' [22:14:27] yep, better [22:18:07] mdholloway: https://gerrit.wikimedia.org/r/#/q/topic:T186927+(status:open+OR+status:merged) [22:18:07] T186927: mw-parser-output divs leaking into mobileview output again - https://phabricator.wikimedia.org/T186927 [22:34:58] eh, probably not a good time to wait for CI [22:37:24] tgr: yeah, the MF tests are quite slow [22:44:45] mdholloway: do we need to purge MCS or RESTBase as well? [22:45:31] the fix should be live btw [22:45:35] tgr: checking now [22:48:06] mhurd: is the iOS app getting extracts from RB these days or still from MW API? [22:50:54] we may want to consider purging/regenerating /page/summary in RB as well /cc bearND Pchelolo [22:51:08] https://en.wikipedia.org/api/rest_v1/page/summary/Cragside is still giving bad extracts after a varnish purge [22:54:26] ah, looking better now [23:01:06] mdholloway: I seem to recall RB getting a few minutes expiration time after the problems with featured X vandalism [23:03:31] it actually seems to be taking a minute for the TextExtract fix to come through, even from MW API, though I'm not sure why that would be [23:03:33] e.g., https://en.wikipedia.org/w/api.php?action=query&prop=extracts&format=json&formatversion=2&exsentences=5&exintro=true&titles=2018_Winter_Paralympics [23:03:55] (i just purged that specific URL, and still not fixed) [23:04:03] same with Cragside though that's better now [23:08:37] weird [23:09:22] these are not cached in varnish, and the code change should have been immediate [23:09:29] I don't see any other caching layer [23:09:45] tgr: and now that's suddenly fixed [23:10:01] tgr: oh, those aren't cached even for GETs? [23:10:25] no, the MW API is not, there would be no way to invalidate it [23:10:37] you can request caching manually via &smaxage=... [23:13:53] so, do you think it is fixed for good? [23:14:02] i'm just clicking around in the Android app. every now and then i find an article with an extract that hasn't been regenerated, and it's coming from the MW API. when it finally comes through RB seems to update immediately fwiw [23:14:30] tgr: still finding faulty extracts: https://en.wikipedia.org/w/api.php?action=query&prop=extracts&format=json&formatversion=2&exsentences=5&exintro=true&titles=Seoul_Capital_Area [23:14:59] the good news is that the extracts are coming back, albeit gradually... [23:17:14] tgr: interesting... 2018 Winter Paralympics was edited right around the time the extract suddenly came back https://en.wikipedia.org/w/index.php?title=2018_Winter_Paralympics&action=history [23:18:15] not true for Cragside, though... so much for that theory [23:20:16] the cache version is still 1 on that box [23:20:24] did something go wrong with the deploy? [23:21:22] hmm :[ [23:23:21] oh duh [23:23:30] I forgot the submodule update didn't I [23:28:19] tgr: that's better! [23:28:30] sorry about that :/ [23:29:29] ha, no worries [23:29:56] as long as the underlying extracts are back we can probably wait for stuff to age out of RB IMO [23:30:08] there don't seem to be a large number of articles affected [23:34:37] tgr: thanks again for fixing this up for us. time for me to wander away from the computer. [23:34:39] have a good weekend! [23:34:57] you too [23:35:28] @mdholloway sorry i missed your question - ya we're using RB extracts [23:36:30] mhurd: that's ok! seems like RB is still catching up on a few summaries but is everything looking better overall? [23:37:40] huh, Crust_(geology) is semi-protected for some reason [23:37:48] very controversial apparently [23:38:45] @mdholloway ya the article i 1st noticed the issue on looks good! others too! [23:38:49] the android app seems in pretty good shape [23:39:25] @mdholloway @tgr thx so much for such quick fixes! :) [23:40:17] mhurd: awesome. tgr gets the credit for this one. i'm off, have a good weekend! [23:40:22] sorry for breeking it in the first place :) [23:40:57] @mdholloway thx you too! [23:41:34] @tgr hehe no worries! [23:44:02] Thanks tgr! [23:51:30] everything seems to be not on fire so I'm off to find some food [23:51:38] I'll see IRC pings