On Wed, Sep 27, 2017 at 01:34:12PM +0200, Paolo Bonzini wrote: > On 27/09/2017 11:15, Daniel P. Berrange wrote: > > On Wed, Sep 27, 2017 at 09:19:22AM +0200, Paolo Bonzini wrote: > >> Are you planning to include only submodules, or also "QEMU-native" > >> firmware such as linuxboot, kvmvapic, s390-ccw, spapr-rtas, etc.? > > > > The submodules make sense to split out because distro vendors buld them > > independently of QEMU, and would rather not have them in the tarballs, > > so they have a clearer path to license compliance and legal export > > certification. > > > > The other bits of mention are all built normally as part of QEMU and > > not subject to these problems, so I don't see a benefit to splitting > > them out of QEMU. > > They aren't rebuilt in general. You end up with x86 builds of > qemu-system-x86 rebuilding linuxboot, ppc builds of qemu-system-ppc > rebuilding spapr-rtas, etc. (search configure for "roms="). In fact, > QEMU has a special exception in Fedora just because these are too hard > to untangle. > > So the advantage would be the ability to introduce better infrastructure > for cross compilation, without complicating further the QEMU build system.
Ah I see. > > > putting those bits in the qemu-firmware > > repo would re-introduce the problem we're trying to solve because > > distros would then need to get linuxboox, kvmvapi etc from a tarball > > of qemu-firmware which would once again include all the bits they > > don't want to have. > > This is true. We could distribute a qemu-firmware tarball with just the > QEMU-specific bits, and a qemu-firmware-all tarball with also those that > are built separately. Yep, as long as there's a tarball for the QEMU bits that does not contain the 3rd party bits, that would work. Regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
