Subject: Re: CVS commit: pkgsrc
To: David Brownlee <abs@netbsd.org>
From: Johnny C. Lam <lamj@stat.cmu.edu>
List: pkgsrc-changes
Date: 05/17/2001 13:48:49
David Brownlee <abs@netbsd.org> writes:
> > >
> > > Actually, I think it should actually look like:
> > >
> > > 	CONFIGURE_ENV+=		MAKE="${MAKE_PROGRAM}"
> > >
> > > For some packages that _require_ gmake, the configure script fails to
> > > detect it because it just invokes "make" and doesn't find GNU make,
> > > much like you're describing above.  I believe this solves your problem
> > > as well, since ${MAKE_PROGRAM} is set to bmake on Linux, right?
> >
> > I mean for the above line to be added to bsd.pkg.mk, not just to the
> > apache Makefile.
> 
> 	If its in bsd.pkg.mk then presumably it could be conditional
> 	on USE_GMAKE?

No, we should need this all the time.  If USE_GMAKE is not defined,
then MAKE_PROGRAM is correctly set to "make" or
"/usr/local/bsd/bin/bmake" on zoularis platforms.  If it _is_ defined,
then MAKE_PROGRAM is set to "gmake".  In either case, GNU configure
scripts should know which make program will be invoked during the
configure process, and unconditionally passing MAKE="${MAKE_PROGRAM}"
accomplishes this.  There are also some stupid packages out there that
directly substitute @MAKE@ in their Makefile.in files, so as a side
effect, we would require fewer patches to fix this kind of stupidity.

> 	btw - would you be OK with addin the following to the apache
> 	Makefile?
> 
> .if defined(APACHE_HARD_SERVER_LIMIT)
> CFLAGS+=                -DHARD_SERVER_LIMIT=${APACHE_HARD_SERVER_LIMIT}
> .endif

I'll answer this in a private email.

     -- Johnny C. Lam <lamj@stat.cmu.edu>
        Department of Statistics, Carnegie Mellon University
        http://www.stat.cmu.edu/~lamj/