Re: AC_FC_DUMMY_MAIN broken with gcc4/gfortran

[Bill Northcott]

On 24/08/2006, at 11:03 PM, Peter O'Gorman wrote:
> > Following my last post I have done some digging on AC_FC_DUMMY_MAIN.
> >
> >
> > Using gcc4 and gfortran FLIBS include '-lgfortranbegin - 
> > lgfortran'.  libgfortranbegin defines _main and needs _MAIN__
> >
> > Since the test program also defines _main it will never link  
> > because of the multiply defined symbol!  So the test for a dummy  
> > main will never succeed.
>
> This seems odd, my libgfortranbegin is a static archive, is yours?  
> Are you passing odd flags like -all_load which would force the main  
> defined in libgfortranbegin to be loaded?
>
> What happens for me is that the main defined in the object is used  
> and the one in libgfortranbegin is ignored because a main has  
> already been found. At least that is what used to happen for me, I  
> haven't looked at it in quite a while...

Funny that.  I am definitely not using -all_load and libgfortranbegin  
is static, or more correctly the libgfortranbegins are static.   The  
compiler is the one distributed with the R 2.3.1 MacOS binary  
package.  So it has fat (ppc i386) compiler executables and three  
possible targets (ppc ppc64 and i386).  It is built from FSF not  
Apple sources.  You could get the compiler and try it yourself.

I will hack about with it some more.

Bill