| |||||||||||||
The timer API: size or type safety?The timer API: size or type safety?Posted Dec 7, 2006 16:16 UTC (Thu) by pimlott (guest, #1535)Parent article: The timer API: size or type safety? Wow--what Al has done is write a type-safe parametrically polymorphic function in C! The signature of SETUP_TIMER is effectively
void SETUP_TIMER((*function)(T), T data)
where the T can be any (pointer) type. While you can't write such a signature in C, Al guarantees that the two instances of T are in fact the same with a clever bit of code that has no run-time impact. Lovely!
(Log in to post comments)
The timer API: size or type safety? Posted Dec 7, 2006 20:34 UTC (Thu) by jzbiciak (✭ supporter ✭, #5246) [Link] Is typeof() standard C or a GNU extension?
The timer API: size or type safety? Posted Dec 8, 2006 18:20 UTC (Fri) by giraffedata (subscriber, #1954) [Link] Is typeof() standard C or a GNU extension?I'ts GNU C. But it looks like it isn't necessary for this technique.
The timer API: size or type safety? Posted Dec 8, 2006 19:42 UTC (Fri) by jzbiciak (✭ supporter ✭, #5246) [Link] Are you certain? Let's look at that closely:
typeof(*data) *p = data;
timer->function = (void (*)(void *)) func;
timer->data = (void *) p;
(void)(0 && (func(p), 0));
I guess your statement is that the last line, (func(p)) could be rewritten as (func(data)) instead. I can see that.
The timer API: size or type safety? Posted Jan 3, 2007 19:48 UTC (Wed) by rjbell4 (guest, #35764) [Link] The reason you wouldn't necessarily do that is that "data" may actually be an expression that has a side effect, so you don't want to reference it twice.
The timer API: size or type safety? Posted Jan 3, 2007 19:59 UTC (Wed) by jzbiciak (✭ supporter ✭, #5246) [Link] So, I guess typeof(*data) doesn't evaluate data then? I guess that makes sense.
The timer API: size or type safety? Posted Jan 3, 2007 20:01 UTC (Wed) by jzbiciak (✭ supporter ✭, #5246) [Link] Actually, hold on... (0 && func(data)) shouldn't evaluate data a second time under any circumstances anyway.
The timer API: size or type safety? Posted Dec 8, 2006 11:46 UTC (Fri) by viro (subscriber, #7872) [Link] Heh... The current version in my tree actually gets you a constantarithmetic expression. It's built around the following: #define check_callback_type(f,x) (sizeof((0 ? (f)(x) : (void)0), 0)) Feel free to torture students with it... It actually checks even more - namely, that f returns void. And since it never evaluates f or x, we are free to use it in macros without any concerns about side effects. Can be used in global initializers, too (and yes, it _is_ valid C99; we get past the constraints since the expression that appears to be problematic is an argument of sizeof and doesn't have a variably-modified type).
The timer API: size or type safety? Posted Dec 8, 2006 17:10 UTC (Fri) by liljencrantz (subscriber, #28458) [Link] That is so clever it hurts.
Am I correct in thinking that since check_callback_type never evaluates f or x, we also don't need to use typeof anymore, as we only truly reference f or x once? I.e. we can write this as:
check_callback_type( func, data );
|
|||||||||||||