gdb 2004-02-20-cvs solaris fails to build "this" properly

[Robert R. Henry]

This describes a problem with gdb evaluating function member function calls
from the "print" command.

The gdb I'm using was built from the sourceforge cvs checkout
2004-02-20-cvs, with gdb configured using
  ./configure --host=sparc64-sun-solaris2.8
as built with gcc 3.2.2.

Consider the program shown at the bottom. That program is compiled
using g++ 3.2.2 -g.

If I use the buggy gdb to put a breakpoint on the line "p->dump()",
and then do:
  (gdb) print p
    $1 = 0x20be0
and then use the print command to evaluate a call to p->dump()
  (gdb) print p->dump()
it appears as if the formal parameter "this" in Thing::dump
is initialized with a bogus actual parameter (0 in this case,
but I've seen 1 in my real application)

    Program received signal SIGSEGV, Segmentation fault.
    0x0001084c in Thing::dump (this=0x0) at x.cc:20
    20        printf("a=%d\n", a);
    The program being debugged was signaled while in a function called from GDB.
    GDB remains in the frame where the signal was received.
    To change this behavior use "set unwindonsignal on"
    Evaluation of the expression containing the function (Thing::dump()) will 
be abandoned.


#include <stdio.h>
class Thing {
private:
  int a;
  int b;
public:
  Thing(void);
  ~Thing(void);
  void dump(void);
};
typedef Thing * ThingP;
int main(void)
{
  ThingP p = new Thing();
  p->dump();
  return 0;
}
void Thing::dump(void)
{
  printf("a=%d\n", a);
  printf("b=%d\n", b);
}
Thing::Thing(void)
{
  a = 0;
  b = 0;
}
Thing::~Thing(void)
{
}