When dealing with UTF-8, you usually end up having to exhaustively track and verify the format of your text at each layer. With Perl, you will also often have to worry about whether strings are considered "utf8" or not by Perl.

When things are working properly, you shouldn't have to worry about Perl's "utf8" flag for strings. And the p5p community (the authors of Perl) tends to cater toward the "things working properly" case and so just being able to tell what format the string is really in and whether the "utf8" flag is on keeps getting trickier.

One of the easiest tools to aid such investigations, in my experience, is Devel::Peek. It will nicely dump the byte format of a string, whether the "utf8" flag is on, and, if it is on, the UTF-8 char format of the string.

So you might want to update your code to something like:

#!/usr/bin/perl -w
use strict;
use warnings;
use Encode;
use DBI;
use feature 'unicode_strings';
use Devel::Peek     qw< Dump >;
binmode(STDOUT, ":utf8");
my $db      = "...";
my $user    = "...";
my $pw      = '...';
my $host    = "...";
my $options = {RaiseError => 1, AutoCommit => 1, mysql_enable_utf8 => 
+1};

my $dbh = DBI->connect("DBI:mysql:$db:$host", $user, "$pw", $options) 
+|| die "Could not connect to database!";
my $sql = 'select post_id, post_text from phpbb_posts where post_id = 
+24358';
my $sth = $dbh->prepare($sql);
$sth->execute() or die "Error: $DBI::errstr";
my ($id, $text) = $sth->fetchrow_array;

print "id: $id text: $text";
Dump( $text );
$text =~ s/&#9500;â&#9516;ñ/ä/g;
$text =~ s/&#9500;â&#9516;Â/ö/g;
$text =~ s/&#9500;â&#9532;©/ß/g;
$text =~ s/&#9500;â&#9516;&#9565;/ü/g;

print "\nid: $id text: $text";
Dump( $text );
[download]

You should also do some tests where you put a specific UTF-8 string into your DB and then pull it back out, using Dump() on both what you sent in and what you got back.

You may need to fiddle with more MySQL-specific options or even update DBD::mysql, or, if you are using something very old, upgrade Perl itself. You might even find that DBD::mysql misuses Perl's "utf8" flag for strings so you may end up testing taking a proper UTF-8 string but telling Perl that it is a string of bytes ("utf8" off) before sending it off to MySQL. Or find that MySQL is giving back to you such a string.

A while ago I changed my webhoster and setup a new MySQL db there which uses UTF8. I exported the old db (which didn't use UTF8) and imported it to the new db.

So the DB import was messed up. You should repair and repeat the DB import. Another option: Export the current messed-up DB to a text file, repair the encoding in the text file, and re-import it.

Alexander

With a while ago I meant "six months ago". Since it's a forum DB there have been many changes and new posts. So redoing the import is no option.
Exporting the text and repair the encoding is what I am trying to do right now. The text is now UTF8, so simply changing the encoding does not work.

Exporting the text and repair the encoding is what I am trying to do right now.

Export using current database tools, not perl. This makes sure you won't get encoding problems at that level, plus often, a bulk import / export is faster than working on table rows.

The text is now UTF8, so simply changing the encoding does not work.

There is no text, there are only database columns, and they contain a mix of properly UTF-8 encoded characters and mojibake.

Exporting, fixing mojibake, and importing takes DBI, DBD::mysql, and the MySQL client libraries used by DBD::mysql out of the picture. Only a UTF-8 encoded text file is left. Much easier to handle. The export should show proper encoded characters in a UTF-8 capable editor for fresh postings. After fixing it (maybe using perl or some other tool), the entire export should show proper encoded characters in the editor. Importing it back to the DB will fix the problem.

Use a second DB (and a second forum installation) to develop and test the process. Once the process works, shut down the forum, export, fix, import, restart the forum.

(This is what you should have done six months ago.)

Alexander

--
Today I will gladly share my knowledge and experience, for there are no sweeter words than "I told you so". ;-)