Page MenuHomePhabricator

Flagged revisions not properly marked
Closed, ResolvedPublic

Description

After upgrade to MediaWiki 1.22 on plwikisource new pages are not marked as "checked" on recent changes and their view/edit/history pages. It hits either all or at least most users that has administrator/editor rights. It can be fixed for a single revision by an empty edit, even made by a user without editor rights. After the empty edit, the initial page is always "checked" by its creator. Some caching problem? Similar effect appears earlier accidentally, but never persists for longer than few hours. Currently it persists for about 3 days.


Version: unspecified
Severity: critical
URL: https://www.mediawiki.org/wiki/Help:Test_page2

Details

Reference
bz46941

Related Objects

StatusSubtypeAssignedTask
ResolvedNone
ResolvedNone

Event Timeline

bzimport raised the priority of this task from to Unbreak Now!.Nov 22 2014, 1:17 AM
bzimport set Reference to bz46941.
bzimport added a subscriber: Unknown Object (MLST).

Is it the same on all namespaces?
There were similar problems on wikis whose Module: pages weren't flagged, and IIRC the solution was adding Module namespace to $wgFlaggedRevsNamespaces. Your current configuration is:

$wgFlaggedRevsNamespaces = array( NS_MAIN, NS_TEMPLATE, NS_FILE, NS_CATEGORY, NS_HELP, 100, 102, 104 );

https://noc.wikimedia.org/conf/highlight.php?file=flaggedrevs.php

At least im Main, Category and Page namespaces. There was no need to create new pages in other namespaces last days. Is it possible to test whether flagging Module namespace would help in this case?

I noticed this efect also on ruwikisource in Main and Template namespaces. But cannot provide more tests as I have not editor rigts there. I hope this to be solved before the efect appears in MediaWiki 1.22 on wikipedias.

I was able to replicate this on mediawiki.org: https://www.mediawiki.org/wiki/Help:Test_page2 . Raising severity, since this will most likely also cause similar troubles on all Wikipedias using FlaggedRevs.

Normal vs fake load balancer testing shows that this is obviously a problem with loading data from a slave DB.

Related URL: https://gerrit.wikimedia.org/r/58118 (Gerrit Change I833481d9200e834518bb74cac0606d3d4550212e)