[OE-core] DEPENDS_virtclass-native = "libpng jpeg" not extended to libpng-native jpeg-native

[Richard Purdie]

On Tue, 2012-09-11 at 10:03 +0200, Martin Jansa wrote:
> Hi,
> 
> in meta-openembedded/meta-oe/recipes-extended/libwmf/libwmf_0.2.8.4.bb:
> 
> DEPENDS_virtclass-native = "libpng jpeg"
> DEPENDS = "libpng jpeg expat gtk+"
> 
> BBCLASSEXTEND = "native"
> 
> Did it work (at least at some point of time) that DEPENDS for
> libwmf-native were expanded to libpng-native and jpeg-native?

I don't think that has ever worked. Its assumed that if you're going to
use class overrides, you put the right thing in place.

> Because now it does not:
> OE tuna at shr ~/shr-core $ bitbake-diffsigs
> stamps.1347348593/nokia900/x86_64-linux/libwmf-native-0.2.8.4-r1.do_configure.sigdata.315a83efebb27040d6bf0aaead16671e
> stamps.1347348593/om-gta02/x86_64-linux/libwmf-native-0.2.8.4-r1.do_configure.sigdata.0f8349ada0c8a18a6d6ed7b7841ec955
> Hash for dependent task libpng_1.2.49.bb.do_populate_sysroot changed
> from 640001ee0530e51ef0aefb300c34f7dd to 7ba7d74a27e1e0af253ddac00a2be1c2
> Hash for dependent task libjpeg-turbo_svn.bb.do_populate_sysroot changed
> from 3fe6eae3a6fd1af40233d548680c5bab to ff852ac3d5e826ff74e68b5ddc4ac3e1
> 
> So native recipe depending on target checksum -> rebuilding with each
> MACHINE switch.
> 
> I can fix it by:
> DEPENDS_virtclass-native = "libpng-native jpeg-native"
> but maybe there is more cases like this and this could be checked by
> native.bbclass or something like that.

That is the correct thing to do, the code assumes when you write
explicit class overrides, you mean what you say...

Cheers,

Richard