Re: RFC: Override expansion - changing the behaviour

[Mark Hatle <mark.hatle@windriver.com>]

On 6/7/11 5:10 PM, Joshua Lock wrote:
> On Wed, 2011-06-01 at 00:00 +0100, Richard Purdie wrote:
>> Currently if you do:
>>
>> OVERRIDES = "z"
>> DEPENDS_prepend = "a "
>> DEPENDS = "b"
>> DEPENDS_z = "c"
>>
>> d.update_data()
>> d.getVar("DEPENDS") gives "a c"
>> d.update_data()
>> d.getVar("DEPENDS") gives "c"
>>
>> I'm proposing we should change the bitbake behaviour such that at the
>> time bitbake expands the DEPENDS_z override, it removes "DEPENDS_z" from
>> the data store.
> 
> FWIW I think this change is absolutely the right thing to do. The
> current behaviour always catches me out (and I hope I'm not alone) and
> changing it as suggested makes the behaviour much less surprising.
> 
> Cheers,
> Joshua

We will likely need to audit the classes.  I've seen places (in the past, they
may all be gone now) where we've got things like getVar("DEPENDS"), and later
getVar("DEPENDS_${PN}"... where PN is the override...

But more consistent behavior is certainly better..  (Maybe simply clearing the
OVERRIDES, once its been applied is an alternative?)

--Mark