Bit-fields question

["Salman Khilji" <salmankhilji at hotmail dot com>]

Suppose I have this statement in the debuggee (target) source code.

*((float*)((void*)(address))) = new_val

<address> is of type long and <new_val> is of type float.  This technique is 
used in real-time steering of a target application.  We can generaralize 
this to other data types---we just need to know the address of a symbol and 
its type to be able to steer an application at run time.  We can then use 
TCP/IP to transfer the address to the debuggee's (target's) memory and have 
an agent execute the above kind of statement to change the value of a symbol 
without having to stop the application.

Now the question is: if you have a structure with bit-fields, how would you 
do this?  I can't find a portable way of doing this.  I have a generic API 
that can manipulate bit-fields in an unsigned char array.  You just pass int 
the bit-position, the bit-size and the value, the API takes care of the 
rest.  Lets say I have the structure:

struct BITS {
  unsigned b1:1;
  int b2:4;
};

struct BITS Object;

Now if I do this:

unsigned char * p = (unsigned char*) (&Object);

Then my generic API does not work for p.  i.e modifying bits in the 
character array does not modify the required bits.  This is most certainly 
due to byte-swapping and endianness issues.


So if we were to do this for bit-fields, what kind of information would be 
necessary to download to the target?  I know you would need the address, the 
bit-offset and the bit size.  But how would the statement (or function) look 
like that would perform actually poking into the required address?  Any 
ideas??  code??

Salman



_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp.