Packages changed:
  Mesa (20.2.4 -> 20.3.2)
  Mesa-drivers (20.2.4 -> 20.3.2)
  NetworkManager
  busybox (1.32.0 -> 1.32.1)
  cloud-init
  openssh
  xen (4.14.1_02 -> 4.14.1_05)

=== Details ===

==== Mesa ====
Version update (20.2.4 -> 20.3.2)
Subpackages: Mesa-libEGL1 Mesa-libGL1 Mesa-libglapi0 libgbm1

- update to 20.3.2
  * third bugfix release for the 20.3 branch
- drm and surfaceless are not specified as platforms anymore,
  remove them from egl_platforms

==== Mesa-drivers ====
Version update (20.2.4 -> 20.3.2)
Subpackages: Mesa-dri Mesa-gallium

- update to 20.3.2
  * third bugfix release for the 20.3 branch
- drm and surfaceless are not specified as platforms anymore,
  remove them from egl_platforms

==== NetworkManager ====
Subpackages: libnm0 typelib-1_0-NM-1_0

- Second attempt to exclude systemd.automount from nfs processing:
  fix boo#1116625

==== busybox ====
Version update (1.32.0 -> 1.32.1)

- Update to version 1.32.1
  - fixes a case where in ash, "wait" never finishes.
- sendmail-ignore-F-option.patch: ignore -F option as used by
  cron (workaround for [bbn#13426])

==== cloud-init ====

- Update cloud-init-write-routes.patch (bsc#1180176)
  + Follow up to previous changes. Fix order of operations
    error to make gateway comparison between subnet configuration and
    route configuration valuable rather than self-comparing.
- Add cloud-init-sle12-compat.patch (jsc#PM-2335)
  - Python 3.4 compatibility in setup.py
  - Disable some test for mock version compatibility

==== openssh ====
Subpackages: openssh-clients openssh-common openssh-server

- Support /usr/etc/pam.d

==== xen ====
Version update (4.14.1_02 -> 4.14.1_05)

- Update libxl.set-migration-constraints-from-cmdline.patch
  Remove code which handled --max_factor. The total amount of
  transferred data is no indicator to trigger the final stop+copy.
  This should have been removed during upgrade to Xen 4.7.
  Fix off-by-one in --max_iters, it caused one additional copy cycle.
  Reduce default value of --max_iters from 5 to 2.
  The workload within domU will continue to produce dirty pages.
  It is unreasonable to expect any slowdown during migration.
  Now there is one initial copy of all memory, one instead of four
  iteration for dirty memory, and a final copy iteration prior move.