Thread: Testing for int64 (was Re: [COMMITTERS] pgsql-server/ /configure /configure.in...)

petere@postgresql.org (Peter Eisentraut - PostgreSQL) writes:
>     The code that checks for the 64-bit int type now gives more reasonable
>     results when cross-compiling: In that case we just take the compiler's
>     information and trust that the arithmetic works.  Disabling int64 is too
>     pessimistic.

It's not so much that we can't trust the arithmetic as that we shouldn't
trust that the platform's s(n)printf supports int64.  This situation
used to be a reality on older machines with gcc but no int64 type in the
native compiler, and I suspect there are still some of them out there.

I think a reasonable choice in cross-compiling situations would be to
assume int64 works if we have a long long int datatype, but to force use
of our own snprintf rather than trusting to luck with the platform's.

(It didn't look like that's what happens right now, but I might be
missing something in the autoconf spaghetti.)

            regards, tom lane

Tom Lane writes:

> I think a reasonable choice in cross-compiling situations would be to
> assume int64 works if we have a long long int datatype, but to force use
> of our own snprintf rather than trusting to luck with the platform's.

That's approximately what's happening.  Formerly it insisted on doing a
run check to detect the int64 type.  Now it does a compile check when
cross-compiling.

For the snprintf format detection we obviously don't have that chance.  I
just refactored the code a little and added a cache variable so the
advanced cross-compiling user can override the check with known values.

--
Peter Eisentraut   peter_e@gmx.net