Re: [PATCH 0/2] psplash fixes

[Martin Jansa <martin.jansa@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2935 bytes --]

On Mon, Dec 19, 2011 at 07:48:48PM +0100, Koen Kooi wrote:
> 
> Op 19 dec. 2011, om 19:32 heeft Paul Eggleton het volgende geschreven:
> 
> > On Monday 19 December 2011 19:27:06 Koen Kooi wrote:
> >> Op 19 dec. 2011, om 19:22 heeft Paul Eggleton het volgende geschreven:
> >>> On Monday 19 December 2011 18:50:35 Koen Kooi wrote:
> >>>> Op 19 dec. 2011, om 18:43 heeft Paul Eggleton het volgende geschreven:
> >>>>> Set the default psplash image to the OpenEmbedded logo, and provide
> >>>>> a
> >>>>> script to allow people to use their own custom image.
> >>>> 
> >>>> What I did in OE-classic and meta-angstrom is to use
> >>>> update-alternatives to provide different psplash images. This way you
> >>>> can choose a different splash for each image instead of having a
> >>>> distro wide one. Is something like that suitable for oe-core as well?
> >>> 
> >>> Sounds like a useful capability, however, does this mean that when you
> >>> want to override it in the image you end up with both psplash versions
> >>> installed?
> >> in a splashless image you can just do 'IMAGE_INSTALL += psplash-angstrom and
> >> it will only install that one. If you want to reuse an existing, unmodified
> >> image with psplash and add your own, then you will end up with both.
> > 
> > Although I guess another way to do it would be to do a VIRTUAL-RUNTIME type 
> > thing like we do for other such selections. At the moment in OE-core, psplash 
> > is brought in via task-core-console, and whilst it is a separate variable that 
> > could be overridden it's a task which once built really makes it impossible to 
> > customise per-image.
> > 
> > What about a "splash" IMAGE_FEATURE and then a VIRTUAL-RUNTIME_splash to 
> > select which psplash to install?
> 
> I will say again: I absolutely HATE that virtual-runtime nonsense. If you need to change a task, change the task, don't introduce things that make it non deterministic. Guess what happens when you change a virtual-runtime *after* you have built the task already.

The advantage of virtual-runtime is that something like initscripts, is
pulled by more than one recipe, so in order to change initscripts pulled
to distro images you have to change 3-4 recipes by .bbappend and bump PR
in all.

With VIRTUAL-RUNTIME you have to select VIRTUAL-RUNTIME_initscripts by
distro config and add .bbappends only with PRINC bumps for recipes using
this variable (easy to find with git grep).
And if someone else adds new recipe also using initscripts then he
should use right one wrt VIRTUAL-RUNTIME_initscripts not expecting you
to notice that other recipe pulling initscripts is there (I've used
BLACKLIST to notice those before, but VIRTUAL-RUNTIME is better).

And it's like DISTRO_FEATURES, nobody should change VIRTUAL-RUNTIME
settings without at least PRINC bumps.

Cheers,

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 205 bytes --]