Peter Eisentraut <peter@eisentraut.org> writes:
> I do suggest some kind of comment, that we're using size_t as a
> convenient proxy for the most suitable chunk/step size for the platform,
> not to actually measure a size.
Yeah. There's not actually anything wrong with using "long" here,
except that we believe that on Win64 it's probably not the processor's
native word width. Size/size_t is more likely to match that.
([u]intptr_t is even more likely to match, but I don't think that's a
good choice because it invites confusion with the usage of uintptr_t
for address arithmetic elsewhere in the macro. That's quite unrelated
to the choice of copy step size.)
regards, tom lane