Peter Eisentraut <peter.eisentraut@enterprisedb.com> writes:
> Per discussion in [0], here is a patch set that allows pfree() to accept
> a NULL argument, like free() does.
So the question is, is this actually a good thing to do?
If we were starting in a green field, I'd be fine with defining
pfree(NULL) as okay. But we're not, so there are a couple of big
objections:
* Code developed to this standard will be unsafe to back-patch
* The sheer number of places touched will create back-patching
hazards.
I'm not very convinced that the benefits of making pfree() more
like free() are worth those costs.
We could ameliorate the first objection if we wanted to back-patch
0002, I guess.
(FWIW, no objection to your 0001. 0004 and 0005 seem okay too;
they don't touch enough places to create much back-patching risk.)
regards, tom lane
