Routes to be withdrawn ignored when UPDATE contains both WITHDRAWN ROUTES and NLRI fields.

First UPDATE advertised a feasible routes (8.0.1.0/28, 8.0.1.16/28) in NLRI variable. Second UPDATE announced next two routes in NLRI and the 8.0.1.0/28 as to be withdrawn from service in WITHDRAWN ROUTES variable. The withdrawal is not processed but it shall be (RFC 4271, chapter 4.3, page 20). For details see the attached karaf.log, please.

robot/report/log after the 8.0.1.0/28 withdrawal - single UPDATE:
Count prefixes in {"topology":[{"topology-id":"example-ipv4-topology","topology-types":{},"node":[{"node-id":"192.0.2.1","l3-unicast-igp-topology:igp-node-attributes":{"prefix":[

{"prefix":"8.0.1.48/28"}

,

{"prefix":"8.0.1.16/28"}

,

{"prefix":"8.0.1.0/28"}

,

{"prefix":"8.0.1.32/28"}

]}}],"server-provided":true}]}.

Reproducible for the distribution-karaf-0.3.1-Lithium-SR1.zip as well for the latest distribution-karaf-0.3.2-20151003.125909-114.zip.

Note: the same scenario using dedicated UPDATES for NLRI and WITHDRAWN ROUTES variables works fine:
robot/report/log after the 8.0.1.0/28 withdrawal - dedicated UPDATEs:
Count prefixes in {"topology":[{"topology-id":"example-ipv4-topology","topology-types":{},"node":[{"node-id":"192.0.2.1","l3-unicast-igp-topology:igp-node-attributes":{"prefix":[

{"prefix":"8.0.1.48/28"}

,

{"prefix":"8.0.1.16/28"}

,

{"prefix":"8.0.1.32/28"}

]}}],"server-provided":true}]}.