Re: pgwin32_safestat weirdness

Tom Lane wrote:
> Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
> > Cygwin is now building happily, but MinGW is now broken on libpq.
> > It looks like libpq now needs dirmod.o or maybe libpgport.a. What I
> > really don't understand though is why MinGW is broken but MSVC
> > isn't.
>
> I don't think we should import dirmod.o into libpq; it's too big.

Is it really big enough to matter? Where would you in general "draw the
line" for including?

> I suggest either
>
> (1) Assume that we don't need "safe" stat for frontend code, and
> compile the safestat stuff only when !defined(FRONTEND)
>
> (2) Split safestat into its own file and include that in libpq.

Is there not a (3) which has it included in all frontend code *except*
libpq? Do we have a define to do that off?

Because I agree with your comments in the other mail that there may be
other frontend stuff that might need it.

In libpq, it's only used in one place to check if a file is present,
and one then in the SSL code to determine permissions and such (which
means it's being ignored on win32).

//Magnus