[00:00:48] has anyone tried making a PHPStorm project with many MW extensions? I assume it will croak, one way or another, if you try to make a project with 2 million lines of code in it [07:36:08] <_joe_> TimStarling: interesting find on the GC contention. [07:43:46] TimStarling: I have 21 extensions in my project at the moment, seems fine [07:52:24] legoktm: I'm around if you want to chat about the ExtensionsRegistry stuff (just submitted another patch) or we can keep the conversation in Gerrit, either way [14:26:56] TimStarling: Re phpstorm... It works fine with all the WMF hosted extenions [14:27:12] On numerous occasions JetBrains have checked them all out to reproduce and fix perf issues with this setup [14:27:51] At worse, you might need to bump memory options a bit, but it's a while since I've had to do that, and have mostly been using stock settings for numerous versions [14:30:49] I do turn off a load of plugins that I don't need though, which may make a bit of difference [14:38:01] It's set to 2048MB for max heap size for me [15:13:23] James_F: hey! can you get me on the Staff channel? [15:13:39] I think we tried this again, but apparently an invite wasn't enough, and then I forgot about it [15:13:45] duesen: I have, aheas ago. [15:13:50] RoanKattouw: https://gerrit.wikimedia.org/r/c/mediawiki/core/+/528824 [15:13:56] Also, ages ago. [15:14:03] * James_F will poke again. [15:14:28] James_F: doesn't work, it'S telling me the channel is invite only [15:15:38] duesen: Try now. [15:17:06] James_F: works now, thank you! [15:17:09] what was the issue? [15:46:29] duesen: No idea, sorry. [15:47:23] * anomie runs into tests/phpunit/unit/ for the first time and ughs over having tests for a component split across two directories [15:49:23] It used to be that if I'm poking at e.g. includes/Rest/, I could run phpunit over tests/phpunit/includes/Rest/ and tests would be run. Now I have to do both that and tests/phpunit/unit/includes/Rest/. In two separate runs because phpunit only seems to allow one path on the command line. [15:54:05] anomie: that'S annoying, but on the other hand, you can now run tests against tests/phpunit/unit/includes/Rest/ without having MediaWiki set up at all [15:54:31] if all tests for the component were in the same place, that wouldn't be possible [15:54:58] duesen: I can't see myself every needing that use case. [15:55:03] s/every/ever/ [15:55:04] having the tests split by test runner isn't great, but acceptable, i think. i mean, we also put qunit tests and api integration tests and selenium tests into different places, because they use different rest runners [15:55:36] This is the same runner, though: phpunit. The only difference is the parent class of the test classes. [15:56:09] anomie: you, personally, by hand - no. But the pure unit tests are a lot faster. CI now runs them first, without any DB setup and stuff. If they fail, it doesn't bother to run the integration tests [15:56:25] that generates feedback much more quickly in the negative case [15:56:46] anomie: it'S not the same runner. one uses actual phpunit, the other uses fake mw flavored phpunit [15:56:54] Doesn't the CI normally work by @group annotations than subdirectory though? [15:57:13] that i don't know [15:57:23] would have been possible to do this by group as well, i suppose [15:57:39] though that's probably harder to make work with `composer test`. [15:58:58] anomie: hack mediawiki's phpunit runner so it accepts multiple pathes? then you can do tests/*/includes/Rest, or something [15:59:43] Probably I'm just going to have to get used to the annoyance. [16:02:23] James_F: can you help me to get this backport in? https://gerrit.wikimedia.org/r/c/mediawiki/core/+/529442 [16:02:47] If I read the test failure correctly, phpcs is broken, or incompatible with the php version [16:03:35] I'm confused about why this happens now, and didn't happen before. can we somehow ignore that? [16:14:40] duesen: It's legoktm's code being clever; it's a lint failure under php73 but not earlier. So release-quibble-composer-mysql-php70-docker passed but release-quibble-composer-mysql-php73-docker failed. [16:18:42] duesen: Also, you really shouldn't make back-ports of code that hasn't merged in master unless it's an emergency, as now old branches and master will drift from each other. :-P [16:19:48] Oh, right, the failures on REL1_31 which is testing 73 for some reason? [16:20:09] * James_F sighs. If only there was an RfC in front of TechCom to make this simpler. ;-) [17:29:26] James_F: as far as I can tell, it is merged on master? Am I missing something? https://gerrit.wikimedia.org/r/c/mediawiki/core/+/505351 [17:29:52] I'm blurry on the RFC, to be honest [17:35:57] duesen: It was, it was just cherry-picked before it had been merged in master. I fixed the pick and force-merged it for now. When I'm back in the office I'll investigate. [17:36:10] duesen: And a Final Call on the RfC would be smashing, if you can swing it. ;-)( [17:38:20] James_F: the original cherry-pick wasn't mine, when I went to create one, I found that one already existed. [17:39:12] * James_F nods. [17:39:57] James_F: you mean this one? https://phabricator.wikimedia.org/T228342 [17:40:45] Yup. [17:42:19] James_F: on a related note, there will be an IRC meeting on improving the RFC process tonight [17:43:55] I put the RFC up as requesting last call. Last call could start next week, per the current (newish) process of giving one week for internal review. [17:44:17] Right. I might not make that meeting, sadly. [19:29:53] anomie: can you take a look at https://gerrit.wikimedia.org/r/#/c/mediawiki/core/+/529199/ ? [19:31:35] AaronSchulz: I can't right now. I see it's on the list for someone from CPT to theoretically look at. [21:51:10] thanks Reedy and kostajh [21:52:24] TimStarling: They're pretty good with complex bug reports too if you find some weird edge cases etc... Seems I've got 930+ extensions checked out currently [21:52:35] good to know [21:53:15] I see they hired nikic recently, I was wondering who was going to grab him since he was finishing his studies [21:53:39] he's probably the smartest PHP core contributor [21:54:20] so evidently there are some actual brains at JetBrains