On Tue, Sep 20, 2022 at 9:18 PM Michael Paquier <michael@paquier.xyz> wrote:
>
> On Tue, Sep 20, 2022 at 08:01:20AM -0400, James Coleman wrote:
> > I don't have access to a Windows machine for testing, but re-reading
> > the documentation it looks like the issue is that our noreturn macro
> > is used after the definition while the MSVC equivalent is used before.
>
> A CI setup would do the job for example, see src/tools/ci/README that
> explains how to set up things.
That's a good reminder; I've been meaning to set that up but haven't
taken the time yet.
> > I've removed that for now (and commented about it); it's not as
> > valuable anyway since it's mostly an indicator for code analysis
> > (human and machine).
>
> Except for the fact that the patch missed to define
> pg_attribute_noreturn() in the MSVC branch, this looks fine to me. I
> have been looking at what you meant with packing, and I can see the
> business with PACK(), something actually doable with gcc.
>
> That's a first step, at least, and I see no reason not to do it, so
> applied.
Thanks!
James Coleman