[bitbake-devel] [PATCH] cooker.py: allow dangling bbappends if explicitly whitelisted

[Patrick Ohly]

On Thu, 2017-05-25 at 22:01 +0200, Patrick Ohly wrote:
> On Thu, 2017-05-25 at 08:50 -0700, Christopher Larson wrote:
> > 
> > On Mon, May 22, 2017 at 3:50 AM, Patrick Ohly <patrick.ohly at intel.com>
> > wrote:
> >         Having a .bbappend without corresponding .bb file triggers an
> >         error or
> >         at least warning, depending on the global
> >         BB_DANGLINGAPPENDS_WARNONLY.
> >         
> >         Some layers (for example, meta-freescale) avoid that message
> >         by only
> >         adding .bbappends to the BBFILES when the layers they apply to
> >         are
> >         present. Others (like intel-iot-refkit) avoid such .bbappends
> >         by
> >         falling back to global assignments with _pn-<recipe> as
> >         override. Both
> >         is complicated.
> >         
> >         Now the warning code checks BBAPPENDS_DANGLING_WHITELIST and
> >         ignores
> >         all bbappends which match a file pattern in that list. This is
> >         an
> >         easier way to have bbappends which may or may not apply to an
> >         existing recipe.
> > 
> > IMHO this makes it too easy to miss legitimate problems. By adding
> > them only when the layers are present, you ensure that if that layer
> > removes the recipe you’re appending, you’ll immediately know it. This
> > hides such actual problems by suppressing for all optional appends
> > across the board.
> 
> I understand that it is a double-edged sword. It's meant for layer
> maintainers who know what they are doing. If a layer maintainer prefers
> to be warned about missing recipes, they can simply ignore the whitelist
> feature. Layer maintainers who use it need to ensure that they have
> tests in place which will fail when the bbappend is no longer used.

Also note that the alternative solution currently used by layers like
meta-freescale is brittle (depends on ordering of entries in
bblayers.conf). This is not just a theoretic problem; there's currently
a mail thread on the Yocto mailing list ("dynamic-layers") where someone
had the order wrong and had to spend time on debugging the problem.

-- 
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.