Re: Support for 8-byte TOAST values (aka the TOAST infinite loop problem)

On Sun, Jan 4, 2026 at 2:23 PM Michael Paquier <michael@paquier.xyz> wrote:
>
>
> One thing that stands out for me is that I am going to need the oid8
> data type anyway, to be store the 8-byte values in the catalogs, with
> a C type to mark some of the TOAST apis with it across the board, so I
> would like to apply that to move this part out of the way (0001 only)
> and move the needle.  If there are any objections to that, please feel
> free.
>

Hi Michael,

Two quick notes from my review:

Semantics: the regression test accepts '-1040' and wraps modulo 2^64.
Is that intentional? If yes, please document it; if not, oid8in()
should reject negative input and the test should expect an error.

Nits: oid8.c appears to include <ctype.h> / <limits.h> without using
them, and there’s an empty src/test/regress/expected/oid8.sql
(intentional?).
--
Nikhil Veldanda