[17:30:52] Can I get a C+2 on https://gerrit.wikimedia.org/r/c/mediawiki/extensions/UploadWizard/+/634293 to unbreak the ability to merge into the repo? :-) [17:31:17] (I'll file a follow-up task to properly re-do the tests later.) [17:32:49] hello, I am the wrong preson to do this but let me poke someone about it [17:34:40] I have poked $someone. [17:35:51] i bonked it [17:37:08] I'm guessing we're going to be finding failures like that all over the place for a while [17:38:08] Thanks MatmaRex. [17:38:12] Reedy: Yeah. :-( [17:38:22] I have unpoked that $someone [17:38:45] can someone give me the tl'dr version? [17:38:45] someone should just run the extensions' tests locally before meging breaking changes… [17:38:59] ^ [17:39:07] I'll +2 it [17:39:18] apergos: basically, a change to core tests mean they now fail if they try to access the interwebs [17:39:35] all of them? [17:39:36] but extensions haven't been tested/verified for failures [17:40:08] whats the commit that did it? [17:40:11] apergos: Yeah, the core test framework now errors on HTTP request attempts. [17:40:20] https://github.com/wikimedia/mediawiki/commit/aea8cb17bd2a9bf873deebb7bd3022add58308a7#diff-59830ebc3a4184110566bf1a290d08473dfdcbd492ce498b14cd1a5e2fa2e441 [17:40:23] maybe these tests should be run now, at least give us a heads up [17:40:23] T262443 [17:40:24] T262443: Do not hit actual urls during phpunit tests - https://phabricator.wikimedia.org/T262443 [17:40:26] Id708cdfa3d3c27f30543c6bf212df96e6b0a69e1 [17:43:01] it's a good idea certainly but em [17:43:13] Yeah. [17:49:32] hey duesen [17:49:56] if you have a few minutes we have gotten reports of extensions failing because of https://gerrit.wikimedia.org/r/c/mediawiki/core/+/632825 [17:50:10] eg uploadwizard [17:50:23] would it be possible to run the tests across all the extensions to catch any other outliers? [17:50:33] apergos: hey! I'm on a train, my connection is wonkey, and irc isn't very smart about that. so if i vanish, you know why ;) [17:50:41] heh ok [17:50:49] are you on vacay now? [17:51:03] if so point me at "how to run those tests" and i'll try to muddle trough [17:51:05] through [17:51:22] hm at 9 at night, ah well that's what I get for replying in the channel :-D [17:51:28] taking a "family supprt" day today. i'm working tomorrow. actual vacation next week [17:51:46] ah [17:51:55] still you're off and that's the important thing [17:52:11] there isn't any thing special to run - the MediaWikiIntegrationTest base class now makes sure no HTTP requests happen. [17:52:17] Tests that do, fail. [17:52:25] So all that is needed is to run whatever tests exist. [17:52:54] that's what I don't know how to do, to run all the tests on all the extensions (remember how I'm not really a PHP dev... yet? :-P) [17:53:33] apergos: There's no real way. Just install every extension you can think of and go for it. Except lots of extensions (even WMF prod ones) trample over each others' tests. [17:53:40] (Yes, there's a task about that.) [17:53:45] ugh [17:53:49] I don't know how to do that either :P I can only tell you that it involves *installing* all extensions. Which is... not so easy. [17:53:52] Welcome to MediaWiki. [17:54:06] Yea. That's why I haven't don it ;) [17:54:12] I have the extensions that come with checking out core, of course they are not "installed" in any real sense nope [17:54:33] how about beta, might they all be there? [17:54:35] If only we were based on slim loosely-coupled services instead of a monolithic PHP blob of hooks… ;-) [17:54:48] We run core CI with quite a few extensions enabled, so all of them are already known to be ok. [17:54:53] didn't krinkle have some magic test environment [17:55:00] Yeah, core CI runs with ~30 extensions. [17:55:02] otherwise there's grep -r :-P :-P [17:55:08] Pretty much. [17:55:26] But e.g. the Math ones are breaking because the "real" code they're calling does indeed do HTTP requests to RESTbase. [17:55:29] Or just make no-op verify jobs and give jerkins some work [17:55:45] James_F: thanks for looking into the uploadwizard thing btw. is there a patch that needs merging? [17:56:07] Reedy: ok, that's actually a pretty nice way to do it! [17:56:35] I know hashar does it occasionally for some stuff [17:56:51] duesen: No worries, already handled. [17:56:53] apergos: as Reedy suggested, it's enough to make a nearly-empty path for each test repo with a "do not merge" comment. CI will run tests on it and tell you if there is an issue. [17:57:11] We should almost have a pre-generated script for that [17:57:14] "for each test repo" [17:57:15] (Please don't overload CI at important times, though. ;-)) [17:57:19] yeah sounds like we should [17:57:21] though, git submodule foreach shouldn't make it hard [17:57:31] As above, I imagine Antoine probably has it [17:57:39] Or can we just run libup and see what breaks there? :P [17:58:26] Did we find any other tests besides UploadFromUrlTest that was *actually* doing HTTP requests to real hosts? [18:00:11] 184 extensions with mw core, of which grep -r Http:: .| grep -i test gives only ApiFlickrBlacklistTest.php and two false positives [18:00:31] hm I woner how up to date my copy of those is though [18:00:35] *wonder [18:00:54] https://codesearch.wmcloud.org/search/ [18:02:19] they probably wouldn't do the actual HTTP calls directly from the test code, but rather from the code being tested [18:02:49] so probably it's not easily greppable like that [18:04:17] can't get te regex for the path to work anyways [18:05:05] Yeah. [18:14:26] when you can't get submodules to work properly, for file in $dirs (cd in, git pull and get out) [18:14:47] I swear 99.99% of my tiem is spent figuring out a workaround for something I don't know how to do or is broken [18:20:23] Reedy: can you remember to ash hashar about the no-op across all extensions thing if I forget? [20:11:51] Running libup is tbh the easiest way to find all these failures [20:12:05] We're due for a new codesniffer release anyways