fix for libf2c installation

Jeffrey A Law
Mon Jun 22 23:11:00 GMT 1998

  In message < >you write:
  > >>>>>> "Manfred" == Manfred Hollstein <> writes:
  > >
  > > Manfred> It was the patch below. It's not fixing anything in Fortran
  > > Manfred> specific files, but in the toplevel files, which now pass
  > > Manfred> down the particular vars. Perhaps this is why you didn't
  > > Manfred> recognize it.
  > >
  > >Yes.  I think autoconf facilities are the right thing, but I'm not in
  > >charge.  Unlike the gcc directory, you can't now do `make install' in
  > >the libf2c one.
  > Dave raises a good point here.  Though libf2c's can't
  > elegantly intuit the appropriate information, perhaps it should
  > AC_SUBST it, which means the top-level configure would have to pass
  > it down there instead of later via the Makefile process.
Sounds reasonable to me.

  > That way, once configured, builds and installs would work from the
  > lower-level directory.  Personally, I'd love this.

  > Manfred, does this make sense?  Could you consider doing it if you
  > understand the relevant bits of top-level configure?  Feel free
  > to fix up the gcc, g77, and libf2c configure/Makefile stuff accordingly
  > (e.g. the gcc configure would have to substitute the $(libsubdir)
  > macro as well in the gcc Makefile it builds, right?).
Agreed.  Fixing it once and for all would be good :-)  The only thing
to be careful about would be setting "version" in a subdir that
has its own notion of version #s (libstdc++, or any non-egcs package
like binutils that someone might put in the egcs tree to set up a
single tree build environment).


More information about the Gcc-patches mailing list