') ... [14:25:38] <^demon> I wonder if the GWT 2.3 move happened yet. [14:25:41] my IE begs to differ re:WMF's gerrit, "A script is preventing the web page at wikimedia.org from responding" and it takes ages for the project list to load (Win7/IE9) [14:26:14] (and I would like to know who to blame for choosing gerrit over extending CR, just for the sake of it) [14:27:16] <^demon> "Extending CR" was abandoned as an idea well before Gerrit was chosen. [14:27:23] ^demon: I have no idea how to verify that. gerrit-gwtui/pom.xml simply list com.google.gwt [14:27:30] <^demon> hashar: Looks like no [14:27:42] <^demon> ashley: Many of the early switching to Git discussions all agreed that extending CR was a non-option [14:28:23] ashley: yup the community mostly agreed it was not worth the time to switch CR to get git support [14:28:33] would take too many engineers days to got something [14:35:22] https://test2.wikipedia.org/ [14:35:25] No localisation cache found for English. Please run maintenance/rebuildLocalisationCache.php. [14:35:27] Damn you MW. [15:13:01] sumanah: https://twitter.com/jonty/status/252765497127485440 [15:13:41] Reedy: Have you ever heard of Sturgeon's Law? [15:17:02] hashar: Is there a nice way to pass key value pairs as a test provider? Without iterating over it to make it array( $key, $value )? [15:17:40] $output = array(); foreach( $foo as $key => $value ) { $output[] = array( $key, $value ); } [15:18:18] I've rewritten it somewhat https://gerrit.wikimedia.org/r/#/c/22936/9/tests/phpunit/includes/api/ApiGeneratorTest.php,unified [15:20:41] sumanah: sounds about right ;) [15:30:25] Reedy: given gerrit is dead .. hm I will have trouble answering :-) [15:31:10] hashar: don't really need gerrit to answer that question. You don't need to visit the page ;) [15:32:53] Reedy: so the data providing functions are a huge array [15:32:59] yeah [15:33:02] each entry is an array of paramaeters [15:33:08] I know :p [15:33:18] it doesn't like being passed array( foo => bar, baz => boo ); [15:33:28] yeah a huge array [15:33:32] which only contains array [15:33:42] it wants array( array( foo, bar ), ( baz, boo ) ); [15:33:43] it wants array( array( foo, bar ), ( baz, boo ) ); [15:33:58] so should be: array( array(case1, foo) , array(case2, bar) ) [15:34:22] the way I remember about that is by using a $cases = array(); in my provider [15:34:46] then I append each cases to it: $cases[] = array( 'para1', 'para2' ); [15:34:57] http://p.defau.lt/?knkZApouMLsvnoHlwCw4lg [15:35:00] ^ just seems stupid [15:35:06] $cases[] = array( 'para1', 'parasometing' ); [15:35:23] that one should work [15:35:26] isn't it ? [15:36:23] array( 'foo' => 'bar', 'baz' => 'boo' ); [15:36:26] ^ that won't 'work [15:36:35] PHPUnit does something like: foreach( dataProvider() as $value ) { call_user_func_array( 'testMethod', $value ); } [15:36:58] ahh hm [15:37:13] Damnnit. Why isn't setUp called before it gets the data from your provider? [15:37:20] $output = array( array( 'foo' => 'bar', 'baz' => 'boo' ) ); [15:37:29] $output[] = array( array( 'foo' => 'bar', 'baz' => 'boo' ) ); [15:37:42] which will pass as first parameter: array( 'foo' => 'bar', 'baz' => 'boo' ) [15:38:17] setUp is run before each test [15:38:24] and the provider gives the tests [15:38:28] so setUp must be run after [15:38:38] you could use __construct() or setupBeforeClass() [15:38:49] the later is run when the class is ibuild [15:40:01] there is even assertPreConditions and assertPostConditions which are run before/after each asserts but I never used them myself [15:40:55] ahh Reedy doc is at http://www.phpunit.de/manual/current/en/fixtures.html#fixtures.more-setup-than-teardown [15:41:22] hashar: http://p.defau.lt/?HZMcxB1czy_3wvj0YLMozQ [15:41:26] I might just keep that ;) [15:41:53] uggllly :-D [15:42:23] I will give a poke this evening [15:42:25] http://p.defau.lt/?R70wIM_HqLVr_KuTfY_cWQ [15:42:29] ^ And that is better? :p [15:42:51] yeah that one looks nicer [15:42:57] o_0 [15:44:41] Reedy: would "instaceof" work ? [15:44:57] no, we're not instaniating them [15:45:04] though, doing so might not be much more expensive [15:45:44] and this way you could use the assertInstanceOf() from PHPUnit [15:45:51] mmm [15:45:54] * Reedy shrugs [15:45:55] but I guess we don't need to initialize the objects [15:45:59] that will slow down the tests a bit [15:46:02] so that is fine [15:46:15] I am looking how we can mark a test as passing [15:47:00] if ( $obj instanceof ApiQueryGeneratorBase ) { [15:47:00] $retval['generator'] = ''; [15:47:00] } [15:47:03] ^ we do that in ParamINfo [15:47:26] <^demon> is_subclass_of() is such a silly function name. [15:47:28] * ^demon whacks php [15:48:17] 90 minutes later [15:48:20] SCAP IS STILL FUCKING RUNNING [15:49:41] <^demon> I haven't run scap in so long. [15:49:48] <^demon> When did it get so fucking slow? [15:49:52] * marktraceur watches and popcorns [15:50:00] It's usually half an hour [15:50:15] or nearer 1 if you're deploying a new version/rebuilding 2 l10n caches [15:50:26] * Reedy hits enter a few more times [15:50:41] Button mashing: Will usually make things load faster [15:51:01] IT wouldn't be the first time it's fixed things [15:51:06] hanging ssh prompts and stuff [15:51:11] snapshot1001 [15:51:19] * marktraceur thanks Nintendo for bad misconceptions about how software works [15:51:47] aha, it's on the last few hosts now [15:51:48] * Reedy dies [15:52:56] <^demon> marktraceur: You mean pulling it out, blowing dust off it, and plugging it back in again doesn't work in other situations? [15:53:16] <^demon> Granted, that's a hardware issue :p [15:53:53] ^demon: No, I find that works with a lot of hardware (there's a joke in there somewhere, but it's too obvious) [15:54:04] "I can't get it up..." [15:54:42] Hmmm, the HTTPS padlocks have disappeared [15:54:46] https://test2.wikipedia.org/wiki/Special:Version [16:01:39] test2 is now on 1.20wmf1 if anyone cares [16:04:15] <^demon> test2? [16:04:18] <^demon> That's my favorite wiki. [16:17:54] 1.21wmf1 :P) [16:17:56] :) [16:19:44] * aude cares [16:23:07] need to wait till its on test and mediawiki.org to get some proper usage [16:27:47] chrismcmahon, zeljkof__, if you need any help with my hackish tests let me know :) [16:27:53] hi zeljkof__ [16:34:59] hi marktraceur and sumanah :) [16:35:06] hi zeljkof__! welcome [16:35:23] thanks [18:03:17] 1.21wmf1 on testwiki, test2wiki and mediawikiwiki [18:03:21] please test, etc [19:14:33] !g I2521b78c7de94c183b7de7e7d4b2b510ab0c7770 [19:14:33] https://gerrit.wikimedia.org/r/#q,I2521b78c7de94c183b7de7e7d4b2b510ab0c7770,n,z [19:15:31] !g If8d16b066015ed1bcaf38408511ac3713eaa6540 [19:15:31] https://gerrit.wikimedia.org/r/#q,If8d16b066015ed1bcaf38408511ac3713eaa6540,n,z [19:16:49] Stop spamming your keyboard [19:17:51] is there a cat walking across hashar's keyboard? ;-) [19:18:16] ;-D [19:19:54] bsitu. I can't find you on Trello, username? [19:21:20] DarTar: it's bsitu [19:21:44] k - invite sent [19:41:33] hashar: https://gerrit.wikimedia.org/r/#/c/22936/11/tests/phpunit/includes/api/ApiGeneratorTest.php,unified [19:41:34] There :p [19:42:38] Reedy: the first line is still tooooo long :] [19:42:58] I shortened both that, and the upper lines [19:43:06] 65? Pfft [19:43:08] you need a blank line too :-) [19:43:34] example: https://android-review.googlesource.com/#/c/43692/ :-] [19:47:22] Damn [19:47:23] $this->assertFalse( isset( $prefixes[$prefix] ), "Module prefix '{$prefix}' is shared between {$class} and {$prefixes[$prefix]}" ); [19:47:33] ^ can't do that, beacuse it whinges about undefined index making the message [19:47:34] lol [19:48:14] nvm [19:48:45] Reedy: so you want to assert that $prefixes as a key named $prefix probably [19:48:57] _doesn't_ have [19:49:15] s/as/has/ [19:49:17] $this->fail( "Module prefix '{$prefix}' is shared between {$class} and {$prefixes[$prefix]}" ); [19:49:17] } [19:49:21] bah [19:49:25] if ( isset( $prefixes[$prefix] ) ) { [19:49:25] $this->fail( "Module prefix '{$prefix}' is shared between {$class} and {$prefixes[$prefix]}" ); [19:49:25] } [19:49:56] assertArrayNotHasKey( mixed $key, array $array[, string $message = '']) [19:50:00] Reedy: ^ [19:50:35] bah we really need badges [19:50:46] I could give marktraceur a "testing stuff" one :-] [19:50:56] and sam too :) [19:51:42] Still gives an undefined index error for "Module prefix '{$prefix}' is shared between {$class} and {$prefixes[$prefix]}" due to teh last bit.. [19:53:16] so that is the message having an undefined key obviously [19:53:16] Oh [19:53:20] I see [19:53:30] Can it accept a delegate for what to call to get the message if it fails? :p [19:53:33] Reedy: PHP is trying to expand the variables in your string [19:53:39] Reedy: Escape the $ [19:54:39] All of them? [19:55:27] simply state that "$prefix is expected to be found in the list of prefixes" [19:55:55] you can't reference $prefixes[$prefix] since the message is shown when it does not exist :-) [19:56:06] well, then the developer has to work out where it's in use [19:56:14] i could wrap the building of the message in an isset itself [19:56:16] * Reedy barfs [19:56:27] grr [19:56:33] we need an ether pad with syntax highlighting [19:56:39] that would make that stuff easier [19:56:40] orrr [19:56:47] we can Skype and screen share :-] [19:57:12] Module prefix 'cl' is shared between ApiQueryCategories and {$prefixes[$prefix]} [19:58:24] $message = ''; [19:58:24] if ( isset( $prefixes[$prefix] ) ) { [19:58:24] $message = "Module prefix '{$prefix}' is shared between {$class} and {$prefixes[$prefix]}"; [19:58:26] * Reedy facepalms [19:58:38] $this->assertArrayNotHasKey( $prefix, $prefixes, $message ); [19:59:06] replacing the last bit with "and another class" decreases its usefulness [19:59:40] Reedy: wanna screen share over Skype ? [19:59:43] or hangout [19:59:45] or pigeons [20:00:16] $this->assertArrayNotHasKey( $prefix, $prefixes, [20:00:17] isset( $prefixes[$prefix] ) [20:00:17] ? "Module prefix '{$prefix}' is shared between {$class} and {$prefixes[$prefix]}" [20:00:17] : '' [20:00:17] ); [20:00:20] I.. [20:00:21] if using pigeon, please print your code with a color printer so I received a syntax highlighted copy of your code [20:00:34] I only own colour printers ;) [20:04:22] * hashar send a pigeon to Sam [20:04:24] I think I'm gonna leave the code as is [20:04:45] It's not great, but it doesn't look as evil as doing something like that [20:04:55] or only then conditionally running the assert when we know it's going to fail [20:05:25] which is evil [20:05:26] ;) [20:05:32] Yeah [20:05:37] as I said on the comment, it get the stuff done [20:05:43] The current isn't great, but that's a lot worse [20:06:03] there is probably a nicer way to test that but it is probably not worth spending hours on it :) [20:06:19] Yup, exactly [20:06:24] Onwards with "proper" bugs [20:06:37] you still need to fix the first commit line :-) [20:06:49] 2nd line need to be empty or gerrit get lost :) [20:07:18] Now only if gitweb wasn't so crappy [20:09:22] it's quicker to push it to a github clone, and use that [20:09:28] Reedy: not sure if you're the right person, but I assigned https://bugzilla.wikimedia.org/40668 to you [20:09:36] Yeah, I'm just looking at it [20:09:59] Oh look, it's an already known bad commit [20:10:14] Reedy is fast :) [20:10:19] I think [20:10:21] https://github.com/reedy/MediaWiki/commit/0a1e291036fac2376b4645f9d070bc3e76ca73d4 [20:10:37] https://gerrit.wikimedia.org/r/#/c/25529/ [20:13:25] interesting. I would have had one day to discover that on beta labs, but instead I was working on hiring the guy to do the automation that would have discovered it (if we were looking for it). step by step by step... [20:14:02] Hmm, maybe not then.. [20:14:24] I figure it would have been visible on Friday, but maybe not. [20:14:33] Amusingly, if you logout, that message is correct [20:14:41] Return to MediaWiki. [20:14:56] and the tooltip/hover thingie is correct also [20:16:46] And (unsuprisingly) we apparently have multiple ways to add a return to.. [20:19:37] $link = $this->msg( 'returnto' )->rawParams( [20:19:37] Linker::link( $title, $text, array(), $query, $options ) )->escaped(); [20:19:41] that still looks likely [20:21:28] I should debug on testwiki, not mw.org [20:21:29] mehh [20:21:55] Linker::link( $title, $text, array(), $query, $options ) [20:21:58] returns Array [20:32:06] chrismcmahon: found it [20:32:10] $this->getOutput()->addReturnTo( $returnToTitle, $returnToQuery, $options ); [20:32:18] becomes to fix it [20:32:19] $this->getOutput()->addReturnTo( $returnToTitle, $returnToQuery, null, $options ); [20:33:24] Caused by https://github.com/reedy/MediaWiki/commit/565014a8cb578889de1666256d1dba6a7ecc8629 [20:39:04] Reedy: makes sense, thanks [20:47:44] ori-l: gwicke tells me you may need a hand with automated testing on the dumps, care to point me in the right direction? [20:48:26] marktraceur: subbu just brought up the same thing in #mediawiki-parsoid [20:48:50] gwicke: ori-l is not there, hence, the more different channel [20:58:06] I am out, see you later [22:54:46] Ryan_Lane: So we're setting up CI for Parsoid, and we'd like to SSH into Gerrit--but we may need a new user and/or to use the existing Jenkins user. If you have a good grasp of what to do, I'd like to hear it, my instinct is to create a new user (ParsoidJenkins or so) [22:55:11] doesn't jenkins already have a user? [22:55:24] Ryan_Lane: Yes, but this is going to be a new instance for now [22:55:28] new instance? [22:55:34] Ryan_Lane: Of Jenkins [22:55:37] please tell me you aren't planning on doing this from inside of labs [22:55:55] Ryan_Lane: We needed a newer version of Node, as I recall, than what was available on the current Jenkins [22:56:03] then we should look at upgrading [22:56:18] labs isn't intended for production work [22:56:20] K, how should I go about doing that? [22:56:30] Ryan_Lane: Parsoid isn't a production-grade product yet [22:56:36] but testing it is [22:56:50] The tests don't even pass, we're just trying to make sure we don't lose tests [22:56:59] doesn't matter [22:57:06] we have a testing server. in production. [22:57:29] marktraceur: you should work with hashar [22:57:29] Ryan_Lane: K, then let's do that. [22:57:33] he has a test instance in labs [22:57:39] the upgrade can be tested there [22:57:45] then we can upgrade production [22:57:45] Ryan_Lane: A) hashar is asleep B) hashar is doing other things this week [22:57:58] So reason 2 for doing this was so I could get it working sooner [22:58:08] well, this can't run in labs [22:58:21] from a security perspective it's not really OK [22:58:48] * marktraceur backs out from attempts [22:59:26] hashar manages jenkins.... [22:59:34] I don't know who can help on the ops side [22:59:43] no clue who's familiar with it [23:00:15] Well, I really don't need any help managing it, just a Gerrit user with Verify privs on the Parsoid repo. I have the rest set up. [23:06:40] Ryan_Lane: Wait, can I configure a job in Jenkins? There's an email from hashar that says people in our LDAP group can, but I'm not sure I have that... [23:07:11] I don't know [23:07:16] I don't really use jenkins much [23:07:37] Noted [23:07:39] marktraceur: it's not a matter of needing help managing it. it's that we need to make sure he's involved [23:08:00] it would suck to upgrade the production instance just to break everything and not have hashar around [23:08:28] I didn't intend to touch the production instance, until you freaked out and told me I couldn't run Jenkins on labs [23:08:42] it's not a matter of freaking out. It's against the policy [23:08:53] Now I feel all inelegant, have to run tests through a web interface [23:09:01] ...if I can log in, which I apparently can't [23:09:16] what's your labsconsole user name? [23:09:21] it uses that username/password [23:09:28] or well, it should [23:09:39] There we go, yes [23:09:44] Excellent [23:10:34] All right, hack it up to run tests on the labs instance, then I'll make a job. [23:10:54] heh. also shouldn't be running tests against the labs instance ;) [23:11:07] I'm not going to completely say no to that, though [23:11:35] Ryan_Lane: We don't have any other way to do it, so until we upgrade the production Jenkins, this is kind of it [23:11:36] in general we try to keep operational dependencies away from labs [23:11:43] yeah. it's fine for now [23:11:53] as you mentioned, parsoid isn't production yet [23:11:59] when it is, we need to have this solved [23:12:15] It should be solved within two weeks, but hashar isn't available to solve it yet [23:12:19] * Ryan_Lane nods [23:12:20] sounds good [23:12:53] I have to be kind of strict about stuff like this, or we'll end up with half of production relying on labs [23:13:19] and mark will want to kill me [23:14:44] Well, that's an awful idea [23:14:49] I agree [23:15:11] But the only place Parsoid is deployed is Labs, so I didn't figure it was an issue....*shrug* this should work, I guess [23:15:24] * Ryan_Lane nods