| Commit message (Collapse) | Author | Age |
|
|
|
|
| |
With the move to pytest remove remainings documentation or comments that
refer to Trial.
|
|
|
|
|
|
|
|
|
|
|
| |
Notable changes:
* Prettier >=2.3 with more consistent js assignments
* Black now formats docstrings
* Added isort to list of autoformaters
* Update flake8 config for v4
Ref: https://prettier.io/blog/2021/05/09/2.3.0.html
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Convert all the twisted.trial tests to pytest_twisted. Also move off of unittest.TestCase as well. Seems there were several tests which weren't actually testing what they should, and also some code that wasn't doing what the broken test said it should.
Goals:
Remove twisted.trial tests
Move to pytest fixtures, rather than many classess and subclasses with setup and teardown functions
Move away from self.assertX to assert style tests
FIx broken tests
Going forward I think these should be the goals when adding/modifying tests:
* Don't use BaseTest or set_up tear_down methods any more. Fixtures should be used either in the test module/class, or make/improve the ones available in conftest.py
* For sure don't use unittest or twisted.trial, they mess up the pytest stuff.
* Prefer pytest_twisted.ensureDeferred with an async function over inlineCallbacks.
- I think the async function syntax is nicer, and it helps catch silly mistakes, e.g. await None is invalid, but yield None isn't, so if some function returns an unexpected thing we try to await on, it will be caught earlier. (I struggled debugging a test for quite a while, then caught it immediately when switching to the new syntax)
- Once the maybe_coroutine PR goes in, using the async syntax can also improve tracebacks when debugging tests.
Things that should probably be cleaned up going forward:
* Remove BaseTestCase
* Remove the subclasses like DaemonBase in favor of new fixtures.
* I think there are some other utility subclasses that could be removed too
* Perhaps use parameterization in the ui_entry tests, rather that the weird combination of subclasses and the set_var fixture I mixed in.
* Convert some of the callback stuff to pytest_twisted.ensureDeferred tests, just for nicer readability
Details relating to pytest fixtures conftest.py in root dir:
* https://github.com/pytest-dev/pytest/issues/5822#issuecomment-697331920
* https://docs.pytest.org/en/latest/deprecations.html#pytest-plugins-in-non-top-level-conftest-files
Closes: https://github.com/deluge-torrent/deluge/pull/354
|
|
|
|
|
|
|
|
| |
Added pyupgrade utility with manual stage to pre-commit and run on all
files.
Ref: https://github.com/asottile/pyupgrade
Closes: deluge-torrent/deluge#326
|
|
|
|
|
|
|
|
|
| |
* Removed all __future__ imports from code
* Removed all six dependencies
* Removed all future_builtins imports
* Removed all Python 2 related code
Closes: deluge-torrent/deluge#325
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fixed the Stats GTKUI test not updated to the new GTK3 dir layout.
Fixed pygobject cairo ImageSurface error:
AttributeError: 'gi.repository.cairo' object has no attribute 'ImageSurface'
The documentation seems to suggest that using `import cairo` is the correct
usage and this fixed the issue, along with adding suggested gi.require_foreign
call.
References:
https://pygobject.readthedocs.io/en/latest/guide/cairo_integration.html
|
|
|
|
|
|
|
| |
Fixed logs flooded with:
[WARNING ][deluge.core.core :655 ] Session status key not valid: num_connections
[WARNING ][deluge.core.core :655 ] Session status key not valid: dht_cache_nodes
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Fixed black hook requiring Py3.6 to installed locally. Will now assume
Py3.6+ in installed.
* Added isort traceback in pre-commit flake8 hook fails
* Updated versions of Black, Prettier and isort
* Keep Flake8 at 3.7.9 due to E402 issue: https://gitlab.com/pycqa/flake8/-/issues/638
* New pyproject config for isort v5 with fixes for Python 2 imports.
* Fixed travis config to run Python 3.6 for lint run. Replaced the
virtualenv with_system_site_packages config with Travis specific Python
config value so lint run doesn't attempt to append with_system_site_packages
to Python 3.6 command.
|
|
This commit reverts namespace for the plugins and uses a module prefix
"deluge_" in it's place. The distribution package name remains the same
for now but will also be considered to use a prefix to help find the
third-party plugins e.g. Deluge-{Plugin} and the pluginmanager will
strip the prefix for displaying.
The change is a result of problems trying to package Deluge with
pyinstaller and the pkg_resources namespaces is not compatible.
Testing alternatives to using the pkgutil or PEP420 (native) namespaces
did not yield any joy either as importing eggs with namespaces does not
work. [1]
At this point importable eggs are considered deprecated but there is no
viable alternative yet. [2]
[1] https://github.com/pypa/packaging-problems/issues/212
[2] https://github.com/pypa/packaging-problems/issues/244
|