On Wed, 2006-01-11 at 02:58 -0500, Tom Lane wrote:
> One comment is that there are worse things than small memory leaks in
> seldom-followed code paths, especially if those paths are only taken in
> error cases.
While I agree the problem isn't a showstopper, I think it is still
worthy of concern: the mbutils example was chosen for being clearly
broken, not as being the most serious instance of the problem. The issue
might occur in *any* situation in which we're allocating memory in a
memory context whose lifetime exceeds the current transaction -- I
haven't looked to see what other places might need fixing.
-Neil