RE: [AUH][PATCH v2 1/9] upgrade-helper.py: Add compatibility with Yocto scarthgap

> -----Original Message-----
> From: Alexander Kanavin <[email protected]>
> Sent: Monday, 27 April 2026 13:02
> To: Daniel Turull <[email protected]>
> Cc: [email protected]; [email protected];
> [email protected]; [email protected]
> Subject: Re: [AUH][PATCH v2 1/9] upgrade-helper.py: Add compatibility with
> Yocto scarthgap
> 
> On Fri, 24 Apr 2026 at 13:46, <[email protected]> wrote:
> > -        pkg_ctx['recipe_dir'] = os.path.dirname(pkg_ctx['env']['FILE'])
> > +        pkg_ctx['recipe_dir'] =
> > + os.path.realpath(os.path.dirname(pkg_ctx['env']['FILE']))
> 
> This needs to be further explained.

In our layout, meta is a symlink (meta -> openembedded-core/meta), so pkg_ctx['env']['FILE'] returns a path through the symlink. Later, git add on that path fails because the git repo is at the resolved openembedded-core/ path. I'll expand the commit message to explain this.
> 
> > -        bb.process.run("bitbake-config-build enable-fragment
> machine/{}".format(machine))
> > -        return self._cmd(recipe, env_var=env)
> > +        try:
> > +            bb.process.run("bitbake-config-build enable-fragment
> machine/{}".format(machine))
> > +        except bb.process.ExecutionError:
> > +            # bitbake-config-build not available (e.g. scarthgap), use MACHINE
> env var
> > +            env = "MACHINE={} {}".format(machine, env).strip()
> > +        return self._cmd(recipe, env_var=env if env else None)
> 
> The condition needs to be 'bitbake-config-build doesn't exist', not that it
> failed to execute (e.g. returned a non-zero).

I'll change it to check shutil.which('bitbake-config-build') instead of catching ExecutionError.

> 
> >              for group in pkggroups:
> > +
> > +                # Scarthgap returns flat tuples; normalize to list-of-dicts
> > +                # so the existing loop handles both formats.
> > +                if not isinstance(group, (list, tuple)) or not group or not
> isinstance(group[0], dict):
> > +                    pn, status, cur_ver, next_ver, maintainer, revision,
> no_upgrade_reason = group
> > +                    group = [{'pn': pn, 'status': status, 'cur_ver': cur_ver,
> > +                              'next_ver': next_ver, 'maintainer': maintainer,
> > +                              'revision': revision,
> > + 'no_upgrade_reason': no_upgrade_reason}]
> 
> Can the condition be inverted, e.g. be checked for the 'old' format?
> I'm not sure I understand why 'old' format is determined by any of three
> different conditions being true, especially 'not group' looks incorrect.

Sure, Scarthgap's _get_recipe_upgrade_status returns a flat tuple of 7 elements, while master returns a list of dicts. I'll simplify it.
> 
> Alex

Thanks for the feedback. I'll updated in v3

Daniel