[01:28:24] looking through the TimedMediaHandler extension, it seems that videos are transcoded the first time they're requested. is that right? if so, is there an advantage of transcoding then as opposed to transcoding upon completing the file's upload? [01:35:45] Why transcode something if there's no need? [01:36:06] I suspect it's the same as everything [01:36:16] There's cost/benefits of doing things lazy or greedy [01:39:57] it would add a lot of latency to do it on demand, wouldn't it? [01:40:37] only the first time ;) [01:40:43] that's why i was curious [01:41:52] I know the BBC transcode to many different quality/sizes etc [01:42:09] If you were streaming it, on demand and writing to disk at the same time would probably work [01:43:05] oh, so transcoding while streaming? that's neat, and would make a good amount of sense of the decision to defer transcoding until request then [01:43:36] I'm not sure if TMH supports that though [01:45:49] for an extension i'm working on i'll need to convert a plain-text file to a ray-traced static image. this takes about 30 seconds to complete on a medium-grade laptop. would it be best to put this task into a job queue upon completing upload of the plain-text file? [01:47:41] That's the sort of thing the jobqueue is designed for [01:48:11] there's no mechanism for polling for job completion or anything like that [01:48:25] although I suppose that could be added [01:48:44] is the job parallelisable? [01:51:08] i don't think it's feasibly parallelizable -- the expensive operation is ray-traced. TMH seems to have some polling/observer-pattern like mechanism built into its jobs, so i could perhaps use that [01:52:10] ray tracing is parallelizable in principle [01:52:52] what ray tracer is it? [01:58:26] i'm not sure. the code i've got ray-traces by interacting with the pymol API (http://en.wikipedia.org/wiki/PyMOL) [02:00:02] if it were feasibly parallelizable, would that be ok to do outside of a job task? [02:01:41] interesting, the manual is behind a paywall [02:02:45] http://www.ccp4.ac.uk/newsletters/newsletter40/11_pymol.html notes that pymol uses a built-in ray tracer, but also supports PovRay (http://en.wikipedia.org/wiki/POV-Ray) [02:04:43] yeah, I was reading the same document as a PDF [02:04:53] it probably doesn't support parallel processing [02:05:26] "One example use of this feature would be to farm out rendering of a molecular animation to a cluster of Linux workstations for parallel processing." [02:05:49] implies that each job would be a separate frame [02:05:58] no parallel processing for an individual frame [02:06:04] do you want single frames or animations? [02:08:02] just single frames. i'm generating the ray-traced static image as a graceful degradation for clients that don't support webgl -- the main purpose of the extension is to enable interactive 3D models of proteins and dna [02:09:08] so you will just have to wait [02:10:00] ok, i don't imagine that would be a major issue [02:10:10] doing it with the job queue is possible, but you would need to have a dedicated job runner for that job type so that the jobs don't have to wait behind jobs of other types [02:11:32] you should try povray, it might be faster [02:16:29] will do. any impression on how fast the model image generation would need to be on my medium-grade laptop to avoid needing to add the task to a job queue? [02:17:07] Suppose it depends on your browser timeout [02:23:00] i imagine the browser's network request timeout would be a hard upper limit on that. are there any rough values/guidelines for how long a request can hang for user-experience? that constraint seems like it could likely be stricter. [02:25:04] actually, would the browser/ux time constraints really determine whether the task had to be added to a job queue? isn't the job queue much more for managing system resource allocation on the server? [02:26:21] the point of using the job queue is to allow the HTTP request to be completed [02:26:38] then the rendering job can continue in the background and you can poll for progress using JS [02:26:44] showing a progress bar or whatever [02:30:09] Whoa [02:30:11] May 8 02:13:17 10.0.8.30 apache2[29174]: PHP Fatal error: Call to undefined method languages::getCode() in /usr/local/apache/common-local/php-1.20wmf2/includes/cache/MessageCache.php on line 616 [02:30:22] Preceded by [02:30:24] May 8 02:13:17 10.0.8.30 apache2[29174]: PHP Warning: call_user_func_array() expects parameter 1 to be a valid callback, class 'languages' does not have a method 'getCode' in /usr/local/apache/common-local/php-1.20wmf2/includes/StubObject.php on line 59 [02:30:52] APC corruption? May 8 00:55:53 10.0.11.47 apache2[5048]: PHP Fatal error: Invalid host name (wikipedia), can't determine language.#012 in /usr/local/apache/common-local/multiversion/MWMultiVersion.php on line 342 [02:31:07] Heh and with that I'm going home [02:32:54] ah, ok. in that case i imagine it'd be best to use a job queue if the task takes ~ 30 seconds [04:46:36] is MessagesQqq plain text or wikitext? [04:52:31] oh hi TimStarling :) [04:57:42] hi ToAruShiroiNeko [04:58:07] I guess by your comments on bug 32077 that you've been renaming yourself again [10:44:59] anyone able to deploy a quick fix? [10:45:04] the fix being https://gerrit.wikimedia.org/r/6915 [10:45:09] basically no tables are visible on the mobile site [10:45:28] :o [10:46:20] yikes [10:46:49] * ToAruShiroiNeko cries for being left out [10:49:21] Reedy, Nikerabbit, any other shell user around? [10:50:19] :( [10:50:34] apergos? [10:51:50] I don't know that I'm a deployer [10:52:11] if it was on the ops side then for sure [10:52:18] You did develop my personality :p [10:52:34] :-/ [10:52:52] "developer" doesn't specify what is being developed. [10:52:53] and I can't +2 it [10:52:59] a baker is a cake developer. [10:53:14] apergos: who here might be deployer? [10:53:25] * apergos looks at the clock [10:53:37] who can +2? [10:53:39] oh people were tlaking about deployers, I feel dumb :( [10:53:46] most folks that come to mind are in the wrong time zone [10:54:03] reedy? what tz are you in? :-P [10:54:29] reedy is in SF? [10:54:33] I dunno [10:54:36] maybe [10:54:39] i think so [10:54:49] could be wrong [10:59:14] loking at the git lgos gives me no inspiration either [11:05:39] oh! [11:05:44] there is a ^demon in another channel [11:05:46] try him [11:06:00] (I just saw a review of his go by) [11:06:05] ^demon ^demon ^demon [11:06:12] othe rchannel, not here [11:06:16] :) [11:06:34] * Katie_WMDE doesn't see him in any channel yet but saw him reviewing [11:06:40] rats [11:06:48] maybe he didn't come on irc yet [11:06:57] yeah [12:10:11] how do I make parser functions check if an input is a number or not? I want an ifexpr where it first checks if {{{1}}} is a number or not, if not it goes to else condition [16:30:18] MaxSem: did you get it sorted? [16:30:22] And yes, I'm in SF [16:30:32] (I saw you in your meeting yesterday morning ;)) [16:30:57] Katie_WMDE: ? [16:31:10] Reedy, Chad tried hard, but Gerrit repelled all HIS attacks! [16:31:23] Reedy: problem with gerrit [16:31:30] not deployed yet :( [16:31:31] Lol [16:32:44] hi Reedy [16:32:53] Ohai [16:33:29] Reedy: has there been any scap in the past 7 days? [16:33:44] I think so [16:33:55] Roan/raindrift deployed a new extension yesterday [16:34:15] Reedy: okay, so if I now enable TranslationNotifications, the i18n files should be already synced [16:34:15] Hmmm. SAL doesn't suggest so [16:34:35] You're not going to break anything by running scap [16:34:42] A few issues have been fixed up with it now [16:35:26] want me to run it? :p [16:36:03] Reedy: if you want to [16:36:24] let me update the extension code first [16:36:28] ok [16:39:27] Reedy: https://gerrit.wikimedia.org/r/6935 [16:39:36] <^demon> If that merges, I'm gonna be pissed. [16:39:58] ^demon: pardon? [16:40:01] <^demon> Because I did the exact same stuff for MobileFrontend, but gerrit won't merge it :( [16:40:28] well, try it? :o [16:41:10] I see two ticks and +1 and +2?? [16:41:12] <^demon> Yours won't merge either. [16:41:13] <^demon> Hrm. [16:41:42] c00l [16:41:44] Gerrit is b0rked? [16:41:57] new shiny 2.3! [16:42:07] <^demon> It's not 2.3...we've been doing merges all week [16:42:50] Nothing useful in the error log? [16:43:10] Gerrit has an error log? [16:43:29] I know it's got something [16:43:34] as it was spammed with crap before [16:43:35] <^demon> Got it [16:43:46] <^demon> http://p.defau.lt/?a58233Vhyf5XY941dpOgxw [16:44:01] I see [16:45:06] ^demon: I can't merge in Gerrit what have you broke ? :D [16:45:32] <^demon> Is this breaking for everything, not just wmf branches? [16:45:44] Let's try [16:46:08] Just wmf [16:46:14] https://gerrit.wikimedia.org/r/#/c/6922/ - I merged that a little while ago [16:46:20] that looks like Java [16:46:27] <^demon> It is java. [16:46:32] RUN. [16:46:46] Missing blob [16:46:59] Missing green blob [16:47:29] only ~100 lines in stacktrace? they're not using Spring! [16:53:15] ^demon: let me know if you get it fixed before my deployment window ends (in 40 minutes) [17:12:43] prepping for 20% checkin. [17:13:12] I know Trevor is doing OnlineStatusBar review. [17:15:33] Morning sumanah [17:15:39] Hi there RoanKattouw [17:15:49] RoanKattouw: how is your day going so far? [17:16:19] I'm doing some Wikimania video legwork now, then CR probably [17:16:33] RoanKattouw: Thank you. [17:17:11] rmoen: How is your day going? [17:17:18] I believe brion is probably still ill? [17:17:51] He's in the office today [17:17:55] * RoanKattouw spots a brion in the office [17:18:06] sumanah: going well :) You ? [17:18:48] rmoen: Not bad, considering :) And do you have 20%-related plans today? I have a request, in case you have an interest [17:19:35] sumanah: You're in luck :) I'm taking requests [17:19:37] https://www.mediawiki.org/wiki/Wikimedia_Foundation_engineering_project_documentation/Status_helper [17:20:06] rmoen: brion - would either of you be interested in implementing this today? or in working on the deployment communications/coordination calendar stuff we've been talking about on the engineering list? [17:20:18] sumanah, i'm back [17:20:21] what's all this? [17:20:27] although - robla correct me if I am wrong - the deployment calendar stuff might need more speccing out before someone ought to actually make tools [17:20:35] Hi brion! glad you are feeling better [17:20:51] \o/ [17:20:52] brion: I was wondering whether, as part of your & rmoen's 20% time today, you'd be able to help with some toolmaking that would benefit teammates [17:21:22] rmoen: the "status helper" would help tfinc & others who want to update onwiki status msgs more often [17:21:57] i was planning to mostly just catch up on review [17:21:59] because right now, the adding-a-status step as described in https://www.mediawiki.org/wiki/Wikimedia_Engineering/Project_documentation_howto#How_to_add_a_new_status_update_to_a_project is a little annoying & cumbersome [17:22:06] hmm [17:22:15] brion: seems reasonable to me, review in good health :D [17:22:21] :) [17:22:25] sounds handy though [17:23:17] rmoen: guillom and tfinc will both be immediately grateful if some engineer can make the statushelper (a gadget I presume?), but if you think it's not a reasonable thing for you to take on for May, just say so [17:23:27] sumanah: is this because the engineering report gets generated by the date manually entered ? [17:23:28] A Gadget yeah [17:23:33] Talk to robla for more info [17:23:59] rmoen: It now gets generated by the status that's dated "year-month-monthly" like "2012-04-monthly" [17:24:04] rmoen: but basically, yes [17:24:20] sumanah: gotcha [17:24:53] rmoen: also check out https://www.mediawiki.org/wiki/Wikimedia_Platform_Engineering and https://www.mediawiki.org/wiki/Wikimedia_Features_engineering and https://www.mediawiki.org/wiki/Mobile -- those are hubs that we aim to have people use to see, at a glance, what's up in each subdept [17:25:00] more frequently updated than the monthly report [17:25:17] so the sorta-tweet-sized Status updates are propagated in a lot of places. [17:26:13] hi drecodeam! [17:26:25] hi sumanah [17:26:46] sumanah: ok. this sounds fun. I'll have a go at it today. [17:27:01] brb [17:28:13] * AaronSchulz is glad brion is back [17:28:40] Got it! ok, thanks rmoen! I shan't bother you about it -- I figure if you don't finish today, you'll start up again on it next week [17:28:40] Thanks all [17:29:19] sumanah: sorry i am a little busy right now, preparing my visa documents [17:29:23] have my interview tomorrow [17:31:25] sumanah: sounds good :) [17:37:17] good luck, drecodeam [18:07:09] if folks are looking for the git/gerrit training - PM saper [18:29:02] or go to #git-gerrit :-) [18:50:25] Reedy, Daniel_WMDE_: why wikidata.org? [18:50:30] rather than data.wikimedia.rog? [18:50:33] *org [18:50:50] I've no idea [18:50:51] it's a single multilingual wiki, right? [18:51:20] Lydia asked me for it [18:51:55] can we have a conversation about it before we decide on the domain name? [18:52:00] lemme start an email thread [19:22:33] RoanKattouw: did you have a chance to look at that refactoring change on gerrit? https://gerrit.wikimedia.org/r/#/c/6319/ [19:23:22] if we can get the queue reviewed and merged, i can submit the central feedback log [19:23:43] OK, I'll look at that nowish [19:28:23] thanks [19:28:38] Approved [19:29:23] 6422 and 6532 need rebasing onto the version of 6319 that actually got merged, doing that now [19:35:27] thanks [19:36:21] 6422 approved [19:36:24] thanks! [20:01:36] hi chrismcmahon & hexmode [20:01:42] should I be dialling into a particular number or URI? [20:02:03] sumanah: g+ invite sent [20:02:07] Thanks! [20:02:15] do you have it? [20:03:10] hm, re-checking [20:03:36] aha! [20:24:56] rsterbin: https://gerrit.wikimedia.org/r/#/c/6532/ has problems , see review [20:25:36] ok, will fix -- one sec [20:32:07] RoanKattouw_away -- fixed [20:45:31] rsterbin: merged [20:45:37] thank you [22:05:02] Hey AaronSchulz, I've got a question about files when you have a min.. [22:47:04] RoanKattouw: set-group-write just does /h/w/common not the conf dir :-] [22:48:00] hashar: Yeah, fixed the perms just now (see -tech) [22:49:24] danke [22:51:07] RoanKattouw: apparently did not catch some dir. Could you please chmod -R g+w /home/wikipedia/conf/httpd/broken/.svn/tmp [22:51:23] taking a short break. will be back soon [22:53:29] hashar: Doen [22:55:27] RoanKattouw: great WFM! [23:01:01] Hah, lucky that include() doesn't cause a fatal for a missing file [23:01:10] A maintenance script I was running just hit this: [23:01:12] Warning: include(/home/wikipedia/common/php-1.20wmf2/../wmf-config/secret-projects.php): failed to open stream: No such file or directory in /usr/local/apache/common-local/wmf-config/CommonSettings.php on line 1854 [23:02:05] Oh and now I get this, this is actually broken: [23:02:06] catrope@fenari:/home/wikipedia/conf/httpd$ mwscript eval.php testwiki [23:02:08] PHP Warning: include(IP/extensions/OpenSearchXml/OpenSearchXml.php): failed to open stream: No such file or directory in /home/wikipedia/common/wmf-config/CommonSettings.php on line 1853 [23:02:09] Reedy: ---^^ [23:16:13] Yeah [23:16:14] Fixed [23:16:18] Mising $ [23:16:19] Thanks [23:16:33] Ah yes the next run didn't get any errors [23:16:38] For some reason "secret-projects.php" only added in OpenSearchXML [23:16:44] It's not exactly a secret now... [23:16:58] I'm running importImages.php for single files (WM11 videos) at 15-min intervals [23:19:35] I need to go go round and merge the git repo into common and wmf-config [23:19:41] make a bit of a mess [23:20:50] You mean actually switch fenari to use the git repo? [23:20:53] yup [23:20:53] Yeah that'll be fun [23:21:02] We need git review on fenari ideally too.. [23:21:18] Hmm yeh [23:21:19] logged an RT ticket for someone to do that [23:21:31] They'll want to have git-review packaged probably, not sure if that's been done yet [23:21:38] indeed [23:22:14] There's a ppa by the look of it [23:22:16] I guess we'd allow direct push for the wmf-deployment group in the meantime? [23:22:25] Hmm, how recent is the version in the PPA? [23:22:39] https://launchpad.net/~managedit/+archive/git-review [23:22:46] 1.14-0mit1~65.gbp1ed350 [23:22:51] Hmm [23:23:05] That's three months old [23:23:07] I'm running 1.16 [23:23:28] I suspect creating a deb for it isn't going too difficult [23:23:33] Probably not [23:23:43] but yeah, need to enable direct push for the wmf group etc [23:23:44] Someone's already done it, could look at that as an exampel [23:23:48] yeah [23:23:52] But even then it's just a script and a man page [23:24:30] I was thinking we should just checkout the git repo at /common [23:24:39] add in some .gitignores to quieten it down [23:24:59] then we get wmf-config in the same place, with .gitignore to keep the private files out [23:25:04] Hmm ... [23:25:21] Yeah that could work [23:25:22] We could symlink the world, but it's just making a mess for no real reason [23:26:17] I should probably email people to tell them this is happening ;) [23:26:30] aye [23:26:32] Are you on ops-l? [23:26:36] yeah [23:26:53] Oh [23:26:54] https://launchpad.net/ubuntu/+source/git-review [23:26:57] 1.16 [23:27:24] 1.16 in QQ... [23:29:26] Sweet [23:29:34] No idea why their precise package is 1.12 [23:29:55] We'll need to copy the quetzal package to our repo and rebuild it for precise and lucid I think [23:30:03] *quantal [23:32:04] http://packages.ubuntu.com/quantal/all/git-review/filelist [23:32:27] It's not going to be too hard! ;) [23:32:57] /usr/share/pyshared/git_review-1.12.egg-info/PKG-INFO [23:33:01] wtf, I thought it was 1.16? [23:39:34] lol