Fri Jun 3 12:20:17 IDT 2005
============================
As noted in the NEWS file, as of 3.1.5, gawk uses character values instead
of byte values for `index', `length', `substr' and `match'. This works
in multibyte and unicode locales.
Wed Jun 18 16:47:31 IDT 2003
============================
Multibyte locales can cause occasional weirdness, in particular with
ranges inside brackets: /[....]/. Something that works great for ASCII
will choke for, e.g., en_US.UTF-8. One such program is test/gsubtst5.awk.
By default, the test suite runs with LC_ALL=C and LANG=C. You
can change this by doing (from a Bourne-style shell):
$ GAWKLOCALE=some_locale make check
Then the test suite will set LC_ALL and LANG to the given locale.
As of this writing, this works for en_US.UTF-8, and all tests
pass except gsubtst5.
For the normal case of RS = "\n", the locale is largely irrelevant.
For other single byte record separators, using LC_ALL=C will give you
much better performance when reading records. Otherwise, gawk has to
make several function calls, *per input character* to find the record
terminator. You have been warned.