Common location for contributing guidelines is "./CONTRIBUTING*" and
even github encourages us to use it. Let's symlink our documentation to
simplify first-time contributions (and also to make github happier).
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

ensure returns of service manager to be of type 'str'.
Signed-off-by: Nlolyu <lolyu@redhat.com>

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

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

Avocado requires the logging streams to be set before some calls. Most
critical ones are 'avocado.Test' related ones as without initialization
the:

    No handlers could be found for logger...

messages can cause infinite recursion. Anyway there are other places
where avocado loggers are called so let's set it on (hopefully) all
places (we have to treat some utils as well as for backward
compatibility reasons these utils use avocado loggers directly).

Note in functional tests this is not required as Avocado always setup
logging streams prior using those utils (otherwise the failure would be
a real bug).
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

Without this, the data files associated with the avocado_result_html
module won't be installed.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

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

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

On some situations, the same information may be given on tags that can
be reused during test runtime.  Let's create a property to allow a
test to read them.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Up to this point, the matching and thus filtering of tags was flat.
Now, the individual value of the required keys are checked.

For instance, if you have a test such as:

    class My(Test):
        def test(self):
            """
            🥑 tags=classification:tier1
            """
            ...

Both of these filter conditions will match:

 $ avocado run -t classification test.py
 $ avocado run -t classification:tier1 test.py

But this won't match:

 $ avocado run -t classification:tier2 test.py

Also, similarly to the how there's support for including tests that
have no tags at all, there's support for including tests that are
not tagged with a specific keys.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

Let's make the parsing of those tags a utility function so that its
behavior can be tested in detail.

Also, instead of parsing it for every test in a test suite, let's
parse those only once, and using during the filter process.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

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

The "include_empty" parameter to test the filtering utility function
is meant to act when a test with no tags is part of a test suite that
also includes tests with tags.

There's a current corner case when filtering with tags that must not
be present and tests with no tags, when "include_empty" is active.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

In the previous addition to the AVOCADO_TEST_TAGS content, a new
Avocado test to be used on unittests was added, but it was not added
to assertions on some of the unittests.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

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

This change introduces two new utility functions related with
filesystems. One function returns all filesystem types supported by the
running kernel and the other returns the filesystem type of a given
mount point.
Signed-off-by: NPraveen K Pandey <praveen@linux.vnet.ibm.com>
Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

This basically allows us to use the same set interface for checking if
a tag is in a set (the dict keys), and at the same time, allows us to
have room for the values given to key:val tags.

For "flat" tags, that is, ones that do not contain any value, the
dictionary will just contain None, and for the key:val tags, the
dictionary will contain a set of the values.  The set is useful
because duplicate values add no value, and a common use case
is to have a base value at the class level, and have it extended
in the
Signed-off-by: NCleber Rosa <crosa@redhat.com>

So that a failure will be easier to spot.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Let's loop over the tags once, and create the "must" and "must not"
set of tags.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

added utility which takes as argument a file and pattern
and return true or false respectively when pattern matches in
file or not

>>> from avocado.utils import genio
>>> genio.find_pattern("/home/praveen/file1", "praveen")
True
>>> genio.find_pattern("/home/praveen/file1", "Praveen")
False
>>> genio.find_pattern("/home/praveen/file12", "Praveen")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "avocado/utils/genio.py", line 219, in find_pattern
    % (filename, pattern))
avocado.utils.genio.GenIOError: invalid file /home/praveen/file12 to match pattern Praveen
>>>
Signed-off-by: NPraveen K Pandey <praveen@linux.vnet.ibm.com>

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

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

Signed-off-by: NCaio Carrara <ccarrara@redhat.com>

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

Previous recursive implementation of waiting for children is not really
possible, because parent has to treat the children exit, otherwise we
get temporary zombie status until the parent process is resumed.

This implementation also uses recursive way of killing the children, but
instead of waiting for them to die it reports the killed processes and
only in the main function it resumes all of them and watches them to
die.

As a benefit people might now get list of killed processes as return.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

We do quite a few loops, let's use boolean instead of string comparison.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

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

When binding to 0.0.0.0 all local addresses are used and these are
not covered by "localhost" binding. Let's use "" which makes sure we can
bind on all local addresses (including eg. 192.168.122.1).
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Binding port won't complain when port already opened for UDP packets and
we are trying to bind it for TCP packets. Let's do a combination.

Note I also tried:

    tuple(getattr(socket, _) for _ in dir(socket)
          if _.startswith('SOCK_')

to cover all of the protocols, but all the time they just reported
incorrect combination so let's just stick with the most common ones.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The "is_port_free" currently only checks IPv4 connections which returns
free ports even when IPv6 connections are established. Usually people
want completelly free ports, not just for IPv4.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>