After observing some racing conditions related to the execution
of processes on sub shells, step back and provide users
the ability to chose whether they want their command to
be executed in a subshell (in most cases, you'll want it
only if you want to use shell constructs such as pipes in
the resulting command).

Add the param shell to the SubProcess constructor (shell=False),
and modify avocado to use shell=True in the few commands that
actually require it.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Now that we do not initialize the subprocess upon class
initialization, add a utility method .start(), that will
start running the command, useful for background commands.

In order to make things symmetrical, add a .stop() method,
that will send a sigterm to the process, then wait on it.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Currently, as soon as we initialize a process.SubProcess
object, a subprocess.Popen() object is created. This breaks
the user's expectation that the process will start when
run() is called. Therefore, move process execution to
a private method, and call that method in all public
methods of the class.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

The reasoning here being that we don't want this part of
the public API. We don't need users to interact directly
with the Popen object, since the interaction will be
done by means of the public methods.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Introduce avocado.utils.script module [V3]

Remote machine plugin [V4]

Introduce the documentation regarding the new remote plugin.
Include information about using the remote plugin in the manual page.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Introduce a new unit test for remote plugin.

Since the remote plugin and the vm plugin share the same RemoteTest
class, we need to update the vm_unittest to use this class.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Introduce remote plugin that allows one to run tests on a remote
machine, by using the SSH network connection provided in module
avocado.utils.remote.

Since the remote plugin and the vm plugin share the same RemoteTest
test class, we're going to define this class inside avocado.test module.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Introduce avocado.utils.script module to handle the creation
of scripts stored in the file system.

The module includes a flexible class for creating generic scripts and
 includes utilities functions too. Some examples:

    CONTENT = """
    #!/bin/sh
    echo "Hi There!"
    """

    x = avocado.utils.script.Script('/var/tmp/test.sh', CONTENT)
    x.save()
    ... do something with x, like run it ...
    x.remove()

    t = avocado.utils.script.make_temp_script('test.sh', CONTENT)
    ... do something with y, like run it ...

Avocado functional tests makes use of script creation,
so we will handle these creation using this proper module.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

avocado.job: Move sysinfo collection to test result object

We found a bug: The sysinfo collection was overwriting the
post information in jobs run through the vm plugin. Fix
the problem by moving the information collection to the
test result object. This is transparent to plugins, one
important requirement of the solution.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

avocado: Bump release to 0.14.0

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

examples: Be didactic with mux-environment.yaml.

avocado.core.tree: Match whole key in filters.

When filtering nodes, use the whole key for comparing.

So if you have /long and /longest and you're filtering for /long,
then it will match /long alone.

The previous behaviour was using the partial
string match, so both /long and /longest were being returned,
which is wrong.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

The example for multiplex must be didactic, using one
that would make sense in a real world usage.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

examples/mux-environmentyaml: Don't use local filters.

Remove lines that contains local filters, which is a feature
that is disabled by now (it's broken).
Signed-off-by: NRudá Moura <rmoura@redhat.com>

GDB stdout support

Manpage: Update information regarding global filters and multiplexer.

avocado.utils.remote: Changes to the .run() method

Avocado's GDB support when using avocado.utils.process API
now properly populates the command result stdout attribute.

This now allows tests that run programs and check its output
(stdout only, not stderr) to work transparently under GDB.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

The init method on SubProcess got new parameters. Even if not
implemented in GDBSubProcess, those two ducks must quack alike.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Update the manual page of avocado to match the current status of the multiplexer,
regarding command line options for global filters.

Fix information on command line option -c | --content. It's about variables.

Plus:  cosmetic fixes around.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

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

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

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

For convenience, save the contents of the whiteboard
into a file 'whiteboard', inside the test results.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Make the run() method to return a CmdResult object,
to make it symmetrical to the process.run() result.
Also, turn on logging of outputs for commands executed
on remote connections.

The reason why we want to do that is to increase the
debugging output we have for commands executed over
the fabric SSH connections. This way we can obtain
better, richer outputs in test logs that use our
fabric wrappers.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Sprint 14 pendencies

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

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

Basically, trying to encode information of the variants themselves
in the tag name is a lost battle, since:

1) We need unique test ids, and sometimes only the last part of
   the name of each variant combined is not going to cut it
2) Coding the entire path of each variant combined would create
   in many cases enormous tags and the point would be lost

So go back and use sequential numbers that match the output
of the multiplex plugin. This way it'll be easy to correlate
variants and tag numbers.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

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

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

Multiplex refinements: filters

In the current state of the code, the local filters are broken, so it's
better to keep it disable for now.