Configure 'test' workflow to cancel running jobs for the branch

This avoids testing twice when pushing to the main repository, once for the push, and another for a PR.

.github/workflows/deploy.yml

@@ -0,0 +1,47 @@
+name: deploy
+
+on:
+  workflow_dispatch:
+    inputs:
+      version:
+        description: 'Release version'
+        required: true
+        default: '1.2.3'
+
+jobs:
+
+  package:
+    runs-on: ubuntu-latest
+    env:
+      SETUPTOOLS_SCM_PRETEND_VERSION: ${{ github.event.inputs.version }}
+
+    steps:
+    - uses: actions/checkout@v3
+
+    - name: Build and Check Package
+      uses: hynek/build-and-inspect-python-package@v1.5
+
+  deploy:
+    needs: package
+    runs-on: ubuntu-latest
+    environment: deploy
+    permissions:
+      id-token: write  # For PyPI trusted publishers.
+      contents: write  # For tag.
+
+    steps:
+    - uses: actions/checkout@v3
+
+    - name: Download Package
+      uses: actions/download-artifact@v3
+      with:
+        name: Packages
+        path: dist
+
+    - name: Publish package to PyPI
+      uses: pypa/gh-action-pypi-publish@v1.8.5
+
+    - name: Push tag
+      run: |
+        git tag --annotate --message=v${{ github.event.inputs.version }} v${{ github.event.inputs.version }} ${{ github.sha }}
+        git push origin v${{ github.event.inputs.version }}

.github/workflows/main.yml

@@ -1,81 +0,0 @@
-name: build
-
-on: [push, pull_request]
-
-jobs:
-  build:
-
-    runs-on: ${{ matrix.os }}
-
-    strategy:
-      fail-fast: false
-      matrix:
-        tox_env:
-          - "py36-pytestlatest"
-          - "py37-pytestlatest"
-          - "py38-pytestlatest"
-          - "py39-pytestlatest"
-          - "py310-pytestlatest"
-          - "py38-pytestmain"
-          - "py38-psutil"
-          - "py38-setproctitle"
-
-        os: [ubuntu-latest, windows-latest]
-        include:
-          - tox_env: "py36-pytestlatest"
-            python: "3.6"
-          - tox_env: "py37-pytestlatest"
-            python: "3.7"
-          - tox_env: "py38-pytestlatest"
-            python: "3.8"
-          - tox_env: "py39-pytestlatest"
-            python: "3.9"
-          - tox_env: "py310-pytestlatest"
-            python: "3.10-dev"
-          - tox_env: "py38-pytestmain"
-            python: "3.8"
-          - tox_env: "py38-psutil"
-            python: "3.8"
-          - tox_env: "py38-setproctitle"
-            python: "3.8"
-
-    steps:
-    - uses: actions/checkout@v2
-    - name: Set up Python
-      uses: actions/setup-python@v2
-      with:
-        python-version: ${{ matrix.python }}
-    - name: Install tox
-      run: |
-        python -m pip install --upgrade pip
-        pip install tox
-    - name: Test
-      run: |
-        tox -e ${{ matrix.tox_env }}
-
-  deploy:
-
-    if: github.event_name == 'push' && startsWith(github.event.ref, 'refs/tags')
-
-    runs-on: ubuntu-latest
-
-    needs: build
-
-    steps:
-    - uses: actions/checkout@v2
-    - name: Set up Python
-      uses: actions/setup-python@v2
-      with:
-        python-version: "3.7"
-    - name: Install wheel
-      run: |
-        python -m pip install --upgrade pip
-        pip install wheel
-    - name: Build package
-      run: |
-        python setup.py sdist bdist_wheel
-    - name: Publish package to PyPI
-      uses: pypa/gh-action-pypi-publish@master
-      with:
-        user: __token__
-        password: ${{ secrets.pypi_token }}

.github/workflows/test.yml

@@ -0,0 +1,81 @@
+name: test
+
+on:
+  push:
+    branches:
+      - master
+      - "test-me-*"
+
+  pull_request:
+    branches:
+      - "*"
+
+# Cancel running jobs for the same workflow and branch.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+
+  check-package:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v3
+    - name: Build and Check Package
+      uses: hynek/build-and-inspect-python-package@v1.5
+
+  test:
+
+    runs-on: ${{ matrix.os }}
+
+    strategy:
+      fail-fast: false
+      matrix:
+        tox_env:
+          - "py37-pytestlatest"
+          - "py38-pytestlatest"
+          - "py39-pytestlatest"
+          - "py310-pytestlatest"
+          - "py310-pytestmain"
+          - "py311-pytestlatest"
+          - "py311-pytestmain"
+          - "py310-psutil"
+          - "py310-setproctitle"
+
+        os: [ubuntu-latest, windows-latest]
+        include:
+          - tox_env: "py37-pytestlatest"
+            python: "3.7"
+          - tox_env: "py38-pytestlatest"
+            python: "3.8"
+          - tox_env: "py39-pytestlatest"
+            python: "3.9"
+          - tox_env: "py310-pytestlatest"
+            python: "3.10"
+          - tox_env: "py310-pytestmain"
+            python: "3.10"
+          - tox_env: "py311-pytestlatest"
+            python: "3.11"
+          - tox_env: "py311-pytestmain"
+            python: "3.11"
+          - tox_env: "py310-psutil"
+            python: "3.10"
+          - tox_env: "py310-setproctitle"
+            python: "3.10"
+
+    steps:
+    - uses: actions/checkout@v3
+      with:
+        # Needed to fetch tags, which are required by setuptools-scm.
+        fetch-depth: 0
+    - name: Set up Python
+      uses: actions/setup-python@v2
+      with:
+        python-version: ${{ matrix.python }}
+    - name: Install tox
+      run: |
+        python -m pip install --upgrade pip
+        pip install tox
+    - name: Test
+      run: |
+        tox -e ${{ matrix.tox_env }}

.pre-commit-config.yaml

@@ -1,22 +1,32 @@
 repos:
+-   repo: https://github.com/PyCQA/autoflake
+    rev: v2.1.1
+    hooks:
+    -   id: autoflake
+        args: ["--in-place", "--remove-unused-variables", "--remove-all-unused-imports"]
 -   repo: https://github.com/psf/black
-    rev: 21.12b0
+    rev: 23.3.0
     hooks:
     -   id: black
         args: [--safe, --quiet, --target-version, py35]
+-   repo: https://github.com/asottile/blacken-docs
+    rev: 1.13.0
+    hooks:
+    -   id: blacken-docs
+        additional_dependencies: [black==23.1.0]
 -   repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v4.0.1
+    rev: v4.4.0
     hooks:
     -   id: trailing-whitespace
     -   id: end-of-file-fixer
     -   id: check-yaml
     -   id: debug-statements
 -   repo: https://github.com/PyCQA/flake8
-    rev: 4.0.1
+    rev: 6.0.0
     hooks:
     -   id: flake8
 -   repo: https://github.com/asottile/pyupgrade
-    rev: v2.29.1
+    rev: v3.4.0
     hooks:
     -   id: pyupgrade
         args: [--py3-plus]
@@ -29,7 +39,7 @@ repos:
         language: python
         additional_dependencies: [pygments, restructuredtext_lint]
 -   repo: https://github.com/pre-commit/mirrors-mypy
-    rev: v0.910-1
+    rev: v1.2.0
     hooks:
     -   id: mypy
         files: ^(src/|testing/)

CHANGELOG.rst

@@ -1,6 +1,154 @@
+pytest-xdist 3.3.0 (2023-05-12)
+===============================
+
+Features
+--------
+
+- `#555 <https://github.com/pytest-dev/pytest-xdist/issues/555>`_: Improved progress output when collecting nodes to be less verbose.
+
+
+pytest-xdist 3.2.1 (2023-03-12)
+===============================
+
+Bug Fixes
+---------
+
+- `#884 <https://github.com/pytest-dev/pytest-xdist/issues/884>`_: Fixed hang in ``worksteal`` scheduler.
+
+
+pytest-xdist 3.2.0 (2023-02-07)
+===============================
+
+Improved Documentation
+----------------------
+
+- `#863 <https://github.com/pytest-dev/pytest-xdist/issues/863>`_: Document limitations for debugging due to standard I/O of workers not being forwarded. Also, mention remote debugging as a possible workaround.
+
+
+Features
+--------
+
+- `#855 <https://github.com/pytest-dev/pytest-xdist/issues/855>`_: Users can now configure ``load`` scheduling precision using ``--maxschedchunk`` command
+  line option.
+
+- `#858 <https://github.com/pytest-dev/pytest-xdist/issues/858>`_: New ``worksteal`` scheduler, based on the idea of `work stealing <https://en.wikipedia.org/wiki/Work_stealing>`_. It's similar to ``load`` scheduler, but it should handle tests with significantly differing duration better, and, at the same time, it should provide similar or better reuse of fixtures.
+
+
+Trivial Changes
+---------------
+
+- `#870 <https://github.com/pytest-dev/pytest-xdist/issues/870>`_: Make the tests pass even when ``$PYTEST_XDIST_AUTO_NUM_WORKERS`` is set.
+
+
+pytest-xdist 3.1.0 (2022-12-01)
+===============================
+
+Features
+--------
+
+- `#789 <https://github.com/pytest-dev/pytest-xdist/issues/789>`_: Users can now set a default distribution mode in their configuration file:
+
+  .. code-block:: ini
+
+      [pytest]
+      addopts = --dist loadscope
+
+- `#842 <https://github.com/pytest-dev/pytest-xdist/issues/842>`_: Python 3.11 is now officially supported.
+
+
+Removals
+--------
+
+- `#842 <https://github.com/pytest-dev/pytest-xdist/issues/842>`_: Python 3.6 is no longer supported.
+
+
+pytest-xdist 3.0.2 (2022-10-25)
+===============================
+
+Bug Fixes
+---------
+
+- `#813 <https://github.com/pytest-dev/pytest-xdist/issues/813>`_: Cancel shutdown when a crashed worker is restarted.
+
+
+Deprecations
+------------
+
+- `#825 <https://github.com/pytest-dev/pytest-xdist/issues/825>`_: The ``--rsyncdir`` command line argument and ``rsyncdirs`` config variable are deprecated.
+
+  The rsync feature will be removed in pytest-xdist 4.0.
+
+- `#826 <https://github.com/pytest-dev/pytest-xdist/issues/826>`_: The ``--looponfail`` command line argument and ``looponfailroots`` config variable are deprecated.
+
+  The loop-on-fail feature will be removed in pytest-xdist 4.0.
+
+
+Improved Documentation
+----------------------
+
+- `#791 <https://github.com/pytest-dev/pytest-xdist/issues/791>`_: Document the ``pytest_xdist_auto_num_workers`` hook.
+
+- `#796 <https://github.com/pytest-dev/pytest-xdist/issues/796>`_: Added known limitations section to documentation.
+
+- `#829 <https://github.com/pytest-dev/pytest-xdist/issues/829>`_: Document the ``-n logical`` option.
+
+
+Features
+--------
+
+- `#792 <https://github.com/pytest-dev/pytest-xdist/issues/792>`_: The environment variable ``PYTEST_XDIST_AUTO_NUM_WORKERS`` can now be used to
+  specify the default for ``-n auto`` and ``-n logical``.
+
+- `#812 <https://github.com/pytest-dev/pytest-xdist/issues/812>`_: Partially restore old initial batch distribution algorithm in ``LoadScheduling``.
+
+  pytest orders tests for optimal sequential execution - i. e. avoiding
+  unnecessary setup and teardown of fixtures. So executing tests in consecutive
+  chunks is important for optimal performance.
+
+  In v1.14, initial test distribution in ``LoadScheduling`` was changed to
+  round-robin, optimized for the corner case, when the number of tests is less
+  than ``2 * number of nodes``. At the same time, it became worse for all other
+  cases.
+
+  For example: if some tests use some "heavy" fixture, and these tests fit into
+  the initial batch, with round-robin distribution the fixture will be created
+  ``min(n_tests, n_workers)`` times, no matter how many other tests there are.
+
+  With the old algorithm (before v1.14), if there are enough tests not using
+  the fixture, the fixture was created only once.
+
+  So restore the old behavior for typical cases where the number of tests is
+  much greater than the number of workers (or, strictly speaking, when there
+  are at least 2 tests for every node).
+
+
+Removals
+--------
+
+- `#468 <https://github.com/pytest-dev/pytest-xdist/issues/468>`_: The ``--boxed`` command-line option has been removed. If you still need this functionality, install `pytest-forked <https://pypi.org/project/pytest-forked>`__ separately.
+
+
+Trivial Changes
+---------------
+
+- `#468 <https://github.com/pytest-dev/pytest-xdist/issues/468>`_: The ``py`` dependency has been dropped.
+
+- `#822 <https://github.com/pytest-dev/pytest-xdist/issues/822>`_: Replace internal usage of ``py.log`` with a custom solution (but with the same interface).
+
+- `#823 <https://github.com/pytest-dev/pytest-xdist/issues/823>`_: Remove usage of ``py._pydir`` as an rsync candidate.
+
+- `#824 <https://github.com/pytest-dev/pytest-xdist/issues/824>`_: Replace internal usages of ``py.path.local`` by ``pathlib.Path``.
+
+
 pytest-xdist 2.5.0 (2021-12-10)
 ===============================
 
+Deprecations and Removals
+-------------------------
+
+- `#468 <https://github.com/pytest-dev/pytest-xdist/issues/468>`_: The ``--boxed`` command line argument is deprecated. Install `pytest-forked <https://pypi.org/project/pytest-forked>`__ and use ``--forked`` instead. pytest-xdist 3.0.0 will remove the ``--boxed`` argument and ``pytest-forked`` dependency.
+
+
 Features
 --------
 

README.rst

@@ -13,7 +13,7 @@ pytest-xdist
     :alt: Python versions
     :target: https://pypi.python.org/pypi/pytest-xdist
 
-.. image:: https://github.com/pytest-dev/pytest-xdist/workflows/build/badge.svg
+.. image:: https://github.com/pytest-dev/pytest-xdist/workflows/test/badge.svg
     :target: https://github.com/pytest-dev/pytest-xdist/actions
 
 .. image:: https://img.shields.io/badge/code%20style-black-000000.svg
@@ -25,520 +25,9 @@ tests across multiple CPUs to speed up test execution::
     pytest -n auto
 
 With this call, pytest will spawn a number of workers processes equal to the number of available CPUs, and distribute
-the tests randomly across them. There is also a number of `distribution modes`_ to choose from.
+the tests randomly across them.
 
-**NOTE**: due to how pytest-xdist is implemented, the ``-s/--capture=no`` option does not work.
+Documentation
+=============
 
-.. contents:: **Table of Contents**
-
-Installation
-------------
-
-Install the plugin with::
-
-    pip install pytest-xdist
-
-
-To use ``psutil`` for detection of the number of CPUs available, install the ``psutil`` extra::
-
-    pip install pytest-xdist[psutil]
-
-
-Features
---------
-
-* Test run parallelization_: tests can be executed across  multiple CPUs or hosts.
-  This allows to speed up development or to use special resources of `remote machines`_.
-
-* ``--looponfail``: run your tests repeatedly in a subprocess.  After each run
-  pytest waits until a file in your project changes and then re-runs
-  the previously failing tests.  This is repeated until all tests pass
-  after which again a full run is performed.
-
-* `Multi-Platform`_ coverage: you can specify different Python interpreters
-  or different platforms and run tests in parallel on all of them.
-
-  Before running tests remotely, ``pytest`` efficiently "rsyncs" your
-  program source code to the remote place.
-  You may specify different Python versions and interpreters. It does not
-  installs/synchronize dependencies however.
-
-  **Note**: this mode exists mostly for backward compatibility, as modern development
-  relies on continuous integration for multi-platform testing.
-
-.. _parallelization:
-
-Running tests across multiple CPUs
-----------------------------------
-
-To send tests to multiple CPUs, use the ``-n`` (or ``--numprocesses``) option::
-
-    pytest -n 8
-
-Pass ``-n auto`` to use as many processes as your computer has CPU cores. This
-can lead to considerable speed ups, especially if your test suite takes a
-noticeable amount of time.
-
-The test distribution algorithm is configured with the ``--dist`` command-line option:
-
-.. _distribution modes:
-
-* ``--dist load`` **(default)**: Sends pending tests to any worker that is
-  available, without any guaranteed order.
-
-* ``--dist loadscope``: Tests are grouped by **module** for *test functions*
-  and by **class** for *test methods*. Groups are distributed to available
-  workers as whole units. This guarantees that all tests in a group run in the
-  same process. This can be useful if you have expensive module-level or
-  class-level fixtures. Grouping by class takes priority over grouping by
-  module.
-
-* ``--dist loadfile``: Tests are grouped by their containing file. Groups are
-  distributed to available workers as whole units. This guarantees that all
-  tests in a file run in the same worker.
-
-* ``--dist loadgroup``: Tests are grouped by the ``xdist_group`` mark. Groups are
-  distributed to available workers as whole units. This guarantees that all
-  tests with same ``xdist_group`` name run in the same worker.
-
-  .. code-block:: python
-
-      @pytest.mark.xdist_group(name="group1")
-      def test1():
-          pass
-
-      class TestA:
-          @pytest.mark.xdist_group("group1")
-          def test2():
-              pass
-
-  This will make sure ``test1`` and ``TestA::test2`` will run in the same worker.
-  Tests without the ``xdist_group`` mark are distributed normally as in the ``--dist=load`` mode.
-
-* ``--dist no``: The normal pytest execution mode, runs one test at a time (no distribution at all).
-
-
-Running tests in a Python subprocess
-------------------------------------
-
-To instantiate a ``python3.9`` subprocess and send tests to it, you may type::
-
-    pytest -d --tx popen//python=python3.9
-
-This will start a subprocess which is run with the ``python3.9``
-Python interpreter, found in your system binary lookup path.
-
-If you prefix the --tx option value like this::
-
-    --tx 3*popen//python=python3.9
-
-then three subprocesses would be created and tests
-will be load-balanced across these three processes.
-
-.. _boxed:
-
-Running tests in a boxed subprocess
------------------------------------
-
-This functionality has been moved to the
-`pytest-forked <https://github.com/pytest-dev/pytest-forked>`_ plugin, but the ``--boxed`` option
-is still kept for backward compatibility.
-
-.. _`remote machines`:
-
-Sending tests to remote SSH accounts
-------------------------------------
-
-Suppose you have a package ``mypkg`` which contains some
-tests that you can successfully run locally. And you
-have a ssh-reachable machine ``myhost``.  Then
-you can ad-hoc distribute your tests by typing::
-
-    pytest -d --tx ssh=myhostpopen --rsyncdir mypkg mypkg
-
-This will synchronize your :code:`mypkg` package directory
-to a remote ssh account and then locally collect tests
-and send them to remote places for execution.
-
-You can specify multiple :code:`--rsyncdir` directories
-to be sent to the remote side.
-
-.. note::
-
-  For pytest to collect and send tests correctly
-  you not only need to make sure all code and tests
-  directories are rsynced, but that any test (sub) directory
-  also has an :code:`__init__.py` file because internally
-  pytest references tests as a fully qualified python
-  module path.  **You will otherwise get strange errors**
-  during setup of the remote side.
-
-
-You can specify multiple :code:`--rsyncignore` glob patterns
-to be ignored when file are sent to the remote side.
-There are also internal ignores: :code:`.*, *.pyc, *.pyo, *~`
-Those you cannot override using rsyncignore command-line or
-ini-file option(s).
-
-
-Sending tests to remote Socket Servers
---------------------------------------
-
-Download the single-module `socketserver.py`_ Python program
-and run it like this::
-
-    python socketserver.py
-
-It will tell you that it starts listening on the default
-port.  You can now on your home machine specify this
-new socket host with something like this::
-
-    pytest -d --tx socket=192.168.1.102:8888 --rsyncdir mypkg mypkg
-
-
-.. _`atonce`:
-.. _`Multi-Platform`:
-
-
-Running tests on many platforms at once
----------------------------------------
-
-The basic command to run tests on multiple platforms is::
-
-    pytest --dist=each --tx=spec1 --tx=spec2
-
-If you specify a windows host, an OSX host and a Linux
-environment this command will send each tests to all
-platforms - and report back failures from all platforms
-at once. The specifications strings use the `xspec syntax`_.
-
-.. _`xspec syntax`: https://codespeak.net/execnet/basics.html#xspec
-
-.. _`socketserver.py`: https://raw.githubusercontent.com/pytest-dev/execnet/master/execnet/script/socketserver.py
-
-.. _`execnet`: https://codespeak.net/execnet
-
-
-When tests crash
-----------------
-
-If a test crashes a worker, pytest-xdist will automatically restart that worker
-and report the test’s failure. You can use the ``--max-worker-restart`` option
-to limit the number of worker restarts that are allowed, or disable restarting
-altogether using ``--max-worker-restart 0``.
-
-
-How-tos
--------
-
-Identifying the worker process during a test
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-*New in version 1.15.*
-
-If you need to determine the identity of a worker process in
-a test or fixture, you may use the ``worker_id`` fixture to do so:
-
-.. code-block:: python
-
-    @pytest.fixture()
-    def user_account(worker_id):
-        """ use a different account in each xdist worker """
-        return "account_%s" % worker_id
-
-When ``xdist`` is disabled (running with ``-n0`` for example), then
-``worker_id`` will return ``"master"``.
-
-Worker processes also have the following environment variables
-defined:
-
-* ``PYTEST_XDIST_WORKER``: the name of the worker, e.g., ``"gw2"``.
-* ``PYTEST_XDIST_WORKER_COUNT``: the total number of workers in this session,
-  e.g., ``"4"`` when ``-n 4`` is given in the command-line.
-
-The information about the worker_id in a test is stored in the ``TestReport`` as
-well, under the ``worker_id`` attribute.
-
-Since version 2.0, the following functions are also available in the ``xdist`` module:
-
-.. code-block:: python
-
-    def is_xdist_worker(request_or_session) -> bool:
-        """Return `True` if this is an xdist worker, `False` otherwise
-
-        :param request_or_session: the `pytest` `request` or `session` object
-        """
-
-     def is_xdist_controller(request_or_session) -> bool:
-        """Return `True` if this is the xdist controller, `False` otherwise
-
-        Note: this method also returns `False` when distribution has not been
-        activated at all.
-
-        :param request_or_session: the `pytest` `request` or `session` object
-        """
-
-    def is_xdist_master(request_or_session) -> bool:
-        """Deprecated alias for is_xdist_controller."""
-
-    def get_xdist_worker_id(request_or_session) -> str:
-        """Return the id of the current worker ('gw0', 'gw1', etc) or 'master'
-        if running on the controller node.
-
-        If not distributing tests (for example passing `-n0` or not passing `-n` at all)
-        also return 'master'.
-
-        :param request_or_session: the `pytest` `request` or `session` object
-        """
-
-
-Identifying workers from the system environment
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-*New in version 2.4*
-
-If the `setproctitle`_ package is installed, ``pytest-xdist`` will use it to
-update the process title (command line) on its workers to show their current
-state.  The titles used are ``[pytest-xdist running] file.py/node::id`` and
-``[pytest-xdist idle]``, visible in standard tools like ``ps`` and ``top`` on
-Linux, Mac OS X and BSD systems.  For Windows, please follow `setproctitle`_'s
-pointer regarding the Process Explorer tool.
-
-This is intended purely as an UX enhancement, e.g. to track down issues with
-long-running or CPU intensive tests.  Errors in changing the title are ignored
-silently.  Please try not to rely on the title format or title changes in
-external scripts.
-
-.. _`setproctitle`: https://pypi.org/project/setproctitle/
-
-
-Uniquely identifying the current test run
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-*New in version 1.32.*
-
-If you need to globally distinguish one test run from others in your
-workers, you can use the ``testrun_uid`` fixture. For instance, let's say you
-wanted to create a separate database for each test run:
-
-.. code-block:: python
-
-    import pytest
-    from posix_ipc import Semaphore, O_CREAT
-
-    @pytest.fixture(scope="session", autouse=True)
-    def create_unique_database(testrun_uid):
-        """ create a unique database for this particular test run """
-        database_url = f"psql://myapp-{testrun_uid}"
-
-        with Semaphore(f"/{testrun_uid}-lock", flags=O_CREAT, initial_value=1):
-            if not database_exists(database_url):
-                create_database(database_url)
-
-    @pytest.fixture()
-    def db(testrun_uid):
-        """ retrieve unique database """
-        database_url = f"psql://myapp-{testrun_uid}"
-        return database_get_instance(database_url)
-
-
-Additionally, during a test run, the following environment variable is defined:
-
-* ``PYTEST_XDIST_TESTRUNUID``: the unique id of the test run.
-
-Accessing ``sys.argv`` from the controller node in workers
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-To access the ``sys.argv`` passed to the command-line of the controller node, use
-``request.config.workerinput["mainargv"]``.
-
-
-Specifying test exec environments in an ini file
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-You can use pytest's ini file configuration to avoid typing common options.
-You can for example make running with three subprocesses your default like this:
-
-.. code-block:: ini
-
-    [pytest]
-    addopts = -n3
-
-You can also add default environments like this:
-
-.. code-block:: ini
-
-    [pytest]
-    addopts = --tx ssh=myhost//python=python3.9 --tx ssh=myhost//python=python3.6
-
-and then just type::
-
-    pytest --dist=each
-
-to run tests in each of the environments.
-
-
-Specifying "rsync" dirs in an ini-file
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-In a ``tox.ini`` or ``setup.cfg`` file in your root project directory
-you may specify directories to include or to exclude in synchronisation:
-
-.. code-block:: ini
-
-    [pytest]
-    rsyncdirs = . mypkg helperpkg
-    rsyncignore = .hg
-
-These directory specifications are relative to the directory
-where the configuration file was found.
-
-.. _`pytest-xdist`: http://pypi.python.org/pypi/pytest-xdist
-.. _`pytest-xdist repository`: https://github.com/pytest-dev/pytest-xdist
-.. _`pytest`: http://pytest.org
-
-
-Making session-scoped fixtures execute only once
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-
-``pytest-xdist`` is designed so that each worker process will perform its own collection and execute
-a subset of all tests. This means that tests in different processes requesting a high-level
-scoped fixture (for example ``session``) will execute the fixture code more than once, which
-breaks expectations and might be undesired in certain situations.
-
-While ``pytest-xdist`` does not have a builtin support for ensuring a session-scoped fixture is
-executed exactly once, this can be achieved by using a lock file for inter-process communication.
-
-The example below needs to execute the fixture ``session_data`` only once (because it is
-resource intensive, or needs to execute only once to define configuration options, etc), so it makes
-use of a `FileLock <https://pypi.org/project/filelock/>`_ to produce the fixture data only once
-when the first process requests the fixture, while the other processes will then read
-the data from a file.
-
-Here is the code:
-
-.. code-block:: python
-
-    import json
-
-    import pytest
-    from filelock import FileLock
-
-
-    @pytest.fixture(scope="session")
-    def session_data(tmp_path_factory, worker_id):
-        if worker_id == "master":
-            # not executing in with multiple workers, just produce the data and let
-            # pytest's fixture caching do its job
-            return produce_expensive_data()
-
-        # get the temp directory shared by all workers
-        root_tmp_dir = tmp_path_factory.getbasetemp().parent
-
-        fn = root_tmp_dir / "data.json"
-        with FileLock(str(fn) + ".lock"):
-            if fn.is_file():
-                data = json.loads(fn.read_text())
-            else:
-                data = produce_expensive_data()
-                fn.write_text(json.dumps(data))
-        return data
-
-
-The example above can also be use in cases a fixture needs to execute exactly once per test session, like
-initializing a database service and populating initial tables.
-
-This technique might not work for every case, but should be a starting point for many situations
-where executing a high-scope fixture exactly once is important.
-
-
-How does xdist work?
---------------------
-
-``xdist`` works by spawning one or more **workers**, which are
-controlled by the **controller**. Each **worker** is responsible for
-performing a full test collection and afterwards running tests as
-dictated by the **controller**.
-
-The execution flow is:
-
-1. **controller** spawns one or more **workers** at the beginning of the
-   test session. The communication between **controller** and **worker**
-   nodes makes use of `execnet <https://codespeak.net/execnet/>`__ and
-   its
-   `gateways <https://codespeak.net/execnet/basics.html#gateways-bootstrapping-python-interpreters>`__.
-   The actual interpreters executing the code for the **workers** might
-   be remote or local.
-
-2. Each **worker** itself is a mini pytest runner. **workers** at this
-   point perform a full test collection, sending back the collected
-   test-ids back to the **controller** which does not perform any
-   collection itself.
-
-3. The **controller** receives the result of the collection from all
-   nodes. At this point the **controller** performs some sanity check to
-   ensure that all **workers** collected the same tests (including
-   order), bailing out otherwise. If all is well, it converts the list
-   of test-ids into a list of simple indexes, where each index
-   corresponds to the position of that test in the original collection
-   list. This works because all nodes have the same collection list, and
-   saves bandwidth because the **controller** can now tell one of the
-   workers to just *execute test index 3* index of passing the full test
-   id.
-
-4. If **dist-mode** is **each**: the **controller** just sends the full
-   list of test indexes to each node at this moment.
-
-5. If **dist-mode** is **load**: the **controller** takes around 25% of
-   the tests and sends them one by one to each **worker** in a round
-   robin fashion. The rest of the tests will be distributed later as
-   **workers** finish tests (see below).
-
-6. Note that ``pytest_xdist_make_scheduler`` hook can be used to
-   implement custom tests distribution logic.
-
-7. **workers** re-implement ``pytest_runtestloop``: pytest’s default
-   implementation basically loops over all collected items in the
-   ``session`` object and executes the ``pytest_runtest_protocol`` for
-   each test item, but in xdist **workers** sit idly waiting for
-   **controller** to send tests for execution. As tests are received by
-   **workers**, ``pytest_runtest_protocol`` is executed for each test.
-   Here it worth noting an implementation detail: **workers** always
-   must keep at least one test item on their queue due to how the
-   ``pytest_runtest_protocol(item, nextitem)`` hook is defined: in order
-   to pass the ``nextitem`` to the hook, the worker must wait for more
-   instructions from controller before executing that remaining test. If
-   it receives more tests, then it can safely call
-   ``pytest_runtest_protocol`` because it knows what the ``nextitem``
-   parameter will be. If it receives a “shutdown” signal, then it can
-   execute the hook passing ``nextitem`` as ``None``.
-
-8. As tests are started and completed at the **workers**, the results
-   are sent back to the **controller**, which then just forwards the
-   results to the appropriate pytest hooks: ``pytest_runtest_logstart``
-   and ``pytest_runtest_logreport``. This way other plugins (for example
-   ``junitxml``) can work normally. The **controller** (when in
-   dist-mode **load**) decides to send more tests to a node when a test
-   completes, using some heuristics such as test durations and how many
-   tests each **worker** still has to run.
-
-9. When the **controller** has no more pending tests it will send a
-   “shutdown” signal to all **workers**, which will then run their
-   remaining tests to completion and shut down. At this point the
-   **controller** will sit waiting for **workers** to shut down, still
-   processing events such as ``pytest_runtest_logreport``.
-
-FAQ
----
-
-**Question**: Why does each worker do its own collection, as opposed to having the
-controller collect once and distribute from that collection to the
-workers?
-
-If collection was performed by controller then it would have to
-serialize collected items to send them through the wire, as workers live
-in another process. The problem is that test items are not easily
-(impossible?) to serialize, as they contain references to the test
-functions, fixture managers, config objects, etc. Even if one manages to
-serialize it, it seems it would be very hard to get it right and easy to
-break by any small change in pytest.
+Documentation is available at `Read The Docs <https://pytest-xdist.readthedocs.io>`__.

RELEASING.rst

@@ -32,14 +32,8 @@ To publish a new release ``X.Y.Z``, the steps are as follows:
 
     $ tox -e release -- X.Y.Z
 
-#. Commit and push the branch for review.
+#. Commit and push the branch to ``upstream`` and open a PR.
 
-#. Once PR is **green** and **approved**, create and push a tag::
+#. Once the PR is **green** and **approved**, start the ``deploy`` workflow manually from the branch ``release-VERSION``, passing ``VERSION`` as parameter.
 
-    $ export VERSION=X.Y.Z
-    $ git tag v$VERSION release-$VERSION
-    $ git push git@github.com:pytest-dev/pytest-xdist.git v$VERSION
-
-That will build the package and publish it on ``PyPI`` automatically.
-
-#. Merge the release PR to `master`.
+#. Merge the release PR to ``master``.

changelog/468.deprecation.rst

@@ -1,3 +0,0 @@
-The ``--boxed`` commmand line argument is deprecated.
-Install pytest-forked and use ``--forked`` instead.
-pytest-xdist 3.0.0 will remove the ``--boxed`` argument and pytest-forked dependency.

docs/.gitignore

@@ -0,0 +1 @@
+_build/

docs/changelog.rst

@@ -0,0 +1,5 @@
+=========
+Changelog
+=========
+
+.. include:: ../CHANGELOG.rst

docs/conf.py

@@ -0,0 +1,55 @@
+# Configuration file for the Sphinx documentation builder.
+#
+# This file only contains a selection of the most common options. For a full
+# list see the documentation:
+# https://www.sphinx-doc.org/en/master/usage/configuration.html
+
+# -- Path setup --------------------------------------------------------------
+
+# If extensions (or modules to document with autodoc) are in another directory,
+# add these directories to sys.path here. If the directory is relative to the
+# documentation root, use os.path.abspath to make it absolute, like shown here.
+#
+# import os
+# import sys
+# sys.path.insert(0, os.path.abspath('.'))
+
+
+# -- Project information -----------------------------------------------------
+
+project = "pytest-xdist"
+copyright = "2022, holger krekel and contributors"
+author = "holger krekel and contributors"
+
+master_doc = "index"
+
+# -- General configuration ---------------------------------------------------
+
+# Add any Sphinx extension module names here, as strings. They can be
+# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
+# ones.
+extensions = [
+    "sphinx_rtd_theme",
+    "sphinx.ext.autodoc",
+]
+
+# Add any paths that contain templates here, relative to this directory.
+templates_path = ["_templates"]
+
+# List of patterns, relative to source directory, that match files and
+# directories to ignore when looking for source files.
+# This pattern also affects html_static_path and html_extra_path.
+exclude_patterns = ["_build", "Thumbs.db", ".DS_Store"]
+
+
+# -- Options for HTML output -------------------------------------------------
+
+# The theme to use for HTML and HTML Help pages.  See the documentation for
+# a list of builtin themes.
+#
+html_theme = "sphinx_rtd_theme"
+
+# Add any paths that contain custom static files (such as style sheets) here,
+# relative to this directory. They are copied after the builtin static files,
+# so a file named "default.css" will overwrite the builtin "default.css".
+# html_static_path = ['_static']

docs/crash.rst

@@ -0,0 +1,7 @@
+When tests crash
+================
+
+If a test crashes a worker, pytest-xdist will automatically restart that worker
+and report the test’s failure. You can use the ``--max-worker-restart`` option
+to limit the number of worker restarts that are allowed, or disable restarting
+altogether using ``--max-worker-restart 0``.

docs/distribution.rst

@@ -0,0 +1,95 @@
+.. _parallelization:
+
+Running tests across multiple CPUs
+==================================
+
+To send tests to multiple CPUs, use the ``-n`` (or ``--numprocesses``) option::
+
+    pytest -n auto
+
+This can lead to considerable speed ups, especially if your test suite takes a
+noticeable amount of time.
+
+With ``-n auto``, pytest-xdist will use as many processes as your computer
+has CPU cores.
+
+Use ``-n logical`` to use the number of *logical* CPU cores rather than
+physical ones. This currently requires the ``psutil`` package to be installed;
+if it is not, pytest-xdist will fall back to ``-n auto`` behavior.
+
+Pass a number, e.g. ``-n 8``, to specify the number of processes explicitly.
+
+To specify a different meaning for ``-n auto`` and ``-n logical`` for your
+tests, you can:
+
+* Set the environment variable ``PYTEST_XDIST_AUTO_NUM_WORKERS`` to the
+  desired number of processes.
+
+* Implement the ``pytest_xdist_auto_num_workers``
+  `pytest hook <https://docs.pytest.org/en/latest/how-to/writing_plugins.html>`__
+  (a ``pytest_xdist_auto_num_workers(config)`` function in e.g. ``conftest.py``)
+  that returns the number of processes to use.
+  The hook can use ``config.option.numprocesses`` to determine if the user
+  asked for ``"auto"`` or ``"logical"``, and it can return ``None`` to fall
+  back to the default.
+
+If both the hook and environment variable are specified, the hook takes
+priority.
+
+
+Parallelization can be configured further with these options:
+
+* ``--maxprocesses=maxprocesses``: limit the maximum number of workers to
+  process the tests.
+
+* ``--max-worker-restart``: maximum number of workers that can be restarted
+  when crashed (set to zero to disable this feature).
+
+The test distribution algorithm is configured with the ``--dist`` command-line option:
+
+.. _distribution modes:
+
+* ``--dist load`` **(default)**: Sends pending tests to any worker that is
+  available, without any guaranteed order. Scheduling can be fine-tuned with
+  the `--maxschedchunk` option, see output of `pytest --help`.
+
+* ``--dist loadscope``: Tests are grouped by **module** for *test functions*
+  and by **class** for *test methods*. Groups are distributed to available
+  workers as whole units. This guarantees that all tests in a group run in the
+  same process. This can be useful if you have expensive module-level or
+  class-level fixtures. Grouping by class takes priority over grouping by
+  module.
+
+* ``--dist loadfile``: Tests are grouped by their containing file. Groups are
+  distributed to available workers as whole units. This guarantees that all
+  tests in a file run in the same worker.
+
+* ``--dist loadgroup``: Tests are grouped by the ``xdist_group`` mark. Groups are
+  distributed to available workers as whole units. This guarantees that all
+  tests with same ``xdist_group`` name run in the same worker.
+
+  .. code-block:: python
+
+      @pytest.mark.xdist_group(name="group1")
+      def test1():
+          pass
+
+
+      class TestA:
+          @pytest.mark.xdist_group("group1")
+          def test2():
+              pass
+
+  This will make sure ``test1`` and ``TestA::test2`` will run in the same worker.
+  Tests without the ``xdist_group`` mark are distributed normally as in the ``--dist=load`` mode.
+
+* ``--dist worksteal``: Initially, tests are distributed evenly among all
+  available workers. When a worker completes most of its assigned tests and
+  doesn't have enough tests to continue (currently, every worker needs at least
+  two tests in its queue), an attempt is made to reassign ("steal") a portion
+  of tests from some other worker's queue. The results should be similar to
+  the ``load`` method, but ``worksteal`` should handle tests with significantly
+  differing duration better, and, at the same time, it should provide similar
+  or better reuse of fixtures.
+
+* ``--dist no``: The normal pytest execution mode, runs one test at a time (no distribution at all).

docs/how-it-works.rst

@@ -0,0 +1,90 @@
+How it works?
+=============
+
+``xdist`` works by spawning one or more **workers**, which are
+controlled by the **controller**. Each **worker** is responsible for
+performing a full test collection and afterwards running tests as
+dictated by the **controller**.
+
+The execution flow is:
+
+1. **controller** spawns one or more **workers** at the beginning of the
+   test session. The communication between **controller** and **worker**
+   nodes makes use of `execnet <https://codespeak.net/execnet/>`__ and
+   its
+   `gateways <https://codespeak.net/execnet/basics.html#gateways-bootstrapping-python-interpreters>`__.
+   The actual interpreters executing the code for the **workers** might
+   be remote or local.
+
+2. Each **worker** itself is a mini pytest runner. **workers** at this
+   point perform a full test collection, sending back the collected
+   test-ids back to the **controller** which does not perform any
+   collection itself.
+
+3. The **controller** receives the result of the collection from all
+   nodes. At this point the **controller** performs some sanity check to
+   ensure that all **workers** collected the same tests (including
+   order), bailing out otherwise. If all is well, it converts the list
+   of test-ids into a list of simple indexes, where each index
+   corresponds to the position of that test in the original collection
+   list. This works because all nodes have the same collection list, and
+   saves bandwidth because the **controller** can now tell one of the
+   workers to just *execute test index 3* index of passing the full test
+   id.
+
+4. If **dist-mode** is **each**: the **controller** just sends the full
+   list of test indexes to each node at this moment.
+
+5. If **dist-mode** is **load**: the **controller** takes around 25% of
+   the tests and sends them one by one to each **worker** in a round
+   robin fashion. The rest of the tests will be distributed later as
+   **workers** finish tests (see below).
+
+6. Note that ``pytest_xdist_make_scheduler`` hook can be used to
+   implement custom tests distribution logic.
+
+7. **workers** re-implement ``pytest_runtestloop``: pytest’s default
+   implementation basically loops over all collected items in the
+   ``session`` object and executes the ``pytest_runtest_protocol`` for
+   each test item, but in xdist **workers** sit idly waiting for
+   **controller** to send tests for execution. As tests are received by
+   **workers**, ``pytest_runtest_protocol`` is executed for each test.
+   Here it worth noting an implementation detail: **workers** always
+   must keep at least one test item on their queue due to how the
+   ``pytest_runtest_protocol(item, nextitem)`` hook is defined: in order
+   to pass the ``nextitem`` to the hook, the worker must wait for more
+   instructions from controller before executing that remaining test. If
+   it receives more tests, then it can safely call
+   ``pytest_runtest_protocol`` because it knows what the ``nextitem``
+   parameter will be. If it receives a “shutdown” signal, then it can
+   execute the hook passing ``nextitem`` as ``None``.
+
+8. As tests are started and completed at the **workers**, the results
+   are sent back to the **controller**, which then just forwards the
+   results to the appropriate pytest hooks: ``pytest_runtest_logstart``
+   and ``pytest_runtest_logreport``. This way other plugins (for example
+   ``junitxml``) can work normally. The **controller** (when in
+   dist-mode **load**) decides to send more tests to a node when a test
+   completes, using some heuristics such as test durations and how many
+   tests each **worker** still has to run.
+
+9. When the **controller** has no more pending tests it will send a
+   “shutdown” signal to all **workers**, which will then run their
+   remaining tests to completion and shut down. At this point the
+   **controller** will sit waiting for **workers** to shut down, still
+   processing events such as ``pytest_runtest_logreport``.
+
+FAQ
+---
+
+**Question**: Why does each worker do its own collection, as opposed to having the
+controller collect once and distribute from that collection to the
+workers?
+
+If collection was performed by controller then it would have to
+serialize collected items to send them through the wire, as workers live
+in another process. The problem is that test items are not easily
+(impossible?) to serialize, as they contain references to the test
+functions, fixture managers, config objects, etc. Even if one manages to
+serialize it, it seems it would be very hard to get it right and easy to
+break by any small change in pytest.

docs/how-to.rst

@@ -0,0 +1,232 @@
+How-tos
+-------
+
+This section show cases how to accomplish some specialized tasks with ``pytest-xdist``.
+
+Identifying the worker process during a test
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+*New in version 1.15.*
+
+If you need to determine the identity of a worker process in
+a test or fixture, you may use the ``worker_id`` fixture to do so:
+
+.. code-block:: python
+
+    @pytest.fixture()
+    def user_account(worker_id):
+        """use a different account in each xdist worker"""
+        return "account_%s" % worker_id
+
+When ``xdist`` is disabled (running with ``-n0`` for example), then
+``worker_id`` will return ``"master"``.
+
+Worker processes also have the following environment variables
+defined:
+
+.. envvar:: PYTEST_XDIST_WORKER
+
+The name of the worker, e.g., ``"gw2"``.
+
+.. envvar:: PYTEST_XDIST_WORKER_COUNT
+
+The total number of workers in this session, e.g., ``"4"`` when ``-n 4`` is given in the command-line.
+
+The information about the worker_id in a test is stored in the ``TestReport`` as
+well, under the ``worker_id`` attribute.
+
+Since version 2.0, the following functions are also available in the ``xdist`` module:
+
+
+.. autofunction:: xdist.is_xdist_worker
+.. autofunction:: xdist.is_xdist_controller
+.. autofunction:: xdist.is_xdist_master
+.. autofunction:: xdist.get_xdist_worker_id
+
+Identifying workers from the system environment
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+*New in version 2.4*
+
+If the `setproctitle`_ package is installed, ``pytest-xdist`` will use it to
+update the process title (command line) on its workers to show their current
+state.  The titles used are ``[pytest-xdist running] file.py/node::id`` and
+``[pytest-xdist idle]``, visible in standard tools like ``ps`` and ``top`` on
+Linux, Mac OS X and BSD systems.  For Windows, please follow `setproctitle`_'s
+pointer regarding the Process Explorer tool.
+
+This is intended purely as an UX enhancement, e.g. to track down issues with
+long-running or CPU intensive tests.  Errors in changing the title are ignored
+silently.  Please try not to rely on the title format or title changes in
+external scripts.
+
+.. _`setproctitle`: https://pypi.org/project/setproctitle/
+
+
+Uniquely identifying the current test run
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+*New in version 1.32.*
+
+If you need to globally distinguish one test run from others in your
+workers, you can use the ``testrun_uid`` fixture. For instance, let's say you
+wanted to create a separate database for each test run:
+
+.. code-block:: python
+
+    import pytest
+    from posix_ipc import Semaphore, O_CREAT
+
+
+    @pytest.fixture(scope="session", autouse=True)
+    def create_unique_database(testrun_uid):
+        """create a unique database for this particular test run"""
+        database_url = f"psql://myapp-{testrun_uid}"
+
+        with Semaphore(f"/{testrun_uid}-lock", flags=O_CREAT, initial_value=1):
+            if not database_exists(database_url):
+                create_database(database_url)
+
+
+    @pytest.fixture()
+    def db(testrun_uid):
+        """retrieve unique database"""
+        database_url = f"psql://myapp-{testrun_uid}"
+        return database_get_instance(database_url)
+
+
+Additionally, during a test run, the following environment variable is defined:
+
+.. envvar:: PYTEST_XDIST_TESTRUNUID
+
+The unique id of the test run.
+
+Accessing ``sys.argv`` from the controller node in workers
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+To access the ``sys.argv`` passed to the command-line of the controller node, use
+``request.config.workerinput["mainargv"]``.
+
+
+Specifying test exec environments in an ini file
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+You can use pytest's ini file configuration to avoid typing common options.
+You can for example make running with three subprocesses your default like this:
+
+.. code-block:: ini
+
+    [pytest]
+    addopts = -n3
+
+You can also add default environments like this:
+
+.. code-block:: ini
+
+    [pytest]
+    addopts = --tx ssh=myhost//python=python3.9 --tx ssh=myhost//python=python3.6
+
+and then just type::
+
+    pytest --dist=each
+
+to run tests in each of the environments.
+
+
+Specifying "rsync" dirs in an ini-file
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+In a ``tox.ini`` or ``setup.cfg`` file in your root project directory
+you may specify directories to include or to exclude in synchronisation:
+
+.. code-block:: ini
+
+    [pytest]
+    rsyncdirs = . mypkg helperpkg
+    rsyncignore = .hg
+
+These directory specifications are relative to the directory
+where the configuration file was found.
+
+.. _`pytest-xdist`: http://pypi.python.org/pypi/pytest-xdist
+.. _`pytest-xdist repository`: https://github.com/pytest-dev/pytest-xdist
+.. _`pytest`: http://pytest.org
+
+
+Making session-scoped fixtures execute only once
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+``pytest-xdist`` is designed so that each worker process will perform its own collection and execute
+a subset of all tests. This means that tests in different processes requesting a high-level
+scoped fixture (for example ``session``) will execute the fixture code more than once, which
+breaks expectations and might be undesired in certain situations.
+
+While ``pytest-xdist`` does not have a builtin support for ensuring a session-scoped fixture is
+executed exactly once, this can be achieved by using a lock file for inter-process communication.
+
+The example below needs to execute the fixture ``session_data`` only once (because it is
+resource intensive, or needs to execute only once to define configuration options, etc), so it makes
+use of a `FileLock <https://pypi.org/project/filelock/>`_ to produce the fixture data only once
+when the first process requests the fixture, while the other processes will then read
+the data from a file.
+
+Here is the code:
+
+.. code-block:: python
+
+    import json
+
+    import pytest
+    from filelock import FileLock
+
+
+    @pytest.fixture(scope="session")
+    def session_data(tmp_path_factory, worker_id):
+        if worker_id == "master":
+            # not executing in with multiple workers, just produce the data and let
+            # pytest's fixture caching do its job
+            return produce_expensive_data()
+
+        # get the temp directory shared by all workers
+        root_tmp_dir = tmp_path_factory.getbasetemp().parent
+
+        fn = root_tmp_dir / "data.json"
+        with FileLock(str(fn) + ".lock"):
+            if fn.is_file():
+                data = json.loads(fn.read_text())
+            else:
+                data = produce_expensive_data()
+                fn.write_text(json.dumps(data))
+        return data
+
+
+The example above can also be use in cases a fixture needs to execute exactly once per test session, like
+initializing a database service and populating initial tables.
+
+This technique might not work for every case, but should be a starting point for many situations
+where executing a high-scope fixture exactly once is important.
+
+
+Creating one log file for each worker
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+To create one log file for each worker with ``pytest-xdist``, you can leverage :envvar:`PYTEST_XDIST_WORKER`
+to generate a unique filename for each worker.
+
+Example:
+
+.. code-block:: python
+
+    # content of conftest.py
+    def pytest_configure(config):
+        worker_id = os.environ.get("PYTEST_XDIST_WORKER")
+        if worker_id is not None:
+            logging.basicConfig(
+                format=config.getini("log_file_format"),
+                filename=f"tests_{worker_id}.log",
+                level=config.getini("log_file_level"),
+            )
+
+
+When running the tests with ``-n3``, for example, three files will be created in the current directory:
+``tests_gw0.log``, ``tests_gw1.log`` and ``tests_gw2.log``.

docs/index.rst

@@ -0,0 +1,63 @@
+pytest-xdist
+============
+
+The `pytest-xdist`_ plugin extends pytest with new test execution modes, the most used being distributing
+tests across multiple CPUs to speed up test execution::
+
+    pytest -n auto
+
+With this call, pytest will spawn a number of workers processes equal to the number of available CPUs, and distribute
+the tests randomly across them.
+
+.. note::
+    Due to how pytest-xdist is implemented, the ``-s/--capture=no`` option does not work.
+
+
+Installation
+------------
+
+Install the plugin with::
+
+    pip install pytest-xdist
+
+
+To use ``psutil`` for detection of the number of CPUs available, install the ``psutil`` extra::
+
+    pip install pytest-xdist[psutil]
+
+Features
+--------
+
+* Test run :ref:`parallelization`: tests can be executed across  multiple CPUs or hosts.
+  This allows to speed up development or to use special resources of :ref:`remote machines`.
+
+* ``--looponfail``: run your tests repeatedly in a subprocess.  After each run
+  pytest waits until a file in your project changes and then re-runs
+  the previously failing tests.  This is repeated until all tests pass
+  after which again a full run is performed (DEPRECATED).
+
+* :ref:`Multi-Platform` coverage: you can specify different Python interpreters
+  or different platforms and run tests in parallel on all of them.
+
+  Before running tests remotely, ``pytest`` efficiently "rsyncs" your
+  program source code to the remote place.
+  You may specify different Python versions and interpreters. It does not
+  installs/synchronize dependencies however.
+
+  **Note**: this mode exists mostly for backward compatibility, as modern development
+  relies on continuous integration for multi-platform testing.
+
+
+
+.. toctree::
+    :maxdepth: 2
+    :caption: Contents:
+
+    distribution
+    subprocess
+    remote
+    crash
+    how-to
+    how-it-works
+    known-limitations
+    changelog

docs/known-limitations.rst

@@ -0,0 +1,71 @@
+Known limitations
+=================
+
+pytest-xdist has some limitations that may be supported in pytest but can't be supported in pytest-xdist.
+
+Order and amount of test must be consistent
+-------------------------------------------
+
+It is not possible to have tests that differ in order or their amount across workers.
+
+This is especially true with ``pytest.mark.parametrize``, when values are produced with sets or other unordered iterables/generators.
+
+
+Example:
+
+.. code-block:: python
+
+    import pytest
+
+
+    @pytest.mark.parametrize("param", {"a", "b"})
+    def test_pytest_parametrize_unordered(param):
+        pass
+
+In the example above, the fact that ``set`` are not necessarily ordered can cause different workers
+to collect tests in different order, which will throw an error.
+
+Workarounds
+~~~~~~~~~~~
+
+A solution to this is to guarantee that the parametrized values have the same order.
+
+Some solutions:
+
+* Convert your sequence to a ``list``.
+
+  .. code-block:: python
+
+    import pytest
+
+
+    @pytest.mark.parametrize("param", ["a", "b"])
+    def test_pytest_parametrize_unordered(param):
+        pass
+
+* Sort your sequence, guaranteeing order.
+
+  .. code-block:: python
+
+    import pytest
+
+
+    @pytest.mark.parametrize("param", sorted({"a", "b"}))
+    def test_pytest_parametrize_unordered(param):
+        pass
+
+Output (stdout and stderr) from workers
+---------------------------------------
+
+The ``-s``/``--capture=no`` option is meant to disable pytest capture, so users can then see stdout and stderr output in the terminal from tests and application code in real time.
+
+However, this option does not work with ``pytest-xdist`` because `execnet <https://github.com/pytest-dev/execnet>`__ the underlying library used for communication between master and workers, does not support transferring stdout/stderr from workers.
+
+Currently, there are no plans to support this in ``pytest-xdist``.
+
+Debugging
+~~~~~~~~~
+
+This also means that debugging using PDB (or any other debugger that wants to use standard I/O) will not work. The ``--pdb`` option is disabled when distributing tests with ``pytest-xdist`` for this reason.
+
+It is generally likely best to use ``pytest-xdist`` to find failing tests and then debug them without distribution; however, if you need to debug from within a worker process (for example, to address failures that only happen when running tests concurrently), remote debuggers (for example, `python-remote-pdb <https://github.com/ionelmc/python-remote-pdb>`__ or `python-web-pdb <https://github.com/romanvm/python-web-pdb>`__) have been reported to work for this purpose.

docs/remote.rst

@@ -0,0 +1,83 @@
+
+.. _`Multi-Platform`:
+.. _`remote machines`:
+
+Sending tests to remote SSH accounts
+====================================
+
+.. deprecated:: 3.0
+
+.. warning::
+
+    This feature is deprecated because the support for ``rsync`` is faulty
+    in terms of reproducing the development environment in the remote
+    worker, and there is no clear solution moving forward.
+
+    This feature is scheduled to be removed in release 4.0, to let the team
+    focus on a smaller set of features.
+
+Suppose you have a package ``mypkg`` which contains some
+tests that you can successfully run locally. And you
+have a ssh-reachable machine ``myhost``.  Then
+you can ad-hoc distribute your tests by typing::
+
+    pytest -d  --rsyncdir mypkg --tx ssh=myhostpopen mypkg/tests/unit/test_something.py
+
+This will synchronize your :code:`mypkg` package directory
+to a remote ssh account and then locally collect tests
+and send them to remote places for execution.
+
+You can specify multiple :code:`--rsyncdir` directories
+to be sent to the remote side.
+
+.. note::
+
+  For pytest to collect and send tests correctly
+  you not only need to make sure all code and tests
+  directories are rsynced, but that any test (sub) directory
+  also has an :code:`__init__.py` file because internally
+  pytest references tests as a fully qualified python
+  module path.  **You will otherwise get strange errors**
+  during setup of the remote side.
+
+
+You can specify multiple :code:`--rsyncignore` glob patterns
+to be ignored when file are sent to the remote side.
+There are also internal ignores: :code:`.*, *.pyc, *.pyo, *~`
+Those you cannot override using rsyncignore command-line or
+ini-file option(s).
+
+
+Sending tests to remote Socket Servers
+--------------------------------------
+
+Download the single-module `socketserver.py`_ Python program
+and run it like this::
+
+    python socketserver.py
+
+It will tell you that it starts listening on the default
+port.  You can now on your home machine specify this
+new socket host with something like this::
+
+    pytest -d --tx socket=192.168.1.102:8888 --rsyncdir mypkg
+
+
+
+Running tests on many platforms at once
+---------------------------------------
+
+The basic command to run tests on multiple platforms is::
+
+    pytest --dist=each --tx=spec1 --tx=spec2
+
+If you specify a windows host, an OSX host and a Linux
+environment this command will send each tests to all
+platforms - and report back failures from all platforms
+at once. The specifications strings use the `xspec syntax`_.
+
+.. _`xspec syntax`: https://codespeak.net/execnet/basics.html#xspec
+
+.. _`execnet`: https://codespeak.net/execnet
+
+.. _`socketserver.py`: https://raw.githubusercontent.com/pytest-dev/execnet/master/execnet/script/socketserver.py

docs/subprocess.rst

@@ -0,0 +1,16 @@
+Running tests in a Python subprocess
+====================================
+
+To instantiate a ``python3.9`` subprocess and send tests to it, you may type::
+
+    pytest -d --tx popen//python=python3.9
+
+This will start a subprocess which is run with the ``python3.9``
+Python interpreter, found in your system binary lookup path.
+
+If you prefix the --tx option value like this::
+
+    --tx 3*popen//python=python3.9
+
+then three subprocesses would be created and tests
+will be load-balanced across these three processes.

pyproject.toml

@@ -1,6 +1,5 @@
 [build-system]
 requires = [
-  # sync with setup.py until we discard non-pep-517/518
   "setuptools>=45.0",
   "setuptools-scm[toml]>=6.2.3",
   "wheel",
@@ -17,32 +16,23 @@ directory = "changelog/"
 title_format = "pytest-xdist {version} ({project_date})"
 template = "changelog/_template.rst"
 
-  [[tool.towncrier.type]]
-  directory = "removal"
-  name = "Deprecations and Removals"
-  showcontent = true
+[tool.towncrier.fragment.removal]
+name = "Removals"
 
-  [[tool.towncrier.type]]
-  directory = "feature"
-  name = "Features"
-  showcontent = true
+[tool.towncrier.fragment.deprecation]
+name = "Deprecations"
 
-  [[tool.towncrier.type]]
-  directory = "bugfix"
-  name = "Bug Fixes"
-  showcontent = true
+[tool.towncrier.fragment.feature]
+name = "Features"
 
-  [[tool.towncrier.type]]
-  directory = "vendor"
-  name = "Vendored Libraries"
-  showcontent = true
+[tool.towncrier.fragment.bugfix]
+name = "Bug Fixes"
 
-  [[tool.towncrier.type]]
-  directory = "doc"
-  name = "Improved Documentation"
-  showcontent = true
+[tool.towncrier.fragment.vendor]
+name = "Vendored Libraries"
 
-  [[tool.towncrier.type]]
-  directory = "trivial"
-  name = "Trivial Changes"
-  showcontent = true
+[tool.towncrier.fragment.doc]
+name = "Improved Documentation"
+
+[tool.towncrier.fragment.trivial]
+name = "Trivial Changes"

setup.cfg

@@ -1,7 +1,8 @@
 [metadata]
 name = pytest-xdist
-description = pytest xdist plugin for distributed testing and loop-on-failing modes
+description = pytest xdist plugin for distributed testing, most importantly across multiple CPUs
 long_description = file: README.rst
+long_description_content_type = text/x-rst
 license = MIT
 author = holger krekel and contributors
 author_email = pytest-dev@python.org,holger@merlinux.eu
@@ -24,23 +25,26 @@ classifiers =
     Programming Language :: Python
     Programming Language :: Python :: 3
     Programming Language :: Python :: 3 :: Only
-    Programming Language :: Python :: 3.6
     Programming Language :: Python :: 3.7
     Programming Language :: Python :: 3.8
     Programming Language :: Python :: 3.9
     Programming Language :: Python :: 3.10
+    Programming Language :: Python :: 3.11
 license_file = LICENSE
+project_urls =
+    Documentation=https://pytest-xdist.readthedocs.io/en/latest
+    Changelog=https://pytest-xdist.readthedocs.io/en/latest/changelog.html
+    Source=https://github.com/pytest-dev/pytest-xdist
+    Tracker=https://github.com/pytest-dev/pytest-xdist/issues
 
 [options]
 packages = find:
 package_dir = =src
 zip_safe = False
-python_requires = >=3.6
+python_requires = >=3.7
 install_requires =
     execnet>=1.1
     pytest>=6.2.0
-    pytest-forked
-setup_requires = setuptools_scm>=6.0
 
 [options.packages.find]
 where = src
@@ -57,6 +61,8 @@ psutil = psutil>=3.0
 setproctitle = setproctitle
 
 [flake8]
+# Ignore any errors related to formatting, let black worry/fix them.
+ignore = E501, W503, E203
 max-line-length = 100
 
 [mypy]

setup.py

@@ -1,4 +0,0 @@
-from setuptools import setup
-
-if __name__ == "__main__":
-    setup()

src/xdist/_path.py

@@ -0,0 +1,19 @@
+import os
+from itertools import chain
+from pathlib import Path
+from typing import Callable, Iterator
+
+
+def visit_path(
+    path: Path, *, filter: Callable[[Path], bool], recurse: Callable[[Path], bool]
+) -> Iterator[Path]:
+    """
+    Implements the interface of ``py.path.local.visit()`` for Path objects,
+    to simplify porting the code over from ``py.path.local``.
+    """
+    for dirpath, dirnames, filenames in os.walk(path):
+        dirnames[:] = [x for x in dirnames if recurse(Path(dirpath, x))]
+        for name in chain(dirnames, filenames):
+            p = Path(dirpath, name)
+            if filter(p):
+                yield p

src/xdist/dsession.py

@@ -1,6 +1,11 @@
-import py
+from __future__ import annotations
+import sys
+from enum import Enum, auto
+from typing import Sequence
+
 import pytest
 
+from xdist.remote import Producer
 from xdist.workermanage import NodeManager
 from xdist.scheduler import (
     EachScheduling,
@@ -8,6 +13,7 @@ from xdist.scheduler import (
     LoadScopeScheduling,
     LoadFileScheduling,
     LoadGroupScheduling,
+    WorkStealingScheduling,
 )
 
 
@@ -34,9 +40,7 @@ class DSession:
 
     def __init__(self, config):
         self.config = config
-        self.log = py.log.Producer("dsession")
-        if not config.option.debug:
-            py.log.setconsumer(self.log._keywords, None)
+        self.log = Producer("dsession", enabled=config.option.debug)
         self.nodemanager = None
         self.sched = None
         self.shuttingdown = False
@@ -102,6 +106,7 @@ class DSession:
             "loadscope": LoadScopeScheduling,
             "loadfile": LoadFileScheduling,
             "loadgroup": LoadGroupScheduling,
+            "worksteal": WorkStealingScheduling,
         }
         return schedulers[dist](config, log)
 
@@ -171,7 +176,7 @@ class DSession:
         """
         self.config.hook.pytest_testnodedown(node=node, error=None)
         if node.workeroutput["exitstatus"] == 2:  # keyboard-interrupt
-            self.shouldstop = "{} received keyboard-interrupt".format(node)
+            self.shouldstop = f"{node} received keyboard-interrupt"
             self.worker_errordown(node, "keyboard-interrupt")
             return
         if node in self.sched.nodes:
@@ -225,13 +230,14 @@ class DSession:
             self.triggershutdown()
         else:
             self.report_line("\nreplacing crashed worker %s" % node.gateway.id)
+            self.shuttingdown = False
             self._clone_node(node)
         self._active_nodes.remove(node)
 
     @pytest.hookimpl
     def pytest_terminal_summary(self, terminalreporter):
         if self.config.option.verbose >= 0 and self._summary_report:
-            terminalreporter.write_sep("=", "xdist: {}".format(self._summary_report))
+            terminalreporter.write_sep("=", f"xdist: {self._summary_report}")
 
     def worker_collectionfinish(self, node, ids):
         """worker has finished test collection.
@@ -250,14 +256,16 @@ class DSession:
         self._session.testscollected = len(ids)
         self.sched.add_node_collection(node, ids)
         if self.terminal:
-            self.trdist.setstatus(node.gateway.spec, "[%d]" % (len(ids)))
+            self.trdist.setstatus(
+                node.gateway.spec, WorkerStatus.CollectionDone, tests_collected=len(ids)
+            )
         if self.sched.collection_is_completed:
             if self.terminal and not self.sched.has_pending:
                 self.trdist.ensure_show_status()
                 self.terminal.write_line("")
                 if self.config.option.verbose > 0:
                     self.terminal.write_line(
-                        "scheduling tests via %s" % (self.sched.__class__.__name__)
+                        f"scheduling tests via {self.sched.__class__.__name__}"
                     )
             self.sched.schedule()
 
@@ -283,6 +291,17 @@ class DSession:
         """
         self.sched.mark_test_complete(node, item_index, duration)
 
+    def worker_unscheduled(self, node, indices):
+        """
+        Emitted when a node fires the 'unscheduled' event, signalling that
+        some tests have been removed from the worker's queue and should be
+        sent to some worker again.
+
+        This should happen only in response to 'steal' command, so schedulers
+        not using 'steal' command don't have to implement it.
+        """
+        self.sched.remove_pending_tests_from_node(node, indices)
+
     def worker_collectreport(self, node, rep):
         """Emitted when a node calls the pytest_collectreport hook.
 
@@ -333,7 +352,7 @@ class DSession:
         if rep.failed:
             self.countfailures += 1
             if self.maxfail and self.countfailures >= self.maxfail:
-                self.shouldstop = "stopping after %d failures" % (self.countfailures)
+                self.shouldstop = f"stopping after {self.countfailures} failures"
 
     def triggershutdown(self):
         self.log("triggering shutdown")
@@ -346,7 +365,7 @@ class DSession:
         # XXX count no of failures and retry N times
         runner = self.config.pluginmanager.getplugin("runner")
         fspath = nodeid.split("::")[0]
-        msg = "worker {!r} crashed while running {!r}".format(worker.gateway.id, nodeid)
+        msg = f"worker {worker.gateway.id!r} crashed while running {nodeid!r}"
         rep = runner.TestReport(
             nodeid, (fspath, None, fspath), (), "failed", msg, "???"
         )
@@ -360,34 +379,51 @@ class DSession:
         self.config.hook.pytest_runtest_logreport(report=rep)
 
 
+class WorkerStatus(Enum):
+    """Status of each worker during creation/collection."""
+
+    # Worker spec has just been created.
+    Created = auto()
+
+    # Worker has been initialized.
+    Initialized = auto()
+
+    # Worker is now ready for collection.
+    ReadyForCollection = auto()
+
+    # Worker has finished collection.
+    CollectionDone = auto()
+
+
 class TerminalDistReporter:
-    def __init__(self, config):
+    def __init__(self, config) -> None:
         self.config = config
         self.tr = config.pluginmanager.getplugin("terminalreporter")
-        self._status = {}
+        self._status: dict[str, tuple[WorkerStatus, int]] = {}
         self._lastlen = 0
         self._isatty = getattr(self.tr, "isatty", self.tr.hasmarkup)
 
-    def write_line(self, msg):
+    def write_line(self, msg: str) -> None:
         self.tr.write_line(msg)
 
-    def ensure_show_status(self):
+    def ensure_show_status(self) -> None:
         if not self._isatty:
             self.write_line(self.getstatus())
 
-    def setstatus(self, spec, status, show=True):
-        self._status[spec.id] = status
+    def setstatus(
+        self, spec, status: WorkerStatus, *, tests_collected: int, show: bool = True
+    ) -> None:
+        self._status[spec.id] = (status, tests_collected)
         if show and self._isatty:
             self.rewrite(self.getstatus())
 
-    def getstatus(self):
+    def getstatus(self) -> str:
         if self.config.option.verbose >= 0:
-            parts = [
-                "{} {}".format(spec.id, self._status[spec.id]) for spec in self._specs
-            ]
-            return " / ".join(parts)
-        else:
-            return "bringing up nodes..."
+            line = get_workers_status_line(list(self._status.values()))
+            if line:
+                return line
+
+        return "bringing up nodes..."
 
     def rewrite(self, line, newline=False):
         pline = line + " " * max(self._lastlen - len(line), 0)
@@ -399,40 +435,44 @@ class TerminalDistReporter:
         self.tr.rewrite(pline, bold=True)
 
     @pytest.hookimpl
-    def pytest_xdist_setupnodes(self, specs):
+    def pytest_xdist_setupnodes(self, specs) -> None:
         self._specs = specs
         for spec in specs:
-            self.setstatus(spec, "I", show=False)
-        self.setstatus(spec, "I", show=True)
+            self.setstatus(spec, WorkerStatus.Created, tests_collected=0, show=False)
+        self.setstatus(spec, WorkerStatus.Created, tests_collected=0, show=True)
         self.ensure_show_status()
 
     @pytest.hookimpl
-    def pytest_xdist_newgateway(self, gateway):
-        if self.config.option.verbose > 0:
-            rinfo = gateway._rinfo()
+    def pytest_xdist_newgateway(self, gateway) -> None:
+        rinfo = gateway._rinfo()
+        is_local = rinfo.executable == sys.executable
+        if self.config.option.verbose > 0 and not is_local:
             version = "%s.%s.%s" % rinfo.version_info[:3]
             self.rewrite(
                 "[%s] %s Python %s cwd: %s"
                 % (gateway.id, rinfo.platform, version, rinfo.cwd),
                 newline=True,
             )
-        self.setstatus(gateway.spec, "C")
+        self.setstatus(gateway.spec, WorkerStatus.Initialized, tests_collected=0)
 
     @pytest.hookimpl
-    def pytest_testnodeready(self, node):
-        if self.config.option.verbose > 0:
-            d = node.workerinfo
+    def pytest_testnodeready(self, node) -> None:
+        d = node.workerinfo
+        is_local = d.get("executable") == sys.executable
+        if self.config.option.verbose > 0 and not is_local:
             infoline = "[{}] Python {}".format(
                 d["id"], d["version"].replace("\n", " -- ")
             )
             self.rewrite(infoline, newline=True)
-        self.setstatus(node.gateway.spec, "ok")
+        self.setstatus(
+            node.gateway.spec, WorkerStatus.ReadyForCollection, tests_collected=0
+        )
 
     @pytest.hookimpl
-    def pytest_testnodedown(self, node, error):
+    def pytest_testnodedown(self, node, error) -> None:
         if not error:
             return
-        self.write_line("[{}] node down: {}".format(node.gateway.id, error))
+        self.write_line(f"[{node.gateway.id}] node down: {error}")
 
 
 def get_default_max_worker_restart(config):
@@ -447,3 +487,36 @@ def get_default_max_worker_restart(config):
         # if --max-worker-restart was not provided, use a reasonable default (#226)
         result = config.option.numprocesses * 4
     return result
+
+
+def get_workers_status_line(
+    status_and_items: Sequence[tuple[WorkerStatus, int]]
+) -> str:
+    """
+    Return the line to display during worker setup/collection based on the
+    status of the workers and number of tests collected for each.
+    """
+    statuses = [s for s, c in status_and_items]
+    total_workers = len(statuses)
+    workers_noun = "worker" if total_workers == 1 else "workers"
+    if status_and_items and all(s == WorkerStatus.CollectionDone for s in statuses):
+        # All workers collect the same number of items, so we grab
+        # the total number of items from the first worker.
+        first = status_and_items[0]
+        status, tests_collected = first
+        tests_noun = "item" if tests_collected == 1 else "items"
+        return f"{total_workers} {workers_noun} [{tests_collected} {tests_noun}]"
+    if WorkerStatus.CollectionDone in statuses:
+        done = sum(1 for s, c in status_and_items if c > 0)
+        return f"collecting: {done}/{total_workers} {workers_noun}"
+    if WorkerStatus.ReadyForCollection in statuses:
+        ready = statuses.count(WorkerStatus.ReadyForCollection)
+        return f"ready: {ready}/{total_workers} {workers_noun}"
+    if WorkerStatus.Initialized in statuses:
+        initialized = statuses.count(WorkerStatus.Initialized)
+        return f"initialized: {initialized}/{total_workers} {workers_noun}"
+    if WorkerStatus.Created in statuses:
+        created = statuses.count(WorkerStatus.Created)
+        return f"created: {created}/{total_workers} {workers_noun}"
+
+    return ""

src/xdist/looponfail.py

@@ -6,11 +6,17 @@
     processes) otherwise changes to source code can crash
     the controlling process which should best never happen.
 """
-import py
+import os
+from pathlib import Path
+from typing import Dict, Sequence
+
 import pytest
 import sys
 import time
 import execnet
+from _pytest._io import TerminalWriter
+
+from xdist._path import visit_path
 
 
 @pytest.hookimpl
@@ -29,7 +35,6 @@ def pytest_addoption(parser):
 
 @pytest.hookimpl
 def pytest_cmdline_main(config):
-
     if config.getoption("looponfail"):
         usepdb = config.getoption("usepdb", False)  # a core option
         if usepdb:
@@ -38,9 +43,12 @@ def pytest_cmdline_main(config):
         return 2  # looponfail only can get stop with ctrl-C anyway
 
 
-def looponfail_main(config):
+def looponfail_main(config: "pytest.Config") -> None:
     remotecontrol = RemoteControl(config)
-    rootdirs = [py.path.local(root) for root in config.getini("looponfailroots")]
+    config_roots = config.getini("looponfailroots")
+    if not config_roots:
+        config_roots = [Path.cwd()]
+    rootdirs = [Path(root) for root in config_roots]
     statrecorder = StatRecorder(rootdirs)
     try:
         while 1:
@@ -71,7 +79,7 @@ class RemoteControl:
 
     def setup(self, out=None):
         if out is None:
-            out = py.io.TerminalWriter()
+            out = TerminalWriter()
         if hasattr(self, "gateway"):
             raise ValueError("already have gateway %r" % self.gateway)
         self.trace("setting up worker session")
@@ -129,7 +137,7 @@ class RemoteControl:
 
 
 def repr_pytest_looponfailinfo(failreports, rootdirs):
-    tr = py.io.TerminalWriter()
+    tr = TerminalWriter()
     if failreports:
         tr.sep("#", "LOOPONFAILING", bold=True)
         for report in failreports:
@@ -137,7 +145,7 @@ def repr_pytest_looponfailinfo(failreports, rootdirs):
                 tr.line(report, red=True)
     tr.sep("#", "waiting for changes", bold=True)
     for rootdir in rootdirs:
-        tr.line("### Watching:   {}".format(rootdir), bold=True)
+        tr.line(f"### Watching:   {rootdir}", bold=True)
 
 
 def init_worker_session(channel, args, option_dict):
@@ -225,16 +233,16 @@ class WorkerFailSession:
 
 
 class StatRecorder:
-    def __init__(self, rootdirlist):
+    def __init__(self, rootdirlist: Sequence[Path]) -> None:
         self.rootdirlist = rootdirlist
-        self.statcache = {}
+        self.statcache: Dict[Path, os.stat_result] = {}
         self.check()  # snapshot state
 
-    def fil(self, p):
-        return p.check(file=1, dotfile=0) and p.ext != ".pyc"
+    def fil(self, p: Path) -> bool:
+        return p.is_file() and not p.name.startswith(".") and p.suffix != ".pyc"
 
-    def rec(self, p):
-        return p.check(dotfile=0)
+    def rec(self, p: Path) -> bool:
+        return not p.name.startswith(".") and p.exists()
 
     def waitonchange(self, checkinterval=1.0):
         while 1:
@@ -243,34 +251,34 @@ class StatRecorder:
                 return
             time.sleep(checkinterval)
 
-    def check(self, removepycfiles=True):  # noqa, too complex
+    def check(self, removepycfiles: bool = True) -> bool:  # noqa, too complex
         changed = False
-        statcache = self.statcache
-        newstat = {}
+        newstat: Dict[Path, os.stat_result] = {}
         for rootdir in self.rootdirlist:
-            for path in rootdir.visit(self.fil, self.rec):
-                oldstat = statcache.pop(path, None)
+            for path in visit_path(rootdir, filter=self.fil, recurse=self.rec):
+                oldstat = self.statcache.pop(path, None)
                 try:
-                    newstat[path] = curstat = path.stat()
-                except py.error.ENOENT:
+                    curstat = path.stat()
+                except OSError:
                     if oldstat:
                         changed = True
                 else:
-                    if oldstat:
+                    newstat[path] = curstat
+                    if oldstat is not None:
                         if (
-                            oldstat.mtime != curstat.mtime
-                            or oldstat.size != curstat.size
+                            oldstat.st_mtime != curstat.st_mtime
+                            or oldstat.st_size != curstat.st_size
                         ):
                             changed = True
                             print("# MODIFIED", path)
-                            if removepycfiles and path.ext == ".py":
-                                pycfile = path + "c"
-                                if pycfile.check():
-                                    pycfile.remove()
+                            if removepycfiles and path.suffix == ".py":
+                                pycfile = path.with_suffix(".pyc")
+                                if pycfile.is_file():
+                                    os.unlink(pycfile)
 
                     else:
                         changed = True
-        if statcache:
+        if self.statcache:
             changed = True
         self.statcache = newstat
         return changed

src/xdist/newhooks.py

@@ -24,12 +24,16 @@ def pytest_xdist_newgateway(gateway):
     """called on new raw gateway creation."""
 
 
-@pytest.hookspec()
+@pytest.hookspec(
+    warn_on_impl="rsync feature is deprecated and will be removed in pytest-xdist 4.0"
+)
 def pytest_xdist_rsyncstart(source, gateways):
     """called before rsyncing a directory to remote gateways takes place."""
 
 
-@pytest.hookspec()
+@pytest.hookspec(
+    warn_on_impl="rsync feature is deprecated and will be removed in pytest-xdist 4.0"
+)
 def pytest_xdist_rsyncfinish(source, gateways):
     """called after rsyncing a directory to remote gateways takes place."""
 

src/xdist/plugin.py

@@ -1,9 +1,8 @@
 import os
 import uuid
 import sys
-from pathlib import Path
+import warnings
 
-import py
 import pytest
 
 
@@ -14,6 +13,15 @@ _sys_path = list(sys.path)  # freeze a copy of sys.path at interpreter startup
 
 @pytest.hookimpl
 def pytest_xdist_auto_num_workers(config):
+    env_var = os.environ.get("PYTEST_XDIST_AUTO_NUM_WORKERS")
+    if env_var:
+        try:
+            return int(env_var)
+        except ValueError:
+            warnings.warn(
+                "PYTEST_XDIST_AUTO_NUM_WORKERS is not a number: {env_var!r}. Ignoring it."
+            )
+
     try:
         import psutil
     except ImportError:
@@ -86,7 +94,15 @@ def pytest_addoption(parser):
         "--dist",
         metavar="distmode",
         action="store",
-        choices=["each", "load", "loadscope", "loadfile", "loadgroup", "no"],
+        choices=[
+            "each",
+            "load",
+            "loadscope",
+            "loadfile",
+            "loadgroup",
+            "worksteal",
+            "no",
+        ],
         dest="dist",
         default="no",
         help=(
@@ -99,6 +115,8 @@ def pytest_addoption(parser):
             "loadfile: load balance by sending test grouped by file"
             " to any available environment.\n\n"
             "loadgroup: like load, but sends tests marked with 'xdist_group' to the same worker.\n\n"
+            "worksteal: split the test suite between available environments,"
+            " then rebalance when any worker runs out of tests.\n\n"
             "(default) no: run tests inprocess, don't distribute."
         ),
     )
@@ -135,11 +153,6 @@ def pytest_addoption(parser):
         metavar="GLOB",
         help="add expression for ignores when rsyncing to remote tx nodes.",
     )
-    group.addoption(
-        "--boxed",
-        action="store_true",
-        help="backward compatibility alias for pytest-forked --forked",
-    )
     group.addoption(
         "--testrunuid",
         action="store",
@@ -150,6 +163,21 @@ def pytest_addoption(parser):
             "on every test run."
         ),
     )
+    group.addoption(
+        "--maxschedchunk",
+        action="store",
+        type=int,
+        help=(
+            "Maximum number of tests scheduled in one step for --dist=load. "
+            "Setting it to 1 will force pytest to send tests to workers one by "
+            "one - might be useful for a small number of slow tests. "
+            "Larger numbers will allow the scheduler to submit consecutive "
+            "chunks of tests to workers - allows reusing fixtures. "
+            "Due to implementation reasons, at least 2 tests are scheduled per "
+            "worker at the start. Only later tests can be scheduled one by one. "
+            "Unlimited if not set."
+        ),
+    )
 
     parser.addini(
         "rsyncdirs",
@@ -164,8 +192,7 @@ def pytest_addoption(parser):
     parser.addini(
         "looponfailroots",
         type="paths" if PYTEST_GTE_7 else "pathlist",
-        help="directories to check for changes",
-        default=[Path.cwd() if PYTEST_GTE_7 else py.path.local()],
+        help="directories to check for changes. Default: current directory.",
     )
 
 
@@ -188,7 +215,19 @@ def pytest_addhooks(pluginmanager):
 
 @pytest.hookimpl(trylast=True)
 def pytest_configure(config):
-    if config.getoption("dist") != "no" and not config.getvalue("collectonly"):
+    config_line = (
+        "xdist_group: specify group for tests should run in same session."
+        "in relation to one another. Provided by pytest-xdist."
+    )
+    config.addinivalue_line("markers", config_line)
+
+    # Skip this plugin entirely when only doing collection.
+    if config.getvalue("collectonly"):
+        return
+
+    # Create the distributed session in case we have a valid distribution
+    # mode and test environments.
+    if config.getoption("dist") != "no" and config.getoption("tx"):
         from xdist.dsession import DSession
 
         session = DSession(config)
@@ -196,20 +235,21 @@ def pytest_configure(config):
         tr = config.pluginmanager.getplugin("terminalreporter")
         if tr:
             tr.showfspath = False
-    if config.getoption("boxed"):
+
+    # Deprecation warnings for deprecated command-line/configuration options.
+    if config.getoption("looponfail", None) or config.getini("looponfailroots"):
         warning = DeprecationWarning(
-            "The --boxed commmand line argument is deprecated. "
-            "Install pytest-forked and use --forked instead. "
-            "pytest-xdist 3.0.0 will remove the --boxed argument and pytest-forked dependency."
+            "The --looponfail command line argument and looponfailroots config variable are deprecated.\n"
+            "The loop-on-fail feature will be removed in pytest-xdist 4.0."
         )
         config.issue_config_time_warning(warning, 2)
-        config.option.forked = True
 
-    config_line = (
-        "xdist_group: specify group for tests should run in same session."
-        "in relation to one another. " + "Provided by pytest-xdist."
-    )
-    config.addinivalue_line("markers", config_line)
+    if config.getoption("rsyncdir", None) or config.getini("rsyncdirs"):
+        warning = DeprecationWarning(
+            "The --rsyncdir command line argument and rsyncdirs config variable are deprecated.\n"
+            "The rsync feature will be removed in pytest-xdist 4.0."
+        )
+        config.issue_config_time_warning(warning, 2)
 
 
 @pytest.hookimpl(tryfirst=True)

src/xdist/remote.py

@@ -6,11 +6,12 @@
     needs not to be installed in remote environments.
 """
 
+import contextlib
 import sys
 import os
 import time
+from typing import Any
 
-import py
 import pytest
 from execnet.gateway_base import dumps, DumpError
 
@@ -24,6 +25,29 @@ except ImportError:
         pass
 
 
+class Producer:
+    """
+    Simplified implementation of the same interface as py.log, for backward compatibility
+    since we dropped the dependency on pylib.
+    Note: this is defined here because this module can't depend on xdist, so we need
+    to have the other way around.
+    """
+
+    def __init__(self, name: str, *, enabled: bool = True):
+        self.name = name
+        self.enabled = enabled
+
+    def __repr__(self) -> str:
+        return f"{type(self).__name__}({self.name!r}, enabled={self.enabled})"
+
+    def __call__(self, *a: Any, **k: Any) -> None:
+        if self.enabled:
+            print(f"[{self.name}]", *a, **k, file=sys.stderr)
+
+    def __getattr__(self, name: str) -> "Producer":
+        return type(self)(name, enabled=self.enabled)
+
+
 def worker_title(title):
     try:
         setproctitle(title)
@@ -33,16 +57,31 @@ def worker_title(title):
 
 
 class WorkerInteractor:
+    SHUTDOWN_MARK = object()
+    QUEUE_REPLACED_MARK = object()
+
     def __init__(self, config, channel):
         self.config = config
         self.workerid = config.workerinput.get("workerid", "?")
         self.testrunuid = config.workerinput["testrunuid"]
-        self.log = py.log.Producer("worker-%s" % self.workerid)
-        if not config.option.debug:
-            py.log.setconsumer(self.log._keywords, None)
+        self.log = Producer(f"worker-{self.workerid}", enabled=config.option.debug)
         self.channel = channel
+        self.torun = self._make_queue()
+        self.nextitem_index = None
         config.pluginmanager.register(self)
 
+    def _make_queue(self):
+        return self.channel.gateway.execmodel.queue.Queue()
+
+    def _get_next_item_index(self):
+        """Gets the next item from test queue. Handles the case when the queue
+        is replaced concurrently in another thread.
+        """
+        result = self.torun.get()
+        while result is self.QUEUE_REPLACED_MARK:
+            result = self.torun.get()
+        return result
+
     def sendevent(self, name, **kwargs):
         self.log("sending", name, kwargs)
         self.channel.send((name, kwargs))
@@ -71,38 +110,63 @@ class WorkerInteractor:
     def pytest_collection(self, session):
         self.sendevent("collectionstart")
 
+    def handle_command(self, command):
+        if command is self.SHUTDOWN_MARK:
+            self.torun.put(self.SHUTDOWN_MARK)
+            return
+
+        name, kwargs = command
+
+        self.log("received command", name, kwargs)
+        if name == "runtests":
+            for i in kwargs["indices"]:
+                self.torun.put(i)
+        elif name == "runtests_all":
+            for i in range(len(self.session.items)):
+                self.torun.put(i)
+        elif name == "shutdown":
+            self.torun.put(self.SHUTDOWN_MARK)
+        elif name == "steal":
+            self.steal(kwargs["indices"])
+
+    def steal(self, indices):
+        indices = set(indices)
+        stolen = []
+
+        old_queue, self.torun = self.torun, self._make_queue()
+
+        def old_queue_get_nowait_noraise():
+            with contextlib.suppress(self.channel.gateway.execmodel.queue.Empty):
+                return old_queue.get_nowait()
+
+        for i in iter(old_queue_get_nowait_noraise, None):
+            if i in indices:
+                stolen.append(i)
+            else:
+                self.torun.put(i)
+
+        self.sendevent("unscheduled", indices=stolen)
+        old_queue.put(self.QUEUE_REPLACED_MARK)
+
     @pytest.hookimpl
     def pytest_runtestloop(self, session):
         self.log("entering main loop")
-        torun = []
-        while 1:
-            try:
-                name, kwargs = self.channel.receive()
-            except EOFError:
-                return True
-            self.log("received command", name, kwargs)
-            if name == "runtests":
-                torun.extend(kwargs["indices"])
-            elif name == "runtests_all":
-                torun.extend(range(len(session.items)))
-            self.log("items to run:", torun)
-            # only run if we have an item and a next item
-            while len(torun) >= 2:
-                self.run_one_test(torun)
-            if name == "shutdown":
-                if torun:
-                    self.run_one_test(torun)
-                break
+        self.channel.setcallback(self.handle_command, endmarker=self.SHUTDOWN_MARK)
+        self.nextitem_index = self._get_next_item_index()
+        while self.nextitem_index is not self.SHUTDOWN_MARK:
+            self.run_one_test()
         return True
 
-    def run_one_test(self, torun):
+    def run_one_test(self):
+        self.item_index = self.nextitem_index
+        self.nextitem_index = self._get_next_item_index()
+
         items = self.session.items
-        self.item_index = torun.pop(0)
         item = items[self.item_index]
-        if torun:
-            nextitem = items[torun[0]]
-        else:
+        if self.nextitem_index is self.SHUTDOWN_MARK:
             nextitem = None
+        else:
+            nextitem = items[self.nextitem_index]
 
         worker_title("[pytest-xdist running] %s" % item.nodeid)
 
@@ -128,7 +192,7 @@ class WorkerInteractor:
                     if len(mark.args) > 0
                     else mark.kwargs.get("name", "default")
                 )
-                item._nodeid = "{}@{}".format(item.nodeid, gname)
+                item._nodeid = f"{item.nodeid}@{gname}"
 
     @pytest.hookimpl
     def pytest_collection_finish(self, session):

src/xdist/report.py

@@ -14,7 +14,8 @@ def report_collection_diff(from_collection, to_collection, from_id, to_id):
     error_message = (
         "Different tests were collected between {from_id} and {to_id}. "
         "The difference is:\n"
-        "{diff}"
+        "{diff}\n"
+        "To see why this happens see Known limitations in documentation"
     ).format(from_id=from_id, to_id=to_id, diff="\n".join(diff))
     msg = "\n".join(x.rstrip() for x in error_message.split("\n"))
     return msg

src/xdist/scheduler/__init__.py

@@ -3,3 +3,4 @@ from xdist.scheduler.load import LoadScheduling  # noqa
 from xdist.scheduler.loadfile import LoadFileScheduling  # noqa
 from xdist.scheduler.loadscope import LoadScopeScheduling  # noqa
 from xdist.scheduler.loadgroup import LoadGroupScheduling  # noqa
+from xdist.scheduler.worksteal import WorkStealingScheduling  # noqa

src/xdist/scheduler/each.py

@@ -1,5 +1,4 @@
-from py.log import Producer
-
+from xdist.remote import Producer
 from xdist.workermanage import parse_spec_config
 from xdist.report import report_collection_diff
 

src/xdist/scheduler/load.py

@@ -1,8 +1,8 @@
 from itertools import cycle
 
-from py.log import Producer
 from _pytest.runner import CollectReport
 
+from xdist.remote import Producer
 from xdist.workermanage import parse_spec_config
 from xdist.report import report_collection_diff
 
@@ -64,6 +64,7 @@ class LoadScheduling:
         else:
             self.log = log.loadsched
         self.config = config
+        self.maxschedchunk = self.config.getoption("maxschedchunk")
 
     @property
     def nodes(self):
@@ -185,7 +186,9 @@ class LoadScheduling:
                     # so let's rather wait with sending new items
                     return
                 num_send = items_per_node_max - len(node_pending)
-                self._send_tests(node, num_send)
+                # keep at least 2 tests pending even if --maxschedchunk=1
+                maxschedchunk = max(2 - len(node_pending), self.maxschedchunk)
+                self._send_tests(node, min(num_send, maxschedchunk))
         else:
             node.shutdown()
 
@@ -245,16 +248,34 @@ class LoadScheduling:
         if not self.collection:
             return
 
+        if self.maxschedchunk is None:
+            self.maxschedchunk = len(self.collection)
+
         # Send a batch of tests to run. If we don't have at least two
         # tests per node, we have to send them all so that we can send
         # shutdown signals and get all nodes working.
-        initial_batch = max(len(self.pending) // 4, 2 * len(self.nodes))
+        if len(self.pending) < 2 * len(self.nodes):
+            # Distribute tests round-robin. Try to load all nodes if there are
+            # enough tests. The other branch tries sends at least 2 tests
+            # to each node - which is suboptimal when you have less than
+            # 2 * len(nodes) tests.
+            nodes = cycle(self.nodes)
+            for i in range(len(self.pending)):
+                self._send_tests(next(nodes), 1)
+        else:
+            # Send batches of consecutive tests. By default, pytest sorts tests
+            # in order for optimal single-threaded execution, minimizing the
+            # number of necessary fixture setup/teardown. Try to keep that
+            # optimal order for every worker.
 
-        # distribute tests round-robin up to the batch size
-        # (or until we run out)
-        nodes = cycle(self.nodes)
-        for i in range(initial_batch):
-            self._send_tests(next(nodes), 1)
+            # how many items per node do we have about?
+            items_per_node = len(self.collection) // len(self.node2pending)
+            # take a fraction of tests for initial distribution
+            node_chunksize = min(items_per_node // 4, self.maxschedchunk)
+            node_chunksize = max(node_chunksize, 2)
+            # and initialize each node with a chunk of tests
+            for node in self.nodes:
+                self._send_tests(node, node_chunksize)
 
         if not self.pending:
             # initial distribution sent all tests, start node shutdown

src/xdist/scheduler/loadfile.py

@@ -1,5 +1,5 @@
 from .loadscope import LoadScopeScheduling
-from py.log import Producer
+from xdist.remote import Producer
 
 
 class LoadFileScheduling(LoadScopeScheduling):

src/xdist/scheduler/loadgroup.py

@@ -1,5 +1,5 @@
 from .loadscope import LoadScopeScheduling
-from py.log import Producer
+from xdist.remote import Producer
 
 
 class LoadGroupScheduling(LoadScopeScheduling):

src/xdist/scheduler/loadscope.py

@@ -1,7 +1,7 @@
 from collections import OrderedDict
 
 from _pytest.runner import CollectReport
-from py.log import Producer
+from xdist.remote import Producer
 from xdist.report import report_collection_diff
 from xdist.workermanage import parse_spec_config
 
@@ -213,13 +213,11 @@ class LoadScopeScheduling:
 
         # A new node has been added later, perhaps an original one died.
         if self.collection_is_completed:
-
             # Assert that .schedule() should have been called by now
             assert self.collection
 
             # Check that the new collection matches the official collection
             if collection != self.collection:
-
                 other_node = next(iter(self.registered_collections.keys()))
 
                 msg = report_collection_diff(
@@ -361,12 +359,12 @@ class LoadScopeScheduling:
         extra_nodes = len(self.nodes) - len(self.workqueue)
 
         if extra_nodes > 0:
-            self.log("Shuting down {} nodes".format(extra_nodes))
+            self.log(f"Shutting down {extra_nodes} nodes")
 
             for _ in range(extra_nodes):
                 unused_node, assigned = self.assigned_work.popitem(last=True)
 
-                self.log("Shuting down unused node {}".format(unused_node))
+                self.log(f"Shutting down unused node {unused_node}")
                 unused_node.shutdown()
 
         # Assign initial workload

src/xdist/scheduler/worksteal.py

@@ -0,0 +1,325 @@
+from collections import namedtuple
+
+from _pytest.runner import CollectReport
+
+from xdist.remote import Producer
+from xdist.workermanage import parse_spec_config
+from xdist.report import report_collection_diff
+
+
+NodePending = namedtuple("NodePending", ["node", "pending"])
+
+# Every worker needs at least 2 tests in queue - the current and the next one.
+MIN_PENDING = 2
+
+
+class WorkStealingScheduling:
+    """Implement work-stealing scheduling.
+
+    Initially, tests are distributed evenly among all nodes.
+
+    When some node completes most of its assigned tests (when only one pending
+    test remains), an attempt is made to reassign ("steal") some tests from
+    other nodes to this node.
+
+    Attributes:
+
+    :numnodes: The expected number of nodes taking part.  The actual
+       number of nodes will vary during the scheduler's lifetime as
+       nodes are added by the DSession as they are brought up and
+       removed either because of a dead node or normal shutdown.  This
+       number is primarily used to know when the initial collection is
+       completed.
+
+    :node2collection: Map of nodes and their test collection.  All
+       collections should always be identical.
+
+    :node2pending: Map of nodes and the indices of their pending
+       tests.  The indices are an index into ``.pending`` (which is
+       identical to their own collection stored in
+       ``.node2collection``).
+
+    :collection: The one collection once it is validated to be
+       identical between all the nodes.  It is initialised to None
+       until ``.schedule()`` is called.
+
+    :pending: List of indices of globally pending tests.  These are
+       tests which have not yet been allocated to a chunk for a node
+       to process.
+
+    :log: A py.log.Producer instance.
+
+    :config: Config object, used for handling hooks.
+
+    :steal_requested_from_node: The node to which the current "steal" request
+       was sent. ``None`` if there is no request in progress. Only one request
+       can be in progress at any time, the scheduler doesn't send multiple
+       simultaneous requests.
+    """
+
+    def __init__(self, config, log=None):
+        self.numnodes = len(parse_spec_config(config))
+        self.node2collection = {}
+        self.node2pending = {}
+        self.pending = []
+        self.collection = None
+        if log is None:
+            self.log = Producer("workstealsched")
+        else:
+            self.log = log.workstealsched
+        self.config = config
+        self.steal_requested_from_node = None
+
+    @property
+    def nodes(self):
+        """A list of all nodes in the scheduler."""
+        return list(self.node2pending.keys())
+
+    @property
+    def collection_is_completed(self):
+        """Boolean indication initial test collection is complete.
+
+        This is a boolean indicating all initial participating nodes
+        have finished collection.  The required number of initial
+        nodes is defined by ``.numnodes``.
+        """
+        return len(self.node2collection) >= self.numnodes
+
+    @property
+    def tests_finished(self):
+        """Return True if all tests have been executed by the nodes."""
+        if not self.collection_is_completed:
+            return False
+        if self.pending:
+            return False
+        if self.steal_requested_from_node is not None:
+            return False
+        for pending in self.node2pending.values():
+            if len(pending) >= MIN_PENDING:
+                return False
+        return True
+
+    @property
+    def has_pending(self):
+        """Return True if there are pending test items
+
+        This indicates that collection has finished and nodes are
+        still processing test items, so this can be thought of as
+        "the scheduler is active".
+        """
+        if self.pending:
+            return True
+        for pending in self.node2pending.values():
+            if pending:
+                return True
+        return False
+
+    def add_node(self, node):
+        """Add a new node to the scheduler.
+
+        From now on the node will be allocated chunks of tests to
+        execute.
+
+        Called by the ``DSession.worker_workerready`` hook when it
+        successfully bootstraps a new node.
+        """
+        assert node not in self.node2pending
+        self.node2pending[node] = []
+
+    def add_node_collection(self, node, collection):
+        """Add the collected test items from a node
+
+        The collection is stored in the ``.node2collection`` map.
+        Called by the ``DSession.worker_collectionfinish`` hook.
+        """
+        assert node in self.node2pending
+        if self.collection_is_completed:
+            # A new node has been added later, perhaps an original one died.
+            # .schedule() should have
+            # been called by now
+            assert self.collection
+            if collection != self.collection:
+                other_node = next(iter(self.node2collection.keys()))
+                msg = report_collection_diff(
+                    self.collection, collection, other_node.gateway.id, node.gateway.id
+                )
+                self.log(msg)
+                return
+        self.node2collection[node] = list(collection)
+
+    def mark_test_complete(self, node, item_index, duration=None):
+        """Mark test item as completed by node
+
+        This is called by the ``DSession.worker_testreport`` hook.
+        """
+        self.node2pending[node].remove(item_index)
+        self.check_schedule()
+
+    def mark_test_pending(self, item):
+        self.pending.insert(
+            0,
+            self.collection.index(item),
+        )
+        self.check_schedule()
+
+    def remove_pending_tests_from_node(self, node, indices):
+        """Node returned some test indices back in response to 'steal' command.
+
+        This is called by ``DSession.worker_unscheduled``.
+        """
+        assert node is self.steal_requested_from_node
+        self.steal_requested_from_node = None
+
+        indices_set = set(indices)
+        self.node2pending[node] = [
+            i for i in self.node2pending[node] if i not in indices_set
+        ]
+        self.pending.extend(indices)
+        self.check_schedule()
+
+    def check_schedule(self):
+        """Reschedule tests/perform load balancing."""
+        nodes_up = [
+            NodePending(node, pending)
+            for node, pending in self.node2pending.items()
+            if not node.shutting_down
+        ]
+
+        def get_idle_nodes():
+            return [node for node, pending in nodes_up if len(pending) < MIN_PENDING]
+
+        idle_nodes = get_idle_nodes()
+        if not idle_nodes:
+            return
+
+        if self.pending:
+            # Distribute pending tests evenly among idle nodes
+            for i, node in enumerate(idle_nodes):
+                nodes_remaining = len(idle_nodes) - i
+                num_send = len(self.pending) // nodes_remaining
+                self._send_tests(node, num_send)
+
+            idle_nodes = get_idle_nodes()
+            # No need to steal anything if all nodes have enough work to continue
+            if not idle_nodes:
+                return
+
+        # Only one active stealing request is allowed
+        if self.steal_requested_from_node is not None:
+            return
+
+        # Find the node that has the longest test queue
+        steal_from = max(
+            nodes_up, key=lambda node_pending: len(node_pending.pending), default=None
+        )
+
+        if steal_from is None:
+            num_steal = 0
+        else:
+            # Steal half of the test queue - but keep that node running too.
+            # If the node has 2 or less tests queued, stealing will fail
+            # anyway.
+            max_steal = max(0, len(steal_from.pending) - MIN_PENDING)
+            num_steal = min(len(steal_from.pending) // 2, max_steal)
+
+        if num_steal == 0:
+            # Can't get more work - shutdown idle nodes. This will force them
+            # to run the last test now instead of waiting for more tests.
+            for node in idle_nodes:
+                node.shutdown()
+            return
+
+        steal_from.node.send_steal(steal_from.pending[-num_steal:])
+        self.steal_requested_from_node = steal_from.node
+
+    def remove_node(self, node):
+        """Remove a node from the scheduler
+
+        This should be called either when the node crashed or at
+        shutdown time.  In the former case any pending items assigned
+        to the node will be re-scheduled.  Called by the
+        ``DSession.worker_workerfinished`` and
+        ``DSession.worker_errordown`` hooks.
+
+        Return the item which was being executing while the node
+        crashed or None if the node has no more pending items.
+
+        """
+        pending = self.node2pending.pop(node)
+
+        # If node was removed without completing its assigned tests - it crashed
+        if pending:
+            crashitem = self.collection[pending.pop(0)]
+        else:
+            crashitem = None
+
+        self.pending.extend(pending)
+
+        # Dead node won't respond to "steal" request
+        if self.steal_requested_from_node is node:
+            self.steal_requested_from_node = None
+
+        self.check_schedule()
+        return crashitem
+
+    def schedule(self):
+        """Initiate distribution of the test collection
+
+        Initiate scheduling of the items across the nodes.  If this
+        gets called again later it behaves the same as calling
+        ``.check_schedule()`` on all nodes so that newly added nodes
+        will start to be used.
+
+        This is called by the ``DSession.worker_collectionfinish`` hook
+        if ``.collection_is_completed`` is True.
+        """
+        assert self.collection_is_completed
+
+        # Initial distribution already happened, reschedule on all nodes
+        if self.collection is not None:
+            self.check_schedule()
+            return
+
+        if not self._check_nodes_have_same_collection():
+            self.log("**Different tests collected, aborting run**")
+            return
+
+        # Collections are identical, create the index of pending items.
+        self.collection = list(self.node2collection.values())[0]
+        self.pending[:] = range(len(self.collection))
+        if not self.collection:
+            return
+
+        self.check_schedule()
+
+    def _send_tests(self, node, num):
+        tests_per_node = self.pending[:num]
+        if tests_per_node:
+            del self.pending[:num]
+            self.node2pending[node].extend(tests_per_node)
+            node.send_runtest_some(tests_per_node)
+
+    def _check_nodes_have_same_collection(self):
+        """Return True if all nodes have collected the same items.
+
+        If collections differ, this method returns False while logging
+        the collection differences and posting collection errors to
+        pytest_collectreport hook.
+        """
+        node_collection_items = list(self.node2collection.items())
+        first_node, col = node_collection_items[0]
+        same_collection = True
+        for node, collection in node_collection_items[1:]:
+            msg = report_collection_diff(
+                col, collection, first_node.gateway.id, node.gateway.id
+            )
+            if msg:
+                same_collection = False
+                self.log(msg)
+                if self.config is not None:
+                    rep = CollectReport(
+                        node.gateway.id, "failed", longrepr=msg, result=[]
+                    )
+                    self.config.hook.pytest_collectreport(report=rep)
+
+        return same_collection

src/xdist/workermanage.py

@@ -3,12 +3,14 @@ import os
 import re
 import sys
 import uuid
+from pathlib import Path
+from typing import List, Union, Sequence, Optional, Any, Tuple, Set
 
-import py
 import pytest
 import execnet
 
 import xdist.remote
+from xdist.remote import Producer
 from xdist.plugin import _sys_path
 
 
@@ -33,7 +35,7 @@ class NodeManager:
     EXIT_TIMEOUT = 10
     DEFAULT_IGNORES = [".*", "*.pyc", "*.pyo", "*~"]
 
-    def __init__(self, config, specs=None, defaultchdir="pyexecnetcache"):
+    def __init__(self, config, specs=None, defaultchdir="pyexecnetcache") -> None:
         self.config = config
         self.trace = self.config.trace.get("nodemanager")
         self.testrunuid = self.config.getoption("testrunuid")
@@ -52,7 +54,7 @@ class NodeManager:
             self.specs.append(spec)
         self.roots = self._getrsyncdirs()
         self.rsyncoptions = self._getrsyncoptions()
-        self._rsynced_specs = set()
+        self._rsynced_specs: Set[Tuple[Any, Any]] = set()
 
     def rsync_roots(self, gateway):
         """Rsync the set of roots to the node's gateway cwd."""
@@ -81,7 +83,7 @@ class NodeManager:
     def _getxspecs(self):
         return [execnet.XSpec(x) for x in parse_spec_config(self.config)]
 
-    def _getrsyncdirs(self):
+    def _getrsyncdirs(self) -> List[Path]:
         for spec in self.specs:
             if not spec.popen or spec.chdir:
                 break
@@ -101,16 +103,16 @@ class NodeManager:
         pytestpath = get_dir(pytest.__file__)
         pytestdir = get_dir(_pytest.__file__)
         config = self.config
-        candidates = [py._pydir, pytestpath, pytestdir]
+        candidates = [pytestpath, pytestdir]
         candidates += config.option.rsyncdir
         rsyncroots = config.getini("rsyncdirs")
         if rsyncroots:
             candidates.extend(rsyncroots)
         roots = []
         for root in candidates:
-            root = py.path.local(root).realpath()
-            if not root.check():
-                raise pytest.UsageError("rsyncdir doesn't exist: {!r}".format(root))
+            root = Path(root).resolve()
+            if not root.exists():
+                raise pytest.UsageError(f"rsyncdir doesn't exist: {root!r}")
             if root not in roots:
                 roots.append(root)
         return roots
@@ -160,18 +162,24 @@ class NodeManager:
 class HostRSync(execnet.RSync):
     """RSyncer that filters out common files"""
 
-    def __init__(self, sourcedir, *args, **kwargs):
-        self._synced = {}
-        ignores = kwargs.pop("ignores", None) or []
-        self._ignores = [
-            re.compile(fnmatch.translate(getattr(x, "strpath", x))) for x in ignores
-        ]
-        super().__init__(sourcedir=sourcedir, **kwargs)
+    PathLike = Union[str, "os.PathLike[str]"]
 
-    def filter(self, path):
-        path = py.path.local(path)
+    def __init__(
+        self,
+        sourcedir: PathLike,
+        *,
+        ignores: Optional[Sequence[PathLike]] = None,
+        **kwargs: object
+    ) -> None:
+        if ignores is None:
+            ignores = []
+        self._ignores = [re.compile(fnmatch.translate(os.fspath(x))) for x in ignores]
+        super().__init__(sourcedir=Path(sourcedir), **kwargs)
+
+    def filter(self, path: PathLike) -> bool:
+        path = Path(path)
         for cre in self._ignores:
-            if cre.match(path.basename) or cre.match(path.strpath):
+            if cre.match(path.name) or cre.match(str(path)):
                 return False
         else:
             return True
@@ -184,26 +192,34 @@ class HostRSync(execnet.RSync):
         if self._verbose > 0:
             path = os.path.basename(self._sourcedir) + "/" + modified_rel_path
             remotepath = gateway.spec.chdir
-            print("{}:{} <= {}".format(gateway.spec, remotepath, path))
+            print(f"{gateway.spec}:{remotepath} <= {path}")
 
 
-def make_reltoroot(roots, args):
+def make_reltoroot(roots: Sequence[Path], args: List[str]) -> List[str]:
     # XXX introduce/use public API for splitting pytest args
     splitcode = "::"
     result = []
     for arg in args:
         parts = arg.split(splitcode)
-        fspath = py.path.local(parts[0])
-        if not fspath.exists():
+        fspath = Path(parts[0])
+        try:
+            exists = fspath.exists()
+        except OSError:
+            exists = False
+        if not exists:
             result.append(arg)
             continue
         for root in roots:
-            x = fspath.relto(root)
+            x: Optional[Path]
+            try:
+                x = fspath.relative_to(root)
+            except ValueError:
+                x = None
             if x or fspath == root:
-                parts[0] = root.basename + "/" + x
+                parts[0] = root.name + "/" + str(x)
                 break
         else:
-            raise ValueError("arg {} not relative to an rsync root".format(arg))
+            raise ValueError(f"arg {arg} not relative to an rsync root")
         result.append(splitcode.join(parts))
     return result
 
@@ -230,12 +246,10 @@ class WorkerController:
         }
         self._down = False
         self._shutdown_sent = False
-        self.log = py.log.Producer("workerctl-%s" % gateway.id)
-        if not self.config.option.debug:
-            py.log.setconsumer(self.log._keywords, None)
+        self.log = Producer(f"workerctl-{gateway.id}", enabled=config.option.debug)
 
     def __repr__(self):
-        return "<{} {}>".format(self.__class__.__name__, self.gateway.id)
+        return f"<{self.__class__.__name__} {self.gateway.id}>"
 
     @property
     def shutting_down(self):
@@ -286,6 +300,9 @@ class WorkerController:
     def send_runtest_all(self):
         self.sendcommand("runtests_all")
 
+    def send_steal(self, indices):
+        self.sendcommand("steal", indices=indices)
+
     def shutdown(self):
         if not self._down:
             try:
@@ -296,11 +313,11 @@ class WorkerController:
 
     def sendcommand(self, name, **kwargs):
         """send a named parametrized command to the other side."""
-        self.log("sending command {}(**{})".format(name, kwargs))
+        self.log(f"sending command {name}(**{kwargs})")
         self.channel.send((name, kwargs))
 
     def notify_inproc(self, eventname, **kwargs):
-        self.log("queuing {}(**{})".format(eventname, kwargs))
+        self.log(f"queuing {eventname}(**{kwargs})")
         self.putevent((eventname, kwargs))
 
     def process_from_remote(self, eventcall):  # noqa too complex
@@ -322,7 +339,7 @@ class WorkerController:
                 return
             eventname, kwargs = eventcall
             if eventname in ("collectionstart",):
-                self.log("ignoring {}({})".format(eventname, kwargs))
+                self.log(f"ignoring {eventname}({kwargs})")
             elif eventname == "workerready":
                 self.notify_inproc(eventname, node=self, **kwargs)
             elif eventname == "internal_error":
@@ -345,6 +362,8 @@ class WorkerController:
                 self.notify_inproc(eventname, node=self, ids=kwargs["ids"])
             elif eventname == "runtest_protocol_complete":
                 self.notify_inproc(eventname, node=self, **kwargs)
+            elif eventname == "unscheduled":
+                self.notify_inproc(eventname, node=self, **kwargs)
             elif eventname == "logwarning":
                 self.notify_inproc(
                     eventname,
@@ -375,7 +394,7 @@ class WorkerController:
                     location=kwargs["location"],
                 )
             else:
-                raise ValueError("unknown event: {}".format(eventname))
+                raise ValueError(f"unknown event: {eventname}")
         except KeyboardInterrupt:
             # should not land in receiver-thread
             raise

testing/acceptance_test.py

@@ -101,7 +101,12 @@ class TestDistribution:
             """
         )
         result = pytester.runpytest(p1, "-v", "-d", "--tx=popen", "--tx=popen")
-        result.stdout.fnmatch_lines(["*1*Python*", "*2 failed, 1 passed, 1 skipped*"])
+        result.stdout.fnmatch_lines(
+            [
+                "created: 2/2 workers",
+                "*2 failed, 1 passed, 1 skipped*",
+            ]
+        )
         assert result.ret == 1
 
     def test_n1_fail_minus_x(self, pytester: pytest.Pytester) -> None:
@@ -151,10 +156,14 @@ class TestDistribution:
         """
         )
         result = pytester.runpytest(p1, "-d", "-v")
-        result.stdout.fnmatch_lines(["*2*Python*", "*2 failed, 1 passed, 1 skipped*"])
+        result.stdout.fnmatch_lines(
+            [
+                "created: 3/3 workers",
+                "*2 failed, 1 passed, 1 skipped*",
+            ]
+        )
         assert result.ret == 1
 
-    @pytest.mark.xfail("sys.platform.startswith('java')", run=False)
     def test_dist_tests_with_crash(self, pytester: pytest.Pytester) -> None:
         if not hasattr(os, "kill"):
             pytest.skip("no os.kill")
@@ -238,9 +247,6 @@ class TestDistribution:
         assert result.ret == 0
         result.stdout.fnmatch_lines(
             [
-                "*0* *cwd*",
-                # "RSyncStart: [G1]",
-                # "RSyncFinished: [G1]",
                 "*1 passed*",
             ]
         )
@@ -277,7 +283,11 @@ class TestDistribution:
         p1 = pytester.makepyfile("def test_func(): pass")
         result = pytester.runpytest("-v", p1, "-d", "--tx=popen")
         result.stdout.fnmatch_lines(
-            ["*0*Python*", "*calculated result is 49*", "*1 passed*"]
+            [
+                "created: 1/1 worker",
+                "*calculated result is 49*",
+                "*1 passed*",
+            ]
         )
         assert result.ret == 0
 
@@ -394,14 +404,14 @@ class TestTerminalReporting:
         out = result.stdout.str()
         if verbosity == "-v":
             assert "scheduling tests" in out
-            assert "gw" in out
+            assert "1 worker [1 item]" in out
         elif verbosity == "-q":
             assert "scheduling tests" not in out
             assert "gw" not in out
             assert "bringing up nodes..." in out
         else:
             assert "scheduling tests" not in out
-            assert "gw" in out
+            assert "1 worker [1 item]" in out
 
     def test_pass_skip_fail(self, pytester: pytest.Pytester) -> None:
         pytester.makepyfile(
@@ -603,11 +613,10 @@ def test_config_initialization(
 def test_crashing_item(pytester, when) -> None:
     """Ensure crashing item is correctly reported during all testing stages"""
     code = dict(setup="", call="", teardown="")
-    code[when] = "py.process.kill(os.getpid())"
+    code[when] = "os._exit(1)"
     p = pytester.makepyfile(
         """
         import os
-        import py
         import pytest
 
         @pytest.fixture
@@ -742,7 +751,7 @@ def test_sub_plugins_disabled(pytester, plugin) -> None:
             pass
     """
     )
-    result = pytester.runpytest(p1, "-n1", "-p", "no:%s" % plugin)
+    result = pytester.runpytest(p1, "-n1", "-p", f"no:{plugin}")
     assert result.ret == 0
     result.stdout.fnmatch_lines("*1 passed*")
 
@@ -1101,8 +1110,9 @@ def test_color_yes_collection_on_non_atty(pytester, request) -> None:
     result = pytester.runpytest(*args)
     assert "test session starts" in result.stdout.str()
     assert "\x1b[1m" in result.stdout.str()
-    assert "gw0 [10] / gw1 [10]" in result.stdout.str()
-    assert "gw0 C / gw1 C" not in result.stdout.str()
+    assert "created: 2/2 workers" in result.stdout.str()
+    assert "2 workers [10 items]" in result.stdout.str()
+    assert "collecting:" not in result.stdout.str()
 
 
 def test_without_terminal_plugin(pytester, request) -> None:
@@ -1245,22 +1255,14 @@ class TestFileScope:
             "test_b.py::TestB", result.outlines
         )
 
-        assert (
-            test_a_workers_and_test_count
-            in (
-                {"gw0": 10},
-                {"gw1": 0},
-            )
-            or test_a_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
-        )
-        assert (
-            test_b_workers_and_test_count
-            in (
-                {"gw0": 10},
-                {"gw1": 0},
-            )
-            or test_b_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
-        )
+        assert test_a_workers_and_test_count in (
+            {"gw0": 10},
+            {"gw1": 0},
+        ) or test_a_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
+        assert test_b_workers_and_test_count in (
+            {"gw0": 10},
+            {"gw1": 0},
+        ) or test_b_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
 
     def test_by_class(self, pytester: pytest.Pytester) -> None:
         pytester.makepyfile(
@@ -1285,22 +1287,14 @@ class TestFileScope:
             "test_a.py::TestB", result.outlines
         )
 
-        assert (
-            test_a_workers_and_test_count
-            in (
-                {"gw0": 10},
-                {"gw1": 0},
-            )
-            or test_a_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
-        )
-        assert (
-            test_b_workers_and_test_count
-            in (
-                {"gw0": 10},
-                {"gw1": 0},
-            )
-            or test_b_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
-        )
+        assert test_a_workers_and_test_count in (
+            {"gw0": 10},
+            {"gw1": 0},
+        ) or test_a_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
+        assert test_b_workers_and_test_count in (
+            {"gw0": 10},
+            {"gw1": 0},
+        ) or test_b_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
 
     def test_module_single_start(self, pytester: pytest.Pytester) -> None:
         """Fix test suite never finishing in case all workers start with a single test (#277)."""
@@ -1347,22 +1341,14 @@ class TestGroupScope:
             "test_b.py::TestA", result.outlines
         )
 
-        assert (
-            test_a_workers_and_test_count
-            in (
-                {"gw0": 5},
-                {"gw1": 0},
-            )
-            or test_a_workers_and_test_count in ({"gw0": 0}, {"gw1": 5})
-        )
-        assert (
-            test_b_workers_and_test_count
-            in (
-                {"gw0": 5},
-                {"gw1": 0},
-            )
-            or test_b_workers_and_test_count in ({"gw0": 0}, {"gw1": 5})
-        )
+        assert test_a_workers_and_test_count in (
+            {"gw0": 5},
+            {"gw1": 0},
+        ) or test_a_workers_and_test_count in ({"gw0": 0}, {"gw1": 5})
+        assert test_b_workers_and_test_count in (
+            {"gw0": 5},
+            {"gw1": 0},
+        ) or test_b_workers_and_test_count in ({"gw0": 0}, {"gw1": 5})
         assert (
             test_a_workers_and_test_count.items()
             == test_b_workers_and_test_count.items()
@@ -1392,22 +1378,14 @@ class TestGroupScope:
             "test_a.py::TestB", result.outlines
         )
 
-        assert (
-            test_a_workers_and_test_count
-            in (
-                {"gw0": 10},
-                {"gw1": 0},
-            )
-            or test_a_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
-        )
-        assert (
-            test_b_workers_and_test_count
-            in (
-                {"gw0": 10},
-                {"gw1": 0},
-            )
-            or test_b_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
-        )
+        assert test_a_workers_and_test_count in (
+            {"gw0": 10},
+            {"gw1": 0},
+        ) or test_a_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
+        assert test_b_workers_and_test_count in (
+            {"gw0": 10},
+            {"gw1": 0},
+        ) or test_b_workers_and_test_count in ({"gw0": 0}, {"gw1": 10})
         assert (
             test_a_workers_and_test_count.items()
             == test_b_workers_and_test_count.items()
@@ -1576,3 +1554,40 @@ class TestAPI:
         assert xdist.get_xdist_worker_id(fake_request) == "gw5"
         del fake_request.config.workerinput
         assert xdist.get_xdist_worker_id(fake_request) == "master"
+
+
+def test_collection_crash(testdir):
+    p1 = testdir.makepyfile(
+        """
+        assert 0
+    """
+    )
+    result = testdir.runpytest(p1, "-n1")
+    assert result.ret == 1
+    result.stdout.fnmatch_lines(
+        [
+            "created: 1/1 worker",
+            "1 worker [[]0 items[]]",
+            "*_ ERROR collecting test_collection_crash.py _*",
+            "E   assert 0",
+            "*= 1 error in *",
+        ]
+    )
+
+
+def test_dist_in_addopts(testdir):
+    """Users can set a default distribution in the configuration file (#789)."""
+    testdir.makepyfile(
+        """
+        def test():
+            pass
+        """
+    )
+    testdir.makeini(
+        """
+        [pytest]
+        addopts = --dist loadscope
+        """
+    )
+    result = testdir.runpytest()
+    assert result.ret == 0

testing/test_dsession.py

@@ -1,7 +1,13 @@
-from xdist.dsession import DSession, get_default_max_worker_restart
+from __future__ import annotations
+from xdist.dsession import (
+    DSession,
+    get_default_max_worker_restart,
+    get_workers_status_line,
+    WorkerStatus,
+)
 from xdist.report import report_collection_diff
-from xdist.scheduler import EachScheduling, LoadScheduling
-from typing import Optional
+from xdist.scheduler import EachScheduling, LoadScheduling, WorkStealingScheduling
+from typing import Sequence
 
 import pytest
 import execnet
@@ -17,6 +23,7 @@ class MockGateway:
 class MockNode:
     def __init__(self) -> None:
         self.sent = []  # type: ignore[var-annotated]
+        self.stolen = []  # type: ignore[var-annotated]
         self.gateway = MockGateway()
         self._shutdown = False
 
@@ -26,6 +33,9 @@ class MockNode:
     def send_runtest_all(self) -> None:
         self.sent.append("ALL")
 
+    def send_steal(self, indices) -> None:
+        self.stolen.extend(indices)
+
     def shutdown(self) -> None:
         self._shutdown = True
 
@@ -115,44 +125,93 @@ class TestLoadScheduling:
         # assert not sched.tests_finished
         sent1 = node1.sent
         sent2 = node2.sent
-        assert sent1 == [0, 2]
-        assert sent2 == [1, 3]
+        assert sent1 == [0, 1]
+        assert sent2 == [2, 3]
         assert sched.pending == [4, 5]
         assert sched.node2pending[node1] == sent1
         assert sched.node2pending[node2] == sent2
         assert len(sched.pending) == 2
         sched.mark_test_complete(node1, 0)
-        assert node1.sent == [0, 2, 4]
+        assert node1.sent == [0, 1, 4]
         assert sched.pending == [5]
-        assert node2.sent == [1, 3]
-        sched.mark_test_complete(node1, 2)
-        assert node1.sent == [0, 2, 4, 5]
+        assert node2.sent == [2, 3]
+        sched.mark_test_complete(node1, 1)
+        assert node1.sent == [0, 1, 4, 5]
         assert not sched.pending
 
-    def test_schedule_fewer_tests_than_nodes(self, pytester: pytest.Pytester) -> None:
+    def test_schedule_maxchunk_none(self, pytester: pytest.Pytester) -> None:
         config = pytester.parseconfig("--tx=2*popen")
         sched = LoadScheduling(config)
         sched.add_node(MockNode())
         sched.add_node(MockNode())
+        node1, node2 = sched.nodes
+        col = [f"test{i}" for i in range(16)]
+        sched.add_node_collection(node1, col)
+        sched.add_node_collection(node2, col)
+        sched.schedule()
+        assert node1.sent == [0, 1]
+        assert node2.sent == [2, 3]
+        assert sched.pending == list(range(4, 16))
+        assert sched.node2pending[node1] == node1.sent
+        assert sched.node2pending[node2] == node2.sent
+        sched.mark_test_complete(node1, 0)
+        assert node1.sent == [0, 1, 4, 5]
+        assert sched.pending == list(range(6, 16))
+        sched.mark_test_complete(node1, 1)
+        assert node1.sent == [0, 1, 4, 5]
+        assert sched.pending == list(range(6, 16))
+
+        for i in range(7, 16):
+            sched.mark_test_complete(node1, i - 3)
+            assert node1.sent == [0, 1] + list(range(4, i))
+            assert node2.sent == [2, 3]
+            assert sched.pending == list(range(i, 16))
+
+    def test_schedule_maxchunk_1(self, pytester: pytest.Pytester) -> None:
+        config = pytester.parseconfig("--tx=2*popen", "--maxschedchunk=1")
+        sched = LoadScheduling(config)
+        sched.add_node(MockNode())
+        sched.add_node(MockNode())
+        node1, node2 = sched.nodes
+        col = [f"test{i}" for i in range(16)]
+        sched.add_node_collection(node1, col)
+        sched.add_node_collection(node2, col)
+        sched.schedule()
+        assert node1.sent == [0, 1]
+        assert node2.sent == [2, 3]
+        assert sched.pending == list(range(4, 16))
+        assert sched.node2pending[node1] == node1.sent
+        assert sched.node2pending[node2] == node2.sent
+
+        for complete_index, first_pending in enumerate(range(5, 16)):
+            sched.mark_test_complete(node1, node1.sent[complete_index])
+            assert node1.sent == [0, 1] + list(range(4, first_pending))
+            assert node2.sent == [2, 3]
+            assert sched.pending == list(range(first_pending, 16))
+
+    def test_schedule_fewer_tests_than_nodes(self, pytester: pytest.Pytester) -> None:
+        config = pytester.parseconfig("--tx=3*popen")
+        sched = LoadScheduling(config)
+        sched.add_node(MockNode())
+        sched.add_node(MockNode())
         sched.add_node(MockNode())
         node1, node2, node3 = sched.nodes
         col = ["xyz"] * 2
         sched.add_node_collection(node1, col)
         sched.add_node_collection(node2, col)
+        sched.add_node_collection(node3, col)
+        assert sched.collection_is_completed
         sched.schedule()
         # assert not sched.tests_finished
-        sent1 = node1.sent
-        sent2 = node2.sent
-        sent3 = node3.sent
-        assert sent1 == [0]
-        assert sent2 == [1]
-        assert sent3 == []
+        assert node1.sent == [0]
+        assert node2.sent == [1]
+        assert node3.sent == []
         assert not sched.pending
 
     def test_schedule_fewer_than_two_tests_per_node(
         self, pytester: pytest.Pytester
     ) -> None:
-        config = pytester.parseconfig("--tx=2*popen")
+        config = pytester.parseconfig("--tx=3*popen")
         sched = LoadScheduling(config)
         sched.add_node(MockNode())
         sched.add_node(MockNode())
@@ -161,14 +220,13 @@ class TestLoadScheduling:
         col = ["xyz"] * 5
         sched.add_node_collection(node1, col)
         sched.add_node_collection(node2, col)
+        sched.add_node_collection(node3, col)
+        assert sched.collection_is_completed
         sched.schedule()
         # assert not sched.tests_finished
-        sent1 = node1.sent
-        sent2 = node2.sent
-        sent3 = node3.sent
-        assert sent1 == [0, 3]
-        assert sent2 == [1, 4]
-        assert sent3 == [2]
+        assert node1.sent == [0, 3]
+        assert node2.sent == [1, 4]
+        assert node3.sent == [2]
         assert not sched.pending
 
     def test_add_remove_node(self, pytester: pytest.Pytester) -> None:
@@ -217,6 +275,169 @@ class TestLoadScheduling:
         assert "Different tests were collected between" in rep.longrepr
 
 
+class TestWorkStealingScheduling:
+    def test_ideal_case(self, pytester: pytest.Pytester) -> None:
+        config = pytester.parseconfig("--tx=2*popen")
+        sched = WorkStealingScheduling(config)
+        sched.add_node(MockNode())
+        sched.add_node(MockNode())
+        node1, node2 = sched.nodes
+        collection = [f"test_workstealing.py::test_{i}" for i in range(16)]
+        assert not sched.collection_is_completed
+        sched.add_node_collection(node1, collection)
+        assert not sched.collection_is_completed
+        sched.add_node_collection(node2, collection)
+        assert sched.collection_is_completed
+        assert sched.node2collection[node1] == collection
+        assert sched.node2collection[node2] == collection
+        sched.schedule()
+        assert not sched.pending
+        assert not sched.tests_finished
+        assert node1.sent == list(range(0, 8))
+        assert node2.sent == list(range(8, 16))
+        for i in range(8):
+            sched.mark_test_complete(node1, node1.sent[i])
+            sched.mark_test_complete(node2, node2.sent[i])
+        assert sched.tests_finished
+        assert node1.stolen == []
+        assert node2.stolen == []
+
+    def test_stealing(self, pytester: pytest.Pytester) -> None:
+        config = pytester.parseconfig("--tx=2*popen")
+        sched = WorkStealingScheduling(config)
+        sched.add_node(MockNode())
+        sched.add_node(MockNode())
+        node1, node2 = sched.nodes
+        collection = [f"test_workstealing.py::test_{i}" for i in range(16)]
+        sched.add_node_collection(node1, collection)
+        sched.add_node_collection(node2, collection)
+        assert sched.collection_is_completed
+        sched.schedule()
+        assert node1.sent == list(range(0, 8))
+        assert node2.sent == list(range(8, 16))
+        for i in range(8):
+            sched.mark_test_complete(node1, node1.sent[i])
+        assert node2.stolen == list(range(12, 16))
+        sched.remove_pending_tests_from_node(node2, node2.stolen)
+        for i in range(4):
+            sched.mark_test_complete(node2, node2.sent[i])
+        assert node1.stolen == [14, 15]
+        sched.remove_pending_tests_from_node(node1, node1.stolen)
+        sched.mark_test_complete(node1, 12)
+        sched.mark_test_complete(node2, 14)
+        assert node2.stolen == list(range(12, 16))
+        assert node1.stolen == [14, 15]
+        assert sched.tests_finished
+
+    def test_steal_on_add_node(self, pytester: pytest.Pytester) -> None:
+        node = MockNode()
+        config = pytester.parseconfig("--tx=popen")
+        sched = WorkStealingScheduling(config)
+        sched.add_node(node)
+        collection = [f"test_workstealing.py::test_{i}" for i in range(5)]
+        sched.add_node_collection(node, collection)
+        assert sched.collection_is_completed
+        sched.schedule()
+        assert not sched.pending
+        sched.mark_test_complete(node, 0)
+        node2 = MockNode()
+        sched.add_node(node2)
+        sched.add_node_collection(node2, collection)
+        assert sched.collection_is_completed
+        sched.schedule()
+        assert node.stolen == [3, 4]
+        sched.remove_pending_tests_from_node(node, node.stolen)
+        sched.mark_test_complete(node, 1)
+        sched.mark_test_complete(node2, 3)
+        assert sched.tests_finished
+        assert node2.stolen == []
+
+    def test_schedule_fewer_tests_than_nodes(self, pytester: pytest.Pytester) -> None:
+        config = pytester.parseconfig("--tx=3*popen")
+        sched = WorkStealingScheduling(config)
+        sched.add_node(MockNode())
+        sched.add_node(MockNode())
+        sched.add_node(MockNode())
+        node1, node2, node3 = sched.nodes
+        col = ["xyz"] * 2
+        sched.add_node_collection(node1, col)
+        sched.add_node_collection(node2, col)
+        sched.add_node_collection(node3, col)
+        sched.schedule()
+        assert node1.sent == []
+        assert node1.stolen == []
+        assert node2.sent == [0]
+        assert node2.stolen == []
+        assert node3.sent == [1]
+        assert node3.stolen == []
+        assert not sched.pending
+        assert sched.tests_finished
+
+    def test_schedule_fewer_than_two_tests_per_node(
+        self, pytester: pytest.Pytester
+    ) -> None:
+        config = pytester.parseconfig("--tx=3*popen")
+        sched = WorkStealingScheduling(config)
+        sched.add_node(MockNode())
+        sched.add_node(MockNode())
+        sched.add_node(MockNode())
+        node1, node2, node3 = sched.nodes
+        col = ["xyz"] * 5
+        sched.add_node_collection(node1, col)
+        sched.add_node_collection(node2, col)
+        sched.add_node_collection(node3, col)
+        sched.schedule()
+        assert node1.sent == [0]
+        assert node2.sent == [1, 2]
+        assert node3.sent == [3, 4]
+        assert not sched.pending
+        assert not sched.tests_finished
+        sched.mark_test_complete(node1, node1.sent[0])
+        sched.mark_test_complete(node2, node2.sent[0])
+        sched.mark_test_complete(node3, node3.sent[0])
+        sched.mark_test_complete(node3, node3.sent[1])
+        assert sched.tests_finished
+        assert node1.stolen == []
+        assert node2.stolen == []
+        assert node3.stolen == []
+
+    def test_add_remove_node(self, pytester: pytest.Pytester) -> None:
+        node = MockNode()
+        config = pytester.parseconfig("--tx=popen")
+        sched = WorkStealingScheduling(config)
+        sched.add_node(node)
+        collection = ["test_file.py::test_func"]
+        sched.add_node_collection(node, collection)
+        assert sched.collection_is_completed
+        sched.schedule()
+        assert not sched.pending
+        crashitem = sched.remove_node(node)
+        assert crashitem == collection[0]
+
+    def test_different_tests_collected(self, pytester: pytest.Pytester) -> None:
+        class CollectHook:
+            def __init__(self):
+                self.reports = []
+
+            def pytest_collectreport(self, report):
+                self.reports.append(report)
+
+        collect_hook = CollectHook()
+        config = pytester.parseconfig("--tx=2*popen")
+        config.pluginmanager.register(collect_hook, "collect_hook")
+        node1 = MockNode()
+        node2 = MockNode()
+        sched = WorkStealingScheduling(config)
+        sched.add_node(node1)
+        sched.add_node(node2)
+        sched.add_node_collection(node1, ["a.py::test_1"])
+        sched.add_node_collection(node2, ["a.py::test_2"])
+        sched.schedule()
+        assert len(collect_hook.reports) == 1
+        rep = collect_hook.reports[0]
+        assert "Different tests were collected between" in rep.longrepr
+
+
 class TestDistReporter:
     @pytest.mark.xfail
     def test_rsync_printing(self, pytester: pytest.Pytester, linecomp) -> None:
@@ -258,7 +479,7 @@ def test_report_collection_diff_equal() -> None:
 def test_default_max_worker_restart() -> None:
     class config:
         class option:
-            maxworkerrestart: Optional[str] = None
+            maxworkerrestart: str | None = None
             numprocesses: int = 0
 
     assert get_default_max_worker_restart(config) is None
@@ -290,7 +511,8 @@ def test_report_collection_diff_different() -> None:
         " bbb\n"
         "+XXX\n"
         " ccc\n"
-        "-YYY"
+        "-YYY\n"
+        "To see why this happens see Known limitations in documentation"
     )
 
     msg = report_collection_diff(from_collection, to_collection, "1", "2")
@@ -311,3 +533,70 @@ def test_pytest_issue419(pytester: pytest.Pytester) -> None:
     reprec = pytester.inline_run("-n1")
     reprec.assertoutcome(passed=2)
     assert 0
+
+
+Created = WorkerStatus.Created
+Initialized = WorkerStatus.Initialized
+ReadyForCollection = WorkerStatus.ReadyForCollection
+CollectionDone = WorkerStatus.CollectionDone
+
+
+@pytest.mark.parametrize(
+    "status_and_items, expected",
+    [
+        (
+            [],
+            "",
+        ),
+        (
+            [(Created, 0)],
+            "created: 1/1 worker",
+        ),
+        (
+            [(Created, 0), (Created, 0)],
+            "created: 2/2 workers",
+        ),
+        (
+            [(Initialized, 0), (Created, 0)],
+            "initialized: 1/2 workers",
+        ),
+        (
+            [(Initialized, 0), (Initialized, 0)],
+            "initialized: 2/2 workers",
+        ),
+        (
+            [(ReadyForCollection, 0), (Created, 0)],
+            "ready: 1/2 workers",
+        ),
+        (
+            [(ReadyForCollection, 0), (ReadyForCollection, 0)],
+            "ready: 2/2 workers",
+        ),
+        (
+            [(CollectionDone, 12), (Created, 0)],
+            "collecting: 1/2 workers",
+        ),
+        (
+            [(CollectionDone, 12), (CollectionDone, 12)],
+            "2 workers [12 items]",
+        ),
+        (
+            [(CollectionDone, 1), (CollectionDone, 1)],
+            "2 workers [1 item]",
+        ),
+        (
+            [(CollectionDone, 1)],
+            "1 worker [1 item]",
+        ),
+        # Different number of tests collected will raise an error and should not happen in practice,
+        # but we test for it anyway.
+        (
+            [(CollectionDone, 1), (CollectionDone, 12)],
+            "2 workers [1 item]",
+        ),
+    ],
+)
+def test_get_workers_status_line(
+    status_and_items: Sequence[tuple[WorkerStatus, int]], expected: str
+) -> None:
+    assert get_workers_status_line(status_and_items) == expected

testing/test_looponfail.py

@@ -1,4 +1,6 @@
-import py
+import unittest.mock
+from typing import List
+
 import pytest
 import shutil
 import textwrap
@@ -16,7 +18,7 @@ class TestStatRecorder:
         tmp = tmp_path
         hello = tmp / "hello.py"
         hello.touch()
-        sd = StatRecorder([py.path.local(tmp)])
+        sd = StatRecorder([tmp])
         changed = sd.check()
         assert not changed
 
@@ -56,15 +58,12 @@ class TestStatRecorder:
         tmp = tmp_path
         tmp.joinpath("dir").mkdir()
         tmp.joinpath("dir", "hello.py").touch()
-        sd = StatRecorder([py.path.local(tmp)])
-        assert not sd.fil(py.path.local(tmp / "dir"))
+        sd = StatRecorder([tmp])
+        assert not sd.fil(tmp / "dir")
 
-    def test_filechange_deletion_race(
-        self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch
-    ) -> None:
+    def test_filechange_deletion_race(self, tmp_path: Path) -> None:
         tmp = tmp_path
-        pytmp = py.path.local(tmp)
-        sd = StatRecorder([pytmp])
+        sd = StatRecorder([tmp])
         changed = sd.check()
         assert not changed
 
@@ -76,16 +75,20 @@ class TestStatRecorder:
         p.unlink()
         # make check()'s visit() call return our just removed
         # path as if we were in a race condition
-        monkeypatch.setattr(pytmp, "visit", lambda *args: [py.path.local(p)])
-
-        changed = sd.check()
+        dirname = str(tmp)
+        dirnames: List[str] = []
+        filenames = [str(p)]
+        with unittest.mock.patch(
+            "os.walk", return_value=[(dirname, dirnames, filenames)], autospec=True
+        ):
+            changed = sd.check()
         assert changed
 
     def test_pycremoval(self, tmp_path: Path) -> None:
         tmp = tmp_path
         hello = tmp / "hello.py"
         hello.touch()
-        sd = StatRecorder([py.path.local(tmp)])
+        sd = StatRecorder([tmp])
         changed = sd.check()
         assert not changed
 
@@ -100,7 +103,7 @@ class TestStatRecorder:
         self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch
     ) -> None:
         tmp = tmp_path
-        sd = StatRecorder([py.path.local(tmp)])
+        sd = StatRecorder([tmp])
 
         ret_values = [True, False]
         monkeypatch.setattr(StatRecorder, "check", lambda self: ret_values.pop())
@@ -123,7 +126,7 @@ class TestRemoteControl:
         failures = control.runsession()
         assert failures
         control.setup()
-        item_path = item.path if PYTEST_GTE_7 else Path(item.fspath)  # type: ignore[attr-defined]
+        item_path = item.path if PYTEST_GTE_7 else Path(str(item.fspath))  # type: ignore[attr-defined]
         item_path.write_text("def test_func():\n assert 1\n")
         removepyc(item_path)
         topdir, failures = control.runsession()[:2]
@@ -141,7 +144,11 @@ class TestRemoteControl:
         control = RemoteControl(modcol.config)
         control.loop_once()
         assert control.failures
-        modcol_path = modcol.path if PYTEST_GTE_7 else Path(modcol.fspath)  # type: ignore[attr-defined]
+        if PYTEST_GTE_7:
+            modcol_path = modcol.path  # type:ignore[attr-defined]
+        else:
+            modcol_path = Path(str(modcol.fspath))
+
         modcol_path.write_text(
             textwrap.dedent(
                 """

testing/test_newhooks.py

@@ -94,3 +94,28 @@ class TestCrashItem:
         res = pytester.runpytest("-n2", "-s")
         res.stdout.fnmatch_lines_random(["*HOOK: pytest_handlecrashitem"])
         res.stdout.fnmatch_lines(["*3 passed*"])
+
+    def test_handlecrashitem_one(self, pytester: pytest.Pytester) -> None:
+        """Test pytest_handlecrashitem hook with just one test."""
+        pytester.makeconftest(
+            """
+            test_runs = 0
+
+            def pytest_handlecrashitem(crashitem, report, sched):
+                global test_runs
+
+                if test_runs == 0:
+                    sched.mark_test_pending(crashitem)
+                    test_runs = 1
+                else:
+                    print("HOOK: pytest_handlecrashitem")
+        """
+        )
+        res = pytester.runpytest("-n1", "-s", "-k", "test_b")
+        res.stdout.fnmatch_lines_random(["*HOOK: pytest_handlecrashitem"])
+        res.stdout.fnmatch_lines(
+            [
+                "FAILED test_handlecrashitem_one.py::test_b",
+                "FAILED test_handlecrashitem_one.py::test_b",
+            ]
+        )

testing/test_plugin.py

@@ -1,5 +1,7 @@
 from contextlib import suppress
 from pathlib import Path
+import sys
+import os
 
 import execnet
 from xdist.workermanage import NodeManager
@@ -7,6 +9,15 @@ from xdist.workermanage import NodeManager
 import pytest
 
 
+@pytest.fixture
+def monkeypatch_3_cpus(monkeypatch: pytest.MonkeyPatch):
+    """Make pytest-xdist believe the system has 3 CPUs"""
+    # block import
+    monkeypatch.setitem(sys.modules, "psutil", None)  # type: ignore
+    monkeypatch.delattr(os, "sched_getaffinity", raising=False)
+    monkeypatch.setattr(os, "cpu_count", lambda: 3)
+
+
 def test_dist_incompatibility_messages(pytester: pytest.Pytester) -> None:
     result = pytester.runpytest("--pdb", "--looponfail")
     assert result.ret != 0
@@ -41,9 +52,10 @@ def test_dist_options(pytester: pytest.Pytester) -> None:
 def test_auto_detect_cpus(
     pytester: pytest.Pytester, monkeypatch: pytest.MonkeyPatch
 ) -> None:
-    import os
     from xdist.plugin import pytest_cmdline_main as check_options
 
+    monkeypatch.delenv("PYTEST_XDIST_AUTO_NUM_WORKERS", raising=False)
+
     with suppress(ImportError):
         import psutil
 
@@ -91,6 +103,7 @@ def test_auto_detect_cpus_psutil(
 
     psutil = pytest.importorskip("psutil")
 
+    monkeypatch.delenv("PYTEST_XDIST_AUTO_NUM_WORKERS", raising=False)
     monkeypatch.setattr(psutil, "cpu_count", lambda logical=True: 84 if logical else 42)
 
     config = pytester.parseconfigure("-nauto")
@@ -102,6 +115,22 @@ def test_auto_detect_cpus_psutil(
     assert config.getoption("numprocesses") == 84
 
 
+def test_auto_detect_cpus_os(
+    pytester: pytest.Pytester, monkeypatch: pytest.MonkeyPatch, monkeypatch_3_cpus
+) -> None:
+    from xdist.plugin import pytest_cmdline_main as check_options
+
+    monkeypatch.delenv("PYTEST_XDIST_AUTO_NUM_WORKERS", raising=False)
+
+    config = pytester.parseconfigure("-nauto")
+    check_options(config)
+    assert config.getoption("numprocesses") == 3
+
+    config = pytester.parseconfigure("-nlogical")
+    check_options(config)
+    assert config.getoption("numprocesses") == 3
+
+
 def test_hook_auto_num_workers(
     pytester: pytest.Pytester, monkeypatch: pytest.MonkeyPatch
 ) -> None:
@@ -122,20 +151,105 @@ def test_hook_auto_num_workers(
     assert config.getoption("numprocesses") == 42
 
 
-def test_boxed_with_collect_only(pytester: pytest.Pytester) -> None:
+def test_hook_auto_num_workers_arg(
+    pytester: pytest.Pytester, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    # config.option.numprocesses is a pytest feature,
+    # but we document it so let's test it.
     from xdist.plugin import pytest_cmdline_main as check_options
 
-    config = pytester.parseconfigure("-n1", "--boxed")
+    pytester.makeconftest(
+        """
+        def pytest_xdist_auto_num_workers(config):
+            if config.option.numprocesses == 'auto':
+                return 42
+            if config.option.numprocesses == 'logical':
+                return 8
+    """
+    )
+    config = pytester.parseconfigure("-nauto")
     check_options(config)
-    assert config.option.forked
+    assert config.getoption("numprocesses") == 42
 
-    config = pytester.parseconfigure("-n1", "--collect-only")
+    config = pytester.parseconfigure("-nlogical")
     check_options(config)
-    assert not config.option.forked
+    assert config.getoption("numprocesses") == 8
 
-    config = pytester.parseconfigure("-n1", "--boxed", "--collect-only")
+
+def test_hook_auto_num_workers_none(
+    pytester: pytest.Pytester, monkeypatch: pytest.MonkeyPatch, monkeypatch_3_cpus
+) -> None:
+    # Returning None from a hook to skip it is pytest behavior,
+    # but we document it so let's test it.
+    from xdist.plugin import pytest_cmdline_main as check_options
+
+    monkeypatch.delenv("PYTEST_XDIST_AUTO_NUM_WORKERS", raising=False)
+
+    pytester.makeconftest(
+        """
+        def pytest_xdist_auto_num_workers():
+            return None
+    """
+    )
+    config = pytester.parseconfigure("-nauto")
     check_options(config)
-    assert config.option.forked
+    assert config.getoption("numprocesses") == 3
+
+    monkeypatch.setenv("PYTEST_XDIST_AUTO_NUM_WORKERS", "5")
+
+    config = pytester.parseconfigure("-nauto")
+    check_options(config)
+    assert config.getoption("numprocesses") == 5
+
+
+def test_envvar_auto_num_workers(
+    pytester: pytest.Pytester, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    from xdist.plugin import pytest_cmdline_main as check_options
+
+    monkeypatch.setenv("PYTEST_XDIST_AUTO_NUM_WORKERS", "7")
+
+    config = pytester.parseconfigure("-nauto")
+    check_options(config)
+    assert config.getoption("numprocesses") == 7
+
+    config = pytester.parseconfigure("-nlogical")
+    check_options(config)
+    assert config.getoption("numprocesses") == 7
+
+
+def test_envvar_auto_num_workers_warn(
+    pytester: pytest.Pytester, monkeypatch: pytest.MonkeyPatch, monkeypatch_3_cpus
+) -> None:
+    from xdist.plugin import pytest_cmdline_main as check_options
+
+    monkeypatch.setenv("PYTEST_XDIST_AUTO_NUM_WORKERS", "fourscore")
+
+    config = pytester.parseconfigure("-nauto")
+    with pytest.warns(UserWarning):
+        check_options(config)
+    assert config.getoption("numprocesses") == 3
+
+
+def test_auto_num_workers_hook_overrides_envvar(
+    pytester: pytest.Pytester, monkeypatch: pytest.MonkeyPatch, monkeypatch_3_cpus
+) -> None:
+    from xdist.plugin import pytest_cmdline_main as check_options
+
+    monkeypatch.setenv("PYTEST_XDIST_AUTO_NUM_WORKERS", "987")
+    pytester.makeconftest(
+        """
+        def pytest_xdist_auto_num_workers():
+            return 2
+    """
+    )
+    config = pytester.parseconfigure("-nauto")
+    check_options(config)
+    assert config.getoption("numprocesses") == 2
+
+    config = pytester.parseconfigure("-nauto")
+    check_options(config)
+    assert config.getoption("numprocesses") == 2
 
 
 def test_dsession_with_collect_only(pytester: pytest.Pytester) -> None:

testing/test_remote.py

@@ -1,5 +1,4 @@
 import pprint
-import py
 import pytest
 import sys
 import uuid
@@ -26,15 +25,14 @@ class EventCall:
         self.name, self.kwargs = eventcall
 
     def __str__(self):
-        return "<EventCall {}(**{})>".format(self.name, self.kwargs)
+        return f"<EventCall {self.name}(**{self.kwargs})>"
 
 
 class WorkerSetup:
-    use_callback = False
-
     def __init__(self, request, pytester: pytest.Pytester) -> None:
         self.request = request
         self.pytester = pytester
+        self.use_callback = False
         self.events = Queue()  # type: ignore[var-annotated]
 
     def setup(self) -> None:
@@ -42,7 +40,7 @@ class WorkerSetup:
         # import os ; os.environ['EXECNET_DEBUG'] = "2"
         self.gateway = execnet.makegateway()
         self.config = config = self.pytester.parseconfigure()
-        putevent = self.use_callback and self.events.put or None
+        putevent = self.events.put if self.use_callback else None
 
         class DummyMananger:
             testrunuid = uuid.uuid4().hex
@@ -61,7 +59,7 @@ class WorkerSetup:
             ev = EventCall(data)
             if name is None or ev.name == name:
                 return ev
-            print("skipping {}".format(ev))
+            print(f"skipping {ev}")
 
     def sendcommand(self, name, **kwargs):
         self.slp.sendcommand(name, **kwargs)
@@ -108,7 +106,7 @@ class TestWorkerInteractor:
         assert ev.name == "collectionstart"
         assert not ev.kwargs
         ev = worker.popevent("collectionfinish")
-        assert ev.kwargs["topdir"] == py.path.local(worker.pytester.path)
+        assert ev.kwargs["topdir"] == str(worker.pytester.path)
         ids = ev.kwargs["ids"]
         assert len(ids) == 1
         worker.sendcommand("runtests", indices=list(range(len(ids))))
@@ -222,6 +220,91 @@ class TestWorkerInteractor:
         ev = worker.popevent()
         assert ev.name == "errordown"
 
+    def test_steal_work(self, worker: WorkerSetup, unserialize_report) -> None:
+        worker.pytester.makepyfile(
+            """
+            import time
+            def test_func(): time.sleep(1)
+            def test_func2(): pass
+            def test_func3(): pass
+            def test_func4(): pass
+        """
+        )
+        worker.setup()
+        ev = worker.popevent("collectionfinish")
+        ids = ev.kwargs["ids"]
+        assert len(ids) == 4
+        worker.sendcommand("runtests_all")
+
+        # wait for test_func setup
+        ev = worker.popevent("testreport")
+        rep = unserialize_report(ev.kwargs["data"])
+        assert rep.nodeid.endswith("::test_func")
+        assert rep.when == "setup"
+
+        worker.sendcommand("steal", indices=[1, 2])
+        ev = worker.popevent("unscheduled")
+        assert ev.kwargs["indices"] == [2]
+
+        reports = [
+            ("test_func", "call"),
+            ("test_func", "teardown"),
+            ("test_func2", "setup"),
+            ("test_func2", "call"),
+            ("test_func2", "teardown"),
+        ]
+
+        for func, when in reports:
+            ev = worker.popevent("testreport")
+            rep = unserialize_report(ev.kwargs["data"])
+            assert rep.nodeid.endswith(f"::{func}")
+            assert rep.when == when
+
+        worker.sendcommand("shutdown")
+
+        for when in ["setup", "call", "teardown"]:
+            ev = worker.popevent("testreport")
+            rep = unserialize_report(ev.kwargs["data"])
+            assert rep.nodeid.endswith("::test_func4")
+            assert rep.when == when
+
+        ev = worker.popevent("workerfinished")
+        assert "workeroutput" in ev.kwargs
+
+    def test_steal_empty_queue(self, worker: WorkerSetup, unserialize_report) -> None:
+        worker.pytester.makepyfile(
+            """
+            def test_func(): pass
+            def test_func2(): pass
+        """
+        )
+        worker.setup()
+        ev = worker.popevent("collectionfinish")
+        ids = ev.kwargs["ids"]
+        assert len(ids) == 2
+        worker.sendcommand("runtests_all")
+
+        for when in ["setup", "call", "teardown"]:
+            ev = worker.popevent("testreport")
+            rep = unserialize_report(ev.kwargs["data"])
+            assert rep.nodeid.endswith("::test_func")
+            assert rep.when == when
+
+        worker.sendcommand("steal", indices=[0, 1])
+        ev = worker.popevent("unscheduled")
+        assert ev.kwargs["indices"] == []
+
+        worker.sendcommand("shutdown")
+
+        for when in ["setup", "call", "teardown"]:
+            ev = worker.popevent("testreport")
+            rep = unserialize_report(ev.kwargs["data"])
+            assert rep.nodeid.endswith("::test_func2")
+            assert rep.when == when
+
+        ev = worker.popevent("workerfinished")
+        assert "workeroutput" in ev.kwargs
+
 
 def test_remote_env_vars(pytester: pytest.Pytester) -> None:
     pytester.makepyfile(

testing/test_workermanage.py

@@ -1,11 +1,14 @@
 import execnet
-import py
 import pytest
 import shutil
 import textwrap
+import warnings
 from pathlib import Path
+from util import generate_warning
 from xdist import workermanage
-from xdist.workermanage import HostRSync, NodeManager
+from xdist._path import visit_path
+from xdist.remote import serialize_warning_message
+from xdist.workermanage import HostRSync, NodeManager, unserialize_warning_message
 
 pytest_plugins = "pytester"
 
@@ -154,12 +157,9 @@ class TestHRSync:
         source.joinpath("somedir").mkdir()
         source.joinpath("somedir", "editfile~").touch()
         syncer = HostRSync(source, ignores=NodeManager.DEFAULT_IGNORES)
-        files = list(py.path.local(source).visit(rec=syncer.filter, fil=syncer.filter))
-        assert len(files) == 3
-        basenames = [x.basename for x in files]
-        assert "dir" in basenames
-        assert "file.txt" in basenames
-        assert "somedir" in basenames
+        files = list(visit_path(source, recurse=syncer.filter, filter=syncer.filter))
+        names = {x.name for x in files}
+        assert names == {"dir", "file.txt", "somedir"}
 
     def test_hrsync_one_host(self, source: Path, dest: Path) -> None:
         gw = execnet.makegateway("popen//chdir=%s" % dest)
@@ -326,3 +326,80 @@ class TestNodeManager:
         )
         (rep,) = reprec.getreports("pytest_runtest_logreport")
         assert rep.passed
+
+
+class MyWarning(UserWarning):
+    pass
+
+
+@pytest.mark.parametrize(
+    "w_cls",
+    [
+        UserWarning,
+        MyWarning,
+        "Imported",
+        pytest.param(
+            "Nested",
+            marks=pytest.mark.xfail(reason="Nested warning classes are not supported."),
+        ),
+    ],
+)
+def test_unserialize_warning_msg(w_cls):
+    """Test that warning serialization process works well"""
+
+    # Create a test warning message
+    with pytest.warns(UserWarning) as w:
+        if not isinstance(w_cls, str):
+            warnings.warn("hello", w_cls)
+        elif w_cls == "Imported":
+            generate_warning()
+        elif w_cls == "Nested":
+            # dynamic creation
+            class MyWarning2(UserWarning):
+                pass
+
+            warnings.warn("hello", MyWarning2)
+
+    # Unpack
+    assert len(w) == 1
+    w_msg = w[0]
+
+    # Serialize and deserialize
+    data = serialize_warning_message(w_msg)
+    w_msg2 = unserialize_warning_message(data)
+
+    # Compare the two objects
+    all_keys = set(vars(w_msg).keys()).union(set(vars(w_msg2).keys()))
+    for k in all_keys:
+        v1 = getattr(w_msg, k)
+        v2 = getattr(w_msg2, k)
+        if k == "message":
+            assert type(v1) == type(v2)
+            assert v1.args == v2.args
+        else:
+            assert v1 == v2
+
+
+class MyWarningUnknown(UserWarning):
+    # Changing the __module__ attribute is only safe if class can be imported
+    # from there
+    __module__ = "unknown"
+
+
+def test_warning_serialization_tweaked_module():
+    """Test for GH#404"""
+
+    # Create a test warning message
+    with pytest.warns(UserWarning) as w:
+        warnings.warn("hello", MyWarningUnknown)
+
+    # Unpack
+    assert len(w) == 1
+    w_msg = w[0]
+
+    # Serialize and deserialize
+    data = serialize_warning_message(w_msg)
+
+    # __module__ cannot be found!
+    with pytest.raises(ModuleNotFoundError):
+        unserialize_warning_message(data)

testing/util.py

@@ -0,0 +1,9 @@
+import warnings
+
+
+class MyWarning2(UserWarning):
+    pass
+
+
+def generate_warning():
+    warnings.warn(MyWarning2("hello"))

tox.ini

@@ -1,11 +1,11 @@
 [tox]
 envlist=
   linting
-  py{36,37,38,39,310}-pytestlatest
-  py38-pytestmain
-  py38-psutil
-  py38-setproctitle
-
+  py{37,38,39,310,311}-pytestlatest
+  py310-pytestmain
+  py310-psutil
+  py310-setproctitle
+isolated_build = true
 [testenv]
 extras = testing
 deps =
@@ -14,14 +14,14 @@ deps =
 commands=
   pytest {posargs}
 
-[testenv:py38-psutil]
+[testenv:py310-psutil]
 extras =
   testing
   psutil
 commands =
   pytest {posargs:-k psutil}
 
-[testenv:py38-setproctitle]
+[testenv:py310-setproctitle]
 extras =
   testing
   setproctitle
@@ -39,15 +39,24 @@ commands = pre-commit run --all-files --show-diff-on-failure
 
 [testenv:release]
 changedir=
-decription = do a release, required posarg of the version number
-basepython = python3.7
+description = do a release, required posarg of the version number
+basepython = python3.10
 skipsdist = True
 usedevelop = True
 passenv = *
 deps =
   towncrier
 commands =
-  towncrier --version {posargs} --yes
+  towncrier build --version {posargs} --yes
+
+[testenv:docs]
+basepython = python3.10
+usedevelop = True
+deps =
+    sphinx
+    sphinx_rtd_theme
+commands =
+    sphinx-build -W --keep-going -b html docs docs/_build/html {posargs:}
 
 [pytest]
 # pytest-services also defines a worker_id fixture, disable