Re: c++/10405: [3.3/3.4 regression] Segfault in setup_class_bindings

Re: c++/10405: [3.3/3.4 regression] Segfault in setup_class_bindings

[nathan]

Synopsis: [3.3/3.4 regression] Segfault in setup_class_bindings

State-Changed-From-To: analyzed->closed
State-Changed-By: nathan
State-Changed-When: Sun Apr 20 12:18:57 2003
State-Changed-Why:
    fixed

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10405

Re: c++/10405: [3.3/3.4 regression] Segfault in setup_class_bindings

[bangerth]

Old Synopsis: [3.3/3.4 regression] [x86-64] Recently introduced segfault in setup_class_bindings
New Synopsis: [3.3/3.4 regression] Segfault in setup_class_bindings

Responsible-Changed-From-To: unassigned->nathan
Responsible-Changed-By: bangerth
Responsible-Changed-When: Tue Apr 15 16:09:01 2003
Responsible-Changed-Why:
    Author of a patch that might have caused this.
State-Changed-From-To: open->analyzed
State-Changed-By: bangerth
State-Changed-When: Tue Apr 15 16:09:01 2003
State-Changed-Why:
    Man, this was a tough one (looks like memory corruption, since
    changing names makes the ICE go away sometimes):
    ----------------------------
    struct Empty {};
    struct Y {
        Y();
    };
    
    struct S : Y, Empty {
        enum {};
        bool b1, b2, b3, b4, b5, b6, SAMENAME, b7, b8, b9;
        enum {};
        enum SAMENAME { n };
    };
    S s;
    --------------------------
    This crashes 3.3 and mainline (3.4). It doesn't in 3.2.3.
    
    x/z> /home/bangerth/bin/gcc-3.4-pre/bin/c++ -c x.cc
    x.cc:12: internal compiler error: Segmentation fault
    Please submit a full bug report,
    with preprocessed source if appropriate.
    See <URL:http://gcc.gnu.org/bugs.html> for instructions.
    
    The ICE happens here:
    (gdb) bt
    #0  setup_class_bindings (name=0x401989c0, type_binding_p=1)
        at ../../gcc-3.4-CVS/gcc/cp/search.c:2052
    #1  0x0812d060 in dfs_push_type_decls (binfo=0x40198740, data=0x0)
        at ../../gcc-3.4-CVS/gcc/cp/search.c:2104
    #2  0x0812b2d8 in dfs_walk_real (binfo=0x40198740, prefn=0, 
        postfn=0x812cfc0 <dfs_push_type_decls>, 
        qfn=0x812c8a0 <unmarked_pushdecls_p>, data=0x0)
        at ../../gcc-3.4-CVS/gcc/cp/search.c:1615
    #3  0x0812b41e in dfs_walk (binfo=0x0, fn=0, qfn=0, data=0x0)
        at ../../gcc-3.4-CVS/gcc/cp/search.c:1629
    #4  0x0812d4b4 in push_class_decls (type=0x2)
        at ../../gcc-3.4-CVS/gcc/cp/search.c:2169
    #5  0x080cc67b in pushclass (type=0x4019b000, modify=true)
        at ../../gcc-3.4-CVS/gcc/cp/class.c:5594
    
    This is here:
      /* First, deal with the type binding.  */
      if (type_binding_p)
        {
          type_binding = lookup_member (current_class_type, name,
    				    /*protect=*/2, /*want_type=*/true);
          if (TREE_CODE (type_binding) == TREE_LIST 
    	  && TREE_TYPE (type_binding) == error_mark_node)
    	/* NAME is ambiguous.  */
    	push_class_level_binding (name, type_binding);
          else
    	pushdecl_class_level (type_binding);
        }
    It crashes in the inner if clause, since type_binding==0.
    
    Nathan, the assignment to type_binding was last changed by
    this patch:
    ----------------------------
    revision 1.252
    date: 2003/02/15 18:03:22;  author: nathan;  state: Exp;  lines: +135 -273
    	* search.c: ANSIfy function declarations and definitions.
    	* cp-tree.h (lookup_field, lookup_member): Last parameter is a bool.
    	* call.c (build_method_call, resolve_scoped_fn_name,
    	build_java_interface_fn_ref): Adjust lookup_field, lookup_member
    	calls.
    	* class.c (handle_using_decl): Likewise.
    	* decl.c (make_typename_type, make_unmound_class_template,
    	start_decl, compute_array_index_type): Likewise.
    	* decl2.c (build_expr_from_tree, build_call_from_tree): Likewise.
    	* init.c (expand_member_init, build_member_call): Likewise.
    	* pt.c (tsubst_copy, tsubst_copy_and_build, do_decl_instantiation,
    	resolve_typename_type): Likewise.
    	* typeck.c (lookup_destructor, finish_class_member_access_exprm
    	build_prememfunc_access_expr): Likewise.
    
    Would you mind taking a look?
    
    W.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10405