On Mon, Oct 14, 2024 at 10:05 AM Jes Sorensen <jes@xxxxxxxxxxxxxxxxxx> wrote: > > On 10/14/24 10:48, Rob Herring wrote: > > On Mon, Oct 14, 2024 at 9:41 AM Jes Sorensen <jes@xxxxxxxxxxxxxxxxxx> wrote: > >> > >> On 10/11/24 18:46, Rob Herring wrote: > >>> On Fri, Oct 11, 2024 at 9:05 AM Jes Sorensen <jes.sorensen@xxxxxxxxx> wrote: > >>>> > >>>> From: Jes Sorensen <jes@xxxxxxxxxxxxxxxxxx> > >>>> > >>>> For Android it is common to use a prebuilt dtc, speficied via DTC=. In > >>>> this case building dtc as part of the kernel is not necessary, and even > >>>> unwanted to avoid mix and match between two different versions of dtc. > >>>> > >>>> Signed-off-by: Jes Sorensen <jes@xxxxxxxxxxxxxxxxxx> > >>>> --- > >>>> scripts/dtc/Makefile | 3 +++ > >>>> 1 file changed, 3 insertions(+) > >>> > >>> Well, this one is much simpler than a prior attempt[1] and may be > >>> acceptable. But I wonder if something is not handled here. > >>> > >>> I think lack of external fdtoverlay support makes this incomplete. > >> > >> So I was thinking about that. The only places I see in the tree that > >> reference fdtoverlay are scripts/make_fit.py, scripts/Makefile.dtbs, and > >> scripts/dtc/Makefile. I don't see anything calling make_fit.py. > > > > "make dtbs" will use fdtoverlay if there are any base+overlay targets > > which there are an increasing number of. > > > >> We can either exclude fdtoverlay from this, or assume that if one > >> explicitly sets prebuilts, fdtoverlay is there too. > > > > Currently, if you set DTC you'll get external dtc plus in-tree > > fdtoverlay. With your patch, you'll get external dtc plus a build > > error because $(objtree)/scripts/dtc/fdtoverlay does not exist. > > Neither case is ideal. > > Agree, what path calls fdtoverlay ? I didn't see anything calling that > script. It is the same magic used used to combine object files for a module. Anywhere you have "foo-dtbs = base.dtb overlay.dtbo \n dtbs-y += foo.dtb", foo.dtb will be built by applying the overlay with fdtoverlay. Rob
