[s1mp3-dev] Linux patch to s1mp3

[M. Warner Losh]

In message: <7.0.1.0.2.20060619233014.03b63408 at gmail.com>
            Bluechip <csbluechip at gmail.com> writes:
: >Actually, NULL should be defined as a plain '0'.  Comparison of
: >pointers against 0 is a special case in 'C' as is casting 0 to a
: >pointer.  Both should be the same.  In this context, the definition of
: >NULL doesn't matter.  It only matters when calling functions
: >w/old-school K&R prototypes or through "...".
: >
: >I agree we should compare against NULL if the functionr returns a
: >pointer.
: 
: I disagree.  I personally never use NULL because 
: every damn system defines it differently and 
: compilers belch all over the shop.  This is an 
: except from one of my standard include files:
: 
: #define cpNULL    ((char*)0)
: #define vpNULL    ((void*)0)
: #define bpNULL    ((UI8 *)0)
: #define wpNULL    ((UI16*)0)
: #define dpNULL    ((UI32*)0)
: #define sNULL     ((char)0)

That's ugly beyond belief.  Why use a non-standard thing when a
standard one will do?  You need a special #define for each type, which
is an insane explosion of defines, all of which are non-standard.

: // The empty string: 'empty'
: // (Didn't really seem worthy of a global definition)
: static const char scach1_NULL[1] = {'\0'};
: #define empty ((char*)scach1_NULL)

Ugg.  That's even worse.

const char *empty="";

is a lot clearer, and you'll wind up with only one copy of the empty
string rather than one per file.

: You can also choose whether to debate #define 
: over typedef ...which will ALSO cause compilers to behave differently.

Please, no #defined types.  They are always wrong.  They haven't been
right since typedef was introduced into the language around v7.

Warner