Dean Roddey wrote in a message to Wim Veldhuis:

WV> Just out of curiosity, I used to think that the above syntax is
WV> just used to  tell the compiler that the bytes starting at the
WV> address  symbolized by p16Pointer have to be interpreted as a void *
WV>  WITHOUT any altering of those bytes. Since OS/2 specifies  special
WV> API functions to convert pointers this would not be  

 DR> Actually the bytes pointed to are irrelevant. The whole
 DR> casting thing changes the content of the pointer itself,
 DR> from a flag 32 bit offset to a segmented address. 

I think this is bad. It breaks the rule that a typecast does not alter the
actual bytes, just the interpretation of them. If CSet does this it would
almost surely mean that the code you produced is non-portable, even not to
another compiler on the same platform !
I would prefer the API calls since that is standard for C++.

OTOH if it is done during the assignment, it would mean an overloaded
operator= is used which is standard C++ (although not all RTL supply it),
but it would not require a cast.

mvg/wr
  
--- timEd/2 1.01.g3+