test_run_command_string_unsafe_with_pipe unit test failing in devel on Mac

[sivel]

Issue Type:

Bug Report

Ansible Version:

ansible 1.6 (devel f96568a) last updated 2014/03/17 07:40:34 (GMT -500)

Environment:

Control: Mac OS, virtualenv, python 2.7.5

Target: N/A

Summary:

======================================================================
FAIL: test_run_command_string_unsafe_with_pipe (TestModuleUtilsBasic.TestModuleUtilsBasic)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/matt/python_venvs/ansibledev/ansible/test/units/TestModuleUtilsBasic.py", line 100, in test_run_command_string_unsafe_with_pipe
    self.assertEqual(out, 'foo bar')
AssertionError: '-n foo bar\n' != 'foo bar'

----------------------------------------------------------------------

Steps To Reproduce:

1. Run unit tests on Mac

Expected Results:

Tests Pass

Actual Results:

test_run_command_string_unsafe_with_pipe is failing, and it appears to be a difference in how /bin/sh works on Mac with respect to echo.

A simple test of the following shows the behavior:

>>> p = subprocess.Popen('echo -n "foo bar"', stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True)
>>> p.communicate()
('-n foo bar\n', '')

Whereas specifying executable=/bin/bash gives what we expect:

>>> p = subprocess.Popen('echo -n "foo bar"', stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True, executable='/bin/bash')
>>> p.communicate()
('foo bar', '')

This appears to be a result of the /bin/sh in OS X having been configured with --enable-xpg-echo-default, so unless you disable xpg_echo, echo interprets escape sequences and does not support any options.

[sivel]

Probably just acceptable to make the following change:

diff --git a/test/units/TestModuleUtilsBasic.py b/test/units/TestModuleUtilsBasic.py
index 3d85b61..d1ee95a 100644
--- a/test/units/TestModuleUtilsBasic.py
+++ b/test/units/TestModuleUtilsBasic.py
@@ -95,9 +95,9 @@ class TestModuleUtilsBasic(unittest.TestCase):

     # test run_command with a command pipe (with both use_unsafe_shell=True|False)
     def test_run_command_string_unsafe_with_pipe(self):
-        (rc, out, err) = self.module.run_command('echo -n "foo bar" | cat', use_unsafe_shell=True)
+        (rc, out, err) = self.module.run_command('echo "foo bar" | cat', use_unsafe_shell=True)
         self.assertEqual(rc, 0)
-        self.assertEqual(out, 'foo bar')
+        self.assertEqual(out, 'foo bar\n')

     # test run_command with a shell redirect in (with both use_unsafe_shell=True|False)
     def test_run_command_string_unsafe_with_redirect_in(self):

Since this test wasn't meant to test echo -n functionality, I think the above is an ok solution. Perhaps we can create another test and see if it possible to fix the echo -n on a bash shell that has xpg_echo enabled.

[jctanner]

@sivel thanks!