On Jul 2, 2008, at 22:14, Tom Lane wrote:
> The "leak" is irrelevant for larger/smaller. The only place where
> it's
> actually useful to do PG_FREE_IF_COPY is in a btree or hash index
> support function. In other cases you can assume that you're being
> called in a memory context that's too short-lived for it to matter.
Stupid question: for the btree index support function, is that *only*
the function referenced in the OPERATOR CLASS, or does it also apply
to functions that implement the operators in that class? IOW, do I
need to worry about memory leaks in citext_eq, citext_ne, citext_gt,
etc., or only in citext_cmp()?
Thanks,
David