Re: [patch] wrong second parameter on sched_set/getaffinity

[Robert Love <rml at tech9 dot net>]

On Mon, 2002-11-25 at 21:28, Roland McGrath wrote:
> I believe all the 64-bit platforms (i.e. long!=int so this might matter at
> all) have register calling conventions where this change wouldn't matter.
> But someone responsible for each of those platforms should speak up.

Ulrich, I sent this awhile back and the above is as far as I ever got. 
I am no expert on arch ABI's, but the long->int change should be safe,
no?

Basically, both the sched_getaffinity and sched_setaffinity calls are
prototyped in the kernel and existing applications as:

	(pid_t, unsigned int, unsigned long *)

but glibc prototypes them as

	(pid_t, unsigned long, unsigned long *)

which trivially breaks the compile even on 32-bit architectures.

Is there a reason we cannot merge this fix?  Thanks.

	Robert Love

Attachment: affinity-wrong-param-rml-glibc-1.patch
Description: Text document