Re: [patch 07/66] btrfs: clear_extent_bit error push-up

[Chris Mason <chris.mason@oracle.com>]

On Mon, Oct 31, 2011 at 02:30:42PM +0200, Ilya Dryomov wrote:
> On Wed, Oct 26, 2011 at 12:13:46PM -0400, Jeff Mahoney wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> > 
> > On 10/26/2011 12:09 PM, David Sterba wrote:
> > > On Wed, Oct 26, 2011 at 11:18:42AM -0400, Jeff Mahoney wrote:
> > >>>> extent_io_tree *tree, u64 start, u64 end, -		       gfp_t
> > >>>> mask); +		       gfp_t mask) __must_check;
> > >>> ^^^^^^^^^^^^ shouldn't this be placed at the beginning of the 
> > >>> prototype?
> > >> 
> > >> I don't see why that would need to be the case. It needs to be at
> > >> the beginning of the prototype if it's the actual function
> > >> definition but not when it's a prototype.
> > > 
> > > I'm not aware of a general recommendation, but the __must_check is 
> > > related to return value and makes sense to place it there. There is
> > > no other instance of __must_check placed at the end of 
> > > declaratin/definition so I'm applying the "be consistent with 
> > > surrounding code" rule.
> > 
> > Putting it at the beginning means indenting the entire prototype,
> > which I'm not a fan of.
> > 
> > - -Jeff
> 
> I think it would be much better to put it at the beginning.  For one
> it's conventional and the rest of the kernel puts it at the beginning.
> The other, and much more important, thing is that it greatly confuses C
> indexing tools like cscope.  Now you could argue that those tools' regexes
> should be tweaked but since it's just a matter of taste why don't put it
> at the beginning ?

The rest of the kernel seems to use the beginning.  I don't really have
a preference, but it's good to use the same standard as the rest of the
kernel.

(Sorry Jeff)

-chris