Re: [PATCH 1/1] license.bbclass: Splitting out licenses

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Tue, 2011-07-26 at 08:15 -0700, Flanagan, Elizabeth wrote:
> On Tue, Jul 26, 2011 at 7:57 AM, Richard Purdie
> <richard.purdie@linuxfoundation.org> wrote:
> > On Mon, 2011-07-25 at 16:24 -0700, Flanagan, Elizabeth wrote:
> >> Adding a bit more functionality here:
> >> 1. Adding some more SPDX Maps to take care of + licenses
> >> 2. Strip out -native and -cross package license wrangling.
> >>    If it doesn't go on the image, we shouldn't wrangle it.
> >> 3. Split out the license destination directory to a
> >>    IMAGE_NAME time stamped dir in
> >>    /tmp/deploy/licenses/${IMAGE_NAME}/<stamp>
> >>
> >> I've removed the handler from my previous Pull as license
> >> manifest needs more discussion and I don't want these
> >> bug fixes to be held up by an added feature.
> >
> > I obviously don't understand this code :/
> >
> > What happens when I run "bitbake core-image-minimal core-image-sato",
> > i.e. when I build two images in one build?
> >
> > I suspect this current approach is flawed and we actually need to
> > postprocess the installed package list after do_rootfs completes at
> > image generation time to build the *real* list based on the installed
> > packages?
> >
> 
> Yes. I've found that this approach is definitely flawed. It works
> great for a single image build. Outside of that it acts funny, like
> you mentioned and can return incorrect results. I'm planning on
> revisting this soon.
> 
> I'm suspecting that your suggested approach is the way it'll have to
> be. Looking at what is generated by various runs and I see issues.
> 
> > This code is obviously still needed as it would provide the basis so the
> > code can get the licenses it needs to pull together...
> 
> Yes, agreed.

So with this in mind, where does that put this pull request? :)

Cheers,

Richard