This example file should have been remove together with commit
7b969527. Better late than never,
so let's do it now.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

Fix the docstring that doesn't reflect that the result parameter
should actually be a TestResultProxy.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

Some tests might require additional cleanups not just after the test,
but after the whole job. This patch adds another message called
"func_at_exit", which can be pushed by the test and will get registered
and called when all tests finish.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The test_status is handled in TestRunner.run_test quite chaotically.
This patch extracts test-related communication into a separate class,
and provides simple abstraction, making it simpler to extend it of
other messages.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The current probe is not identifying the version numbers
for CentOS 7.X. Let's use a more relaxed regexp to fix
that problem.
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

It turns out we forgot to state fabric dependencies in the
debian package. Let's fix that problem.
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

Let's remove an artifact from old times. The current
process APIs should handle stdout and stderr just fine.
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

Add the gdebi-core dependency that our ubuntu users might need
to if they are going to use the avocado software manager
library.
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

It turns out that apt doesn't really handle local installs
graciously as dnf/yum, but thankfully there is a program
named gdebi, shipped in Debian and Ubuntu, that would let
us do that. Use it when available (we can make gdebi a
dependency for the avocado debian package to ensure it'll
always be there).
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

Fix the command used to check package status (there was
a missing space there) and the logic for checking installation
(verify OK string in the entire command output).
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

process.py: CmdError class backward compatibility fix

In the newer CmdError class implementation the __init__() doesn't
have anymore the third non-mandatory argument "additional_text".
But still in the different parts of the source code the
CmdError is used with the third argument. If there is no special
reason behind deleting the "additional_text" then I would suggest
to keep for having the backward compatibility.
(for example got argument amount errors while running particular
libvirt tests which had to raise a CmdError exception)

avocado/core/virt.py: avoid crashing when trying to detect VM IP address

The current code assumes that libvirt will always return information
about the VM in a given format, and also assumes that the VM will have
a NIC, which may not be the case.

This adds better checks if the conditions are not met, skipping the
IP detection feature, instead of crashing.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

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

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

Together with the asset files, we'll now host the RPM repos at
a couple of OpenShift instances, named after the Avocado Project.

The two main repo files are located at:

 * https://repos-avocadoproject.rhcloud.com/static/avocado-fedora.repo
 * https://repos-avocadoproject.rhcloud.com/static/avocado-el.repo

And they point to the same base URL for individual repositories.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

With recent refactorings of the loading code, some of
the logic became unreachable (and irrelevant to avocado).
Let's remove the unreachable code.
Signed-off-by: NLucas Meneghel Rodrigues <lmr@redhat.com>

Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

To sync with the latest changes in Avocado.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The path in logdir has to be turned into a safe-path.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This function can be used in other parts of this project (eg. remote).
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

this was removed from the "Log remaining buffer on exit" fix, but it's
in-deed necessary (bfr is filled in a loop).
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The SubProcess logs output only when it finishes with '\n' to avoid
partial outputs. Anyway we have to print the remaining buffer on
exit, otherwise we miss the unterminated output (eg on "echo -n")
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

When the user provides url, which is not directory, this piece of code
tries to discover it as usual. When this fails it splits the command
using "shlex.split" and checks whether the first argument is executable.
If so it it reports it as a simple test.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Currently SimpleTest escapes the test command. External runner used
a workaround to be able to override the command with it's custom
one. Instead of this hack we can simply expect user to supply command
already escaped so it can be simply executed.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

We base our log-dir on test name, which can contain various characters
including '/', causing multiple directories being created instead of one
log directory. This patch modifies the `self.logdir` to use '_' instead
'/'.

Note this approach is not safe on windows so in the future we might
implement a `utils` function to make filenames "safe". (similarly to
django, which uses whitelist)
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

We have a nice mechanism to notify user about job failures, let's use
it instead of custom sys.exit().

Additionally add '\n' before the first message on exception, because
quite often there are already some characters written.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>