Re: [PATCH v2 0/2] introduce isar-image class

[Jan Kiszka <jan.kiszka@siemens.com>]

On 31.10.18 07:10, Cedric Hombourger wrote:
> Reduce the previous patch series to only include the first two patches.
> The changes will allow users to easily create their own image recipes.
> Documentation was updated with instructions.
> 
> The second patch refactors do_rootfs() into smaller functions to reduce
> its complexity (one function == one purpose/goal). It was also changed
> to become a Python function so it may be extended by other classes with
> Python code.
> 
> Patches introducing ROOTFS_*_COMMAND are on hold. I will investigate use
> of e.g. IMAGE_TRANSIENT_PACKAGES to implement some of the features we are
> currently carrying in our own project layers.
> 
> Cedric Hombourger (2):
>    isar-image-base: introduce and use isar-image class
>    isar-image: refactor do_rootfs()
> 
>   doc/user_manual.md                                 | 33 ++++++++++
>   meta-isar/classes/isar-image.bbclass               | 72 ++++++++++++++++++++++
>   .../files => conf/distro}/debian-configscript.sh   |  0
>   .../files => conf/distro}/raspbian-configscript.sh |  0
>   meta-isar/recipes-core/images/isar-image-base.bb   | 57 +----------------
>   5 files changed, 108 insertions(+), 54 deletions(-)
>   create mode 100644 meta-isar/classes/isar-image.bbclass
>   rename meta-isar/{recipes-core/images/files => conf/distro}/debian-configscript.sh (100%)
>   rename meta-isar/{recipes-core/images/files => conf/distro}/raspbian-configscript.sh (100%)
> 

Before we move things in meta-isar around: I'd like to first of all get rid of 
the dependency on meta-isar for building typical images. Everything that is not 
of demonstration purpose (specific boards, example recipes etc.) should go to 
the core layer (meta). Therefore, we should not introduce further generic 
classes to meta-isar.

One background: I already ran into conflict with machine conf files in meta-isar 
that I wanted to replace with own configs in a downstream layer. While config 
overloading is possible, it's not as obvious as bbappend, nor does it work well 
in the presence of kas. So the better approach is avoiding it in the first place 
- e.g. by not pulling meta-isar.

Jan

-- 
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux