RunnerClient name=value parsing does not work as intended

[whiteinge]

Bug identified in #55 by @jfroche.

The problem can be clearly seen via the following command. In addition to the name=value parsing, the arguments in RunnerClient are not being run through the YAML parser like in LocalClient.

pepper --client=runner test.arg foo bar=Bar baz=true qux='{qux: Qux, quux: false}'

[whiteinge]

This appears to be a deficiency in RunnerClient.cmd_sync() and RunnerClient.cmd_async() which are the methods that salt-api calls. RunnerClient.cmd() works correctly which is what the salt-run CLI command uses.

>>> import salt.config
>>> import salt.runner
>>> opts = salt.config.master_config('/etc/salt/master')
>>> runner = salt.runner.RunnerClient(opts)
>>> runner.cmd('test.arg', ['foo', 'bar=Bar', 'baz=True'])
{'args': ('foo',), 'kwargs': {'bar': 'Bar', 'baz': True}}
>>> runner.cmd_sync({'fun': 'test.arg', 'args': ['foo', 'bar=Bar', 'baz=True'], 'username': 'saltd
ev', 'password': 'saltdev', 'eauth': 'auto'})
{'args': ['foo', 'bar=Bar', 'baz=True'], 'kwargs': {}}

It looks like the two methods above are not processing the input through salt.utils.args.parse_input() which both LocalClient and RunnerClient.cmd() use.

@jfroche does this bug sound like something you can tackle from here? My available Salt time is limited at the moment. If not we can tag the core Salt team on the issue and they can add it to their sprint backlog.

[whiteinge]

Er. That shell session may not be 100% accurate. Pepper may be sending slightly differently low data when using --client=runner. Must verify what Pepper is sending first, but parse_input() is still the likely fix.

[ashald]

+1

[ashald]

So the PR with the fix is merged into Salt's develop branch. Let's wait for its release. ;)

[whiteinge]

@ashald thank you very much for fixing that!