Re: [PATCH] usbinit, weston-init, run-postinsts, qt-demo-init: Drop allarch

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

On Mon, 2014-06-23 at 14:44 +0200, Martin Jansa wrote:
> On Mon, Jun 23, 2014 at 02:39:22PM +0200, Martin Jansa wrote:
> > On Mon, Jun 23, 2014 at 01:12:25PM +0100, Burton, Ross wrote:
> > > On 19 June 2014 14:54, Paul Eggleton <paul.eggleton@linux.intel.com> wrote:
> > > >> * update-rc.d.bbclass now adds dependency on TUNE_PKGARCH initscripts, so
> > > >> these cannot be allarch
> > > >
> > > > As I've said before, I really dislike this change. The system should be made
> > > > to handle the situation properly instead of working around the issue.
> > > 
> > > Isn't this what the abi exclude thingy is for?
> > 
> > Do we want to maintain list of all recipes which inherit update-rc.c in
> > layer.conf?
> > 
> > That's what I did in one of our layers and the list is rather long.
> 
> Ah sorry for possible confusion, I had to list *all* of them because we
> have MACHINE_ARCH initscripts.
> 
> To fix current issues in oe-core we only need to list all recipes which
> inherit allarch *and* update-rc.d (unless we want to exclude it
> everywhere just for consistency).
> 
> > abi exclude thingy
> 
> is a bit dangerous, I still sometimes see issues when something is
> supposed to be abisafe and in some cases it isn't and I need to manually
> bump PR in the recipe which wasn't rebuilt.
> 
> So I'm trying to use SIGGEN_EXCLUDE_SAFE_RECIPE_DEPS only in cases where
> it's really safe or when the dependent recipe takes forever to build (so
> I wan't to build it just once or once per TUNE_PKGARCH). For simple and
> quickly built recipe I really don't mind "building" them multiple times
> if it prevents "rebuilding" them every single time I change MACHINE.

Can't we just add "initscripts" to the list there. There shouldn't be a
need to rebuild initscripts any time something depending on it changes,
it should have a safe ABI in that sense?

Cheers,

Richard