Options:
- # Session Start: Thu Mar 21 00:00:00 2013
- # Session Ident: #testing
- # [00:07] * Quits: glenn (~gadams@public.cloak) (Client closed connection)
- # [00:31] * Joins: glenn (~gadams@public.cloak)
- # [03:38] * Joins: glenn_ (~gadams@public.cloak)
- # [03:38] * Quits: glenn (~gadams@public.cloak) (Client closed connection)
- # [03:49] * Quits: glenn_ (~gadams@public.cloak) (Client closed connection)
- # [04:47] * Joins: glenn (~gadams@public.cloak)
- # [06:49] * Quits: glenn (~gadams@public.cloak) (Client closed connection)
- # [07:54] * Joins: tobie (tobie@public.cloak)
- # [08:00] * Joins: glenn (~gadams@public.cloak)
- # [08:04] * Quits: glenn (~gadams@public.cloak) (Ping timeout: 60 seconds)
- # [08:29] * Quits: tobie (tobie@public.cloak)
- # [08:50] * Joins: tobie (tobie@public.cloak)
- # [08:52] * Quits: tobie (tobie@public.cloak)
- # [09:24] * Joins: darobin (rberjon@public.cloak)
- # [09:44] * Joins: zcorpan (~zcorpan@public.cloak)
- # [10:10] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [10:11] * Joins: zcorpan (~zcorpan@public.cloak)
- # [10:14] * Joins: AutomatedTester (~AutomatedTester@public.cloak)
- # [10:43] * Quits: AutomatedTester (~AutomatedTester@public.cloak) (Client closed connection)
- # [11:25] * Joins: abarsto (~abarsto@public.cloak)
- # [11:25] * abarsto is now known as ArtB
- # [12:51] * Joins: darobin_ (rberjon@public.cloak)
- # [12:51] * Quits: darobin (rberjon@public.cloak) (Client closed connection)
- # [12:58] * Joins: plh (plehegar@public.cloak)
- # [13:01] * Joins: tobie (tobie@public.cloak)
- # [13:01] * darobin_ is now known as darobin
- # [13:33] * Quits: shepazu (schepers@public.cloak) ("is sleepy")
- # [15:29] * Joins: shepazu (schepers@public.cloak)
- # [15:55] * Joins: AutomatedTester (~AutomatedTester@public.cloak)
- # [15:59] * Quits: AutomatedTester (~AutomatedTester@public.cloak) (Ping timeout: 60 seconds)
- # [16:34] * Joins: AutomatedTester (~AutomatedTester@public.cloak)
- # [17:13] * Joins: jhammel (~jhammel@public.cloak)
- # [17:15] * Joins: glenn (~gadams@public.cloak)
- # [17:19] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [17:53] * Quits: stearns (~anonymous@public.cloak) (stearns)
- # [17:56] * Joins: stearns (~anonymous@public.cloak)
- # [17:57] * Quits: shepazu (schepers@public.cloak) (Ping timeout: 60 seconds)
- # [18:00] * Joins: shepazu (schepers@public.cloak)
- # [18:02] * Joins: Ms2ger (~Ms2ger@public.cloak)
- # [18:03] * Quits: AutomatedTester (~AutomatedTester@public.cloak) (Client closed connection)
- # [18:06] * Quits: darobin (rberjon@public.cloak) (Client closed connection)
- # [18:06] * Joins: darobin (rberjon@public.cloak)
- # [18:30] * Quits: darobin (rberjon@public.cloak) (Client closed connection)
- # [18:53] * Joins: zcorpan (~zcorpan@public.cloak)
- # [18:55] * Quits: glenn (~gadams@public.cloak) (Client closed connection)
- # [19:02] * Quits: shepazu (schepers@public.cloak) (Ping timeout: 60 seconds)
- # [19:02] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [19:03] * Joins: zcorpan (~zcorpan@public.cloak)
- # [19:07] * Quits: zcorpan (~zcorpan@public.cloak) (Ping timeout: 60 seconds)
- # [19:13] * Joins: AutomatedTester (~AutomatedTester@public.cloak)
- # [19:17] * Quits: AutomatedTester (~AutomatedTester@public.cloak) (Ping timeout: 60 seconds)
- # [19:38] * Joins: zcorpan (~zcorpan@public.cloak)
- # [19:54] <ArtB> tobie - re your "Review of tests upstream …" thread ...
- # [19:55] <ArtB> in the GH world, does a spec still have directories for submissions?
- # [19:56] <ArtB> e.g. webstorage/submissions/Tobie/foo.html
- # [19:58] <Ms2ger> No
- # [20:03] <ArtB> oh, that's right the "submissions" are in someone else's GH home?
- # [20:05] * Joins: glenn (~gadams@public.cloak)
- # [20:05] <tobie> ArtB: pretty much, yes.
- # [20:09] <ArtB> so, in the case of the upstream streamlining, you propose those tests be automatically approved if the upstream contributor can show the tests have had "adequate review"?
- # [20:09] * Quits: glenn (~gadams@public.cloak) (Client closed connection)
- # [20:09] * Joins: glenn (~gadams@public.cloak)
- # [20:16] <jgraham> ArtB: Yes, that's the proposal AIUI
- # [20:19] <ArtB> jgraham, but that doesn't replace the WG's responsibility to approve tests used to determine CR exit criteria. Correct?
- # [20:19] <tobie> ArtB: yes. We'd combine this with a bunch of automated tests ran on the submission (which the person merging the submission would have access to).
- # [20:19] <jgraham> ArtB: Well, I think the CR exit criteria are defined on a per-WG basis.
- # [20:20] <jgraham> So you can more or less choose the process you want there
- # [20:20] <jgraham> (all choices are bad though)
- # [20:24] <ArtB> well I am not aware of any Silver Bullets here
- # [20:26] <tobie> ArtB: there's a proposal to extend the notion of test facilitators for all specs in the web-platform-test repo.
- # [20:26] <tobie> ArtB: what if you made the facilitator responsible for deciding whether coverage is adequate or not?
- # [20:27] <tobie> ArtB: we'll have coverage data available too.
- # [20:27] <ArtB> well, we would need to quibble on the defn of "responsible"
- # [20:27] <ArtB> I think a TF should be qualified enuf to make a proposal re coverage
- # [20:27] <ArtB> but it's really a WG decision
- # [20:27] <tobie> ArtB: oh uh. You do that. :) I'm not part of WebApps. :)
- # [20:27] * Joins: AutomatedTester (~AutomatedTester@public.cloak)
- # [20:28] * Ms2ger doesn't care about CR exit criteria
- # [20:28] <ArtB> I hear all that time Ms2ger from Members with no IP ;-)
- # [20:28] <tobie> Ms2ger: that was a given.
- # [20:29] <jgraham> Well we have had the problem in WebApps where company A submits a testsuite for a feature and then company A's representative is asked "is your testsuite sufficient to go to CR?"
- # [20:29] <jgraham> Which, in the case that company A only cares about IP, is always answered "yes"
- # [20:29] <Ms2ger> And then two vendors submit their test suites, and suddenly duplicated tests
- # [20:30] <ArtB> so a fundamental problem here is certainly that different Members want different things from test suites
- # [20:30] * jgraham doesn't really care about duplicated tests
- # [20:30] <Ms2ger> ArtB, for all I care, you could publish recs from caniuse data
- # [20:30] * tobie doesn't care about what people don't care about.
- # [20:30] * tobie cares about what people care about.
- # [20:31] <Ms2ger> ArtB, that would allow people to work on interoperability testing without process pressure, and would get IP protection much sooner
- # [20:32] <tobie> heh
- # [20:32] * Ms2ger sees no downsides
- # [20:33] <jgraham> It would discourage some parties from submitting tests at all
- # [20:33] <Ms2ger> jgraham, so, git fetch w3c && git merge --ff-only, right?
- # [20:34] <jgraham> Yes, if you don't have any local changes
- # [20:34] <Ms2ger> No, doesn't work
- # [20:34] <jgraham> Where does it fail?
- # [20:34] <Ms2ger> $ git merge --ff-only
- # [20:34] <Ms2ger> usage: git merge [options] [<commit>...]
- # [20:34] <jgraham> Which git version?
- # [20:34] <jgraham> git --version
- # [20:34] <Ms2ger> 1.7.5.4
- # [20:35] <Ms2ger> --ff-only is listed in the help message
- # [20:35] <Ms2ger> And on the point of discouraging these parties; I'm not sure if that would be positive or negative overall
- # [20:36] <jgraham> Well they do sometimes contribute good tests
- # [20:36] <Ms2ger> News to me ;)
- # [20:36] <jgraham> So we have this problem
- # [20:36] <jgraham> Well we have lots of problems
- # [20:36] <Ms2ger> Amen
- # [20:37] <jgraham> I wonder if I am supposed to say this. Oh well.
- # [20:38] <jgraham> Microsoft are obviously heavily influenced by IPR concerns and seem to submit tests for that reason.
- # [20:38] <jgraham> Opera have submitted lots of tests in the past, but aren't developing their own engine anymore
- # [20:39] <jgraham> Mozilla submit fewer tests than one might hope for given their Misson (but can probably be encouraged to submit more tests)
- # [20:39] <jgraham> Google and Apple basically never submit tests
- # [20:39] <jgraham> So
- # [20:40] * ArtB noticed the "fire sale" on Opera tests ;-)
- # [20:40] * Joins: mdas (~mdas@public.cloak)
- # [20:40] <Ms2ger> Hi mdas :)
- # [20:41] <jgraham> Some serious work is needed to change this situation
- # [20:42] <jgraham> Ms2ger: Oh you probably want to say git merge --ff-only w3c/master
- # [20:42] <Ms2ger> Worked, thanks
- # [20:42] <Ms2ger> And yeah, I'm aware we're not doing as well as I'd like
- # [20:43] <jgraham> Otherwise there is a danger that we will end up with more specs where we have the minimum of tests needed to satisfy the Process
- # [20:43] <jgraham> but not enough to actually cover the corner cases
- # [20:43] <jgraham> Making the Process heavier isn't going to go down well (after all, everyone likes the IPR stuff)
- # [20:43] <Ms2ger> Aryeh's tried to get us to default to testharness.js, but that failed on "I don't like its API" grounds, unfortunately
- # [20:44] <ArtB> there is indeed a mismatch between caring about a set of tests that qualify all of the QA/customer requirements and a set that allows a spec to get IP commitments
- # [20:44] <ArtB> so dangerous in what way?
- # [20:44] <ArtB> is anyone blocking the deployment of their browser b/c the W3C test suite is not complete?
- # [20:45] <Ms2ger> Dangerous in the sense that the conformance test suite sucks
- # [20:45] <Ms2ger> And that interoperability sucks as a result
- # [20:45] <jgraham> Dangerous in the "interop. suffers and eventually the platform fails" kind of way (where ending up with a single rendering engine is a particular failure mode)
- # [20:46] <ArtB> well it must not suck enuf for the vendors James enumerated above to care
- # [20:47] <Ms2ger> Well
- # [20:47] <Ms2ger> Vendors benefit from having other vendors write tests
- # [20:47] <Ms2ger> Not so much from submitting tests themselves
- # [20:47] <mdas> Ms2ger: oh hello
- # [20:47] <jgraham> In the short term, submitting tests is a pain
- # [20:48] <jgraham> You have to move to someone else's setup for writing tests
- # [20:48] <jgraham> And inherit some of their problems (exposed as API decisions)
- # [20:49] <Ms2ger> (trivial PR up at https://github.com/w3c/web-platform-tests/pull/39)
- # [20:49] <jgraham> So for *today* it can seem like a bad idea
- # [20:50] <jgraham> And if you are relatively resource happy and don't really care about the long term success of the platform (or have good marketshare and only care about your own continued success) it can seem like giving a free leg up to your competitors
- # [20:50] <ArtB> so, the prevailing attitude here (when a browser group was in my building) was that W3C tests were a joke and they had to do ALL of the testing themselves. That attitude seems to be shared by most of the browser vendors in WebApps.
- # [20:50] <ArtB> I say most because Opera might see it differntly?
- # [20:51] <jgraham> Well the thing is
- # [20:51] <ArtB> (or perhaps "used to"
- # [20:51] <jgraham> The W3C doesn't have tests
- # [20:51] <Ms2ger> They are a joke :)
- # [20:51] <jgraham> People have tests
- # [20:51] <jgraham> That they contribute to the W3C pool
- # [20:51] <jgraham> (this would be a bit different if the W3C employed people to write tests, but they don't)
- # [20:51] <ArtB> but I do see some members willing to contribute to a "Test Suite for the Attorneys" ;-)
- # [20:51] <jgraham> The people with the most tests are the vendors
- # [20:52] <jgraham> But they largely aren't contributing them to the pool
- # [20:52] <jgraham> And where they are (excluding Opera), it is somewhat obvious that they have the lawyers in mind
- # [20:53] <tobie> Ms2ger: what parts of the IP was disliked would be useful info.
- # [20:53] <tobie> s/IP/API/
- # [20:53] <Ms2ger> tobie, it all happened in a public newsgroup
- # [20:53] * tobie is tired.
- # [20:53] <Ms2ger> tobie, mostly "more typing", IIRC
- # [20:53] <ArtB> so you talk about the "test suite for attorney" as if that's a Bad Thing
- # [20:53] <Ms2ger> As in, mochitests have ok/is
- # [20:54] <Ms2ger> And no test()
- # [20:54] <ArtB> another way to look at it is that without IP commitments, the OWP is at risk ;)
- # [20:54] <Ms2ger> Office of Water Programs? Oxford Wheels Project?
- # [20:55] <Ms2ger> Organizacja Wyzwolenia Palestyny?
- # [20:55] <tobie> Ms2ger: we could get the code's AST and rewrite the tests.
- # [20:55] <Ms2ger> tobie, as in, automatic rewriting from mochitest to th.js?
- # [20:56] <tobie> yeah
- # [20:56] <Ms2ger> That would lead to pretty unidiomatic th.js, I'm afraid :/
- # [20:56] <tobie> that, or write a shim.
- # [20:56] <tobie> Ms2ger: haven't looked at the code, so difficult to say.
- # [20:56] <tobie> links?
- # [20:57] <Ms2ger> The main issue is that th.js expects you to wrap logical blocks in test()
- # [20:57] <Ms2ger> Whereas mochitest just lets you write code in the global scope
- # [20:57] <tobie> Sure.
- # [20:58] <Ms2ger> (A smaller issue is that is/ok don't map cleanly to th.js assertions)
- # [20:58] <tobie> sounds like that implies one mochitest page -> one th.js test function
- # [20:59] <Ms2ger> (is does == instead of ===, and ok tests truthiness instead of ===true)
- # [20:59] <tobie> big deal.
- # [20:59] <tobie> We probably need both anyway.
- # [20:59] <Ms2ger> ==? Not in my experience
- # [21:00] <tobie> ok, yeah. That one sucks.
- # [21:00] <tobie> I'm actually more concerned about mapping these tests to specs.
- # [21:01] <Ms2ger> That's another problem, yes
- # [21:01] <Ms2ger> Also, using non-standard or cutting-edge JS
- # [21:01] <tobie> meh
- # [21:01] <tobie> AST rewrites.
- # [21:01] <Ms2ger> Might work, yes
- # [21:01] <Ms2ger> Though I'd like to see it for 'yield'
- # [21:02] <tobie> yeah, not to sure how that desugars. :-/
- # [21:06] * Joins: darobin (rberjon@public.cloak)
- # [21:07] * Quits: darobin (rberjon@public.cloak) (Client closed connection)
- # [21:09] <jgraham> I think trying to auto-rewrite mochitests is a terrible idea
- # [21:09] <tobie> Ms2ger: could you perhaps point to where in the Moz codebase I could find examples of those tests?
- # [21:09] <Ms2ger> tobie, which kind?
- # [21:09] <tobie> oh, man. how many kinds are there!?
- # [21:10] <Ms2ger> Mochitests, you mean?
- # [21:10] <tobie> jgraham: you're probably right.
- # [21:10] <tobie> Ms2ger: I suppose so.
- # [21:10] <tobie> jgraham: but it's a fun exercise.
- # [21:12] <Ms2ger> http://mxr.mozilla.org/mozilla-central/find?string=test_.*\.html%24&tree=mozilla-central&hint= :)
- # [21:13] <Ms2ger> You probably want the ones in /dom or /content, but not in /dom/imptests
- # [21:13] <Ms2ger> (The latter are, in fact, th.js tests)
- # [21:14] <Ms2ger> tobie, ^
- # [21:14] <tobie> ty
- # [21:15] * Quits: ArtB (~abarsto@public.cloak) ("Leaving.")
- # [21:18] * Joins: shepazu (schepers@public.cloak)
- # [21:24] <tobie> Ms2ger: It doesn't look that bad, actually. Even after an automatic transformation, there's be a significant amount of work triaging the tests.
- # [21:24] <tobie> Wonder if this something we should look into.
- # [21:25] <Ms2ger> Well, yes, I'm sure a lot of our tests just aren't very good :)
- # [21:26] <Ms2ger> Or not testing anything backed by a spec
- # [21:26] <Ms2ger> And then there's licensing \o/
- # [21:27] <tobie> Just because it's open-source doesn't mean yada yada.
- # [21:28] <tobie> oh well.
- # [21:28] <Ms2ger> See https://bugzilla.mozilla.org/show_bug.cgi?id=788511 for example
- # [21:29] <tobie> well, that was fun.
- # [21:30] <tobie> while it lasted.
- # [21:30] <tobie> Back to work now.
- # [21:30] <tobie> Evenin' folks.
- # [21:31] <Ms2ger> Night
- # [21:31] * Quits: glenn (~gadams@public.cloak) (Client closed connection)
- # [21:37] * Quits: tobie (tobie@public.cloak)
- # [21:59] * Joins: glenn (~gadams@public.cloak)
- # [22:10] * Quits: glenn (~gadams@public.cloak) (Client closed connection)
- # [22:10] * Joins: glenn (~gadams@public.cloak)
- # [22:12] * Joins: glenn_ (~gadams@public.cloak)
- # [22:12] * Quits: glenn (~gadams@public.cloak) (Client closed connection)
- # [22:27] * Quits: glenn_ (~gadams@public.cloak) (Client closed connection)
- # [22:28] * Joins: glenn (~gadams@public.cloak)
- # [22:46] * Quits: shepazu (schepers@public.cloak) ("is sleepy")
- # [22:54] * Quits: zcorpan (~zcorpan@public.cloak) (Client closed connection)
- # [22:56] * Joins: zcorpan_ (~zcorpan@public.cloak)
- # [23:09] * Quits: glenn (~gadams@public.cloak) (Ping timeout: 60 seconds)
- # [23:10] * Quits: plh (plehegar@public.cloak) ("Leaving")
- # [23:12] * Joins: glenn (~gadams@public.cloak)
- # [23:31] * Quits: Ms2ger (~Ms2ger@public.cloak) ("nn")
- # [23:48] * Joins: shepazu (schepers@public.cloak)
- # [23:51] * Quits: AutomatedTester (~AutomatedTester@public.cloak) (Client closed connection)
- # Session Close: Fri Mar 22 00:00:00 2013
The end :)