NikhilS <nikkhils(at)gmail(dot)com> writes:
> The attached patch should fix this.
And break other things, no doubt. needed = 0 is a perfectly valid
edge case and mustn't be rejected here. (In fact, I doubt you'd
even get through the regression tests with this patch ... how much
did you test it?)
The real problem with what you describe is that you should have used
makeStringInfo().
> I also found the absence of a function like resetStringInfo() a bit
> puzzling.
CVS HEAD is way ahead of you.
regards, tom lane