Re: Reduce the number of special cases to build contrib modules on windows

Thank you for having another look at this.

On Tue, 12 Jan 2021 at 20:18, Michael Paquier <michael@paquier.xyz> wrote:
>
> On Wed, Dec 30, 2020 at 10:07:29AM +1300, David Rowley wrote:
> > -#ifdef LOWER_NODE
> > +/*
> > + * Below we ignore the fact that LOWER_NODE is defined when compiling with
> > + * MSVC.  The reason for this is that earlier versions of the MSVC build
> > + * scripts failed to define LOWER_NODE.  More recent version of the MSVC
> > + * build scripts parse makefiles which results in LOWER_NODE now being
> > + * defined.  We check for _MSC_VER here so as not to break pg_upgrade when
> > + * upgrading from versions MSVC versions where LOWER_NODE was not defined.
> > + */
> > +#if defined(LOWER_NODE) && !defined(_MSC_VER)
> >  #include <ctype.h>
> >  #define TOLOWER(x)   tolower((unsigned char) (x))
> >  #else
>
> While on it, do you think that it would be more readable if we remove
> completely LOWER_NODE and use only a check based on _MSC_VER for those
> two files in ltree?  This could also be handled as a separate change.

I'm hesitant to touch that.  If anyone is running an instance compiled
with a non-default LOWER_NODE then we might give them some trouble if
they pg_upgrade their database later.

> > +     foreach my $line (split /\n/, $mf)
> > +     {
> > +             if ($line =~ /^[A-Za-z0-9_]*\.o:\s(.*)/)
> > +             {
> > +                     foreach my $file (split /\s+/, $1)
> > +                     {
> > +                             foreach my $proj (@projects)
> > +                             {
> > +                                     $proj->AddFileConditional("$subdir/$n/$file");
> > +                             }
> > +                     }
> > +             }
> > +     }
> Looking closer at this change, I don't think that this is completely
> correct and that could become a trap.  This is adding quite a bit of
> complexity to take care of contrib_extrasource getting empty, and it
> actually overlaps with the handling of OBJS done in AddDir(), no?

hmm. I'm not quite sure if I know what you mean by "trap" here.

contrib/cube/Makefile has an example of what this is trying to catch:

# cubescan is compiled as part of cubeparse
cubeparse.o: cubescan.c

I don't really see what other options there are apart from just not
get rid of $contrib_extrasource.

Can you give an example of what sort of scenario you've got in mind
where it'll cause issues?

I've attached a rebased patch.

David