On 7 Feb 2014, at 11:42, Ian Hinder ian.hinder@aei.mpg.de wrote:
On 7 Feb 2014, at 03:43, Frank Loeffler knarf@cct.lsu.edu wrote:
On Fri, Feb 07, 2014 at 02:35:11AM +0000, ian.hinder@aei.mpg.de wrote:
Error: Syntax error on line 5 of /home/jenkins/workspace/EinsteinToolkit/simfactory/etc/defs.local.ini: no whitespace allowed in section identifier: []Hi
Is this file in svn/git somewhere, or how else can we change it?
This file is created dynamically. The section header in question comes from the output of "sim whoami", which was clearly never designed to be used from scripts. For some reason, on the caltech build node, it is now outputting
Unknown machine name caltechcactusjenkins.spdns.org Unknown machine name caltechcactusjenkins.spdns.org Current machine: None
and the build system is only expecting the last line (as the machine is not defined in the simfactory database). The warnings should be on stderr, but they are not. Maybe the hostname of the VM host changed? Anyway, I've modified the scripts to try to get around this (I just take the last line of output).
Bzzzt. Wrong answer. In fact, the script would have worked fine in this case. My debugging was complicated by the fact that this particular problem was only visible with a partially-written defs.local.ini file, so I didn't notice it when testing.
The real problem was that Barry committed a new machine definition with a missing basedir entry. This was probably in his defs.local.ini file, so simfactory never complained on his machine. In the test system, the entry remained undefined, and simfactory aborted with an error about this during "sim whoami". This error should have gone to stderr, but it went to stdout, because the dprint function in simfactory has a special case for "early" calls before simenv is initialised, and this branch of the function ignores the instruction to output to stderr. Since the output of "sim whoami" was going into a pipe, the error message was lost, and the machine name was interpreted as "" instead of None, which is what the script expected.
I've fixed the machine entry with a dummy value; Barry, you probably want to put something sensible there.
I've also fixed simfactory to output to stderr even in the pre-simenv branch of dprint, so this shouldn't happen again.