[00:03:29] yay when opensource works [00:03:30] https://github.com/JamesHeinrich/getID3/pull/181 [00:09:20] Reedy: nice, now to get a release as well? [00:09:40] They've only just done one.. :P [00:09:50] That's how I knew they had some more stuff creeping in [00:13:58] I'll rebase the rest when that one merges [00:14:47] k [00:15:03] Looks like the vendor test suite now doesn't check core's composer.json anymore, which makes things a lot easier. [00:15:16] So that one can land safely in the queue, and then the core (with depends-on) right after it. [00:15:19] no more force merging [00:15:23] or circular deps [00:15:41] Just gotta make sure to not merge anything in vendor without the core patch [00:15:58] yay for workflow not sucking too badly [00:35:11] It [00:35:13] Bah. [00:35:14] It [00:35:28] It's been this way around (vendor not checking, core checking) for a few years. [00:35:52] It means it's possible to totally break all MW repos by merging something in vendor without the corresponding patch in core, which sucks. [00:36:15] But better than the previous experience of people habitually force-merging patches in both. [00:37:03] (We still /have/ a circular dependency, we just have a Sword of Damocles which cuts the Gordian Knot and sometimes our fingers, to abuse the metaphors beyond belief.) [00:39:01] I just stacked the next 3 vendor patches [00:39:06] I saw. [00:39:18] :) [00:46:21] Krinkle: Sorry, /vendor merging broke the cherry-pick merge of Iee44c2e51e2 to wmf.16 because, well, /vendor and wmf branches... [00:46:49] ic, np [00:47:56] Reedy: Is https://gerrit.wikimedia.org/r/c/mediawiki/vendor/+/346351 going anywhere? ;-) [00:48:13] lol, probably not [00:48:26] Of course, the people that added the functionality to core et al have no interest in maintaining it [00:48:38] Most of them aren't around any more, I believe. [00:54:08] yay techdebt legacy [01:00:12] /var/www/wiki/mediawiki/vendor$ composer outdated | wc -l [01:00:12] 27 [01:01:12] See if I can tackle another 5-10 of those later today [01:01:20] Nice. [01:01:37] At least 5 are wikidata related [01:02:02] gd-text gets to go away when Zero dies [01:02:13] ^ James_F Make sure you have that on your list of things to clean up when we undeploy :) [01:02:38] Reedy: Phabricator task or it won't happen. ;-P [01:03:01] lol [01:05:24] https://phabricator.wikimedia.org/T215745 [01:05:54] Probably in the wrong place in the tree, but meh :P [01:08:11] * James_F grins. [01:29:00] Krinkle: https://gerrit.wikimedia.org/r/c/mediawiki/extensions/AbuseFilter/+/489125 probably is fine to land now, if you're around to un-dirty production by syncing it. :-) [01:38:57] James_F: I don't see that patch deployed or checked out on deploy1001. [01:39:33] Krinkle: Yes, but you C+2'ed it. [01:39:42] So presumably you planned to do so? [01:39:57] No, and I didn't :P [01:40:17] I removed a +2 [01:40:30] The one by Tim from 3 days ago, so that it wouldn't silently merge on Sunday. [01:40:30] Oh, never mind, I didn't read gerrit's output closely enough. :-) [01:40:33] Yeah. [01:40:50] I'll let him reland tomorrow . [01:45:29] I mean, it is already Monday for him. :-) But yes. [01:57:47] https://github.com/cssjanus/php-cssjanus/issues/14 [01:57:55] ^ we should probably look at that for our PHP 7 support [01:58:00] "Yeah, the workaround that initially seemed to work didn't do the trick in the end. We fixed our problem by forking cssjanus and applying the patch in #13." [02:00:05] https://github.com/cssjanus/php-cssjanus/pull/18 adds some .gitattributes too :P [02:23:45] 3 more vendor patches up [02:24:21] It seems we've got some dependancies for at-ease and assert that require other depedancy bumps too, so leaving those for now, but made a css-sanitizer patch to allow us to upgrade at-ease [02:27:37] Last commit 9 months ago by Krinkle :P [02:30:24] https://phabricator.wikimedia.org/T215746 filed for PHP 7.0 support [02:30:30] Probably needs tagging on that task tree [03:00:39] I think zuul died [03:06:58] zuul reloaded [03:16:17] James_F: There's another ~10 patches to go ;P [03:16:41] There always is. ;-) [03:21:42] Reedy: https://gerrit.wikimedia.org/r/c/mediawiki/vendor/+/489596 needs a corresponding core patch. [03:22:14] It's not in cores composer.json [03:22:39] It's a dependancy for psysh, but it was being weird, so I removed it out of the other bump patch I did for it [03:22:48] Oh, OK. [03:23:16] It's a bit confusing, fo sho [03:23:23] But we should pin. [03:23:41] Future problem. [03:24:22] Hmm, hang on [03:24:40] No, we shouldn't merge that one [03:25:10] We seem to put "jakub-onderka/php-console-highlighter": "0.3.2", in numerous libraries directly [03:25:21] Yes, it's part of the standard CI package. [03:25:41] Should pull that one out of the chain then [03:25:50] Un-C+2'ed. [03:26:02] But if it was going to have caused problems it'd have failed gate, I imagine. [03:26:22] (Do we gate with composer, or just unit tests? I forget.) [03:50:13] Hmm, ipset [03:53:29] I hate that composer goes nah, can't resolve one dependacy, and gives up [03:53:33] You fix that, and then try again [03:53:39] And it's like, nah, can't resolve that either [03:58:52] I removed a +2 [03:58:52] The one by Tim from 3 days ago, so that it wouldn't silently merge on Sunday. [03:58:57] it's not sunday here [03:59:30] I'm trying again to merge it so I can run the script with --dry-run [04:06:06] looks like it should go through this time [04:07:22] I will try to stay on task this time and not forget about it and go do something else for the third time :) [04:08:02] merged! [07:08:03] cheers James_F [07:14:25] Reedy: Eurgh, EventBus patch died. [07:15:26] Looks like upstream npmpackage release or something [07:16:06] Npm flakes from time to time. [07:16:23] At our scale, it’s a few times a day. [07:16:34] For normal people, it’s once a year. [14:12:03] James_F: Re the API stuff: Looks like we hard-deprecated those in 1.33, I486c4ccab. So the deprecation policy wants us to wait for 1.34 at least, it looks like. Other than that, removing them is fine with me. [14:17:55] James_F: There's no need to wait for T183420 to clean up those deprecated User-class methods, and I doubt we'll ever actually do T183420 in a way that works for generic LAMP-stack third-party installations, and outside of unit tests it seems unlikely to reliably replace existing uses. I note that search is picking up calls to non-User methods also named setPassword(), BTW. [14:17:57] T183420: Authentication data should not be available through the normal DB abstraction layer - https://phabricator.wikimedia.org/T183420 [14:18:05] James_F: You could also look at finally killing AuthPlugin if you want. [14:59:05] Right. [14:59:11] Thanks. :-) [19:19:14] * anomie got nerd-sniped by T215746 in the backscroll here [19:22:56] nice [21:40:34] yo [21:42:44] Hey eprodromou. [22:02:19] This was a rich social interaction [22:03:30] Time to hide for the rest of the week now? [22:15:46] 260 [22:15:48] Hello [22:17:01] hi dereckson