Re: OOPS: 2.6.16-rc6 cpufreq_conservative

[Willy Tarreau <willy@w.ods.org>]

On Sun, Mar 19, 2006 at 10:26:30PM -0800, Linus Torvalds wrote:
> 
> 
> On Mon, 20 Mar 2006, Willy Tarreau wrote:
> > 
> > Now, does removing the macro completely change the output code ?
> > I think that if something written like this produces the same
> > code, it would be easier to read :
> > 
> > #define for_each_cpu_mask(cpu, mask)			\
> > 	for ((cpu) = 0; (cpu) < NR_CPUS; (cpu)++) {	\
> > 		unsigned long __bits = (mask).bits[0] >> (cpu); \
> > 		if (!__bits)				\
> > 			break;				\
> > 		if (!__bits & 1)			\
> > 			continue;			\
> > 		else
> 
> Absolutely, but now it has a dangling "{" that didn't get closed. So the 
> above would definitely be more readable, it just doesn't actually work.
> 
> Unless you'd do the "end_for_each_cpu" define (to close the statement), 
> and update the 300+ places that use this. Which might well be worth it.
> 
> So the subtle "break from the middle of a statement expression" was just a 
> rather hacky way to avoid having to change all the users of this macro.
>
> 			Linus

Oh, you're right, now I understand your motivation in doing this.
Then perhaps using your trick but applying it to the whole for loop
would make it easier to read ?

#define for_each_cpu_mask(cpu, mask)			\
	for ((cpu) = 0; (cpu) < NR_CPUS; (cpu)++) 	\
	    ({	unsigned long __bits = (mask).bits[0] >> (cpu); \
		if (!__bits)				\
			break;				\
		if (!__bits & 1)			\
			continue;			\
		else					\
                ...					\
            })

Please note that I've not read the rest of the code, so there
may be some problems left. However, if the above works, I find
it easier to read. And in this case, yes, it's interesting to
be able to break from within an expression.

Cheers,
Willy