"Michael S. Tsirkin" <m...@redhat.com> wrote: > On Wed, Nov 28, 2018 at 08:46:15AM +0100, Juan Quintela wrote: >> "Michael S. Tsirkin" <m...@redhat.com> wrote: >> > On Mon, Nov 26, 2018 at 08:59:53PM +0100, Juan Quintela wrote: >> >> Hi >> >> >> >> V2: >> >> >> >> - Rebase on top of master >> >> >> >> Please review. >> >> >> >> Later, Juan. >> > >> > And just making sure the point is so that one >> > can disable virtio-pci and still build virtio? >> > >> > And the point of that is mostly as a cleanup on >> > generic principles, right? Or is there a project that >> > actually wants to do it? >> >> This allows "make check" to work with them disabled. Without this >> changes, introspection fails, because we have the "name" (kind of) >> "virtio-pci-blk", but not the real virtio device, so it tries to use it >> while it is not there. See the equivalent change for virtio-ccw-*. >> >> Thanks, Juan. >> >> PD. Without this change, you can build a qemu with virtio devices >> disabled (i.e. not compiled in), but it fails "make check". With >> the changes, it also works for "make check". > > And just to complete the picture the point of that is I guess enabling > testing of random configs?
Not yet. But for developing, I try _not_ to compile the bits that I am not using. Here I just went for being general. After this patches, you can disable all CONFIG_VIRTIO_* stuff, and "make check" still works. (Ok CONFIG_VIRTIO_GPU/VGA can't be disable, it needs changes inside the tests themselves). Later, Juan.