[Bug target/49909] New: FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[Bug target/49909] New: FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[howarth at nitro dot med.uc.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49909

           Summary: FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: howarth@nitro.med.uc.edu


The bootstrap of FSF gcc 4.6.1 fails on Core Solo/Core Duo processors when
configure is executed within the x86_64 multilib and attempts to execute x86_64
code on a non-EMT64 processor. This failure in the config.log from the x86_64
multilib of libquadmath is...

configure:3307: checking for C compiler default output file name
configure:3329: /sw/src/fink.build/gcc46-4.6.1-1000/darwin_objdir/./gcc/xgcc
-B/sw/src/fink.build/gcc46-4.6.1-1000/darwin_objdir/./gcc/
-B/sw/lib/gcc4.6/i386-apple-darwin10.8.0/bin/
-B/sw/lib/gcc4.6/i386-apple-darwin10.8.0/lib/ -isystem
/sw/lib/gcc4.6/i386-apple-darwin10.8.0/include -isystem
/sw/lib/gcc4.6/i386-apple-darwin10.8.0/sys-include  -m64 -g -O2   conftest.c 
>&5
configure:3333: $? = 0
configure:3370: result: a.out
configure:3386: checking whether the C compiler works
configure:3395: ./a.out
../../../../gcc-4.6.1/libquadmath/configure: line 3397: ./a.out: Bad CPU type
in executable
configure:3399: $? = 126
configure:3406: error: in
`/sw/src/fink.build/gcc46-4.6.1-1000/darwin_objdir/i386-apple-darwin10.8.0/x86_64/libquadmath':
configure:3410: error: cannot run C compiled programs.
If you meant to cross compile, use `--host'.
See `config.log' for more details.

[Bug target/49909] FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49909

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-07-29 23:35:13 UTC ---
Use --disable-multilib.

[Bug target/49909] FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[howarth at nitro dot med.uc.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49909

--- Comment #2 from Jack Howarth <howarth at nitro dot med.uc.edu> 2011-07-29 23:35:46 UTC ---
Created attachment 24867
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24867
config.log file from x86_64 multilib directory on Core Duo (non-EMT64)

[Bug target/49909] FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[howarth at nitro dot med.uc.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49909

--- Comment #3 from Jack Howarth <howarth at nitro dot med.uc.edu> 2011-07-29 23:43:12 UTC ---
This section of configure appears unprepared for the possibility of compiling
on i386-apple-darwin* with a multilib that can't be executed on the non-EMT64
cpu...


# Check that the compiler produces executables we can run.  If not, either
# the compiler is broken, or we cross compile.
{ $as_echo "$as_me:${as_lineno-$LINENO}: checking whether the C compiler works"
>&5
$as_echo_n "checking whether the C compiler works... " >&6; }
# If not cross compiling, check that we can run a simple program.
if test "$cross_compiling" != yes; then
  if { ac_try='./$ac_file'
  { { case "(($ac_try" in
  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
  *) ac_try_echo=$ac_try;;
esac
eval ac_try_echo="\"\$as_me:${as_lineno-$LINENO}: $ac_try_echo\""
$as_echo "$ac_try_echo"; } >&5
  (eval "$ac_try") 2>&5
  ac_status=$?
  $as_echo "$as_me:${as_lineno-$LINENO}: \$? = $ac_status" >&5
  test $ac_status = 0; }; }; then
    cross_compiling=no
  else
    if test "$cross_compiling" = maybe; then
        cross_compiling=yes
    else
        { { $as_echo "$as_me:${as_lineno-$LINENO}: error: in \`$ac_pwd':" >&5
$as_echo "$as_me: error: in \`$ac_pwd':" >&2;}
as_fn_error "cannot run C compiled programs.
If you meant to cross compile, use \`--host'.

See \`config.log' for more details." "$LINENO" 5; }
    fi
  fi
fi
{ $as_echo "$as_me:${as_lineno-$LINENO}: result: yes" >&5
$as_echo "yes" >&6; }

rm -f -r a.out a.out.dSYM a.exe conftest$ac_cv_exeext b.out conftest.out
ac_clean_files=$ac_clean_files_save
# Check that the compiler produces executables we can run.  If not, either
# the compiler is broken, or we cross compile.
{ $as_echo "$as_me:${as_lineno-$LINENO}: checking whether we are cross
compiling" >&5
$as_echo_n "checking whether we are cross compiling... " >&6; }
{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $cross_compiling" >&5
$as_echo "$cross_compiling" >&6; }

[Bug target/49909] FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[howarth at nitro dot med.uc.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49909

--- Comment #4 from Jack Howarth <howarth at nitro dot med.uc.edu> 2011-07-30 02:58:20 UTC ---
(In reply to comment #1)
> Use --disable-multilib.

Not an option for us on fink since the same packaging script has to both list
the built shared libraries
and run on both non-EMT64 and EMT64 processors. For now, I am forced to just
short circuit the
test locally with the hack...

--- gcc-4.6.1/libquadmath/configure.orig        2011-07-29 21:31:51.000000000
-0400
+++ gcc-4.6.1/libquadmath/configure     2011-07-29 21:32:36.000000000 -0400
@@ -3395,7 +3395,7 @@
 eval ac_try_echo="\"\$as_me:${as_lineno-$LINENO}: $ac_try_echo\""
 $as_echo "$ac_try_echo"; } >&5
   (eval "$ac_try") 2>&5
-  ac_status=$?
+  ac_status=0
   $as_echo "$as_me:${as_lineno-$LINENO}: \$? = $ac_status" >&5
   test $ac_status = 0; }; }; then
     cross_compiling=no

so that libquadmath can be built like all the other 64-bit binaries regardless
of EMT64 support.

[Bug target/49909] FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49909

--- Comment #5 from Andreas Schwab <schwab@linux-m68k.org> 2011-07-30 07:51:29 UTC ---
If you want to build x86-64 multilib natively you must use a capable host. 
Otherwise force cross compilation.

[Bug target/49909] FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49909

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID

--- Comment #6 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-07-30 09:38:14 UTC ---
indeed

[Bug target/49909] FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[howarth at nitro dot med.uc.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49909

--- Comment #7 from Jack Howarth <howarth at nitro dot med.uc.edu> 2011-07-30 14:06:08 UTC ---
How would I perform a cross-compilation in this situation? I considered that
option but isn't cross-compilation designed for the situation when neither the
code generated by the base compiler or the multi lib can execute on the host?
Also, it seemed that the cross compilation depended on host not being the same
as target or build (which isn't the situation here). The i386-apple-darwin10
triplet is perfectly correct for all three. So how do you do a
cross-compilation for EMT64 code (i.e. -m64) in the multi lib for a non-EMT64
processor (both of which have the same i386-apple-darwin triplet).

[Bug target/49909] FSF gcc 4.6.1 fails to bootstrap on Core Solo/Core Duo

[howarth at nitro dot med.uc.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49909

--- Comment #8 from Jack Howarth <howarth at nitro dot med.uc.edu> 2011-07-30 14:22:07 UTC ---
Also,  while I don't have a Core Solo or Core Duo machine to test this on, I
would expect this issue to exist on other i386-*-* targets. The default build
will create a multilibs for x86_64 but  the build will fail in the x86_64
multilib for libquadmath (which is unique in the FSF gcc multilibs as its
configure tries to execute generated code). Surely this late failure in the
bootstrap is considered a bug.