[tree-ssa] Bootstrap failure on powerpc64-linux-gnu

[tree-ssa] Bootstrap failure on powerpc64-linux-gnu

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 1219 bytes --]


FYI, current CVS fails for me:

/usr/src/packages/BUILD/gcc-3.5-tree-ssa-20031213/obj-powerpc64-suse-linux/gcc/gfortran -B/usr/src/packages/BUILD/gcc-3.5-tree-ssa-20031213/obj-powerpc64-suse-linux/gcc/ -B/opt/gcc-tree-ssa/powerpc64-suse-linux/bin/ -B/opt/gcc-tree-ssa/powerpc64-suse-linux/lib/ -isystem /opt/gcc-tree-ssa/powerpc64-suse-linux/include -isystem /opt/gcc-tree-ssa/powerpc64-suse-linux/sys-include -Wall -fno-repack-arrays -fno-underscoring -c ../../../libgfortran/intrinsics/selected_kind.f90  -fPIC -DPIC -o .libs/selected_kind.o
../../../libgfortran/intrinsics/selected_kind.f90: In function `selected_int_kin
d':

../../../libgfortran/intrinsics/selected_kind.f90:46: internal compiler error: in rs6000_output_function_epilogue, at config/rs6000/rs6000.c:12532
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.suse.de/feedback> for instructions.
make[3]: *** [selected_kind.lo] Error 1

Unfortunatly gdb seems unable to debug this :-(

Andreas
-- 
 Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj
  SuSE Linux AG, Deutschherrnstr. 15-19, 90429 Nürnberg, Germany
   GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: [tree-ssa] Bootstrap failure on powerpc64-linux-gnu

[law]

In message <u84qw4mu6g.fsf@gromit.moeb>, Andreas Jaeger writes:
 >--=-=-=
 >Content-Type: text/plain; charset=iso-8859-1
 >Content-Transfer-Encoding: quoted-printable
 >
 >
 >FYI, current CVS fails for me:
 >
 >/usr/src/packages/BUILD/gcc-3.5-tree-ssa-20031213/obj-powerpc64-suse-linux/=
 >gcc/gfortran -B/usr/src/packages/BUILD/gcc-3.5-tree-ssa-20031213/obj-powerp=
 >c64-suse-linux/gcc/ -B/opt/gcc-tree-ssa/powerpc64-suse-linux/bin/ -B/opt/gc=
 >c-tree-ssa/powerpc64-suse-linux/lib/ -isystem /opt/gcc-tree-ssa/powerpc64-s=
 >use-linux/include -isystem /opt/gcc-tree-ssa/powerpc64-suse-linux/sys-inclu=
 >de -Wall -fno-repack-arrays -fno-underscoring -c ../../../libgfortran/intri=
 >nsics/selected_kind.f90  -fPIC -DPIC -o .libs/selected_kind.o
 >../../../libgfortran/intrinsics/selected_kind.f90: In function `selected_in=
 >t_kin
 >d':
 >
 >../../../libgfortran/intrinsics/selected_kind.f90:46: internal compiler err=
 >or: in rs6000_output_function_epilogue, at config/rs6000/rs6000.c:12532
 >Please submit a full bug report,
 >with preprocessed source if appropriate.
 >See <URL:http://www.suse.de/feedback> for instructions.
 >make[3]: *** [selected_kind.lo] Error 1
 >
 >Unfortunatly gdb seems unable to debug this :-(
Let's deal with the x86_64 problem first.  I doubt the underlying cause
of the two bugs is the same, but you never know :-)



jeff

Re: [tree-ssa] Bootstrap failure on powerpc64-linux-gnu

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 502 bytes --]

law@redhat.com writes:

>  >Unfortunatly gdb seems unable to debug this :-(
> Let's deal with the x86_64 problem first.  I doubt the underlying cause
> of the two bugs is the same, but you never know :-)

Fine with me - I'll try on monday with a colleague to debug this on
powerpc64...

Andreas
-- 
 Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj
  SuSE Linux AG, Deutschherrnstr. 15-19, 90429 Nürnberg, Germany
   GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: [tree-ssa] Bootstrap failure on powerpc64-linux-gnu

[law]

In message <u8r7z7lvww.fsf@gromit.moeb>, Andreas Jaeger writes:
 >--=-=-=
 >Content-Type: text/plain; charset=iso-8859-1
 >Content-Transfer-Encoding: quoted-printable
 >
 >law@redhat.com writes:
 >
 >>  >Unfortunatly gdb seems unable to debug this :-(
 >> Let's deal with the x86_64 problem first.  I doubt the underlying cause
 >> of the two bugs is the same, but you never know :-)
 >
 >Fine with me - I'll try on monday with a colleague to debug this on
 >powerpc64...
OK.  In the mean time I'll start a build/test cycle on my ppc32 box in
the hopes of triggering some kind of failure.  I'll hold off on the next
improvement to the jump threading stuff until we get this resolved.

jeff

Re: [tree-ssa] Bootstrap failure on powerpc64-linux-gnu

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 2369 bytes --]

law@redhat.com writes:

> In message <u8r7z7lvww.fsf@gromit.moeb>, Andreas Jaeger writes:
>  >--=-=-=
>  >Content-Type: text/plain; charset=iso-8859-1
>  >Content-Transfer-Encoding: quoted-printable
>  >
>  >law@redhat.com writes:
>  >
>  >>  >Unfortunatly gdb seems unable to debug this :-(
>  >> Let's deal with the x86_64 problem first.  I doubt the underlying cause
>  >> of the two bugs is the same, but you never know :-)
>  >
>  >Fine with me - I'll try on monday with a colleague to debug this on
>  >powerpc64...
> OK.  In the mean time I'll start a build/test cycle on my ppc32 box in
> the hopes of triggering some kind of failure.  I'll hold off on the next
> improvement to the jump threading stuff until we get this resolved.

It works for me on ppc32 - it just seems to fail on ppc64.

Looking at the error:
../../../libgfortran/intrinsics/selected_kind.f90:46: internal compiler error: in rs6000_output_function_epilogue, at config/rs6000/rs6000.c:12532

I see the following construct:

      if (! strcmp (language_string, "GNU C")
	  || ! strcmp (language_string, "GNU Objective-C"))
	i = 0;
      else if (! strcmp (language_string, "GNU F77"))
	i = 1;
      else if (! strcmp (language_string, "GNU Ada"))
	i = 3;
      else if (! strcmp (language_string, "GNU Pascal"))
	i = 2;
      else if (! strcmp (language_string, "GNU C++"))
	i = 9;
      else if (! strcmp (language_string, "GNU Java"))
	i = 13;
      else if (! strcmp (language_string, "GNU CHILL"))
	i = 44;
      else
	abort ();
      fprintf (file, "%d,", i);


Since Fortran defines:
#define LANG_HOOKS_NAME                 "GNU F95"

Similar constructs are also in dwarfout.c and dwarf2out.c (just search
for language_string).  dwarf2out.c handles Fortran95 but dwarfout.c
not as far as I can see.  Can any of the PowerPC folks look at
rs6000.c and fix that piece of code to handle Fortran95 (and I suggest
to remove GNU CHILL at the same time).

Jeff, so this does not seem to be related to any of your changes.  I
just guess that I'm the first to try Fortran95 on ppc64-linux and it
never worked before...

Cheers,
Andreas
-- 
 Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj
  SuSE Linux AG, Deutschherrnstr. 15-19, 90429 Nürnberg, Germany
   GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: [tree-ssa] Bootstrap failure on powerpc64-linux-gnu

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 629 bytes --]

Andreas Jaeger <aj@suse.de> writes:

> Similar constructs are also in dwarfout.c and dwarf2out.c (just search
> for language_string).  dwarf2out.c handles Fortran95 but dwarfout.c
> not as far as I can see.  Can any of the PowerPC folks look at
> rs6000.c and fix that piece of code to handle Fortran95 (and I suggest
> to remove GNU CHILL at the same time).

I've just send patches for these and test them now,

Andreas
-- 
 Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj
  SuSE Linux AG, Deutschherrnstr. 15-19, 90429 Nürnberg, Germany
   GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]