Cumulusci

Breaking changes:

* We changed the default path for the mapping file created by the ``generate_dataset_mapping`` task to ``datasets/mapping.yml`` so that it matches the defaults for ``extract_dataset`` and ``load_dataset``

* We changed the ``extract_dataset`` and ``load_dataset`` tasks to default to storing data in an SQL file, ``datasets/sample.sql``, instead of a binary SQLite database file.

Changes:

* ``run_tests`` can now detect and optionally retry two classes of concurrency issues with Apex unit tests. ``run_tests`` should always report an accurate total of test methods run, in parallel or serial mode.

* Added the task ``generate_data_dictionary``. This task indexes the fields and objects created in each GitHub release for the project and generates a data dictionary in CSV format.

* Added a ``devhub`` service. This can be used to switch a project to a non-default sfdx Dev Hub using ``cci service connect devhub --project``

* Added a predefined ``qa`` scratch org. It uses the same scratch org definition file as the ``dev`` org, but makes it easier to spin up a second org for QA purposes without needing to first create it using ``cci org scratch``.

* The ``database_url`` option for the ``extract_dataset`` and ``load_dataset`` tasks is no longer required. Either ``database_url`` or ``sql_path`` must be specified. If both are specified, the ``sql_path`` will be ignored.

* Developers can now directly execute CumulusCI from the Python command line using ``python -m cumulusci`` or ``python cumulusci/__main__.py``

Issues closed:

* A problem with how ``run_tests`` performed Apex test retries when ``retry_always`` is set to True has been corrected.

New features:

* After connecting an org with ``cci org connect``, the browser now shows the message
"Congratulations! Your authentication succeeded." instead of "OK"
* External GitHub sources can now specify ``release: latest``, ``release: latest_beta``,
or ``release: previous`` instead of a commit, branch, or tag.
* The ``execute_anon`` task has been revised to detect when a gack occurred during execution.

Issues closed:

* When importing a scratch org from sfdx using ``cci org import``, the org's ``days``
is now set correctly from the org's actual expiration date. (1101)
* The package API version from ``cumulusci.yml`` is now validated to make sure
it's in the "XX.0" format expected by the API. (1134)
* Fixed an error deploying new setting objects using the ``org_settings`` task in Winter '20.
* Fixed a bug in processing preflight check tasks for MetaDeploy.
* Fixed path handling in the ``update_admin_profile`` task when run in a cross-project flow.

Breaking changes:

* The ``metadeploy_publish`` task now requires setting ``-o publish True`` in order to automatically set the Version's is_listed flag to True. (This is backwards incompatible in order to provide a safer default.)

New features:

* Python 3.8 is now officially supported.

* Flows can now include tasks or flows from a different project. See [Using Tasks and Flows from a Different Project](https://cumulusci.readthedocs.io/en/latest/features.html#using-tasks-and-flows-from-a-different-project) for details.

* In the ``metadeploy_publish`` task it is now possible to specify a commit hash with ``-o commit [sha]``, instead of a tag. This is useful while MetaDeploy plans are in development.

* Bulk data:

* Added support for mapping Record Types between orgs (by Developer Name) during bulk data extract and load.
* Added support for Record Type mapping in the ``generate_dataset_mapping`` task.
* Added [new documentation](https://cumulusci.readthedocs.io/en/latest/bulk_data.html) for bulk data tasks.

* Robot Framework:

* The sample ``create_contact.robot`` test that is created when initializing a new project with ``cci project init`` now makes use of page objects.
* The page objects library has two new keywords, ``wait for modal`` and ``wait for page object``, which wait for a new page object to appear.
* ``cumulusci.robotframework.utils`` now has a decorator named ``capture_screenshot_on_error`` which can be used to automatically capture a screenshot when a keyword fails.
* Prior to this change, ``Go to page Detail Contact`` required you to use a keyword argument for the object id (eg: ``Go to page Detail Contact object_id=${object_id}``). You can now specify the object id as a positional parameter (eg: ``Go to page Detail Contact ${object_id}``).

* ``OrgConfig`` objects now have a ``latest_api_version`` property which can be used to check what Salesforce API version is available.

Issues closed:

* Updated the scratch org definition files generated by ``cci project init`` to the new recommended format for org settings. Thanks to umeditor for the fix.

* The ``create_unmanaged_ee_src`` task (part of the ``unmanaged_ee`` flow) has been revised to remove the Protected setting on Custom Objects, to ensure that projects using this setting can be deployed to an Enterprise Edition org.

* The Salesforce REST API client used by many tasks will now automatically retry requests on certain connection and HTTP errors.

* Fixed an issue where posts to the Metadata API could reuse an existing connection and get a connection reset error if Salesforce had closed the connection.

* Disabled use of PyOpenSSL by the Python requests library, since it is no longer needed in the versions of Python we support.

Issues closed:

* Fixed a bug in deploying email templates and dashboards that was introduced
in 3.0.1.
* Removed broken ``config_qa`` flow from the ``cci project init`` template.

New features:

* Added support for new metadata types when generating ``package.xml``
from a directory of metadata using the ``update_package_xml`` task.

* The ``ci_feature`` flow now supports generating change notes for a
parent feature branch's pull request from the notes on child pull requests.
The parent pull request description will be overwritten with the new notes
after a child branch is merged to the parent if the parent pull request has
a special label, ``Build Change Notes``.

* When running Apex tests with the ``run_tasks`` task, if there is a single
remaining class being run, its name will be logged.

* Apex test failures that happen while uploading a package are now logged.

* In the ``robot_libdoc`` task, wildcards can now be used in the ``path`` option.

* Added an ``org_settings`` task which can deploy scratch org settings
from a scratch org definition file.

Issues closed:

* Added a workaround for an issue where refreshing the access token for a sandbox
or scratch org could fail if the user's credentials were new and not fully propagated.

Breaking change:

* CumulusCI 3.0.0 removes support for Python 2 (which will reach end of life at the end of 2019). If you're still running Python 2 you can use an older version of CumulusCI, but we recommend upgrading to Python 3. See our [installation instructions](https://cumulusci.readthedocs.io/en/latest/install.html) for your platform.