A question about program entry point

A question about program entry point

[qinwei]

Dears,

When printing a c++ class method using function "cp_print_class_method", 
and the program entry point is 
set over 0x80000000(for example 0xa0001000), gdb will treat the "addr" as 
virtual address and finally
goes into the "else" branch at the following, so gdb will not print out 
the correct class method.
 
 common: 
  if (i < len)
    {
      char *demangled_name;

      fprintf_filtered (stream, "&");
      fputs_filtered (kind, stream);
      demangled_name = cplus_demangle (TYPE_FN_FIELD_PHYSNAME (f, j), 
                       DMGL_ANSI | DMGL_PARAMS);
      if (demangled_name == NULL)
    fprintf_filtered (stream, "<badly mangled name %s>",
              TYPE_FN_FIELD_PHYSNAME (f, j));
      else 
    {
      fputs_filtered (demangled_name, stream);
      xfree (demangled_name);
    }
    }
  else
    {
      fprintf_filtered (stream, "(");
      type_print (type, "", stream, -1);
      fprintf_filtered (stream, ") %d", (int) addr >> 3);
    }

I see some macros defined in "cp-support.h". Does that mean gdb support 
printing class method
only if the program entry point is less than 0x80000000? Thanks!

/* Pointer to member function.  Depends on compiler implementation.  */

#define METHOD_PTR_IS_VIRTUAL(ADDR)  ((ADDR) & 0x80000000)
#define METHOD_PTR_FROM_VOFFSET(OFFSET) (0x80000000 + (OFFSET))
#define METHOD_PTR_TO_VOFFSET(ADDR) (~0x80000000 & (ADDR)) 


Best regards,
Qinwei
Mail  qinwei@sunnorth.com.cn
Phone +86-010-62981668-2708
Fax   +86-010-62985972

Re: A question about program entry point

[Daniel Jacobowitz]

On Fri, Dec 29, 2006 at 11:26:33AM +0800, qinwei@sunnorth.com.cn wrote:
> Dears,
> 
> When printing a c++ class method using function "cp_print_class_method", 
> and the program entry point is 
> set over 0x80000000(for example 0xa0001000), gdb will treat the "addr" as 
> virtual address and finally
> goes into the "else" branch at the following, so gdb will not print out 
> the correct class method.

Take a look at the patch I just posted to gdb-patches.  I deleted the
check that's causing you trouble.

> /* Pointer to member function.  Depends on compiler implementation.  */
> 
> #define METHOD_PTR_IS_VIRTUAL(ADDR)  ((ADDR) & 0x80000000)
> #define METHOD_PTR_FROM_VOFFSET(OFFSET) (0x80000000 + (OFFSET))
> #define METHOD_PTR_TO_VOFFSET(ADDR) (~0x80000000 & (ADDR)) 

Note to myself.  I meant to delete these macros with my patch but
forgot to do so.  I deleted all the uses.

-- 
Daniel Jacobowitz
CodeSourcery