[U-Boot] [PATCH v2 0/8] fdt: Allow indicating a node is for U-Boot proper only

[Simon Glass]

Hi Patrick,

On Tue, 21 May 2019 at 10:07, Patrick DELAUNAY <patrick.delaunay at st.com> wrote:
>
> Hi Simon,
>
> I will reply for the serie
>
> > Hi Patrick,
> >
> > On Mon, 20 May 2019 at 07:00, Patrick Delaunay <patrick.delaunay at st.com>
> > wrote:
> > >
> > >
> > > Hi,
> > >
> > > I create this v2 serie with:
> > >
> > > 1/ documentation update for previous patch
> > >    [U-Boot,v2] dm: remove pre reloc properties in SPL and TPL device tree
> > >    http://patchwork.ozlabs.org/patch/1081155/
> > >
> > >    PS: Code is already merged in commit commit c7a88dae997f ("dm: remove
> > >        pre reloc properties in SPL and TPL device tree")
> > >        but not the documentation.
> > >
> > > 2/ missing part for (patch 1/2)
> > >    http://patchwork.ozlabs.org/project/uboot/list/?series=89846
> > >
> > > 3/ new tests for pre-reloc propertie in SPL as suggested by Simon
> > >    (http://patchwork.ozlabs.org/patch/1081155/#2156621)
> > >
> > > Tell me if is better to split the serie.
> >
> > Somehow this cover letter appears in a patch, hence some of my confusion. I see
> > what you are doing and it is a comprehensive approach.
> >
> > But please see my comments about comparing the .dtb file instead of making
> > sandbox print it out.
>
> I will change the test to only compare the device tree : it is more simple.
>
> => v3 in few days
>
> My first approach was complicated but it is to allow
> - check if sandbox SPL with devicetree and libfdt in working as is already done for platdata
> - check if sandbox SPL can start U-Boot in booth case
>   (by executing the simple test_000_version in ./py/tests/test_000_version.py)
> - split test and normal device tree (I move the "spl-tests" nodes in test.dts)

Yes, certainly this is useful and it does provide an end-to-end sanity check.

But if we do this I think it should be *in addition* to smaller test.

So could we start with the simpler, smaller test and then see how far
that gets us? I am not saying that the functional test is bad, but if
something goes wrong with the test, there are a lot of pieces to look
at to figure out what went wrong.

>
> But it is too complicated just the purpose of this test.
>
> NB: the executable "u-boot-spl" for sandbox_spl_defconfig already
>        include the devicetree information, with platdata.

Yes.

Regards,
Simon