[oe] forcing dependency on genext2fs-native

Denys Dmytriyenko denis at denix.org
Wed Nov 9 01:32:04 UTC 2011


On Tue, Nov 08, 2011 at 02:00:28PM -0600, Peter Bigot wrote:
> On Tue, Nov 8, 2011 at 1:23 PM, Tom Rini <tom.rini at gmail.com> wrote:
> > On Tue, Nov 8, 2011 at 11:25 AM, Peter Bigot <bigotp at acm.org> wrote:
> >> I've added a DEPENDS += "genext2fs-native" for a recipe in meta-ti,
> >> but the corresponding package is not getting built.  bitbake -DDD
> >> says:
> >>
> >> DEBUG: Added dependency genext2fs-native for
> >> /prj/oe/angstrom/sources/meta-ti/recipes-misc/images/cloud9-gfx-image.bb
> >> DEBUG: providers for genext2fs-native are: ['genext2fs-native']
> >> DEBUG: sorted providers for genext2fs-native are:
> >> ['virtual:native:/prj/oe/angstrom/sources/openembedded-core/meta/recipes-devtools/genext2fs/genext2fs_1.4.1.bb']
> >> DEBUG: adding virtual:native:/prj/oe/angstrom/sources/openembedded-core/meta/recipes-devtools/genext2fs/genext2fs_1.4.1.bb
> >> to satisfy genext2fs-native
> >>
> >> I'm guessing the "virtual:native:" means somebody thinks genext2fs
> >> should already be installed on the host.  It's not available on
> >> Fedora, so this decision is wrong.  I haven't been able to figure out
> >> what's causing this to happen.
> >
> > No, it means that the "virtual:native" (coming from BBCLASSEXTEND in
> > the recipe file) is being selected to provide "genext2fs-native"
> > (since in many cases we don't have a separate -native recipe, we just
> > use BBCLASSEXTEND).  What's probably happening is that you need
> > genext2fs to exist for a specific task and just "DEPENDS +=
> > genext2fs-native" really means "make do_configure depend on
> > genext2fs-native being in the sysroot".  Did you modify
> > cloud9-gfx-image?  It should already get the DEPENDS right for
> > populating the image.
> 
> The build had failed in meta-ti/classes/sdcard_image.bbclass which
> used genext2fs to build an sdimg image.
> 
> I had added DEPENDS += "genext2fs-native" to
> meta-ti/recipes-misc/images/ti-hw-bringup-image.bb since it was what
> inherited sdcard_image.
> 
> On further investigation cued by your response, it seems that the
> correct change is instead to add:
> 
> IMAGE_DEPENDS_sdimg = "genext2fs-native"
> 
> analogously to what's done for the other image types.  With that the
> bitbake of cloud9-gfx-image does the right thing.

Thanks for your report! Can you please submit a patch to meta-ti mailing list 
at meta-ti at yoctoproject.org
Thanks.

-- 
Denys





More information about the Openembedded-devel mailing list