> On 27.1.2015, at 21.49, Burton, Ross <ross.bur...@intel.com> wrote:
>
> There really isn't - systemd has a newer udev that had an ABI change, so the
> version was increased.
>
> There's no way that your tmp/deploy/rpms can contain both systemd and udev
> packages (specifically, libudev1 and libudev0) unless you don't delete it
> when enabling the systemd DISTRO_FEATURE. The systemd and udev packages
> conflict with each other and they both can't be built at the same time.
>
> A copy of the cooker log for the image would be useful after you've deleted
> tmp and -ccleansstate's google-chrome if you are certain that this is what
> has happened.
>
Ross,
yeah, you were right - I mistook the existing libudev and udev rpms not to be
from systemd before I checked the contents (they DO contain only udevstuff
"generated" by systemd). So the image build itself was ok all the time, but
libudev.so.0 just doesn't exist when using systemd.
However, about my actual problem: I managed to package Google Chrome
successfully by adding SKIP_FILEDEPS_${PN} = "1" and by creating a symlink
/opt/google/chrome/libudev.so.0 -> ../../../lib/libudev.so.1 in do_install. A
bit dirty, but this is what the official deb from Google seems to do to manage
the different versions. Also skipping file deps inspection is not clearly an
optimal solution, but works for me in this case.
Thanks for your help,
Timo
--
_______________________________________________
yocto mailing list
yocto@yoctoproject.org
https://lists.yoctoproject.org/listinfo/yocto
