[14:00:21] Krinkle: Do you have thoughts on this? https://phabricator.wikimedia.org/T230211 [14:00:35] specifically deferred post-send updates https://phabricator.wikimedia.org/T230211#5534081 [14:02:55] duesen: normal users won't have CP wait for post-send updates, either. Is there an example of something you want to assert? [14:03:23] Krinkle: entries in recentchanges after an edit, for instance. [14:03:44] might be easier to just use a single-db rather than replica. At this rate, there is not much added value of using a replica if we always wait and disable most of CP. [14:03:49] duesen: Thats pre-send. [14:03:53] Krinkle: generally, there needs to be a way to assert *all* the things we do post-send. there would just be no way to test them otherwise [14:04:15] I just janked out a 30s wait loop from Wdio tests for the same reason [14:04:21] was also waiting for recent changes [14:05:55] Yeah, I understand the general idea, but I'm suspicious of specifics. Seems to me that either most things would be pre-send that matter to end-users. Or else, it's something that either shouldn't be asserted (imho) or might be better to assert by e.g. just waiting 1-2 second blindly and requiring that to be enough. [14:07:36] Krinkle: that makes for really really slow tests. We weant to have nots of API tests. thousands [14:07:55] Krinkle: i'm in a meeting now, perhaps let's have this conversation on the ticket [14:13:27] duesen: done [16:15:56] Krinkle: so, you would prefer a config setting to disable post-send over a request header? [16:16:35] That would work, the thing I don't like about it is the fact that them, tests won't run reliably against a default install. YOu have to change config to make tests pass. [16:16:59] duesen_: Fair, but you're also trying to make it behave in non-default ways :D [16:17:18] Will be simpler after https://gerrit.wikimedia.org/r/#/c/mediawiki/core/+/538362/ [16:17:21] which abstracts this already [16:18:00] Krinkle: my proposal is to make it standard for clients to be able to wait for deferred updates... [16:19:06] I should probably survey the uses of deferred updates [16:19:15] I fallback to needing concrete use case. [16:19:20] post-send ones, i mean [16:19:44] I'm having trouble seeing a use case for it that is appropriate for testing but not real-world. [16:20:20] deploying now, then afak for 1-2h. Will follow on task. [16:21:21] Krinkle: I'm back to looking at RC. notifyEdit() calls $rc->save() in a POSTSEND callback. [16:21:42] am I missing something [16:21:59] oh, ok. I also have to go to a meeting. [20:38:02] davidwbarratt: no rush, but as I understand it https://gerrit.wikimedia.org/r/#/c/mediawiki/core/+/534933/ is currently awaiting your feedback wrt to the trackBlockWithCookie() logic change. [20:38:27] I've got it around hte top of my list to finalise, so do let me know where I can help :) [20:41:01] Krinkle ah! thanks!