[Bug c++/24780] New: ICE set_mem_attributes_minus_bitpos

[Bug c++/24780] New: ICE set_mem_attributes_minus_bitpos

[ericboesch at hotmail dot com]

In a file that defined several templates, I received the following error
message:

internal compiler error: in set_mem_attributes_minus_bitpos, at emit-rtl.c:1539

Bug 20073 involved the same error message, but that was marked "fixed" last
February.  The .ii file compiles in 3.4.4 but fails in 4.0.2.  I don't have
enough disk space to check out and test newer versions.

The .ii file is at http://mysite.verizon.net/eric.boesch1/TicTacToe.ii.bz2

(Apologies if this report is redundant or less than helpful.)


-- 
           Summary: ICE set_mem_attributes_minus_bitpos
           Product: gcc
           Version: 4.0.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ericboesch at hotmail dot com
  GCC host triplet: i686-pc-cygwin


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2005-11-10 17:52 -------
Reducing (this means I can reproduce it).


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|target                      |c++
           Keywords|                            |ice-on-valid-code
      Known to fail|                            |4.1.0 4.0.3
      Known to work|                            |3.4.0
            Summary|ICE                         |[4.0/4.1 Regression] ICE
                   |set_mem_attributes_minus_bit|set_mem_attributes_minus_bit
                   |pos                         |pos
   Target Milestone|---                         |4.0.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2005-11-10 21:11 -------
Confirmed, reduced testcase:
template<typename S=int>
struct Move {
  Move();
  static Move<S> const MOVES[2][2];
};
template<typename S>
  Move<S> const Move<S>::MOVES[2][2]={};
typedef Move<int> const MoveClass;
void moves(int x, int y) {
  &MoveClass::MOVES[x][y];
}


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |critical
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
 GCC target triplet|i686-pc-cygwin              |
   Last reconfirmed|0000-00-00 00:00:00         |2005-11-10 21:11:19
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[janis at gcc dot gnu dot org]

------- Comment #3 from janis at gcc dot gnu dot org  2005-11-10 22:53 -------
A regression hunt identified this patch:

r95217 | jakub | 2005-02-18 06:58:40 +0000 (Fri, 18 Feb 2005) | 9 lines

        PR c++/19813
        * emit-rtl.c (set_mem_attributes_minus_bitpos): Add assertion
        that ref to be marked MEM_READONLY_P doesn't have base that needs
        constructing.

        * decl.c (start_decl_1): Clear TREE_READONLY flag if
        its type has TYPE_NEEDS_CONSTRUCTING.
        (complete_vars): Likewise.

That's not terribly useful, of course, because it's when the ICE check was
added, not when an existing check started triggering.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2005-11-10 23:20 -------
Mark do you know if your patch for PR 20912 would fix this?


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mmitchel at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[mark at codesourcery dot com]

------- Comment #5 from mark at codesourcery dot com  2005-11-11 06:26 -------
Subject: Re:  [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

pinskia at gcc dot gnu dot org wrote:
> ------- Comment #4 from pinskia at gcc dot gnu dot org  2005-11-10 23:20 -------
> Mark do you know if your patch for PR 20912 would fix this?

I don't know one way or the other.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[jakub at gcc dot gnu dot org]

------- Comment #6 from jakub at gcc dot gnu dot org  2005-11-11 09:31 -------
Here complete_vars is called (via
complete_type->instantiate_class_template->finish_struct_1 call chain) before
TYPE_NEEDS_CONSTRUCTING flag is set
(mark_used->instantiate_decl->tsubst->build_cplus_array_type call chain),
so TREE_READONLY is set on the decl and later on its type becomes
TYPE_NEEDS_CONSTRUCTING.

complete_vars calls complete_type before calling cp_apply_type_quals_to_decl,
and complete_type should presumably set TYPE_NEEDS_CONSTRUCTING flag
(typeck.c:112), but in this case TYPE_NEEDS_CONSTRUCTING is still set only
on the inner array (t) and not on TYPE_MAIN_VARIANT (t).

A quick hack could be to dive recursively in cp_apply_type_quals_to_decl
into ARRAY_TYPEs, as in:
...
      type_quals &= ~TYPE_QUAL_CONST;
  else
    for (; TREE_CODE (type) == ARRAY_TYPE; type = TREE_TYPE (type))
      if (TYPE_NEEDS_CONSTRUCTNG (type) || TYPE_HAS_MUTABLE_P (type))
        {
          type_quals &= ~TYPE_QUAL_CONST;
          break;
        }

  c_apply_type_quals_to_decl (type_quals, decl);

or in complete_type looking at both t and its main variant:

           TYPE_NEEDS_CONSTRUCTING (type)
             = TYPE_NEEDS_CONSTRUCTING (t) || TYPE_NEEDS_CONSTRUCTING
(TYPE_MAIN_VARIANT (t));
           TYPE_HAS_NONTRIVIAL_DESTRUCTOR (type)
             = TYPE_HAS_NONTRIVIAL_DESTRUCTOR (t) ||
TYPE_HAS_NONTRIVIAL_DESTRUCTOR (TYPE_MAIN_VARIANT (t));

Note that complete_type doesn't update TYPE_NEEDS_CONSTRUCTING etc. for
main variant, just for the type that has been requested.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[jakub at gcc dot gnu dot org]

------- Comment #7 from jakub at gcc dot gnu dot org  2005-11-11 09:39 -------
Or broadcast both flags to all type variants similarly how
layout_type->finalize_type_size broadcasts type size etc.
I.e.
-      TYPE_NEEDS_CONSTRUCTING (type)
-        = TYPE_NEEDS_CONSTRUCTING (TYPE_MAIN_VARIANT (t));
-      TYPE_HAS_NONTRIVIAL_DESTRUCTOR (type)
-        = TYPE_HAS_NONTRIVIAL_DESTRUCTOR (TYPE_MAIN_VARIANT (t));
+      unsigned need_constructing = TYPE_NEEDS_CONSTRUCTING (TYPE_MAIN_VARIANT
(t));
+      unsigned nontriv_destructor = TYPE_HAS_NONTRIVIAL_DESTRUCTOR
(TYPE_MAIN_VARIANT (t));
+      for (t = TYPE_MAIN_VARIANT (type); t; t = TYPE_NEXT_VARIANT (t))
+         {
+           TYPE_NEEDS_CONSTRUCTING (t) = need_constructing;
+           TYPE_HAS_NONTRIVIAL_DESTRUCTOR (t) = nontriv_destructor;
+         }


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[jakub at gcc dot gnu dot org]

------- Comment #8 from jakub at gcc dot gnu dot org  2005-11-12 20:44 -------
Subject: Bug 24780

Author: jakub
Date: Sat Nov 12 20:44:55 2005
New Revision: 106833

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=106833
Log:
        PR c++/24780
        * typeck.c (complete_type): Set TYPE_NEEDS_CONSTRUCTING
        and TYPE_HAS_NONTRIVIAL_DESTRUCTOR flags for all variants
        of array type.

        * g++.dg/opt/pr24780.C: New test.

Added:
    trunk/gcc/testsuite/g++.dg/opt/pr24780.C
Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/typeck.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[jakub at gcc dot gnu dot org]

------- Comment #9 from jakub at gcc dot gnu dot org  2005-11-12 20:45 -------
Subject: Bug 24780

Author: jakub
Date: Sat Nov 12 20:45:47 2005
New Revision: 106834

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=106834
Log:
        PR c++/24780
        * typeck.c (complete_type): Set TYPE_NEEDS_CONSTRUCTING
        and TYPE_HAS_NONTRIVIAL_DESTRUCTOR flags for all variants
        of array type.

        * g++.dg/opt/pr24780.C: New test.

Added:
    branches/gcc-4_0-branch/gcc/testsuite/g++.dg/opt/pr24780.C
Modified:
    branches/gcc-4_0-branch/gcc/cp/ChangeLog
    branches/gcc-4_0-branch/gcc/cp/typeck.c
    branches/gcc-4_0-branch/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780

[Bug c++/24780] [4.0/4.1 Regression] ICE set_mem_attributes_minus_bitpos

[pinskia at gcc dot gnu dot org]

------- Comment #10 from pinskia at gcc dot gnu dot org  2005-11-12 22:14 -------
Fixed.


-- 

pinskia at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24780