Hi Raj,
My apologies for the delayed reply.
On Fri, Aug 31, 2018 at 11:39:54PM +0000, Mani, Rajmohan wrote:
...
> > > +struct ipu3_uapi_af_meta_data {
> > > + __u8 y_table[IPU3_UAPI_AF_Y_TABLE_MAX_SIZE] IPU3_ALIGN;
> >
> > Here IPU3_ALIGN is put at the end...
> >
> > > +} __packed;
> > > +
> > > +/**
> > > + * struct ipu3_uapi_af_raw_buffer - AF raw buffer
> > > + *
> > > + * @meta_data: raw buffer &ipu3_uapi_af_meta_data for auto focus meta
> > data.
> > > + */
> > > +struct ipu3_uapi_af_raw_buffer {
> > > + IPU3_ALIGN struct ipu3_uapi_af_meta_data meta_data;
> >
> > ... and here at the start. Is that due to the difference between an array and a
> > struct?
> >
>
> No.
>
> When preparing uAPI kernelDoc using "make htmldocs",
> the kernel-doc encounters two type of error/warnings
> caused by IPU3_ALIGN.
>
> case 1:
> struct IPU3_ALIGN ipu3_uapi_dummy {
> ...
> } __packed;
>
> "error: Cannot parse struct or union!"
>
> case 2:
> struct ipu3_uapi_dummy {
> struct ipu3_uapi_x x IPU3_ALIGN;
> } __packed;
>
> "warning: Function parameter or member 'IPU3_ALIGN' not
> described in 'ipu3_uapi_dummy'"
>
> Positioned the attribute syntax without altering the
> mem layout of the structs, while also making "make htmldocs" to
> compile fine.
>
> Let us know if it's okay to ignore Sphinx warnings.
I looked a bit at different options for handling this in scripts/kernel-doc
but the difficulty in macro substitution comes in determining where to do
the substitution and where not to. That seems unaddressable in the kernel-doc
script; most of the time you want the definitions as-is while this is
likely the only case where something else might be appropriate. Making
IPU3_ALIGN a special case probably wouldn't really fly either.
In this particular case I'd just write open the alignment requirement so
kernel-doc can correctly parse it.
--
Kind regards,
Sakari Ailus
sakari.ailus@xxxxxxxxxxxxxxx
