There's a difference on how Python 2 and 3 represents a set(), which is true
for the set derived FilterSet.  It just happens the a tree fingerprint
depends on its environment, which depends on the FilterSet representation.

Let's provide a predictive __repr__ implementation, and consequently, a
predictive fingerprint for Trees across Python versions.

This fixes the following unittest failure on Python 3:

   FAIL: test_fingerprint (selftests.unit.test_tree.TreeNode)
   ----------------------------------------------------------------------
   Traceback (most recent call last):
     File "/home/cleber/src/avocado/avocado/selftests/unit/test_tree.py", line 63, in test_fingerprint
       "/foo{},{},FilterSet([]),FilterSet([])")
   AssertionError: '/foo{},{},FilterSet(),FilterSet()' != '/foo{},{},FilterSet([]),FilterSet([])'
   - /foo{},{},FilterSet(),FilterSet()
   + /foo{},{},FilterSet([]),FilterSet([])
   ?                     ++            ++
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

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

Count WARN in to calculate the success rate in the HTML report.
Signed-off-by: NAmador Pahim <apahim@redhat.com>

Signed-off-by: NAmador Pahim <apahim@redhat.com>

Signed-off-by: NAmador Pahim <apahim@redhat.com>

Recently we introduced "in-package-config" which is packaged by default
with Avocado and in the original version if exists the system-wide
config is not used at all. I think this is unexpected behavior as users
are used to set configs in /etc so this patch changes it to first use
in-package-config, then in-etc-config and last the user-config.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Unless a lower version is set as a requirement, the RPM building
process will attempt to download from PyPI, which is not allowed.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

In case no suitable jobdir location can be created (user basically can't
write anywhere) Avocado crashes without a meaningful message, let's
provide one.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The user-config location ("~/.config/avocado.conf") might not be
writable by users (usually when executing in docker without user
home-dir). Let's use the file if available, but don't fail in case we
fail to produce it.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Signed-off-by: NAmador Pahim <apahim@redhat.com>

Signed-off-by: NAmador Pahim <apahim@redhat.com>

Signed-off-by: NAmador Pahim <apahim@redhat.com>

Some tweaks here and there found during the upkeep.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The remote execution is an optional plugin, let's just mention the ways
and link the description to the remote_runners there.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Add the description of the recent changes of Avocado packaging to
LTS.next documentation.
Signed-off-by: NLukáš Doktor <ldoktor@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>

With the recent changes the "libexec" dir is only available when
installed via RPM. In order to have "libexec" files available
as standard package we have to include it in  the package_files
dir (avocado/). The benefit is that we can actually use package
resources to get the location in "avocado exec-path" command.

For packaging purposes let's simply symlink $libexec/avocado to the
packaged $avocado/libexec location so there is the usual location, but
keep using the python-location by default.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

With the python3 patches we have another core dependency on six. Let's
include it in setup.py as without it Avocado is unusable.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

In case on of the first cleanups fails the following ones are not
performed, leaving uncleaned files. Lets perform always all steps and
then check for occurred errors.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

PEP461 is only available on Python 3.5.  Our CI (Travis) is set to run
on Python 3.4, and there's no reason not to be compatible with those
older Python versions IMO.

Let's then use a syntax that will produce the same content, but will
work on all Python 3.x versions.

Reference: https://docs.python.org/3.5/whatsnew/3.5.htmlSigned-off-by: NCleber Rosa <crosa@redhat.com>

The prevailing coding style of Avocado (unittest.TestCase) tests is to
use the specialized assert methods, such as assertNotRegexpMatches.

Unfortunately, the regular expression related assertion methods have
been through a messy renaming.  In theory, assertNotRegexpMatches and
assertRegexpMatches should be available in Python 3.0 and 3.1, but not
in Python >= 3.2, where it got renamed to assertRegex and
assertNotRegex.

In practice, in my system with Python 3.6.3, I see:

>>> unittest.TestCase.assertNotRegexpMatches
<function TestCase._deprecate.<locals>.deprecated_func at 0x7f3b85b709d8>

But in Travis, with Python 3.4, it's not available:

ERROR: test_directives_regex (selftests.unit.test_safeloader.DocstringDirectives)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/build/avocado-framework/avocado/selftests/unit/test_safeloader.py", line 135, in test_directives_regex
    self.assertNotRegexpMatches(directive, safeloader.DOCSTRING_DIRECTIVE_RE)
AttributeError: 'DocstringDirectives' object has no attribute 'assertNotRegexpMatches'

Let's simplify things and just use a regex match and check for its
result.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Signed-off-by: NAmador Pahim <apahim@redhat.com>

Signed-off-by: NAmador Pahim <apahim@redhat.com>

We increased the failed_once threshold, let's see whether Travis will
behave better now.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The "realpath" is being deprecated in favor of "readlink -f", which
should mainly work the same way. This is required by Travis as it's
version of Ubuntu does not contain "realpath".
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

In Travis more tests are failing when running in parallel, but still 20
failed tests in parallel re-ran in series are faster than all tests in
series.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

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

* https://github.com/avocado-framework/avocado:
  utils.cpu: Update cpu info pattern for aarch64

arm64 or aarch64 pattern is not easy to identify in the
cpuinfo. For v7 there is model name in the cpuinfo and it
could be regonize as arm for 32 bit. But in v8 only could
found 'CPU architecture' in the cpuinfo and current code
will fail to match and return i386 which is wrong.

For fix this, if 'CPU architecture' found in cpuinfo, use
platform.machine to return the arch name.
Signed-off-by: NWayne Sun <gsun@redhat.com>

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

* https://github.com/avocado-framework/avocado:
  Packaging: add bash RPM sub package
  Packaging: remove man page sources from binary RPM packages
  Packaging: remove examples from Python side of installation
  Packaging: remove unused function _get_resource_files()
  Settings: remove variables from the main module namespace
  Packaging: exclude optional_plugins tests in addition to selftests
  Packaging: add missing configuration files for RPM packages
  Packaging: make config files data files of the avocado module

To remove some of the bloat (even if little) of the base
python-avocado package, let's move content that is not useful to
Python developers out of the way.

This should also make the experience of content more similar across
users of the Python module (say, from a `pip install`) and users of
the python-avocado package.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

There's little to no point in providing these files, since the
generated man pages are already packaged.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Leaving them only in the source distribution generation (if one runs
python setup.py sdist) and in the "example" RPM package.

Because the copying of those files won't be performed by the Python
module installation code, it must be done in the SPEC file.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

The main module namespace has a lot of variables which are only
used in the main class initialization method.  Let's avoid poluting
the module namespace and move those to where they're used.  If
we want to put variables at the global module namespace, we
should consider if they're useful to other modules, make them
all caps, and document them.

A nicer refactor could be done here, but IMO, this is still an
improvement.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

The selftests from the main source directory (and package) are already
excluded from being considered modules to be distributed.  Let's also
exclude the tests that have recently been moved to their own modules.

Without this change, the "tests" module will be installed in Python's
library location, and will contain the files from all of those
directories.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

These files weren't being installed by the setup.py code, and thus
ended up missing from RPM packages.
Signed-off-by: NCleber Rosa <crosa@redhat.com>