On Fri, May 27, 2022 at 10:43:17AM -0700, Nathan Bossart wrote:
> Makes sense. Here's a new patch set. 0001 is the part intended for
> back-patching, and 0002 is the rest (i.e., adding pg_attribute_nonnull()).
> I switched to using __has_attribute to discover whether nonnull was
Okay, I have looked at 0001 this morning and applied it down to 12.
The change in GetConfigOptionByNum() is not required in 10 and 11, as
the strings of pg_show\all_settings() have begun to be translated in
12~.
> supported, as that seemed cleaner. I didn't see any need for a new
> configure check, but maybe I am missing something.
And I've learnt today that we enforce a definition of __has_attribute
at the top of c.h, and that we already rely on that. So I agree that
what you are doing in 0002 should be enough. Should we wait until 16~
opens for business though? I don't see a strong argument to push
forward with that now that we are in beta mode on HEAD.
--
Michael