Hello again,
pgcrypto 0.4 had a nasty obscure bug, when after including
postgres.h the BYTE_ORDER is not defined, it compiles code for
both big-endian and little-endian machines. This makes MD5 and
SHA1 algorithms in internal library produce garbage, ofcourse
now HMAC and crypt-md5 produce junk too. Also Rijndael on
big-endian machine may be broken. If you used any external
library, you are fine.
Affected platforms: FreeBSD *, probably more.
Not affected: Linux/glibc 2.2, NetBSD 1.5.
http://www.l-t.ee/marko/pgsql/pgcrypto-0.4.2.tar.gz
Morale: _do_ run regression tests before real use...
--
marko