Re: Yet more ARM breakage in linux-next

[Stephen Hemminger <shemminger@vyatta.com>]

On Thu, 4 Dec 2008 11:57:14 +1030
Rusty Russell <rusty@rustcorp.com.au> wrote:

> On Thursday 04 December 2008 10:07:44 Randy Dunlap wrote:
> > Rusty Russell wrote:
> > > (Yes, classic useless kerneldoc documentation doesn't actually *say*
> > > this clearly).
> >
> > oh fud.  That's not a fault of kernel-doc, just of whoever wrote it.
> > It's only as good as someone makes it.
> 
> Sorry that this came out wrong.  kernel-doc provides structure, but it can't 
> provide content.  And authors seem unable to think from the POV of someone 
> *using* the API.
> 
> With some work, I tracked it back to Stephen Hemminger for this comment in 
> 12d9c8420b9daa1da3d9e090640fb24bcd0deba2.  It's since been fixed and moved,
> but it's still:
> 
>  * __fls: find last set bit in word
>  * @word: The word to search
>  *
>  * Undefined if no set bit exists, so code should check against 0 first.
> 
> Which would be *fine* if fls() didn't have such confusing bit numbering and 
> the exact same one-line description.
> 
> Thanks,
> Rusty.

I think the idea was that fls was supposed to match ffs which had stupid
bit numbering inherited from BSD. and __ffs and __fls were same
but undefined if word is 0 so that they could just be one line asm.