* [WIP] become plugins
Move from hardcoded method to plugins for ease of use, expansion and overrides
- load into connection as it is going to be the main consumer
- play_context will also use to keep backwards compat API
- ensure shell is used to construct commands when needed
- migrate settings remove from base config in favor of plugin specific configs
- cleanup ansible-doc
- add become plugin docs
- remove deprecated sudo/su code and keywords
- adjust become options for cli
- set plugin options from context
- ensure config defs are avaialbe before instance
- refactored getting the shell plugin, fixed tests
- changed into regex as they were string matching, which does not work with random string generation
- explicitly set flags for play context tests
- moved plugin loading up front
- now loads for basedir also
- allow pyc/o for non m modules
- fixes to tests and some plugins
- migrate to play objects fro play_context
- simiplify gathering
- added utf8 headers
- moved option setting
- add fail msg to dzdo
- use tuple for multiple options on fail/missing
- fix relative plugin paths
- shift from play context to play
- all tasks already inherit this from play directly
- remove obsolete 'set play'
- correct environment handling
- add wrap_exe option to pfexec
- fix runas to noop
- fixed setting play context
- added password configs
- removed required false
- remove from doc building till they are ready
future development:
- deal with 'enable' and 'runas' which are not 'command wrappers' but 'state flags' and currently hardcoded in diff subsystems
* cleanup
remove callers to removed func
removed --sudo cli doc refs
remove runas become_exe
ensure keyerorr on plugin
also fix backwards compat, missing method is attributeerror, not ansible error
get remote_user consistently
ignore missing system_tmpdirs on plugin load
correct config precedence
add deprecation
fix networking imports
backwards compat for plugins using BECOME_METHODS
* Port become_plugins to context.CLIARGS
This is a work in progress:
* Stop passing options around everywhere as we can use context.CLIARGS
instead
* Refactor make_become_commands as asked for by alikins
* Typo in comment fix
* Stop loading values from the cli in more than one place
Both play and play_context were saving default values from the cli
arguments directly. This changes things so that the default values are
loaded into the play and then play_context takes them from there.
* Rename BECOME_PLUGIN_PATH to DEFAULT_BECOME_PLUGIN_PATH
As alikins said, all other plugin paths are named
DEFAULT_plugintype_PLUGIN_PATH. If we're going to rename these, that
should be done all at one time rather than piecemeal.
* One to throw away
This is a set of hacks to get setting FieldAttribute defaults to command
line args to work. It's not fully done yet.
After talking it over with sivel and jimi-c this should be done by
fixing FieldAttributeBase and _get_parent_attribute() calls to do the
right thing when there is a non-None default.
What we want to be able to do ideally is something like this:
class Base(FieldAttributeBase):
_check_mode = FieldAttribute([..] default=lambda: context.CLIARGS['check'])
class Play(Base):
# lambda so that we have a chance to parse the command line args
# before we get here. In the future we might be able to restructure
# this so that the cli parsing code runs before these classes are
# defined.
class Task(Base):
pass
And still have a playbook like this function:
---
- hosts:
tasks:
- command: whoami
check_mode: True
(The check_mode test that is added as a separate commit in this PR will
let you test variations on this case).
There's a few separate reasons that the code doesn't let us do this or
a non-ugly workaround for this as written right now. The fix that
jimi-c, sivel, and I talked about may let us do this or it may still
require a workaround (but less ugly) (having one class that has the
FieldAttributes with default values and one class that inherits from
that but just overrides the FieldAttributes which now have defaults)
* Revert "One to throw away"
This reverts commit 23aa883cbed11429ef1be2a2d0ed18f83a3b8064.
* Set FieldAttr defaults directly from CLIARGS
* Remove dead code
* Move timeout directly to PlayContext, it's never needed on Play
* just for backwards compat, add a static version of BECOME_METHODS to constants
* Make the become attr on the connection public, since it's used outside of the connection
* Logic fix
* Nuke connection testing if it supports specific become methods
* Remove unused vars
* Address rebase issues
* Fix path encoding issue
* Remove unused import
* Various cleanups
* Restore network_cli check in _low_level_execute_command
* type improvements for cliargs_deferred_get and swap shallowcopy to default to False
* minor cleanups
* Allow the su plugin to work, since it doesn't define a prompt the same way
* Fix up ksu become plugin
* Only set prompt if build_become_command was called
* Add helper to assist connection plugins in knowing they need to wait for a prompt
* Fix tests and code expectations
* Doc updates
* Various additional minor cleanups
* Make doas functional
* Don't change connection signature, load become plugin from TaskExecutor
* Remove unused imports
* Add comment about setting the become plugin on the playcontext
* Fix up tests for recent changes
* Support 'Password:' natively for the doas plugin
* Make default prompts raw
* wording cleanups. ci_complete
* Remove unrelated changes
* Address spelling mistake
* Restore removed test, and udpate to use new functionality
* Add changelog fragment
* Don't hard fail in set_attributes_from_cli on missing CLI keys
* Remove unrelated change to loader
* Remove internal deprecated FieldAttributes now
* Emit deprecation warnings now
* allow shells to have per host options, remote_tmp
added language to shell
removed module lang setting from general as plugins have it now
use get to avoid bad powershell plugin
more resilient tmp discovery, fall back to `pwd`
add shell to docs
fixed options for when frags are only options
added shell set ops in t_e and fixed option frags
normalize tmp dir usag4e
- pass tmpdir/tmp/temp options as env var to commands, making it default for tempfile
- adjusted ansiballz tmpdir
- default local tempfile usage to the configured local tmp
- set env temp in action
add options to powershell
shift temporary to internal envvar/params
ensure tempdir is set if we pass var
ensure basic and url use expected tempdir
ensure localhost uses local tmp
give /var/tmp priority, less perms issues
more consistent tempfile mgmt for ansiballz
made async_dir configurable
better action handling, allow for finally rm tmp
fixed tmp issue and no more tempdir in ballz
hostvarize world readable and admin users
always set shell tempdir
added comment to discourage use of exception/flow control
* Mostly revert expand_user as it's not quite working.
This was an additional feature anyhow.
Kept the use of pwd as a fallback but moved it to a second ssh
connection. This is not optimal but getting that to work in a single
ssh connection was part of the problem holding this up.
(cherry picked from commit 395b714120522f15e4c90a346f5e8e8d79213aca)
* fixed script and other action plugins
ensure tmpdir deletion
allow for connections that don't support new options (legacy, 3rd party)
fixed tests
* Update module_utils.six to latest
We've been held back on the version of six we could use on the module
side to 1.4.x because of python-2.4 compatibility. Now that our minimum
is Python-2.6, we can update to the latest version of six in
module_utils and get rid of the second copy in lib/ansible/compat.
* Replace pipes.quote for shlex_quote
* More migration of pipes.quote to shlex_quote
Note that we cannot yet move module code over. Modules have six-1.4
bundled which does not have shlex_quote. This shouldn't be a problem as
the function is still importable from pipes.quote. It's just that this
has become an implementation detail that makes us want to import from
shlex instead.
Once we get rid of the python2.4 dependency we can update to a newer
version of bundled six module-side and then we're free to use
shlex_quote everywhere.
This should fix issues with fish shell users as && and || are
not valid syntax, fish uses actual 'and' and 'or' programs.
Also updated to allow for fish backticks pushed quotes to subshell,
fish seems to handle spaces w/o them.
Lastly, removed encompassing subshell () for fish compatibility.
fixes#13199
Pipelining is a *significant* performance benefit, because each task can
be completed with a single SSH connection (vs. one ssh connection at the
start to mkdir, plus one sftp and one ssh per task).
Pipelining is disabled by default in Ansible because it conflicts with
the use of sudo if 'Defaults requiretty' is set in /etc/sudoers (as it
is on Red Hat) and su (which always requires a tty).
We can (and already do) make sudo/su happy by using "ssh -t" to allocate
a tty, but then the python interpreter goes into interactive mode and is
unhappy with module source being written to its stdin, per the following
comment from connections/ssh.py:
# we can only use tty when we are not pipelining the modules.
# piping data into /usr/bin/python inside a tty automatically
# invokes the python interactive-mode but the modules are not
# compatible with the interactive-mode ("unexpected indent"
# mainly because of empty lines)
Instead of the (current) drastic solution of turning off pipelining when
we use a tty, we can instead use a tty but suppress the behaviour of the
Python interpreter to switch to interactive mode. The easiest way to do
this is to make its stdin *not* be a tty, e.g. with cat|python.
This works, but there's a problem: ssh will ignore -t if its input isn't
really a tty. So we could open a pseudo-tty and use that as ssh's stdin,
but if we then write Python source into it, it's all echoed back to us
(because we're a tty). So we have to use -tt to force tty allocation; in
that case, however, ssh puts the tty into "raw" mode (~ICANON), so there
is no good way for the process on the other end to detect EOF on stdin.
So if we do:
echo -e "print('hello world')\n"|ssh -tt someho.st "cat|python"
…it hangs forever, because cat keeps on reading input even after we've
closed our pipe into ssh's stdin. We can get around this by writing a
special __EOF__ marker after writing in_data, and doing this:
echo -e "print('hello world')\n__EOF__\n"|ssh -tt someho.st "sed -ne '/__EOF__/q' -e p|python"
This works fine, but in fact I use a clever python one-liner by mgedmin
to achieve the same effect without depending on sed (at the expense of a
much longer command line, alas; Python really isn't one-liner-friendly).
We also enable pipelining by default as a consequence.
Matt Martz
Brian Coca
Dag Wieers
Toshio Kuratomi
Brian Coca
Toshio Kuratomi
nitzmahone
Abhijit Menon-Sen
James Cammarata
Marius Gedminas
Lukas Pirl