🚀 go-pugleaf

Here is the problem:

gdb-4.6 dumps core setting breakpoint on destructor when command 
completion is not used.

My system configuration:

gdb-4.6 running on Interactive UNIX 2.2.1
configure i386-none-sysv32
trad-core patch, and patches for debug register use.
#define CPLUS_MARKER '.'
None of these patches are responsible for this problem.

Here is what I did.

Concider this c++ program compiled thusly:

> gcc -v -g -o test test.cc
Reading specs from /usr/local/lib/gcc-lib/i386-isc/2.2.2/specs
gcc version 2.2.2
 /usr/local/lib/gcc-lib/i386-isc/2.2.2/cpp -lang-c++ -v -undef -D__GNUC__=2 -D__
GNUG__=2 -D__cplusplus -Dunix -Di386 -D__unix__ -D__i386__ -D__unix -D__i386 -g
test.cc /usr/tmp/cca06705.i
GNU CPP version 2.2.2 (80386, ATT syntax)
 /usr/local/lib/gcc-lib/i386-isc/2.2.2/cc1plus /usr/tmp/cca06705.i -quiet -dumpb
ase test.cc -g -version -o /usr/tmp/cca06705.s
GNU C++ version 2.2.2 (80386, ATT syntax) compiled by GNU C version 2.2.2.
 as -o test.o /usr/tmp/cca06705.s
 ld -o test /lib/crt1.o /usr/local/lib/gcc-lib/i386-isc/2.2.2/crtbegin.o -L/usr/
local/lib/gcc-lib/i386-isc/2.2.2 -L/usr/local/lib test.o -lgcc -lc -lg -lgcc /us
r/local/lib/gcc-lib/i386-isc/2.2.2/crtend.o /lib/crtn.o
>
//------ test.cc ---------------------------------------------------------
class tree_element {
  int value;
  tree_element *larger;
  tree_element *smaller;
  tree_element *parent;
public:
  int get() {return value;};
  tree_element(void) {value = 0;  parent = larger = smaller = 0;};
  tree_element(int i) {value = i; parent = larger = smaller = 0;};
  ~tree_element();
};

tree_element::~tree_element()
{
  if(larger) delete larger;
  if(smaller) delete smaller;
}

int main(int argc, char *argv[])
{

  tree_element e(3);

  int i = e.get() + 1;
  return(0);
}
//------------------------------------------------------------------

set a break point using command completion works fine.

first I do this b tree_ele<TAB>::~tree<TAB><CR>

> gdb test
GDB is free software and you are welcome to distribute copies of it
 under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.6, Copyright 1992 Free Software Foundation, Inc...
(gdb) b 'tree_element::~tree_element(void)'
Breakpoint 1 at 0x1ff: file test.cc, line 16.
(gdb)

If I try this with out using command completion (because I am using
gdb mode in emacs).

GDB is free software and you are welcome to distribute copies of it
 under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.6, Copyright 1992 Free Software Foundation, Inc.
(gdb) file test
Reading symbols from test...done.
(gdb) b tree_element::~tree_element

Program received signal 11, Segmentation fault
0x23179 in decode_line_1 (argptr=0x7ffffcf4, funfirstline=1,
    default_symtab=0x0, default_line=0) at symtab.c:1611
symtab.c:1611: No such file or directory.
(top-gdb) bt
#0  0x23179 in decode_line_1 (argptr=0x7ffffcf4, funfirstline=1,
    default_symtab=0x0, default_line=0) at symtab.c:1611
#1  0x708f in break_command_1 (arg=0x42d025 "", tempflag=0, from_tty=1)
    at breakpoint.c:1596
#2  0x7314 in break_command (arg=0x42d00a "tree_element::~tree_element",
    from_tty=1) at breakpoint.c:1681
#3  0x1659 in execute_command (p=0x42d00a "tree_element::~tree_element",
    from_tty=1) at main.c:915
#4  0x17dd in command_loop () at main.c:976
#5  0x1496 in main (argc=1, argv=0x7ffffe20) at main.c:853

(top-gdb) l
1606                  if (destructor_name_p (copy, t))
1607                    {
1608                      /* destructors are a special case.  */
1609                      struct fn_field *f = TYPE_FN_FIELDLIST1 (t, 0);
1610                      int len = TYPE_FN_FIELDLIST_LENGTH (t, 0) - 1;
1611                      char *phys_name = TYPE_FN_FIELD_PHYSNAME (f, len);
1612                      physnames[i1] = (char *)alloca (strlen (phys_name) + 1
);
1613                      strcpy (physnames[i1], phys_name);
1614                      sym_arr[i1] =
1615                        lookup_symbol (phys_name, SYMBOL_BLOCK_VALUE (sym_cl
ass),
(top-gdb) print *t
$1 = {code = TYPE_CODE_STRUCT, name = 0x430528 "struct tree_element",
  length = 16, objfile = 0x443e08, target_type = 0x0, pointer_type = 0x4460e4,
  reference_type = 0x0, function_type = 0x0, flags = 0, nfields = 4,
  fields = 0x446118, vptr_basetype = 0x0, vptr_fieldno = -1, type_specific = {
    arg_types = 0x425020, cplus_stuff = 0x425020}}


(top-gdb) print t->type_specific.cplus_stuff
$2 = (struct <opaque> *) 0x425020

(top-gdb) print *(struct cplus_struct_type *)t->type_specific.cplus_stuff
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 
WHY DO I HAVE TO DO THIS????????????????????????
THIS MAKES ME SOOOOOOOO MADDD. (a limitation of coff???)


$3 = {virtual_field_bits = 0x0, private_field_bits = 0x0,
  protected_field_bits = 0x0, nfn_fields = 0, n_baseclasses = 0,
  nfn_fields_total = 0, fn_fieldlists = 0x0, via_protected = 0 '\000',
  via_public = 0 '\000'}

(top-gdb) print &cplus_struct_default
$4 = (struct cplus_struct_type *) 0x425020

As you can see here the cplus_stuff is set to the default.  I think the
default values need to be chosen a bit better.  I am not sure what to
do about this so I did this 

1606                  if (destructor_name_p (copy, t) && 0)

Now I get a message saying I should be using command completion, but at
least I don't blow core.

Dave Daney     daney@mermaid.micro.umn.edu