Auto generate api docs

Instead of manually updating the API rst files upon
new module additions, auto generate the .rst files
through a trick on conf.py.

With this we have reduced maintenance and always
up to date API docs.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

We'll introduce code in docs/source/conf.py to call
sphinx-apidoc to generate them on the fly, ensuring
API docs will always be complete on each new build
triggered on read the docs, due to the web commit
hooks configured on github.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

docs: API - Update new modules recently added to avocado

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

Report link

Class names can be found on instance.__class__.__name__,
not instance.__class__.name. Very silly bug that is
fixed here.

Also, Print error headers to provide visual cues of
the error that happened, also, provide a bug report
link for avocado crashes. Example:

Avocado crashed: ValueError: BUG!
Traceback (most recent call last):

  File "/home/lmr/Code/avocado/avocado/job.py", line 117, in _run
    raise ValueError("BUG!")

ValueError: BUG!

Please include the traceback info and command line used on your bug report
Report bugs visiting https://github.com/avocado-framework/avocado/issues/newSigned-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

In order to color error messages more prominently,
introduce fail headers, colored red, to provide
visual cues of what's going on.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Fix assorted bugs, make avocado return codes and functional test more complete

Avocado should handle gracefully and not crash.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

We have to tell apart 'expected' avocado error from
plain crashes. Refactor the Job.run() method, separating
it to an unhandled exception auxiliary method, and a
handled one, that properly sets the following return
codes for avocado execution:

* AVOCADO_ALL_OK (0)
    Both job and tests PASSed
* AVOCADO_TESTS_FAIL (1)
    Job went fine, but some tests FAILed or ERRORed
* AVOCADO_JOB_FAIL (2)
    Something went wrong with the Job itself, by explicit
    :class:`avocado.core.exceptions.JobError` exception.
* AVOCADO_CRASH (3)
    Something else went wrong and avocado plain crashed.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Add a mapping of possible return codes for avocado,
to the meaning of them. We need to differentiate
simple jobs from tests failed, jobs that failed, and
avocado crashes.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

For the (yet non existing) cases where we want to throw
a specific JobError, that will be managed by the runner
plugin.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

When we skip a test, that is not a failure. Only the
test was not applicable for the given test environment.
Therefore, we should not report those as failures.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

When tests don't exist, current avocado breaks on
a very spectacular fashion:

Traceback (most recent call last):
  File "scripts/avocado", line 29, in <module>
    sys.exit(app.run())
  File "/home/lmr/Code/avocado2/avocado/cli/app.py", line 72, in run
    return self.args.func(self.args)
  File "/home/lmr/Code/avocado2/avocado/plugins/runner.py", line 106, in
run_tests
    return job_instance.run()
  File "/home/lmr/Code/avocado2/avocado/job.py", line 111, in run
    test_instance = self.run_test(url)
  File "/home/lmr/Code/avocado2/avocado/job.py", line 76, in run_test
    test_instance = self._load_test_instance(url)
  File "/home/lmr/Code/avocado2/avocado/job.py", line 63, in
_load_test_instance
    f, p, d = imp.find_module(url, [test_module_dir])
ImportError: No module named bogustest

We should be of course more classy when such a problem happens.
Introduce a NonExistentTest class to handle such problems,
and use it when the ImportError happens. The end result is
a lot better:

DEBUG LOG: /home/lmr/avocado/logs/run-2014-04-29-13.12.34/debug.log
TOTAL TESTS: 1
(1/1) bogustest.1:  ERROR (0.10 s)
TOTAL PASSED: 0
TOTAL ERROR: 1
TOTAL FAILED: 0
TOTAL SKIPPED: 0
TOTAL WARNED: 0
ELAPSED TIME: 0.10 s
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

An oversight happened when writing the output plugins made
the ERROR status to be ignored and reported as PASS.
This patch fixes this problem.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Add new rule 'build-rpm-all' to the Makefile.

Add rule to build both source and binary RPMs.
Signed-off-by: NRuda Moura <rmoura@redhat.com>

Display init docs

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>

We had to set the autoclass_content param in conf.py.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Fix tmp dir

Introduce the --keep-tmp-files flag to the test runner,
that will keep them intact if for some reason we want to,
but by default clean test artifacts at the end of a test
job.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

In newer systems, such as Fedora 19 onwards, the /tmp dir
is mounted on a tmpfs by default, and most users won't
ever disable this 'feature'. Some test suites were
not quite designed to run under tmpfs, hence they might
fail. Not to mention that sometimes the test suite might
be large, or generate large binary files so that tmpfs
won't be able to handle it.

So instead of storing test artifacts such as test suite
source code copies, and compiled test suites under /tmp,
let's use /var/tmp, that is not supposed to be wiped
through reboots, and usually is not mounted on tmpfs.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

If we run avocado tests through nosetests, logs were
being generated in ~/avocado instead of ~/avocado/logs/run-[timestamp],
as normal tests executed by avocado do.

So, refactor the code that figures out the job directory
in job.Job() and move it to data_dir(). Make tests that
don't have a base_logdir set (such as instances executed
by nosetests) to use the same code, and we're all set.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

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

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

.travis.yml: Documentation build test

In order to avoid errors and garbage creeping up in the
code base, look at the sphinx build output for errors and
warnings, and fail if we found any.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Let's certify that Sphinx documentation can be properly
built as well.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

.travis.yml: Add some functional testing of avocado

Let's actually execute avocado in scenarios where we
know what the produced outcome is:

* Execute a list of tests that pass, expected rc == 0
* Execute a list of tests where some fail, expected rc == 1

Although very simple, it's already a good start of
functional testing of the code during each travis job.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

docs: Fix install instructions in GetStartedGuide

The commands described did not work under some circumstances,
so change them to safer versions.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

More docstring improvements

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>