If you look at https://en.wikipedia.org/w/api.php?action=sitematrix&format=xml
the Bihari wiki is there under code bh but has no name attribute
Description
Details
Related Objects
- Mentioned In
- T233096: Operation 'login' ended with result 'NeedToken'.
T39459: [Task] Don't try to add labels in non-existing languages: restrict to Language::isKnownLanguageTag
T99888: Link adaptation does not happen when the language code involved has a different domain name
rMW6e8094da2b51: Revert "Remove invalid language code "bh" (alias of "bho")" - Mentioned Here
- T91240: "bh" is not a valid language code
Event Timeline
The "bh" language code was removed (T91240: "bh" is not a valid language code) and bhwiki uses "bho" but sitematrix incorrectly believes the wiki is in "bh" language. There is no name for the incorrect code.
Where does sitematrix fetch the incorrect code from?
On the cluster, it appears it gets the list of codes from the file langlist in the operations/mediawiki-config repo. But since the wiki itself is still "bh.wikipedia.org" with database name "bhwiki", I'd be wary that simply changing that would break something.
IMO https://gerrit.wikimedia.org/r/193802 should be reverted. Names.php contains a bunch of "invalid"/"wrong" language codes (e.g. als, be-x-old) that are needed for backward compatibility and also to make the WMF cluster work in its current configuration.
Whether this fixes the problem or not, I can't say, but it should still happen regardless. A proper solution to the mess of legacy language codes needs to be figured out, not just a piecemeal approach that breaks various things along the way.
There is a short discussion on the problem here too https://en.wikipedia.org/wiki/Wikipedia_talk:AutoWikiBrowser/Bugs#TypeInitializationException_Unable_to_log_in
I get the feeling that the people who are involved here fail to grasp that AWB users all over the world are hors de combat. We need to have https://gerrit.wikimedia.org/r/193802 reversed
I have uploaded an AWB snapshot release to https://sourceforge.net/projects/autowikibrowser/files/autowikibrowser/Snapshots/ named 5.5.6.0-S, this is the 5.5.6.0 release with a workaround for the Bihari wiki not having a name.
Well, I was doing the same in parallel... :)
http://tools.freeside.sk/download/AWB-5561-sitematrix-fix.zip
Names.php contains a bunch of "invalid"/"wrong" language codes
"A bunch" is an exaggeration. Most of the incorrect language codes are valid under at least one standard and/or are not used as a wiki's language code. "bh" is a unique case of egregious typo.
I am temp admin on bhwiki. I list here some old discussions (I never participated, I am still not much technical to understand these things):
https://phabricator.wikimedia.org/T41968
https://bugzilla.redhat.com/show_bug.cgi?id=677570
plz see!
Change 211119 had a related patch set uploaded (by Nemo bis):
Fetch the actual wgLanguageCode when we miserably failed
Change 211906 had a related patch set uploaded (by Nemo bis):
Revert "Remove invalid language code "bh" (alias of "bho")"
Change 211906 merged by jenkins-bot:
Revert "Remove invalid language code "bh" (alias of "bho")"
Is the priority of this task still correct or should it be lowered?
This task has been "Unbreak now" priority but no activity for nearly one month.
https://gerrit.wikimedia.org/r/#/c/211906/ was reverted in https://gerrit.wikimedia.org/r/#/c/211906/
(And Nemo_bis' patch in https://gerrit.wikimedia.org/r/#/c/211119/ for SiteMatrix is still awaiting review.)
Somebody needs to review Nemo_bis' patch in https://gerrit.wikimedia.org/r/#/c/211119/ for SiteMatrix .
CC'ed as reviewers in Gerrit are @Anomie and @Nikerabbit - is that correct or are there better candidates?
Has this been fixed? https://en.wikipedia.org/w/api.php?action=sitematrix shows a name after the "code": "bh"
Change 211119 abandoned by Thiemo Kreuz (WMDE):
[mediawiki/extensions/SiteMatrix@master] Fetch the actual wgLanguageCode when we miserably failed
Reason:
5 years old. The linked ticket is closed as resolved.