------- Comment #7 from steven at gcc dot gnu dot org 2005-12-18 16:04 -------
comment #6 says "invalid code". So is this an ICE on valid, or _invalid_ code?
Anyway,
Starting program: /abuild/stevenb/build/gcc/cc1plus t.C
A<T>::A()
Breakpoint 4, expand_member_init (name=0x401c9958) at init.c:950
950 if (!current_class_ref)
(gdb) cont
Continuing.
A<T>::A() [with T = int*]
Breakpoint 4, expand_member_init (name=0x40153c38) at init.c:950
950 if (!current_class_ref)
(gdb) next
953 if (!name)
(gdb) p debug_generic_expr(name)
intD.2 *
$9 = void
(gdb) next
973 else if (TYPE_P (name))
(gdb)
975 basetype = TYPE_MAIN_VARIANT (name);
(gdb)
976 name = TYPE_NAME (name);
(gdb)
983 if (basetype)
(gdb) p debug_generic_expr(name)
$10 = void
(gdb) p debug_generic_expr(basetype)
intD.2 *
$11 = void
(gdb)
So we have nullified name. Then we go on to the error message without a name:
(gdb) b init.c:1025
Breakpoint 5 at 0x815c5fb: file init.c, line 1025.
(gdb) cont
Continuing.
Breakpoint 5, expand_member_init (name=0x0) at init.c:1025
1025 error ("type %qD is not a direct base of %qT",
(gdb) l
1020 {
1021 if (CLASSTYPE_VBASECLASSES (current_class_type))
1022 error ("type %qD is not a direct or virtual base of %qT",
1023 name, current_class_type);
1024 else
1025 error ("type %qD is not a direct base of %qT",
1026 name, current_class_type);
1027 return NULL_TREE;
1028 }
1029
(gdb) p name
$12 = 0x0
(gdb)
Obviously we ICE when we try to print the NULL name.
--
steven at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords|ice-on-valid-code |ice-on-invalid-code
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24278
