Re: RFA: Fix compile time warnings building iq2000-tdep.c

[Nick Clifton <nickc@redhat.com>]

Hi Daniel,

> Looks OK except...
> 
>> *************** iq2000_extract_return_value (struct type
>> *** 558,564 ****
>>        returned in a register, and if larger than 8 bytes, it is 
>>        returned in a stack location which is pointed to by the same
>>        register.  */
>> !   CORE_ADDR return_buffer;
>>     int len = TYPE_LENGTH (type);
>>   
>>     if (len <= (2 * 4))
>> --- 558,564 ----
>>        returned in a register, and if larger than 8 bytes, it is 
>>        returned in a stack location which is pointed to by the same
>>        register.  */
>> !   gdb_byte return_buffer;
>>     int len = TYPE_LENGTH (type);
>>   
>>     if (len <= (2 * 4))
> 
> That?  Aren't you going to run off the end of that if it's only a
> single byte?

Urk, yes.  How about this version instead:

*************** iq2000_extract_return_value (struct type
*** 558,564 ****
        returned in a register, and if larger than 8 bytes, it is
        returned in a stack location which is pointed to by the same
        register.  */
-   CORE_ADDR return_buffer;
     int len = TYPE_LENGTH (type);

     if (len <= (2 * 4))
--- 558,563 ----
*************** iq2000_extract_return_value (struct type
*** 582,598 ****
       }
     else
       {
         /* Return values > 8 bytes are returned in memory,
   	 pointed to by FN_RETURN_REGNUM.  */
!       regcache_cooked_read (regcache, E_FN_RETURN_REGNUM, & 
return_buffer);
!       read_memory (return_buffer, valbuf, TYPE_LENGTH (type));
       }
   }

--- 581,605 ----
       }
     else
       {
+       CORE_ADDR dummy;
+       union
+       {
+ 	gdb_byte bytes[sizeof dummy];
+ 	CORE_ADDR addr;
+       }
+       return_buffer;
+
         /* Return values > 8 bytes are returned in memory,
   	 pointed to by FN_RETURN_REGNUM.  */
!       regcache_cooked_read (regcache, E_FN_RETURN_REGNUM, 
return_buffer.bytes);
!       read_memory (return_buffer.addr, valbuf, TYPE_LENGTH (type));
       }
   }

(Rest of the patch as before).

This makes the puning of gdb_byte and CORE_ADDR explicit, which I think 
is a good thing, but maybe there is a preferred gdb way of doing this ? 
  (Also I could not think of a clean way of obtaining the size of a 
CORE_ADDR without using a dummy variable).

Cheers
   Nick