test_run_sudo() actually needs sudo, because it attempts to run it.
Let's check for the file at the usual location, and skip the test
unless it exists.

   ERROR: test_run_sudo (selftests.unit.test_utils_process.TestProcessRun)
   ----------------------------------------------------------------------
   Traceback (most recent call last):
     File "/home/travis/virtualenv/python2.7_with_system_site_packages/local/lib/python2.7/site-packages/mock/mock.py", line 1305, in patched
       return func(*args, **keywargs)
     File "selftests/unit/test_utils_process.py", line 180, in test_run_sudo
       p = process.run(cmd='ls -l', sudo=True, ignore_status=True)
     File "avocado/utils/process.py", line 1272, in run
       cmd_result = sp.run(timeout=timeout)
     File "avocado/utils/process.py", line 781, in run
       self._init_subprocess()
     File "avocado/utils/process.py", line 571, in _init_subprocess
       raise details
   OSError: [Errno 2] No such file or directory (/bin/sudo -n ls -l)
Signed-off-by: NCleber Rosa <crosa@redhat.com>

This change introduces the following requirement: the command to
be run by the avocado.utils.process functions should be given as unicode.
It brings a clearer interface, in which there's no need to deal with
encodings internally.

For simple test, it brings changes to respect the of passing unicode.
And, it changes the presentation of simple tests with unicode names
explicitly expreseed as such.

The way to deal with the process data that may be generated on
stdout/stderr is unchanged.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

On the basis of the non-existence of the "true" command on the system
running tests.  Most of the tests do not even reference the "true"
command, and those that do, do not require the binary to exist, but
only a valid command line to be given.

In that aspect, a fictional command is used where needed.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

The sudo related test are currently broken in the following aspects:

 1) it doesn't require TRUE_CMD
 2) the monkey patching of find_command() produces and adverse effect,
    and makes the binary we look for ("sudo") become "true"
 3) the expected results reflect the wrong sudo binary

Let's fix those aspects, making sure the process.py library looks for
sudo only, and that it prepends that (the found sudo binary) to the
generated command.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

And only if the method is not available, attempt to fallback to
the stdout itself if it is a string.

Before 30617c1e, the following kind of check was done:

   >>> type(bytes('foo')) in string_types
   False

And after it, the check was changed to:

   >>> isinstance(bytes('foo'), string_types)
   True

That made the bytes decoding not to be performed at times.  While the
new check is correct, we should favor the attempt to decode, and
only if not available, fallback to the original stdout if it looks
like a string.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Sometimes, the command to be executed is a string encoded a specific
encoding, different than the system or hard coded default.

Let's make the encoding configurable, and it will apply also to the
content produced by the command.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

The `path.find_command` raises exception in case it fails to find the
binary. Let's use the same approach as in
`selftests.functional.test_basic` to get binary or skip individual
tests.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

The FDDrainer can fail to flush due to some other conditions,
including when the stream handler is a plain StreamHandler,
and not a FileHandler.

Let's add tests that cover this situation, and another one that's
already fixed that deals with closed streams.

Reference: https://github.com/avocado-framework/avocado/pull/2512Signed-off-by: NCleber Rosa <crosa@redhat.com>

In situations where a user of the CmdResult class manually sets the
stdout/stderr attributes to a string, instead of the recommended and
documented content of type "bytes", it won't be possible to "decode"
the bytes into a string of a given encoding, and we'll end up with
a crash.

Since the goal of std{out,err}_text is to return a text version
of stdout, if itself already is of such a type, let's just return
it.

Additionally, if the data cannot be "decode()d", let's raise an
explicit TypeError on this location, rather than later in the code
when its value is attempted to be used.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

That's the only sane way to handle data, as bytes.  Let's not assume
any enconding on them.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

The FDDrainer reads from the PIPEs and writes to both the internal
StringIO instances and to stream logging handlers.  There are no
guarantees that the logging handlers will flush the content before the
FDDrainer finishes.

Let's explicitly close all the handlers (which should really be
FileHandlers) associated with the stream loggers, which should flush
and sync the content on the files themselves.

Additionally, the combined drainer was being missed from the flush
process, and was causing some output in that mode to be unavailable on
generated output.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

This is refactor of the current locks/file/threads created in the
SubProcess class to a more contained an isolated (and tested)
block of functionality.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

And raise an exception early if an invalid value is given.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Mock is slightly lighter than MagicMock, which is only necessary
when mocking "magic" methods.  So, let's revert to the lighter
Mock class.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

This module already uses mock, but doesn't import it in a way
that is usable on Python 3.  Also, let's make the style the
same as in the other modules that import conditionally for
Python 2 and 3.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Some Unixes won't have /bin/true, so let's make life
easier for those other OS to run the unittests.
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

The unittest standard library on Python 2.7 an later has everything
that the unittest2 backport is supposed to have.  Let's then drop all
the conditional imports of unittest2 and stick with unittest.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Recent pylint update is more pedantic about module level spacing. There
are no changes to code, only couple of extra spaces to make it happy.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This unittest patches find_command to return /usr/bin/sudo and then runs
it and expects some results. This works only with correctly set sudo,
but it does not work without sudo (as the binary is not found) or with
incorrectly set sudo. Let's use "/bin/true" instead.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

This function is useful, when one executes shell=True commands where
he sets values and then executes command. He can use this function
to get the actual binary path instead of the first variable=value
assignment.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

Some commands do require admin privileges to run, but we
can't use the 'root' user in some testing scenarios [1].

So let's add a 'sudo' parameter to process.run(), process.system()
and process.system_output() interfaces, that will append
a sudo command to the original command string, causing it
to be executed under sudo. The basic assumption here is
that the user executing the tests was added to the sudoers
file in such a way that it can run administrative commands
without a password.

[1] One specific environment where we can't use the root
user is when avocado runs on an AWS (Amazon Web Services)
instance.
Signed-off-by: NLucas Meneghel Rodrigues <lookkas@gmail.com>

shlex.split() can fail on some strange commands (eg. on an unfinished
quotation). This patch fallbacks to simple string.split() in such cases,
which might return incorrect results, but is safer to use. Although it
logs warning about this notifying the user about possible issue.
Signed-off-by: NLukáš Doktor <ldoktor@redhat.com>

It has been reported that avocado crashes when a malformed command
is attempted to be executed, even when using `shell=True`.

The reason is that `shlex.split` will raise `ValueError` on those
commands. Since improving `shlex.split` may be too complex, and
possibly not the best place to "fix" this, let's just assume safely
that this command should be run on GDB.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

To be honest, our test code could still be kept in the same directories and
have the same names. But I think we can improve two things here:

1) Flatten a little bit the directory structure of selftests. Two path
components are being dropped here: "all" and "avocado". So that
"selftests/all/functional/avocado" becomes simply
"selftests/functional".

2) File names match what is, by default, recognized by unittest (the
Standard Library module) based discovery of tests. That means that
doc_build_test.py becomes test_doc_build.py. Not a big deal IMHO.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

Instead of having a "common ground" that two namespaces would import and
share runtime information, let's just make the plugin set the defaults
for the module.

This is quite appropriate, both regarding the namespace where the
configurations are kept (in the avocado.utils.gdb module itself) and
also regarding the single direction of module layering.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

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

Self test files used to be executable, but that changed on 143b574c.
So, it doesn't make much sense to keep the shebangs + python commands
on them.
Signed-off-by: NCleber Rosa <crosa@redhat.com>

So that any code using our process API can be run through a debugger.
Signed-off-by: NCleber Rosa <crosa@redhat.com>