Whenever cairo expects utf8-encoded strings, automatically "upgrade" the
user-supplied string to utf8.  This makes it unnecessary (but harmless)
for users to call utf8::upgrade() on their strings.

Note, however, that calling utf8::encode() will now yield doubly-encoded
and thus broken strings.  Also, having utf8-encoded literals in your
code without an accompanying "use utf8" line will also lead to double
encoding.

https://rt.cpan.org/Ticket/Display.html?id=73177