> Ok, I can't resist. Please keep in mind I don't know anything about
> the coding conventions for this project. But wouldn't it be better
> to adopt a convention that uses something like WORD16 and WORD32 for
> dealing with memory, since int, long, and short can all be compiler
> dependent?
As long as these are kept to the lowest level possible (only accessing
physical device structures), it's not a problem.
Typedefs like this often cause more problems than they solve, because
eventually someone is going to try to write to an "int" sized memory
location using a "WORD32*" or to a "WORD16" sized memory location using
and "int*".
>
> e.g.:
>
> -----<in the appropriate header:>-----
>
> typedef unsigned long WORD32;
> typedef unsigned int WORD16;
^^^
short is better.
> typedef unsigned short BYTE; /* isn't (unsigned char) safer? */
^^^^^
short is wrong, char is correct.
> -----<in fs/select.c:>-----
>
> WORD16 peekw( BYTE *x );
>
> WORD32 peekd( BYTE *x )
> {
> return peekw( x ) | ( peekw( x + 2 ) << 16 );
> /* or return peekw( &x[0] ) | ( peekw( &x[2] ) << 16 ); */
> }
This would not work as BYTE is defined above.
>
>
> I'm not sure whether typedefs or #defines would be better.
typedefs are definitely better than defines.
Eric
