[20:22:29] <anomie>	 tgr: I made an error in one of the API i18n messages. Fix is https://gerrit.wikimedia.org/r/#/c/325619/
[20:38:20] <tgr>	 anomie: can you think of teams who are affected by the change, apart from the mobile apps? multimedia due to the upload api changes, is there anyone else?
[20:45:18] <tgr>	 anomie: I'm a bit unsure about ** make sure you set errorsuselocal if you interact with user-defined
[20:45:50] <tgr>	 (linux clipboard snafu)
[20:46:03] <tgr>	 ....about errorsuselocal defaulting to false
[20:46:46] <tgr>	 it will affect anything that used manual hacks and passed parsed strings to the API
[20:47:04] <anomie>	 tgr: Can't think of anyone in particular. Not sure what you're asking about errorsuselocal.
[20:47:30] <tgr>	 so it's a b/c break to some extent either way, but maybe defaulting to true would be the smaller one
[20:47:54] <tgr>	 esp. with api modules that interact with user-defined messages, like abusefilter
[20:48:45] <tgr>	 maybe bc mode should use the db when the message does not exist otherwise?
[20:54:48] <anomie>	 tgr: Well, we probably don't want BC mode to use the database since it's liable to wind up returning something very unexpected if wikis start customizing some of these messages for whatever gadgets and user scripts start actually using the new localized error messages.
[20:58:24] <tgr>	 anomie: that makes sense. still, with messages that only exist in the database it's worth making an exception IMO; currently those are all things that can be displayed via the web as well
[20:59:40] <tgr>	 abusefilter, titleblacklist, spamblacklist are the ones I can think of
[20:59:54] <anomie>	 tgr: Well, something making a web UI would probably want to be using &errorformat=html&errorsuselocal=1 anyway. Is there any sane way to tell a message "only use the database as a last resort"?
[21:01:22] <tgr>	 check if it exists and set the db flag based on that
[21:01:33] <tgr>	 which is not really sane considering parameter chains...
[21:01:44] <tgr>	 I'll write a patch for that
[21:02:49] <anomie>	 tgr: At least for AbuseFilter it doesn't seem that this actually changed anything, both before and after it doesn't seem to be loading messages from the DB.
[21:03:36] <tgr>	 I ran into it while testing TitleBlacklist IIRC
[21:05:03] <tgr>	 I guess that will mostly just affect file uploads since nothing else uses page creating APIs heavily, and UploadWizard etc. needs to be fixed anyway
[21:21:42] <MatmaRex>	 anomie: tgr: for a good start, you broke VE. ;) https://phabricator.wikimedia.org/T152547
[21:48:02] <anomie>	 MatmaRex, tgr: I put up https://gerrit.wikimedia.org/r/#/c/325641/ that should fix VE, although I haven't ruled out changing the ApiParse check back to ignoring most duplicates if that sort of thing turns out to be more widespread.
[21:48:38] <ostriches>	 MatmaRex: So nothing important? ;-)
[21:49:01] <anomie>	 tgr, bd808: Any chance you could look in the kafka thing to see whether action=parse is often called with more than one of 'oldid', 'page', and 'pageid'?
[21:50:01] <tgr>	 anomie: I have a meeting but can do in an hour or so
[21:50:34] <anomie>	 tgr: Ok, thanks. I'll probably have to see it in IRC scrollback tomorrow then.