[00:10:40] <quiddity>	 http://img02.deviantart.net/690d/i/2012/347/4/8/log_from_blammo___ren_and_stimpy_pepakura_by_hebi69-d5nyq2h.png
[00:13:33] <ori>	 TimStarling: regarding the Parser-bug-that-wasn't, I think it is caused by T109540, which (happily for you) also falls within your area of expertise. There is a bug in ebernhardson's strtr() code (since merged upstream) that I can't reproduce, even when simulating high concurrency. It has to do with eviction of pattern cache items. See https://phabricator.wikimedia.org/T109540#1591281. The code is https://github.com/wikimedia/operatio
[00:13:33] <ori>	 ns-debs-hhvm/blob/c55d1350da27efa48f02e5defdec8fe8e2d0954c/debian/patches/D40473-strtr.patch . You're not obliged to look at that one but I would greatly appreciate it, especially since the cause my be readily apparent to you.
[00:15:50] <ori>	 *may
[00:16:46] <legoktm>	 RoanKattouw: bd808 has a bunch of open patches for this, the bug is https://phabricator.wikimedia.org/T89169
[00:17:21] <ori>	 oh cool
[00:19:12] <RoanKattouw>	 "via an undocumented extension of the error handler callback data"
[00:19:15] <RoanKattouw>	 oh good
[00:25:43] <legoktm>	 ori: I would have protected the page, but it looks like the edit war has stopped :/
[00:44:38] <Krinkle>	 TimStarling: Ah, so it's a matter of server configuration (or choice of web server software). 
[00:45:07] <TimStarling>	 yeah
[00:45:13] <Krinkle>	 I imagine in that case, a bunch of things would redirect, right? Also certain index.php and api.php urls
[00:45:48] <Krinkle>	 Although it'd be annoying for the guy running that server having all load.php urls redirect - especially since the redirect is UA-agnostic.
[00:45:58] <Krinkle>	 kills performance.
[00:46:27] <TimStarling>	 if there is a query string with a dot in the last parameter, yes the redirect will be hit
[00:47:08] <TimStarling>	 when roan wrote that patch, before r89248, the dot could be pretty much anywhere in the query string
[00:47:59] <TimStarling>	 but after r89248 it needs to be near the end with alphabetic characters after the dot
[00:48:06] <Krinkle>	 TimStarling: Hm.. okay, yeah, then we're fine for load.php
[00:48:15] <Krinkle>	 Since urls typically end with &version=
[00:48:31] <TimStarling>	 yeah, should be fine then
[00:48:35] <Krinkle>	 Thx
[00:49:16] <Krinkle>	 We may want to emit some kind of warning if haveUndecodedRequestUri doesn't work on the current server, so the site admin can fix their server or consider using different cgi software
[00:50:20] <TimStarling>	 maybe in the installer
[00:50:29] <Krinkle>	 Yeah
[00:50:32] <Krinkle>	 Too bad it's a whitelist
[00:51:03] <TimStarling>	 well, that is because we were limited in what we could easily test
[00:51:27] <TimStarling>	 if the installer tested for the actual behaviour, then it could report when a server should be whitelisted but isn't
[00:53:38] <TimStarling>	 I'm not sure if it's worth doing a big lot of work for this though
[00:53:59] <TimStarling>	 once China stops using IE6, we can remove the whole thing
[00:54:26] <TimStarling>	 they are down to 3.1% now
[01:02:42] <Krinkle>	 TimStarling: yeah
[01:02:56] <Krinkle>	 Or potentially do it in Varnish for MSIE6 ua's only.
[01:03:03] <Krinkle>	 :sick:
[01:04:09] <TimStarling>	 1.3% according to http://gs.statcounter.com/#browser_version_partially_combined-CN-monthly-201107-201507
[01:04:25] <TimStarling>	 maybe it is time to do it already
[01:04:32] <Krinkle>	 http://dev.modern.ie/ie6countdown/
[01:05:17] <TimStarling>	 yeah, that's where I got 3.1% from
[01:05:43] <TimStarling>	 but I wanted to see a time series, and statcounter has that
[01:06:07] <Krinkle>	 Hm.. it says .66%
[01:06:11] <Krinkle>	 Oh, it's 3.1 in list view
[01:06:12] <Krinkle>	 odd
[01:06:33] <Krinkle>	 Ah, 0.66 of world usage
[01:06:39] <Krinkle>	 within China
[01:06:46] <Krinkle>	 0.99% total
[01:07:47] <Krinkle>	 nice
[01:09:10] <Krinkle>	 IE11 surpassed Firefox 5+
[01:09:20] <Krinkle>	 both only 5%
[01:09:22] <Krinkle>	 that seems odd
[01:09:35] <Krinkle>	 Oh within China
[01:09:36] <Krinkle>	 right
[01:09:47] <Krinkle>	 I guess Firefox isn't popular there at all
[01:27:14] <tgr>	 anomie: do we hold the team meeting tomorrow? it's the two of us
[18:05:25] <legoktm>	 anomie: are you coming to the API namespace office hour?
[18:05:34] <anomie>	 legoktm: Already in the channel
[18:05:45] <legoktm>	 :)
[21:22:55] * robla starts figuring out Architecture Committee/RFC meeting logistics for this week
[21:54:05] <TimStarling>	 why does the mysql driver in HHVM give strings for integer columns? that is a compatibility break isn't it?
[21:55:30] <robla>	 mornin' Tim!  in answer to your question: *shrug*
[21:57:59] <TimStarling>	 hmm, actually it looks like it is that way in PHP as well
[21:58:04] <TimStarling>	 I guess it is a version difference
[22:38:01] <legoktm>	 TimStarling: I think https://gerrit.wikimedia.org/r/#/c/210856/ is now ready for review and hopefully merging :)
[23:01:11] * AaronSchulz sends https://gerrit.wikimedia.org/r/#/c/234837/ to csteipp
[23:02:12] <csteipp>	 AaronSchulz: since I haven't read the docs-- when are DeferredUpdates actually run, end of web request?
[23:02:23] <AaronSchulz>	 yeah
[23:02:46] <AaronSchulz>	 after the user gets the response
[23:02:50] <csteipp>	 Will the user get an exception or soemthign if it fails?
[23:03:18] <AaronSchulz>	 no
[23:03:48] <csteipp>	 Are we logging it somewhere?
[23:05:43] <AaronSchulz>	 the usual exception logs yeah
[23:16:00] <csteipp>	 AaronSchulz: I'll look at it later
[23:17:09] <AaronSchulz>	 heh