| 08:10 | <kennyluck> | Any Opera folk here? |
| 08:30 | <Ms2ger> | kennyluck, fwiw, I think the fail-if-unsupported requirement is silly |
| 08:34 | <kennyluck> | Ms2ger, someone at the event suggests we mimic border-radius with <canvas> feature or something I just can't remember... |
| 08:34 | <Ms2ger> | Ugh |
| 08:34 | <kennyluck> | But right, it's silly. |
| 10:57 | <gsnedders> | FWIW, though he's gone now, I think fail-if-unsupported is nice-to-have, but if not easily possible not a hard requirement. |
| 11:16 | <annevk> | GPHemsley: btw, please ask AryehGregor for access to the wiki |
| 11:17 | <annevk> | GPHemsley: be totally cool if you update it |
| 11:53 | AryehGregor | isn't sure he remembers how to give access to the wiki |
| 11:53 | <jgraham> | I wonder if tests should come with a priority label, like bugs |
| 11:54 | <jgraham> | So something testing a minor edge condition that doesn't matter in practice could be a P5, but a test to check a feature works at all would be a P1 |
| 11:55 | <jgraham> | Seems like a lot of work and possibly the source of fights, but much more useful than a setup where all tests are considered essential |
| 11:57 | <jgraham> | And kind of reenforces the idea that interoperability is a continuum |
| 11:58 | <Velmont> | Yes. I've often thought about that. I sometimes feel a bit reluctant to actually write some of the tests and/or upload them because they're checking super stupid stuff like ordering of exceptions that probably noone cares much about anyway. |
| 11:58 | <Velmont> | When some of the other failures are rather severe, I mean :-) |
| 12:00 | <jgraham> | I guess one problem is that it's possible to fail a P1 test for a P3 reason e.g. a security test failing sounds like a P1, but if the problem is "throws the wrong exception name", that's a tiny problem compared to "doesn't throw" |
| 12:01 | <Velmont> | Yea. So for e.g. the CORS tests, they were using XHR and checking for NETWORK_ERR. But when we made the feature for assert_throws to just check for throwing I changed them. |
| 12:02 | <Velmont> | Because almost all the tests were failing on gecko because of that single "no we use NS_STUPID_LONG_NETWORK_ERROR instead" |
| 12:18 | <jgraham> | Well that is pretty silly |
| 12:18 | <jgraham> | I mean not using assert_throws just because a browser has bugs |
| 12:19 | <jgraham> | I can't think of a good way that one could bake this into testharness, really |
| 12:33 | <SimonSapin> | TabAtkins: do we have something like "pull requests" for specs? I want to make editorial changes to css3-values (have the definition for dpi say "CSS inch" instead of just "inch".) I guess just making the edit myself would be less work for you, but I still want to get some review before pushing to the ED. |
| 12:36 | <jgraham> | For the specs that are on github (not that one, I guess), we have something very like pull requests :) |
| 12:40 | <SimonSapin> | jgraham: yes, I can imagine that :) |
| 12:40 | <SimonSapin> | but the CSSWG uses a private mercurial server http://dvcs.w3.org/hg/csswg/ |
| 12:41 | <jgraham> | Yeah, I know |
| 12:41 | <jgraham> | I suspect that filing a bug and making a patch is the right thing to do |
| 12:42 | <SimonSapin> | we use the mailing list more that the bug tracker, but I guess I could send patches by email |
| 12:42 | <jgraham> | Could work |
| 12:42 | <jgraham> | I know the CSSWG at least has components in the bugtracker these days |
| 12:43 | <jgraham> | I'm not sure what they are actually used for |
| 12:43 | <SimonSapin> | I would love to use github for this but I guess it’s not gonna happen, having just switched to hg and all |
| 12:43 | <jgraham> | Well there are noises about moving more stuff to github |
| 12:44 | <jgraham> | Not necessarily for CSSWG though, since plinss is quite oppsoed to it |
| 12:44 | <jgraham> | *opposed |
| 12:44 | <SimonSapin> | do you know why he is opposed? |
| 12:45 | <AryehGregor> | jgraham, if conformance tests are ways for implementers to spot bugs, rather than ways for them to brag about how they implement the standards, prioritization doesn't seem necessary. |
| 12:45 | <AryehGregor> | If they're only to brag, then just don't test things that aren't so important. |
| 12:45 | <AryehGregor> | The problem is when the two goals are conflated. |
| 12:46 | <jgraham> | SimonSapin: I can't actually find confirmation he is opposed. So maybe I am misstating his position |
| 12:46 | <jgraham> | So now I will stop. But I thought someone, possibly him, thought that git would be too confusing for normal humans |
| 12:47 | <jgraham> | AryehGregor: The problem is that those are both goals |
| 12:47 | <AryehGregor> | That's your problem, then. |
| 12:47 | <AryehGregor> | Also: yes, git is definitely too confusing for normal humans. We tend to assume spec-writers are supernormal. |
| 12:48 | <AryehGregor> | (also, they can always use hg-git, like Ms2ger does) |
| 12:48 | <jgraham> | My problem in particular, or just a problem? |
| 12:49 | <jgraham> | Anyway, it is not obvious to me that having both as goals is bad |
| 12:49 | Philip` | guesses implementers would find it quite useful for the test results to somehow indicate e.g. "these 500 test cases are failing because of bug 123456 in your browser" (so they don't have to check every single case to make sure there's not a really important unknown bug), which would be a superset of the functionality needed for prioritising failures |
| 12:50 | <AryehGregor> | I meant "That's your problem, then" as "Well, then, mister, there's yer problem right there, starin' ya in the face", not "Thankfully, that's your problem, not mine". |
| 12:50 | <AryehGregor> | Philip`, in practice, that's done by implementers reviewing the failures. |
| 12:50 | <AryehGregor> | (which I've done lots of) |
| 12:50 | <jgraham> | So one problem I currently see is that W3C tests are too often ignored |
| 12:51 | <jgraham> | Because there is no kudos associated with doing well in them |
| 12:51 | <jgraham> | Whereas crap like html5test.com gets loads of attention |
| 12:51 | <Ms2ger> | Velmont, I care about the ordering of exceptions ;) |
| 12:52 | <jgraham> | On the other hand, making W3C tests a big PR thing has really bad side effects |
| 12:52 | <jgraham> | people start submitting tests to look good or to make others look bad |
| 12:52 | <AryehGregor> | Both are fair points. |
| 12:52 | <AryehGregor> | One thing we could do is hope that people care about actual interop in practice and thus consult any high-quality test suite. |
| 12:52 | <jgraham> | e.g. by making every test fail for some browser for some tiny bug |
| 12:53 | <Ms2ger> | jgraham, also, I do think git is too confusing for humans :) |
| 12:53 | <jgraham> | (that could be hard to fix) |
| 12:53 | <jgraham> | Ms2ger: You aren't human, so what would you know? |
| 12:54 | <jgraham> | AryehGregor: I would like to believe that, but look at the amount of attention that html5test gets compared to any proper testsuite |
| 12:54 | <jgraham> | People want nice simple answers |
| 12:54 | <Ms2ger> | And it looks fancy! |
| 12:56 | <Philip`> | It seems a lot of people have never used any VCS other than git and get on perfectly fine with it, and hate being forced to use anything else, so I think the problem is just that git is too confusing for any human who has used any other VCS before |
| 12:57 | <AryehGregor> | Philip`, maybe all the normal people are scared away from VCSes in general when asked to use git. |
| 12:58 | <Philip`> | (If you've used e.g. CVS/SVN, then Hg is a totally different underlying model but still has familiar syntax and terminology, whereas Git has the different model and also has needlessly incompatible syntax and terminology) |
| 12:59 | <jgraham> | Anyway, so I think my point is that it would be nice if there was a middle ground; some way to encode the idea that not all failures are as serious as others - a fact that is known to anyone that has ever developed software - into the testsuite itself |
| 12:59 | <jgraham> | So that a browser could advertise "we pass all the P1 and P2 tests for feature X" |
| 13:00 | <jgraham> | And not be penalised for the fact that it throws the wrong exception name sometimes, causing P3 failures |
| 13:00 | <SimonSapin> | Philip`: I used svn and then hg before switching to git. Now I have developed lots of habits with git and git-related tools, and I hate being forced to use hg or svn |
| 13:01 | <jgraham> | I think git has a learning curve like / but mercurial has a learning curve like _| |
| 13:01 | <Ms2ger> | The other way around, you mean :) |
| 13:02 | <jgraham> | Ms2ger: Git has a learning curve like \? :p |
| 13:02 | <Ms2ger> | Well... yes :) |
| 13:02 | <Ms2ger> | And you have to climb it upside-down |
| 13:04 | Ms2ger | really doesn't like git |
| 13:04 | jgraham | is really surprised by this |
| 13:04 | <jgraham> | Or I would be if I didn't already know it |
| 13:07 | <jgraham> | (that wasn't sarcasm btw, I really have no idea why you find it so distasteful) |
| 13:11 | <SimonSapin> | Ms2ger: is hg-git good enough, or do you really prefer the upstream to use hg? |
| 13:13 | <Ms2ger> | I can live with hg-git, as long as upstream doesn't do anything weird :) |
| 13:13 | <SimonSapin> | what kind of weird thing? |
| 13:14 | <Ms2ger> | Anything I don't know how to deal with through my hg experience :) |
| 13:16 | <SimonSapin> | I guess that would also apply to git users with a git upstream |
| 13:16 | <SimonSapin> | or any other combination |
| 13:16 | <Ms2ger> | True |
| 14:15 | <GPHemsley> | AryehGregor: If you ever remember how to give people access to the wiki, I'll be happy to have some. :) |
| 14:15 | <Velmont> | jgraham: I am using assert_throws(null, { sync xhr }). I don't reeeaally need to specify WHAT it throws because that should be checked in XHR tests, not for CORS. |
| 14:15 | <Velmont> | jgraham: I'm just using XHR because it's easiest (and most thorough) way to test CORS. |
| 14:16 | <Velmont> | jgraham: So WHAT exception it returns is not that interesting, that it doesn't succeed is the important part (for CORS). |
| 14:16 | <GPHemsley> | Ms2ger: Maybe one day you could walk me through getting hg-git installed? When I tried it a few months ago, it wasn't working. |
| 14:16 | <Velmont> | Ms2ger: Yeeaaaaaa. Not essential though. |
| 14:17 | <Ms2ger> | GPHemsley, on what OS? |
| 14:17 | <GPHemsley> | Mac 10.6 |
| 14:17 | <Ms2ger> | Mm |
| 14:19 | <Ms2ger> | easy_install, according to http://hg-git.github.com/ |
| 14:24 | <GPHemsley> | Ms2ger: I thought I followed the instructions properly, but I got an error when I tried to use it. (Particularly, I don't think hggit was even found.) |
| 14:24 | <GPHemsley> | So either I or they missed a step somewhere. |
| 14:25 | <Ms2ger> | How did you use it? You just need to hg clone git://... |
| 14:34 | <GPHemsley> | When I add `hggit = ` to my .hgrc, it tells me it can't find the module |
| 14:35 | <Ms2ger> | Strange |
| 15:35 | <SimonSapin> | GPHemsley: make sure that you run easy_install with the same python as the one used by hg |
| 17:57 | <GPHemsley> | SimonSapin: How do I do that? |
| 18:18 | <SimonSapin> | GPHemsley: I’m not sure … Try this: head -c1 $(which hg) |
| 18:18 | <SimonSapin> | it should give you the path to the python interpreter |
| 18:19 | <GPHemsley> | -n1, but yeah, thanks |
| 18:19 | <SimonSapin> | yeah, -n1 |
| 18:19 | <GPHemsley> | ... and it is indeed not the same as `which python` |
| 18:20 | <SimonSapin> | GPHemsley: do you have easy_install in the same directory? |
| 18:21 | <GPHemsley> | easy_install is in a third directory >_< |
| 18:21 | <GPHemsley> | and it seems that `which python` is actually the older one (with regard to version number) |
| 18:21 | <GPHemsley> | I have a feeling I have a bunch of different versions of python on this computer |
| 18:21 | <SimonSapin> | how did you install hg? |
| 18:22 | <GPHemsley> | good question |
| 18:23 | <zewt> | i never touch easy_install in linux, that's the package manager's job |
| 18:23 | <SimonSapin> | This is the general direction to look into but I’m not familiar with OS X so I can’t help much more |
| 18:23 | <zewt> | no idea on osx |
| 18:24 | <GPHemsley> | SimonSapin: Looks like the current one was installed via MacPorts |
| 18:24 | <SimonSapin> | the current what? |
| 18:24 | <GPHemsley> | hg |
| 18:25 | <GPHemsley> | and I believe the python it is using |
| 18:25 | <GPHemsley> | which is not the same as `which python` |
| 18:25 | <GPHemsley> | so I'm thinking maybe I should just switch `which python` |
| 18:25 | <GPHemsley> | (how do I do that?) |
| 18:28 | <GPHemsley> | SimonSapin: http://pastebin.mozilla.org/1873683 |
| 18:54 | <GPHemsley> | oh, what a mess |
| 18:54 | <GPHemsley> | I have so many different versions of Python on this computer |
| 19:04 | <jgraham> | GPHemsley: Oh, OSX, yeah bad luck |
| 19:08 | <jgraham> | GPHemsley: Do you have easy_install in ports? (or pip for that matter) |
| 19:10 | <GPHemsley> | jgraham: Hang on, I'm reinstalling stuff |
| 19:22 | <annevk> | GPHemsley: you could also just install the GitHub app |
| 19:22 | <annevk> | I used command line thus far, but that app is quite convenient |
| 19:43 | <GPHemsley> | w00t |
| 19:43 | <GPHemsley> | got it working |
| 19:43 | <GPHemsley> | it was a conflict between python versions |
| 19:43 | <GPHemsley> | so thanks SimonSapin, et al. |
| 19:46 | <GPHemsley> | annevk: I'll stick to the command line, now that it works again :) |
| 20:18 | <zewt> | holy hell |
| 20:18 | <zewt> | chrome now refuses to install greasemonkey scripts normally: "extensions, apps, and user scripts can only be added from the chrome web store" |
| 21:29 | <jgraham> | zewt: have fun being upsold |
| 21:30 | <zewt> | fun having features silently auto-upgraded into "go to our store!" |