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

Add parameter to specify the job timeout a number of seconds
or a number followed by a suffix for seconds, minutes or hours.

Examples:

avocado run --job-timeout 10 synctest sleeptest ...
avocado run --job-timeout 2m synctest sleeptest ...
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Allows to specify the total amount of time (in seconds)
that a test suite is allow to execute in a job execution.

When the timeout happens in a running test, then the test
is marked as error and the following test cases are marked
as skipped.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

run_suite() will accept an optional parameter to set (in seconds)
the timeout for the test suite to execute.

If the timeout happens in the middle of the test, it will report
as an error. The other tests (if any) will be marked as skipped.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Use TimeOutSkipTest as a placeholder for a test, to mark
that it will not execute due a job timeout.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Remove redundant initialization in classes:
* MissingTest
* BuggyTest
* NotATest
Signed-off-by: NRudá Moura <rmoura@redhat.com>

There are python traps (they can't be called scripts) which on import
modify stdin/out/err. This even on discovery makes avocado behave very
badly and it's hard to discover what and when it happened. This patch
provides dummy stdin/out/err when discovering the tests.

This solution only takes care of stdin/out/err related issues, safer
method would be to spawn another process and gather the info there.
Still we can't avoid modifying the OS during the discovery, but that's
IMO acceptable for a test-framework.

The examples/tests/failtest_ugly.py is used in basic_tests unittest
using "avocado list" (without this patch it waits for input, than it
prints the test's outputs and exits without listing the tests).
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Some very ugly modules use sys.exit() on imports. These are not catched
by Exception thus BaseException is required.

I did not omit KeyboardInterrupt as users get mad when ctrl+c doesn't
work. On the other hand nasty python code raising KeyboardInterrupt on
import can interrupt test-discovery (avocado list nasty.py => Interrupted
by user)
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This patch cleanups the view even on BaseException errors.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Usage of "except:" without listing exceptions catches BaseException,
which is not a good practise. This patch uses "except Exception" or
list of exceptions instead.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Get rid of test_parameters_simple and test_parameters_name
and use only test_parameters variable.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

To avoid making SimpleTest a special case, use the same
parameters that any avocado test class should use,
so that the parameter `path` is now `name`.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Sysinfo: make it configurable [v2]

As the sysinfo configuration moves from hard coded to configurable,
unittests should not count on any existing condition, such as the
existence of files or commands collected.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

This change adds flexibility to the current sysinfo implementation in
the sense that the commands to be executed and files to be collected
are no longer hard coded, but configurable.

The configuration file gains a section for that, and the files
themselves are plain text files that can be easily customized by users
to include extra commands/files.

Also, the the configuration for the profilers has been adjusted for a
uniform look, feel and experience.

Changes from v1:
 * Return the default `profiler` parameter value to None, so that it
   looks at the configuration value by default
Signed-off-by: NCleber Rosa <crosa@redhat.com>

A test may or may not have a data dir. Also, a test might
be on a read only location, so forcing the creation of
one will raise an OSError. This fixes a problem introduced
with 133b60e8.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

A simple wrapper for open() + readlines() that do not raise errors
in case of missing or inaccessible files.

Changes from v1:
 * Fix typo in docstring
Signed-off-by: NCleber Rosa <crosa@redhat.com>

First, this is purely a name change, no change of behaviour should
happen because of this.

Second, this is a naming issue, which means it can sound like a great
or awful idea and both opinions would be right about it.

Finally, IMHO, the name collectible is better suited here for a couple
of reasons:

 * It breaks apart from the already poluted use of log, logfile,
   logging, etc
 * It does not suggest that something will be logged in the
   traditional sense, that is, it will not go through the `logging`
   mechanism
 * It's more closely related to what us humans want, collect the files
   or command outputs for later analysis
Signed-off-by: NCleber Rosa <crosa@redhat.com>

We have mentioned a couple of times that symmetry is desirable on sysinfo
collectibles, so let's enforce that and simplify the code.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

When a relative file (or directory) operation is performed, it will
always be based on the current working directory. This means there's
no need to call `os.getcwd()` for that.

Let's skip that extra call and thus optmize this piece of code.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Avocado has no support for iterations other than running a given test
multiple times, so it makes sense to also remove this never used
portion of the sysinfo code.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

The default commands and files to be executed/collected at the start
and end of tests have never been defined, so they're safe to be
removed.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

After all, this is not a unittest of avocado per se, but
an example test. Besides, when executing it as a unittest, we
create an undesirable job directory in ~/avocado/job-results.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Let's use regular, triple quoted strings.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

It turns out that the test classes defined at the top of the
unittest file were being loaded, therefore directories were
being created at the ~/avocado/job-results directory. We can
fix that by moving the auxiliary class definitions to the
unittest class itself.
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>

Use a try/finally block, when at the end of the test,
we restore the old settings object to not mess around with
the behavior of future unittests.

Changes from v1:
 * Use a try/finally block instead of simply cleaning up,
   since if a bug/error happens, that cleanup wouldn't be
   executed, per ldoktor's comments.
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>

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

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

The flag -r tells to the stand alone tests to clean their
temporary results at the end of the run, so that we don't
leave traces of their execution.

Changes from v1:
 * Wrote an actual commit message, explaining the purpose
   of the commit, per ldoktor's suggestion.
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>

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

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