We've tried running LNST tests with a ppc64 machine as slave (another
machine, which acted as master+slave was x86_64) and got the following
traceback with all tests (even the most basic ping test):
2015-04-23 12:42:29 (secondone) - DEBUG: Executing: "ping
10.16.167.142 -c 10 -I eth1"
2015-04-23 12:42:29 (secondone) - WARNING: decoding NDA_DST
2015-04-23 12:42:29 (secondone) - WARNING:
Traceback (most recent call last):
line 1084, in decode_nlas
line 1215, in decode
family = self.family_map[self.parent['family']]
These tracebacks don't seem to affect test in any way.
I've digged into this a bit and the issue seems to be caused by an unexpected
byteorder when decoding a message in pyrote2.netlink.nlmsg_base.decode().
This happens when we get a 'family' of 4 bytes, so we read 4 bytes:
0x02 0x00 0x00 0x00
And struct.unpack() translates it into 33554432 instead of 2.
I realize that I might be in a wrong list for this, so I wanted to ask
if thats something our setup (x86_64 mixed with ppc64) or LNST might be
responsible for or should I take this to pyroute2 devs?
P.S. If thats a pyroute2 bug I would be greatful if someone would help
me out to reproduce this with a simple script instead of LNST.