Magnus Hagander <magnus@hagander.net> writes:
> On Tue, Aug 4, 2009 at 16:10, Peter Eisentraut<peter_e@gmx.net> wrote:
>> Well, there is nothing outright wrong with this patch, but without any
>> measurable effect, it is too early to commit it. �At least I would like to see
>> the Datum typedef to be changed to use intptr_t and the fallout from that
>> cleaned up.
> +1.
> I think it's good that it was posted for a quick review of the general
> idea, but I agree that it's too early to commit it until we can see
> some actual benefit. And I expect the Datum changes to be much larger
> than this, so we can just review/apply them as one when the time
> comes.
The other thing that I would say is a non-negotiable minimum requirement
is that the patch include the necessary configure pushups so it does not
break machines without uintptr_t. I think we could just do a
conditional
typedef unsigned long uintptr_t;
and proceed from there; then machines without the typedef are no worse
off than before.
regards, tom lane