Recently I found that ERRNO ( "$!" or POSIX::strerror($!+0) ) is actually locale-dependant.

perllocale:

Note especially that the string value of $! and the error messages given by external utilities may be changed by LC_MESSAGES . If you want to have portable error codes, use %! . See Errno.

Example: the following commands print different binary strings. First prints "File not found" message translated to Russian in WINDOWS-1251, second in UTF-8:

LANG=ru_RU LANGUAGE=ru_RU:ru LC_ALL=ru_RU.CP1251 LC_MESSAGES=ru_RU.CP1
+251 perl  -e 'open my $f, "<", "notafile" or print $!'
[download]

LANG=ru_RU LANGUAGE=ru_RU:ru LC_ALL=ru_RU.utf8 perl  -e 'open my $f, "
+<", "notafile" or print $!'
[download]

I would like to convert binary strings to character strings (and print it on the screen in the end with other information):

my $errno = decode("SOMEENCODING", $!)
[download]

Problem is: what is the most convenient and reliable way to determine $! encoding?

The is 'use locale', but seems it's obsolete way to work with single byte-encoding and binary strings, and it's not a good idea to mix it with Unicode.

Note that I have already binmode STDOUT, ":encoding(UTF-8)" and option for user to override terminal encoding, and that works fine. However it won't work with $! for me, because I have a testsuite which tests my error handling code and testsuite can be ran on different locales without way to specify encoding. Obviously decode("UTF-8, $!) can break if encoding is wrong (not every byte sequence is correct character sequence in UTF-8 encoding).