Re: Convert a latin1 db imported in utf8...

[Donal Lynch]

For a start, taking the équipe fière row (or even just é will suffice), can you post the result of select charset(txt1), hex(txt1), and also the result of SHOW VARIABLES LIKE 'char%'. This will show exactly in what way the data is broken. With that information, there's probably something you can do with convert to fix the data.

You want to end up with the bytes matching the encoding of the character, so é stored as latin1 should be 0xe9, and é stored as utf8 should be 0xc3a9. If things get mixed up, and c3a9 gets stored in a latin1 column, it'll come out as é. If the same thing happens again, you end up with é, but you could be seeing that because your client is trying to read utf8-encoded bytes as latin1.