Are you perhaps doing any of the following?

• Using the :utf8 layer for input (open, binmode)
• Using _utf8_on anywhere
• Forgetting to decode input
• Forgetting to encode output

Thanks for this, AFAIK we are using utf8 functions correctly. However, it is very likely that somewhere in 200K lines of perl a simple string concat is happening between a UTF8 string and a non-UTF8 string. My idea is to try to trap this so we can find it. The only way I can think of doing is to overload the concat operator. I'm also keen to know if it is possible. We tried this but it does not work...

use strict;

package UNIVERSAL;

use overload "." => \&concat;


sub concat {
    my ($a, $b) = @_;
    return join "XXX", $a, $b;

}


package Foo;

my $thing = "cat" . "dog";

print "\n\n\n$thing\n\n\n";
[download]

This produces catdog

Have a look at encoding::warnings :)

Could you please provide a piece of code that shows your described behaviour (the "trouble"), along with your perl version?

Update: And what is ut8dom?

-- Frank