Bug 30132 - [4.2 Regression] ICE in find_lattice_value, at tree-complex.c:133
Summary: [4.2 Regression] ICE in find_lattice_value, at tree-complex.c:133
Alias: None
Product: gcc
Classification: Unclassified
Component: tree-optimization (show other bugs)
Version: 4.1.2
: P1 normal
Target Milestone: 4.3.0
Assignee: Not yet assigned to anyone
URL: http://gcc.gnu.org/ml/gcc-patches/200...
Keywords: ice-on-valid-code, patch
Depends on: 20280
Blocks: 30142 30165
  Show dependency treegraph
Reported: 2006-12-09 11:31 UTC by Debian GCC Maintainers
Modified: 2009-03-30 20:16 UTC (History)
5 users (show)

See Also:
Known to work: 3.4.0 4.3.0
Known to fail: 4.1.2 4.2.0 4.2.5
Last reconfirmed: 2008-01-09 14:45:38

preprocessed source (1023 bytes, application/x-gzip)
2006-12-09 11:33 UTC, Debian GCC Maintainers
patch which I need to retest and fix the last regression (582 bytes, patch)
2007-03-13 16:12 UTC, Andrew Pinski
Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Debian GCC Maintainers 2006-12-09 11:31:31 UTC
[forwarded from http://bugs.debian.org/400484]

seen on i486-linux-gnu with 3.4, 4.1.2, 4.3 20061022, works with -O0

$ gcc -c -O1 complex.c 
complex.c: In function 'testit':
complex.c:4: internal compiler error: in find_lattice_value, at tree-complex.c:133
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
Comment 1 Debian GCC Maintainers 2006-12-09 11:33:27 UTC
Created attachment 12774 [details]
preprocessed source

#include <complex.h>

void testit(double complex* t, double* b)
  b[0] = t[0]==0.0?0.0:-t[0];

  static double complex k = 5;
  static double b;

The following modification fixes the problem:

void testit(double complex* t, double* b)
  b[0] = t[0]==0.0?0.0:__real__(-t[0]);
Comment 2 Andrew Pinski 2006-12-09 13:20:26 UTC
The gimplifier is messing up ...
Comment 3 Andrew Pinski 2006-12-09 14:20:41 UTC
Note, what we have is invalid gimple which we don't catch until tree-complex.c which is wrong, I have a patch to catch it earlier.

Comment 4 Andrew Pinski 2006-12-09 14:22:06 UTC
Also note moving catching it earlier makes this a middle-end (gimplifier) issue rather than an optimization issue.
Comment 5 Andrew Pinski 2006-12-09 14:24:43 UTC
After gimplification:
  D.1853 = *t;
  if (D.1853 == __complex__ (0.0, 0.0))
      D.1854 = __complex__ (0.0, 0.0);
      iftmp.0 = &D.1854;
      D.1855 = *t;
      D.1856 = -D.1855;
      iftmp.0 = &D.1856;
  D.1857 = REALPART_EXPR <*iftmp.0>;
  *b = D.1857;

  *b = REALPART_EXPR <*t == __complex__ (0.0, 0.0) ? __complex__ (0.0, 0.0) : -*t>;

I have not figured out why we need to take the address of the temp variable yet.
Comment 6 Andrew Pinski 2006-12-09 19:11:33 UTC
We have:
      if ((fallback & fb_lvalue) == 0)
	  result = tmp2 = tmp = create_tmp_var (TREE_TYPE (expr), "iftmp");
	  ret = GS_ALL_DONE;

But fallback is "fallback=fb_either" so the above if is false.
So we have two problems, an ineffient gimplification and a gimplification that is wrong.

The ineffient gimplification is easy to fix.  I will now go to figure out why we get an invalid one.
Comment 7 Andrew Pinski 2006-12-09 19:24:45 UTC
The second issue (the invalid gimple), comes from marking a variable as ADDRESSABLE late after the orginal gimplification to that variable had happened:
	  /* Mark the RHS addressable.  */
	  lang_hooks.mark_addressable (TREE_OPERAND (expr, 0));
Comment 8 Andrew Pinski 2006-12-09 19:36:15 UTC
I now have a fix for both issues (fixing the first problem of ineffient gimplification is enough to fix this bug but might not be enough to fix others).
Comment 9 Andrew Pinski 2006-12-10 22:47:47 UTC
The patch which fixed PR 20280 causes this invalid gimple to show up.
Comment 10 Geir Johansen 2007-01-22 19:24:16 UTC
Here is a Fortran test case:

$ cat bug2737.f90
      PROGRAM get_tst_inc_complex
      implicit none
      external subrrg, checkrr
      complex :: vrr, trr
      call subrr ( )
    subroutine subrr ()
      call checkrr (trr)
    end subroutine subrr
$ gfortran --version
GNU Fortran 95 (GCC) 4.1.1 20060524 (rpm:5)
Copyright (C) 2006 Free Software Foundation, Inc.
GNU Fortran comes with NO WARRANTY, to the extent permitted by law.
You may redistribute copies of GNU Fortran
under the terms of the GNU General Public License.
For more information about these matters, see the file named COPYING
$ gfortran -c -O0 bug2737.f90
$ gfortran -c -O1 bug2737.f90
bug2737.f90: In function 'MAIN__':
bug2737.f90:1: internal compiler error: in find_lattice_value, at tree-complex.c:133
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
Comment 11 Andrew Pinski 2007-01-22 19:30:54 UTC
(In reply to comment #10)
> Here is a Fortran test case:
I think that Fortran issue is something unrelated to this bug.
In fact that Fortran testcase was fixed for 4.1.2 by PR 27889.
Comment 12 Andrew Pinski 2007-03-13 16:12:08 UTC
Created attachment 13201 [details]
patch which I need to retest and fix the last regression
Comment 13 Andrew Pinski 2007-11-02 04:18:45 UTC
Updated patch send:
Comment 14 Steven Bosscher 2007-11-10 16:40:07 UTC
Patch is waiting for approval of the C++ bits.
Comment 15 Andrew Pinski 2008-01-03 16:23:00 UTC
This was caused by the patch which fixed PR 20280 and the 2nd iteration of the patch was rejected so unassigning from me.
Comment 16 Andrew Pinski 2008-01-03 16:26:42 UTC
4.0.0 really did not work either, it just did not cause a crash as the checks for invalid gimple was not there.
Comment 17 Richard Biener 2008-01-09 14:45:38 UTC
Comment 18 Richard Biener 2008-01-09 15:47:35 UTC
Subject: Bug 30132

Author: rguenth
Date: Wed Jan  9 15:46:49 2008
New Revision: 131430

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=131430
2008-01-09  Richard Guenther  <rguenther@suse.de>
	Andrew Pinski  <andrew_pinski@playstation.sony.com>

        PR middle-end/30132
	* gimplify.c (gimplify_cond_expr): Do not create an addressable
	temporary if an rvalue is ok or an lvalue is not required.

	* gcc.c-torture/compile/pr30132.c: New testcase.


Comment 19 Richard Biener 2008-01-09 15:47:38 UTC
Fixed on the trunk.
Comment 20 Joseph S. Myers 2008-07-04 21:45:41 UTC
Closing 4.1 branch.
Comment 21 Joseph S. Myers 2009-03-30 20:16:29 UTC
Closing 4.2 branch, fixed in 4.3.