[08:01:20] join #wikimedia-toolserver [13:40:49] for those who plan on deploying wmf9 today. DONT [13:41:01] tablesorting is broken. [13:41:36] Reedy: is it you who does that normally ? [15:59:41] thedj[work]: it's still considered a blocker it seems https://bugzilla.wikimedia.org/show_bug.cgi?id=38803 [16:01:08] Reedy: includes/GlobalFunctions.php:line 528: $serverHasProto = $bits && is_set( $bits['scheme'] ); [16:03:03] isset [16:17:37] Nemo_bis: because I added it :D [16:20:10] thedj[work]: :D [16:23:20] gwicke: Morning! [16:23:44] marktraceur: good morning Mark! [16:24:25] gwicke: It's the weekly spontaneous work-from-home day for me :) [16:24:44] ah, ok [16:24:52] I just arrived in the office.. [16:25:25] This much I figured, yes. It's the time when all the magical WMF fairies flock to downtown and begin the day :) [16:25:48] at least in this part of the third floor [16:25:56] Right [16:26:08] Well, it's an hour-long process most days [16:26:28] did a bit of CPP research over the weekend- there sure are some nice libraries out there for our purposes [16:26:58] folly::dynamic for example [16:27:16] Cool! Has work begun on implementation, or are you fishing for volunteers? [16:27:43] right now we need to figure out the memory management strategy and basic data structures / interfaces [16:28:10] plus see if we want to actually use plain make as our build system [16:28:29] Make is pretty awesome, I must say [16:28:30] our dependencies use automake, I personally kind of like cmake [16:28:55] but right now we don't need anything too fancy, so it is not too pressing [16:28:57] cmake doesn't seem overly problematic, but I haven't written cmakefiles in the past [16:28:57] simple, wide implementation == ++ [16:29:24] Amgine: It's getting pretty trivial to install cmake, actually [16:29:49] Any addition is a complication. If it's not necessary, why complicate? [16:29:50] I just apt-get installed it the last time I used it [16:30:16] anyway, we should get away with plain make for a while [16:30:29] Amgine: So why use make at all? :) [16:30:51] on that I happen to agree, but clearly you do not. [16:31:18] Amgine: At some point, there's a nice balance between ease of development and user frustration [16:31:48] As long as users can install the dependencies, I'd say that running the same number of commands with some perks is an acceptable thing for them [16:31:50] non-dev users should just apt-get install libmediawiki or somesuch [16:32:23] It depends on your goals. If your goal is ubiquity, no there is no balance: zero user frustration is the only possible measure. [16:33:19] Amgine: somebody who has a full CPP build system plus boost libraries installed *might* have make installed as well [16:33:56] Yes, gwicke, just maybe. [16:34:44] * marktraceur doesn't always use Boost to write shorter C++, but when I do, I use the time I saved to manually run g++ on each file. [16:35:03] (-: [16:56:10] marktraceur: http://www.rasterbar.com/products/luabind/docs.html looks quite elegant for Lua bindings [18:07:28] https://twitter.com/shit_hn_says/status/234856345579446272 [19:53:20] hmm, someone pushed wmf9 to en.wp ? [19:53:41] we broke all sortable tables in certain browsers. [19:55:10] https://bugzilla.wikimedia.org/show_bug.cgi?id=39296 [19:55:52] tried to warn about this this morning, but we have no real workflow for reporting issue in the deploy cycle. [19:57:06] thedj: Start highlighting random people who might have done it? :P [19:57:10] robla: Reedy: AaronSchulz: ---^^ [19:57:11] No, don't, hold on [19:58:04] thedj: Reviewing https://gerrit.wikimedia.org/r/#/c/19216/ now [19:59:25] Looks good to me, deploying shortly [19:59:53] I already tried highlighting folks a couple of hours ago, but I guess SF wasn't awake enough :D [20:00:25] and i marked the bug as blocker on the deploy. [20:06:58] marktraceur: You work a lot on the uploadwizard, right? [20:08:26] multichill: "worked" might be more accurate, but I know it pretty well, yes [20:09:30] After Wiki Loves Monuments in september 2011 a lot of bugs seem to have been introduced related to the licensing and i18n [20:10:13] multichill: Yeah, there are a lot of bugs still open, a lot of patches still awaiting review [20:11:20] multichill: https://gerrit.wikimedia.org/r/#/q/project:mediawiki/extensions/UploadWizard+is:open,n,z might include some fixes for the open bugs [20:11:37] You can help me pester people into reviewing them, maybe [20:12:09] Though there are plans for a review sprint in later August, so maybe just wait for it? [20:12:14] Ouch [20:12:34] I know the goal is definitely "deploy it for WLM" [20:12:56] But sadly that's overlapping with several goals like "finish project X by end of August", at least for me [20:16:09] yeah..... [20:19:45] multichill: Oh, you're Maarten? Hi! :) [20:20:36] Yup, that's me :-) [20:20:56] thanks RoanKattouw and thedj [20:21:17] multichill: It's sometimes frustrating to not know which usernames match up to others, but getting better all the time [20:22:08] Yes, that can be hard [20:33:57] OK actually cherry-picking that fix now [20:34:09] I got distracted by the lunch conversation around here [20:39:40] RoanKattouw: thx [20:40:05] Deployed now [20:40:11] thedj: Is it working? [20:43:22] for me it is [20:43:57] Cool [20:46:32] hmm, but multiple sortkeys is broken [20:47:03] 'undefined' is not an object (evaluating 'parsers[j].format') [21:20:30] WTF [21:20:35] marktraceur: Were you talking about this error? [21:20:47] /var/lib/parsoid/Parsoid/js/node_modules/jquery/lib/node-jquery.js:10 window.XMLHttpRequest.prototype.withCredentials = false; ^TypeError: Cannot read property 'prototype' of undefined [21:24:48] RoanKattouw: Yeah, you're getting it too now? [21:24:58] So maybe it's not my failure to install node? [21:25:55] Yeah I'm getting this on the new box [21:26:02] I'm gonna try to downgrade to jquery 1.7.2 [21:26:41] Oh right, new jQuery came out [21:27:15] Yeah it was running 1.7.3 [21:27:29] It's now warning me that jQuery 1.7.2 wants node 0.6 and I'm running node 0.8.something [21:27:30] Translate has something broken [21:27:32] So we'll see how well that goes [21:28:55] RoanKattouw: From my testing, not well [21:29:26] Oh FFS [21:29:31] Now it's giving me the navigator crap [21:29:40] I think I may have to downgrade node too [21:30:07] :( sadness [21:42:29] marktraceur: AHA [21:42:35] I found the problem [21:42:37] I didn't have g++ installed [21:42:45] RoanKattouw: Really? That's it? [21:42:55] Wait, I don't have g++ installed? [21:43:01] Well, just saying [21:43:05] No, I have g++ [21:43:12] When I downgraded node to 0.6 it didn't want to build [21:43:19] Oh, that problem [21:43:20] K [21:43:29] contextify was failing ot build [21:43:39] Somewhere in the maze of confusing error messages was something about a C++ compiler [21:43:44] So I installed g++ on the machine and now it's working [21:43:49] Running jq 1.7.3, but still with node 0.6 [21:43:54] Gonna try with node 0.8 now [21:45:50] OK now it's running in 0.8 [21:45:55] Let's see if it'll build from scratch in 0.8 [21:51:08] marktraceur: Hah, and installing make helps too :) [21:52:57] WHOO [21:53:05] I got stuff running with the new node and the new jquery [21:57:26] RoanKattouw: What were the problems? Just g++? [21:59:41] marktraceur: g++ and make [21:59:50] Hm. I have both of those things. [22:00:01] In node 0.6.12 I got nice flashy explodey build failures when they were missing [22:00:17] In node 0.8 I think the build just sort of silently half-succeeded and thnigs exploded when trying to run jQuery [22:00:26] Are you sure you had those installed when you ran npm install? [22:00:40] Pretty sure [22:00:46] Hmm [22:01:01] And I'm definitely using node 0.8 [22:01:13] 0.8.6 [22:01:18] For me (on wtp1.pmtpa.wmnet), installing g++ and make (or just build-essential, really), then destroying the node_modules directory and running npm install to rebuild it fixed it [22:01:30] I'm on node 0.8.2 [22:03:07] Aha, that did it, thanks [22:04:00] You know what, I'm glad you guys are porting this to C++. Node is a PITA to deploy [22:04:11] And how. [22:04:26] Well, apparently ruby is worse [22:04:41] Haha, I'm sure that conversation could be a long one [22:04:53] Faidon pinged me last week saying "I just installed gitlabs. Well, sort of. I installed 87 gems and it's still not working" [22:04:58] Trashing languages A-Y until finally Z is the only one left [22:05:57] And Z turns out to be something ridiculous like LOLCODE [22:06:10] Say what you will about the language, it's damn fun to use [22:08:33] * marktraceur can finally actually work on this bug as opposed to trying to get node installed [22:12:22] ....why MW API no return status codes? [22:12:57] There's a bug about that [22:13:24] But yeah unless something is terribly wrong, the API will return a 200 and any error will be in the body [22:14:34] Euuugghhhh. [22:14:47] I think setting status codes could be nice, but one of the main rules of the MW API is that it should be usable without any "special" HTTP semantics [22:15:04] By which I mean, it should be usable if all you can do is GET and POST and you can't access response codes or headers [22:15:15] Hm. [22:15:31] A) You can return the same response body with a different response code [22:15:40] B) Why would you not be able to access response codes? [22:15:56] *shrug* [22:16:35] The response code thing is less defensible than the GET/POST thing, because browsers typically can't perform exotic HTTP methods like PUT and DELETE [22:17:03] That much I believe [22:17:13] But 1) a lot of HTTP libraries suck and 2) all information is conveyed in one format (JSON), which makes stuff easier to deal with IMO [22:17:15] (well, without some javascript console work anyway) [22:17:37] I'm open to implementing response codes as an addition to the error mechanism but not as a replacement [22:17:51] *nod* I'll maybe look into it later this week