I turned on the multibyte option by default and the world failed to end. Two localised problems were revealed by the tests. One, in fact, I'm no longer testing for: this was for invalid characters (actually ISO-8859-1 but any single byte character set would have done) with multibyte mode turned on. It ought to be done properly within D07multibyte.ztst, but I haven't at the moment got the energy for generating deliberately invalid UTF-8 sequences. The intention is such sequences should be compared as raw bytes (I found yet another STOUC() was necessary). There's also a test for things like [[:upper:]] with Unicode characters. I found one failed on the Fedora Core 3 system here. As it's accented Greek it could put a strain on many ever so slightly flakey implementations. I've sent this as uunencoded because one file includes the old ISO-8859-1 sequence and another UTF-8, which mailers probably don't like. To access the latest news from CSR copy this link into a web browser: http://www.csr.com/email_sig.php
Attachment:
multibyte_on_by_default.dif
Description: Binary data