🚀 go-pugleaf

In article <4h75af$4ln@solutions.solon.com>
           chrisw@leland.stanford.edu "Chris Wilkins" writes:

 > 
 > I was wondering if someone could tell me exactly what behaviour is guaranteed 
 > for the mem library functions when n=0. I presume the answers are:
 > 
 > memcpy()  -> do nothing
 > memmove() -> do nothing

The wording in the standard for both of these is 'copies n characters from the
object'. There is nothing to make n==0 illegal so, yes.

 > memset()  -> do nothing

The standard wording is similar wording so this is correct.

 > memcmp()  -> return 0

Probably. However the result is defined by some rather unfortunate wording in
the standard:

"The memcmp function returns an integer greater than, equal to, or less than
 zero, accordingly as the object pointed to by s1 is greater than, equal to,
 or less than the object pointed to by s2"

The problem here is that in the standard the term 'object' means
something with a well-defined non-zero size. An object also has a type.
It doesn't make much sense to describe the action of memcmp as comparing
two objects since that doesn't provide any meaning to the comparison result.
IMHO this is a defect in the standard. I've cross-posted to comp.std.c.

 > memchr()  -> return NULL

Again, probably. However there is a similar problem here to that with
memcmp.

"The memchr function returns a pointer to the located character, or a null
 pointer if the character does not occur in the object"

If this had said "... does not occur in the initial n characters of the
object" then fine. However it doesn't say that and you're back to trying
to force the term 'object' to mean something it really can't.

 > But are these guaranteed?
 > 
 > Also, when n=0 is it guaranteed that these functions will not dereference 
 > their other arguments? 

There is no explicit guarantee.

 > Basically, I'm interested whether I can safely use something like:
 > 
 > memcpy(ptr,NULL,0);

No. It is explicit that the pointer arguments to these functions have to
point to objects. The functions could legally contain code to the effect of:

    if (arg == NULL) abort();

 > or whether I have to wrap it:
 > 
 > if(len>0) {
 >     memcpy(dest,src,len);
 > }

Yes, assuming that len>0 guarantees that dest and src are legal pointers
to objects of suitable size (e.g. not null pointers).

-- 
-----------------------------------------
Lawrence Kirby | fred@genesis.demon.co.uk
Wilts, England | 70734.126@compuserve.com
-----------------------------------------