On Thu, Jul 19, 2012 at 08:51:30AM -0700, Ben Pfaff wrote: > On Thu, Jul 19, 2012 at 08:43:16AM -0700, Ben Pfaff wrote: > > On Thu, Jul 19, 2012 at 03:27:58PM +0900, Simon Horman wrote: > > > On Thu, Jul 19, 2012 at 10:08:50AM +0900, Simon Horman wrote: > > > > On Wed, Jul 18, 2012 at 10:21:50AM -0700, Ben Pfaff wrote: > > > > > On Wed, Jul 18, 2012 at 12:02:19PM +0900, Simon Horman wrote: > > > > > > Signed-off-by: Simon Horman <ho...@verge.net.au> > > > > > > > > > > When I apply this to master (currently 333be161abe78), I get the > > > > > following test failure. It isn't immediately obvious to me why this > > > > > change should have an effect on dl_vlan_pcp, so I'll leave it to you > > > > > to take a look and let me know. > > > > > > > > Thanks. I thought I had checked that all the tests pass. > > > > I'll look into this. > > > > > > After much confusion on my part I believe that I have found the problem. > > > > > > Expanding struct flow_wildcards causes the value hash returned > > > by flow_wildcards_hash() to change. This in turn causes the order > > > of flows in a NXST_FLOW_MONITOR reply (and probably elsewhere) to > > > be changed. > > > > > > I applied the _hack_ shown at the end of this email to test this theory: > > > it > > > should return the same hash as before the ipv6_label_mask element was > > > sdded > > > to struct flow_wildcards. The result was that the test in question passes. > > > > > > So witht his in mond I propose adding a hunk to this patch to > > > correct the test for the new hash in a new version of this patch. > > > > > > Alternateively, perhaps the test could be reverted to the state > > > it was in before I recently expanded it. Perhaps that would lead to > > > a stable test. > > > > I'd rather not revert it. Here's a workaround. > > I applied the workaround and then patch 1/3 to master. Thank you!
Thanks. FWIW, Acked-by: Simon Horman <ho...@verge.net.au> _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
