[22:42:01] bd808: anomie: what do you think about delaying the sessionmanager patch a week? hashar is worried about too many big changes being deployed at the same time due to the code freeze [22:43:00] we wouldn't be deploying the authmanager patch next week anyway, so it should not upset anything [22:43:31] tgr: it pushes back the date that anomie can announce bot passwords [22:43:39] and ton of code is going to land because of the freeze [22:43:52] so that might make the debugging harder for you guys when the train happen [22:44:04] does authmanager generally feel risky to you tgr? [22:44:16] bd808: you mean the time between the bot password announcement and the login api changing becoming too short? [22:44:18] I have no clue of the impact anyway, and I guess if you have time to baby sit once it lands in production it is fine to land [22:44:20] or are you just responding to hashar's general concern? [22:44:29] just wanted to warn about the huge amount of code which is going to land on prod [22:44:32] authmanager or sessionmanager? [22:44:40] sessionmanager, sorry [22:44:53] I wouldn't expect any major problem [22:44:57] edge cases, probably [22:45:13] beside that, I have no concern of it landing :-} [22:45:13] hashar: *nod* its worth bringing up but the damage is mostly already done [22:45:15] Shouldn't we let the sessionmanager code sit in beta for a while? [22:45:45] and yeah beta might help you give some low hanging fruit. Saves you time for the actual prod deploy [22:46:12] but as I said in #wikimedia-releng Don't take all the above as a veto , more like I am being very careful for potentially no reason [22:46:16] it probably would be safest to deploy it just after the branch cut for a long beta soak if we (sarcastic we) are worried about it causing problems [22:46:19] so it is your call really :) [22:46:40] if you don't anticipate any troubles, just deploy [22:46:47] there is no point in delaying just for the sake of it [22:47:21] I'm not too worried about it causing big problems, it got a fair amount of review and testing [22:48:04] OTOH if it does cause small problems, wich is almost certain to happen, debugging is less painful if only one week's worth of changes coincide with it [22:48:34] tgr: anomie probably won't see pings till later (maybe not even until morning), so it's a bit up to your judgement [22:48:44] I am be fine with merging today, if we have a reason to do [22:49:14] so why does it matter if the bot password change gets out a week later? [22:50:03] if we want to make sure the bot community has time to adopt, that makes sense [22:50:14] but they have a month or so anyway, right: [22:50:16] ? [22:50:17] it wouldn't be the end of the world. It might make us hold a bit longer before dropping authmanager. I don't remember how long anomie wanted that to be available [22:50:59] but I agree that we will quite likely still have another month at best [22:52:16] I'll delay it then, unless anomie comes back and says its a bad idea [22:52:21] tgr: gotta sleep, again no worries from me if you C+2 [22:52:53] the other benefit is that the "new extensions + old core" configuration gets some testing in production this way [22:53:10] we don't need that but third parties might [22:53:25] thanks for fixing the CI pipeline hashar! [22:53:56] tgr: I got some phab tasks to make those maintenances tasks automatic :) [22:54:02] hopefully in 2016 [22:54:49] have a good afternoon [23:47:43] bd808: What do we do about symfony-process then? [23:48:13] Shipping a version in the vendor repo that MW core can't technically support... Feels icky :) [23:49:36] Reedy: hmmm... well... [23:49:48] Yes, technically it's fine for silver/hhvm etc [23:49:55] it's not that core can't support having 5.3+ code coexist with it [23:50:01] yeah [23:50:29] it's more a problem of anyone not the WMF trying to use mediawiki/vendor.git I suppose [23:50:56] The fact 2.6 supports 5.3.3, and 2.7 is 5.3.9... For an arbitary subversion jump, suggests a reason for doing so [23:51:16] I don't remember what thing we have that needs symfony-process. Was it for geshi? [23:51:40] yup [23:51:43] which is targetting "symfony/process": "~2.5" [23:52:09] ah and we don't pin to lowest so that's <3.0 [23:52:36] it should be ok to downgrade then [23:53:21] version 3 bumps support to 5.5.9, which is a blocker till we get core/tin done [23:53:54] 2.8.1, 2.7.8 or 2.6.11 [23:54:10] Whichever way, the commit summary IS wrong [23:54:33] oh look, 2.6.12 is out :P [23:54:45] Reedy: yeah was just writing that you should fix the summary for https://gerrit.wikimedia.org/r/#/c/260157/ and I can merge it for you [23:55:28] I changed the version in the extension to ~2.6 in https://gerrit.wikimedia.org/r/#/c/260161/ too [23:55:30] I guess as long as your being OCD about new hotness you can pick up the latest 2.6.x if that's their 5.3 compat branch [23:55:40] Yeah, might aswell [23:55:53] needs a rebase too