[Bug c++/34895] New: ICE at dwarf2out.c:13742

[Bug c++/34895] New: ICE at dwarf2out.c:13742

[theodore dot papadopoulo at sophia dot inria dot fr]

The (soon to be attached) code ICEs when compiled with options -g -O2.
This is still true as of gcc version 4.3.0 20080118 (x86_64-unknown-linux-gnu),
it used to pass but I do not know if this was due to the use of a previous
version of the compiler (3.4) or the use of a i386 target (this is most
probable, I will check).


-- 
           Summary: ICE at dwarf2out.c:13742
           Product: gcc
           Version: 4.1.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: theodore dot papadopoulo at sophia dot inria dot fr
 GCC build triplet: x86_64-redhat-linux
  GCC host triplet: x86_64-redhat-linux
GCC target triplet: x86_64-redhat-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34895

[Bug c++/34895] ICE at dwarf2out.c:13742

[theodore dot papadopoulo at sophia dot inria dot fr]

------- Comment #1 from theodore dot papadopoulo at sophia dot inria dot fr  2008-01-21 03:15 -------
Created an attachment (id=14984)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14984&action=view)
The code that triggers the ICE

Just compile with -g -O2 to see the problem.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34895

[Bug debug/34895] [4.1/4.2/4.3 Regression] CE in force_type_die, at dwarf2out.c:12923

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2008-01-21 10:15 -------
On trunk, ICEs with plain -g.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
      Known to fail|                            |4.1.3 4.3.0
      Known to work|                            |3.3.6
           Priority|P3                          |P2
   Last reconfirmed|0000-00-00 00:00:00         |2008-01-21 10:15:23
               date|                            |
            Summary|ICE at dwarf2out.c:13742    |[4.1/4.2/4.3 Regression] CE
                   |                            |in force_type_die, at
                   |                            |dwarf2out.c:12923
   Target Milestone|---                         |4.2.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34895

[Bug debug/34895] [4.1/4.2/4.3 Regression] ICE in force_type_die, at dwarf2out.c:12923

[matz at gcc dot gnu dot org]

------- Comment #3 from matz at gcc dot gnu dot org  2008-01-22 18:28 -------
The problem is, that gen_type_die() generates a DIE only for the main variant
of the given type.  But lookup_type_die looks in the type itself.  In the
example the type in question is "const A", for which no DIE exists already.
So force_type_die() goes into the if(), calls gen_type_die which generated
a DIE for "A" (in fact that already exists, so it just returns).  But
a DIE for "const A" still doesn't exist of course.  The example can be made to
compile to force such existence, e.g. by adding a member to "A":

  struct A {
    ...
    void cmem() const;
  };

There are two ways of fixing this:
1) making sure, that "const A" (the tree node) refers to the DIE for "A", i.e.
   losing the qualifiers.  One could ensure this in either force_type_die
   or even gen_type_die by calling equate_type_number_to_die.
2) ensure that force_type_die() really creates a DIE which reflects type
   (and not just the main variant).

With the above test one can see, that the intention was, that the AT_import
decl really comes inside the DIE for "const A".  So, that speaks for option 2.
This is also the natural choice as a function named "force_type_die" surely
should generate a DIE for that very type and not some other variant of it.

The below patch would implement this.

Index: dwarf2out.c
===================================================================
--- dwarf2out.c (revision 131712)
+++ dwarf2out.c (working copy)
@@ -13736,11 +13736,8 @@ force_type_die (tree type)
       else
        context_die = comp_unit_die;

-      type_die = lookup_type_die (type);
-      if (type_die)
-       return type_die;
-      gen_type_die (type, context_die);
-      type_die = lookup_type_die (type);
+      type_die = modified_type_die (type, TYPE_READONLY (type),
+                                   TYPE_VOLATILE (type), context_die);
       gcc_assert (type_die);
     }
   return type_die;


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34895

[Bug debug/34895] [4.1/4.2/4.3 Regression] ICE in force_type_die, at dwarf2out.c:12923

[jason at gcc dot gnu dot org]

------- Comment #4 from jason at gcc dot gnu dot org  2008-01-22 20:06 -------
That patch looks right to me.  This seems to be a bug caused by the deceptive
name "gen_type_die" which really should be called something like
"gen_main_variant_type_die".


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34895

[Bug debug/34895] [4.1/4.2/4.3 Regression] ICE in force_type_die, at dwarf2out.c:12923

[matz at gcc dot gnu dot org]

------- Comment #5 from matz at gcc dot gnu dot org  2008-01-23 16:45 -------
Subject: Bug 34895

Author: matz
Date: Wed Jan 23 16:44:27 2008
New Revision: 131761

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=131761
Log:
        PR debug/34895
        * dwarf2out.c (force_type_die): Use modified_type_die instead of
        gen_type_die.

        * g++.dg/debug/pr34895.cc: New testcase.

Added:
    trunk/gcc/testsuite/g++.dg/debug/pr34895.C
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/dwarf2out.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34895

[Bug debug/34895] [4.1/4.2/4.3 Regression] ICE in force_type_die, at dwarf2out.c:12923

[matz at gcc dot gnu dot org]

------- Comment #6 from matz at gcc dot gnu dot org  2008-01-23 16:54 -------
Fixed.


-- 

matz at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34895