[Bug c/12931] New: ICE on valid code while parsing

["tm at kloo dot net" <gcc-bugzilla at gcc dot gnu dot org>]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12931

           Summary: ICE on valid code while parsing
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: tm at kloo dot net
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i386-linux
  GCC host triplet: i386-linux
GCC target triplet: sh-elf

(gdb) file cc1
Reading symbols from cc1...done.
(gdb) set args -O2 -m4 quatern.i
(gdb) run
Starting program: /home/tm/gcc-build-sh-elf/gcc/cc1 -O2 -m4 quatern.i
quatern.i:480: warning: conflicting types for built-in function 'nan'
quatern.i:556: warning: conflicting types for built-in function 'nanf'

Program received signal SIGSEGV, Segmentation fault.
0x0807a535 in tagged_types_tu_compatible_p (t1=0x40201144, t2=0x4016cec4,
flags=0) at ../../gcc/gcc/c-typeck.c:680
680	    t2 = DECL_ORIGINAL_TYPE (TYPE_NAME (t2));
(gdb) bt
#0  0x0807a535 in tagged_types_tu_compatible_p (t1=0x40201144, t2=0x4016cec4,
flags=0) at ../../gcc/gcc/c-typeck.c:680
#1  0x0807a25a in comptypes (type1=0x40201144, type2=0x4016cec4, flags=0) at
../../gcc/gcc/c-typeck.c:575
#2  0x0807aff4 in type_lists_compatible_p (args1=0x40202f64, args2=0x401760a0,
flags=0) at ../../gcc/gcc/c-typeck.c:902
#3  0x0807ae4b in function_types_compatible_p (f1=0x402043cc, f2=0x401a2360,
flags=0) at ../../gcc/gcc/c-typeck.c:861
#4  0x08079f87 in comptypes (type1=0x402043cc, type2=0x401a2360, flags=0) at
../../gcc/gcc/c-typeck.c:523
#5  0x08069544 in duplicate_decls (newdecl=0x40204438, olddecl=0x401a23cc,
different_binding_level=0, different_tu=0)
    at ../../gcc/gcc/c-decl.c:809
#6  0x0806c23d in pushdecl (x=0x40204438) at ../../gcc/gcc/c-decl.c:1721
#7  0x0806deba in start_decl (declarator=0x401fe1f8, declspecs=0x40202e38,
initialized=0, attributes=0x0)
    at ../../gcc/gcc/c-decl.c:2706
#8  0x0804c9d8 in yyparse () at c-parse.y:1332
#9  0x0804fc73 in c_parse_file () at c-parse.y:3033
#10 0x080a542c in c_common_parse_file (set_yydebug=0) at
../../gcc/gcc/c-opts.c:1227
#11 0x0826cb72 in compile_file () at ../../gcc/gcc/toplev.c:1836
#12 0x08270e43 in do_compile () at ../../gcc/gcc/toplev.c:4691
#13 0x08270eca in toplev_main (argc=4, argv=0xbffff934) at
../../gcc/gcc/toplev.c:4731
#14 0x4003ab5c in __libc_start_main (main=0x80bf530 <main>, argc=4,
ubp_av=0xbffff934, init=0x80493b4 <_init>, 
    fini=0x831b1b0 <_fini>, rtld_fini=0x4000d634 <_dl_fini>,
stack_end=0xbffff92c) at ../sysdeps/generic/libc-start.c:129
(gdb) 

It looks like the second loop in tagged_types_tu_compatible_p:

static int
tagged_types_tu_compatible_p (tree t1, tree t2, int flags)
{
  tree s1, s2;
  bool needs_warning = false;
 
  /* We have to verify that the tags of the types are the same.  This
     is harder than it looks because this may be a typedef, so we have
     to go look at the original type.  It may even be a typedef of a
     typedef...  */
  while (TYPE_NAME (t1) && TREE_CODE (TYPE_NAME (t1)) == TYPE_DECL)
    t1 = DECL_ORIGINAL_TYPE (TYPE_NAME (t1));

  while (TYPE_NAME (t2) && TREE_CODE (TYPE_NAME (t2)) == TYPE_DECL)
    t2 = DECL_ORIGINAL_TYPE (TYPE_NAME (t2));
...

winds up dereferencing a null pointer when trying to find the original type.

Toshi