Re: typedefs and structs

[Dmitry Torokhov]

On Tuesday 08 November 2005 19:59, Douglas McNaught wrote:
> linas <linas@xxxxxxxxxxxxxx> writes:
> 
> > On Tue, Nov 08, 2005 at 07:37:20PM -0500, Douglas McNaught was heard to remark:
> >> 
> >> Yeah, but if you're trying to read that code, you have to go look up
> >> the declaration to figure out whether it might affect 'foo' or not.
> >> And if you get it wrong, you get silent data corruption.
> >
> > No, that is not what "pass by reference" means. You are thinking of
> > "const", maybe, or "pass by value"; this is neither.  The arg is not 
> > declared const, the subroutine can (and usually will) modify the contents 
> > of the structure, and so the caller will be holding a modified structure
> > when the callee returns (just like it would if a pointer was passed).
> 
> Right.  My point is only that it's not clear from looking at the call
> site whether a struct passed by reference will be modified by the
> callee (some people pass by reference just for "efficiency").  And if
> the called function modifies the data without the caller's knowledge,
> it leads to obscure bugs.  Whereas if you pass a pointer, it's
> immediately clear that the called function can modify the pointed-to
> object.
>

A structure is almost never passed by value, no matter whether it is C
or C++. So both languages require you either use descriptive naming or
look up declaration/implementation:

C:
	struct str {
		char buf[1024];
		int count;
	};
	struct str s;

	do_something_with_s(&s);
	do_something_else_with_s(&s);

Which one modufies s?

-- 
Dmitry
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/