0

I help run a small "wiki farm" with a wiki in several languages, including English (the primary), German, Spanish, French, Portuguese, Chinese, Korean, and Russian.

Recently I upgraded the server software as follows:

MediaWiki from 1.14 to 1.25.1 PHP from about 5.2.5 to 5.5.27 MySQL to 5.6.23

For most of the wikis, page loading still works.

The Russian site is the problem. Before the update, any user typing in Cyrillic characters and Cyrillic page titles could always find them. Furthermore, Interwiki links to the Russian version of a page in another language always worked; the pages always loaded. But I always knew the page titles themselves were stored as UTF-8 and were human-unreadable. Because whenever I would get a watchlist notice of a change on a page on that wiki, the link would come through as UTF-8 and display as a lot of weird-looking characters that definitely were not Cyrillic. But the links were always good and would display as Cyrillic in my browser.

But now: no more. If I follow a link, it goes to the special search page. That page brings up several results, all of them with titles I can't read. Usually the first title on the list, is the page I'm looking for.

But how do I get the system to find the page by its Cyrillic title, as it always could before?

The other wikis work just fine, as before. But not the Russian one.

Terry Hurlbut
  • 1
  • Have you a special configuration parameter in $wgLegacyEncoding (different than false)? And is the server config (Apache/nginx) the same for the Russian and the other wikis? – Seb35 Jul 31 '15 at 15:42
  • Can you show us some of the "can't read" text. – Rick James Aug 01 '15 at 05:29
  • Here's an example of the non-readable text: Либерализм_в_хриÑтианÑтве The server is the same server for the Russian as for all the others. Lately I realized I could add another page to this wiki, and MediaWiki would store the title as readable Cyrillic script. So that means going through every page, figuring out how the title should read (I can get a few clues from Interwiki links, MediaWiki search, and an on-line decoder), and "moving" the page from the unreadable text to something a Russian, at least, can read. – Terry Hurlbut Aug 01 '15 at 12:07

1 Answers1

0

Because whenever I would get a watchlist notice of a change on a page on that wiki, the link would come through as UTF-8 and display as a lot of weird-looking characters that definitely were not Cyrillic. But the links were always good and would display as Cyrillic in my browser.

That sounds like your wiki was always broken, and after the update it is broken in slightly different ways. I would start by looking at $wgDBmysql5 and $wgDBTableOptions and what charsets used in the database (page and text tables for example).

Tgr
  • 27,442
  • 12
  • 81
  • 118