[ovs-dev] [PATCH 0/3] Fix testsuite failures on big endian platforms.

[Daniele Di Proietto]

openvswitch branch-2.6 fails to build on Debian on big endian platforms,
as reported in https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=840770.

This series aims to fix some endianness related failures:

 1168: ofproto-dpif.at:8503 ofproto-dpif - conntrack - ICMP related
 1174: ofproto-dpif.at:8795 ofproto-dpif - conntrack - tcp pick up
 2235: ovn.at:599         ovn -- action parsing
 2249: ovn.at:3508        ovn -- dhcpv4 : 1 HV, 2 LS, 2 LSPs/LS dhcpv4

This should allow the build to pass at least on ppc64.  The mips builder
reports another failure

 2239: ovn.at:1651        ovn -- vtep: 3 HVs, 1 VIFs/HV, 1 GW, 1 LS vtep

but I haven't been able to reproduce that locally using qemu usermode
emulation of mips.  It's unlikely to be and endianness issue, because
it is failing also on mipsel.

I haven't actually run the whole testsuite on master, because qemu usermode
is too slow, I just run the above tests.

Daniele Di Proietto (3):
  csum: Fix csum_continue() on big endian with an odd number of bytes.
  ovn: Encode dhcpv6 PACKET_IN userdata as big endian.
  tests: Ignore options order in dhcpv4 ovn test.

 lib/csum.c               |  4 ++++
 ovn/controller/pinctrl.c | 28 ++++++++++++-------------
 ovn/lib/actions.c        | 19 ++++++++++-------
 ovn/lib/ovn-dhcp.h       |  5 +++--
 tests/ovn.at             | 54 +++++++++++++++++++++++++++++++++++++++++-------
 5 files changed, 78 insertions(+), 32 deletions(-)

-- 
2.10.2