Re: [PATCH 16/18] mtd: rawnand.h: use nested union kernel-doc markups

[Boris Brezillon]

On Mon, 7 May 2018 08:32:32 -0300
Mauro Carvalho Chehab <mchehab+samsung@xxxxxxxxxx> wrote:

> Hi Boris,
> 
> Em Mon, 7 May 2018 11:46:50 +0200
> Boris Brezillon <boris.brezillon@xxxxxxxxxxx> escreveu:
> 
> > Hi Mauro,  
> 
> > > diff --git a/include/linux/mtd/rawnand.h b/include/linux/mtd/rawnand.h
> > > index 5dad59b31244..b986f94906df 100644
> > > --- a/include/linux/mtd/rawnand.h
> > > +++ b/include/linux/mtd/rawnand.h
> > > @@ -740,8 +740,9 @@ enum nand_data_interface_type {
> > >  
> > >  /**
> > >   * struct nand_data_interface - NAND interface timing
> > > - * @type:	type of the timing
> > > - * @timings:	The timing, type according to @type
> > > + * @type:	 type of the timing
> > > + * @timings:	 The timing, type according to @type
> > > + * @timings.sdr: Use it when @type is %NAND_SDR_IFACE.    
> > 
> > Hm, really feels weird to do that. I mean, either we describe
> > timings.sdr or timings, but not both. I this case, I agree that
> > describing timings.sdr would make more sense than generically
> > describing any possible entries in the timings union.  
> 
> This struct is funny, as the union has just one item. I assume
> that the original intend is to be ready to have more timing
> types (or you had it in the past). By the time you have a
> second value there, describing the union would make more
> sense.
> 
> > Is there a simple
> > way we can get rid of the warning we have when not describing timings
> > but all of its fields?  
> 
> There's no direct way. It won't be hard to add a way to do it,
> like applying the enclosed patch with ends by declaring timings as:
> 
> 	* @timings:	 -- undescribed --
> 
> IMHO, this is uglier.

Yep, don't like it either. I'll just take your initial patch.

Thanks,

Boris