tests/empty-hostname.sh fails: expected hostname "localhost" not found in logs

[mbiebl]

Bug-Debian: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=844914

I'm getting a test-suite failure

======================================================================
------------------------------------------------------------
Test: ./empty-hostname.sh
------------------------------------------------------------
rsyslogd started with pid  7403
imdiag[13500]: mainqueue empty
expected hostname "localhost" not found in logs, rsyslog.out.log is:
2016-11-20T14:10:26.205937+01:00 foo rsyslogd: [origin software="rsyslogd" swVersion="8.23.0" x-pid="7403" x-info="http://www.rsyslog.com"] start
2016-11-20T14:10:28.238432+01:00 foo rsyslogd: [origin software="rsyslogd" swVersion="8.23.0" x-pid="7403" x-info="http://www.rsyslog.com"] exiting on signal 15.
FAIL empty-hostname.sh (exit status: 1)

To reproduce the issue setup a /etc/hosts like this

127.0.0.1	foo localhost

If one uses

127.0.0.1 localhost
127.0.1.1 foo

the test succeeds. So I asume it makes certain (incorrect) assumptions about how /etc/hosts is setup

[rgerhards]

I have added some doc to the test. The assumptions on the hostname are valid, the probable root cause is that the preload library is not loaded for some reason. This is also clear from the hostname being returned.

[mbiebl]

I added some printf debug statements to runtime/net.c

Looking at getLocalHostname(), we do get an empty hostname from
gethostname() (from the overriden library) and at
https://github.com/rsyslog/rsyslog/blob/master/runtime/net.c#L1157

hnbuf is set to localhost (as expected).

At https://github.com/rsyslog/rsyslog/blob/master/runtime/net.c#L1164
dot is NULL so it calls gethostbyname() and iterates over the aliases.
This is where things go wrong. If I have an entry like this

127.0.0.1 foo localhost

fqdn is set to foo and this is what's returned from getLocalHostname.

[mbiebl]

On a related note: if gethostname() returns a non-emtpy string, why do we not use that name but set it to localhost at
https://github.com/rsyslog/rsyslog/blob/master/runtime/net.c#L1151

UPDATE: Forget what I said here. The != 0 check is testing the return code and not the string length. So all fine

[rgerhards]

Thanks, I was obviously too focussed on the gethostname() call. Using localhost as a default obviously also brings some confusion. I think we should rename it to localhost-empty-hostname, which would be an indicator of the real problem. I'll change the code accordingly and create a PR.

[rgerhards]

Note: I have also reconfigured the debian buildbot slave to showcase that problem. So hostname-related issues are hopefully sooner detected in the future.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.