On Thu, May 30, 2002 at 06:02:14PM +0000, Christos Zoulas wrote:

 > Well, I've been watching this format-zero-length thingy for a while,
 > and it occured to me that we are going the wrong way about this.
 > 
 > It really should not be a compiler flag, but an attribute. What
 > happens if you turn -Wno-format-zero-length on, and then you
 > have a function that does not allow zero format and you want to
 > check it?
 > 
 > I suggest something:
 > 
 > 	__attribute__((__format__(__printf__,m,n[,
 > 	    __nullformat__[,__emptyformat__]])));
 > 
 > to show that a function allows null and/or empty format strings.

I'll take this up with the GCC folks.  I expect to get pushback on it.
I'd also have to think about how it might be implemented.

Note I implemented the "__nullformat__" thing in gcc-current .. it's generic
null argument checking ... you specify that an argument is to be non-null
like:

	void my_memcpy(void *dst, const void *src, size_t len)
	   __attribute__((__nonnull__(1,2)));

..and the printf, etc. built-ins have nonnull attributes chained to the
format attributes.

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>