* Dave Rolsky ([EMAIL PROTECTED]) [10 Aug 2003 00:38]:
> On Sat, 9 Aug 2003, Iain Truskett wrote:
> > This make sense? make's default $(CC) may have no relation to
> > the one Perl was compiled with, so shouldn't we use what
> > knowledge we can get?
> >
> >
> > Index: Makefile.PL
> > ===================================================================
> > RCS file: /cvsroot/perl-date-time/modules/DateTime.pm/Makefile.PL,v
> > retrieving revision 1.31
> > diff -u -r1.31 Makefile.PL
> > --- Makefile.PL 24 Jul 2003 03:54:50 -0000 1.31
> > +++ Makefile.PL 9 Aug 2003 13:13:55 -0000
> > @@ -31,7 +31,7 @@
> > close F or no_cc();
> >
> > my $make = $^O =~ /win32/i ? 'nmake' : 'make';
> > - system("$make test$Config{obj_ext}") and no_cc();
> > + system("CC=$Config{cc} $make test$Config{obj_ext}") and no_cc();
> >
> > if ( -d 'CVS' )
> > {
> Will this match what MakeMaker generates? If so, good, if not, the patch
> won't achieve much.
Good question. Not sure. At the very least it means it gets it correct
for one Solaris box I have access to.
It has a perl that was compiled with gcc, but the system default is cc.
They are incompatible compilers and with the patch Makefile.PL correctly
finds out that a suitable compiler isn't available. cc does exist, but
is wrong.
If it's not what MakeMaker generates, it's at least *closer* to
what MM does.
[I've added "research for compiler checking" to my todo list, but
there's no guarantee I'll get near it]
cheers,
--
Iain.
