[yocto] [PATCH V4 06/10] update.py: update layers orderly

[Paul Eggleton]

Hi Robert,

On Thursday, 22 June 2017 1:04:23 PM CEST Robert Yang wrote:
> On 06/21/2017 10:21 PM, Paul Eggleton wrote:
> > On Tuesday, 13 June 2017 4:36:47 AM CEST Robert Yang wrote:
> >> --- a/layerindex/update_layer.py
> >> +++ b/layerindex/update_layer.py
> ...
> >> +def get_layer_var(config_data, var):
> >> +    collection = config_data.getVar('BBFILE_COLLECTIONS', True)
> >
> > This makes the assumption that the layer.conf only adds one collection
> > to BBFILE_COLLECTIONS. Probably a valid assumption 99.9% of the time these
> > days, but it is worth noting.
> 
> Should we check all BBFILE_COLLECTIONS, please ?

I would say let's split the value and use the first one, but if there is more 
than one then warn e.g. "Multiple collections found, handling first one (%s) 
only" but otherwise continue processing. That way at least we'll know if this 
situation occurs.

> >> +def explode_dep_versions2(bitbakepath, deps):
> >> +    bblib = bitbakepath + '/lib'
> >> +    if not bblib in sys.path:
> >> +        sys.path.insert(0, bblib)
> >> +    import bb.utils
> >> +    return bb.utils.explode_dep_versions2(deps)
> >
> > I'm not particularly happy with this (the wrapper nor calling bitbake code
> 
> update.py needs explode_dep_versions2 to get dependencies and then calculate
> them, I had tried to move the code into update_layer.py, and let
> update_layer.py print explode_dep_versions2() to update.py, but it is a
> OderedDic, and not easy to process it manually, so I leave the code in
> update.py.

Right, I understand the limitations and I don't really have a better 
alternative unfortunately, so we'll go with your current implementation. We 
may need to re-evaluate this in future at least structure-wise if we find the 
need to call other bitbake code in a similar way.

Cheers,
Paul


-- 

Paul Eggleton
Intel Open Source Technology Centre