[00:57:06] https://www.mediawiki.org/wiki/Special:Code/MediaWiki/97029 [00:57:14] for sumana and aaron [01:00:52] aww, it's like a greeting card! [01:02:29] Did something just change with logins? [01:02:47] There used to be a log me in to other projects button [01:03:28] (radio/checkbox I guess, I forget the input type) [01:03:49] PiRSquared17: hmm, check the deployments calendar? [01:04:08] I couldn't log in for a few attempts [01:04:16] PiRSquared17: https://wikitech.wikimedia.org/history/Software_deployments [01:04:40] You don't need the checkbox [01:05:05] It was removed as it shouldn't have been added in the first place [01:05:12] Is there currently a phishing WMF site? [01:05:29] (using https/http) [01:05:50] phishing wmf site? [01:06:13] Yes [01:06:21] "Phishing" WMF-look-alike [01:06:52] Anything running MediaWiki and vector looks like most wmf sites [01:07:36] Yes, but it was "https://simple.wikipedia.org/w/index.php" [01:08:04] (or something *VERY* close) [01:08:15] And? [01:09:06] I couldn't log in [01:09:47] How's this got anything to do with the checkbox? [01:09:58] (well, it took me to the "logged in" page, but after that logged me out) [01:10:02] IDK [01:10:13] It doesn't protect you from phishing sites [01:10:22] I thought some one would knoe [01:10:26] use a browser with phishing site blocking, and hope they've been reported [01:10:36] *someone, w [01:11:02] sorry [01:16:16] TimStarling, AaronSchulz, ^d, are we ok with commit usernames that have hyphens in them? [01:16:46] I see happy-melon has one so I think we are, just wanted to check with you first [01:17:03] I guess so [01:17:14] sure [01:17:18] UNIX usernames are really very broad, the form that they usually take is by convention [01:17:53] ok [01:18:07] of course if you use a space in a username then every shell script in existence will start trying to delete your root partition [01:18:29] but a lot of other characters could theoretically be used [08:16:11] TimStarling: good evening Tim! [15:06:22] hi apergos [15:07:28] is robots.txt in svn? if not I am wondering why... in that case more people could update it based on requests on wikipedia [15:07:41] hello [15:07:45] hey :) [15:08:15] re https://en.wikipedia.org/wiki/Wikipedia:Village_pump_(technical)#Why_is_this_talk_page_showing_up_in_a_Google_search.3F [15:09:28] I think some config files could be in svn or gerrit so that other devs without shell acess have a chance to update them, if the change is wrong, it wouldn't be deployed either [15:09:48] petan|wk: Individual wikis can use MediaWiki:Robots.txt [15:09:58] And yes, the config files should be in git in the future [15:10:02] They are now in a private SVN repo [15:10:03] ah, that's something I didn't know ^^ [15:11:34] btw I read some irc logs from demon's git talk and I wasn't really sure about difference between extensions-wmf and extensions-all? [15:12:04] but I didn't read everything so far maybe it was explained later :P [15:13:35] also are you going to restrict core access in same way as it is now, or it's going to be more open? [15:14:15] Well there [15:14:20] There's access and there's access [15:15:07] huh [15:15:26] I think we're gonna restrict who can review and accept changes [15:15:38] We're probably gonna allow "everyone" to submit changes for review, and to create branches [15:15:56] ok so it would be possible for everyone to commit / push changes to core or not [15:16:06] ah, ok [15:16:23] cool [15:21:56] hi roan, do you know where i can find the code that generates the dumps.wikimedia.org/other/pageviews-raw/ files? i am trying to locate apergos but i am not successful [15:22:07] you're trying to locate me? [15:22:10] yes! [15:22:11] I said hi and everything [15:22:21] sorry [15:22:23] :-D [15:22:24] missed it :) [15:22:35] the code is in svn, it's the same code that's been running forever [15:22:44] and where in svn? [15:22:50] we just copy them to ds2 instead of domas hosting them [15:22:55] lemme dig it up [15:23:03] i have a question about the encoding of the titles [15:28:15] http://svn.wikimedia.org/viewvc/mediawiki/trunk/webstatscollector/ [15:28:34] domas is the person for that.. he wrote these back in 2008 :-D [15:36:43] Why does svn co svn+ssh://??? always ask for my passphrase twice? [15:36:53] thx apergos [15:37:09] sure [15:37:12] are you familiar with the code at all? [15:37:18] I have looked at it [15:37:29] not in depth [15:37:53] GrafZahl_: Use ssh-agent [15:38:08] can you help me figure out how the titles are encoded? [15:38:28] btw, this is diederik [15:40:27] RoanKattouw: Thanks, I'll look into it [15:51:03] apergos: URL's are encoded using UTF8, right? [15:51:17] I assume so [15:52:09] so i am looking at collector.c [15:52:37] but it doesn't look like titles are handled as UTF8 [16:12:21] you know the other person who will know a ton about these is erik z [16:12:27] since he works the heck out of [16:12:28] them [18:14:42] alolita: i've become itnerested in the visual editor, and sent a (simple) patch. i see there is no image support. is anyone working on that//what things can I do to help? fixing bugs? thanks [18:22:56] joancreus: There are plenty of VE bugs in Bugzilla, patches for any and all of them are appreciated. So yeah, reporting and fixing bugs is how you can help [18:23:14] joancreus: thanks - you can indeed help; can you ping flipzagging to do more on visual editor bug fixes, documentation and testing [18:23:33] alolita: I go by neilk_ [18:24:03] neilk_: that works :-) flipzagging is now neilk_ [18:24:33] "now" [18:24:39] joancreus: nobody is working on it per se. I was thinking of adapting some of our existing image uploading tech to work with it. Will be happy to either give you a plan for doing this or help you through it [18:24:41] Since like, a year ago? :) [18:24:49] well, since birth actually [18:25:15] joancreus: where's your patch? [18:25:40] neilk_: grayed out buttons without visual feedback [18:25:45] let me look for the link [18:25:52] it was added to the SVN afaik [18:26:28] joancreus: I think I saw that [19:07:01] good morning, nerds. [19:07:36] straw poll for those of you who live in the bay area: which says "home" more to you: Sutro Tower or the Golden Gate Bridge? [19:08:37] I have nicer feelings about the GGB, Sutro Tower is more like an in-crowd reference. [19:08:50] like, I would use Sutro tower to say "this is for locals" [19:09:02] +1 [19:09:10] This goes to Tina's thing about how movies get SF wrong [19:09:22] 1) No, the Golden Gate Bridge isn't in your back yard [19:09:47] 2) You never see Sutro Tower. In the real city, you can see Sutro Tower almost everywhere, when it's not foggy [19:09:55] we kind of want it to be "this is a locals" thing. [19:10:05] s/never see Sutro Tower/never see the GGB/ ? [19:10:14] oh you mean you never see it in the movie [19:10:17] i used the phrase "shibboleth" [19:10:22] sutro is a shibboleth for locals. [19:10:29] but I can't pronounce shibboleth!! [19:10:38] shi-bo-leth [19:11:02] sorry, in-joke for people who know bronze age Jewish lore. [19:11:13] i never see ggb. but sutro tower overlooks everything. [19:11:18] Exactly [19:11:18] it's like the eye of sauron. [19:11:22] And in the movies, it's the other way around [19:11:51] Sutro Tower has appeared in one movie that I know of -- Escape from Alcatraz. However, in that case it's an anachronism since it wasn't built at the time the movie documents. [19:12:05] i love car chases in sf. [19:12:17] There was also this (canceled?) TV series that took place in SF and was actually shot in SF [19:12:22] I forgot the name but Trevor knows [19:12:24] neilk_: unfortunately I know the in joke cause I was part of the discussions around the shibiboleth access control stuff :-/ [19:12:26] we're downtown! take a left! oh, now we're on geary! take a left! now we're at golden gate! [19:12:28] The Streets of San Francisco [19:12:49] i was really depressed when i found out that even the "on location" shots in "monk" were actually greenscreened. [19:12:59] Some people have posted some hilarious analyses of the car chase in Bullitt [19:13:04] It's got stuff like a major car pile-up on 280 near Potrero Hill, and a sniper in a building on Union Square (or was it Civic Center, I forget) [19:14:28] jorm: I saw Hitchcock's "The Birds" on TV once. It's like 30 seconds into the movie that they reveal it takes place in SF (shop sign with a GGB picture), but I instantly recognized that the opening shot was one of the corners of Union Square crossing Powell Street, even before that reveal [19:14:30] ggb for me. sutro tower looks like it would walk over the hill to kill people. not a warm fuzzy feeling [19:14:30] :) [19:17:00] took my cat to the vet today. [19:17:08] she's been overgrooming. it's probably an allergy. [19:17:17] so now she's on a strict diet of "rabbit, and only rabbit." [19:17:19] I think of Sutro Tower as being the source of the evil paganist radiation [19:17:44] also, i'm going to point out something: [19:17:58] number of people killed by sutro tower: 0. number of people killed by ggb: ~75 [19:18:04] or is it up to 100? [19:18:15] it reminds me of the wicker man. [19:18:32] By "killed" do you mean only $killer !== $victim cases? [19:18:43] Or are you just counting suicides [19:19:40] man. people really, REALLY want to bikeshed on the diff colors thing. and i'm just not getting into it. [19:19:45] *neilk_ ponders $killer != $victim versus $killer !== $victim [19:19:58] frickin nerds [19:20:07] can'tlivewithemcan'tshootem? [19:20:07] heh [19:20:34] i have a feeling that this one guy really *really* wants to have a patch commit under his belt. [19:20:43] and that's his major thing. [19:20:53] jorm: You talking about Edokter? [19:21:07] He's got a number of patches committed already [19:21:09] >6 at lesat [19:21:36] neilk_: I guess the loose comparison case could apply for schizophrenia maybe? [19:21:45] yeah! [19:22:02] i'm guessing. [19:22:15] *apergos tries to parse lesat at least 5 tims before realizing that it was *not* a foreign word [19:22:27] RoanKattouw: Your honor, it's true I tried to kill him, but I was coerced! [19:22:36] groannnnn [19:23:14] apergos: The fact that you typo "times" as "tims" in a sentence that contains "parse" is almost like a Freudian slip [19:23:23] almost eh? [19:23:33] except that I have a typo (at least one) in practically every sentence [19:23:44] unless I'm vewy vewy caweful [19:23:59] *apergos grabs elmer fudd's rifle [19:24:02] typo hunting! [19:43:17] rmoen: ping [20:06:37] hashar: Awesome news about Testswarm [20:06:48] Great job [20:07:13] hashar: Where can I find the script that submits the jobs to testswarm ? [20:07:14] Krinkle: thx :p [20:07:27] brion raised an issue though, the dashboard on the main page is missing. Was it a local hack? [20:07:33] Yes [20:07:36] among many other things [20:08:09] There is an if (false) {} statement in the homepage code but enabling that does not give you what I have on the Toolserver. [20:08:27] plus the original one disabled in (false) is very inefficient and would be bad to run live on the homepage [20:08:35] http://integration.mediawiki.org/checkouts/mw/trunk/r106834/index.php/Main_Page [20:08:38] something looks wrong though [20:08:43] "Jump to: navigation, search" [20:08:45] do you have patches for your local hackes? Would be great to push them upstream [20:08:51] and the empty category links element [20:09:09] the job submitter is in subversion IIRC [20:09:10] hashar: No, they are not compatible in any way. Totally out of sync with upstream code. [20:09:16] doh [20:09:27] There are open feature requests / bugs about all I did though, and I will push them upstream [20:09:50] good to know! [20:09:57] the fetching script is at svn.wikimedia.org/svnroot/mediawiki/trunk/tools/testswarm [20:10:19] but they are deployed by puppet! [20:10:27] sure [20:10:31] so I usually do the development using git puppet repository [20:10:43] and from time to time copy that to my svn working path and svn commit that [20:10:46] hashar: I'd like to change it so that it doesn't submit jobs for all browsers that are ever known in testswarm [20:10:47] it is ugly [20:11:12] compare with here: http://toolserver.org/~krinkle/testswarm/user/MediaWiki/ [20:11:18] that feature is built-in [20:11:32] http://integration.mediawiki.org/testswarm/user/MediaWiki/ [20:11:49] although we can enable it for all browsers one day if we want, right now it makes it near impossible for a revision to be marked 'complete' [20:11:51] I have hardcoded something like mobile + beta [20:11:55] yeah [20:12:37] toolserver uses 'popularbeta' [20:12:46] https://svn.wikimedia.org/viewvc/mediawiki/trunk/tools/testswarm/scripts/testswarm-mediawiki-svn.php?revision=105930&view=markup#l89 [20:12:57] let me change that [20:13:18] popularbetamobile [20:13:30] I have added mobile because someone wanted mobile devices too [20:13:52] maybe later :) [20:13:57] I wouldn't recommend doing that now [20:14:05] and mediawiki doesn't support that for the default skin anyway. [20:14:26] would only result in revisions being marked broken everywhere [20:16:56] later we could add additional runs for other skins and other platforms, but that needs better support in TestSwarm first. There's proposals/RFC's for that on the jQuery Testing wiki (TestSwarm falls under the jQuery Testing working group of which I'm a member as well), we should probably look at that and do it and commit upstream (I have commit access so we can skip maintaining patches and commit right away with pull request, no need to deploy ou [20:16:56] branch first). [20:17:18] I think first priority now is to make our JSTesting branch work and make testing work with Special:JavaScriptTest (instead of qunit/index.html) [20:17:44] do you remember the state of that ? [20:19:09] yes [20:19:14] sorry on phone with my mother :-))) [20:22:56] so many stuff to handle [20:24:48] Krinkle: fixing to use popularbeta :) [20:24:54] ok [20:25:06] doing so will likely uncover a bug, but I'll show you if and when that happends [20:25:20] I fixed that at toolserver though. [20:27:03] *brion waves [20:27:41] oh and somehow testswarm fetched all revisions since r105000 or something :D [20:27:44] that was a lot of jobs [20:28:05] Krinkle: here is the ops request : https://gerrit.wikimedia.org/r/1673 [20:28:20] yeah, got that (stalk word on "testswarm") [20:28:30] hashar: ow.. [20:29:07] saved in svn too [20:29:27] hashar: hm.. I thought I would see a diff where it changes the svn revision to check out of that script, but it's not doing that, is it ? [20:29:37] brion: the old testswarm dashboard is no more :) Meanwhile you can use http://integration.mediawiki.org/testswarm/user/MediaWiki/ [20:29:38] duplicated/copy ? [20:29:46] tx [20:29:51] brion: MediaWiki is the user submitting jobs to testswarm [20:30:14] brion: hashar: the dashboard never existed in testswarm repo, it was a local hack I did on TS. [20:30:17] Krinkle: the git and svn are duplicates yes. [20:30:20] ko [20:30:22] ok* [20:30:37] still not making it through all tests? :( [20:30:39] Krinkle: the script is really in puppet git now. I will probably delete the svn one [20:31:05] hashar: I'll need to be set up properly to make commits there. [20:31:16] brion: I believe it is a race condition between the cronjob that submit a new job every minute and the clients that only do 3 or 4 modules per minutes :-) [20:31:17] haven't done that ever yet. [20:31:37] :( [20:31:42] brion: what do you mean ? [20:32:02] http://integration.mediawiki.org/testswarm/job/213/ <- only a small subset of tests seem to have been run [20:32:14] from memory: TestSwarm distributes tests to clients, starting with the latest one, and then working it's way back. [20:32:23] Krinkle: you will need to have your account enabled on labs then some doc is at https://labsconsole.wikimedia.org/wiki/Git [20:32:38] and if latest X tests are complete it'll start at the back instead. [20:32:45] not quite sure what the exact logic is. [20:32:59] ops can assist you to setup your puppet clone :) [20:35:32] ssh: Could not resolve hostname ssh: nodename nor servname provided, or not known [20:35:39] fatal: The remote end hung up unexpectedly [20:35:57] oh, I see [20:36:10] two slashes, not one in the protocol :D [20:39:47] Wow, it's taking forever to check it out [20:39:54] 0-76% went super fast [20:39:59] Receiving objects: 76% (7267/9514), 30.93 MiB | 182 KiB/s [20:40:59] Krinkle: There are two big objects, 15-20MB [20:41:01] Blame Ryan [20:41:42] mistakes that were reverted ? [20:43:14] No [20:43:16] But even if they were [20:43:25] That object is gonna be in the repo forever [20:43:46] (There's a .jar file and another file of a similar type, each contains an entire software package) [20:44:21] 28M GeoIPcity.dat [20:44:41] 18M gerrit-2.2.1.war [20:44:43] Those two [20:45:58] k [20:49:22] RoanKattouw: can't we edit the history and pretend the commit / object never existed? [20:49:30] (then everyone repo will be screwed up probably) [20:50:03] Yes [20:50:12] The SHA-1 hash of every single commit will change [20:50:17] Essentially, we'd be switching to a whole new repo [20:51:47] does the sha1 change because it includes the parent sha1? [20:52:24] brion: the testswarm job submitter catched up with our SVN repo so we have box filling in green now : http://integration.mediawiki.org/testswarm/user/MediaWiki/ [20:54:59] yay! [20:57:44] hashar: Yeah it depends on its parent somehow [20:57:47] *RoanKattouw needs to look up how [21:32:51] hashar: Hm.. any idea as to why the layout is wrong in production mode ? [21:32:55] http://integration.mediawiki.org/checkouts/mw/trunk/r106834/index.php/Main_Page?debug=true [21:32:59] http://integration.mediawiki.org/checkouts/mw/trunk/r106834/index.php/Main_Page [21:34:03] Krinkle: looking [21:34:27] also for me, load.php requests seem to be taking for ever. [21:34:37] perhaps we can do something towards caching ? [21:35:17] if we're going to use Special:JavaScriptTest in the future, things need to load a bit faster then they do now or the server likely won't perform well. [21:35:30] cache invalidation would theoretically never happen. [21:36:32] found it [21:36:35] mediawiki.legacy.shared isn't loaded [21:36:39] it is loaded in debug but not in production mode [21:36:50] mw.loader.getState('mediawiki.legacy.shared') // "registered" [21:37:57] maybe enable APC [21:38:08] I am not sure there is any cache enabled [21:38:09] memc? [21:38:16] or memc yeah [21:38:17] I wonder why that module isn;t loaded though [21:38:20] both [21:38:22] Do you have Varnish/Squid? [21:38:23] http://integration.mediawiki.org/checkouts/mw/trunk/r106834/load.php?debug=false&lang=en&modules=mediawiki.legacy.shared&only=styles&skin=vector&* [21:38:30] looks good, mw-jump style is there [21:38:33] If not, you may want to consider using Aaron's RL file cache [21:38:37] no every request hit PHP [21:39:54] RoanKattouw_away: Strange, 'mediawiki.legacy.shared' is loaded and applied in debug mode, but getState keeps saying "registered" [21:40:04] hm [21:40:17] http://integration.mediawiki.org/checkouts/mw/trunk/r106834/index.php/Main_Page?debug=true [21:40:17] I got Unexpected non-MediaWiki exception encountered, of type "PDOException" :))) [21:40:22] on http://integration.mediawiki.org/checkouts/mw/trunk/r106834/index.php/Main_Page [21:40:55] also, in debug mode it loads it through only=styles, whereas if I load it manually from console, it loads the css file directly [21:41:42] error is here: http://integration.mediawiki.org/checkouts/mw/trunk/r106834/load.php?debug=false&lang=en&modules=mediawiki.legacy.commonPrint%2Cshared%7Cskins.vector&only=styles&skin=vector&* [21:41:52] Did you forget to run maintenance/update.php after upgrading? See: https://www.mediawiki.org/wiki/Manual:Upgrading#Run_the_update_script [21:41:52] Query: REPLACE INTO module_deps (md_module,md_skin,md_deps) VALUES ('mediawiki.legacy.shared','vector','["\/var [21:41:56] yep, that's it [21:42:14] Function: DatabaseSqlite::replace/single-row [21:42:15] Error: 5 database is locked [21:42:15] ' in /var/lib/testswarm/mediawiki-trunk/checkouts/r106834/includes/db/Database.php:919 [21:43:06] too many queries at the same time maybe [21:43:14] and SQLite can not keep up ? [21:44:51] SQLite has no transaction handling worth mentioning [21:45:05] Its transaction handling consists of throwing fatal errors like that [21:45:12] Well I do remember from working with wikibots that an sqlfile cannot be queried by two processes simultaneously. Because it has to open and close the file contiously [21:45:25] whereas mysql is all in-memory, not directly onto the file (afaik). [21:45:37] sqlite-file* that is [21:46:10] Well so what's going on here is [21:46:30] Suppose you have two processes that read then write (or write than read) to the same table, in a transaction [21:46:47] MySQL will actually fake the table state, much like branching in a VCS [21:46:57] And then 'merge' the 'commits' when you run COMMIT [21:47:05] yeah, mysql can do that because it's approached as a server, not as a file. [21:47:19] you could do that in SQLite [21:47:23] okay [21:47:31] So if a transaction modifies a table that is already modified in another uncommitted transaction, MySQL does its branching thing. SQLite just kills the 2nd transaction with an error [21:47:47] so??? what do we do about it [21:48:04] just ignore it ? [21:48:19] or open yet another bug to have testswarm checkouts installed on a mysql DB :-) [21:49:06] SQLite should be good enough for now [21:51:11] well, obviously not. [21:51:37] both modules to be tested and the test suites are going to be loaded through ResourceLoade [21:51:38] r [21:51:50] The Special-page needs to be loaded properly, not have modules missing or broken. [21:52:09] also, either the error is happening every time, or it's cached. [21:52:17] (server side) [21:52:31] the testswarm install is working ok right now because it's loading the files statically from HTML [21:53:29] got it [21:53:45] so that will be a problem for the JSTesting branch right? [21:54:13] but since most queries are probably only doing read / select it might be just fine [21:54:28] resource loader uses database tables for some data [21:54:32] and it won't work otherwise [21:54:52] it'll refuse to load certain modules in that case, as we see now [21:57:03] well we will have to talk about that next year :D [21:57:16] my daughter just woke up, need to take care of it [21:57:37] please drop me emails or open bugs if you have any idea / issues! [21:59:10] have a good night!