Restore functionality in docker entrypoint

[zsistla]

In speaking further with Johannes, he recalled the original functionality allowed for starting something in the container that was not our daemon.

I removed the last clause of the case statement to not assume that our daemon will be run and to allow the caller to run another command besides newrelic-daemon. It functionally behaves like the previous (non POSIX compatible) entry point.

However…
The functionality of this(old and new) is somewhat limited because it only works for commands that don't use preceding flags. For instance, both instances can run a command the following:

  docker build -t c-daemon .
  name=temp
  docker run -d --name $name c-daemon ls
  docker logs $name

and

  docker build -t c-daemon .
  name=temp
  docker run -d --name $name c-daemon ls -l
  docker logs $name

And you'll see the output in the logs. However, if you try to call any command that uses flags preceding the command, like:

docker run -d --name $name c-daemon -lt /bin/sh

We call the newrelic daemon with all flags and anything following the flags, so we'd call newrelic-daemon with -lt /bin/sh arguments.

The three use cases are then:
. Running the daemon with the full path: docker run -d --name $name c-daemon /usr/bin/newrelic-daemon --loglevel debug
. Running the daemon without specifying the daemon binary: docker run -d --name $name c-daemon --loglevel debug
. Running an arbitrary command that is NOT preceded by a flag: docker run -d --name $name c-daemon cat /var/log/newrelic-daemon.log

This has been tested by tests/image_tests.sh.

image_tests.sh has also been updated to test functionality of usage case #3.

[pyohannes]

Shouldn't this clause be removed?

[zsistla]

Yes, thanks for catching that. It looks like that commit didn't make it for some reason. I'll redo the commit for that one again.