Am 16.04.2013 um 10:36 hat Stefan Hajnoczi geschrieben: > On Mon, Apr 15, 2013 at 09:38:28AM -0600, Eric Blake wrote: > > On 04/15/2013 09:17 AM, Stefan Hajnoczi wrote: > > > Test that qemu-img convert -c works when input image length is not a > > > multiple of the cluster size. > > > > > > Previously an error message would be produced: > > > > > > qemu-img: error while compressing sector 0: Input/output error > > > > > > Now that qcow2 and qcow handle this case the test passes successfully. > > > > > > Signed-off-by: Stefan Hajnoczi <stefa...@redhat.com> > > > --- > > > > > +echo > > > +echo "== Creating single sector image ==" > > > + > > > +_make_test_img 512 > > > +$QEMU_IO -c "write -P0xa 0 512" $TEST_IMG | _filter_qemu_io > > > +mv $TEST_IMG $TEST_IMG.orig > > > + > > > +echo > > > +echo "== Converting the image, compressed ==" > > > + > > > +$QEMU_IMG convert -c -O $IMGFMT $TEST_IMG.orig $TEST_IMG > > > +_check_test_img > > > + > > > +# success, all done > > > > Is it worth also testing that qemu-img info on the converted image still > > reports that the guest sees a size of 512, to prove that the virtual > > size was not expanded as a result of compression tail padding? > > Good point. Let's beef up the test case. We can also read the first > sector to verify to still contains 0xa bytes.
Maybe just qemu-img compare -s both images? And you can add the info as well just in case that compare is broken... Kevin