapper around the selenium-webdriver gem
19:26
<jarib>
it can
19:26
<jarib>
the watir api is just more high level
19:28
<jarib>
i maintain both of those gems, feel free to pm/email me if you need any assistance
19:40
<Lachy>
Hixie, Requiem 1.9.7 has been released now. It should work for iTunes 9.2.1. Won't work for iTunes 10, if you've upgraded already
19:44
<gsnedders>
jarib: Assistance is only useful if I reach the point of bothering to automate them in the general case :P
19:49
<jarib>
yep. don't think i can help you reach that point :)
19:50
<jarib>
out of curiosity though, what tests are you referring to?
19:50
<gsnedders>
http://dvcs.w3.org/hg/html/
19:52
<jarib>
thanks
19:54
AryehGregor
discovers that fall classes start in two days, drat
19:55
<gsnedders>
heh
19:55
<gsnedders>
I have another two weeks till that panic
19:55
<gsnedders>
But then it'll be like, "uh, studying? I stopped doing that before when I realized I hated that. Why am I here?"
19:58
<AryehGregor>
I was hoping to finish up my reflection test cases and submit them to the W3C before school started.
19:58
<AryehGregor>
Well, with luck I'll be able to do that before I have too much coursework to do.
19:59
<AryehGregor>
(not to mention all the other things I planned to do . . .)
20:11
<Hixie>
Lachy: k
21:03
<jgraham>
gsnedders: Except for special cases loading tests in an iframe should work rather well, no?
21:04
<jgraham>
The special cases being things that require a top-level browsing context
21:04
<jgraham>
In fact I was somewhat under the impression that that was what the Microsoft thing already did
21:04
<jgraham>
Although I confess I haven't really looked at it hard
21:07
<gsnedders>
jgraham: But then you have to then aggregate the results from different browsers separately
21:07
<jgraham>
gsnedders: That doesn't seem like a problem, really
21:08
<jgraham>
This isn't like the CSS case where you have thousands of manual tests
21:08
<gsnedders>
jgraham: Indeed
21:08
<gsnedders>
But it's still more work than it need be
21:08
<jgraham>
in the ideal case you open the browser, start the testsuite, it generates a json file of all the results for you
21:08
gsnedders
tries scanning some of his old 35mm film
21:08
<AryehGregor>
CSS has thousands of manual tests? Yikes.
21:09
<jgraham>
CSS2.1 does
21:09
<gsnedders>
AryehGregor: Almost all of them are visual tests
21:09
<AryehGregor>
Oh, so you can just skim through to see what's green?
21:09
<jgraham>
For CSS 3 they will mostly use reftests which can be auomated more easilly
21:09
<AryehGregor>
(unless you're color-blind, oh well)
21:09
<AryehGregor>
Yay reftests.
21:09
<gsnedders>
AryehGregor: Well, yeha. But it's really way more time-consuming than that.
21:10
<gsnedders>
AryehGregor: MS guy in the F2F said it took 16 hours for all the HTML copies for one browser. There are XHTML copies of all the tests too.
21:10
<jgraham>
although browsers don't generally expose a consistent API for taking a screenshot so it is non-trivial to make a reftest harness
21:10
<gsnedders>
So the full testsuite would take 32 hours to run.
21:10
<AryehGregor>
jgraham, but that could be standardized pretty easily.
21:10
<AryehGregor>
At least in privileged contexts.
21:10
<jgraham>
that would be good
21:11
gsnedders
wants to get around 800 tests in the CSS 2.1 testsuite coverted to reftests in the next week
21:13
<gsnedders>
(there are around 9k)
21:13
<gsnedders>
(source files, that is)
21:13
<gsnedders>
(then there are generated from that 9k HTML copies)
21:13
<Philip`_>
AryehGregor: That sounds like it probably wouldn't help in the case of e.g. mobile browsers (when you're not a developer of that browser, e.g. you're someone who just wants to check standards compliance)
21:13
<gsnedders>
(but all that matters is the number of source files)
21:14
<Philip`_>
AryehGregor: or at least it doesn't sound pretty easy to standardise and implement in that context
21:15
<AryehGregor>
Philip`_, the obvious thing would be to just do a screenshot API that only works if nothing in the page is cross-domain. Is there any reason you know of not to do that?
21:15
<AryehGregor>
It would probably have other applications too.
21:16
<Philip`_>
AryehGregor: That seems unsafe for non-privileged code, because you could do stuff like read vlink colours
21:16
<AryehGregor>
Oh, bleh.
21:17
<Philip`_>
and the text in <input type=file>s etc
21:17
<AryehGregor>
I guess so.
21:17
<AryehGregor>
Hmm.
21:18
<AryehGregor>
Maybe you could have a command-line flag where the browser will retrieve a URL, execute it non-interactively with a fresh profile with the screenshot feature enabled, and spit out a result.
21:18
<Philip`_>
Same issues as with exposing Mozilla's context2d.drawWindow to web content
21:19
<AryehGregor>
If you can get the user to type an arbitrary command line, you can pwn them anyway, right?
21:19
<Philip`_>
Does a command-line flag help if you want to test the browser on your mobile phone which doesn't have a command line?
21:20
<AryehGregor>
Mobile browsers are so hopelessly broken anyway that they can be set aside for now.
21:20
<AryehGregor>
Perfect is the enemy of good, you know.
21:23
<Philip`_>
It could be e.g. Opera on Wii, which probably isn't hopelessly broken, where some user wants to run the test suite to see how well they can rely on various features
21:23
<Philip`_>
and reftests would be less useful than some other pure JS style of testing
21:24
<gsnedders>
Philip`_: It does use Presto 1, so I expect it isn't exactly bug-free wrt stuff fixed in the past few years
21:24
<AryehGregor>
It doesn't have to be command-line. Every browser just has to provide some reasonably secure way to expose the functionality.
21:24
<AryehGregor>
You could try exposing this to non-privileged JS, but it would be a minefield.
21:25
<gsnedders>
AryehGregor: Presto 2.1 and above with Scope protocol?
21:25
<gsnedders>
(or maybe the screenshot API was only added later)
21:25
<AryehGregor>
Maybe?
21:26
<gsnedders>
STP/1 certainly supports screenshots being taken
21:26
<gsnedders>
http://dragonfly.opera.com/app/scope-interface/services/Exec/Exec_2_0.html#setupscreenwatcher
22:21
<TheLifelessOne>
Soo