Allow environment variable substitution in conf file

[delitescere]

• Enabled with the --enable-env command line option
• Opt-in ensures current conf files don't get into ConfigParser
  recursion issues, e.g. home: %(HOME)s
• Provide nicer error on exit if the conf file has bad interpolation
• New unit tests with and without command line option present

[delitescere]

The build failure is because it is claimed in the README that python 2.7 is ok to use, but CI is building with python 2.6 which doesn't support the assertRaises as a context manager (see http://docs.python.org/2/library/unittest.html#unittest.TestCase.assertRaises).

[delitescere]

Wow this is one flakey CI. It can't make up it's mind if it's running python 2.6 or 2.7, and randomly fails anyway.

Suffice it to say, this patch is clean.

Cheers,
Josh.

[LeoCavaille]

Thanks @delitescere
We will review this change for the 5.3 release! (5.2 is now frozen for imminent release)
Regarding the CI, it's running BOTH python 2.6 and 2.7 ! Because non-EOL'd OS still run python 2.6, and anyone should be able to install from source the agent..

We have made huge steps towards non-flakiness, but I cannot agree more it sucks. Relying on so many 3rd party mirrors and services for integration testing is hard, next up will be some more caching to avoid that.

[LeoCavaille]

Cross reference #1243

[vincenzovitale]

+1

[LeoCavaille]

This is a really nice feature, we'll work on it soon for the future release we didn't have time to really work on it yet. And we want to add some more work on top so that we revamp the config.py code and the check config code in the agent to allow using the environment for more things. Setting milestone to 5.4.

[LeoCavaille]

Still some more work to do, working a leo/configcleanup branch. Will submit new PR for 5.5.

[mattes]

+1

[irabinovitch]

@LeoCavaille Should we close this in favor of the branch you are working on?

[irabinovitch]

@LeoCavaille Should we close this in favor of the branch you mentioned in #1343 (comment) ?

[irabinovitch]

@LeoCavaille checking back in?

[LeoCavaille]

@irabinovitch wow somehow I missed that thread 3 times in a row. I will coordinate with the rest of the team (cc @remh) to decide what to do next for this feature and re-update the PR this week.
I think this has been milestoned for 6.X as that code has been refactored quite a lot since that PR was opened. My former branch with updates to this initial patch cannot easily be merged either.

[jonhosmer]

@delitescere @LeoCavaille
You may want to add ConfigParser.ConfigParser.optionxform = str before config = ConfigParser.ConfigParser(os.environ) so that the keys in os.environ are treated as case sensitive.

For example:

>>> from ConfigParser import ConfigParser
>>> import os
>>> os.environ['XY'] = 'foo'
>>> os.environ['Xy'] = 'bar'
>>> os.environ['xY'] = 'baz'
>>> print '\n'.join(sorted([':'.join((k, v)) for k, v in os.environ.items() if k.lower() == 'xy']))
XY:foo
Xy:baz
xY:bar

>>> cfgp = ConfigParser(os.environ)
>>> print '\n'.join(sorted([':'.join((k, v)) for (k, v) in cfgp.defaults().items() if k.lower() == 'xy']))
xy:foo

>>> ConfigParser.optionxform = str
>>> cfgp_cs = ConfigParser(os.environ)
>>> print '\n'.join(sorted([':'.join((k, v)) for (k, v) in cfgp_cs.defaults().items() if k.lower() == 'xy']))
XY:foo
Xy:baz
xY:bar

>>> assert([(k, v) for (k, v) in cfgp_cs.defaults().items() if k.lower() == 'xy'] == [(k, v) for (k, v) in os.environ.items() if k.lower() == 'xy'])
>>> assert([(k, v) for (k, v) in cfgp.defaults().items() if k.lower() == 'xy'] == [(k, v) for (k, v) in os.environ.items() if k.lower() == 'xy'])
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-43-1f97baeb1e97> in <module>()
----> 1 assert([(k, v) for (k, v) in cfgp.defaults().items() if k.lower() == 'xy'] == [(k, v) for (k, v) in os.environ.items() if k.lower() == 'xy'])

AssertionError:

[delitescere]

/unsubscribe