[01:37:53] I've moved @wmf-apps-ci to be a member of the wikimedia/mobile team. If there are any additional repos it needs access to, let me know :) [09:22:16] hashar: morning [09:22:59] hashar: is there a way to bypass the unrelted failures mentioned in https://phabricator.wikimedia.org/T117710#3047699 ? Happening to me in https://gerrit.wikimedia.org/r/#/c/339126/ on the hhvm job https://integration.wikimedia.org/ci/job/mwext-testextension-hhvm/38083/console [09:23:10] on the gate-and-submit phase [09:24:13] joakino: good morning [09:24:25] yeah let me clear that one out right now [09:24:30] until I find the proper solution [09:25:16] hashar: no rush, these patches are not critical [09:27:37] joakino: yeah but CI randomly failing on bunch of patches is critical :-} [09:27:40] I cleared the skins [09:27:52] and I guess I am going to migrate the jobs [09:28:52] salt -v '*slave*' cmd.run 'rm -fR /srv/jenkins-workspace/workspace/mwext-testextension*/src/skins/*' [09:28:54] salt is magic :) [09:36:00] haha [09:37:29] i'm in tracy [09:37:43] woops [09:38:12] hashar: hhvm and php passed on the gate-and-submit 👍 [09:38:37] gonna monitor the followups [09:44:37] joakino: so the issue is [09:44:59] some mediawiki/skins/* repositories have an experimental pipeline that runs some mwext-testextension* jobs [09:45:13] aha [09:45:16] those mwext-testextension* jobs keep their workspaces between builds [09:45:28] so if one check experimental on the skin , the job keep that skin behind [09:45:39] ooh [09:45:53] why doesn't it clear it? [09:46:29] fault :D [09:46:44] I will migrate those jobs to the instances that get deleted once build is completed [09:46:55] I am tired of fighting those dirty states [09:47:08] yeah [09:47:23] get rid of them! [09:47:27] thanks for your help! [10:18:08] I wish I could but not this week sorry :( [10:18:14] unless you are blocked on something [10:51:56] so summary would be: fill some task, lets write down the workflow of what need to be achieved [10:52:03] then come out a few use cases scenario [10:52:13] pass those scenario through the workflow and refine until happy [10:52:25] then we can look whether it is worth implementing [10:52:50] the CI (Jenkins/Zuul) I am not worried about it beside the potential race condition when patches merges in a short sequence [10:53:11] I am more worried about how to fit in the system that cut the wmf branches out of master and assumes that extensions are in repo named mediawiki/extensions/Foo [10:53:43] the wmf branching script cut from master, but I am pretty sure it supports cutting from a different one (eg: deploy , stable or whatever) [10:54:07] I don't think the wmf branching script would support mediawiki/extensions/Foo/deploy out of the box [10:54:50] joakino: phuedx ^^ summary :} [10:56:24] hashar: yep, that's it [10:56:50] feel free to reuse some of those bits above in the task [10:56:52] :} [10:56:53] hashar: for the quick merges, before the build script, kill other build scripts running? [10:57:15] what happens in Zuul is [10:57:32] when you CR+2 two changes in a serie lets say changes A, B, C [10:57:41] it put them in a queue A -> B -> C [10:57:44] A is tested [10:57:52] B is tested AS IF A already got merged [10:58:08] so to test B, Zuul get the tip of the branch, merge A and merge B on top of it [10:58:24] and to test C Zuul merges tip of branch + A + B [10:58:32] so C is tested as if A and B already merged [10:58:50] that prevents conflicts/race conditions between patches [10:59:00] in the above example, if A happens to fail [10:59:21] the jobs for B get cancelled. Zuul attempts to merge B agains the tip of the branch (which does not have A since it failed) [10:59:22] and test tip of branch + merge of B [10:59:36] the jobs for C get cancelled. Zuul attempts to merge tip of branch + B + C [10:59:51] and run the jobs of C again as if B had been merged (but without A, it has been dropped) [10:59:57] that is for when you CR+2 [11:00:06] when commits get merged [11:00:21] Gerrit send event to Zuul as the patches get merged [11:00:26] so it would receives: A merged, B merged, C merged [11:00:42] we can then have a build-asset job that is made to NOT run concurrently [11:00:50] which hopefully will build against A , then B then C [11:01:14] but I don't think there is a guarantee that the build job for B will actually happen before C [11:01:23] so you might end up with assets from B instead of C :( [11:01:40] other ways is to have a Jenkins job that is not reacting to merge events. It would instead continuously poll the git repo [11:01:52] notice that A, B, C got merged and build the tip of the branch which is C [11:02:01] that would guarantee you build against the latest tip of the branch [11:02:13] (lot to digest sorry) [11:50:32] thanks hashar, good to know [11:50:45] i'll write the task and we can discuss problems on approaches there [11:50:58] sure :} [12:04:27] joakino: I guess it can be made a subtask of "Use bundler to generate Popups JavaScript code" https://phabricator.wikimedia.org/T156333 [12:04:29] seems related [12:04:32] @Krinkle re: wmf-apps-ci, limiting to just the mobile team is fine. I also enabled 2FA on the account. [20:22:41] joewalsh: Thanks! [21:46:31] bmansurov: phuedx|afk olliv raynor CAKE!! [21:46:45] where ? [21:47:21] I ate too many doughnuts today but there is still some space for cake !!!! [21:47:46] just saw new page previews 👏 [21:48:08] so no cake ?? [21:48:10] ;/ [21:48:15] we need cake [21:48:16] now [21:48:18] so fancy, much shadows! [21:48:20] <3 [21:51:02] https://68.media.tumblr.com/392acfa2620afe89e9a372d4cac4163e/tumblr_inline_noexxs4pvD1sqf2ij_500.gif [21:56:06] niedzielski: mdholloway: OK! So it looks like we'll have the go-ahead for pushing to production by mid-to-late next week. [21:56:13] Do we want to release a Beta from master now? [21:56:30] nice! [21:56:36] * mdholloway takes a quick peek at gerrit [21:56:37] ideally, i'd like to get this in, too: :) https://gerrit.wikimedia.org/r/339426 [21:57:02] dbrant: i was just looking at the revision [21:57:19] there's one other slight (possible) UI regression i noticed [21:57:45] on pre-marshmallow devices, on languages with no news card, there's a pronounced "bounce" effect upon refreshing the feed [21:58:05] to the point where the search bar/card is mostly out of the viewport [21:58:36] even on master, on pre-marshmallow devices, there's a *slight* bounce i notice now, but with this patch it's more pronounced [21:58:53] let me check real quick [21:59:23] i was using tajik and ukrainian as my example languages just because i first noticed on those [21:59:44] it's probably not a deal-breaker but a little odd [22:01:05] can't say I'm seeing it... [22:02:54] hmmm... i'm reproducing it consistently. all on emulators. let me try on a physical device... maybe it's an artifact of touchpad slop or something [22:03:19] i'm testing on emulators, and don't see it [22:03:32] wanna hop on a hangout? [22:05:07] i've reproduced it on API 16, 21, and now 17 [22:07:11] mdholloway: sure, batcave [22:07:56] dbrant: i'm there [22:15:33] dbrant: niedzielski: https://gerrit.wikimedia.org/r/#/c/338107 would be a nice one to get into the next beta, if possible (it's required IMO for us to consider T139377 complete) [22:15:34] T139377: [Feeds] The "Save" button in Featured Article card should be filled in if it's already saved. - https://phabricator.wikimedia.org/T139377 [22:15:45] i can imagine it needing a revision or two, though [22:30:51] mdholloway: o/ sorry i missed your message. i'm using the freenode.net client today because riot is having problems [22:31:14] oh and the wikimedia logs aren't up to date right now either :] [22:33:12] niedzielski: i was saying that we'll likely have the go-ahead for pushing to production by mid-to-late next week, and do we want to a Beta now from master? [22:33:24] *to release a beta [22:36:12] dbrant: well, i'm 0/3 on repro'ing "the bounce" on a physical device [22:36:27] probably safe to disregard [22:37:50] (niedzielskii: i had noticed an issue on pre-marshmallow emulators with https://gerrit.wikimedia.org/r/339426 applied, where the feed would seem to "bounce" when refreshed on languages with no news card. but i can't reproduce it on a physical device.) [22:38:12] mdholloway: i think i have a fix, all the same [22:44:01] dbrant: sorry, was wrapping up this guy: https://gerrit.wikimedia.org/r/#/c/339574/ [22:44:23] mdholloway dbrant: sure, i can get started on michael's patch and see if i can repro the feed issue [22:44:37] niedzielski: no need on the feed issue; fixed. [22:44:46] * mdholloway is looking at dbrant's update [22:45:12] yep, fixed [22:45:48] it's an appcompat easter egg. [22:46:14] the chris bane of our existence :] [22:46:24] lol [22:54:15] * MaxSem makes siren sounds [22:54:24] https://en.m.wikipedia.org/wiki/Sgt._Pepper%27s_Lonely_Hearts_Club_Band#/languages [22:55:17] jdlrobson, joakino, phuedx|afk ^ [23:12:17] dbrant: niedzielskii: i've gotta get packed and stuff for flight tomorrow morning so i guess https://gerrit.wikimedia.org/r/#/c/338107 may have to wait if that's cool with you [23:12:37] 👍 [23:12:46] yep [23:13:13] (thanks for looking btw! i tried the invalidation approach early on and didn't have much success but it's probably worth another look) [23:13:24] cool, have a good one, see you monday! [23:14:08] i hope you have a really nice break! see you soon! [23:15:12] dbrant: so we'll skip that bookmark icon patch. were you aiming to get in these bottom tab bar patches? [23:15:31] niedzielskii: if possible [23:15:50] dbrant: ok cool, i'll check 'em out