[Buildroot] [PATCH v3 0/7] qt5 bump to 5.8.0-rc

[Thomas Petazzoni <thomas.petazzoni@free-electrons.com>]

Hello,

On Wed, 21 Dec 2016 10:13:27 +0100, Andreas Naumann wrote:

> One idea i just had has to do with Peters question how Yocto deals with 
> it. Their meta-qt5 layer has branches for all the past Yocto releases 
> that are maintained. Each of them has a fixed qt-version. So in that 
> regard is not much different from our situation.
> However, if you want to use an older Qt-Version with a newer 
> Yocto-Version it may be a matter of checking out a different meta-qt5 
> branch.
> 
> With that in mind I thought of our br2-external trees, which we now can 
> have multiple of. There would probably be some way to combine mainline 
> buildroot and an older version of the qt5 packages as external, but lots 
> of questions arise here also. Variable names, additional repos, 
> maintenance...

That just doesn't work in Buildroot: you can have a single package with
a given name. If your different packages for the different Qt versions
have different names, then it's a nightmare because all packages that
depend on Qt need to handle the different possibilities.

> So no, I dont think thats a good proposal. But all other solutions I can 
> think of involve some kind of ugly code duplication, increased 
> complexity (ifeq...) and so on. And maintenance will have to be done 
> regardless of the implementation.
> So with that in mind, I turn to my initial thought of last week. If 
> people/companies need maintenance of older versions, why not maintain 
> the buildroot release that has them? With IoT and the need to for long 
> term support, combined efforts for maintenance would be rather helpful, 
> right?

We discussed this with Peter, and we would like for now to support both
versions within the same package, with a "choice ... endchoice" that
allows to choose between the LTS release and the latest release.

If you look at your patch switching to 5.8 there is in fact very, very
little change in the .mk files, so that can be handled with a bunch of
ifeq () conditions.

Thanks!

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com