Make loader registration unique and make dispatcher propagate exceptions

The dispatcher classes weren't propagating exceptions when running
`map_method`, which is basically how the command line plugins methods
`configure` and `run` are called.

The dispatcher should not be silent about possible exceptions thrown
by plugin's `configure` and `run` methods, so let's fix that.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Long story short, after taking a look at:

https://github.com/avocado-framework/avocado-vt/issues/322

I noticed that, on purpose, the VirtTestLoader class would be registered
both by `avocado_vt/plugins/vt.py` and avocado_vt/plugins/vt_list.py`.
The reasoning is simply to keep plugins independent, without depending
on each other.

But, it turns out the registering the loader on both places doubles the
time it takes to find VT tests. Without this patch:

$ /usr/bin/time ./scripts/avocado list > /dev/null
9.52user 0.17system 0:09.68elapsed 100%CPU (0avgtext+0avgdata 189528maxresident)k
0inputs+3744outputs (0major+63564minor)pagefaults 0swaps

With this patch:

$ /usr/bin/time ./scripts/avocado list > /dev/null
5.47user 0.13system 0:05.60elapsed 100%CPU (0avgtext+0avgdata 145716maxresident)k
0inputs+1920outputs (0major+51958minor)pagefaults 0swaps
Signed-off-by: NCleber Rosa <crosa@redhat.com>

loader: make test methods unique when discovering tests [v2]

avocado.core.loader: Use sys.path.remove instead of pop

As per report on issue #952, Avocado currently finds multiple tests
when multiple test names exist in a given Python class. When the
test code itself is loaded and the test class instantiated, only
one test method will remain (others will have been overloaded by
that one).

So, to mimic this in the static, AST based, test discovery, let's
make the test methods unique and try to keep the order in which the
AST parser found them.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

It's safer to use "remove" in case someone adjusted "sys.path".
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

HTML output support: behave when the plugin is not installed

We're still pointing to the old repo location. Let's update that.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

avocado.tests: Port aiostress from autotest [v2]

It looks like a lot of our users are also hackers and want to run
Avocado from source. Since version 0.31.0 this requires one additional
step. Let's properly and permanently document it.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

The previous text "please adapt to your platform" is enough.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

If contributors choose to send patches to the mailing list, we'll
take them.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

So that regular and advanced users (Avocado hackers) know better
where to look for specific information.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

That way no uninstall is attempted, and no (failed) installation of an older
version crashes the builds.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

We currently only set HTML support as disabled if the `html.py`
library fails to import. The truth is that missing files that
are intended to be present on the (separately packaged) HTML
plugin package can cause other kinds of exceptions.

Let's add an extra check, looking if the resource files are
present (currently the template file).

Note: there's one outstanding bug, for which a fix is being
worked on. When Avocado is installed, either from packages
or from source code, the HTML plugin entry point is always
registered. That means the plugin Python module is always
tried to be loaded. If the plugin *package* is not installed,
an error message is given on the command line.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Sometimes remote servers might have MOTD (message of the day)
banners, that will introduce line breaks on the remote output
of the `avocado -v` command. Let's use re.MULTILINE and
re.findall to solve that.
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

Some debian packages are needy and demand our attention
even if we already told apt-get to answer all questions
with 'yes'. We need to export

DEBIAN_FRONTEND='noninteractive'

So that apt-get can be run fully unattended. Let's adjust
accordingly, and pass some default dpkg values.
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

Add new 'aiostress' test based on autotest.
Signed-off-by: NAmador Pahim <apahim@redhat.com>

avocado.tests: Port bonnie from autotest

avocado.tests: Port compilebench from autotest [v2]

Add new 'compilebench' test based on autotest.
Signed-off-by: NAmador Pahim <apahim@redhat.com>

Add new 'bonnie' test based on autotest.
Signed-off-by: NAmador Pahim <apahim@redhat.com>

selftests/functional/test_basic.py: remove references to old Plugin

There are a few leftover references to the old Plugin classes in
the functional tests. Let's remove them.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

avocado.core.loader: Append test directory to sys.path [v2]

If a test has local imports (a fairly common use case),
we have to append the base test dir to sys.path. Otherwise,
we end up having a horrible traceback as a result:

Exception loading test
Traceback (most recent call last):
  File "/home/lmr/Code/avocado.lmr/avocado/core/runner.py", line 200, in _run_test
    instance = loader.load_test(test_factory)
  File "/home/lmr/Code/avocado.lmr/avocado/core/loader.py", line 262, in load_test
    test_module = imp.load_module(module_name, f, p, d)
  File "scylla_longevity.py", line 5, in <module>
    from sdcm.tester import ScyllaClusterTester
ImportError: No module named sdcm.tester

Which is clearly not wanted. We had this problem before
and now we have a regression. It's time for a functional
test to make sure we don't regress again.
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>
Signed-off-by: NCleber Rosa <crosa@redhat.com>

With the same one that comes from the test documentation.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Add new 'stress' test based on autotest.
Signed-off-by: NAmador Pahim <apahim@redhat.com>

RHEL requires epel to be enabled. Let's mention it in the documentation.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Older versions of stevedore does not support `__contains__` method. Use
try/except to handle subcommand handling.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Revert "Pseudo terminals: add utility module"

It was noticed that the Python standard library provides os.ttyname(),
and, while it only returns the Pseudo Terminal file path when given
the slave file descriptor, it looks to be sufficient for most, if not
all, use cases.

This reverts commit 003bfaed.

CC: Lukáš Doktor <ldoktor@redhat.com>
Signed-off-by: NCleber Rosa <crosa@redhat.com>