Re: linux-next: build warning after merge of the pinctrl tree

[Stephen Rothwell]

Hi all,

On Fri, 18 Mar 2022 18:31:01 +1100 Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
>
> On Thu, 20 Jan 2022 14:25:39 +1100 Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
> >
> > On Thu, 28 Oct 2021 08:03:31 +0100 Mauro Carvalho Chehab <mchehab@xxxxxxxxxx> wrote:  
> > >
> > > Em Wed, 27 Oct 2021 12:10:45 +0100
> > > Marc Zyngier <maz@xxxxxxxxxx> escreveu:
> > >     
> > > > On Wed, 27 Oct 2021 12:01:18 +0100, Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:      
> > > > > 
> > > > > After merging the pinctrl tree, today's linux-next build (htmldocs)
> > > > > produced this warning:
> > > > > 
> > > > > include/linux/gpio/driver.h:284: warning: Function parameter or member 'parent_handler_data_array' not described in 'gpio_irq_chip'
> > > > > 
> > > > > Introduced by commit
> > > > > 
> > > > >   cfe6807d82e9 ("gpio: Allow per-parent interrupt data")
> > > > > 
> > > > > But may actually be a problem with the tool :-(        
> > > > 
> > > > I guess the tool doesn't like having two fields that are part of a
> > > > union documented together... Happy to tweak it if someone tells me how
> > > > this should be written.      
> > > 
> > > Yes, that's the case. See, when you do:
> > > 
> > > 	/**
> > > 	 * @parent_handler_data:
> > > 	 * @parent_handler_data_array:
> > > 	 *
> > > 	 * Data associated, and passed to, the handler for the parent
> > > 	 * interrupt. Can either be a single pointer if @per_parent_data
> > > 	 * is false, or an array of @num_parents pointers otherwise.  If
> > > 	 * @per_parent_data is true, @parent_handler_data_array cannot be
> > > 	 * NULL.
> > > 	 */
> > > 	union {
> > > 		void *parent_handler_data;
> > > 		void **parent_handler_data_array;
> > > 	};
> > > 
> > > The tool will understand it as an undocumented "parent_handler_data" and
> > > a documented "parent_handler_data_array".
> > > 
> > > It has to do that, as otherwise it won't get cases where people just adds a
> > > @foo: as a template but actually forgets to fill it.
> > > 
> > > The solution would be to add a description for both, e. g. something
> > > similar to:
> > > 
> > > 	/**
> > > 	 * @parent_handler_data:
> > > 	 *
> > > 	 * If @per_parent_data is false, contains a single pointer 
> > > 	 * with the data associated, and passed to, the handler for the 
> > > 	 * parent interrupt.
> > > 	 *
> > > 	 * @parent_handler_data_array:
> > > 	 *
> > > 	 * If @per_parent_data is true, it should contain an array of 
> > > 	 * @num_parents pointers with the data associated, and passed to,
> > > 	 * the handler for the parent interrupt. Cannot be NULL.
> > > 	 */    
> > 
> > I am still getting this warning.  
> 
> I am still getting this warning.

I am still getting this warning.

-- 
Cheers,
Stephen Rothwell

Attachment: pgpgReiv9VswQ.pgp
Description: OpenPGP digital signature