[yocto-ab] YP Advisory Board: VOTE meta-ti BSP layer YP Compatible

[Otavio Salvador]

On Mon, Mar 5, 2018 at 8:36 PM, Denys Dmytriyenko <denis at denix.org> wrote:
> On Mon, Mar 05, 2018 at 07:19:30PM -0300, Otavio Salvador wrote:
>> On Mon, Mar 5, 2018 at 7:01 PM, Denys Dmytriyenko <denis at denix.org> wrote:
>> > On Mon, Mar 05, 2018 at 04:42:10PM -0500, Denys Dmytriyenko wrote:
>> >> I can push a workaround I mentioned below to meta-ti, but it won't completely
>> >> eliminate all the signature differences I mentioned - both DEFAULTTUNE and
>> >> eudev. We either need to fix the script or have a more detailed guidelines.
>> >
>> > And the workaround is now in.
>> > bitbake-diffsigs still reports differences in eudev do_install and
>> > do_populate_sysroot between TI and non-TI machines...
>> >
>> >
>> >> On Mon, Mar 05, 2018 at 01:34:00PM -0800, Jeff Osier-Mixon wrote:
>> >> > I assume this is a minor issue that can be worked out, so I am not
>> >> > suspending voting at this time. If anyone else can jump in to test, it
>> >> > would be appreciated.
>> >> >
>> >> > On Mon, Mar 5, 2018 at 1:30 PM, akuster <akuster at mvista.com> wrote:
>> >> >
>> >> > >
>> >> > > So what now? Is the voting on hold?
>> >> > >
>> >> > > -armin
>> >> > >
>> >> > >
>> >> > > On 03/05/2018 11:50 AM, Denys Dmytriyenko wrote:
>> >> > > > On Mon, Mar 05, 2018 at 04:25:20PM -0300, Otavio Salvador wrote:
>> >> > > >> On Mon, Mar 5, 2018 at 4:12 PM, akuster <akuster at mvista.com> wrote:
>> >> > > >>>
>> >> > > >>> On 03/05/2018 10:48 AM, Otavio Salvador wrote:
>> >> > > >>>
>> >> > > >>> On Mon, Mar 5, 2018 at 3:41 PM, Denys Dmytriyenko <denis at denix.org>
>> >> > > wrote:
>> >> > > >>>
>> >> > > >>> FWIW, just tried with distroless setup - only oe-core and meta-ti and
>> >> > > still
>> >> > > >>> get a PASS.
>> >> > > >>>
>> >> > > >>> Did you test with qemuarm?
>> >> > > >>>
>> >> > > >>>
>> >> > > >>> The current explanation in the dev-manual does not make that a
>> >> > > requirement
>> >> > > >>> for the Yocto-check-layer to be run. If it is unclear, then we need to
>> >> > > >>> update doc's or clarify things in the docs.
>> >> > > >>>
>> >> > > >>>  If the script is mis-behaving, then please open a bug.
>> >> > > >> A BSP layer cannot change other machines, so it implies that a non-BSP
>> >> > > >> machine must be tested against.
>> >> > > > So, this eudev.bbappend, you are so adamant about, adds couple files to
>> >> > > > SRC_URI, which are not patches, just extra rule files, so the source
>> >> > > doesn't
>> >> > > > change. They are also not installed by default, they are only installed
>> >> > > on TI
>> >> > > > machines...
>> >> > > >
>> >> > > > Ok, running bitbake-diffsigs manually to emulate your error condition, I
>> >> > > get
>> >> > > > this signature difference. Now, changing SRC_URI to only add those 2
>> >> > > files
>> >> > > > for TI machines, so for anything else it would pretty much result in an
>> >> > > empty
>> >> > > > bbappend. And guess what? It still reports the same signature difference!
>> >> > > >
>>
>> So the package must be MACHINE_ARCH; and it also needs fixing.
>
> I was actually trying to avoid that - it is the same for 20+ TI machines...
>
> Moreover, the script still does not complain for me on multiple machines from
> or-core (qemuarm) and meta-ti (beaglebone, etc.):

The design is wrong; instead of adding the rules on eudev itself, add
a new package (e.g: udev-rules-ti) and add this to machine default
recommends or depends. This way you avoid having it changing across
machines.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750