A workaround for Perl was originally added in bug 518522. It broke in bug 792537 and we've since cleaned it up, but that only stops it being added for *new* installs. Please figure out a way to clean it up from the user's postgresql.conf or warn on the bad line being there given the ebuild was responsible for adding it.
Though I'm not familiar with PL/Perl, I can say that "use utf8" is only useful in cases where the source code is to be treated as UTF-8. That is, analogous to Encode::decode('UTF-8', $source_code). $ perl -e '$str = "😀"; printf "%v.X (length = %d)\n", $str, length $str' F0.9F.98.80 (length = 4) $ perl -Mutf8 -e '$str = "😀"; printf "%v.X (length = %d)\n", $str, length $str' 1F600 (length = 1) The second case shows the string containing a single codepoint (whatever encoding Perl uses to store it internally is immaterial). --- That aside, the PostgreSQL docs make the following claim. "Arguments will be converted from the database's encoding to UTF-8 for use inside PL/Perl, and then converted from UTF-8 back to the database encoding upon return." This is ambiguous. Do they mean that strings are converted to UTF-8 bytes (as in the first example above), or that they are properly stored as 'wide' characters (as in the second example above)? I don't know but it should be wholly unrelated to the use of the utf8 pragma. Suffice to say that letting the user deal with all of this is almost certainly a wise decision.
