Options:
- # Session Start: Thu Aug 22 00:00:00 2013
- # Session Ident: #testing
- # [00:00] * Quits: rhauck (~Adium@public.cloak) (Ping timeout: 180 seconds)
- # [00:15] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [00:16] * Joins: zcorpan (~zcorpan@public.cloak)
- # [00:23] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [00:25] * Quits: tobie (tobie@public.cloak) (Client closed connection)
- # [00:26] * Joins: tobie (tobie@public.cloak)
- # [00:46] * Joins: zcorpan (~zcorpan@public.cloak)
- # [00:58] * Quits: rhauck1 (~Adium@public.cloak) ("Leaving.")
- # [01:01] * Joins: rhauck (~Adium@public.cloak)
- # [01:12] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [01:26] * heycam|away is now known as heycam
- # [01:37] * Joins: zcorpan (~zcorpan@public.cloak)
- # [01:47] * Quits: kershaw (~kkershaw@public.cloak) (Ping timeout: 180 seconds)
- # [02:08] * Quits: tobie (tobie@public.cloak)
- # [02:20] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [02:30] * Quits: jhammel (~jhammel@public.cloak) ("leaving")
- # [02:44] * Joins: zcorpan (~zcorpan@public.cloak)
- # [02:54] * Joins: gitbot (~gitbot@public.cloak)
- # [02:54] -gitbot:#testing- [web-platform-tests] zqzhang synchronize pull request #296: Added idlharness test for animation-timing (master...submission/zqzhang/animation-timing-idlharness) https://github.com/w3c/web-platform-tests/pull/296
- # [02:54] * Parts: gitbot (~gitbot@public.cloak) (gitbot)
- # [03:35] * Joins: gitbot (~gitbot@public.cloak)
- # [03:35] -gitbot:#testing- [web-platform-tests] ibelem synchronize pull request #280: draggable_attribute, dropzone_attribute and dragevent (master...submission/ibelem/draganddrop) https://github.com/w3c/web-platform-tests/pull/280
- # [03:35] * Parts: gitbot (~gitbot@public.cloak) (gitbot)
- # [04:23] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [04:33] * heycam is now known as heycam|away
- # [04:54] * Quits: rhauck (~Adium@public.cloak) ("Leaving.")
- # [04:55] * Joins: rhauck (~Adium@public.cloak)
- # [05:02] * Quits: rhauck (~Adium@public.cloak) (Ping timeout: 180 seconds)
- # [05:11] * Joins: kershaw (~kkershaw@public.cloak)
- # [05:11] * heycam|away is now known as heycam
- # [05:29] * Joins: MikeSmith_ (~MikeSmith@public.cloak)
- # [05:30] * Quits: MikeSmith_ (~MikeSmith@public.cloak) ("Make the road by walking. (B-side: Tired of Fighting)")
- # [07:15] * Quits: kershaw (~kkershaw@public.cloak) (Ping timeout: 180 seconds)
- # [07:43] * Joins: zcorpan (~zcorpan@public.cloak)
- # [07:44] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [07:44] * Joins: zcorpan (~zcorpan@public.cloak)
- # [08:08] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [08:08] * Joins: zcorpan (~zcorpan@public.cloak)
- # [08:15] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [08:19] * Joins: Ms2ger (~Ms2ger@public.cloak)
- # [08:21] * Joins: tobie (tobie@public.cloak)
- # [09:09] * Joins: gitbot (~gitbot@public.cloak)
- # [09:09] -gitbot:#testing- [web-platform-tests] ibelem synchronize pull request #280: draggable_attribute, dropzone_attribute and dragevent (master...submission/ibelem/draganddrop) https://github.com/w3c/web-platform-tests/pull/280
- # [09:09] * Parts: gitbot (~gitbot@public.cloak) (gitbot)
- # [09:31] * Joins: gitbot (~gitbot@public.cloak)
- # [09:31] -gitbot:#testing- [web-platform-tests] ibelem synchronize pull request #280: draggable_attribute, dropzone_attribute and dragevent (master...submission/ibelem/draganddrop) https://github.com/w3c/web-platform-tests/pull/280
- # [09:31] * Parts: gitbot (~gitbot@public.cloak) (gitbot)
- # [09:36] * Joins: gitbot (~gitbot@public.cloak)
- # [09:36] -gitbot:#testing- [web-platform-tests] ibelem synchronize pull request #280: draggable_attribute, dropzone_attribute and dragevent (master...submission/ibelem/draganddrop) https://github.com/w3c/web-platform-tests/pull/280
- # [09:36] * Parts: gitbot (~gitbot@public.cloak) (gitbot)
- # [09:48] * Quits: Lachy (~Lachy@public.cloak) ("Computer has gone to sleep.")
- # [09:52] * Joins: Lachy (~Lachy@public.cloak)
- # [09:53] * Quits: Lachy (~Lachy@public.cloak) ("Computer has gone to sleep.")
- # [09:56] * Joins: zcorpan (~zcorpan@public.cloak)
- # [10:22] * Joins: Lachy (~Lachy@public.cloak)
- # [10:49] * heycam is now known as heycam|away
- # [10:58] * Joins: AutomatedTester (~AutomatedTester@public.cloak)
- # [13:40] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [13:40] * Joins: zcorpan (~zcorpan@public.cloak)
- # [13:41] * Joins: Automate_ (~AutomatedTester@public.cloak)
- # [13:47] * Quits: AutomatedTester (~AutomatedTester@public.cloak) (Ping timeout: 180 seconds)
- # [13:47] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [13:48] * Quits: Automate_ (~AutomatedTester@public.cloak) (Ping timeout: 180 seconds)
- # [14:00] * Joins: darobin (rberjon@public.cloak)
- # [14:12] * Joins: zcorpan (~zcorpan@public.cloak)
- # [14:40] * Joins: dom (dom@public.cloak)
- # [14:58] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [15:22] * Joins: zcorpan (~zcorpan@public.cloak)
- # [15:28] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [15:31] * Joins: zcorpan (~zcorpan@public.cloak)
- # [15:41] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [15:41] * Joins: zcorpan (~zcorpan@public.cloak)
- # [15:48] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [16:10] * Quits: dom (dom@public.cloak) ("")
- # [16:12] * Joins: zcorpan (~zcorpan@public.cloak)
- # [16:59] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [17:00] * Joins: zcorpan (~zcorpan@public.cloak)
- # [17:01] * Joins: zcorpan_ (~zcorpan@public.cloak)
- # [17:01] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [17:04] * Joins: kershaw (~kkershaw@public.cloak)
- # [17:08] * Quits: zcorpan_ (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [17:19] <jgraham> darobin: Really? You think that it would work well to require authors to copy an existing file every time they want to reuse a reference is a win compared to just pointing at an existing file?
- # [17:19] <jgraham> Uh, that sentence got lost halfway through
- # [17:19] <jgraham> But you see what I mean
- # [17:20] * darobin looks at timer — and it's a new record for jgraham jumping from email to IRC in anger!!!! The crowd goes wild.
- # [17:20] <Ms2ger> Hey darobin!
- # [17:20] <darobin> hey Ms2ger!
- # [17:20] <jgraham> Also, having to read and hash every file would be much slower than just comparing urls
- # [17:20] * jgraham isn't angry :p
- # [17:21] <darobin> jgraham: to be honest, I don't care about that specific issue :)
- # [17:21] <Ms2ger> darobin, have you had time to look at the canvas code?
- # [17:21] <darobin> we could use naming conventions that identify resources as references without them having the same name as the file they're references for
- # [17:21] <darobin> Ms2ger: I just came back from vacation
- # [17:21] <darobin> (again)
- # [17:21] <darobin> well, from a long weekend
- # [17:21] <darobin> so no
- # [17:21] <Ms2ger> You must be French :)
- # [17:21] <darobin> but I plan to, unless beaten to it
- # [17:21] <darobin> Ms2ger: you guess well :)
- # [17:25] <jgraham> darobin: That doesn't solve anything really. You still need metadata to say which ref goes with which file, which currently seems to be <meta> elements in-file. Which is too slow to read all the time. Therefore you either need a build step to generate manifests or checked-in manifests
- # [17:25] <darobin> yeah
- # [17:25] <darobin> go build step!
- # [17:25] <Ms2ger> Boo build step!
- # [17:25] <darobin> it solves the "identifying which files contain tests" problem
- # [17:25] <darobin> which is the one I care about more
- # [17:25] <darobin> linking to ref tests, well, that can use a build step
- # [17:25] <jgraham> I think my feeling about build step is "meh". It doesn't solve the "I want to specify a URL as a test" problem
- # [17:25] <darobin> I just don't want to have to see conflicts from manifests ever again
- # [17:26] <jgraham> But Mozilla is going to need some hideous thing to work out which tests are expected to fail
- # [17:26] <darobin> for the subset of tests where the mapping from URL to file is not obvious, you can specify that
- # [17:26] <darobin> oh this will require hideous stuff no matter what :)
- # [17:26] <darobin> we're just arguing about which hideous colour to paint the horror
- # [17:28] <jgraham> I guess I could live with file[name] based metadata for most things and explicit MANIFEST files to override the extracted data where required.
- # [17:28] <jgraham> And a checked in tool that builds a full manifest so that everyone agrees on what the actual data is
- # [17:28] <darobin> jgraham: to be frank about the whole thing, given that you seem to have the bandwidth to put in the work I reckon you get to pick the solution anyway
- # [17:29] <jgraham> darobin: There is some truth in that :)
- # [17:29] * Ms2ger pushes jgraham towards manifests
- # [17:29] <darobin> jgraham: yeah, I think that if we can have vernacular filename based information and only resort to manifests when needed I'd be happy, and we'd all be happier in the long run
- # [17:30] <darobin> manifests just don't play well with branches and stuff
- # [17:30] * Joins: zcorpan (~zcorpan@public.cloak)
- # [17:30] <jgraham> I have to say that I can foresee manifests causing merge issues
- # [17:30] * Quits: Lachy (~Lachy@public.cloak) ("Computer has gone to sleep.")
- # [17:30] <darobin> I wonder if it would also be possible to use a file format slightly less ad hoc than the one in manifestdestiny :)
- # [17:31] <jgraham> That aren't hard to solve, but require people to understand how to resolve merge conflicts
- # [17:31] <Ms2ger> Dunno, you only get merge issues if you have a small manifest and a lot of changes to it
- # [17:31] <darobin> jgraham: they will cause merging issues, but perhaps more annoying is the problem of how they will likely become corrupted when a frustrated merger goes through the manifest conflicts too fast
- # [17:31] <jgraham> Manifests out to merge very well
- # [17:31] <jgraham> *ought
- # [17:32] <darobin> that's not my experience :)
- # [17:32] <Ms2ger> Having a check that all files in the repo are mentioned in a manifest should be sufficient, surely
- # [17:32] <jgraham> If they are just lists the merge conflict resolution is almost always "take the change from both branches"
- # [17:34] <darobin> mmmm
- # [17:34] <jgraham> But of course it is always easy to make a mistake when merging
- # [17:34] <darobin> so maybe solve that with CI?
- # [17:34] * Ms2ger already wrote such a script once
- # [17:36] <jgraham> Ms2ger: So what's the argument in favour of explicit manifests vs some sort of generated manifest?
- # [17:37] <Ms2ger> 1. bad experiences with forcing a build step in the css tests
- # [17:37] <jgraham> I seem to recall that was bad too, but I don't remember the specific badness
- # [17:37] <Ms2ger> 2. having manifests that are only used for some tests seems unnecessarily confusing
- # [17:39] <jgraham> Re: 2. I think it's OK to have features that are for advanced users only. People like Aryeh who are writing complex, long-running, tests can deal with the confusion of having to edit a file like override.manifest
- # [17:40] <darobin> yeah I agree
- # [17:41] <Ms2ger> Also, istr that for reftests in particular, the in-test approach is a lot more verbose than a typical manifest
- # [17:42] <jgraham> It is
- # [17:42] <darobin> do you need anything more in-test than a pointer to the ref?
- # [17:42] <Ms2ger> <link rel="reference" href="...
- # [17:42] <jgraham> Well accorind to the CSSWG you should spend about half an hour writing metadata
- # [17:43] <jgraham> per test
- # [17:45] * Ms2ger wonders how to distinguish partially-manual and fully-automatic testharness tests
- # [17:45] <jgraham> (link rel=authou, rel=flags rel=assert, rel=reviewer, rel=match|mismatch
- # [17:45] <jgraham> )
- # [17:45] <jgraham> *author
- # [17:46] <jgraham> (I find rel=reviewer particularly funny. A test gets reviewed OK and then you have to change it and hope that doesn't invalidate the test)
- # [17:47] <jgraham> Ms2ger: All tests that aren't sully automatic are manual, I guess
- # [17:47] <jgraham> *fully
- # [17:47] <Ms2ger> Well, yes
- # [17:47] <Ms2ger> How do you tell?
- # [17:47] <jgraham> Well there are two conventions I know of
- # [17:47] <jgraham> foo-manual.html
- # [17:47] <jgraham> and
- # [17:47] <jgraham> non-automated/foo.html
- # [17:48] <Ms2ger> Neither of those is used in wpt, though
- # [17:48] <jgraham> I thought they were
- # [17:48] <jgraham> (not consistently)
- # [17:49] <jgraham> /workers/non-automated/
- # [17:50] <jgraham> jgraham@luna:~/develop/web-platform-tests$ find . -iname *MANUAL* | wc -l
- # [17:50] <jgraham> 22
- # [17:50] * Joins: rhauck (~Adium@public.cloak)
- # [17:52] <Ms2ger> Also -interactive, apparently
- # [17:53] <Ms2ger> A bunch of the foreign content parsing tests from MS are unmarked
- # [17:54] <Ms2ger> And a few nested-document-write ones
- # [17:54] <jgraham> Yeah, not surprising given that there is no strong convention
- # [17:54] <jgraham> Why are those manual tests?
- # [17:54] <jgraham> I guess the no-manifest thing doesn't really solve the timeout problem
- # [17:55] <jgraham> Quite a few tests need unusually long timeouts
- # [17:55] <jgraham> I'm not super-excited about having to run the files to find this out.
- # [17:58] * Joins: zcorpan_ (~zcorpan@public.cloak)
- # [17:58] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [17:59] * Ms2ger is not super-excited about running files in general
- # [17:59] <jgraham> I quite like running them to get the test results out
- # [17:59] <jgraham> It always struck me as useful
- # [18:00] <Ms2ger> Meh
- # [18:00] * Ms2ger gets dinner
- # [18:00] <darobin> well there's a file naming convention for everything
- # [18:01] <darobin> you could have foo-seems-to-go-on-forever.html
- # [18:01] * Joins: Lachy (~Lachy@public.cloak)
- # [18:01] <darobin> for those, bump the timeout to, I don't know 300s
- # [18:02] <darobin> in fact, thinking about it, I'm realising that the ability to specify a numeric time out is undesirable
- # [18:02] <darobin> you only want two values: "regular" and "might take a fucking long time"
- # [18:03] <darobin> the reason being that you don't know what context your tests are being run in, and so don't know what's reasonable and long
- # [18:03] <tobie> darobin: nah, you want PER DEVICE numeric values
- # [18:03] <jgraham> darobin: Or climb down from my anti-meta stance and say that rather than doing setup({timeout:10000}) do <meta name=timeout value=10> and have the setup function read from that
- # [18:03] * darobin kicks tobie
- # [18:03] <jgraham> And if you can't add <meta> use the manifest
- # [18:04] <darobin> jgraham: honestly, I still don't think that's all that useful due to it being numeric
- # [18:04] <jgraham> darobin: No, the person running the tests applies a device-specific multiplier
- # [18:04] <darobin> mmmmm, maybe
- # [18:04] <jgraham> That was what we did at Opera
- # [18:04] <darobin> that works?
- # [18:04] <jgraham> It wasn't perfect but it was OK
- # [18:04] <jgraham> I mean
- # [18:04] <tobie> what's the use case for long timeouts, btw?
- # [18:05] <jgraham> You don't want to wait 300s for a test that is supposed to take 11s just because the default is 10
- # [18:05] <darobin> tobie: tests that do loads of stuff
- # [18:05] <gsnedders> tobie: Imagine a test that loads a file over XHR, using a script such that only 10 bytes are sent per second.
- # [18:05] <darobin> jgraham: right, but then again if you work from a multiplier it means that everyone has to enter based on the same expectations
- # [18:06] <jgraham> Yes
- # [18:06] <gsnedders> The problem with multipliers is not everything scales equally. IO (for cache behaviour) can be radically different to CPU speed, for example.
- # [18:06] <darobin> jgraham: say I checked my test on my old iphone 3g because I'm that kind of guy, and updated the timeout accordingly... you're going to have a bad time multiplying that
- # [18:06] <tobie> isn't there a way to ping the framework at regular intervals in such cases?
- # [18:07] <gsnedders> tobie: Okay then, sync XHR.
- # [18:07] <jgraham> tobie: that sounds way too invasive
- # [18:07] <gsnedders> tobie: Testing that it doesn't give up after 20s.
- # [18:08] <tobie> gsnedders: that's really QoI tests
- # [18:08] <jgraham> Not at all
- # [18:09] <tobie> jgraham: enlighten me
- # [18:10] * Quits: rhauck (~Adium@public.cloak) ("Leaving.")
- # [18:10] <jgraham> well, everything is testing the QoI to some extent. So there are two ways to look at this: a) apart from the hardware limitations clause there is no provision in the standard for giving up on operations so it is a conformance problem if you do so and b) we actually don't care about QoI vs conformance; we care about interop and should test all the things we know affect interop
- # [18:11] <jgraham> rather than arbitarily rejecting tests (that might well be based on actual browser bugs)
- # [18:11] <Ms2ger> Fwiw, mochitests have a requestLongerTimeout(factor) that just multiplies the default timeout
- # [18:12] <tobie> jgraham: thanks
- # [18:13] <jgraham> Ms2ger: It isn't clear to me that's materially different to a ms-based timeout and a per-platform multiplier
- # [18:14] <Ms2ger> Maybe not
- # [18:14] <Ms2ger> But I guess that means that a 10s timeout doesn't mean you time out after 10s
- # [18:17] * Quits: zcorpan_ (~zcorpan@public.cloak) (Client closed connection)
- # [18:18] * Joins: zcorpan (~zcorpan@public.cloak)
- # [18:25] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [18:30] * Joins: zcorpan (~zcorpan@public.cloak)
- # [18:45] * Joins: rhauck (~Adium@public.cloak)
- # [19:00] * Joins: [1]kershaw (~kkershaw@public.cloak)
- # [19:05] * Quits: kershaw (~kkershaw@public.cloak) (Ping timeout: 180 seconds)
- # [19:05] * [1]kershaw is now known as kershaw
- # [19:05] * Quits: darobin (rberjon@public.cloak) (Client closed connection)
- # [19:06] * Joins: jhammel (~jhammel@public.cloak)
- # [20:38] * Joins: AutomatedTester (~AutomatedTester@public.cloak)
- # [20:57] * Quits: Ms2ger (~Ms2ger@public.cloak) ("nn")
- # [21:20] * Joins: jimevans (~jimevans@public.cloak)
- # [21:42] * Quits: jimevans (~jimevans@public.cloak) ("Colloquy for iPhone - http://colloquy.mobi")
- # [22:22] * heycam|away is now known as heycam
- # [22:23] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [22:24] * Joins: zcorpan (~zcorpan@public.cloak)
- # [22:31] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [22:54] * Joins: zcorpan (~zcorpan@public.cloak)
- # [23:31] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 180 seconds)
- # [23:40] * Quits: shepazu (schepers@public.cloak) ("is sleepy")
- # [23:55] * Joins: zcorpan (~zcorpan@public.cloak)
- # Session Close: Fri Aug 23 00:00:00 2013
The end :)