Age | Commit message (Collapse) | Author |
|
|
|
|
|
Before this change, we displayed two different error messages, one in
_on_error, one in _on_finished. See the screenshot here:
https://www.reddit.com/r/qutebrowser/comments/kwvaw8/setting_vim_as_editor_creates_error_message/
|
|
They will always be present (either stdlib or backport), don't have a
version number, and we can't tell the two variants apart (other than
looking at the Python version) - so there's zero value in having them
there.
|
|
It doesn't actually need to be reusable, also make sure it's cleaned up
when unneeded.
|
|
|
|
Mostly reverts bdab7b35b6b3b7cd807f975d3449ac0d1fc68e55 as we didn't end
up using it in b2fcc270ec5427df1b77ed28c459d295d77dd28f.
|
|
It doesn't really matter if we run a DELETE query on an empty table vs.
running a query to find out whether the table is empty.
|
|
|
|
# Conflicts:
# qutebrowser/app.py
|
|
This reverts commit 32229f8b07a6773bdafff04ce2db582f9eefaf5e.
Seems to act differently on Windows.
|
|
Makes it clearer what the intent is (rather than implicitly falling back
on __len__ via Python) and also happens to be a small (most probably
insignificant) performance improvement.
On my machine, without a __bool__:
Name (time in us) Min Max Median
test_bool_benchmark 144.2940 879.5960 167.2730
With it:
Name (time in us) Min Max Median
test_bool_benchmark 133.3990 876.1080 152.1879
|
|
See https://doc.qt.io/qt-5/qsqlquery.html#numRowsAffected
|
|
|
|
Before this change, we displayed two different error messages, one in
_on_error, one in _on_finished. See the screenshot here:
https://www.reddit.com/r/qutebrowser/comments/kwvaw8/setting_vim_as_editor_creates_error_message/
|
|
|
|
See #6023
|
|
Let's just use "no error message" as "module is available" instead of
saving the same information in a bool and in an optional string.
See #6023
|
|
While we use dataclasses for its base class (interceptors.Request), it
causes more problems than it solves here.
See #6023
|
|
See #6023
|
|
See #6023
|
|
See #6023
|
|
For some reason, some issues were only uncovered by mypy after switching
from attrs to dataclasses.
See #6023
|
|
See #6023
|
|
See #6023
|
|
See #6023
|
|
See #6023
|
|
See #6023
|
|
Update dependencies
|
|
We know that QApplication.instance() will always be non-None for
practical purposes, but the stubs now (correctly) declare it as
Optional.
See https://github.com/stlehmann/PyQt5-stubs/pull/126
|
|
|
|
|
|
See #5551
|
|
|
|
Closes #5555
|
|
Fix brew installation instructions
|
|
|
|
|
|
It has some subtle typing differences compared to 3.6.1, see e.g.
https://docs.python.org/3/library/typing.html#typing.NamedTuple
Also, we can't easily test it, it's unlikely to be in use anywhere
anyways, and various dependencies already dropped support for it or are
going to (e.g. PyQt 6).
See #4800
|
|
See #5395
|
|
See https://github.com/actions/virtual-environments/issues/841
|
|
|
|
See #6010
This reverts commit 0810b3e7b99316861bf9e34c694136c30e1a2266.
|
|
|
|
|
|
|
|
See #6007
|
|
Instead of using an empty table as a trigger to rebuild, use a simple
flag. If the user wants to force a rebuild, they can set the user
version to 0 instead.
|
|
This avoids having to do various unneeded SQL queries on every start,
thus increasing startup time. While all those queries are idempotent,
doing them on every startup is still kind of expensive and might involve
disk I/O.
|
|
|