This looks like consequence of r10458, is still present in r10568, and is easy to reproduce:
[EMAIL PROTECTED]> cat load-test.pir .sub _main :main .local string file_name file_name = iso-8859-1:"foo.pbc" load_bytecode file_name .end [EMAIL PROTECTED]> ./parrot load-test.pir Cross-charset index not supported [EMAIL PROTECTED]> It's easy enough for me to work around, because I don't need ISO-8859-1 file names, but shouldn't load_bytecode be a bit more liberal? I could probably hack something together along those lines, but it might be better have a solution that takes account of what charset(s) the OS would insist upon for file names, true? The other mystery is that I still don't know why I'm generating ISO-8859-1 strings in the first place. Could it be picking this up from the source file of the code that builds the string? If so, how? And why -- the code looks like plain ASCII to me (and to emacs). TIA, -- Bob Rogers http://rgrjr.dyndns.org/