Skip to content

Latest commit

 

History

History
798 lines (631 loc) · 29.8 KB

contributing.asciidoc

File metadata and controls

798 lines (631 loc) · 29.8 KB
Raw

Contributing to qutebrowser

Table of Contents
Important
 Bandwidth for pull request review is currently quite limited. If you want to contribute where it’s most needed, please consider reviewing or testing open pull requests.

I <3 [1] contributors!

This document contains guidelines for contributing to qutebrowser, as well as useful hints when doing so.

If anything mentioned here would prevent you from contributing, please let me know, and contribute anyways! The guidelines are meant to make life easier for me, but if you don’t follow everything in here, I won’t be mad at you. In fact, I will probably change it for you.

If you have any problems, I’m more than happy to help! You can get help in several ways:

Finding something to work on

Chances are you already know something to improve or add when you’re reading this. It might be a good idea to ask on the mailing list or IRC channel to make sure nobody else started working on the same thing already.

If you want to find something useful to do, check the issue tracker. Some pointers:

If you prefer C++ or Javascript to Python, see the relevant issues which involve work in those languages:

  • C++ (mostly work on Qt, the library behind qutebrowser)

  • JavaScript

There are also some things to do if you don’t want to write code:

  • Help the community, e.g., on the mailinglist and the IRC channel.

  • Improve the documentation.

  • Help on the website and graphics (logo, etc.).

Using git

qutebrowser uses git for its development. You can clone the repo like this:

git clone https://github.com/qutebrowser/qutebrowser.git

If you don’t know git, a git cheatsheet might come in handy. Of course, if using git is the issue which prevents you from contributing, feel free to send normal patches instead, e.g., generated via diff -Nur.

Getting patches

The preferred way of submitting changes is to fork the repository and to submit a pull request.

If you prefer to send a patch to the mailinglist, you can generate a patch based on your changes like this:

git format-patch origin/main (1)

  1. Replace main by the branch your work was based on, e.g., origin/develop.

Running qutebrowser

After installing qutebrowser in a virtualenv, you can run .venv/bin/qutebrowser --debug --temp-basedir to test your changes with debug logging enabled and without affecting existing running instances.

Alternatively, you can install qutebrowser’s dependencies system-wide and run python3 -m qutebrowser --debug --temp-basedir.

Useful utilities

Checkers

qutebrowser uses tox to run its unittests and several linters/checkers.

Currently, the following tox environments are available:

  • Tests using pytest:

    • py38, py39, …​: Run pytest for python 3.8/3.9/…​ with the system-wide PyQt.

    • py38-pyqt515, …​, py38-pyqt65: Run pytest with the given PyQt version (py39-* etc. also works).

    • py38-pyqt515-cov: Run with coverage support (other Python/PyQt versions work too).

  • flake8: Run various linting checks via flake8.

  • vulture: Run vulture to find unused code portions.

  • pylint: Run pylint static code analysis.

  • pyroma: Check packaging practices with pyroma.

  • eslint: Run ESLint javascript checker.

  • check-manifest: Check MANIFEST.in completeness with check-manifest.

  • mkvenv: Bootstrap a virtualenv for testing.

  • misc: Run scripts/misc_checks.py to check for:

    • untracked git files

    • VCS conflict markers

    • common spelling mistakes

  • mypy for static type checking:

    • mypy-pyqt5 run mypy with PyQt5 installed

    • mypy-pyqt6 run mypy with PyQt6 installed

The default test suite is run with tox; the list of default environments is obtained with tox -l.

Please make sure the checks run without any warnings on your new contributions.

There’s always the possibility of false positives; the following techniques are useful to handle these:

  • Use _foo for unused parameters, with foo being a descriptive name. Using _ is discouraged.

  • If you think you have a good reason to suppress a message, then add the following comment:

    # pylint: disable=message-name

    Note you can add this per line, per function/class, or per file. Please use the smallest scope which makes sense. Most of the time, this will be line scope.

  • If you really think a check shouldn’t be done globally as it yields a lot of false-positives, let me know! I’m still tweaking the parameters.

Running specific tests

While you are developing you often don’t want to run the full test suite each time.

Specific test environments can be run with tox -e <envlist>.

Additional parameters can be passed to the test scripts by separating them from tox arguments with --.

Examples:

# run only pytest tests which failed in last run:
tox -e py38 -- --lf

# run only the end2end feature tests:
tox -e py38 -- tests/end2end/features

# run everything with undo in the generated name, based on the scenario text
tox -e py38 -- tests/end2end/features/test_tabs_bdd.py -k undo

# run coverage test for specific file (updates htmlcov/index.html)
tox -e py38-cov -- tests/unit/browser/test_webelem.py

Specifying the backend for tests

Tests automatically pick the backend based on what they manage to import. If you have both backends available and you would like the tests to be run with a specific one you can set either of a) the environment variable QUTE_TESTS_BACKEND , or b) the command line argument --qute-backend, to the desired backend (webkit/webengine).

Profiling

In the scripts/dev/ subfolder there’s run_profile.py which profiles the code and shows a graphical representation of what takes how much time.

It uses the built-in Python cProfile module. It launches a qutebrowser instance, waits for it to exit and then shows the graph.

Available methods for visualization are:

You can also save the binary profile data to a file (--profile-tool=none).

Debugging

There are some useful functions for debugging in the qutebrowser.utils.debug module.

When starting qutebrowser with the --debug flag, you also get useful debug logs. You can add --logfilter [!]category[,category,…​] to restrict logging to the given categories.

With --debug there are also some additional debug-* commands available, for example :debug-all-objects and :debug-all-widgets which print a list of all Qt objects/widgets to the debug log — this is very useful for finding memory leaks.

Useful websites

Some resources which might be handy:

Documentation of used Python libraries:

Related RFCs and standards:

HTTP

Hints

Python and Qt objects

For many tasks, there are solutions available in both Qt and the Python standard library.

In qutebrowser, the policy is usually to use the Python libraries, as they provide exceptions and other benefits.

There are some exceptions to that:

  • QThread is used instead of Python threads because it provides signals and slots.

  • QProcess is used instead of Python’s subprocess.

  • QUrl is used instead of storing URLs as string, see the handling URLs section for details.

When using Qt objects, two issues must be taken care of:

  • Methods of Qt objects report their status with their return values, instead of using exceptions.

    If a function gets or returns a Qt object which has an .isValid() method such as QUrl or QModelIndex, there’s a helper function ensure_valid in qutebrowser.utils.qtutils which should get called on all such objects. It will raise qutebrowser.utils.qtutils.QtValueError if the value is not valid.

    If a function returns something else on error, the return value should carefully be checked.

  • Methods of Qt objects have certain maximum values based on their underlying C++ types.

    To avoid passing too large of a numeric parameter to a Qt function, all numbers should be range-checked using qutebrowser.qtutils.check_overflow, or by other means (e.g. by setting a maximum value for a config object).

The object registry

The object registry in qutebrowser.utils.objreg is a collection of dictionaries which map object names to the actual long-living objects.

There are currently these object registries, also called scopes:

  • The global scope, with objects which are used globally (config, cookie-jar, etc.).

  • The tab scope with objects which are per-tab (hintmanager, webview, etc.). Passing this scope to objreg.get() selects the object in the currently focused tab by default. A tab can be explicitly selected by passing tab=tab-id, window=win-id to it.

A new object can be registered by using objreg.register(name, object[, scope=scope, window=win-id, tab=tab-id]). An object should not be registered twice. To update it, update=True has to be given.

An object can be retrieved by using objreg.get(name[, scope=scope, window=win-id, tab=tab-id]). The default scope is global.

All objects can be printed by starting with the --debug flag and using the :debug-all-objects command.

The registry is mainly used for command handlers, but it can also be useful in places where using Qt’s signals and slots mechanism would be difficult.

Logging

Logging is used at various places throughout the qutebrowser code. If you add a new feature, you should also add some strategic debug logging.

Unlike other Python projects, qutebrowser doesn’t use a logger per file, instead it uses custom-named loggers.

The existing loggers are defined in qutebrowser.utils.log. If your feature doesn’t fit in any of the logging categories, simply add a new line like this:

foo = getLogger('foo')

Then in your source files, do this:

from qutebrowser.utils import log
...
log.foo.debug("Hello World")

The following logging levels are available for every logger:

critical

Critical issue, qutebrowser can’t continue to run.

error

There was an issue and some kind of operation was abandoned.

warning

There was an issue but the operation can continue running.

info

General informational messages.

debug

Verbose debugging information.

Commands

qutebrowser has the concept of functions which are exposed to the user as commands.

Creating a new command is straightforward:

from qutebrowser.api import cmdutils

...

@cmdutils.register(...)
def foo():
    ...

The commands arguments are automatically deduced by inspecting your function.

If the function is a method of a class, the @cmdutils.register decorator needs to have an instance=... parameter which points to the (single/main) instance of the class.

The instance parameter is the name of an object in the object registry, which then gets passed as the self parameter to the handler. The scope argument selects which object registry (global, per-tab, etc.) to use. See the object registry section for details.

There are also other arguments to customize the way the command is registered; see the class documentation for register in qutebrowser.api.cmdutils for details.

The types of the function arguments are inferred based on their default values, e.g., an argument foo=True will be converted to a flag -f/--foo in qutebrowser’s commandline.

The type can be overridden using Python’s function annotations:

@cmdutils.register(...)
def foo(bar: int, baz=True):
    ...

Possible values:

  • A callable (int, float, etc.): Gets called to validate/convert the value.

  • A python enum type: All members of the enum are possible values.

  • A typing.Union of multiple types above: Any of these types are valid values, e.g., typing.Union[str, int].

You can customize how an argument is handled using the @cmdutils.argument decorator after @cmdutils.register. This can, for example, be used to customize the flag an argument should get:

@cmdutils.register(...)
@cmdutils.argument('bar', flag='c')
def foo(bar):
    ...

For a str argument, you can restrict the allowed strings using choices:

@cmdutils.register(...)
@cmdutils.argument('bar', choices=['val1', 'val2'])
def foo(bar: str):
    ...

For typing.Union types, the given choices are only checked if other types (like int) don’t match.

The following arguments are supported for @cmdutils.argument:

  • flag: Customize the short flag (-x) the argument will get.

  • value: Tell qutebrowser to fill the argument with special values:

  • value=cmdutils.Value.count: The count given by the user to the command.

  • value=cmdutils.Value.win_id: The window ID of the current window.

  • value=cmdutils.Value.cur_tab: The tab object which is currently focused.

  • completion: A completion function (see qutebrowser.completions.models.*) to use when completing arguments for the given command.

  • choices: The allowed string choices for the argument.

The name of an argument will always be the parameter name, with any trailing underscores stripped and underscores replaced by dashes.

Handling URLs

qutebrowser handles two different types of URLs: URLs as a string, and URLs as the Qt QUrl type. As this can get confusing quickly, please follow the following guidelines:

  • Convert a string to a QUrl object as early as possible, i.e., directly after the user did enter it.

    • Use utils.urlutils.fuzzy_url if the URL is entered by the user somewhere.

    • Be sure you handle utils.urlutils.FuzzyError and display an error message to the user.

  • Convert a QUrl object to a string as late as possible, i.e., before displaying it to the user.

    • If you want to display the URL to the user, use url.toDisplayString() so password information is removed.

    • If you want to get the URL as string for some other reason, you most likely want to add the QUrl.EncodeFully and QUrl.RemovePassword flags.

  • Name a string URL something like urlstr, and a QUrl something like url.

  • Mention in the docstring whether your function needs a URL string or a QUrl.

  • Call ensure_valid from utils.qtutils whenever getting or creating a QUrl and take appropriate action if not. Note the URL of the current page always could be an invalid QUrl (if nothing is loaded yet).

Running valgrind on QtWebKit

If you want to run qutebrowser (and thus QtWebKit) with valgrind, you’ll need to pass --smc-check=all to it or recompile QtWebKit with the Javascript JIT disabled.

This is needed so valgrind handles self-modifying code correctly:

This option controls Valgrind’s detection of self-modifying code. If no checking is done and a program executes some code, overwrites it with new code, and then executes the new code, Valgrind will continue to execute the translations it made for the old code. This will likely lead to incorrect behavior and/or crashes.

…​

Note that the default option will catch the vast majority of cases. The main case it will not catch is programs such as JIT compilers that dynamically generate code and subsequently overwrite part or all of it. Running with all will slow Valgrind down noticeably.

Setting up a Windows Development Environment

  • Install Python 3.9.

  • Install PyQt via pip install PyQt5.

  • Install git from the git-scm downloads page. Try not to enable core.autocrlf, since that will cause flake8 to complain a lot. Use an editor that can deal with plain line feeds instead.

  • Clone your favourite qutebrowser repository.

  • To install tox, open an elevated cmd, enter your working directory and run pip install -rmisc/requirements/requirements-tox.txt.

Note that the flake8 tox env might not run due to encoding errors despite having LANG/LC_* set correctly.

Rebuilding the website

If you want to rebuild the website, run ./scripts/asciidoc2html.py --website <outputdir>.

Chrome URLs

With the QtWebEngine backend, qutebrowser supports several chrome:// urls which can be useful for debugging.

Info pages:

  • chrome://device-log/ (QtWebEngine >= 6.3)

  • chrome://gpu/

  • chrome://sandbox/ (Linux only)

Misc. / Debugging pages:

  • chrome://dino/

  • chrome://histograms/

  • chrome://network-errors/

  • chrome://tracing/ (QtWebEngine >= 5.15.3)

  • chrome://ukm/ (QtWebEngine >= 5.15.3)

  • chrome://user-actions/ (QtWebEngine >= 5.15.3)

  • chrome://webrtc-logs/ (QtWebEngine >= 5.15.3)

Internals pages:

  • chrome://accessibility/

  • chrome://appcache-internals/ (QtWebEngine < 6.4)

  • chrome://attribution-internals/ (QtWebEngine >= 6.4)

  • chrome://blob-internals/

  • chrome://conversion-internals/ (QtWebEngine >= 5.15.3 and < 6.4)

  • chrome://indexeddb-internals/

  • chrome://media-internals/

  • chrome://net-internals/ (QtWebEngine >= 5.15.4)

  • chrome://process-internals/

  • chrome://quota-internals/

  • chrome://serviceworker-internals/

  • chrome://webrtc-internals/

Crash/hang pages:

  • chrome://crash/ (crashes the current renderer process!)

  • chrome://gpuclean/ (crashes the current renderer process!)

  • chrome://gpucrash/ (crashes qutebrowser!)

  • chrome://gpuhang/ (hangs qutebrowser!)

  • chrome://kill/ (kills the current renderer process!)

QtWebEngine internals

This is mostly useful for qutebrowser maintainers to work around issues in Qt - if you don’t understand it, don’t worry, just ignore it.

The hierarchy of widgets when QtWebEngine is involved looks like this:

  • qutebrowser has a WebEngineTab object, which is its abstraction over QtWebKit/QtWebEngine.

  • The WebEngineTab has a _widget attribute, which is the QWebEngineView

  • That view has a QWebEnginePage for everything which doesn’t require rendering.

  • The view also has a layout with exactly one element (which also is its focusProxy()).

  • Qt 5: That element is the RenderWidgetHostViewQtDelegateWidget (it inherits QQuickWidget) - also often referred to as RWHV or RWHVQDW. It can be obtained via sip.cast(tab._widget.focusProxy(), QQuickWidget).

  • Qt 6: That element is the WebEngineQuickWidget (it inherits QQuickWidget). It can be obtained via tab._widget.focusProxy().

  • Calling rootObject() on that gives us the QQuickItem where Chromium renders into (?). With it, we can do things like .setRotation(20).

Style conventions

qutebrowser’s coding conventions are based on PEP8 and the Google Python style guidelines with some additions:

  • The Raise: section is not added to the docstring.

  • Methods overriding Qt methods (obviously!) don’t follow the naming schemes.

  • Everything else does though, even slots.

  • Docstrings should look like described in PEP257 and the google guidelines.

  • Class docstrings have additional Attributes:, Class attributes: and Signals: sections.

  • In docstrings of command handlers (registered via @cmdutils.register), the description should be split into two parts by using // - the first part is the description of the command like it will appear in the documentation, the second part is "internal" documentation only relevant to people reading the sourcecode.

    Example for a class docstring:

    """Some object.

Attributes:
    blub: The current thing to handle.

Signals:
    valueChanged: Emitted when a value changed.
                  arg: The new value
"""

    Example for a method/function docstring:

    """Do something special.

This will do something.

//

It is based on http://example.com/.

Args:
    foo: ...

Return:
    True if something, False if something else.
"""

  • The layout of a module should be roughly like this:

    • Shebang (#!/usr/bin/python, if needed)

    • Copyright

    • GPL boilerplate

    • Module docstring

    • Python standard library imports

    • PyQt imports

    • qutebrowser imports

    • functions

    • classes

  • The layout of a class should be like this:

    • docstring

    • __magic__ methods

    • other methods

    • overrides of Qt methods

  • Type hinting: the qutebrowser codebase uses type hints liberally to enable static type checking and autocompletion in editors.

    • We use mypy in CI jobs to perform static type checking.

    • Not all of the codebase is covered by type hints currently. We encourage including type hints on all new code and even adding type hints to existing code if you find yourself working on some that isn’t already covered. There are some module specific rules in the mypy config file, .mypy.ini, to make type hints strictly required in some areas.

    • More often than not mypy is correct when it raises issues. But don’t be afraid to add # type: ignore[...] statements or casts if you need to. As an optional part of the language not all type information from third parties is always correct. Mypy will raise a new issue if it spots an "ignore" statement which is no longer needed because the underlying issue has been resolved.

    • One area where we have to take particular care is in code that deals with differences between PyQt5 and PyQt6. We try to write most code in a way that will work with either backend but when you need to deal with differences you should use a pattern like:

      if machinery.IS_QT5:
    ... # do PyQt5 specific implementation
else:
    # PyQt6
    ... # do PyQt6 specific implementation

      then you have to tell mypy to treat machinery.IS_QT5 as a constant value then run mypy twice to cover both branches. There are a handful of variables in qutebrowser/qt/machinery.py that mypy needs to know about. There are tox jobs (mypy-pyqt5 and mypy-pyqt6) that take care of telling mypy to use them as constants.

Checklists

These are mainly intended for myself, but they also fit in here well.

New Qt release

  • Run all tests and check nothing is broken.

  • Check the Qt bugtracker and make sure all bugs marked as resolved are actually fixed.

  • Update own PKGBUILDs based on upstream Archlinux updates and rebuild.

  • Update recommended Qt version in README.

  • Grep for WORKAROUND in the code and test if fixed stuff works without the workaround.

  • Check relevant qutebrowser bugs and check if they’re fixed.

New PyQt release

  • See above.

  • Update tox.ini/.github/workflows/ci.yml to test new versions.

qutebrowser release

  • Make sure there are no unstaged changes and the tests are green.

  • Make sure all issues with the related milestone are closed.

  • Mark the milestone as closed.

  • Consider updating the completions for content.headers.user_agent in configdata.yml.

  • Minor release: Consider updating some files from main:

    • misc/requirements/ and requirements.txt

    • scripts/

  • Update changelog in main branch and ensure the correct version number has (unreleased)

  • If necessary: Update changelog in release branch from main.

Automatic release via GitHub Actions (starting with v3.0.0):

  • Double check Python version in .github/workflows/release.yml

  • Run the release workflow on the main branch, e.g. via gh workflow run release -f release_type=major (release_type can be major, minor or patch; you can also override python_version)

Manual release:

  • Make sure Python is up-to-date on build machines.

  • Run ./.venv/bin/python3 scripts/dev/update_version.py {major,minor,patch}.

  • Run the printed instructions accordingly.

Post release:

  • Update qutebrowser-git PKGBUILD if dependencies/install changed.

  • Add unreleased future versions to changelog

  • Update IRC topic

  • Announce to qutebrowser and qutebrowser-announce mailinglist.

  • Post announcement mail to subreddit

  • Post on the website formerly known as Twitter

1. <3 in HTML