Re: [PATCH] packagekit: Updated to 0.8.13

[Paul Eggleton <paul.eggleton@linux.intel.com>]

On Thursday 28 November 2013 11:49:18 Koen Kooi wrote:
> Paul Eggleton schreef op 28-11-13 11:32:
> > Hi Koen / Felipe,
> > 
> > On Thursday 28 November 2013 10:22:31 Koen Kooi wrote:
> >> Felipe F. Tonello schreef op 28-11-13 01:56:
> >>> From: "Felipe F. Tonello" <eu@felipetonello.com> This recipe supports
> >>> the backend for packagekit dynamically based on the IMAGE_PKGTYPE.
> >> 
> >> NAK! IMAGE_FEATURES should *never* change non-image recipe params.
> >> This breaks using feeds horribly.
> > 
> > IMAGE_PKGTYPE is influenced by PACKAGE_CLASSES; so this is not about
> > IMAGE_FEATURES, and correct me if I'm wrong but maintaining package feeds
> > 
> >  would generally preclude switching to an alternative package manager,
> > 
> > right?
> 
> No, it's perfectly possible to build both opkg and rpm, which is what I'm
> currently doing. When doing that the DISTRO.conf does need to make a
> decision on what to support for things like packagekit.
> 
> > Some options:
> > 
> > 1) Apply the patch as-is. Changing the order/value of PACKAGE_CLASSES
> > will mean this and anything that depends upon it will rebuild.
> > 
> > 2) Install the appropriate backend via some code in the image recipe.
> > Obviously this means you have to do this for every image recipe though.
> > 
> > 3) Use non-dynamic PACKAGECONFIG. Of course this means you'll have to
> > remember to change this manually if you change PACKAGE_CLASSES or it'll
> > just be broken at runtime.
> > 
> > Honestly, option 1 sounds like the best course to me here. This is rather
> > a special case compared to other recipes.
> 
> 1) will let you end up with packagekit_1.0.ipk that only supports RPM

Correct, it would. I agree that's not ideal. Neither is having it broken by 
default for some people though (unless you just set all backends to on by 
default, that is.)

> 2) Is what we would really want, but I don't think packagekit supports that
> :(

This patch looks like it is splitting out the backends into separate 
packages...
 
> So that leaves 3, which makes it a clear DISTRO decision, like it should be.

It's worth pointing out that the patch sets PACKAGECONFIG with ??= so it 
doesn't stop you from doing this.

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre