mailarchive of the ptxdist mailing list
 help / color / mirror / Atom feed
From: Michael Olbrich <m.olbrich@pengutronix.de>
To: Alexander Dahl <ada@thorsis.com>
Cc: ptxdist@pengutronix.de
Subject: Re: [ptxdist] target package dependency to platform
Date: Wed, 1 Feb 2023 09:49:22 +0100	[thread overview]
Message-ID: <20230201084922.GM8218@pengutronix.de> (raw)
In-Reply-To: <2489528.Mdqamz9Hy9@ada>

Hi,

On Tue, Jan 31, 2023 at 04:33:00PM +0100, Alexander Dahl wrote:
> I'm stuck with a problem where I can not find an elegant solution for.  
> Let me describe.
> 
> I have a layered BSP.  The base layer has different platformconfigs for 
> hardware targets with different SoC architectures, let's say the 
> PTXCONF_PLATFORM is "pla" on one and "plb" on the other.  (I can not build the 
> whole stuff for one platform only, two platforms are strictly required.)
> 
> There is a more or less ordinary CMake based target application in the base 
> layer, let's say in "base/rules/myapp.{in,make}", because that application is 
> needed on both platforms.
> 
> The two upper layers select the different platforms with different toolchains 
> and both also enable "myapp".  (Yes, I could also have it all in a single 
> layer BSP with two working copies and building one platform per working copy.  
> Doesn't change the problem.)
> 
> myapp depends on multiple libraries, some of them hardware specific.  So for 
> platform "pla" it needs "libmyliba" and for platform "plb" it does not.
> 
> In "base/rules/myapp.make" I pass PTXCONF_PLATFORM as a build time option to 
> CMake like this:
> 
>     MYAPP_CONF_OPT += -DHARDWARE_NAME=$(PTXCONF_PLATFORM)
> 
> Depending on "HARDWARE_NAME" I call `find_package(myliba)` in "CMakeLists.txt" 
> of myapp.
> 
> I could do a "select LIBMYLIBA" in "base/rules/myapp.in" unconditionally, and 
> that would make the build successful.  In the BSP for platform "pla" CMake 
> would look for libmyliba and link and on the target it would be used at 
> runtime just fine.  In the BSP for platform "plb" CMake would not look for it, 
> would not complain, myapp would run happily on the target without using 
> libmyliba just fine.
> 
> However, libmyliba would be built for platform "plb" and it would be copied 
> over to "plb" target and take space, but never be used there.  Build time for 
> it would be completely wasted.  Even worse, user might find it on target and 
> wonder what it means, although it's useless on platform "plb".
> 
> I could _not_ add that "select LIBMYLIBA" and build would fail for target 
> "pla" dependending on package build order.  Also dependencies in opkg packages 
> would not be correct then.
> 
> What also does not work is this (in "base/rules/myapp.in"):
> 
>     select LIBMYLIBA if PLATFORM = "pla"
> 
> This is valid syntax from Kconfig point of view, but PLATFORM is just not 
> defined here at all.  I assume configs/ptxconfig and configs/platform-pla/
> platformconfig are somewhat distinct Kconfig spaces and one can not simply 
> reference to Kconfig options from the other side.
> 
> Interestingly I can do `ptxdist print PTXCONF_PLATFORM` and `ptxdist print 
> PTXCONF_MYAPP` and ptxdist prints variables from both Kconfig spaces.  Also I 
> can successfully use both in base/rules/myapp.make and that's at least 
> somewhat misleading. (Misleading as in: Why can I use both here but not 
> there?)
> 
> That's where I'm stuck.  Of course I looked through ptxdist git sources to 
> find hints on how to solve this dilemma, but without luck.
> 
> Would be happy if anyone could give a hint or suggest a solution.

You can add the dependency in just one layer:
Either copy myapp.in to rules/myapp.in in that layer and add the dependency
or (if that file is sufficiently complex), create rules/myapp.fixup.in
with:
-------------------------------------------
## SECTION=....

config MYAPP
	select LIBMYLIBA
-------------------------------------------
Kconfig will basically just merge the two entries internally.

Then run "ptxdist oldconfig" in that layer. You'll get a modified ptxconfig
and a ptxconfig.diff with just PTXCONF_LIBMYLIBA=y (and comments).
You'll need to run oldconfig whenever the base ptxconfig changes. PTXdist
will complain if they are out of sync.

This is one of the use-cases I had in mind when I designed the layers in
PTXdist.

Michael

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |



      reply	other threads:[~2023-02-01  8:49 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-01-31 15:33 Alexander Dahl
2023-02-01  8:49 ` Michael Olbrich [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230201084922.GM8218@pengutronix.de \
    --to=m.olbrich@pengutronix.de \
    --cc=ada@thorsis.com \
    --cc=ptxdist@pengutronix.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox