[09:39:41] Change on 12mediawiki a page Wikimedia Labs/Tool Labs/List of Toolserver Tools was modified, changed by Nemo bis link https://www.mediawiki.org/w/index.php?diff=800236 edit summary: /* Intro */ +[[mailarchive:toolserver-l/2013-October/006337.html|Top 50 tools by hits]] [13:03:34] valhallasw: around? [13:04:25] valhallasw: Upload failed. Reason: execv() arg 2 must contain only strings (check log above for details) on the gerrit-patch-uploader [13:05:20] valhallasw: also, it should mention somewhere that in git, you have to use the following incantation: git format-patch -1 --stdout HEAD --no-prefix (too me way too long to discover, eh) [13:16:36] Coren|Away, mutante: gluster on bots project went read-only [13:16:42] something I can do to fix it? [13:20:19] MatmaRex: hmmm [13:20:30] MatmaRex: what does the log say? [13:23:07] valhallasw: gah, i closed it already, but there was nothing else [13:23:18] valhallasw: the git commit command and this below [13:23:38] let me try again [13:24:12] MatmaRex: ok, so it was in the git commit... what did you enter in the author field? [13:24:44] http://i.imgur.com/aFpPDxQ.png [13:24:54] could the "ń" be the culprit? [13:25:37] yep, I forgot to encode it as utf-8 [13:26:05] valhallasw: I take it that the new web service scheme is full of win? [13:26:32] I'm especially enjoying the quick web service restarts now ;-) [13:26:39] (unrelated, while i'm here: who do i bribe to get https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/Access_Request/Bartosz_Dziewoński fulfilled?) [13:28:02] MatmaRex: I accept cash an all major credit cards. :-) [13:28:12] ;) [13:28:56] "Successfully added Bartosz Dziewoński to tools." [13:29:32] But, FYI, there is currently an issue with the new NFS server that may end up with your home and/or project directories ending up owned by root. Tell me if that's the case. [13:29:47] (I can fix manually) [13:29:47] MatmaRex: could you try again? should be fixed now [13:30:06] valhallasw: sure, on it [13:30:12] Coren|Away: hmm okay, let me try to ssh in [13:30:53] MatmaRex: and maybe I should move away from patch -p0 to something that accepts more patch styles :/ [13:31:14] valhallasw: i suggest git apply [13:31:26] valhallasw: it has the fine property of accepting output of git diff [13:31:42] but not a simple unified diff, I think [13:31:48] oh, maybe it does [13:32:02] git diff is a simple unified diff? [13:32:17] but with extra headers [13:32:34] and commit IDs, otherwise it's unified [13:33:09] valhallasw: you might just want to try -p0 and then -p1 if it fails [13:33:34] Coren|Away: Creating directory '/home/matmarex'. [13:33:34] Unable to create and initialize directory '/home/matmarex'. [13:33:53] (that was on `ssh matmarex@tools-login.wmflabs.org`) [13:35:00] valhallasw: yay, it worked, but redirected me to https://gerrit.wikimedia.org/r/89514%1B (which is a 404) [13:35:27] matanya: Try ye now. [13:36:04] Coren|Away: yup, worked. thanks [13:37:22] MatmaRex: ugh, ansi escape that wasn't filtered out. Not sure why it worked for me in the first place, though... [13:39:50] ok, and switched to git apply now. [13:53:42] MatmaRex: ok, switched to git apply [13:54:14] and fixed an newlines issue [13:54:20] oh, the joys of software development ;-) [13:54:36] gah, is the info at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/Help#Joining_and_creating_a_Tool_account up-to-date? [13:54:49] `become wikiblame` doesn't work for me (i just created the tool account) [13:55:13] MatmaRex: you need to log out and in again [13:55:14] (a few minutes passed already) [13:55:17] otherwise your groups are not updated [13:55:30] i did that [13:55:32] hm okay, works now. [13:55:33] thanks :) [13:55:53] valhallasw: does your patch uploader support adding new patchset to existing changesets? [13:55:57] (actually, let's test that) [13:59:40] (it does work. sexy.) [14:00:12] valhallasw: i see you added " Allowed patch formats:" on the main page, is that up-to-date now that you're using git apply? [14:00:23] i just submitted a patch generated using `git format-patch -1 --stdout HEAD` (no --no-prefix) and it worked [14:10:08] MatmaRex: I suppose I should also implement git am support ;-) [14:11:10] valhallasw: anyway, that's a really cool thing :D i forwarded your wikitech e-mail to a certain guy who loves adding patches as bugzilla attachments [14:11:19] cool :-) [14:16:02] ok, now. somebody tell me how do i run https://github.com/MatmaRex/wikiblame on labs. :D [14:16:19] 1. is there a way for me to install custom gems (the `gem` command)? [14:16:50] 2. is there a way to set up something to redirect http://tools.wmflabs.org/wikiblame/ to my own little server running on labs? (reverse proxy?) [14:33:34] MatmaRex: for (2): in 'the new web system' you have your own private lighttpd instance [14:33:41] and you can basically tell it to do whatever you want [14:50:25] MatmaRex: ok, now it does git apply, then patch -p0, then patch -p [14:50:27] -p1 [14:50:35] that should allow most patches to 'just work' [14:52:49] yay [16:05:30] hi, per https://wikitech.wikimedia.org/wiki/Help:Instances someone with "Sys Admin" can make a new instance, can I request "Sys Admin" right or someone can make an instance for me? [16:16:52] hmmm, how do i debug what's wrong with http://tools.wmflabs.org/wikiblame/cgi-bin/wikiblame/cgi.rb ? [16:17:44] hmm actually, nevermind for now, i'm getting to something [16:33:46] ok. now, how do i debug it? :D [16:36:10] gah [16:36:17] i hate cgi so much [16:39:12] why? [16:45:37] actually, scratch that [16:45:45] i hate linux package managers. with passion [16:45:53] local-wikiblame@tools-login:~/cgi-bin/wikiblame$ ruby1.9.1 --version [16:45:58] wanna bet what this yields? [16:46:12] (definitely not 1.9.1) [16:47:17] labs apparently has four ruby executables [16:47:26] (and two ruby versions) [16:47:59] and only one gem executable [16:48:06] uh, no [16:48:09] three gem executables [16:48:12] and only one version [16:48:36] and naturally, none of the rubies can find gems installed by any of the gems [16:48:49] which is the direct cause of http://tools.wmflabs.org/wikiblame/cgi-bin/wikiblame/cgi.rb not working. [16:49:01] /rant [16:49:15] MatmaRex: 1.9.3? :P [16:49:31] i have yet to find a server where ruby versions are not hopelessly broken [16:49:38] usually i just install rvm, except that it refuses to work here [16:49:40] meh [16:50:10] MatmaRex: it's also something strange in the ruby world where you're expected to compile your own interpreter [16:50:28] valhallasw: um, what? since when? [16:50:49] i use rvm because all packages of all linux distros i tried are completely fucked up [16:51:08] MatmaRex: rvm compiles your own interpreter [16:51:10] last time I checked [16:51:26] it can compile it, it can look for binary versions too, or something [16:51:34] (neither way worked when i tried on labs a few hours ago) [16:51:42] oh, ok. [16:52:04] still I'm unsure why the packagers seem to do fine with python, but not with ruby [16:52:27] valhallasw: do you use the old python 2.x? [16:52:37] 2.7 at the moment, yes [16:52:45] valhallasw: the ruby situation would be nice if everyone still used 1.8 instead of 1.9 or 2.0 [16:53:05] (the normal 'ruby' packages are still on 1.8.7 afaik) [16:53:16] (which is kinda ancient and *everyone* has moved on to 1.9) [16:53:35] so the `ruby` executable is always useless [16:53:56] there is a ruby19 package, or something [16:54:11] which depending on the weather install randomly named executables in random places [16:54:41] and then ruby gets all confused about lib paths [16:55:03] MatmaRex: oh, ruby 1.8 is *really* ancient [16:55:05] I never realised [16:55:06] throw in gem into the mix, and you lose any hope of unraveling it [16:55:20] it's python 2.3-ancient [16:56:03] i wonder if i can bypass gem and just hardcode paths to all gems i installed. [17:02:13] man, why did i have the bright idea to include some inline C in that [17:02:33] (inline C++, in fact) [17:03:21] also, why does it work when i run the cgi script from command line, but not from, well, cgi [17:07:50] MatmaRex: paths [17:07:53] probably [17:08:04] $PATH in CGI is just /bin, I think [17:08:39] MatmaRex: it's easier if you use lighttpd, as you get error logs then [17:08:54] just putting and empty .lighttpd.conf in ~ should do the trick [17:09:01] and maybe then run webservice start [17:12:43] valhallasw: hm, where do i get those logs? [17:12:58] ~/error.log [17:13:22] local-wikiblame@tools-login:~/cgi-bin/wikiblame$ cat ~/error.log [17:13:22] 2013-10-13 17:12:41: (log.c.166) server started [17:13:25] helpful. :D [17:13:34] hm okay [17:13:40] how do i make it run my cgi scripts? [17:14:04] (or rather, my full-fledged rack application shoehorned into a cgi script) [17:14:44] once lighttpd is started, it should take over [17:15:10] it did, now i'm getting a 404 instead of the 500 that i had earlier [17:15:27] (i stopped it, let me wrestle with cgi some more for now) [17:15:31] (the enemy i know.) [17:18:19] valhallasw: i love you, it was the PATH in fact [17:49:56] hmm [17:50:31] if i have an expensive cgi script that could potentially run for days when triggered with the right parameters, am i expected to implement a timeout myself? or is there one hardcoded? [17:51:18] MatmaRex: I don't think there is a hard-coded one [18:31:03] There is no hardcoded timeout, but clients will usually give up after a while. [23:44:42] Coren|Away, ping