Fixes#18544.
When a loop is over an empty list, the result is set to
{'skipped_reason': u'No items in the list', 'skipped': True, 'changed': False}
which means that accessing `hr._result['results']` throws a `KeyError`.
(cherry picked from commit 200d6bdb23)
Previous changes addressed a corner case, which unfortunately introduced
another bug. This patch adds a new flag to the host state (did_rescue) which
is set to true when the rescue portion of a block completes. This flag is
then checked in _check_failed_state() when the fail_state != FAILED_NONE.
This lead to the discovery of another bug - current strategies are not advancing
hosts to ITERATING_COMPLETE after doing a peek at the next task, leaving the
host state in the run_state of the final task. To address this, before gathering
the list of failed hosts in StrategyBase.run(), a final pass through the iterator
for all hosts is done to ensure each host is in its final state. This way, no
strategy derived from StrategyBase has to worry about it and it's handled.
Fixes#17983
(cherry picked from commit ca5b361ad8)
* Have template action plugin call do_template
Avoids all the magic done for 'inline templating' for ansible plays.
renamed _do_template to do_template in templar to make externally accessible.
fixes#18192
* added backwards compat as per feedback
(cherry picked from commit bd70397e24)
When loading an include statically, we previously were simply doing a
copy() of the TaskInclude object, which recurses up the parents creating
a new lineage of objects. This caused problems when used inside load_list_of_blocks
as the new parent Block of the new TaskInclude was not actually in the list
of blocks being operated on. In most circumstances, this did not cause a
problem as the new parent block was a proper copy, however when used in
combination with PlaybookInclude (which copies conditionals to the list of
blocks loaded) this untracked parent was not being properly updated, leading
to tasks being run improperly.
Fixes#18206
(cherry picked from commit 5b87951d6c)
In some situations, where the Base class defines an Attribute, the
BaseMeta class doesn't properly see the _get_parent_attribute or
_get_attr_<whatever> methods because of multiple layers of subclasses
(ie. Handler, which subclasses Task). This addresses that by merging
the __dict__ of the parent with the current classes __dict__ meaning
all future iterations see available special methods.
Fixes#18378
(cherry picked from commit 4794b5da45)
Text strings and byte strings both have a translate method but the byte
string version is harder to use. It requires a mapping of all 256 bytes
to a translation value. Text strings only require a mapping from the
characters that are changing to the new string. Switching to text
strings on both py2 and py3 allows us to state what we're getting rid of
simply without having to rely on the maketrans() helper function.
(cherry picked from commit ee14e0cc2a)
The traceback is the following:
Traceback (most recent call last):
File \"/tmp/ansible_8s0bj604/ansible_module_setup.py\", line 134, in <module>
main()
File \"/tmp/ansible_8s0bj604/ansible_module_setup.py\", line 126, in main
data = get_all_facts(module)
File \"/tmp/ansible_8s0bj604/ansible_modlib.zip/ansible/module_utils/facts.py\", line 3641, in get_all_facts
File \"/tmp/ansible_8s0bj604/ansible_modlib.zip/ansible/module_utils/facts.py\", line 3584, in ansible_facts
File \"/tmp/ansible_8s0bj604/ansible_modlib.zip/ansible/module_utils/facts.py\", line 1600, in populate
File \"/tmp/ansible_8s0bj604/ansible_modlib.zip/ansible/module_utils/facts.py\", line 1649, in get_memory_facts
TypeError: translate() takes exactly one argument (2 given)
And the swapctl output is this:
# /sbin/swapctl -sk
total: 83090 1K-blocks allocated, 0 used, 83090 available
The only use of the code is to remove prefix in case they are present, so just
replacing them with empty space is sufficient.
(cherry picked from commit df145df962)
* Fix bug (#18355) where encrypted inventories fail
This is first part of fix for #18355
* Make DataLoader._get_file_contents return bytes
The issue #18355 is caused by a change to inventory to
stop using _get_file_contents so that it can handle text
encoding itself to better protect against harmless text
encoding errors in ini files (invalid unicode text in
comment fields).
So this makes _get_file_contents return bytes so it and other
callers can handle the to_text().
The data returned by _get_file_contents() is now a bytes object
instead of a text object. The callers of _get_file_contents() have
been updated to call to_text() themselves on the results.
Previously, the ini parser attempted to work around
ini files that potentially include non-vailid unicode
in comment lines. To do this, it stopped using
DataLoader._get_file_contents() which does the decryption of
files if vault encrypted. It didn't use that because _get_file_contents
previously did to_text() on the read data itself.
_get_file_contents() returns a bytestring now, so ini.py
can call it and still special case ini file comments when
converting to_text(). That also means encrypted inventory files
are decrypted first.
Fixes#18355
(cherry picked from commit dd0189839e)
This allows meta refresh_inventory to work with relative paths
Added option to unfrackpath to not resolv symlinks
fixes#16857
(cherry picked from commit 8217c1c39c)
If there is an intermittent network failure, we might be trying to reach
an URL multiple times. Without this patch, we would be re-adding the same
certificate to the OpenSSL default context multiple times.
Normally, this is no big issue, as OpenSSL will just silently ignore them,
after registering the error in its own error stack.
However, when python-cryptography initializes, it verifies that the current
error stack of the default OpenSSL context is empty, which it no longer is
due to us adding the certificates multiple times.
This results in cryptography throwing an Unknown OpenSSL Error with details:
OpenSSLErrorWithText(code=185057381L, lib=11, func=124, reason=101,
reason_text='error:0B07C065:x509 certificate routines:X509_STORE_add_cert:cert already in hash table'),
Signed-off-by: Patrick Uiterwijk <puiterwijk@redhat.com>
(cherry picked from commit 77af3a68de)
if ANSIBLE_VAULT_PASSWORD_FILE is set, 'ansible-vault rekey myvault.yml'
will fail to prompt for the new vault password file, and will use
None.
Fix is to split out 'ask_vault_passwords' into 'ask_vault_passwords'
and 'ask_new_vault_passwords' to make the logic simpler. And then
make sure new_vault_pass is always set for 'rekey', and if not, then
call ask_new_vault_passwords() to set it.
ask_vault_passwords() would return values for vault_pass and new
vault_pass, and vault cli previously would not prompt for new_vault_pass
if there was a vault_pass set via a vault password file.
Fixes#18247
(cherry picked from commit 309f54b709)
In order to support legacy plugins, the following two method signatures
are allowed for `CallbackBase.v2_playbook_on_start`:
def v2_playbook_on_start(self):
def v2_playbook_on_start(self, playbook):
Previously, the logic to handle this divergence checked to see if the
callback plugin being called supported an argument named `playbook`
in its `v2_playbook_on_start` method. This was fragile in a few ways:
- if a plugin author did not use the literal `playbook` to name their
method argument, their plugin would not be called correctly
- if a plugin author wrapped their `v2_playbook_on_start` method and
by doing so changed the argspec to no longer expose an argument
with that literal name, their plugin would not be called correctly
In order to continue to support both types of callback for backwards
compatibility while making the call more robust for plugin authors,
the logic can be reversed in order to have a positive check for the old
method signature instead of a positive check for the new one.
Signed-off-by: Steve Kuznetsov <skuznets@redhat.com>
(cherry picked from commit 0bc35354ce)
Mitigate the effects of observing the ssh process still running
after seeing an EOF on stdout when using OpenSSH with
ControlPersist, since it does not close the stderr file descriptor
in this case.
(cherry picked from commit 679da00236)
As neon is derived from Ubuntu, ansible_os_family should have the value
"Debian" instead of "Neon". Add a test case for KDE neon and set
os_family correctly for it.
(cherry picked from commit 4ff8890ec1)
- When there is no file at the destination yet, we have no modification time for the `If-Modified-Since`-Header. In this case trust the cache to make the right decision to either serve a cached version or to refresh from origin. This should help with mass-deployment scenarios where you want to use a local cache to relieve your uplink.
- If you don't trust the cache to make the right decision you can still force it to refresh by providing the `force: yes` option.
(cherry picked from commit c05bad9f74)
This limitation of python-3.4 mkstemp() is the final reason we made
python-3.5 our minimum version. Since we know about it, give a nice
error to the user with a hint that Python3.4 could be the issue.
Fixes#18160
(cherry picked from commit fda933723c)