[0/3] kbuild: Add wilddt function instead of listing dtbs

Message

Leonard Crestez Jan. 7, 2019, 7:31 p.m. UTC

The dts makefiles go through a lot of pointless churn when boards are
added. Many SOCs (such as imx) have very simple naming conventions for
all boards using a certain chip and board listings can be easily
collapsed using wildcards.

Add a "wilddt" function and use it for imx6/7/8 and layerscape. This can
be applied to many other soc families later.

Previously sent as RFC: https://lore.kernel.org/patchwork/patch/1022737/

Changes since RFC:
 * Split into 3-part series
 * Move the wilddt to Kbuild.include so that it's available everywhere
 * Use $(srctree)/$(src) instead of $(dtstree)
 * Also use wilddt in arm64/boot/dts/freescale

Series is against next-20190107, conflicts are to be expected as board
list keeps changing.

Leonard Crestez (3):
  kbuild: Add wilddt function
  ARM: dts: imx: Use wilddt function
  arm64: dts: freescale: Use wilddt function

 arch/arm/boot/dts/Makefile             | 201 ++-----------------------
 arch/arm64/boot/dts/freescale/Makefile |  25 +--
 scripts/Kbuild.include                 |   6 +
 3 files changed, 21 insertions(+), 211 deletions(-)

Comments

Leonard Crestez Jan. 21, 2019, 4:16 p.m. UTC | #1

On 1/7/2019 9:31 PM, Leonard Crestez wrote:
> The dts makefiles go through a lot of pointless churn when boards are
> added. Many SOCs (such as imx) have very simple naming conventions for
> all boards using a certain chip and board listings can be easily
> collapsed using wildcards.
> 
> Add a "wilddt" function and use it for imx6/7/8 and layerscape. This can
> be applied to many other soc families later.
> 
> Previously sent as RFC: https://lore.kernel.org/patchwork/patch/1022737/

Any feedback? This should be reviewed by kbuild and arm-soc

I think this is a worthwhile cleanup, are there any corner cases or more 
exotic setup I should try in order to validate this series?

--
Regards,
Leonard

Masahiro Yamada Jan. 24, 2019, 4:03 p.m. UTC | #2

On Tue, Jan 22, 2019 at 1:17 AM Leonard Crestez <[e-mail]> wrote:
>
> On 1/7/2019 9:31 PM, Leonard Crestez wrote:
> > The dts makefiles go through a lot of pointless churn when boards are
> > added. Many SOCs (such as imx) have very simple naming conventions for
> > all boards using a certain chip and board listings can be easily
> > collapsed using wildcards.
> >
> > Add a "wilddt" function and use it for imx6/7/8 and layerscape. This can
> > be applied to many other soc families later.
> >
> > Previously sent as RFC: https://lore.kernel.org/patchwork/patch/1022737/
>
> Any feedback? This should be reviewed by kbuild and arm-soc
>
> I think this is a worthwhile cleanup, are there any corner cases or more
> exotic setup I should try in order to validate this series?



Personally, I prefer explicit listing
because I can browse the Makefile to get a quick idea
of which boards are compiled.

I like this to be consistent.


 - Exploit wildcard for all platforms

    or

- Do not do this at all


But, I am pessimistic about the former
when I look at AT91 platform.

Leonard Crestez Jan. 30, 2019, 1:16 p.m. UTC | #3

On Fri, 2019-01-25 at 01:03 +0900, Masahiro Yamada wrote:
> On Tue, Jan 22, 2019 at 1:17 AM Leonard Crestez <[e-mail]> wrote:
> > 
> > On 1/7/2019 9:31 PM, Leonard Crestez wrote:
> > > The dts makefiles go through a lot of pointless churn when boards are
> > > added. Many SOCs (such as imx) have very simple naming conventions for
> > > all boards using a certain chip and board listings can be easily
> > > collapsed using wildcards.
> > > 
> > > Add a "wilddt" function and use it for imx6/7/8 and layerscape. This can
> > > be applied to many other soc families later.
> > 
> > Any feedback? This should be reviewed by kbuild and arm-soc
> 
> Personally, I prefer explicit listing
> because I can browse the Makefile to get a quick idea
> of which boards are compiled.

Or you can just ls soc-*.dts?

> I like this to be consistent.
> - Exploit wildcard for all platforms
> - Do not do this at all
> 
> But, I am pessimistic about the former
> when I look at AT91 platform.

I can try to resend in a form that uses a wildcard for most platforms
but skips those with very few SOCs or confusing conventions (like
at91).

--
Regards,
Leonard