Re: Further simplification of c.h's #include section

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> On Wed, Nov 15, 2017 at 5:22 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Thoughts?

> Sure, having a win32_port.h as a sub-include of port.h seems fine.

Here's a draft patch for this. I'm not too certain about the interactions
with Cygwin; some of the stuff I moved out of port.h might have to go back
there so that a Cygwin build will see it. There might also be some
declaration ordering dependencies that I failed to spot.

(Speaking of which, I'm wondering why the existing code monkeys around
with _WIN32_WINNT after it's already included a bunch of system headers.
Shouldn't that be set earlier --- in other words, shouldn't that code
move back to win32.h from win32_port.h? But I've not touched that here.
I did remove an at-best-redundant definition from pg_ctl.c though.)

Anybody want to test this manually, or shall we just throw it into the
buildfarm and see what blows up?

regards, tom lane