They are mainly accented characters like à, é or ö. They are read from a utf-8 encoded file using "<:encoding(UTF-8)", which converts them (if I understand it right) to "internal Perl format", presumably also UTF-8. However if I don't perform the encode("utf8",$_) first, they arrive in my C function as Latin-1, not UTF-8. The result is also output to a file in UTF-8 using >:encoding(UTF-8), but if I don't perform the decode("utf8",..) on the string leaving my function, I get "double UTF8" encoded strings! So either:

• Perl's internal string coding is Latin-1, not UTF-8 (at least for the characters I'm currently dealing with), or
• Perl's internal string coding is UTF-8, but is converted to Latin-1 when passed to my C-function, and converted back from Latin-1 when reading the result back.

Or is there another explanation? Anyway, regardless what characters I'm dealing with, is there a way to bypass all this recoding in an elegant and reliable way? Or is the Perl interpreter smart enough to do all the bypassing by itself, so that I incur no speed penalty?