Cross compilation vs. libiberty

Cross compilation vs. libiberty

[Mark Mitchell]

Folks --

  I have run into the problem where libiberty, only configured once,
does not build for the target since the target configuration does not
match the host.  In particular, I have stdlib.h on the host, but not
on the target, but files in libiberty want to include stdlib.h because
the host configury indicates that is the right thing to do.

  Jeff, I remember you were against rerunning configure.  Please
remind me why, or point me at a concise email summarizing your
thinking, if it's available.

  To me, this seems like the right thing to do.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Re: Cross compilation vs. libiberty

[Geoff Keating]

Mark Mitchell <mark@codesourcery.com> writes:

> Folks --
> 
>   I have run into the problem where libiberty, only configured once,
> does not build for the target since the target configuration does not
> match the host.  In particular, I have stdlib.h on the host, but not
> on the target, but files in libiberty want to include stdlib.h because
> the host configury indicates that is the right thing to do.

I don't understand.  libiberty is configured once for the host, and
once for each multilib combination on the target:

Configuring in powerpc-eabisim/libiberty
creating cache ../config.cache
checking host system type... powerpc-unknown-eabisim
checking build system type... sparc-sun-solaris2.6
checking for powerpc-eabisim-ar... powerpc-eabisim-ar
checking for powerpc-eabisim-ranlib... powerpc-eabisim-ranlib
checking for gcc... /sloth/delay/tbox/build-gcc/gcc/xgcc -B/sloth/delay/tbox/build-gcc/gcc/ -B/sloth/delay/tbox/build-gcc/powerpc-eabisim/newlib/ -isystem /sloth/delay/tbox/build-gcc/powerpc-eabisim/newlib/targ-include -isystem /sloth/delay/tbox/cvs-gcc/egcs/newlib/libc/include -B/sloth/delay/tbox/objs/powerpc-eabisim/bin/ -B/sloth/delay/tbox/objs/powerpc-eabisim/lib/ -isystem /sloth/delay/tbox/objs/powerpc-eabisim/include
checking whether we are using GNU C... yes
...
checking for limits.h... yes
checking for stdlib.h... yes
checking for string.h... yes
checking for unistd.h... yes
checking for strings.h... no

[this is significant because solaris does have /usr/include/strings.h.]

The target configuration happens after the compiler is built, of course, 
otherwise it couldn't work.

I think you have some other problem with your build.
-- 
- Geoffrey Keating <geoffk@cygnus.com>

Re: Cross compilation vs. libiberty

[Mark Mitchell]

>>>>> "Geoff" == Geoff Keating <geoffk@cygnus.com> writes:

But pieces of libiberty are also used in libgcc2, and there it is not
configured.  Hmm.

I see that is the direct result of Alex's demangler changes.
Ugh.  I'll talk to Alex.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Re: Cross compilation vs. libiberty

[Alexandre Oliva]

On Aug  2, 2000, Mark Mitchell <mark@codesourcery.com> wrote:

>   I have run into the problem where libiberty, only configured once,
> does not build for the target since the target configuration does not
> match the host.

libiberty is configured and built for the target with CC_FOR_TARGET,
after GCC is built.  Isn't a libiberty directory created within the
target sub-directory of the build tree?

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                  aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist    *Please* write to mailing lists, not to me

Re: Cross compilation vs. libiberty

[Jeffrey A Law]

  In message < 20000802182521R.mitchell@codesourcery.com >you write:
  > 
  > Folks --
  > 
  >   I have run into the problem where libiberty, only configured once,
  > does not build for the target since the target configuration does not
  > match the host.  In particular, I have stdlib.h on the host, but not
  > on the target, but files in libiberty want to include stdlib.h because
  > the host configury indicates that is the right thing to do.
Sounds like a configury problem of some kind.


  >   Jeff, I remember you were against rerunning configure.  Please
  > remind me why, or point me at a concise email summarizing your
  > thinking, if it's available.
I was against rerunning it for the different stages of building gcc itself
because, in theory, there shouldn't be any differences between the
result between stages.

jeff