[OE-core] [PATCH] gtk-doc.bbclass: unbreak dependency-chain for native recipes

[Chris Larson]

On Thu, Jul 26, 2012 at 8:35 AM, Andreas Müller
<schnitzeltony at googlemail.com> wrote:
> On Thu, Jul 26, 2012 at 5:19 PM, Burton, Ross <ross.burton at intel.com> wrote:
>> On 26 July 2012 16:16, Andreas Müller <schnitzeltony at googlemail.com> wrote:
>>> -DEPENDS_virtclass-native_append = " gtk-doc-stub-native"
>>> +DEPENDS_append_virtclass-native = " gtk-doc-stub-native"
>>
>> Was this a typo on my part, or is there a semantic difference between those?
>>
>> Ross
>>
> It is not an answer to your question but
>
> I grepped for DEPENDS_virtclass-native_append in oe-core and did find
> one occurrence in gtk-doc.bbclass. DEPENDS_append_virtclass-native I
> found in xcb-proto.inc so I gave it a try by
>
> bitbake -e atk-native | grep ^DEPENDS

DEPENDS = "alpha"
DEPENDS_virtclass-native_append = "foo"

End result for native: the latter is applied, resulting in a
'DEPENDS_virtclass-native' whose contents are just 'foo'. Then
DEPENDS_virtclass-native overwrites DEPENDS, resulting in the main
variable only containing the bits you appended, not the original
contents.

To sum up, _<override>_append appends to the override, which then
overwrites the original, whereas _append_<override> appends directly
to the main variable, conditionally on the override.

I'm not sure if that clears it up.. I've never been good at describing
this. For reference, I'm pretty sure we put something in the bitbake
manual about this circumstance specifically, since we chose to leave
it this way to retain maximum flexibility when writing the code, but
realized that it was a potential source of confusion.
-- 
Christopher Larson