* Add tests for check_mode at play and task level
These test inheritance of check_mode from the various levels (command
line, as a play attribute and as a task attribute) so they will be
useful for checking that the change to fieldattribute inheritance with
defaults works
* Add a sentinel object
The Sentinel object can be used in place of None when we need to mark an
entry as being special (usually used to mark something as not having
been set)
* Start of using a Sentinel object instead of None.
* Handle edge cases around use of Sentinel
* _get_parent_attribute needs to deal in Sentinel not None
* No need to special case any_errors_fatal in task.py any longer
* Handle more edge cases around Sentinel
* Use Sentinel instead of None in TaskInclude
* Update code to clarify the vars we are copying are class attrs
* Add changelog fragment
* Use a default of Sentinel for delegate_to, this also allows 'delegate_to: ~' now to unset inherited delegate_to
* Explain Sentinel stripping in _extend_value
* Fix ModuleArgsParser tests to compare with Sentinel
* Fixes for tasks inside of roles inheriting from play
* Remove incorrect note. ci_complete
* Remove commented code
* Move relative time handling to module_utils and rewrite it
* Fix cases with no seconds defined
* fix a small typo along the way
* add relative time handling to the ownca provider in openssl_certificate
* add initial integration test for relative time ownca
* quote the documentation to produce valid yaml
* move timespec conversion and validation to the init function
* fix small edge case in conversion function
* add relative timestamp handling to the selfsigned provider
* add get_relative_time_option
* add relative timestamp handling to valid_in
* pep8 fix indentation
* add quotes in error message
* add changelog fragment
* Update changelogs/fragments/50570-relative_time_crypto.yaml
Co-Authored-By: MarkusTeufelberger <mteufelberger@mgit.at>
* Added support for labels for Docker Network
* Fixed missing comma in Docker Network module
* Specified minimal versions for dependencies and added changelog
* Fixes based on suggestions after code review
* Added integration tests
* Add command option
* Fix test task name
* Changelog fragment
* Fix indendation
* Add version_added
* Better command description
* Test passing command as list
* Handle invalid command types
* Cleaner command type checking
* Check that all items in command list are strings
* Better error about command list items
* Better type checking errors
* Add tests for command type checking
* Update command tests
* Fix messed up rebase
* Fix crash when using configs parameter
* Don’t add difference for configs set to None
* Add changelog fragment
* Revert "Add changelog fragment"
This reverts commit 8df497e3881d976e7417d2662a167694fbc33147.
* Enable config tests
* Pass empty list of configs to docker
* Properly clean up after all tests
* Fix publish idempotency when mode is None
* Add changelog fragment
* Python 2.6 compat
* Use self.publish
* Check length of publish before comparing content
* Sort publish lists before zipping
* Enable publish tests
* python3 compat
* Don’t sort by mode as it is not safe
* Document publish suboptions and add them to args
* Add type to publish documentation
* Add choices to publish argument_spec suboptions
* Make tcp the default protocol
* Make documentation reflect protocol default
* Simplify setting mode
* Remove redundant string quoting
* Test order of publish
* Add comment about publish change detection
* Revert "Revert "openssl_csr: Allow to use cryptography as backend (#50324)""
This reverts commit bbd2e31e9f.
* Remove more complicated selection copy'n'pasted from openssl_privatekey.
* Add tests for backend selection.
* Add openssl_csr test for arbitrary string commonName.
* Allow to disable commonName -> SAN copying (fixes#36690).
* Fix subversion integration test on Fedora 29.
This upgrades the sqlite-libs and subversion packages to make sure
that the version of sqlite expected by subversion is installed.
* Fix compatibility with RHEL and CentOS.
* Corner case in which import_role would add another instance of a role with the same signature into roles: when it already existed there.
roles:
- name: a
tasks:
- import_role: name=a
would execute role 'a' 3 times instead of the intended 2 (x2 in roles: phase +1 in tasks:)
* added tests
* Further cleanup of integration test inventory.
* Preserve aci and msc inventory in template.
* Update ansible-test inventory template handling.
* Fix classification of inventory file.
* Integration tests now have their own list of allowed shebangs.
* Use `#!/usr/bin/env bash` instead of `#!/bin/bash`
since the location is different on various platforms.
Some integration test targets have dependencies on files outside
the `test/integration/targets/` directory tree. Changes to these
dependencies can result in unexpected test failures since they do
not trigger integration tests which depend on them.
* Log dependencies at verbosity level 4.
This makes it easier to debug target dependency issues.
* Scan symlinks for target dependencies.
Some test targets use symlinks to files in other test targets.
These dependencies were previously undetected. This could result in
changes made to dependencies without triggering the dependent tests.
* Track missing target deps with `needs/target/*`.
Some existing test targets have untracked dependencies on other
test targets. This can result in changes to those dependencies
not triggering their dependent tests, resulting in test failures
after a PR is merged.
This PR adds the appropriate `needs/target/*` aliases to track
those dependencies, along with appropriate processing in
ansible-test to handle the new aliases.
* Scan meta dependencies in script targets.
Script targets are often former role targets which were converted
to allow custom invocations of ansible-playbook. These targets still
have their meta dependencies, but they were not being detected.
This could result in changes to dependencies not triggering the
targets which depend on them.
When a user home dir is not created with `useradd`, the home dir will now
be created with umask from /etc/login.defs. Also fixed a bug in which
after a local user is deleted, and the same user exists in the central
user management system, the module would create that user's home.
* Added new vmware module for gathering facts for DRS groups
* Update lib/ansible/modules/cloud/vmware/vmware_drs_group_facts.py
* Updated doc
* Fixed integration test with mutual exclusive paramteres
Co-Authored-By: karstenjakobsen <karsten@karstenjakobsen.dk>
* Added new vmware module for creating DRS VM or Host groups in a given cluster
* Fixed typo in module name
* Added better docstrings. Fixed better messaging for existing groups. Added delete example.
* Update doc
* Update lib/ansible/modules/cloud/vmware/vmware_drs_group.py
* Updated logic, so result is populated with correct data even if no changes are made
* Update lib/ansible/modules/cloud/vmware/vmware_drs_group.py
* Improved performance by getting group_obj in init()
* Fixed syntax error and added group_name as required if state is
* Added state= to integration test
Co-Authored-By: karstenjakobsen <karsten@karstenjakobsen.dk>
* Allow to use cryptography as backend for openssl_csr.
* Use different curve.
* Adding changelog.
Includes changelog fragment for #49416, which didn't include one.
* Make TLS available for RabbitMQ
* Use correct path
* Include cleanup
* Also remove unused package
* Don't generate a TLS cert every time if one already has been generated
* Add newlines, clean up repository and clarify block purpose
* One too many new lines :)
* Make TLS certs static
* Move var_blending test inventory into test.
* Remove Amazon specific inventory entry for tests.
* Remove Azure specific inventory entry for tests.
* Move var_precedence test inventory into test.
* Move unicode test inventory into test.
* Remove unused inventory entry.
* Move gathering_facts test inventory into test.
* Move delegate_to test inventory into test.
* Clean up inventory for binary_modules test.
* Clean up integration test inventory.
* Add tests for WANT_JSON and old style modules
* quote source path
* Attempt to further appease tests
* Check for file and not just exists
* omg don't be dumb
* moar fixes
* shellcheck is the worst :)
* Test the custom modules for failure without arg files
Provide toggle flag to allow display of unreachable task to stderr
using default callback plugin.
Fixes: #48069
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
* fixing the exception import from tower modules
* Adding tests for checking tower modules are failing with correct msg
* fixed failing tests
* fixed failing test in tower_team
* Introduce 'insertbefore' and 'insertafter' to specify the position children have to be inserted.
* Added version_added to new options
* Xpath in example needs single quotes
* Added integration tests for insertafter and insertbefore
* Changed version_added to 2.8
* Improve cleanup.
* Add check mode tests.
Failing tests commented out; will be fixed in #49948.
* Add reload and reset tests.
* Add tests for other global state.
* Work around ufw bugs.
* Cloudscale integration test setup
CloudProvider and CloudEnvironment classes for Cloudscale integration
tests. This also contains a cloudscale_common role with common
variables for all tests.
* cloudscale_volume module
New cloud module to manage volumes on the cloudscale.ch IaaS service.
* new module: AIX LVM file system and NFS
This module creates, removes, mount and unmount LVM and NFS file system
for AIX using /etc/filesystems. For LVM file systems is also possible
to resize the file system.
* better parameters options structure
better parameters options structure
* Improved file system resize returns
Added better tratment for return codes for file system resize.
When a resize is not possible because no enough space on lv or
shrink is not allowed.
* improved doc and creation file system return code
- improved doc
- creation file system return code 10 was treated.
* Doc recomendations, dict result, line limit
- Added doc recomendations
- Changed return to dict results on main()
- Using 159 columns for code limit
* wrong changed return when file system is already
Fixed wrong changed return when file system is already mounted.
When the file system is already mounted the return for changed
is False.
* Fixed description and included playbook for tests
- Fixed description
- Included playbook for manual tests
* Various small bits to get this merged ASAP
* Rename test/legacy/aix_filesystem.yml to test/integration/targets/aix_filesystem/tasks/main.yml
Move integration test to its proper location
* Create aliases
* Fix CI issues
* urldecode filter for Jinja2
We needed this in order to deconstruct correct URLs using Jinja2.
And we might as well upstream this.
* Add integration tests
* Fixes for Python 3
* Add urlencode for older Jinja2
* add openssh_cert module
* fix some linter issues
* address review feedback
* add basis for integration tests
* change valid datetime formats
* handle principals as set not as list
* handle principals as set not as list
* handle principals as set not as list
* fix order of dnf api operations so transactions don't fail
Previously dnf.base.fill_sack() was called before
dnf.base.update_cache() which apparently breaks dnf transaction
logic as per https://bugzilla.redhat.com/show_bug.cgi?id=1658694Fixes#49060
Signed-off-by: Adam Miller <admiller@redhat.com>
* add changelog and test case
Signed-off-by: Adam Miller <admiller@redhat.com>
* MSC: Various bugfixes and features enhancements
This PR includes:
- Lookups of roles, labels and domains
- Auto-create new labels
- Improvements to comparing complex datastructures
- Force removal of sites
- Support non top-level queries
- Document internal functions
- Add parameter types to modules
- Fix documentation examples
- Improvements to idempotency wrt. returning changed
- Support site locations
- Update permission list
- Various improvements to integration tests
* Fix Ci issues
* Refactor integration test for lvg module to introduce grow/reduce test
* List correctly current PV in lvg module: fix lvg reduce
Previous behaviour was to only take into account PV passed in 'pvs'
argument. This lead to reduce not working as expecting:
* with state=present and list of wanted pvs, lvg found only the pvs to
add or already present and ignored the pv to remove (obviously absent
from the list of given PV)
* with state=absent and a pv to remove, lvg found that the remaining pvs
list is empty (ignoring possible other PV in the vg) and decides to
remove the vg entirely (as supposely no PV are left anymore to store
lvm metadata)
* Add changelog fragment
* Refactoring of cnos_interface module like what followed by other vendors
* To remove cnos-interface from E326 validation ignore list
* Effect of default parameters impacted UT
* Added organization in the scm_credential get
* Fallback looking for cred in project org
* Tests project with multi org credential
* Fixed CI issue
* Added changelog fragment
* Fix mandatory statement error for junos modules
Fixes#40267
* Add error regex in junos terminal plugin to error out
in case of commit fails
* If commit fails add logic to discard changes before existing
else next task will result in error
* Add integration test
* Minor update
* Add cryptography backend for openssl_privatekey.
* Adding ECC support.
No support for X25519 and X449, since they don't support serialization.
* Improve finterprint calculation to work with Python 3.
* Add fingerprint check.
* Fix typo.
* Use separate curve option for elliptic curves, and use type 'ECC'.
* Using curve names as defined in IANA registry.
* Bump minimal supported cryptography version. Older versions might work as well, but I couldn't test them.
* Improve documentation.
* Change test suite to fit expected behaviour
This reverts some changes from ansible/ansible@723daf3
If a line is found in the file, exactly or via regexp matching, it must
not be added again.
insertafter/insertbefore options are used only when a line is to be
inserted, to specify where it must be added.
* Implement the change in behaviour mentioned in the previous commit
* Fix comment to reflect what the code does
* Set the correct return message.
In these cases, the lines are added, not replaced.
* Add a changelog
* Allow tags to be templated from a variable. Fixes#49825
* Restore _load_tags to ensure we do csv tag splitting
* Add tests for csv tags and templated tags
* evaluate_tags doesn't need to accept strings, because _load_tags handles this
* Windows Privileges - moved util code to it's own C# util
* Rename Enabler class to PrivilegeEnabler to remove ambiguity
* rename Utils to PrivilegeUtil
* fix missing util name changes
* win become: refactor and add support for passwordless become
* make tests more stable
* fix up dep message for Load-CommandUtils
* Add further check for System impersonation token
* re-add support for become with accounts that have no password
* doc fixes and slight code improvements
* fix doc sanity issue
* Fix various bugs related in reboot
- Use format strings for consistency and improve debug log messages
- Use local variables instead of class attributes in order to be thread safe
- Run setup module to get distribution and version
- Run find module to get full path of shutdown command
- Use ansible_os_family and ansible_distribution to find commands and args
- Use same command for all Solaris/SunOS distributions
- Move delay calculations to properties
- Reliably check for module run failure
- Fix bug in run_test_command() that accidentally made the method work properly
- Use better exceptions rather than Exception
- Use dict literals rather than constructors
- Correct _check_delay() so it always returns a value, not None
- Don't store and return result in run_test_command() because it's not used anywhere
- add test for post reboot command that fails
- test negative values for delay parameters
Template lookup plugin now support variable_start_string and
variable_end_string, just like template module.
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
* read_csv: new module to read CSV files
* Add a doc reference to the csvfile lookup plugin
* Enable the use of custom dialect options
* Improve error handling
* Fix PEP8
* Fix more PEP8
* Simplify custom dialect code
* Add integration tests
* Fixes for CI
* Fix for python 2.6
* Add info about loop based on jinja2 loop var
* ansible_loop
* Update test count
* Add extended loop_control that defines whether ansible_loop should be added
* Extended needs to be defaulted
* Revert "Update test count"
This reverts commit f1e93ee469825f4cdcd90fb28667d29aa088275c.
* Add docs about loop_control.extended
* Add revindex and revindex0
* Document ansible_loop in special vars
* Add changelog fragment
* Add tests, change items to allitems so that dot notation works, fix logic error with previtem
Incorporating the abandoned work from PRs #35288 and #45552. Also adding in
the version checking from `docker_container.py`, which should be abstracted
out to `docker_common.py`.
* fix py3 scope for unique filter errors, enable filters integration tests on rhel8 beta
Signed-off-by: Adam Miller <admiller@redhat.com>
* add changelog
Signed-off-by: Adam Miller <admiller@redhat.com>
* create local subversion server for tests
* fix sanity issues
* don't touch system config when bringing up site
* removed original setup files
* fix opensuse config
* Changed $IpAddress to $Gateway
* Tweaked win_route unit tests
Checks to see if new static route has the correct gateway
* Create win_route.yaml
* Fixed incorrect variable name
* Add parameters to module vmware_guest for conversion of disk to thin or thick when vm is cloned or deployed with template
* unit test for convert clone vm
Co-Authored-By: chris93111 <christopheferreira@ymail.com>
* Create python3and2
Adding tests to see if dnf still works when python-2 and python-3 are installed.
* Update main.yml
Include the tests that run on python 2 and python 3, based on the package manager.
* Update dnf.py
Use python3-dnf by default, otherwise python2-dnf.
* Rename python3and2 to python3and2.yml
Rename.
* Support for pids_limit parameter in docker_container module
This add pids_limit parameter support in docker_container module
Fixes#43337
Signed-off-by: Akshay <akshay@localhost.localdomain>
* Add changelog for pids_limit parameter
Signed-off-by: Akshay Gaikwad <akgaikwad001@gmail.com>
* Remove unnecessary lines of code
The map is needed if the names are different.
Signed-off-by: Akshay Gaikwad <akgaikwad001@gmail.com>
* Update pids_limit option tests
It also run for docker-py < 1.10.0
Signed-off-by: Akshay Gaikwad <akgaikwad001@gmail.com>
* Only one exit point.
* Refactoring account handling.
* Add diff support for acme_account.
* Insert public_account_key into acme_account_facts result and into acme_account diff.
* Add changelog.
* Fix lookup_passwordstore test skipping.
Skip all of RHEL instead of specific versions.
Skip all of CentOS < 7 instead of specific versions.
This makes the test more robust when testing newer versions.
Tests could be executed on RHEL if EPEL was installed during the test.
* Added win_partition module
* Fixes sanity tests
* Requested changes minus changes in partition_size
* Removed trailing whitespace and fixed docstring
* Changes to parititon_size to allow variable units
* Stricter regex for partition_size
* Support UpdateService forceNewDeployment in ecs_service module
* Force update to be called if force_new_deployment set
* Fixes for review
* Add force_new_deployment option to ecs_service.py
cherrypicks changes from via/ansible
Adds tests for pull request #42518
fixes backwards compatability with boto<1.8.4
* change version_added to 2.8 for force_new_deployment
* remove extra lines from test
* remove more unnecessary whitespace
* Added feature health_check_grace_period_seconds to ecs_service, this time with a botocore version check and some initial testing
* Only set health_check_grace_period_seconds when loadbalancers are defined
* Removed leftover commas and fix in test
* Removed blank line
* Minor improvements for ecs_service module
* Removed default (30) for health_check_grace_period_seconds param
* Changed botocore version allowed to 1.8.20 for health check param.
* Fix empty healthcheck failure
* Add append_hash functionality to k8s module
append_hash adds a hash based on the contents of a ConfigMap
or Secret to the name - this enables immutable ConfigMaps and
Secrets.
* Provide k8s_config_resource_name plugin
The k8s_config_resource_name filter plugin provides a means of determining
the name of ConfigMaps and Secrets created with append_hash
* Add changelog fragment
* fix failing tests
* Update openshift version needed for append_hash
* win_mapped_drive - refactor module and docs
* Updated code to work with become and split tokens
* use win_credential_manager instead of cmdkey
* updated credential manager module name
* harden the system token impersonation process
* win_credential_manager: new module to manage credentials
* fix sanity issues and removed CredSSP references
* renamed module to win_credential
* fix typo on test variable
* fix sanity ignore line
* win_snmp: Initial commit
* win_snmp: Better handling of lists
* win_snmp: Documentation fixes
* win_snmp: Updated documentation to match parameters
* win_snmp: Added integration tests
* win_snmp: Fixed typo in test
* win_snmp: Adjusted parameter checks to match documentation
* win_snmp: Updated option descriptions to be full sentences
* win_snmp: Better type checking and output suppression
* win_snmp: Fixed unset managers and communities
* win_snmp: Fixed skipping default registry keys
* win_snmp: Migrated to using add/set/remove action from replace
* win_snmp: Fixed check mode
* win_snmp: Fixed setting action and documentation. Expanded tests.
* win_snmp: Efficiency changes and documentation cleanup
* Added example of explicitly setting an empty set of managers to
documentation
* Made sure set will only remove items if there is a list of items
provided. This list can be of length 0
* Improved efficiency in selecting next index for SNMP manager
* Updated tests
* win_snmp: Added output of permitted managers and community strings
* win_snmp: Documentation fix
AWS uses rule type, protocol, port range, and source as an idempotent identifier.
There can only be one rule with that unique combination. Rules that differ only by description are allowed but overwritten by AWS.
Add a test
Co-authored-by: Will Thames <will@thames.id.au>
* Make wait_for return matched groups defined in search_regex. Closes#25020.
* Fix formatting issues.
* Fix issues raised in review.
- Use output_dir instead of hardcoded /tmp for temp files
- Sleep for only 3s instead of 10s
- Revert indent change
* ucs_disk_group_policy module and integration tests
* Additional refactor based on review in other modules.
* Fix issue with automatic config and add virtual_drive config.
Integration tests added for automatic config and virtual_drive config.
* Code review updates (documentation items)
* update version added to 2.8
* Use expect module to copy files
* Remove old and redundant upgrade files
* Return error message instead of code
* Cleanup copy command code
* Fix force issue in nxos_install_os
* new nxos_install_os integration tests
* Uncomment transport tests
* Revert negative test change
* Remove combined option that is no longer required
* Make shippable happy
* Add n5k test files
* Check minimal API and docker-py versions for all docker_* tests.
* Improve docker_swarm creation/destruction for tests.
* Fail when conditions aren't met.
* Don't hardcode address for advertise_addr.
* Add difference tracking tool
* Improve --diff mode for docker_container.
* Improve diffs of sets by ordering the sets.
* Rewrite imports, get rid of HAS_DOCKER_PY_x variables and use docker_version instead.
* Rename container -> active (more generic).
* Add --diff for docker_volume. Change old diff output.
* Add --diff for docker_network. Change old diff output.
* Add --diff for docker_swarm_service.
* Add changelog.
* Add entry for porting guide on docker_network and docker_volume.
* port win-say to use CSharpUtil AnsibleBasic and add warning when requested voice not found
* win_say: fixes following code review: use C# style properties consistently; prefix changelog fragments
* fix invalid yaml in change log fragment
* win_say: fixes following code review: use generic module parameter validation where possible
* remove redundant setting of Result.changed to false, simplified some logic for readability.
* fix serialisation issue when message text is drawn from a file; allows tests to be run not in check mode and fix up some test descriptions
* UCS managed objects module for direct control of any object and properties.
* Avoid checks for parent info or passwords in property compares
* Planned for 2.8
* try except for imports
* Provide Kubernetes resource validation to k8s module
Use kubernetes-validate to validate Kubernetes resource
definitions against the published schema
* Additional tests for kubernetes-validate
* Improve k8s error messages on exceptions
Parse the response body for the message rather than returning
a JSON blob
If we've validated and there are warnings, return those too - they
can be more helpful
```
"msg": "Failed to patch object: {\"kind\":\"Status\",\"apiVersion\":\"v1\",\"metadata\":{},
\"status\":\"Failure\",\"message\":\"[pos 334]: json: decNum: got first char 'h'\",\"code\":500}\n",
```
vs
```
"msg": "Failed to patch object: [pos 334]: json: decNum: got first char 'h'\nresource
validation error at spec.replicas: 'hello' is not of type u'integer'",
```
* Update versions used
In particular openshift/origin:3.9.0
* Add changelog for k8s validate change
* -Change: Include dependency role names in `role_names`.
-Add: `play_role_names` magic variable to include only explicitly named roles (formerly `role_names`).
-Add: `dependent_role_names` magic variable to include all dependency names for all roles.
* -Change: use the ansible_ prefix for new magic variables.
-Change: keep `role_names` as a deprecated variable, using the old functionality.
* -Add: changelog fragment for the role_names rework.
* -Add: Tests for the role_names (and ansible_*role_names) special variables
* -Fix: resolve erroneous documentation snippet that was introduced after rebasing.
* -Fix: explicitly sort to ensure list comparison works in test.
* Improve code structure
* Add author for module
* Now returns some values
* Update module's metadata
* Copy test case of rabbitmq_lookup
* Add test cases for rabbitmq_vhost_limits
* Minor fixes in documentation
* Fix module's return values
* Refactor module
* Improve test case
* Revise English in documentation
* Disable returning values because it's useless & unnecessary
* Work on failures: E261: match PEP8 styles
* Work on failures: E312: add RETURN section in documentation
* Add integration tests for iam_policy
* Fix indentation and ignore errors during clean up
* Mark iam_policy integration tests as unsupported by CI
* Add policies to a temporary folder that is cleaned up
* Add tasks to verify that iam_policy can remove policies from users, roles, and groups
* RabbitMQ basic publisher
* Split out of a module_util. Preparing for binary posts.
* Can now send a file to the queue.
* Allowing an empty queue to be used so RabbitMQ returns a random queue.
* Added RETURN docstring.
* Updated and added tests. Now returns a dictionary with msg, content_type and queue published to.
* Extra tests and introduced a none url method of providing server host details.
* Added testing and errors for url/host parameters.
* Updating RETURN sample
* Added an image file for testing binary publishing.
* Minor changes to test.
* Added filename key/value to headers if a binary file is published.
* Adding ability to specify headers.
* Renaming to rabbitmq_publish
* Changed tests to reflect name, and, preparing for testing headers.
* Updated some documentation
* Minor pip install update
* Modifications after feedback.
* Updates based on feedback.
* Fixing pep8 issue.
* Updating module and module_util name to amqp.
* Reverting back to rabbitmq_publish naming.
* Minor addition to notes.
We don't need to test with `upgrade: dist`, since we're not trying
to test the `apt` module. We just need to make sure the hold set
by the `dpkg_selections` module is working.
This change will avoid updating all the packages on the system,
which is slow, unnecessary, and can cause the installed python
to be changed.
Alex Stephen