On 10.02.2016 20:26, Sascha Silbe wrote: > Dear Max, > > Max Reitz <mre...@redhat.com> writes: > > [tests/qemu-iotests/140] >>> -_launch_qemu -drive >>> if=ide,media=cdrom,id=drv,file="$TEST_IMG",format=$IMGFMT \ >>> - 2> >(_filter_nbd) >>> +_launch_qemu -drive >>> if=none,media=cdrom,id=drv,file="$TEST_IMG",format=$IMGFMT \ >>> + -device virtio-scsi -device scsi-cd,drive=drv 2> >(_filter_nbd) >> >> Why not just omit the device (and the media=cdrom along with it, keeping >> if=none)? This will change the reference output because there is no >> longer any tray to be moved, but this will still test what it's supposed to. >> >> (This may sound hypocritical coming from me, because I wrote this test >> so I could have just done so in the first place; I guess I just didn't >> realize that 'eject' works on device-less drives, too.) > > Is this supposed to work? I.e. can we rely on it?
Let's say I would rely on it. :-) (which is why I proposed it) The test checks that ejecting a BlockDriverState tree from a BlockBackend (that is, a medium from a drive) works even if that BlockBackend is exposed via an NBD server. It doesn't really matter whether the drive has a device or not, the main thing is that the NBD server notices that the medium is ejected and automatically stops offering the drive. So I would think that we are supposed to be able to rely on it; if we cannot, something else is probably broken. > If so, that would > certainly be the easier route for this particular test. Test coverage > should be unaffected as 139 already tests ejection (using virtio, unlike > 118 which is PC-only). > > The aliases patch has a value of its own, but that's a separate > matter. Yes, certainly. Max
signature.asc
Description: OpenPGP digital signature