[Bug bootstrap/45816] New: [4.6 Regression] Bootstrap comparison failure! for powerpc-apple-darwin9 with --enable-checking=release

[Bug bootstrap/45816] New: [4.6 Regression] Bootstrap comparison failure! for powerpc-apple-darwin9 with --enable-checking=release

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

           Summary: [4.6 Regression] Bootstrap comparison failure! for
                    powerpc-apple-darwin9 with --enable-checking=release
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: bootstrap
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: dominiq@lps.ens.fr
                CC: bernds@codesourcery.com
              Host: powerpc-apple-darwin9
            Target: powerpc-apple-darwin9
             Build: powerpc-apple-darwin9


Since revision 164552:

Author:    bernds
Date:    Thu Sep 23 10:04:33 2010 UTC (5 days, 1 hour ago)
Changed paths:    14
Log Message:    
    PR rtl-optimization/44374
    * basic-block.h (enum bb_flags): Add BB_MODIFIED.
    * df-core.c (df_set_bb_dirty): Set it.
    * ifcvt.c (find_memory): Remove function.
    (dead_or_predicable): Use can_move_insns_across.
    * df.h (can_move_insns_across): Declare function.
    * cfgcleanup.c (block_was_dirty): New static variable.
    (try_crossjump_bb, try_forward_edges): Test BB_MODIFIED flag rather
    than df_get_bb_dirty.
    (try_head_merge_bb): New static function.
    (try_optimize_cfg): Call it.  Call df_analyze if block_was_dirty
    is set.
    * df-problems.c: Include "target.h"
    (df_simulate_find_uses): New static function.
    (MEMREF_NORMAL, MEMREF_VOLATILE): New macros.
    (find_memory, find_memory_store): New static functions.
    (can_move_insns_across): New function.
    * Makefile.in (df-problems.o): Update dependencies.

testsuite/
    PR rtl-optimization/44374
    * gcc.target/arm/headmerge-1.c: New test.
    * gcc.target/arm/headmerge-2.c: New test.
    * gcc.target/i386/headmerge-1.c: New test.
    * gcc.target/i386/headmerge-2.c: New test.

bootstrapping powerpc-apple-darwin9 fails with

...
rm -f stage_current
Comparing stages 2 and 3
warning: gcc/cc1-checksum.o differs
Bootstrap comparison failure!
gcc/cfghooks.o differs
make[2]: *** [compare] Error 1
make[1]: *** [stage3-bubble] Error 2
make: *** [all] Error 2

Reverting this commit allows bootstrap to complete

Using built-in specs.
COLLECT_GCC=/opt/gcc/gcc4.6w-rel/bin/gfortran
COLLECT_LTO_WRAPPER=/opt/gcc/gcc4.6w-rel/libexec/gcc/powerpc-apple-darwin9/4.6.0/lto-wrapper
Target: powerpc-apple-darwin9
Configured with: ../gcc-4.6-work/configure --prefix=/opt/gcc/gcc4.6w-rel
--target=powerpc-apple-darwin9 --host=powerpc-apple-darwin9
--build=powerpc-apple-darwin9 --enable-languages=c,fortran,lto --with-gmp=/sw
--with-libiconv-prefix=/usr --with-system-zlib --with-cloog=/sw
--enable-checking=release
Thread model: posix
gcc version 4.6.0 20100927 (experimental) [trunk revision 164648r2] (GCC) 

This does not occurs for x86_64-apple-darwin10.4 nor without
--enable-checking=release

Using built-in specs.
COLLECT_GCC=gfc
COLLECT_LTO_WRAPPER=/opt/gcc/gcc4.6w/libexec/gcc/powerpc-apple-darwin9/4.6.0/lto-wrapper
Target: powerpc-apple-darwin9
Configured with: ../gcc-4.6-work/configure --prefix=/opt/gcc/gcc4.6w
--target=powerpc-apple-darwin9 --host=powerpc-apple-darwin9
--build=powerpc-apple-darwin9
--enable-languages=c,c++,fortran,objc,obj-c++,java,lto --with-gmp=/sw
--with-libiconv-prefix=/usr --with-system-zlib --with-cloog=/sw
Thread model: posix
gcc version 4.6.0 20100927 (experimental) [trunk revision 164648r1] (GCC)

[Bug bootstrap/45816] [4.6 Regression] Bootstrap comparison failure! for powerpc-apple-darwin9 with --enable-checking=release

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

Bernd Schmidt <bernds at gcc dot gnu.org> changed:

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

--- Comment #1 from Bernd Schmidt <bernds at gcc dot gnu.org> 2010-09-28 20:27:11 UTC ---
I don't think I have access to that type of machine.  Can you help narrow it
down as was done in PR45445?

[Bug bootstrap/45816] [4.6 Regression] Bootstrap comparison failure! for powerpc-apple-darwin9 with --enable-checking=release

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

--- Comment #2 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2010-09-28 20:52:27 UTC ---
> I don't think I have access to that type of machine.  

If you send me a public key, I can give you access to my machine.

> Can you help narrow it down as was done in PR45445?

I can try (under supervision!-), but probably not before the end of the week.

[Bug bootstrap/45816] [4.6 Regression] Bootstrap comparison failure! for powerpc-apple-darwin9 with --enable-checking=release

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

--- Comment #3 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2010-09-29 20:43:45 UTC ---
The failing comparison is

[karma] gcc/rel_build% ../_gcc_clean/contrib/compare-debug
stage2-gcc/cfghooks.o stage3-gcc/cfghooks.o
stage2-gcc/cfghooks.o.stripped stage3-gcc/cfghooks.o.stripped differ: char 60,
line 2
stripping off .eh_frame and LTO sections, then retrying
failed to strip off .eh_frame
stage2-gcc/cfghooks.o.stripped stage3-gcc/cfghooks.o.stripped differ: char 60,
line 2

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.6.0
            Summary|[4.6 Regression] Bootstrap  |[4.6 Regression]
                   |comparison failure! for     |--enable-checking=release
                   |powerpc-apple-darwin9 with  |causes a comparison failure
                   |--enable-checking=release   |on powerpc-darwin
           Severity|normal                      |blocker

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

--- Comment #5 from Bernd Schmidt <bernds at gcc dot gnu.org> 2010-10-02 14:18:07 UTC ---
We'll need to find out why stage1 gcc and stage2 gcc produce different output. 
To do that, the easiest thing to do is to copy object files from stage1 to
stage2, rebuild cc1, and then compile cfghooks.o normally.  Once you arrive at
a cfghooks.o which is identical to the one produced by stage1, you've found the
object file that was miscompiled in stage2.

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

--- Comment #6 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2010-10-02 16:07:53 UTC ---
> We'll need to find out why stage1 gcc and stage2 gcc produce different output. 
> To do that, the easiest thing to do is to copy object files from stage1 to
> stage2, rebuild cc1, and then compile cfghooks.o normally.  Once you arrive at
> a cfghooks.o which is identical to the one produced by stage1, you've found the
> object file that was miscompiled in stage2.

Sorry, but I don't understand what you ask to do:
(1) AFAICT stage1 gcc and stage2 gcc have to produce different output
(different compilers, different options and so on, the gcc directories contain
345 binary files that are different between stage 1 and 2);
(2) Even between stage2 and stage3 the raw files are different and have to be
filtered with contrib/compare-debug before doing the comparisons.

I think that before starting a very lengthy (and boring process) it would be
better to do some "postmortem" analysis:
(1) why does contrib/compare-debug chokes on the stage3 cfghooks.o?
(2) then why the bad sequence is generated?
(3) why this depends on the value of --enable-checking=?
(4) what is changed by revision 164552 with this respect?
(5) and so on.

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

--- Comment #7 from Bernd Schmidt <bernds at gcc dot gnu.org> 2010-10-02 16:35:28 UTC ---
The files between stage1 and stage2 are supposed to differ.  What isn't
supposed to differ (after stripping optional debug information) is stage2 and
stage3, which are produced by stage1 and stage2.  The bootstrap comparison
failure shows that gcc/cfghooks.o is the only file that differs in such a way.

The process isn't that lengthy, it's a binary search and shouldn't take more
than half an hour maximum to identify the file that's being miscompiled.  In my
experience none of the other questions can easily be answered without
identifying a testcase in such a way.

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

--- Comment #8 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2010-10-02 21:03:38 UTC ---
> The process isn't that lengthy, it's a binary search and shouldn't take more
> than half an hour maximum to identify the file that's being miscompiled.  In my
> experience none of the other questions can easily be answered without
> identifying a testcase in such a way.

I have already spent half an hour to go nowhere and I can probably spend days
(I don't have) without progress. So if you want to do that your way, I need
precise directives:
-step1 do this
-step2 do that
...
and so on.

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hjl.tools at gmail dot com

--- Comment #9 from H.J. Lu <hjl.tools at gmail dot com> 2010-10-02 23:13:10 UTC ---
This may be related to PR 45865 since the same checkin
caused both PRs. PR 45865 can be reproduced on Linux/ia32.

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

--- Comment #10 from Bernd Schmidt <bernds at gcc dot gnu.org> 2010-10-04 22:13:12 UTC ---
Does
  http://gcc.gnu.org/ml/gcc-patches/2010-10/msg00059.html
fix it?

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

--- Comment #11 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2010-10-05 09:59:01 UTC ---
> Does
>   http://gcc.gnu.org/ml/gcc-patches/2010-10/msg00059.html
> fix it?

Yes it does.

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
                 CC|                            |jakub at gcc dot gnu.org

[Bug bootstrap/45816] [4.6 Regression] --enable-checking=release causes a comparison failure on powerpc-darwin

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45816

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

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

--- Comment #12 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-04 11:56:10 UTC ---
Marking as dup of the other ppc bootstrap compare fail caused by the same
patch and fixed by the same.

*** This bug has been marked as a duplicate of bug 45801 ***