[oe] bitbake -b with BBCLASSEXTENDed recipes Was: [RFC} glib-2.0 cleanup

[Martin Jansa]

On Mon, Mar 01, 2010 at 03:43:12PM +0100, Dr. Michael Lauer wrote:
> > And I want to reiterate the importance of new-style staging and
> > leveraging that for BBCLASSEXTEND.
> 
> Sure. That's certainly a good step, I just think it's not enough.

Is it possible to implement some support for old functionality like
bitbake -c clean -b path.to.recipe.with.BBCLASSEXTEND.bb

Now it fails with error like this:
bitbake at jama ~/build.dev.shr.gta $ bitbake -c clean -b ../dev/recipes/libxml/libxml2_2.7.6.bb
ERROR: Parsing error data_fn virtual:native:/OE/dev/recipes/libxml/libxml2_2.7.6.bb and fn /OE/dev/recipes/libxml/libxml2_2.7.6.bb don't match
Traceback (most recent call last):
  File "/usr/bin/bitbake", line 143, in <module>
    main()
  File "/usr/bin/bitbake", line 140, in main
    cooker.cook()
  File "/usr/lib64/python2.6/site-packages/bb/cooker.py", line 608, in cook
    if not self.buildFile(self.configuration.buildfile):
  File "/usr/lib64/python2.6/site-packages/bb/cooker.py", line 500, in buildFile
    taskdata.add_provider(self.configuration.data, self.status, item)
  File "/usr/lib64/python2.6/site-packages/bb/taskdata.py", line 340, in add_provider
    self.add_provider_internal(cfgData, dataCache, item)
  File "/usr/lib64/python2.6/site-packages/bb/taskdata.py", line 376, in add_provider_internal
    eligible, foundUnique = bb.providers.filterProviders(all_p, item, cfgData, dataCache)
  File "/usr/lib64/python2.6/site-packages/bb/providers.py", line 226, in filterProviders
    eligible = _filterProviders(providers, item, cfgData, dataCache)
  File "/usr/lib64/python2.6/site-packages/bb/providers.py", line 188, in _filterProviders
    sortpkg_pn[pn] = sortPriorities(pn, dataCache, pkg_pn)
  File "/usr/lib64/python2.6/site-packages/bb/providers.py", line 46, in sortPriorities
    priority = dataCache.bbfile_priority[f]
KeyError: 'virtual:native:/OE/dev/recipes/libxml/libxml2_2.7.6.bb'

-b was usefull for quick rebuild of modified recipe (ie to rebuild it in workdir
after updating for new staging or BBCLASSEXTEND to see if everything is "the same")

It's really hard to implement something like prefixing -b path with "virtual:native:"
for native variant and use old behavior (clean non-native version) without virtual: prefix)?
Or just nobody cared yet? In this case I would check what can be done.

Regards,

-- 
uin:136542059                jid:Martin.Jansa at gmail.com
Jansa Martin                 sip:jamasip at voip.wengo.fr 
JaMa