Re: RFC: String cleanup

[Owen Taylor <otaylor redhat com>]

On Fri, 2005-12-16 at 16:37 +0100, Beno�Dejean wrote:
> Le vendredi 16 d�mbre 2005 �6:16 +0100, Xavier Bestel a �it :
> > On Fri, 2005-12-16 at 16:11, Beno�Dejean wrote:
> > 
> > > > Changing from:
> > > >   [const] char *str
> > > > to:
> > > >   [const] char str[]
> > > 
> > > Isn't that exactly the same ?
> > 
> > const char str[] is a string of chars (and a symbol pointing of them).
> > 
> > const char *str is a string of char and a pointer on them (and a
> > symbol).
> 
> I don't get it. To me, it's is exactly the same, the [] notation giving
> extra information that str is going to be used like an array.

If you tried using it in a function, then GCC may have optimized out
the difference because it could see all uses of the variable

But note that if you have 

 char *a_str = "a" and 
 char b_str[] = "b"

then &str is different things in the two cases -

 &a_str is a pointer to a char *, so there actually has to be 4 bytes
   in memory somewhere to point to the string.
 &b_str is the same as b_str ... a pointer to the first character in
   b_str.

(Similarly, sizeof(a_str) == 2, sizeof(b_str) == 4).

So, if they are global variables, then we need to reserve, in the first
case:

 4 bytes for a_str (in a non-shared segment, because of the way shared
   library relocations work. Unless you use prelink)
 2 bytes for "a" (in a shared read-only segment)

In the second case we just need:

 2 bytes for b_str/"b" (in a shared read-only segment

Isn't C wonderful?
						Owen