[05:55:42] <_joe_> legoktm: anom.ie found the culprit, probably [05:55:50] <_joe_> https://gerrit.wikimedia.org/r/plugins/gitiles/mediawiki/extensions/TimedMediaHandler/+/e6356ea7384daf9528de292cb701680cd16e7d1f/includes/TimedMediaHandlerHooks.php#590 [09:39:38] :/ [09:39:44] Has anyone fixed it yet [09:40:26] doesn't look like it [09:42:15] https://gerrit.wikimedia.org/r/#/c/mediawiki/extensions/TimedMediaHandler/+/520395 [09:42:21] Dunno if there's a bug to attach [09:44:42] Ah T226840 [09:44:42] T226840: Consistent HTTP 503 Varnish Error on some urls for some logged-in users (CentralAuth Set-Cookie storm) - https://phabricator.wikimedia.org/T226840 [10:46:12] <_joe_> Reedy: I would say there should be some protection against instantiating a new instance [10:46:23] <_joe_> of RequestContext, I mean [10:47:40] <_joe_> but I see it's used elsewhere [10:48:35] <_joe_> ugh :P [11:37:40] sounds about right :/ [11:37:50] I mean, there's probably a handful of cases it's useful, but beyond that... [12:00:13] in most cases it wouldn't cause problems beyond a small performance hit due to loss of instance caching [12:01:02] except the session object automatically fixes partial sessions (token cookie but no session cookie) on creation [15:42:36] It uses global state, so a new instance is logically incorrect. [15:42:43] Split brain etc [15:42:57] It was migrated from wgUser [15:43:04] But should have used getMain [15:43:28] We can make it private in the next release. [15:43:35] Fun. [15:43:52] Derivative or faux context is fine to create. A new global one makes no sense. [15:44:16] <_joe_> there are a series of such uses in core [15:44:41] Also... the same TMH code *still uses wgUser* [15:45:05] :facepalm: [15:52:59] reedy@deploy1001:/srv/mediawiki-staging/php-1.34.0-wmf.11/extensions$ grep -R wgUser * | wc -l [15:52:59] 749 [15:53:01] Krinkle: yw [15:53:25] AF, CA, CN... [15:56:24] Start with CA, because no-one else dares to.