In article <on7gpug752bj3v9f0v3po33uha7a3s487r@4ax.com> , ABX 
<abx### [at] abxartpl>  wrote:

>> but potentially *breaking* the code on compilers which handle
>> multibyte character constants correctly!!!
>
> how ?

If you mix a programmer defined implementation with an implementation
defined implementation.  The following, while perfectly legal code may
break:

    unsigned long prog = FOURCHARS_TO_INT('A','B','C','D');

    if(prog == 'ABCD')
        puts("equal");
    else
        puts("not equal");

This is perfectly legal and will not break:

     unsigned long prog = 'ABCD';

    if(prog == 'ABCD')
        puts("equal");
    else
        puts("not equal");

Essentially the problem is that what you do actually creates a problem if
the implementation defined implementation differs from the programmer
defined implementation of multicharacter constants.  This may very well
happen depending i.e. on byteorder, but also on many other things.

Multicharacter constants are great if used consistently, but not if one
messes around with them like your macro does.  Some compilers think they
have to always warn you because you _might_ use it incorrectly.

POV-Ray uses it correctly and will work as it should because it never makes
any assumption like the actual representation of the multicharacter
constants, which your macro does.  It only compares multicharacter constants
with other multicharacter constants, which works without any limitations
whatsoever and is perfectly legal and clean code.  Except some compilers
with broken default warning settings warn about it, of course :-(


    Thorsten


____________________________________________________
Thorsten Froehlich, Duisburg, Germany
e-mail: tho### [at] trfde

Visit POV-Ray on the web: http://mac.povray.org

Post a reply to this message