I just committed a new version of the driver that fixes a minor bug in
the infamous lob hack. It doesn't appear that the bug affected
gcc-compiled versions on x86 systems due to a bit of luck in the way
that the compiler allocates memory, so you may not need to update. An
array was allocated a byte too short, not allowing for space for the
terminating null character of the encoded string stuffed into the db
by the lob insertion routines.
I uncovered this as I was adding two new features I needed for a
personal project that can't wait for PG 7.1's built-in large types.
The two new commands, blob_get (returns a Tcl string containing the
lob value) and blob_put (stuffs a Tcl string directly into the lob)
give more flexibility in the use of lobs.
