get_url fails on rhel5 hosts using ansible 2.0.2.0

[MichaelBaydoun]

ISSUE TYPE

• Bug Report

COMPONENT NAME

get_url

ANSIBLE VERSION

ansible 2.0.2.0
  config file = /home/icansible/.ansible.cfg
  configured module search path = Default w/o overrides

CONFIGURATION

[defaults]
callback_whitelist = profile_tasks,timer
forks = 50
host_key_checking = false
inventory = ~/ansible/hosts-dev
pattern = NONE
retry_files_enabled = false
transport = ssh

OS / ENVIRONMENT

control host rhel6
target hosts rhel5

SUMMARY

get_url fails with error on rhel5 targets when using 2.0.2.0, but works on rhel6 targets.
revert back to 2.0.1.0 and the get_url module works on both rhel5 and rhel6 targets

STEPS TO REPRODUCE

---
- hosts: someRhel5Server:someRhel6Server
  tasks:

    - name: demo the error
      get_url:
        url: "https://github.com/ansible/ansible/blob/devel/README.md"
        dest: /tmp/
        validate_certs: no

EXPECTED RESULTS

expected file to be downloaded to both target hosts

ACTUAL RESULTS

on the rhel5 target

fatal: [rhel5server]: FAILED! => {"changed": false, "dest": "/tmp/", "failed": true, "gid": 0, "group": "root", "mode": "01777", "msg": "Request failed", "owner": "root", "response": "An unknown error occurred: addinfourl instance has no attribute 'getcode'", "secontext": "system_u:object_r:tmp_t:s0", "size": 69632, "state": "directory", "status_code": -1, "uid": 0, "url": "https://github.com/ansible/ansible/blob/devel/README.md"}

on the rhel6 target

changed: [rhel6server]

Note: -vvvv and ANSIBLE_DEBUG=y produced no additional useful information

[MichaelBaydoun]

bisect on the main repository says this is the first bad commit.

7158eb489f1817adc01177f9d6d67417883e080b is the first bad commit
commit 7158eb489f1817adc01177f9d6d67417883e080b
Author: Matt Martz <matt@sivel.net>
Date:   Fri Feb 19 12:59:58 2016 -0600

    Add python info to facts

:040000 040000 16828d2099affa1b8afe79eda1667270e754dacf 8369b7e9a1ea0354c0fb890bf4d7edec93b5697c M    lib

tagging @sivel

[MichaelBaydoun]

Going back one more commit in the main repo and the problem is not present. The subrepo reference for both the breaking and working commit is the same

➜  ansible-project git:(7766a9a) git ls-tree devel:lib/ansible/modules          
100644 blob ae8ccff5952585ebf8134e2f7d09dbfb63772976    __init__.py
160000 commit 9be870bb2a5f277ac7ba68e87243b74dcc0ab1ac  core
160000 commit 2e09202aae9e8f7934e4b05a8305f7c1352a61e6  extras

[MichaelBaydoun]

I don't understand how changing facts could introduce this error ... but it appears to have done just that

➜  ansible-project git:(7158eb4) git diff HEAD^ HEAD
diff --git a/lib/ansible/module_utils/facts.py b/lib/ansible/module_utils/facts.py
index fd5729c..53c4133 100644
--- a/lib/ansible/module_utils/facts.py
+++ b/lib/ansible/module_utils/facts.py
@@ -166,6 +166,7 @@ class Facts(object):
             self.get_local_facts()
             self.get_env_facts()
             self.get_dns_facts()
+            self.get_python_facts()

     def populate(self):
         return self.facts
@@ -782,6 +783,21 @@ class Facts(object):
             pass
         return size_total, size_available

+    def get_python_facts(self):
+        self.facts['python'] = {
+            'version': {
+                'major': sys.version_info[0],
+                'minor': sys.version_info[1],
+                'micro': sys.version_info[2],
+                'releaselevel': sys.version_info[3],
+                'serial': sys.version_info[4]
+            },
+            'version_info': list(sys.version_info),
+            'executable': sys.executable,
+            'type': sys.subversion[0]
+        }
+
+
 class Hardware(Facts):
     """
     This is a generic Hardware subclass of Facts.  This should be further

[alikins]

I suspect it's ed35e8b that causes it

commit ed35e8bce21180529e72f5bbe1d853b67ddac870
Author: James Cammarata <jimi@sngx.net>
Date:   Mon Apr 18 13:02:57 2016 -0400

    Don't override the http status code when successful in urls.py

diff --git lib/ansible/module_utils/urls.py lib/ansible/module_utils/urls.py
index 5526921..0c00fe8 100644
--- lib/ansible/module_utils/urls.py
+++ lib/ansible/module_utils/urls.py
@@ -887,8 +887,7 @@ def fetch_url
                      url_password=password, http_agent=http_agent, force_basic_auth=force_basic_auth,
                      follow_redirects=follow_redirects)
         info.update(r.info())
-        info['url'] = r.geturl()  # The URL goes in too, because of redirects.
-        info.update(dict(msg="OK (%s bytes)" % r.headers.get('Content-Length', 'unknown'), status=200))
+        info.update(dict(msg="OK (%s bytes)" % r.headers.get('Content-Length', 'unknown'), url=r.geturl(), status=r.getcode()))
     except NoSSLError, e:
         distribution = get_distribution()
         if distribution is not None and distribution.lower() == 'redhat':

The docs at https://docs.python.org/2/library/urllib.html#urllib.urlopen

Changed in version 2.6: Added getcode() to returned object and support for the no_proxy environment variable.

RHEL5 is python2.4, so it does not have getcode().

paging @jimi-c, @sivel

[sivel]

@alikins it looks like we may have access to r.code back as far as 2.4.