[00:37:43] <bd808>	 anomie: the lua perf data is on wikitech again now so if you want to take a look at https://wikitech.wikimedia.org/wiki/Module:SAL you can actually tell what's going on. I tweaked it a bit over the weekend and cut down the reparse time Server_Admin_Log by a few seconds.
[00:38:18] <bd808>	 I think the overhead that is left is mainly in internal argument marshaling 
[01:02:18] <bd808>	 heh. I'm looking at the action api data we pipe into hadoop for some stats on toollabs usage and found action=somethingdoesntmatter called 5 times. cracked me up
[03:09:38] <tgr>	 bd808: can you look at https://gerrit.wikimedia.org/r/#/c/313932/ again?
[03:42:27] <bd808>	 tgr: heh. sorry for being too quick to +2 before :)
[14:22:21] <anomie>	 bd808: I tried a few things, couldn't make any positive difference.
[15:57:23] <bd808>	 anomie: thanks for taking a look. The biggest difference I made when tweaking it was adding the check for [[ and {{ before handing off to the parser
[15:58:36] <bd808>	 I might try a version that puts multiple log messages in a single template call.
[16:03:38] <anomie>	 bd808: Sanity check: can you just subst the template? Or do you want improvements to the module to apply retroactively to old log messages?
[16:04:30] <bd808>	 subst would work I think. I didn't even think of that :/
[16:04:40] <bd808>	 my wiki fu is weak
[16:23:22] <bd808>	 anomie: hmmm... some log entries blow up the lua with subst:. "Lua error in Module:SAL at line 13: bad argument #1 to 'find' (string expected, got nil)."
[16:23:55] * bd808 pokes around a bit
[16:24:46] <anomie>	 bd808: I see no reason for that error message there... What was the specific template invocation that blew up?
[16:25:32] <bd808>	 {{subst:SAL entry|1=14:20 akosiaris@puppetmaster1001: conftool action :   set/pooled=yes; selector: mw1171.eqiad.wmnet (tags: ['dc=eqiad', 'cluster=appserver', 'service=apache2'])}}
[16:29:10] <anomie>	 bd808: https://wikitech.wikimedia.org/w/index.php?title=Template:SAL_entry&diff=1008815&oldid=958331 might help, so the #invoke gets substed too
[16:30:09] * bd808 tries
[16:31:16] <bd808>	 yeah that voodoo seems to fix it
[16:39:18] <bd808>	 TIL anomie doesn't have a github account
[16:46:04] <anomie>	 bd808: ?
[16:46:51] <bd808>	 I was going to @ping you in the commit message to tools-stashbot but couldn't find a github account for you. Your commits to core are not linked to an account
[16:47:52] <anomie>	 Ah. Yeah, I don't ever use github.
[16:56:13] * greg-g would have assumed as much :)
[18:21:20] <legoktm>	 bd808: https://github.com/anomiex !
[21:16:31] <ostriches>	 Reedy: I did a +2 on your backport for https://phabricator.wikimedia.org/T148957. Is that all that's needed (doc fixes), or do we need anything further? ie: do we want gerrit 322772? 
[21:16:54] <Reedy>	 I didn't manage to find a way to replicate the issue on demand
[21:17:00] <Reedy>	 I think it is, but could do with confirming it