[yocto] Postinstallation script warnings in Yocto 2.4 Rocko

[Diaz de Grenu, Jose]

-----Original Message-----
From: Alexander Kanavin [mailto:alexander.kanavin at linux.intel.com] 
Sent: Monday, January 8, 2018 1:20 PM

On 01/08/2018 01:44 PM, Diaz de Grenu, Jose wrote:

> > Thanks for the information. I guess this fix will be for Yocto 2.5?

> Maybe; I'm working right now on the patchset, and will post it to oe-core list later today. Last time it was a bit controversial, so I'd appreciate if you a) try and test it; b) express your
> support for the changes ;)

> Work in progress here:

> https://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/log/?h=akanavin/dnf-rpm4-postinst-fix

Unfortunately I am not able to follow the conversation in oe-core, as I subscribed after you sent the patchset. 

I have tested your patchset by cherry-picking the commits to the rocko branch. There was only a minor conflict. With those changes I get a message like this:

WARNING: core-image-base-1.0-r0 do_rootfs: The postinstalls for the following packages will be postponed for first boot: init-ifupdown u-boot-fw-utils

The patches work, but I was wondering if we really need to mark this behavior as a WARNING. From my point of view, this may be confusing for a user, as it could lead to think that something is wrong. But I think that, even if image post processing during the build is preferable, there are some image post processing that just cannot be done there. Think for example of any HW identifier, like a MAC address that you might want to use.

I think a bb.note should be enough. At least an option per recipe to skip the warning, so that expected on-target-postprocessing warnings can be hidden.

Also, is there any chance this could end up in rocky? As it is, any postprocessing for the target causes ugly WARNINGS with '-1 error code' messages.

Thanks again for the patches!