[16:17:19] legoktm, Reedy, or anyone: I have a doc fix in https://gerrit.wikimedia.org/r/#/c/380541/ that could use a quick typo-check and merge. [16:51:03] anomie: Reedy: Could one of you confirm regarding https://gerrit.wikimedia.org/r/#/c/370190/ / https://gerrit.wikimedia.org/r/#/c/367698/ toward EddieGP whether it's true or not that the other DBs didn't need a similar patch? I don't think it was wrong per se, but would be nice to maybe reply and acknowledge e.g. yes they still need it (TODO), or no, there's a good reason for this particular case not doing it. [17:03:49] Krinkle: In practice, no one requires schema changes for anything except MySQL/MariaDB and Sqlite, because we don't run unit tests against anything else. Sometimes it's hard to do so even if someone wants to thanks to previous missed schema changes. We may eventually add PostgreSQL to that list if T164898 goes somewhere interesting, but I doubt MSSQL or Oracle will ever get there because we probably can't put them in CI to block merges. As for [17:03:50] 370190 specifically, the primary motivation is because some MySQL/MariaDB maintenance tools require an actual PK instead of just a unique key that could be a PK. Theoretically other DBs could have used the rename too so the USE INDEX in ApiQueryLinks would work, but all of them ignore that option (and don't name the indexes the same as MySQL anyway) so it doesn't actually matter. [17:03:50] T164898: PostgreSQL schema change for consistency with MySQL - https://phabricator.wikimedia.org/T164898 [17:27:41] anomie: lgtm +2, but my laptop isn't turning on :( [17:28:40] legoktm: I self-+2ed based on that comment, thanks. [17:38:17] anomie: also could you take a look at https://gerrit.wikimedia.org/r/#/c/379724/ ? [18:00:13] legoktm: done [18:08:05] The above explains why suddenly https://www.mediawiki.org/w/index.php?title=Development_policy&type=revision&diff=2570658&oldid=2542099 [18:45:38] Nemo_bis: Hm.. good point. [18:45:52] anomie: Aye, we don't have CI indeed, but I've always thought we do require the patch to cover them all. [18:46:04] At least Postgres is continuously tested via Travis CI post-commit [18:46:19] and breakage there has been a reason for UBN and blocking release or reverting bad commits. [18:46:34] anomie: it'd be theoretically possible to get CI for mssql now that they have a linux version (or will in a week), but that would require me to stop being lazy and get that schema up-to-date with all the other missed patches before it's a feasible option. And, I'm not sure if there's a free version for linux or not [18:46:36] At least that's what I've done so far. [18:46:56] Although I suppose in most cases that just means blind patch writing. [18:47:19] Which isn't very different from what we do for sqlite, given we don't use that in CI for master, either. [18:47:26] Krinkle: I noticed that PG doesn't have the recent ip_changes patch, for example. [18:47:35] that's mostly the issue I see, the dbmses are different enough from each other that some patches need more than just trivial changes [18:47:39] even if it passes CI [18:47:56] Do we not run sqlite in CI anymore? [18:48:02] Nope, not for years. [18:48:07] except in some wikibase jobs. [18:48:50] We switched from sqlite to mysql (on tmpfs for speed) after we broke mysql too many times and to make it more prod-like. It was only sqlite in the first place (afaik) for speed reasons. [18:48:55] That happened in 2013 or so. [18:49:23] I knew we did that, but for some reason I thought we had still kept tests against sqlite too. [18:49:36] Adding postgres within wmf ci would be nice, but it would be the first time we test mutiple DBs pre-merge. [18:50:02] I guess we could limit it to just the plain core phpunit job. [18:58:56] anomie: i've just discovered that the JS enhancements for the timezone field in MW preferences were originally written by you. i guess you had the same meticulous attention to detail no one else bothers with nine years ago as you do now :D [19:00:04] MatmaRex: Making the timezone field in preferences use zoneinfo rather than manual offsets that had to be manually updated for DST twice a year was one of my first big patches to MediaWiki. [19:00:05] anomie: people introduced a few bugs into that code in various rewrites over the years. want to review https://gerrit.wikimedia.org/r/380570 ? ;) [20:33:47] anomie: no_justification: Someone familiar with multiversion/mwscript - Review on https://gerrit.wikimedia.org/r/#/c/378007/ would be appreciated :) [20:35:10] looking [20:42:43] +1 [20:56:17] Thanks!