The current ODBC driver has a bug
caused by the loid being cast and
stored in a pointer.
Later on there are two places where
this pointer is dereferenced to
free some memory but there is no
check to see if the current type
is LONG_VAR_BINARY.
I had fixed this in my local copy
of the source but have since had
my laptop's hard disk die. I'll
try to reconstruct the fix in a
few days. (other more important
things to reconstruct in the
meantime) Perhaps someone else
can fix it in the meantime from
the meager description above.
Sam
P.S.
Is anyone else using PG + ODBC from C?