[02:17:11] has anyone successfully incorporated InitialiseSettings / CommonSettings / wgConf.php into a simple test wiki? I want to match production settings but this is daunting. [02:31:31] includes/DefaultSettings.php *and* includes/conf/DefaultSettings.php ? [11:49:47] New patchset: Hashar; "sane default for builddir / sourcedir" [integration/jenkins] (master) - https://gerrit.wikimedia.org/r/28006 [11:50:01] Change merged: Hashar; [integration/jenkins] (master) - https://gerrit.wikimedia.org/r/28006 [14:16:15] 684 Warning: array_map() [function.array-map]: An error occurred while invoking the map callback in /usr/local/apache/common-local/php-1.21wmf1/includes/resourceloader/ResourceLoaderFileModule.php on line 0 [14:16:15] 123 Warning: array_map() [function.array-map]: An error occurred while invoking the map callback in /usr/local/apache/common-local/php-1.21wmf1/includes/resourceloader/ResourceLoaderFileModule.php on line 601 [14:16:19] These are starting to annoy me [14:16:41] That's not good [14:16:52] You mustn't be annoyed [14:17:17] lol [14:17:27] Roans response was before "Yeah, that's supposed to happen" [14:17:38] But yet it only started in the last week or so [14:17:42] Its not the error I'd expect [14:17:46] array_map [14:17:49] <^demon> That's such a lame error. It happens when you throw an exception in an array_map() callback as far as I remember. [14:17:53] Yeah [14:18:07] Something like that, I know we've had it in a few places [14:18:10] <^demon> I believe PHP's response is "Yeah, don't do that then." [14:18:38] lol. [14:19:46] line 0? [14:19:57] Reedy: Is there also [14:19:57] throw new MWException( __METHOD__.": style file not found: \"$localPath\"" ); [14:20:12] that's the only exception thrown from the array_map callback in FileModule:601 [14:20:33] I think array_map suppresses it into a warning [14:20:39] arg [14:20:47] Argh! [14:20:53] we can just change it for a foreach or similar and get rid of array_map [14:21:21] Reedy: Perhaps run tests/phpunit/resources/ResourcesTest.php and see what files are missing [14:21:43] (Yes, we have this now!) [14:21:57] (not in wmf branch yet though) [14:22:41] yeah, just noticed that [14:22:53] I'll steal it from wmf2 in a bit [14:24:02] can be run locally as well of course [14:24:38] Could possibly be an extension too.. [14:24:48] FAILURES! [14:24:48] Tests: 595, Assertions: 595, Failures: 19. [14:24:49] lol. [14:25:12] All mobile frontend related [14:25:22] That probably makes sense why "it just started" [14:26:24] I'll deal with it when my desktop boots.. [14:28:51] Though, i thought I'd done these before... [14:29:48] Wheeeee [14:29:55] And other qunit tests are broken.. [14:30:24] Reedy: "other"" [14:30:35] Non mobile frontend stuff [14:30:42] right [14:30:46] mediawiki.cldr [14:30:55] Right, those are known failures [14:30:59] mediawiki.jqueryMsg: Output matches PHP parser [14:31:08] mediawiki: mw.loader.implement( styles={ "url": { : [url, ..] } } ) [14:31:11] People keep breaking shit [14:31:12] There is some tests I migrated from the Jasmine test suite [14:31:16] turned out they never worked in the first place [14:31:22] planned features (afaik) [14:31:29] I thought Santosh fixed those? [14:31:30] But I just transformed it into QUnit [14:31:34] Yes [14:31:39] Maybe it wasn't merged yet? [14:31:40] Let me check [14:33:13] ResourcesTest::testFileExistence with data set #379 ('/var/www/wiki/mediawiki/extensions/MobileFrontend/javascripts/external/eventlog.js', 'mobile.beta.jquery.eventlog', 'javascripts/external/eventlog.js') [14:33:13] File 'javascripts/external/eventlog.js' referenced by 'mobile.beta.jquery.eventlog' must exist. [14:33:13] Failed asserting that file "/var/www/wiki/mediawiki/extensions/MobileFrontend/javascripts/external/eventlog.js" exists. [14:33:24] Those paths are wrong for my setup [14:34:12] submodule initialized? [14:34:18] nvm, stupid question [14:34:32] (ofc it is, it registered itself) [14:56:52] https://www.mediawiki.org/wiki/MediaWiki_1.21/wmf2 [14:57:50] Krinkle: lol, phpunit is somewhat broken on fenariu [14:58:09] or rather, not installed [14:59:48] lol [14:59:56] And all qunit tests now apparently pass on my dev wiki [15:04:55] Oh. [15:08:40] Krinkle: fixed [15:08:42] Reedy: isn't jenkins running tests for the wmf branches ? [15:08:54] hashar: No idea [15:09:06] But it isn't running them with extensions loaded in? [15:09:17] probably not [15:09:25] :-( [15:09:32] soo many things to do [15:09:45] heh :) [15:09:46] tis ok [15:10:36] https://gerrit.wikimedia.org/r/28030 [15:10:38] daughter time [15:10:42] will be back later tonight [15:10:45] in roughly 4 hours [15:12:27] PHP Warning: file_put_contents(/home/wikipedia/common/wmf-config/ExtensionMessages-1.21wmf2.php): failed to open stream: Permission denied in /home/wikipedia/common/php-1.21wmf2/maintenance/mergeMessageFileList.php on line 119 [15:12:27] Warning: file_put_contents(/home/wikipedia/common/wmf-config/ExtensionMessages-1.21wmf2.php): failed to open stream: Permission denied in /home/wikipedia/common/php-1.21wmf2/maintenance/mergeMessageFileList.php on line 119 [15:12:30] These are new.. [15:17:07] what's the owner of ExtensionMessages-1.21wmf2.php ? [15:20:23] reedy@fenari:/home/wikipedia/common/wmf-config$ ls -al ExtensionMessages-1.21wmf2.php [15:20:23] -rw-rw-r-- 1 reedy wikidev 17409 2012-10-15 14:40 ExtensionMessages-1.21wmf2.php [15:21:12] <^demon> Wasn't that script updated to use mwdeploy? [15:21:17] <^demon> (Or supposed to?) [15:21:40] scap doesnt.. [15:21:47] mw-update-l10n does [15:21:47] sudo -u mwdeploy $BINDIR/mwscript mergeMessageFileList.php --wiki="$mwDbName" \ [15:21:47] --list-file=$SOURCE/wmf-config/extension-list \ [15:21:47] --output=$SOURCE/wmf-config/ExtensionMessages-"$mwVerNum".php [15:21:59] so I guess that's the issue [15:22:55] and I'm not allowed to chown it [15:22:56] wheee [15:34:40] yay, the warnings are going away [16:23:42] Oct 15 16:22:54 10.0.2.191 apache2[16074]: PHP Warning: array_map() [function.array-map]: An error occurred while invoking the map callback in /usr/local/apache/common-local/php-1.21wmf1/includes/resourceloader/ResourceLoaderFileModule.php on line 0 [16:23:42] Oct 15 16:22:55 10.0.2.249 apache2[1621]: PHP Warning: array_map() [function.array-map]: An error occurred while invoking the map callback in /usr/local/apache/common-local/php-1.21wmf1/includes/resourceloader/ResourceLoaderFileModule.php on line 0 [16:23:42] Oct 15 16:22:57 10.0.11.61 apache2[12110]: PHP Warning: array_map() [function.array-map]: An error occurred while invoking the map callback in /usr/local/apache/common-local/php-1.21wmf1/includes/resourceloader/ResourceLoaderFileModule.php on line 601 [16:23:42] Oct 15 16:22:57 10.0.2.191 apache2[16211]: PHP Warning: array_map() [function.array-map]: An error occurred while invoking the map callback in /usr/local/apache/common-local/php-1.21wmf1/includes/resourceloader/ResourceLoaderFileModule.php on line 0 [16:23:44] DAMN IT [16:25:04] Looks like that fixed some of it, but not all... [21:01:13] Krinkle: is there an open bug for RL about it not updating timestamps when resources are removed from a module? i thought i saw it last week but am having trouble locating it [21:01:30] Yes [21:01:38] * Krinkle pulls it up [21:02:13] awjr: https://bugzilla.wikimedia.org/show_bug.cgi?id=37812 [21:02:24] Bug 37812 - Module cache should be invalidated (newer timestamp) when lesser old items are removed or added (scripts, style, messages) [21:02:36] ah ok yeah that's what i saw last week; thanks! [21:02:41] yw [21:23:40] Reedy: https://gerrit.wikimedia.org/r/#/c/22880/ fixes a bug milestoned for 1.20 [21:23:53] ok for backport? [21:24:02] yeah, just go ahead [21:24:25] it seems more than a simple bug fix, it basically fixes the fatal by a refactoring [21:24:31] alrighty [21:25:38] doesn't apply cleanly [21:26:05] hm.. I wonder what else changed that file [21:26:20] there's probably a few dependant commimts [21:55:31] have a good night! [21:56:38] AaronSchulz: http://paste.tstarling.com/p/BDWOup.html [21:57:29] cruel and unusual cpu usage [21:58:50] asher was pinging me about duplicate removal earlier [21:59:08] I said I'd work on it once the job queue stuff is merged [21:59:38] right [22:00:10] 24 patchsets now? is that a record? [22:00:25] TimStarling: Can't possibly be a record. No way. [22:00:27] I feel like I've seen more somewhere, but it's certainly up there [22:00:41] I think I tried to get a diff with my gerrit-differ.php tool, but it timed out [22:02:35] call me sentimental, but I just like to be able to actually see the code I'm approving before I approve it [22:02:42] I'm sure it's fine [22:03:13] "I'm sure it's fine"? Really? That doesn't sound like the Tim I know [22:03:41] I've seen 30+ somewhere [22:04:26] Press rebase a few more times [22:05:14] lately I've been rebasing before I merge each commit, to make the history less convoluted when you view it in gitk or whatever [22:05:30] TimStarling++ [22:05:50] If they've got RELEASE-NOTES additions you've almost always got to do it.. [22:05:53] ++TimStarling is faster [22:05:59] in PHP [22:06:05] but I don't usually wait 5 minutes or so for jenkins to finish testing the rebase before I merge it [22:06:21] AaronSchulz: Hello from the Parsoid team, we rock out in JS and C++ [22:06:31] and yesterday I broke master merging a contenthandler change which tested OK before the rebase [22:06:41] AaronSchulz, wait, really? [22:08:32] yes [22:08:35] I thought the only difference was the order it returns in... huh [22:09:47] the schema change in the job queue patchset is conflicting every time someone changes the schema [22:10:43] easy enough to merge manually though [22:12:32] success, I have a diff from patchset 14 to 24 [22:13:02] "Oh well, the whole class is premature generalisation really" [22:13:07] nice to see you removed that comment [22:13:21] I wrote it when there was only one subclass, when we added the second one it became false [22:21:00] AaronSchulz: have you tested the non-MySQL branch? [22:21:13] in claim()? [22:22:33] hmm, I don't remember, 2 of my test wikis are wiki are sqlite and one at home is PG [22:22:52] maybe I should double-check those in case I didn't test them [22:23:52] + 'job_random' => mt_rand( 0, 2147483647 ) // [0, 2^31 - 1] [22:24:07] usually I write 0x7fffffff for that number [22:24:19] it's easier for humans to read [22:24:20] that looks prettier [22:26:35] PHP Notice: Undefined index: format in /var/www/CephWiki/core/includes/Import.php on line 664 [22:26:40] great, import is broken [22:26:49] -CREATE INDEX /*i*/job_cmd_token ON /*_*/job (job_cmd,job_token,job_id); [22:26:49] +CREATE INDEX /*i*/job_cmd_token ON /*_*/job (job_cmd,job_token,job_random); [22:27:15] never mind [22:28:51] I assume you used base 32 instead of base 36 because you wanted to be different? [22:29:13] I recall Roan pointing out that it fits perfectly [22:29:25] and then we wondered why we did it differently, or something [22:29:26] right, so it's logical too! [22:29:51] Aaron and I had ... some discussion about this, I remember it very poorly [22:30:07] It was kind of an idle-conversation-as-I'm-heading-out thing [22:31:11] I think it should be base 36 [22:31:28] consistency wins over logic [22:31:53] everyone will hate you for introducing yet another way to store hashes in the DB [22:32:05] a fourth, to add to the three already present [22:33:22] base 16, base 36 and binary (in the math table) [22:34:03] TimStarling: and we can use NUMERIC(49) too :) [22:34:14] good idea [22:34:25] TimStarling: meh, I don't have much preference [22:34:31] you can change if you want :) [22:34:54] * marktraceur votes for base 42 (because it's the answer) [22:34:54] 36 feels silly now [22:40:53] done, PS25 uploaded [22:41:00] * TimStarling doesn't mind feeling silly [22:41:36] great, now you ruined it, all the elegance sucked right out [22:44:18] ok to merge? [22:44:40] As long as we get the schema changes done before 1.21wmf3.... [22:44:55] I guess everywhere that isn't zhwiki wont take long [22:45:38] seems to be working for sqlite [22:45:46] excellent [22:46:06] I can recall if I tested PG, I guess I will when I get home [22:46:17] * AaronSchulz is too lazy to install pg on yet another box [22:46:18] even zhwiki only has 275k rows [22:46:25] ah [22:46:27] not so bad then [22:46:33] MySQL should be able to alter that in a few seconds [22:46:38] just truncate it and then add the columns/indexes ;) [22:47:22] where do I record the pending schema change? [22:47:33] it's already there IIRC [22:47:39] yeah [22:47:42] http://wikitech.wikimedia.org/view/Schema_changes [22:48:11] I guess the status can be updated to "merged" [22:56:00] TimStarling: maybe we can start working on duplicate detection now? [22:56:38] can i really just truncate zhwiki.job and then run the alter? plz oh plz [22:57:41] AaronSchulz: yes [22:58:14] binasher: ok [22:58:42] binasher: what about renames? [22:59:12] hmm, maybe just delete all the refreshLinks2 jobs and then wait for it to be cleared [23:01:24] | enotifNotify | 1 | [23:01:24] | htmlCacheUpdate | 174 | [23:01:24] | refreshLinks2 | 2754132 | [23:01:26] Haha [23:02:02] oh yeah, I guess I missed a digit when I said it was 275k [23:02:12] actually 2.75M [23:02:13] it happens [23:02:20] be glad you are not in finance [23:19:18] Template:Str_≥_len [23:27:07] Hmm, there are now 2790267 refreshLinks jobs [23:30:23] large jobs are split into 50 smaller jobs when they are executed [23:30:45] ah [23:30:47] so the number can increase very rapidly when the job runners hit a set of big jobs [23:30:53] Reedy: be aware of the job hydras [23:30:53] do it binasher, I dare you ;) [23:55:39] i wonder how long zhwiki.job will stay at 0 rows [23:57:42] | refreshLinks2 | 1 | [23:57:51] binasher not long [23:57:51] it begins