These are chat archives for Exa-Networks/exabgp

20th
Mar 2016
Thomas Mangin
@thomas-mangin
Mar 20 2016 08:52
Thank you - I will look into this /etc/exabgp issue
You can indeed skip the tests - I run them on my laptop before release
They do not need to be run on the build
Thomas Mangin
@thomas-mangin
Mar 20 2016 09:19
@vincentbernat /etc issue resolved on master - silly pebkac. Exa-Networks/exabgp@2eda799
Thomas Mangin
@thomas-mangin
Mar 20 2016 09:34
Screen Shot 2016-03-20 at 09.34.02.png
Go understand !
sigh !
Thomas Mangin
@thomas-mangin
Mar 20 2016 10:22
@vincentbernat search in /etc/exabgp backported too
Thomas Mangin
@thomas-mangin
Mar 20 2016 10:47
@vincentbernat my laptop and codeship have no issue with running the tests
If it is failing on your machine, could you just find one of the failing test (not green) when your run ./qa/bin/conversation
and then use two terminals
SERVER=<code of the test, i.e.: A> ./qa/bin/conversation
CLIENT=<code> ./qa/bin/conversation
and provide me the outpout so I can see why it is not completing (if they still fail)
If it passes - they it is an issue due to forking and lost likely as you pointed communicate
Vincent Bernat
@vincentbernat
Mar 20 2016 11:11
@thomas-mangin: for example, on test 0, I see the client is trying to connect to port 179 instead of the port specified in exabgp.tcp.port
while the server is correctly bound to port 38833
nothing in the logs (port not displayed), but I see that on tcpdump
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:20
Thank you
Now I have no idea why this is happening
I assume you are running sid
@vincentbernat if so I will make a vm and try to replicate
Vincent Bernat
@vincentbernat
Mar 20 2016 11:22
@thomas-mangin: wouldn't it be possible that exabgp reads my /etc/exabgp/exabgp.env and ignore the environment variables?
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:23
.... It should not but it is a good lead worth investigating
Vincent Bernat
@vincentbernat
Mar 20 2016 11:23
deleting the directory, still the same behavior
should the client use exabgp.tcp.port or does it get the port information from somewhere else?
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:24
I would guess that using exabgp_tcp_port may fix it
It may be due to the way . are handled
Vincent Bernat
@vincentbernat
Mar 20 2016 11:25
yes, it works
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:26
Ok easy fix but not sure why this behaviour is os specific
Vincent Bernat
@vincentbernat
Mar 20 2016 11:27
replacing other vars in the same works too
maybe it's related to the shell
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:27
I will fix and release 4.16 as I have two other bug fixes already
Vincent Bernat
@vincentbernat
Mar 20 2016 11:27
on Debian, /bin/sh is /bin/dash
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:27
Yes, I use bash
Vincent Bernat
@vincentbernat
Mar 20 2016 11:28
sbin/exabgp is #!/bin/sh
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:28
Yes which is bash on OS X
Vincent Bernat
@vincentbernat
Mar 20 2016 11:28
ok
on Ubuntu, /bin/sh is also dash
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:29
I will double check
Vincent Bernat
@vincentbernat
Mar 20 2016 11:29
what distribution is codeship running?
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:29
No idea
Will check
Could be centos
Vincent Bernat
@vincentbernat
Mar 20 2016 11:29
I remember something about dot in environment variables
afaik, you can have dot in environment variables, but maybe shells may feel otherwise because they have to map that to variables names
so maybe some of them are cleaning up the environment
env titi.toto=3 sh -c env | grep titi
env titi.toto=3 bash -c env | grep titi
first one is empty, second one shows the environment variable
that may be a change in recent versions of dash since it is not working with 3.4.13 either while I am pretty sure it worked previously
and Debian is using dash as sh since a long long time
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:34
Yes. Fun fun fun
Vincent Bernat
@vincentbernat
Mar 20 2016 11:35
note that there is no problem of replacing /bin/sh by /bin/bash if you find this change less invasive
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:36
No simpler to use _ I added them as I had this issue around two years ago and it fixed itself
Perhaps more than two .. Times flies
Need to go in 5 minutes so doing it now
Thomas Mangin
@thomas-mangin
Mar 20 2016 11:41
dots with underscore breaks the program on my mac, I will need to look at it tonightk - replacing .
s/./replacing dots with underscore breaks the program on my mac, I will need to look at it tonight.
sigh
ttfn
Thomas Mangin
@thomas-mangin
Mar 20 2016 17:03
@vincentbernat changed, let see what the different CI reports. So travis says, my tree is fine but the exa-networks one is not .. for the same code ..
so it may have helped but it does not look like the whole solution
ok - only test 8 failed on pypy - the rest passed
and this test works on pypy so the issue is something else
The test failed as 'failed, a test is taking more than the allocated time’ which is when my timer (10 minutes) kicks in.
ExaBGP seems to be starting at 16:45:30

using port 32507

Sun, 20 Mar 2016 16:45:30 | DEBUG | 4874 | network | out loop, intialising

Thomas Mangin
@thomas-mangin
Mar 20 2016 17:10
and is killed at 16:45:43
Sun, 20 Mar 2016 16:45:43 | DEBUG | 4874 | wire | session 1 outgoing 127.0.0.1 lost TCP session with peer
So it looks like Travis-CI is limiting the number of parrallel fork and this causing my test to run sequentially and take over 10 minutes ...
Thomas Mangin
@thomas-mangin
Mar 20 2016 17:36
time ./qa/bin/conversation
real 0m52.529s
user 0m23.563s
sys 0m10.846s
Thomas Mangin
@thomas-mangin
Mar 20 2016 18:09
Increasing the build time on travis-ci

collected


0 1 2 3 4 5 6 7 8 A B C D E F G H I J K L M N O P Q R S T U V W X Y a b c - e f g h i

No output has been received in the last 10m0s, this potentially indicates a stalled build or something wrong with the build itself.

The build has been terminated

Sigh !