[Bug bootstrap/47279] New: Bootstrap fails in stage1 with GCC 4.6

[Bug bootstrap/47279] New: Bootstrap fails in stage1 with GCC 4.6

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

           Summary: Bootstrap fails in stage1 with GCC 4.6
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: bootstrap
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: rguenth@gcc.gnu.org
            Target: powerpc64-*-linux


When bootstrapping trunk using a GCC 4.6 (from begin of december) host compiler
using CFLAGS="-O -g" I get the following error when linking cc1 during stage1:


/usr/lib64/gcc/powerpc64-suse-linux/4.6/../../../../powerpc64-suse-linux/bin/ld:
 libbackend.a(cse.o)(.text.unlikely+0x60): sibling call optimization to 
               `.opd' does not allow automatic multiple TOCs; recompile with 
-mminimal-toc or -fno-optimize-sibling-calls, or make `.opd' extern

I don't know if this hints at a generic problem or if this is some intermediate
failure of that particular host compiler version (which of course
bootstrapped just fine at some point).

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |amodra at gmail dot com,
                   |                            |hubicka at gcc dot gnu.org

--- Comment #1 from Jan Hubicka <hubicka at gcc dot gnu.org> 2011-01-27 16:50:43 UTC ---
I can not parse the error, but it seems to be introduced by Alan's patch
http://odin3.kernel.org/git-lewiemann/?p=devel/binutils/hjl/x86.git;a=commitdiff_plain;h=cf32326845120a7fd0d4eecbd42bd9374fa0b74d

Alan, any idea what gets wrong and why?

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[bergner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

Peter Bergner <bergner at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bergner at gcc dot gnu.org

--- Comment #2 from Peter Bergner <bergner at gcc dot gnu.org> 2011-01-27 17:56:49 UTC ---
I just tried a default64 "make CFLAGS=-O -g" build of sources from Jan 24th
(using 2.21 binutils) and it completed with no errors.  What configure options
and binutils version are you using?

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[amodra at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

--- Comment #3 from Alan Modra <amodra at gmail dot com> 2011-01-27 22:52:29 UTC ---
This is odd.  The error is given when a plt call, or a call needing an r2
offsetting stub is made but the code does not have a following nop which can be
replaced with an r2 restoring instruction.  Assuming the linker has not gone
completely crazy, the fact that it is complaining about a call to '.opd' says
the call is to a static function.  Static functions ought to be using the same
toc, so no need for r2 adjusting stub.  Can I have a look at your cse.o?

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-01-28 10:11:28 UTC ---
I am using the last 4.6 that built ok as host compiler, a snapshot from
6th december last year.  Binutils is the 2.11 release, but the build
also failed with the 2.10 release.

I configure with

 CFLAGS='-fmessage-length=0 -O2 -D_FORTIFY_SOURCE=2 -funwind-tables
-fasynchronous-unwind-tables -Wno-unused-but-set-variable -g -U_FORTIFY_SOURCE'
../configure --prefix=/usr --infodir=/usr/share/info --mandir=/usr/share/man
--libdir=/usr/lib64 --libexecdir=/usr/lib64
--enable-languages=c,c++,objc,fortran,obj-c++,java --enable-checking=yes
--with-gxx-include-dir=/usr/include/c++/4.6 --enable-ssp --disable-libssp
--disable-plugin --with-bugurl=http://bugs.opensuse.org/
'--with-pkgversion=SUSE Linux' --disable-libgcj --disable-libmudflap
--with-slibdir=/lib64 --with-system-zlib --enable-__cxa_atexit
--enable-libstdcxx-allocator=new --disable-libstdcxx-pch
--enable-version-specific-runtime-libs --program-suffix=-4.6
--enable-linux-futex --without-system-libunwind --with-plugin-ld=/usr/bin/gold
--with-cpu=power4 --enable-secureplt --with-long-double-128
--build=powerpc64-suse-linux

and build with

make bootstrap-lean 'STAGE1_CFLAGS=-O -g' 'BOOT_CFLAGS=-fmessage-length=0 -O2
-D_FORTIFY_SOURCE=2 -funwind-tables -fasynchronous-unwind-tables
-Wno-unused-but-set-variable -g -U_FORTIFY_SOURCE' -j6 -l12

I will try building the current snapshot with a 4.5 host compiler and see
if that allows using itself as host compiler again.

Unfortunately the builds are automated, so I do not have access to cse.o
at the moment (but I'll recreate the problem locally if it persists
with a current 4.6 host compiler)

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[bergner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

--- Comment #5 from Peter Bergner <bergner at gcc dot gnu.org> 2011-01-28 12:48:34 UTC ---
I assume the binutils 2.11 and 2.10 versions are typos and you meant 2.21 and
2.20?  I'll fire off a build using your configure options and report back.

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

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

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

--- Comment #6 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-01-28 15:10:00 UTC ---
Yes, the versions were typos.  I seem to have been just unlucky with the
specific host compiler version (I was then stuck with), building a current
snapshot with 4.5 and using that built snapshot to build itself seems
to work fine now (I'm now past the failure point and into stage2).

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[amodra at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

Alan Modra <amodra at gmail dot com> changed:

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

--- Comment #7 from Alan Modra <amodra at gmail dot com> 2011-01-28 15:45:37 UTC ---
Is the older 4.6 snapshot compiler still around on your system?  If so, I'd
really like to investigate the linker error, to make sure that powerpc64 ld
isn't doing something stupid with toc partitioning..

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

--- Comment #8 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-01-28 15:48:08 UTC ---
I don't have binaries anymore but it was built from SVN rev. 167488.  So using
a gcc 4.6 with that revision as host compiler should reproduce the problem.

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[amodra at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

--- Comment #9 from Alan Modra <amodra at gmail dot com> 2011-01-31 01:40:15 UTC ---
I can't duplicate the failure, even using 167488 as host compiler.  -Wl,--stats
shows:

/usr/local/powerpc-linux/bin/ld: linker stubs in 2 groups
/usr/local/powerpc-linux/bin/ld:   branch       0
/usr/local/powerpc-linux/bin/ld:   toc adjust   0
/usr/local/powerpc-linux/bin/ld:   long branch  0
/usr/local/powerpc-linux/bin/ld:   long toc adj 0
/usr/local/powerpc-linux/bin/ld:   plt call     165

so no toc adjusting stubs needed, as expected for -mcmodel=medium.  That makes
your ld failure even weirder.

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[amodra at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

--- Comment #10 from Alan Modra <amodra at gmail dot com> 2011-01-31 08:47:16 UTC ---
With enough fiddling around, I finally duplicated the error, in my case when
linking lto1.

libbackend.a(cse.o): In function `insert_const_anchors':
/src/gcc-current/gcc/cse.c:1293: sibling call optimization to `.opd' does not
allow automatic multiple TOCs; recompile with -mminimal-toc or
-fno-optimize-sibling-calls, or make `.opd' extern
/src/gcc-current/gcc/cse.c:1296: sibling call optimization to `.opd' does not
allow automatic multiple TOCs; recompile with -mminimal-toc or
-fno-optimize-sibling-calls, or make `.opd' extern
/home/alan/build/ppc/bin/ld/ld-new: final link failed: Bad value

It is a GNU ld bug.

[Bug bootstrap/47279] Bootstrap fails in stage1 with GCC 4.6

[amodra at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47279

Alan Modra <amodra at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED
         AssignedTo|unassigned at gcc dot       |amodra at gmail dot com
                   |gnu.org                     |

--- Comment #11 from Alan Modra <amodra at gmail dot com> 2011-01-31 22:53:48 UTC ---
http://sourceware.org/ml/binutils/2011-01/msg00403.html