source of powerpc-rtems failure

source of powerpc-rtems failure

[Joel Sherrill]

I have narrowed this down about as far as I can without getting some
advice. :) It is a bug in the bfd/configure.in script and easy to
reproduce in binutils 2.9.

Here's how:

mkdir build
../binutils-2.9/bfd/configure --target=powerpc-rtems

Edit the generated Makefile and search for "tdefaults".  It will look like
this:

tdefaults =  -DDEFAULT_VECTOR=bfd_elf32_powerpc_vec
-DSELECT_VECS='&bfd_elf32_powerpc_vec,&rs6000coff_vec,&bfd_elf32_powerpcle_vec,&bfd_powerpcle_pei_vec,&bfd_powerpc_pei_vec,&bfd_powerpcle_pe_vec,&bfd_powerpc_pe_vec,&ppcboot_vec'
-DSELECT_ARCHITECTURES='&bfd_rs6000_arch,&bfd_powerpc_arch'
-DHAVE_bfd_elf32_powerpc_vec -DHAVE_rs6000coff_vec
-DHAVE_bfd_elf32_powerpcle_vec -DHAVE_bfd_powerpcle_pei_vec
-DHAVE_bfd_powerpc_pei_vec -DHAVE_bfd_powerpcle_pe_vec
-DHAVE_bfd_powerpc_pe_vec -DHAVE_ppcboot_vec

Then clean the build directory and try this one:

../binutils-2.9/bfd/configure --target=powerpc-rtems --enable-targets=all

And the definition of tdefaults is changed to:

tdefaults =  -DDEFAULT_VECTOR=bfd_elf32_powerpc_vec

I am going to try to fix this here but thought this might be enough to
trip someone into know what the problem was.

--joel

Re: source of powerpc-rtems failure

[Ian Lance Taylor]

   Date: Thu, 16 Apr 1998 12:56:10 -0500 (CDT)
   From: Joel Sherrill <joel@OARcorp.com>

   Then clean the build directory and try this one:

   ../binutils-2.9/bfd/configure --target=powerpc-rtems --enable-targets=all

   And the definition of tdefaults is changed to:

   tdefaults =  -DDEFAULT_VECTOR=bfd_elf32_powerpc_vec

That by itself is not a bug.  If you configure with
--enable-targets=all, SELECT_VECS is not supposed to be defined.
SELECT_VECS is only defined if you have a list of selected vectors.

The bug would seem to be that bfd_elf32_powerpcle_vec is not in the
default list of targets in targets.c.

Ian

Index: targets.c
===================================================================
RCS file: /cvs/cvsfiles/devo/bfd/targets.c,v
retrieving revision 1.250
diff -u -r1.250 targets.c
--- targets.c	1998/02/12 23:03:24	1.250
+++ targets.c	1998/04/16 18:46:09
@@ -676,6 +676,7 @@
 	&bfd_elf32_m88k_vec,
 	&bfd_elf32_sparc_vec,
 	&bfd_elf32_powerpc_vec,
+	&bfd_elf32_powerpcle_vec,
 	&bfd_elf32_v850_vec,
 #ifdef BFD64			/* No one seems to use this.  */
 	&bfd_elf64_big_generic_vec,

Re: source of powerpc-rtems failure

[Joel Sherrill]

On Thu, 16 Apr 1998, Ian Lance Taylor wrote:

>    tdefaults =  -DDEFAULT_VECTOR=bfd_elf32_powerpc_vec
> 
> That by itself is not a bug.  If you configure with
> --enable-targets=all, SELECT_VECS is not supposed to be defined.
> SELECT_VECS is only defined if you have a list of selected vectors.

Thanks.  I was trying to track it down but did not understand the magic
going on behind the screen. :)

> The bug would seem to be that bfd_elf32_powerpcle_vec is not in the
> default list of targets in targets.c.

This lets it get further and now it fails in the same way the mips64orion
did.  It builds at least one multilib version of tmplibgcc2.a and then
complains ...:

build-powerpc-tools/gcc/ -O2  -DCROSS_COMPILE -DIN_GCC    -O2 -g
-I./include   -g1 -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -fexceptions
-Dinhibit_libc   -mlittle -mrelocatable-lib -mno-eabi -mstrict-align -I.
-I../../src/gcc -I../../src/gcc/config -c -DL${name} \
      ../../src/gcc/libgcc2.c -o ${name}.o; \
  if [ $? -eq 0 ] ; then true; else exit 1; fi; \
  /usr1/rtems/work/tools/build-powerpc-tools/binutils/ar rc tmplibgcc2.a
${name}.o; \
  rm -f ${name}.o; \
done
_muldi3
_divdi3
/usr1/rtems/work/tools/build-powerpc-tools/binutils/ar: tmplibgcc2.a:
Invalid operation

It does this repeatedly until the rule stmp-multilib-sub fails.  Is this a
related problem but in ar?

I will start to see what I can figure out about this one. :)

--joel