15:03:07 <RRSAgent> RRSAgent has joined #htmlt
15:03:07 <RRSAgent> logging to http://www.w3.org/2012/08/14-htmlt-irc
15:03:37 <krisk> Here is the agenda http://lists.w3.org/Archives/Public/public-html-testsuite/2012Aug/0003.html
15:03:59 <krisk> Though it might be good to talk about github mirror as well
15:04:29 <krisk> This was setup a while back by jgraham (it's just a mirror, no submits)
15:05:27 <krisk> Agenda Item #1 Bugs on Approved Tests
15:05:48 <krisk> Here is the tinyurl link - http://tinyurl.com/6mvghxx
15:06:09 <krisk> with tmpsantos we should look at his bug/update as well
15:06:41 <krisk> see -> https://www.w3.org/Bugs/Public/show_bug.cgi?id=18522
15:08:15 <krisk> Let's look at that update first
15:11:26 <krisk> Indeed it looks like a typo
15:13:14 <krisk> Though no one passes this test looking at other browsers...
15:13:59 <tmpsantos> krisk: it passes now at webkit :)
15:14:12 <krisk> Maybe we don't actually understand the intent of this test?
15:14:32 <tmpsantos> the canvas element at this test is not attached to the DOM
15:14:54 <tmpsantos> 1 sec
15:15:08 <tmpsantos> let me open the test here again
15:15:50 <tmpsantos> so when using relative unit at this particular case
15:16:03 <krisk> I'm looking at the w3c spec for why this should change from the set (ctx2.font = 1000% serif) to the get ctx2.font = 100px serif)
15:16:09 <tmpsantos> it should be applied using the defaults as reference
15:16:47 <tmpsantos> ctx is dangling at this test case
15:17:08 <tmpsantos> it is just there because of the way the test runner/harness was designed
15:18:58 <krisk> I'm looking at http://dev.w3.org/html5/2dcontext/#canvastext
15:20:31 <tmpsantos> this is the important bit
15:20:32 <tmpsantos> "When the 'font-weight' component is set to the relative values 'bolder' and 'lighter', these must be interpreted relative to the computed value of the 'font-weight' property of the corresponding canvas element at the time that the attribute is set. If the computed values are undefined for a particular case (e.g. because the canvas element is not in a Document), then the relative keywords must be interpreted relative to the normal-weight 10px sans-seri
15:20:33 <tmpsantos> f default."
15:21:09 <tmpsantos> sorry, actually just the last sentence
15:23:38 <krisk> Yep I see that...
15:24:52 <krisk> That looks correct
15:24:59 <krisk> does anyone else object?
15:26:03 <krisk> jgraham have you looked at this bug/fix?
15:26:17 <krisk> ms2ger looked already according to bugzilla
15:28:18 <jgraham> krisk: No, I didn't look
15:29:10 <krisk> It looks correct from my point of view (though no interop in browsers today, though a fix exists in some version of webkit)
15:29:45 <krisk> We have one more bug as well https://www.w3.org/Bugs/Public/show_bug.cgi?id=18505
15:30:02 <krisk> Let's look at this while james takes a peek at 18522
15:30:35 <jgraham> krisk: <canvas> isn't my specialist subject, but I trust Ms2ger in this instance :)
15:30:50 <Ms2ger> Uh-oh
15:31:16 <Ms2ger> We should turn tmpsantos into our <canvas> specialist :)
15:31:37 <tmpsantos> Ms2ger: Uh-oh
15:31:58 <krisk> Cool
15:33:26 <krisk> 18505 looks like a bug
15:36:34 <krisk> ms2ger did you have a patch or fix?
15:37:04 <Ms2ger> No, I haven't looked too closely
15:37:05 <tmpsantos> Ms2ger: is it because pause == true when the video is loading?
15:37:15 <krisk> I suspect the test needs to see that the media has loaded and is playing before calling pause
15:37:30 <krisk> Note that all browsers don't 'play' the video
15:37:36 <Ms2ger> Stepping through the Gecko code, paused did seem to be true
15:37:54 <Ms2ger> I ran out of time before figuring out if that matches the spec
15:39:21 <Ms2ger> (I believe a couple more of those tests timed out, but I haven't really looked at those)
15:40:12 <krisk> I added this to the bug
15:41:02 <krisk> let's move on to the next agenda item
15:41:07 <krisk> new test submissions
15:42:19 <krisk> see http://dvcs.w3.org/hg/html/shortlog
15:42:55 <tmpsantos> Ms2ger: we have a slightly modified version of the same test case on webkit
15:43:22 <Ms2ger> That doesn't surprise me, it was a Google submission
15:43:39 <tmpsantos> it calls explicitly play() and doesn't use autoplay
15:43:54 <krisk> two meetings ago we were reviewing ms2ger canvas updates
15:44:00 <krisk> and stopped at http://dvcs.w3.org/hg/html/rev/7145549fcecf
15:44:07 <krisk> let's continue with this rev
15:44:47 <krisk> This fixes a bunch of tests to match the WebIDL spec
15:44:57 <krisk> ...exceptions
15:48:15 <krisk> I'm OK with this change
15:50:47 <krisk> any disagreement?
15:51:50 <krisk> Now looking at http://dvcs.w3.org/hg/html/rev/b0caf9306a9f
15:51:54 <krisk> (the next rev)
15:52:56 <krisk> It looks like the test (and others are broken - timeout???)
15:52:57 <krisk> http://www.w3c-test.org/html/tests/submission/PhilipTaylor/canvas/2d.gradient.object.invalidcolour.html
15:53:50 <krisk> The approved test is at http://www.w3c-test.org/html/tests/approved/canvas/2d.gradient.object.invalidcolour.html
15:54:41 <krisk> ms2ger are you aware of this class of regression in the submitted canvas tests?
15:55:40 <Ms2ger> Oh, right
15:55:55 <Ms2ger> I need to look at that issue some more
15:56:06 <krisk> I see this timeout in alot more tests as well...
15:56:20 <krisk> can you fix this or revert the change that caused this until this can be fixed?
15:56:31 <Ms2ger> It happens if the only tests that run are assert_throws ones, and they all pass
15:56:47 <krisk> ..or submit a fix?
15:57:07 <Ms2ger> I'll try to look at it soon
15:57:50 <krisk> If it's going to be a while to get a true fix in place can you just revert the change until the fix is more complete?
15:58:40 <krisk> I can't tell right now if it's a simple revert or not..
15:59:05 <krisk> ms2ger do you think this is a simple revert?
16:00:01 <Ms2ger> Not sure, but I don't think it'll take long to fix
16:00:41 <krisk> OK - if it's going to be a while and you don't have time send an email to the list so someone else can look at addressing
16:01:37 <krisk> It's after the scheduled meeting time...
16:02:29 <krisk> Though I did want to comment on the public-test-infra@w3.org thread about 'Request to move the test repositories to Github'
16:02:56 <krisk> jgraham did this for the HTML suite (mirrored) already a few months back
16:04:39 <krisk> see -> https://github.com/w3c/html-testsuite
16:05:46 <jgraham> (MikeSmith/W3C set up the infrastructure to make the updates automatic)
16:06:41 <krisk> Others exists as well see https://github.com/w3c/
16:08:07 <krisk> One easy change would be to make the default Hg page to have alot more information
16:08:53 <jgraham> Change for what purpose?
16:08:57 <krisk> e.g http://w3c-test.org/html/ looks pretty bare
16:09:18 <krisk> and same with the github page(s)
16:09:23 <jgraham> Yes, the whole w3c-test.org domain needs to be improved a lot
16:09:41 <jgraham> But that is differently motivated to the use of github
16:10:24 <jgraham> Github is motivated by the fact that there is a large community that is used to git and github in particular compared to a small one that uses hg and the W3C site
16:10:59 <krisk> I'm just looking at the information at https://github.com/w3c/ and the hg pages
16:11:18 <jgraham> Many web-devs experience github as a de-facto part of their profession as basic git(hub) usage is needed to access many common libraries
16:12:02 <krisk> I get that it' just doesn't have any simple information on how to actually participate and submit tests
16:12:02 <jgraham> So moving to the venue where the people are increases our exposure and makes it more likely that people that aren't paid to work on this full-time will contribute
16:12:38 <jgraham> Agreed, fixing w3c-test.org to be user-friendly is something seperate that I have discussed with plh and others
16:13:05 <jgraham> There are no technical barriers, but someone with the right skills needs to step up and do the work
16:13:39 <jgraham> (that excludes me, unfortunatly, since "right skills" in this case involves some graphic design)
16:14:07 <krisk> that seems pretty key - since even the testing-how-to takes you to a page with html markup :)
16:14:11 <krisk> ...see https://github.com/w3c/testing-how-to/blob/master/Overview.html
16:14:18 <tmpsantos> krisk, jgraham: feedback from a newcomer at w3c test (me)... git would be much better than hg
16:14:42 <tmpsantos> technical merits put aside, I don't know a single hg user
16:15:01 <krisk> I'm actually indifferent with either git or hg...
16:15:22 <jgraham> Sure, over time I have come to prefer git to hg
16:16:03 <jgraham> There are some technical difficulties having *both* the current hg repo and the git repo and having them both be writable though
16:16:24 <krisk> Though what ever we use needs to have some type of control for commits 
16:16:55 <jgraham> Yes. As discussed on list that does seem possible to solve with a technical solution
16:16:56 <krisk> I'm not sure how the w3c would manage accounts like they do today with cvs/hg
16:17:59 <krisk> I also don't want to have two spots to submit and create a bunch of merge work...
16:18:14 <jgraham> No, having lots of merges is an explicit non-goal
16:18:39 <jgraham> There either needs to only be one placve to submit, or they need to be kept in sync automatically
16:18:48 <krisk> agree
16:19:29 <krisk> shall we adjourn?
16:20:08 <jgraham> (for the committer thing, the easiest thing would probably be to have a github id associated with a W3C account, let all those people be committers to the repo, and set up the repo to reject commits authored (not committed) by people who are not also committers)
16:20:35 <jgraham> Sure
16:20:39 <krisk> Yeah I don't think it would be problem that could not be solved
16:22:35 <krisk> It looks like we (at least in HTML) could do this....
16:22:36 <krisk> by...
16:23:28 <krisk> A) Getting the w3c to set up the account mapping and create a process to manage this mapping of w3c account -> github account
16:23:44 <krisk> B) Make Hg Read-only (just syncs)
16:24:08 <krisk> C) Making sure that tests sync from git -> hg -> www.w3c-test.org/html
16:24:32 <jgraham> Yes, that would be perfect I think
16:25:22 <krisk> Ok let's adjourn
16:25:50 <krisk> ms2ger please let the list know when you have fixed the canvas test harness that makes a bunch of tests timeout.
16:25:51 <krisk> RRSAgent, make logs public
16:25:57 <Ms2ger> Will do
16:26:50 <krisk> meeting adjourned
16:27:00 <krisk> thanks for participating tmpsantos
17:08:48 <Ms2ger> RRSAgent, please excuse us
17:08:48 <RRSAgent> I see no action items