🚀 go-pugleaf

In article <1991Mar8.171015.2767@uvm.edu> wollman@emily.uvm.edu (Garrett Wollman) writes:
>In <lots of articles>, <lots of people> state that under C, the
>compiler must (In W. Losh's words) "do the right thing" when either 0
>or NULL is used.  This is FALSE.

Not "the right thing", but the *same* thing.

>extern void func_foo();
>
>void main() {
>	/* this */
>	func_foo("banana",0);
>	/* is NOT guaranteed to work as this */
>	func_foo("grape",(char *)0);
>	/* is */
>}
>
>Since most GNU software does not use prototypes, this is a valid
>objection, and one reason why I feel the GNU Coding Standards are an
>abomination.

Using NULL instead of 0 in the above example doesn't fix the bug, since
implementations are permitted to do "#define NULL 0".  In the absence of
prototypes, both NULL and 0 must be cast when being passed to a function
expecting a pointer.

Also, if the function is expecting a pointer type other than char* or
void*, even "#define NULL ((void*)0)" isn't enough to avoid the cast.
Different pointer types may have different representations (e.g.
word-oriented machines may use a less efficient representation for char*
and void* than for int*).  The cast is necessary to make sure that the
void* is converted to the appropriate parameter type.
--
Barry Margolin, Thinking Machines Corp.

barmar@think.com
{uunet,harvard}!think!barmar