[9fans] kenc: casting expression to void not ignored

[9fans] kenc: casting expression to void not ignored

[Pietro Gagliardi]

Hello. I have a simple push macro for a stack machine interpreter. It  
looks like this:

	#define push(s) ((void)(((stackptr >= (stack + stacksize)) ?  
growstack() : (void)0), (*stackptr++ = errcheck(s))))

Does this boggle the mind? It's very simple: it checks for stack  
overflow and makes the stack bigger if so, then checks to see if s, a  
double, is NaN or ±Inf, and finally pushes it onto the stack if not  
(errcheck calls longjmp - how I wish I had try).

However, when I use the push macro, I get tons of warnings like this:

	warning: run.c:170 result of operation not used

Shouldn't the cast to void mean that I don't intend to use the result  
of the operation? Thanks.

Re: [9fans] kenc: casting expression to void not ignored

[Pietro Gagliardi]

I forgot to mention this is with the -w option; I found that turning  
it off also hid warnings I wanted to keep.

On Jun 29, 2008, at 9:53 AM, Pietro Gagliardi wrote:

> Hello. I have a simple push macro for a stack machine interpreter.  
> It looks like this:
>
> 	#define push(s) ((void)(((stackptr >= (stack + stacksize)) ?  
> growstack() : (void)0), (*stackptr++ = errcheck(s))))
>
> Does this boggle the mind? It's very simple: it checks for stack  
> overflow and makes the stack bigger if so, then checks to see if s,  
> a double, is NaN or ±Inf, and finally pushes it onto the stack if  
> not (errcheck calls longjmp - how I wish I had try).
>
> However, when I use the push macro, I get tons of warnings like this:
>
> 	warning: run.c:170 result of operation not used
>
> Shouldn't the cast to void mean that I don't intend to use the  
> result of the operation? Thanks.
>
>

Re: [9fans] kenc: casting expression to void not ignored

[erik quanstrom]

> Hello. I have a simple push macro for a stack machine interpreter. It
> looks like this:
>
> 	#define push(s) ((void)(((stackptr >= (stack + stacksize)) ?
> growstack() : (void)0), (*stackptr++ = errcheck(s))))
>
> Does this boggle the mind? It's very simple: it checks for stack
> overflow and makes the stack bigger if so, then checks to see if s, a
> double, is NaN or ±Inf, and finally pushes it onto the stack if not
> (errcheck calls longjmp - how I wish I had try).
>
> However, when I use the push macro, I get tons of warnings like this:
>
> 	warning: run.c:170 result of operation not used
>
> Shouldn't the cast to void mean that I don't intend to use the result
> of the operation? Thanks.

no.  USED(x) where x is a variable name means you do not intend to use
the result of the operation. why don't you write push like this instead?

void
push(Whatever *s)
{
	if(stackptr >= stack + stacksize)
		growstack();
	*stackptr++ = errcheck(s)
}

this eliminates two wierd void casts, a ternary operator,
a comma operator and your warning.  i find it
difficult to imagine that a function call would be significant
overhead in this case.

on the certainly wrong assumption that the function call
needs to be avoided, there's no rule that one can't use if
statements in macros:

#define push(s)\
	do {\
		if(stackptr >= stack + stacksize)\
			growstack;\
		*stackptr++ = errcheck(s);\
	} while(0)

but for pete's sake, use a function until you know it's
too slow.

- erik

Re: [9fans] kenc: casting expression to void not ignored

[Rob Pike]

> Shouldn't the cast to void mean that I don't intend to use the result of the
> operation? Thanks.

That's not standard C, just a bad convention established by lint(1)
and carried too far for a while before fading away again.

-rob