pytest-jobserver 0.2.2

Limit parallel tests with posix jobserver.

pytest-jobserver

Limit parallel tests with posix jobserver.

Installation

Install with pip:

pip install pytest-jobserver

The plugin officially supports Python >= 3.6 on a Linux OS. You may find other Python 3 versions/MacOS work as well.

Usage

The plugin uses a POSIX jobserver to manage parallel test loading.

Pass the --jobserver argument with a path, where this path points to a named pipe acting as a jobserver. You should probably also use pytest-xdist to enable parallelism in the first place:

pytest -n4 --jobserver /opt/jobserver

The plugin can also listen for an existing jobserver as created by another tool. Currently the environment variables it supports are:

• CARGO_MAKEFLAGS
• MAKEFLAGS
• MFLAGS

The environment variables will only be checked when the cli flag is not set.

Please note: it is not supported to use both make environment variables and use xdist. The execnet protocol used by xdist does not support passing file descriptors to remote child processes.

Implementation

This plugin wraps pytest's call to pytest_runtest_protocol, which is called to actually run a collected item on the worker node.

Development

See the top level Makefile and .circleci/config.yml for the development flow. But in essence:

• make dev to install deps
• make lint to lint code (will change files)
• make test to run lint/unit tests
• make integrate to run integration tests

Changelog

0.2.2

Bugfixes

• move jobserver status on startup to official pytest_report_header hook

0.2.1

Bugfixes

• remove pytest-xdist from runtime dependencies

0.2.0

Features

• can run against a jobserver specified with make environment variables
  • note that this will deliberately not work when also using pytest-xdist

0.1.1

Bugfixes

• fix setup.py package metadata

0.1.0

Features

• can run against a jobserver specified at the cli

TODO

• run against cli jobserver
• integrate with Make jobserver from environment variables
• factor out creation of jobservers to seperate python package