Add the fields 'logdir', 'logfile' and 'fail_reason'.
Those fields will be necessary to expand the HTML
reports to correctly exhibit failure information
from remotely run tests.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Looking in retrospect, we used quite cryptic names for
the logfile and logdir template dict keys. Let's fix that
so that the code is more understandable.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Make sure all brower terminal output is suppressed. In order
to do that, eschew the use of the python module web browser
and use subprocess, executing the utility xdg-open in a subprocess
with stdout/err redirected to /dev/null.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

* Make code lines to fit under 80 columns
* Slight reorganization of commits

No functional changes were made to the code.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Argument parsing job result [V4]

The current Job object initialization is doing many setups at once,
 for example, it creates a new job results dir with an "id" file inside.
 It handles the sysinfo start, update the latest
 link, etc. Things that can be done later.

This changes allows split the initialization in subparts:
* _setup_job_results() - creates a new job results directory with an id inside.
* _update_latest_link() - update the "latest (symbolic) link" to the job results.
* _start_sysinfo() - start the system information module.
* _remove_job_results() - remove the current job results from the directory.

So when running tests that are not found, say `avocado run quuuux`
then the job results will be removed and the link for the previous
result (latest) is preserved. Everything else should work as it was before.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Instead of delaying the parsing of the test list to code that handles
the jobs, let's handle that early, and in the usual Python way, simply
by relying on argparse.

For other, possibly more complex command line requirements, custom code
can be written, but it should still be placed at the argparse level.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Some obvious, but AFAICT not yet existing tests, that check if when
using bad command line arguments, avocado exists with a proper status
code.

Also, a more complex command line behavior test that checks if avocado
errors out early, before it creates the job result directory.
Signed-off-by: NCleber Rosa <crosa@redhat.com>
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Documentation: Add section about installing from git repository.

Add a new section inside the "Getting started guide", about installing Avocado
from the git repository with confidence that the dependences will be installed.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Cleanups sprint 21 - stage 1

docs: Add workshop mindmap to Introduction docs

Add link to the DevConf mindmap to introduction documentation.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

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

Some typos found during automated inspection of the code.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Clean up imports inside the avocado code tree. This
does not bring any functional changes to avocado.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

PEP8 fixes: W503

The new pep8 adds W503 line break before %s check. This is mass style
update of avocado framework without any actual changes.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This commit aims to keep the debian changelog file
as up to date as possible.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

requirements-travis.txt: Require the latest version of inspektor

It'll be necessary so that we can start ignoring error E402
during pylint checking.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

GDB fixes

The path of the binary that is fed into the GDB commands file can be
a relative path, which can cause the behavior described in issue #412.

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

The reason is that runtime GDB settings are set by the GDB plugin, and
when running the test in standalone mode, there's no plugin activation
whatsoever.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

It's a very frequent code pattern to look for a binary in the system
using find_command(), and, if no command is found, trap the exception
generated and return a fallback.

This adds a `default` parameter that makes it possible to follow that
pattern without code outside the function itself.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

avocado.utils.process.find_command(): move from avocado.utils.process to avocado.utils.path

The reasons for the proposed move include:

1) Even though find_command() looks for executables, it doesn't really
   try to execute them as a living process.

2) The implementation looks for it in a $PATH like list, so, out of a
   possibly better choice, avocado.utils.path was chosen.

3) Other modules can use that functionality, but not the execution of
   process.

4) Avoid circular dependencies: this is actually the first motivation
   of the proposed change, and a true story with the GDB plugin/test.
   Since avocado.utils.process depends on avocado.gdb, this module
   can not use find_command() to look for gdb/gdbserver binaries.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

process.find_command() raises process.CmdNotFoundError, and not ValueError.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

Introduce sysinfo profilers (journalctl & vmstat) [v5]

Sysinfo module now allows to define profilers, which are programs
defined in command line, to start as a background processes and log
the output in a proper directory. Example: vmstat 1.

The profilers are started before a job execution and then, stopped at
the end of the job. The output is stored inside `sysinfo/profile/*`
subdirectory, in job result directory.

The profilers can be customized in `avocado.conf`, section
`sysinfo.collect`.  The default configuration define
profilers as the following: `journalctl -f` and `vmstat 1`.

This feature is turned off by default.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

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

Set _DEFAULT_COMMANDS_END_JOB as the same as DEFAULT_COMMANDS_START_JOB.
Set _DEFAULT_FILES_END_JOB as the same as _DEFAULT_FILES_END_JOB.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Define attributes .pre_dir and .post_dir (and for future usage, .profile_dir)
inside Sysinfo class, to handle the paths of those directories at the initialization.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Don't ignore stderr, redirect its output to stdout, in order to catch
commands not installed or malformed command line options.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Detect output to stdout as invalid at the beginning of execution
of the plugin, and not at the end of the html result execution.

Fix problem detected on test_output_incompatible_setup_3 when
sysinfo profile daemons are started and then left running, after
plugin exits with code 2.
Signed-off-by: NRudá Moura <rmoura@redhat.com>

Job standalone mode