Open Hi-Angel opened 7 months ago
That's so weird to see the commits being dated by 2022, whereas I'm certain it is not that old, I would say I only noticed it maybe a few months ago or so…
But FTR, I didn't botch the bisection: I repeated it (this time even in a different way: by checking out good/bad commits along the main
branch whereas the post was done by using stable branches as good/bad points), and I came to the exact same commits list.
Works fine for me. The bisect is misleading, those commits just reflect you switching from Qt 5 to Qt 6. If you run with --qt-wrapper pyqt5
it will probably work fine on main
too.
This might just be a duplicate of #8127 I think, and the breakage was probably introduced by you upgrading to Qt 6.6.2 or maybe Qt 6.7?
The bisect is misleading, those commits just reflect you switching from Qt 5 to Qt 6. If you run with
--qt-wrapper pyqt5
it will probably work fine onmain
too.
Oh, I confirm, running with a --qt-wrapper PyQt5
fixes the problem!
This might just be a duplicate of #8127 I think
Well, from the description I am not so sure… It seems to be about the default behavior of a keybinding being changed. But you know the code better, so if you think the underlying reason is the same, maybe so.
and the breakage was probably introduced by you upgrading to Qt 6.6.2 or maybe Qt 6.7?
Hmm, it is possible. I'm kind of afraid of downgrading it, something might break… but it does sound plausible, given you've recently fixed another bug that was related to Qt regression as well…
Works fine for me.
What Qt version do you have?
The problem is fixed after upgrade. Not sure what fixed (QtWebEngine update or the Qutebrowser), so here's the snippet from my current :version
:
qutebrowser v3.2.0
Git commit:
Backend: QtWebEngine 6.7.1
based on Chromium 118.0.5993.220
with security patches up to 124.0.6367.202 (plus any distribution patches)
(source: api)
Qt: 6.7.1
Not sure either! Nothing changed in qutebrowser, and I can't find anything in the Qt 6.7.1 changelog that sounds related. Given that I can't reproduce, maybe the culprit was somewhere else (a theme?) and fixed.
Actually, I was wrong, the issue still happens but something apparently influences it.
When I closed it I had to use older laptop for some time and I had Wayland KDE there. Now I'm back to the work laptop and after updating the system the problem still there. So it may be due to Wayland difference or something else, I'm not sure…
I'm getting a very similar issue on Wayland, where every shortcut that uses the alt
key does something else.
alt+m
switches to tab 1
alt+y
(in the download prompt), switches to tab 26 (I can't explain this one)
alt+1
, alt+2
, alt+3
switches to tab 10, 20 and 30, but alt+[4-9]
switches to the correct tab.
I haven't changed the bindings for these shortcuts.
Running with --qt-wrapper PyQt5
doesn't work for me as I'm on NixOS, so only the packages applied to qutebrowser during the package build are available.
Version info:
qutebrowser v3.2.1
Git commit:
Backend: QtWebEngine 6.7.2
based on Chromium 118.0.5993.220
with security patches up to 125.0.6422.142 (plus any distribution patches)
(source: api)
Qt: 6.7.2
CPython: 3.12.5
PyQt: 6.7.0.dev2404081550
Qt wrapper info:
PyQt6: success
PyQt5: not imported
-> selected: PyQt6 (via autoselect)
EDIT: Changing the qutebrowser package to use Qt5, does indeed make the shortcuts work again. Version info:
qutebrowser v3.2.1
Git commit:
Backend: QtWebEngine 5.15.17
based on Chromium 87.0.4280.144
with security patches up to 123.0.6312.58 (plus any distribution patches)
(source: UA)
Qt: 5.15.14
CPython: 3.12.5
PyQt: 5.15.10
Qt wrapper info:
PyQt6: ModuleNotFoundError: No module named 'PyQt6'
PyQt5: success
-> selected: PyQt5 (via autoselect)
EDIT 2: Seeing as these are Nix packages, you should be able to get a reproducible build by using nixpkgs-unstable
(revision 574d1eac1c200690e27b8eb4e24887f8df7ac27c
) for Qt6, and the Qt5 build with this setup:
qutebrowser.override {
qtbase = pkgs.qt5.qtbase;
qtwayland = pkgs.qt5.qtwayland;
qtwebengine = pkgs.qt5.qtwebengine;
};
Version info:
Does the bug happen if you start with
--temp-basedir
?: yesDescription
Some time earlier I noticed that I no longer can switch back and forth between two last tabs by pressing Alt with a number of either of them.
I just started digging into it and bisected this down to a number of commits.
Unfortunately these commits apparently haven't been tested, because each one of them either fails or crashes the browser whatsoever. So the code is not bisectable any further than that:
How to reproduce
./qutebrowser.py --temp-basedir
Expected
The browser switches between 1-st and 2-nd tab
Actual
The browser switches to 2nd tab just once, and nothing happens anymore.