Perhaps the easiest way to contribute to Qtile is to report any bugs you run into on the GitHub issue tracker.
Useful bug reports are ones that get bugs fixed. A useful bug report normally has two qualities:
Reproducible. If your bug is not reproducible it will never get fixed. You should clearly mention the steps to reproduce the bug. Do not assume or skip any reproducing step. Describe the issue, step-by-step, so that it is easy to reproduce and fix.
Specific. Do not write an essay about the problem. Be specific and to the point. Try to summarize the problem in a succinct manner. Do not combine multiple problems even if they seem to be similar. Write different reports for each problem.
Ensure to include any appropriate log entries from
xtrace is requested. If that is the case, refer to
capturing an xtrace.
Use a separate git branch to make rebasing easy. Ideally, you would
git checkout -b <my_feature_branch_name> before starting your work.
See also: using git.
Submit a pull request¶
Pull requests are not considered complete until they include all of the following:
Code that conforms to PEP8 and is formatted by black.
Unit tests that pass locally and in our CI environment (More below). Please add unit tests to ensure that your code works and stays working!
Documentation updates on an as needed basis.
Code that does not include unrelated changes. Examples for this are formatting changes, replacing quotes or whitespace in other parts of the code or "fixing" linter warnings popping up in your editor on existing code. Do not include anything like the above!
Widgets don't need to catch their own exceptions, or introduce their own polling infrastructure. The code in
libqtile.widget.base.*does all of this. Your widget should generally only include whatever parsing/rendering code is necessary, any other changes should go at the framework level. Make sure to double-check that you are not re-implementing parts of
Commit messages are more important that Github PR notes, since this is what people see when they are spelunking via
git blame. Please include all relevant detail in the actual git commit message (things like exact stack traces, copy/pastes of discussion in IRC/mailing lists, links to specifications or other API docs are all good). If your PR fixes a Github issue, it might also be wise to link to it with
#1234in the commit message.
PRs with multiple commits should not introduce code in one patch to then change it in a later patch. Please do a patch-by-patch review of your PR, and make sure each commit passes CI and makes logical sense on its own. In other words: do introduce your feature in one commit and maybe add the tests and documentation in a seperate commit. Don't push commits that partially implement a feature and are basically broken.
Others might ban force-pushes, we allow them and prefer them over incomplete commits or commits that have a bad and meaningless commit description.
Feel free to add your contribution (no matter how small) to the appropriate place in the CHANGELOG as well!
We must test each unit of code to ensure that new changes to the code do not break existing functionality. The framework we use to test Qtile is pytest. How pytest works is outside of the scope of this documentation, but there are tutorials online that explain how it is used.
Our tests are written inside the
test folder at the top level of the
repository. Reading through these, you can get a feel for the approach we take
to test a given unit. Most of the tests involve an object called
This is the test manager (defined in test/helpers.py), which exposes a command
manager.c that we use to test a Qtile instance running in a
separate thread as if we were using a command client from within a running
For any Qtile-specific question on testing, feel free to ask on our issue tracker or on IRC (#qtile on irc.oftc.net).
Running tests locally¶
This section gives an overview about
tox so that you don't have to search
its documentation just to get
Checks are grouped in so-called
environments. Some of them are configured to
check that the code works (the usual unit test, e.g.
others make sure that your code conforms to the style guide (
mypy). A third kind of test verifies that the documentation
and packaging processes work (
- The following examples show how to run tests locally:
To run the functional tests, use
tox -e py39(or a different environment). You can specify to only run a specific test file or even a specific test within that file with the following commands:
tox -e py39 # Run all tests with python 3.9 as the interpreter tox -e py39 -- -x test/widgets/test_widgetbox.py # run a single file tox -e py39 -- -x test/widgets/test_widgetbox.py::test_widgetbox_widget
To run style and building checks, use
tox -e docs,packaging,pep8,.... You can use
-p autoto run the environments in parallel.
The CI is configured to run all the environments. Hence it can be time- consuming to make all the tests pass. As stated above, pull requests that don't pass the tests are considered incomplete. Don't forget that this does not only include the functionality, but the style, typing annotations (if necessary) and documentation as well!