[Bug c++/52939] New: ice in gimple_get_virt_method_for_binfo with -O3

[Bug c++/52939] New: ice in gimple_get_virt_method_for_binfo with -O3

[dcb314 at hotmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

             Bug #: 52939
           Summary: ice in gimple_get_virt_method_for_binfo with -O3
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: dcb314@hotmail.com


Created attachment 27137
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27137
gzipped C++ source code

I just tried to compile the package openfst-1.3.1-1
on gcc-4.8 trunk dated 20120411 on an AMD x86_64 box.

The compiler said

determinize.cc: At global scope:
determinize.cc:38:1: internal compiler error: in
gimple_get_virt_method_for_binfo, at gimple-fold.c:3093
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

Preprocessed source code attached. Flag -O3 required.

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-04-12
          Component|c++                         |middle-end
                 CC|                            |jamborm at gcc dot gnu.org
     Ever Confirmed|0                           |1
            Summary|ice in                      |[4.7/4.8 Regression] ice in
                   |gimple_get_virt_method_for_ |gimple_get_virt_method_for_
                   |binfo with -O3              |binfo with -O3
   Target Milestone|---                         |4.7.1
      Known to fail|                            |4.7.0, 4.8.0

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-04-12 09:31:34 UTC ---
Confirmed.  Possibly a regression.

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

Martin Jambor <jamborm at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |jamborm at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #2 from Martin Jambor <jamborm at gcc dot gnu.org> 2012-04-12 15:10:50 UTC ---
Well, what happens is that the target of devirtualization which is
loaded from the VMT is:

<integer_cst 0x7fffae757200 type <pointer_type 0x7fffbf12f0a8 __vtbl_ptr_type>
constant 0>

I'll investigate why.

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

--- Comment #3 from Martin Jambor <jamborm at gcc dot gnu.org> 2012-04-12 17:05:01 UTC ---
Created attachment 27143
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27143
Simple testcase

This should be a simpler testcase.  What happens is that we are
attempting to devirtualize call to a virtual method introduced in a
descendant but with an ancestor which does not have it.  I suppose
this is OK if the call is never executed in run-time.

Dealing with this situation in gimple_get_virt_method_for_binfo would
be easy, but perhaps we want fold_ctor_reference to return NULL is it
is requested to fold an expression from beyond the provided
constructor?

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-04-13 09:58:44 UTC ---
(In reply to comment #3)
> Created attachment 27143 [details]
> Simple testcase
> 
> This should be a simpler testcase.  What happens is that we are
> attempting to devirtualize call to a virtual method introduced in a
> descendant but with an ancestor which does not have it.  I suppose
> this is OK if the call is never executed in run-time.
> 
> Dealing with this situation in gimple_get_virt_method_for_binfo would
> be easy, but perhaps we want fold_ctor_reference to return NULL is it
> is requested to fold an expression from beyond the provided
> constructor?

Well, if we know what the size of the constructor target is, yes.  Otherwise
all non-initialized parts are implicitely initialized to zero.

Thus, you need to handle this case in gimple_get_virt_method_for_binfo
anyway (fails are always better than asserts ...)

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

Martin Jambor <jamborm at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |http://gcc.gnu.org/ml/gcc-p
                   |                            |atches/2012-04/msg00835.htm
                   |                            |l

--- Comment #5 from Martin Jambor <jamborm at gcc dot gnu.org> 2012-04-13 13:55:31 UTC ---
Bah, I posted the patch to fix this with a wrong subject but there it
is:

http://gcc.gnu.org/ml/gcc-patches/2012-04/msg00835.html

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

--- Comment #6 from Martin Jambor <jamborm at gcc dot gnu.org> 2012-04-13 17:57:30 UTC ---
Author: jamborm
Date: Fri Apr 13 17:57:21 2012
New Revision: 186428

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=186428
Log:
2012-04-13  Martin Jambor  <mjambor@suse.cz>

    PR middle-end/52939
    * gimple-fold.c (gimple_get_virt_method_for_binfo): Bail out if
    fold_ctor_reference returns a zero constant.

    * testsuite/g++.dg/ipa/pr52939.C: New test.


Added:
    trunk/gcc/testsuite/g++.dg/ipa/pr52939.C
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/gimple-fold.c
    trunk/gcc/testsuite/ChangeLog

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

--- Comment #7 from Martin Jambor <jamborm at gcc dot gnu.org> 2012-04-16 10:02:12 UTC ---
Author: jamborm
Date: Mon Apr 16 10:02:04 2012
New Revision: 186489

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=186489
Log:
2012-04-16  Martin Jambor  <mjambor@suse.cz>

    PR middle-end/52939
    * gimple-fold.c (gimple_get_virt_method_for_binfo): Bail out if
    fold_ctor_reference returns a zero constant.

    * testsuite/g++.dg/ipa/pr52939.C: New test.


Added:
    branches/gcc-4_7-branch/gcc/testsuite/g++.dg/ipa/pr52939.C
Modified:
    branches/gcc-4_7-branch/gcc/ChangeLog
    branches/gcc-4_7-branch/gcc/gimple-fold.c
    branches/gcc-4_7-branch/gcc/testsuite/ChangeLog

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

Martin Jambor <jamborm at gcc dot gnu.org> changed:

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

--- Comment #8 from Martin Jambor <jamborm at gcc dot gnu.org> 2012-04-16 10:10:48 UTC ---
Fixed.

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[blelbach at cct dot lsu.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

Bryce Lelbach (wash) <blelbach at cct dot lsu.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |blelbach at cct dot lsu.edu

--- Comment #9 from Bryce Lelbach (wash) <blelbach at cct dot lsu.edu> 2012-08-22 23:52:55 UTC ---
Is the fix for this in a released version of GCC?

[Bug middle-end/52939] [4.7/4.8 Regression] ice in gimple_get_virt_method_for_binfo with -O3

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52939

--- Comment #10 from Martin Jambor <jamborm at gcc dot gnu.org> 2012-08-23 09:31:14 UTC ---
(In reply to comment #9)
> Is the fix for this in a released version of GCC?

If I understand our web ViewCVS correctly, the 4.7.1 release was based
on revision 188597 and I have committed the patch to the release
branch as revision 186489.  So yes, it should be in the 4.7.1 release.