Log a message when loading external plugins discovers that
there is no plugin inside a module.

Plus: cosmetic formatting.
Signed-off-by: NRuda Moura <rmoura@redhat.com>

Signed-off-by: NRuda Moura <rmoura@redhat.com>

* Add optional parameters 'name' and 'enabled' to the constructor.
  You can just use the class variables name and enabled.
  Change plugin variable to noname and enabled variable to False.

* Include new attribute 'description' to include short documentation
  about the plugin. By default it uses the docstring of the class
  or fall backs to a message that there is no documentation.

* Add repr() facility.
Signed-off-by: NRuda Moura <rmoura@redhat.com>

avocado.core.tree: Add note about ETE2 and copyright clarification

Add a note in the copyright headers and tree docstrings
recognizing that the base Tree data structure had inspiration
(and the ASCII drawing code) from the ETE2 project, and
hereby list Jaime Huerta Cepas <jhcepas@gmail.com> as one
of the authors of the code, with his express consent.

CC: Jaime Huerta Cepas <jhcepas@gmail.com>
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Ctrl c improvements last

Not necessary or desirable, since we get trinity's
output in the logs already, and the output of the
test is quite extensive.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

1) Revert the test subprocess side SIGINT handler
   to its original form (just raising an appropriate
   TestInterruptError exception)
2) Move the KeyboardInterrupt handling to the test
   run main loop, as we had nested handling blocks
   that ended up creating a deadlock.
3) Start counting the ignore window time after Ctrl+C is
   pressed by the second time, not the first.

With this, we finally get trinity to call its Ctrl+C
handler correctly and cleanup after itself, as well
as avocado cleaning up correctly if enough time is
given to the test subprocesses to end.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Chasing a bug, I found opportunities to improve the
process related code:

*  Populating the CmdResult object on every
   opportunity that wait() or poll() is executed
   in the subprocess.

*  Changing the name of the original wait() method
   to run(), making it consistent with the run()
   function of the module.

Also, to avoid a defunct process on a Ctrl+C, register
the SIGINT signal handler to the SubProcess wait()
method.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

For a number of reasons, the original implementation
called the processes with shell=False (subprocess default),
I noticed later that we lose things like the ability
to use pipes in commands executed and proper job
control and process handling.

So change that and use shell=True.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

If a process is misbehaved, that message will be confusing,
because avocado is actually waiting on child processes to
finish.

Then let's get rid of the message altogether.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Fix small bug in external plugins

Live test state

New multiplexer

Also, remove the references to the old system.
Signed-off-by: NRuda Moura <rmoura@redhat.com>

Signed-off-by: NRuda Moura <rmoura@redhat.com>

That change has to be made all at once for the sake of
bisectability. Update avocado.job and the multiplexer
plugin with the new multiplexing library code.

Also, update functional tests, the avocado sample tests
with the new YAML based config files, and the requirements.txt
file with the new prominent dependency of avocado, PyYAML.
Signed-off-by: NRuda Moura <rmoura@redhat.com>

Based on the new treenode data structure, we
built a new multiplexing algorithm. It has some
different properties and uses a more standard
YAML file format for the configuration files.

This commit is the new base library.
Signed-off-by: NRuda Moura <rmoura@redhat.com>

Introduce a new data structure that will be the base of
our new multiplexing mechanism.
Signed-off-by: NRuda Moura <rmoura@redhat.com>

Which doesn't hold any logic, but serves as an example.

To watch the intended behavior in action, I recommend running the
"one_hundred_cycles" variant of this tests.

The test runner will show that the test is running (yellow throbber
spinning), and once each six seconds (+1% into the test run) a green
throbber will be shown.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

A test writer can write its own check that will be called by default, or just decide
on each call to communicate_progress() if there has been any (progress).

Also, the job main loop now shows a green throbber when there has been real test
progress.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

So that it can report back its state whenever it feels like it.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Including number of sleep cycles, sleep cycle length and finally
the sleep method, either using the python time.sleep() or running
the shell "sleep" command.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

So that multiple test states can be received by the main loop, but
only if the test is finished we break out from the loop.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

And running state. Every time the state is produced and put in the queue,
the elapsed time is updated. Also, a new flag, "running", shows if the
framework has entered or exited the test entry point.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

The throbber that gives visual indication of progress can now show if the progress
includes test specific progress (GREEN) or if the test process is running (YELLOW)
but actual progress is unknown.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NCleber Rosa <crosa@redhat.com>

Add functional tests to explore external plugins in two
bad scenarios: 1) plugin without configure() method implemented,
2) plugin with syntax error.
Signed-off-by: NRuda Moura <rmoura@redhat.com>

Fix the variable to display in log when an external plugin fails
to load.
Signed-off-by: NRuda Moura <rmoura@redhat.com>

setup.py: add missing !#/bin/env python header

avocado.job: Change Ctrl+C handling

Without it, setup.py is interpreted as a bash script.

Instead of sending SIGKILL to child processes right away,
ignore Ctrl+C for a time window of 2 seconds, explaining
to the user that, and that after the ignore window,
new Ctrl+C will send SIGKILL to the children and terminate
the process.

The functional test was removed, and will be re-added
once I figure out how to properly test the interrupt
behavior.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Initial man page support

This page is intended as a basic offline aid for new users of avocado.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NCleber Rosa <crosa@redhat.com>

IMHO, the generated sphinx config file is too noisy and thus hard to
manage. Let's clean it up.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

More usability fixes

Instead of displaying a not very informative 'too few arguments'
message, display the original error message that mentions about
test IDs *and* the run help output. Adjustments to the original
functional test were made accordingly.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>