Re: c++/7788: g++-3.2 internal error: Segmentation fault

[Pop Sébastian <pop at gauvain dot u-strasbg dot fr>]

The following reply was made to PR c++/7788; it has been noted by GNATS.

From: =?iso-8859-1?Q?Pop_S=E9bastian?= <pop@gauvain.u-strasbg.fr>
To: nathan@gcc.gnu.org, gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org,
	jerrysiebe@mindspring.com, nobody@gcc.gnu.org,
	gcc-gnats@gcc.gnu.org
Cc: nathan@codesourcery.com
Subject: Re: c++/7788: g++-3.2 internal error: Segmentation fault
Date: Sun, 29 Sep 2002 11:08:53 +0200

 On Fri, Sep 13, 2002 at 08:39:32PM -0000, nathan@gcc.gnu.org wrote:
 > Synopsis: g++-3.2 internal error: Segmentation fault
 > 
 > State-Changed-From-To: open->analyzed
 > State-Changed-By: nathan
 > State-Changed-When: Fri Sep 13 13:39:32 2002
 > State-Changed-Why:
 >     confirmed as a regression
 > 
 > http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7788
 
 
 The problem comes from cp/rtti.c in unemitted_tinfo_decl_p ().
 
 Associated patch with these changes is:
 
 Date: Fri, 21 Jun 2002 12:20:51 +0100
 From: Nathan Sidwell <nathan@codesourcery.com>
 Organization: Codesourcery LLC
 X-Accept-Language: en
 To: jason@redhat.com
 Cc: gcc-patches@gcc.org
 Subject: [C++ PATCH]: Rework typeinfo objects
 
 2002-06-20  Nathan Sidwell  <nathan@codesourcery.com>
 
 
   int
 ! tinfo_decl_p (t, data)
        tree t;
        void *data ATTRIBUTE_UNUSED;
   {
 !   return TREE_CODE (t) == VAR_DECL
 !          && IDENTIFIER_GLOBAL_VALUE (DECL_NAME (t)) == (t)
 !          && TREE_TYPE (t) == tinfo_decl_type
 !          && TREE_TYPE (DECL_NAME (t));
   }
 
 
 Changed into 
 
 
   int
 ! unemitted_tinfo_decl_p (t, data)
        tree t;
        void *data ATTRIBUTE_UNUSED;
   {
 !   if (/* It's a var decl */
 !       TREE_CODE (t) == VAR_DECL
 !       /* whos name points back to itself */
 !       && IDENTIFIER_GLOBAL_VALUE (DECL_NAME (t)) == t
 !       /* whos name's type is non-null */
 !       && TREE_TYPE (DECL_NAME (t))
 !       /* and whos type is a struct */
 !       && TREE_CODE (TREE_TYPE (t)) == RECORD_TYPE
 !       /* with a first field of our pseudo type info */
 !       && TREE_TYPE (TYPE_FIELDS (TREE_TYPE (t))) == ti_desc_type_node)
 !     return 1;
 !   return 0;
   }
 
 
 I rewrote the condition as :
 
 
 int
 unemitted_tinfo_decl_p (t, data)
      tree t;
      void *data ATTRIBUTE_UNUSED;
 {
   /* It's a var decl */
   if (TREE_CODE (t) == VAR_DECL)
     {
       tree dnt = DECL_NAME (t);
       /* whos name points back to itself */
       if (IDENTIFIER_GLOBAL_VALUE (dnt) == t)
         /* whos name's type is non-null */
         if (TREE_TYPE (dnt))
           {
             tree ttt = TREE_TYPE (t);
             /* and whos type is a struct */
             if (TREE_CODE (ttt) == RECORD_TYPE) 
               {
                 tree tfttt = TYPE_FIELDS (ttt);
                 /* with a first field of our pseudo type info */
 =>              if (TREE_TYPE (tfttt) == ti_desc_type_node)
                   return 1;
               }
           }
     }
   return 0;
 }
 
 
 
 The problem is here => since tfttt is NULL in this bug-report example.
 Nathan could you review this patch and the associated bug-report please?
 
 Thanks, 
 Sebastian