Re: [PATCHv2] yocto-compat-layer.py: Add script to YP Compatible Layer validation

[Patrick Ohly <patrick.ohly@intel.com>]

On Tue, 2017-02-28 at 14:33 -0600, Aníbal Limón wrote:
> 
> On 02/28/2017 02:09 PM, Patrick Ohly wrote:
> > On Mon, 2017-02-20 at 15:12 -0600, Aníbal Limón wrote:
> >> common.test_signatures: Test executed in BSP and DISTRO layers to review
> >>     doesn't comes with recipes that changes the signatures.
> > 
> > I have a question about the goal for this test: is it meant to detect
> > layers which incorrectly change the signatures of allarch recipes or
> > recipes which share the same tune flags with other machines?
> 
> The requirement is DISTRO and BSP layers aren't allowed to automatically
> change the MACHINE or DISTRO variable that causes the signatures to change.

I do not fully understand this explanation. Can you give an example for
the kind of misbehavior what the test is meant to catch?

> > Let's take MACHINE=edison as an example.
> > 
> > Modifying allarch creates a conflict with basically all other machines
> > in a distro. Example for something not allowed:
> > VOLATILE_BINDS_append_pn-volatile-binds_edison = " /var/volatile/foo /var/foo \n"
> > 
> > This can be detected by comparing against OE-core, but only when testing
> > with MACHINE=edison.
> > 
> > More difficult to detect is modifying recipes with the same tune flags,
> > which is the majority of the recipes. MACHINE=edison and
> > MACHINE=intel-core2-32 both compile for the same target architecture, so
> > something like this is incorrect:
> > do_install_append_pn-base-files_edison () {
> >     echo "Built for Edison" >>${D}${sysconfdir}/motd
> > }
> > 
> > This can only be detected when testing with both MACHINE=edison and
> > MACHINE=intel-core2-32 - at least I think MACHINE=qemux86 uses different
> > tune flags (haven't checked).
> > 
> > My point is, the test probably needs to be extended to run with a set of
> > machines, and that set of machines must be broad enough to cover a
> > variety of common tune flags.
> 
> It is expected to change recipe signatures when the machine change, this
> leaves me other question. what signatures are expected to change when
> set a different MACHINE?

Everything that is machine-specific may change, for example image
recipes and kernel (although even that is a slight simplification, as
meta-intel intentionally shares kernels between different MACHINE
configs).

The real-world test is: can two BSP layers be combined in a single
distro so that one can build first for one MACHINE, then the other (or,
when using multiconfig, even in a single build)?  If any shared package
changes as result of changing the MACHINE, then that won't work.

-- 
Best Regards, Patrick Ohly

The content of this message is my personal opinion only and although
I am an employee of Intel, the statements I make here in no way
represent Intel's position on the issue, nor am I authorized to speak
on behalf of Intel on this matter.