[Bug c/54428] New: ICE in gimplify_expr, at gimplify.c:7591

[Bug c/54428] New: ICE in gimplify_expr, at gimplify.c:7591

[kyle.niemeyer at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54428

             Bug #: 54428
           Summary: ICE in gimplify_expr, at gimplify.c:7591
    Classification: Unclassified
           Product: gcc
           Version: 4.7.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: kyle.niemeyer@gmail.com


Created attachment 28108
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28108
preprocessed file

I get the compiler error "internal compiler error: in gimplify_expr, at
gimplify.c:7591" on the line:

rt[i] = f[i] - s_val * cpow(w[i], K) * b[i];

System information:
-gcc version: GNU C (GCC) version 4.7.1 (x86_64-apple-darwin12.0.0)
-system type: x86_64-apple-darwin12.0.0
-gcc build options: ../configure --enable-languages=c,c++,fortran
--prefix=/usr/local/Cellar/gcc/4.7.1/gcc
--datarootdir=/usr/local/Cellar/gcc/4.7.1/share
--bindir=/usr/local/Cellar/gcc/4.7.1/bin --program-suffix=-4.7
--with-gmp=/usr/local/Cellar/gmp/5.0.5 --with-mpfr=/usr/local/Cellar/mpfr/3.1.1
--with-mpc=/usr/local/Cellar/libmpc/1.0 --with-system-zlib
--enable-stage1-checking --enable-plugin --enable-lto --disable-multilib
--disable-nls

-command: gcc -c -o obj/cf.o cf.c -O0 -g3 -fbounds-check -Wunused-variable
-Wunused-parameter -Wall -pedantic -ftree-vrp -std=c99 -Wextra
-fno-strict-aliasing -fwrapv -v -save-temps -I.

-compiler output: cf.c: In function 'cf':
cf.c:290:30: internal compiler error: in gimplify_expr, at gimplify.c:7591

I've attached the preprocessed file cf.i.

Strangely, when compiled with "gcc -c -o obj/cf.o cf.c -O3 -ffast-math -std=c99
-I.", there is no error (and the results are correct).

[Bug c/54428] ICE in gimplify_expr, at gimplify.c:7591

[polacek at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54428

Marek Polacek <polacek at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |polacek at redhat dot com

--- Comment #1 from Marek Polacek <polacek at redhat dot com> 2012-08-31 05:16:38 UTC ---
Reproduced.  Happens with 4.7/trunk, but not 4.6 and older.  Only -std=c99 is
needed, with -ffast-math the ICE does not happen.

[Bug c/54428] ICE in gimplify_expr, at gimplify.c:7591

[polacek at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54428

--- Comment #2 from Marek Polacek <polacek at redhat dot com> 2012-08-31 05:46:07 UTC ---
Reduced.  Note that the typedef is needed to trigger ICE.

typedef double _Complex fftw_complex;
extern fftw_complex *fftw_alloc_complex (int);

int
cf (int n)
{
  int K = 75; 
  int nf = 1024;
  int i;
  fftw_complex *w = fftw_alloc_complex (nf);
  fftw_complex *f = fftw_alloc_complex (nf);
  double S[K * sizeof (double)];
  double s_val = S[n];
  fftw_complex *b = fftw_alloc_complex (nf);
  fftw_complex *rt = fftw_alloc_complex (nf);
  for (i = 0; i < nf; ++i)
    {   
      rt[i] = f[i] - s_val * __builtin_cpow (w[i], K) * b[i];
    }   
}

[Bug c/54428] [4.7/4.8 Regression] ICE in gimplify_expr, at gimplify.c:7591

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54428

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2012-08-31
                 CC|                            |jakub at gcc dot gnu.org
   Target Milestone|---                         |4.7.2
            Summary|ICE in gimplify_expr, at    |[4.7/4.8 Regression] ICE in
                   |gimplify.c:7591             |gimplify_expr, at
                   |                            |gimplify.c:7591
     Ever Confirmed|0                           |1

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-08-31 06:57:24 UTC ---
Slightly more reduced:
typedef double _Complex C;

C
foo (C x, C y, double *s, double z, C w)
{
  return y - z * __builtin_cpow (x, 75) * w;
}

[Bug c/54428] [4.7/4.8 Regression] ICE in gimplify_expr, at gimplify.c:7591

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54428

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|jakub at gcc dot gnu.org    |
         AssignedTo|unassigned at gcc dot       |jakub at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-08-31 07:29:20 UTC ---
Started with http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=178209

[Bug c/54428] [4.7/4.8 Regression] ICE in gimplify_expr, at gimplify.c:7591

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54428

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-08-31 08:37:09 UTC ---
Created attachment 28109
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28109
gcc48-pr54428.patch

Untested fix.

[Bug c/54428] [4.7/4.8 Regression] ICE in gimplify_expr, at gimplify.c:7591

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54428

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-08-31 19:01:05 UTC ---
Author: jakub
Date: Fri Aug 31 19:00:59 2012
New Revision: 190840

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=190840
Log:
    PR c/54428
    * c-convert.c (convert): Don't call fold_convert_loc if
    TYPE_MAIN_VARIANT of a COMPLEX_TYPE is the same, unless e
    is a COMPLEX_EXPR.  Remove TYPE_MAIN_VARIANT check from
    COMPLEX_TYPE -> COMPLEX_TYPE conversion.

    * gcc.c-torture/compile/pr54428.c: New test.

Added:
    trunk/gcc/testsuite/gcc.c-torture/compile/pr54428.c
Modified:
    trunk/gcc/c/ChangeLog
    trunk/gcc/c/c-convert.c
    trunk/gcc/testsuite/ChangeLog

[Bug c/54428] [4.7/4.8 Regression] ICE in gimplify_expr, at gimplify.c:7591

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54428

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-09-03 16:04:25 UTC ---
Author: jakub
Date: Mon Sep  3 16:04:19 2012
New Revision: 190899

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=190899
Log:
    Backported from mainline
    2012-08-31  Jakub Jelinek  <jakub@redhat.com>

    PR c/54428
    * c-convert.c (convert): Don't call fold_convert_loc if
    TYPE_MAIN_VARIANT of a COMPLEX_TYPE is the same, unless e
    is a COMPLEX_EXPR.  Remove TYPE_MAIN_VARIANT check from
    COMPLEX_TYPE -> COMPLEX_TYPE conversion.

    * gcc.c-torture/compile/pr54428.c: New test.

Added:
    branches/gcc-4_7-branch/gcc/testsuite/gcc.c-torture/compile/pr54428.c
Modified:
    branches/gcc-4_7-branch/gcc/ChangeLog
    branches/gcc-4_7-branch/gcc/c-convert.c
    branches/gcc-4_7-branch/gcc/testsuite/ChangeLog

[Bug c/54428] [4.7/4.8 Regression] ICE in gimplify_expr, at gimplify.c:7591

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54428

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

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

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-09-03 16:07:23 UTC ---
Fixed.