[16:16:34] lgaulia: I'll see if there is a MediaWiki::main => Action/SpecialPage/OutputPage => Skin diagram... [16:24:37] TimStarling: speaking of pgsql: [16:24:38] > DO NOT USE ! Simple database layer for postgresql based on existing mysql one. I dont think it even works but I want to share the ongoing work on this. [16:24:54] http://mediawiki.org/wiki/Special:Code/MediaWiki/3908 [19:20:50] do we have a Krinkle around? https://gerrit.wikimedia.org/r/c/mediawiki/extensions/FlaggedRevs/+/824270 UBN train blocker [19:21:30] Amir1: checking [19:28:18] Krinkle: thanks. The failure is unrelated [19:28:22] I try to fix core [19:28:31] (for the phan failure) [19:39:34] it was in the extension itself [19:39:35] https://gerrit.wikimedia.org/r/c/mediawiki/extensions/FlaggedRevs/+/824272 [22:30:45] thanks for the merges overnight [22:30:53] my modtoken patch https://gerrit.wikimedia.org/r/c/mediawiki/core/+/823793 is AFAIK the last remaining blocker for re-enabling multi-DC [22:33:01] TimStarling: I'm trying to write an integration test that e.g. calls to SqlBag instances with differently mocked time and e.g. see that the later time wins even if it was called first. But then meetings happened and end of day. Perhaps tomorrow, or perhaps you'd like to give it a try? [22:33:07] two* [22:36:04] sure, better than nothing [22:39:27] I mean, good to at least have a regression test even if we don't test every party of Aaron's essay on line 1184-1189 [22:39:45] s/party/part/ [23:08:37] ok, I added a regression test, and confirmed that it fails with the old SUBSTR() comparison [23:22:59] I am scratching my head trying to figure out the best way to detect a ~5% latency change on a subset of servers in production, for scaling_governor T315398 [23:22:59] T315398: Set MW appserver scaling_governor to performance - https://phabricator.wikimedia.org/T315398 [23:39:41] I'll probably use the mediawiki_http_requests_duration_sum metric, but I'm going to fiddle with some existing dashboards first to get a feel for it [23:57:06] Yeah that's the one I'd go for, assuming that's what we plot on the RED dashboard latency rows