Mediawiki and SQL funsies

[ysengrin]

Moving the host for the wikis to a different location. As part of that, they need to be on 1.24.1 instead of 1.23.1. This morning I start upgrading them, making the decision to bring them up on nice, clean Mediawiki installs, to weed out the cruft from several previous versions. The two little wikis go pretty smoothly, with only a minor issue about permissions on a couple of binaries. Then I flip the main wiki.

I should be seeing this:

Cúchulainn split it with his sword when it failed to cry out under his protégé ...

Instead I see this:

Cúchulainn split it with his sword when it failed to cry out under his protégé ...

It's a known issue, caused by differing character encoding in MySQL tables. It pops up because UTF8 Unicodes were not supported in older (now ancient) MySQL versions. Unfortunately, the database of the big wiki dates back to those ancient days. I've been plastering over the issue, but this last MediaWiki update refused to let the encoding mismatches slide. In short, the update exposed all those non-conforming bits of text. There were a lot of them, some in article titles.

So, I spent the evening firing up a copy of the wiki from yesterday's backups and invoking my SQL-fu to whip the database into conformity. I seem to have been successful, as the Mediawiki update didn't repeat the encoding snafus.

Now I just need to copy the test DB over the live one, which I'll do in the morning when I'm fresher.

Comments

[schnee.livejournal.com]

MySQL and Unicode, still a complete clusterf*ck. x.x

[porsupah.livejournal.com]

Ah, those long (thankfully) lost days of 8-bit encoding, where Windows used one character set, Mac OS used another, and.. well, pretty much everyone went their own way, eventually grudgingly half-accepting ISO 8859.

(I may occasionally feel a touch nostalgic about home computing's history, but no way I'm giving up this MBP with an SSD running around 1GB/s sustained throughput, or an iPad Air. Ah, to be able to show both to 15-y.o. me =:)