[Bug regression/55167] New: [4.8 Regression]: g++.dg/other/vector-compare.C, ICE

[Bug regression/55167] New: [4.8 Regression]: g++.dg/other/vector-compare.C, ICE

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55167

             Bug #: 55167
           Summary: [4.8 Regression]: g++.dg/other/vector-compare.C, ICE
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: regression
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: hp@gcc.gnu.org
                CC: glisse@gcc.gnu.org
              Host: x86_64-unknown-linux-gnu
            Target: cris-axis-elf


This test previously passed, now it fails.
A patch in the revision range (last_known_working:first_known_failing)
192985:192991
exposed or caused this regression.  Since then it fails as follows:

Running /tmp/hpautotest-gcc0/gcc/gcc/testsuite/g++.dg/dg.exp ...
FAIL: g++.dg/other/vector-compare.C (internal compiler error)
FAIL: g++.dg/other/vector-compare.C (test for excess errors)

In g++.log:

Executing on host:
/tmp/hpautotest-gcc0/cris-elf/gccobj/gcc/testsuite/g++/../../g++
-B/tmp/hpautotest-gcc0/cris-elf/gccobj/gcc/testsuite/g++/../../
/tmp/hpautotest-gcc0/gcc/gcc/testsuite/g++.dg/other/vector-compare.C 
-fno-diagnostics-show-caret  -nostdinc++
-I/tmp/hpautotest-gcc0/cris-elf/gccobj/cris-elf/libstdc++-v3/include/cris-elf
-I/tmp/hpautotest-gcc0/cris-elf/gccobj/cris-elf/libstdc++-v3/include
-I/tmp/hpautotest-gcc0/gcc/libstdc++-v3/libsupc++
-I/tmp/hpautotest-gcc0/gcc/libstdc++-v3/include/backward
-I/tmp/hpautotest-gcc0/gcc/libstdc++-v3/testsuite/util -fmessage-length=0 
-std=gnu++11 -Wall  -S   -isystem
/tmp/hpautotest-gcc0/cris-elf/gccobj/cris-elf/./newlib/targ-include -isystem
/tmp/hpautotest-gcc0/gcc/newlib/libc/include  -o vector-compare.s    (timeout =
300)
/tmp/hpautotest-gcc0/gcc/gcc/testsuite/g++.dg/other/vector-compare.C: In
function 'int main()':

/tmp/hpautotest-gcc0/gcc/gcc/testsuite/g++.dg/other/vector-compare.C:30:21:
internal compiler error: Segmentation fault

0x9fc565 crash_signal

    /tmp/hpautotest-gcc0/gcc/gcc/toplev.c:333

0x802f65 force_operand(rtx_def*, rtx_def*)

    /tmp/hpautotest-gcc0/gcc/gcc/expr.c:6975

0x743ac0 expand_gimple_stmt_1

    /tmp/hpautotest-gcc0/gcc/gcc/cfgexpand.c:2168

0x743ac0 expand_gimple_stmt

    /tmp/hpautotest-gcc0/gcc/gcc/cfgexpand.c:2204

0x744b39 expand_gimple_basic_block

    /tmp/hpautotest-gcc0/gcc/gcc/cfgexpand.c:3979

0x746276 gimple_expand_cfg

    /tmp/hpautotest-gcc0/gcc/gcc/cfgexpand.c:4498


According to <http://gcc.gnu.org/ml/gcc-patches/2012-10/msg02858.html>, this
also happens for i686-linux and revision r192986 is to blame.

Author of suspect patch CC:ed.

[Bug regression/55167] [4.8 Regression]: g++.dg/other/vector-compare.C, ICE

[glisse at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55167

--- Comment #1 from Marc Glisse <glisse at gcc dot gnu.org> 2012-11-01 19:05:13 UTC ---
Thanks.

Since the error happens in expand, I assume this is a dup of PR 55001, for
which I posted a patch earlier today. Note that AFAIK, anything that starts
failing because of this patch was silently miscompiled before.

(I am also investigating a different problem with this testcase, due to the
fact that the C front-end turns (void)(x<0) into if(x<0)(void)0 which vector x
doesn't like)

[Bug regression/55167] [4.8 Regression]: g++.dg/other/vector-compare.C, ICE

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55167

--- Comment #2 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-11-01 19:52:02 UTC ---
(In reply to comment #1)
> Since the error happens in expand, I assume this is a dup of PR 55001, for
> which I posted a patch earlier today.

It could be; that PR lacks a description of the effects, so my pre-report
search did not find it.  (It only speaks of something missing in the code,
could be just a missing optimization at a glance.)

I'll leave dealing with it to you.

[Bug regression/55167] [4.8 Regression]: g++.dg/other/vector-compare.C, ICE

[glisse at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55167

--- Comment #3 from Marc Glisse <glisse at gcc dot gnu.org> 2012-11-01 20:22:14 UTC ---
(In reply to comment #2)
> It could be; that PR lacks a description of the effects, so my pre-report
> search did not find it.  (It only speaks of something missing in the code,
> could be just a missing optimization at a glance.)

Yes, sorry, "dup" probably wasn't the right term, I meant it as: "I think I
know what's going on".

Thanks again for filing the bug and CCing me.

[Bug regression/55167] [4.8 Regression]: g++.dg/other/vector-compare.C, ICE

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55167

Hans-Peter Nilsson <hp at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Depends on|                            |55001

--- Comment #4 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-11-02 05:34:06 UTC ---
I made this PR depend on PR55001, which has a committed patch that solves the
regression in this PR.  So, looks like a dup, for some definition thereof.

[Bug regression/55167] [4.8 Regression]: g++.dg/other/vector-compare.C, ICE

[glisse at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55167

Marc Glisse <glisse at gcc dot gnu.org> changed:

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

--- Comment #5 from Marc Glisse <glisse at gcc dot gnu.org> 2012-11-02 08:30:11 UTC ---
Thanks.
The failure also disappeared between:
http://gcc.gnu.org/ml/gcc-testresults/2012-11/msg00087.html
and:
http://gcc.gnu.org/ml/gcc-testresults/2012-11/msg00095.html

so let's mark this bug as fixed.

PR55001 remains open because the patch makes the compiler produce sub-optimal
code (but correct at least).