[00:32:36] So, who's looking forward to our glorious gitty future the day after tomorrow? [00:35:47] * saper not. [17:15:00] Hi all! https://toolserver.org/~robla/crstats/?report=trunkphase3 the more people review and revert code today, the less we'll have headaches after the Git migration tomorrow. [17:16:16] Nikerabbit: I know today you have been focusing on working with Antoine re the localisation workflow, right? [17:21:45] Brion isn't here yet [17:21:59] and Trevor is probably ooo or otherwise out of commission ... [17:22:06] anyone in the SF office know where Trevor is? [17:22:10] rmoen: ping re 20% time today [17:22:22] MaxSem's day is over [17:22:41] hello sumanah [17:22:57] hi rmoen - what will you concentrate on today? [17:23:03] the more people review and revert code today, the less we'll have headaches after the Git migration tomorrow. [17:23:16] today it's ok to just plain revert stuff that we really don't need in the tree [17:23:20] in fact it's encouraged [17:23:37] sumanah, I'm fine with CR if that is where I am needed [17:23:49] specifically, if it hasn't been reviewed, it's been in there for a few days, and it's not an immediate bugfix needed for 1.19 deploy or release, revert it! [17:25:01] Ok to be sure, you would like me to svn revert it and recommit, thus creating another CR item ? [17:25:26] good question. hexmode what do you think? [17:26:57] sumanah: not at all [17:28:45] Nikerabbit: ? [17:29:21] Nikerabbit: hashar reached out to you to ask you to choose one of the https://www.mediawiki.org/wiki/Git/Conversion/translatewiki options, right? [17:29:21] sumanah: I've graduated today and my mother was visiting [17:29:28] Nikerabbit: congratulations!! [17:30:53] sumanah: sorry office connection went out, any updates from hexmode ? [17:31:11] hexmode: ping? [17:33:02] sumanah: I only see stuff about importing there [17:35:23] rmoen: so, here's my suggestion: do code review and keep pinging hexmode for an answer :-) [17:35:49] of those I'm leaning towards 2 [17:36:25] Nikerabbit: please comment on the bug? [17:36:32] Nikerabbit: https://bugzilla.wikimedia.org/show_bug.cgi?id=34137 [17:45:08] thanks Nikerabbit [17:45:11] sumanah: did I understand correctly that only core is being flipped tomorrow? [17:45:33] Nikerabbit: core, plus extensions that are deployed on WMF sites. ^demon|away is in charge of it and https://www.mediawiki.org/wiki/Git/Conversion#Affected_development_projects has more info [17:45:47] oh [17:45:54] nasty [17:46:14] yes, we will experience a productivity hit during readjustment and training in March and April [17:46:21] but the long-term benefits are worth it [17:47:09] not so sure about that [17:48:05] but I probably need whole rest of week just to adjust all the wikis I maintain to git [17:54:40] 6 dev wikis to re-checkout, plus 2 repo checkouts to do [17:54:45] Oh the fun [17:55:02] rmoen: from #mw: revert and commit to git, would be best I think [17:55:27] hexmode: hmmm ok, sounds fun [17:55:43] rmoen: well, you could just revert [17:56:51] hexmode: question rephrased: if I revert in svn, i have to recommit the changes for it it take effect, thus creating another CR. [17:57:15] hexmode just want to be clear on this [17:57:33] thanks all, have to go to a meeting now [17:57:43] you can mark the rev reverted, and then someone else can easily mark the revert ok [17:57:56] rmoen: what Reedy said :) [17:59:06] Reedy: so not really revert, just mark reverted [17:59:18] No, you do revert [17:59:28] You commit a revert, which is a revision that undoes an earlier revision [17:59:37] Then you go to the rev that is now undone, and you set its status to 'reverted' [17:59:58] Then you get someone else to trivially verify your revert is the exact mirror image of the other rev, and they'll mark it 'ok' [18:01:46] RoanKattouw: i understand the process of reverting. I was just stating that by reverting CR's, to cut down the queue, i'm essentially creating more CR's [18:02:06] Yes [18:02:09] But they're easy ones [18:02:12] RoanKattouw: But I understand that these new CRs can easily be ok'd [18:04:03] KO [18:05:21] btw, do we have a guide online yet for how to convert a live wiki from svn to git? (of course production wikis usually don't run off trunk, but even Wikimedia's install is based on SVN, be it not trunk) [18:05:57] it could just be: checkout from git, copy LocalSettings and images, and checkout extensions. But I'm wondering if there's more too it [18:06:07] and the swapping of directories [18:07:29] I don't think we have one [18:17:20] jorm: STOP [18:18:08] HAMMERTIME [18:20:01] does WMF git require https? [18:22:10] Nikerabbit: not sure.. I think it might, unless you use git:// , although that might use https under ground. no idea [18:22:44] Is there a meeting, like, now? Or was it cancelled, I read something about it but hadn't confirmed [18:23:08] ah, yep, it's gone [18:23:09] Krinkle: or let me put it this way: I have one production server with weird firewall, I can access port 80 and do svn up, but git clone git:// doesn't work. [18:23:36] Nikerabbit: That's because SVN is implemented in a crazy way that makes it work over port 80. Git uses a different port [18:23:40] Nikerabbit: thats because git works over SSH [18:23:56] so will there be way to do git clone over http from WMF? [18:24:03] Github has an anonymous HTTP mode [18:24:07] ^^ [18:24:09] oh, nope, that's https too [18:24:10] lol [18:24:13] https://github.com/jquery/testswarm [18:24:15] "HTTP" [18:24:15] Krinkle: it has http mode, and it works [18:24:20] Ryan_Lane: Does Gerrit/git support cloning over HTTP? [18:24:25] click it, it goes to https [18:24:36] I just did git clone http://github.com/Nikerabbit/TermBank.git [18:24:39] and it worked' [18:24:51] ok [18:25:06] it's going to be PITA if I can't clone MW directly there [18:25:25] https://github.com/blog/92-http-cloning [18:25:31] with the local patches it's not so fun to use tarballs [18:28:26] is the Great Git Migration happening tomorrow - Wednesday, March 21st? [18:28:34] I believe so ye [18:28:50] right on :) [18:32:22] RoanKattouw: kind of [18:32:34] RoanKattouw: but, we should just use https [18:32:37] Ryan_Lane: Well Nikerabbit wants to know [18:32:52] it isn't *that* much slower, and it's more secure [18:32:54] Apparently he's firewalled everything but port 80, and SVN Just Worked cause it uses HTTP in a crazy way [18:33:13] I *hate* that we allow http for svn [18:33:18] it's too late to change that, though [18:33:23] yeah [18:33:29] it's incredibly insecure [18:33:36] Well, HTTPS seems reasonable to me [18:33:45] dude. what place doesn't support https? [18:33:47] I mean if you're opening 80 you should probably also be opening 443 [18:34:02] there's no way in hell somewhere blocks outbound 443 [18:34:06] where is he? china? [18:34:24] wait no. even china doesn't block 443 ;) [18:34:28] On "a production server" somewhere [18:34:50] The default clone command didn't work, but that might be because it tried to ssh [18:35:09] it's possible the server blocks 443 too [18:35:19] he needs to get the admin to open a hole for gerrit on 443 [18:35:33] if they block 443 and not 80 they should be slapped directly in their faces [18:35:44] hard [18:35:46] multiple times [18:37:51] they can snoop on 80, maybe they want to see what he's doing [18:37:55] so they block 443 :-P [18:38:32] that's the dumbest thing I've ever heard :) [18:38:47] well, they are owning us over port 80, but at least we can see *how* they are owning us [18:39:19] And it's flawed [18:39:26] OR, in reality. "hey. they aren't blocking ports 10000-8932789237489723498, let's use one of those [18:39:32] " [18:39:33] If you control a server elsewhere, you just run sshd and run it on 80 [18:39:36] is this incomgin our outgoing? (I admit I did not do the backread) [18:39:40] out [18:39:41] Ryan_Lane: Finland, and I'm not the fw admin [18:39:44] RoanKattouw: nah. that doesn't work [18:39:54] Nikerabbit: ask the fw admin to stop being a moron :) [18:40:05] I only care about read access [18:40:15] there shouldn't be anything secret in the code [18:40:21] Well sure, but you'd like your code to be delivered securely I hope [18:40:29] I refuse to open 80 [18:40:30] sorry [18:40:31] HTTPS is not only for secrecy, it's also to prevent tampering [18:40:44] it's insecure, and I'm tired of us doing things incorrectly [18:40:52] Sending executable code over an insecure connection should scare you [18:40:59] exactly [18:40:59] will you open 443 then? [18:41:04] 443 *is* open [18:41:16] Question is if it's open in your firewall [18:41:24] nope [18:41:37] you can't ask the fw admin to open a rule for you? [18:41:42] Right, that's why we're calling your fw admin a moron, for opening 80 but not 443 [18:41:59] I've asked other person in my project to ask the fw admin to open it, but it's bureaucratic [18:42:28] so both the admin and you are causing lots of annoyance to me [18:42:48] Nikerabbit: Do you have SSH access into the machine? [18:42:49] well, we're trying to ensure your server doesn't get owned [18:43:08] RoanKattouw: I do [18:43:17] If so, you can clone the repo securely from another machine with a non-moronic firewall, then tar up the clone and scp it over [18:43:28] sending executable code across the internet to production services is asking for problems [18:43:41] That'll get the initial checkout done for you [18:43:47] RoanKattouw: like I said, that's just shit with local patches [18:44:04] True, you just made me realize this is not a sustainable proces [18:44:05] s [18:44:12] or, you can always use a proxy system, and pull, then push [18:44:16] I'm not blaming you, I'm just saying that you are making a switch that will annoy me [18:44:28] we're finally doing things in a secure way [18:44:41] and would prefer not to revert to doing things insecurely [18:45:58] RoanKattouw: ssh pipe sounds something that would work with least hassles [18:46:14] until I get the port opened [18:46:52] You would have to take another server that you have full control over, and run sshd on it and have it listen on a port that is open for outbound on your Finnish host (80 or some port >1024) [18:47:01] (maybe 8080 or 2222 or whatever) [18:47:19] Then you can set up port forwarding through that [18:47:36] there's another way too [18:47:37] or right [18:47:43] the idiots have also blocked 22 [18:47:47] git is distributed [18:47:57] so, run a constant pull on an intermediate [18:48:05] then either push or pull from that intermediate [18:48:46] Ryan_Lane: On what port? [18:48:47] some places have policies in place that forbid outgoing ports. I've worked in those environments before. [18:48:48] it sucks [18:48:51] wait [18:48:56] this is *all* systems? [18:49:00] there's no way [18:49:12] it's impossible to use the internet without 443 [18:49:53] One way or another, his machine is gonna have to communicate with some other machine, and you run into the draconian firewall rules. So unless your solution involves port 80 or another port that's magically open on that box (maybe 8080 is), it's dead in the water [18:50:11] that's why I'm saying to use a third box [18:50:14] like his desktop [18:50:21] pull onto the desktop [18:50:23] push to the server [18:50:37] Oh, right, he has SSH inbound [18:50:40] yup [18:50:51] Yeah that would work I guess [18:50:53] I can just initiate ssh proxy that way [18:51:12] I'm not too familiar with using arbitrary machines as repos to pull from and push to [18:51:12] yay for distributed source control :) [18:51:38] I mean I know it can be done, I just have no idea how it works [18:51:45] maybe this will work until the fw admin adds an exception? [18:52:41] what I'm more worried about is whether we do have anything that replaces the "svn up" I'm using, something that can handle mix of git/svn repos [18:52:57] For updating extensions you mean? [18:53:02] That's a ^demon|away question I think [18:53:16] yup [18:53:21] I already asked him [18:53:46] I'm just going to explode without them [19:17:22] Nikerabbit: for the extension distributor I've just made it iterate through directories and run git pull or svn update as necessary [19:17:29] basically looking at if there is a .svn/.git folder [19:19:13] aha [19:19:49] look at is cron.php [19:19:54] line 38 onwards [19:53:59] hexmode: Hey do you still have a script for counting the # of new revs in core + WMF extensions? [19:54:13] We know there's ~300 new in all of /trunk but we're wondering how many of those are relevant to the git switchover [19:54:16] (we = me and RobM) [20:47:52] It's nearer 200 new ;) [20:49:42] 180 [21:49:41] RoanKattouw: If I came up with such a script before (rev counter) I don't know where it is now [21:50:11] Oh, KO [21:50:18] Well I've already decided what to do anyways [21:50:20] wikitech-l post inbound [21:50:25] :) [21:54:24] will we have a new system for tracking re visions in SQL like we had with CR? [22:08:38] RoanKattouw: what is the "pipe-separator" message? [22:08:49] rsterbin: In English it's just '|' [22:09:01] In other languages, it might be other things [22:09:02] well, yes. [22:09:07] * RoanKattouw wonders how much it actually differs [22:09:14] can i just use mw.msg() [22:09:15] ? [22:09:17] Yeah [22:09:21] And add it to the list of used messages [22:09:25] ok [22:09:30] thanks [22:10:03] i'm not sure i understood what you meant about "make it a parsed message" with regard to line breaks [22:10:11] Sorry about that [22:10:27] I meant you'd need to invoke the message with wfMessage( 'foobar' )->params( whatever )->parse() [22:10:31] is there some sort of token i should be using to allow translators to mark a line break? [22:10:33] So ->parse() instead of ->text() [22:10:41] this is all in js [22:10:41] And then you can use wikitext in your message [22:10:46] Crap, really? [22:10:48] yep [22:10:50] hmm right [22:10:59] Yeah then the best you can do is use an HTML message [22:11:08] i.e. you put HTML in your message and you deliberately don't escape it [22:11:14] * RoanKattouw wishes we had a better msg library in JS [22:11:14] i thought we were trying to avoid that :-P [22:11:22] Yes, we avoid that unless it's necessary [22:11:43] i'm running it through my little link-building method anyway [22:12:17] i could potentially put some sort of token in there [22:13:20] like [[BR]] or something [22:13:30] meh, whatever [22:13:42] no need to start inventing new standards for this [22:16:00] but that's ALWAYS the best way [22:25:11] hi guys.... Is there a Mysqli database class already? [22:36:30] jeez louise im getting a lot of password prompts on sync-file as well as 'error reading response length from authentication socket' [22:36:34] awjrichards@fenari:/home/wikipedia/common$ sync-file php/extensions/MobileFrontend/api/ApiQueryExtracts.php r114319 [22:36:35] No syntax errors detected in /home/wikipedia/common/php/extensions/MobileFrontend/api/ApiQueryExtracts.php [22:36:35] copying to apaches [22:36:35] mw74: Error reading response length from authentication socket. [22:36:35] awjrichards@mw74's password: srv293: Error reading response length from authentication socket. [22:36:35] snapshot3: Error reading response length from authentication socket. [22:36:36] srv294: Error reading response length from authentication socket. [22:36:36] srv300: Error reading response length from authentication socket. [22:36:37] snapshot1004: Error reading response length from authentication socket. [22:36:37] searchidx1001: Error reading response length from authentication socket. [22:36:37] awjrichards@snapshot3's password: awjrichards@srv293's password: awjrichards@srv294's password: awjrichards@srv300's password: snapshot1003: Error reading response length from authentication socket. [22:36:38] as a snippet [22:46:18] awjr: fine for me.... [22:46:23] Are you forwarding your agent? [22:49:18] Reedy: yep [23:04:43] RoanKattouw: a froyo? :D [23:05:18] I'm acting like a dictator, there has to be *something* humorous in my post :) [23:51:11] hexmode, Reedy: Re API bug triage, I'll be on a plane at the time it's scheduled for. I have no burning desire to be there but it seems like either Reedy or I should be there. Reedy, are you available at that time (Friday 20:00 UTC) [23:51:13] ? [23:53:42] why does mediawiki never hold my session for any reasonable amount of time? [23:53:47] even on the production cluster? [23:53:59] am I the only one that notices this, ever? [23:54:20] My session gets held for at least a couple months... [23:54:36] I logged into mediawiki.org like an hour ago, and I go back to it and I'm not logged in anymore [23:54:42] this is constant [23:55:06] Its a server conspiracy [23:55:08] shouldn't my session last until my browser closes, if I don't select "remember me"? [23:55:41] One would think, I always select remember me though, so i haven't had much experiance with that