Horton

Re: Cannot find a working 64-bit integer type on Illumos

Thomas Munro <thomas.munro@gmail.com>

View thread View on postgr.es
From: Thomas Munro <thomas.munro@gmail.com>
To: Peter Eisentraut <peter@eisentraut.org>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Japin Li <japinli@hotmail.com>, Andres Freund <andres@anarazel.de>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2024-04-18T21:00:05Z
Lists: pgsql-hackers 
On Thu, Apr 18, 2024 at 8:47 PM Peter Eisentraut <peter@eisentraut.org> wrote:
> Maybe this means something like our int64 is long long int but the
> system's int64_t is long int underneath, but I don't see how that would
> matter for the limit macros.

Agreed, so I don't think it's long vs long long (when they have the same width).

I wonder if this comment is a clue:

static char *
inet_net_ntop_ipv6(const u_char *src, int bits, char *dst, size_t size)
{
    /*
     * Note that int32_t and int16_t need only be "at least" large enough to
     * contain a value of the specified size.  On some systems, like Crays,
     * there is no such thing as an integer variable with 16 bits. Keep this
     * in mind if you think this function should have been coded to use
     * pointer overlays.  All the world's not a VAX.
     */

I'd seen that claim before somewhere else but I can't recall where.
So there were systems using those names in an ad hoc unspecified way
before C99 nailed this stuff down?  In modern C, int32_t is definitely
an exact width type (but there are other standardised variants like
int_fast32_t to allow for Cray-like systems that would prefer to use a
wider type, ie "at least", 32 bits wide, so I guess that's what
happened to that idea?).

Or perhaps it's referring to worries about the width of char, short,
int or the assumption of two's-complement.  I think if any of that
stuff weren't as assumed we'd have many problems in many places, so
I'm not seeing a problem.  (FTR C23 finally nailed down
two's-complement as a requirement, and although C might not say so,
POSIX says that char is a byte, and our assumption that int = int32_t
is pretty deeply baked into PostgreSQL, so it's almost impossible to
imagine that short has a size other than 16 bits; but these are all
assumptions made by the OLD coding, not by the patch I posted).  In
short, I guess that isn't what was meant.

Commits

  1. Move pg_int64 back to postgres_ext.h

  2. pgbench: Make set_random_seed() 64-bit everywhere.

  3. Use PRI?64 instead of "ll?" in format strings (continued).

  4. Fix order of -I switches for building pg_regress.o.

  5. libpq: Deprecate pg_int64.

  6. Use PRI*64 instead of "ll*" in format strings (minimal trial)

  7. Fix header inclusion order in c.h.

  8. Use <stdint.h> and <inttypes.h> for c.h integers.

  9. Remove traces of BeOS.

  10. More correct way to check for existence of types, which allows to specify