Re: [igt-dev] [PATCH i-g-t v7 2/7] lib/xe: Introduce Xe library

["Dixit, Ashutosh" <ashutosh.dixit@intel.com>]

On Thu, 09 Mar 2023 22:41:35 -0800, Mauro Carvalho Chehab wrote:
>
> On Thu, 09 Mar 2023 20:38:03 -0800
> "Dixit, Ashutosh" <ashutosh.dixit@intel.com> wrote:
>
> > On Thu, 09 Mar 2023 20:30:39 -0800, Dixit, Ashutosh wrote:
> > >
> > > On Tue, 07 Mar 2023 14:12:31 -0800, Zbigniew Kempczyński wrote:
> > > >
> > >
> > > Hi Zbyzsek,
> > >
> > > > +#define for_each_gt(__fd, __gt) \
> > > > +	for (__gt = 0; __gt < xe_number_gt(__fd); ++__gt)
> > >
> > > This for_each_gt conflicts with for_each_gt in our internal i915 tree
> > > (which were planning to post upstream soon). How do you want to resolve
> > > this?
> >
> > Can we can just do this?
> >
> > diff --git a/lib/xe/xe_query.h b/lib/xe/xe_query.h
> > index fe1343f6161..72ee6c9478b 100644
> > --- a/lib/xe/xe_query.h
> > +++ b/lib/xe/xe_query.h
> > @@ -66,6 +66,7 @@ struct xe_device {
> >  #define for_each_hw_engine_class(__class) \
> >         for (__class = 0; __class < DRM_XE_ENGINE_CLASS_COMPUTE + 1; \
> >              ++__class)
> > +#undef for_each_gt
> >  #define for_each_gt(__fd, __gt) \
> >         for (__gt = 0; __gt < xe_number_gt(__fd); ++__gt)
> >
> >
>
> My two cents there: this works, but it is painful if the i915-specific
> version is placed on a header used by both Xe and i915 drivers.
>
> I can see two possible solutions:
>
> 1. ensure that the i915-specific version is placed on a lib/i915*
>    header;
>
> 2. name such kind of macros after the name of the driver, e. g.:
>
>	sed s,for_each_gt,xe_for_each_gt, -i lib/xe/* tests/xe/*
>	sed s,for_each_gt,i915_for_each_gt, -i lib/i915/* tests/i915/*
>
> They're not mutually exclusive.

Thanks Mauro, yes we'll have to do something like the two options above.