Traceback (most recent call last):
File "...\electrum\electrum\gui\qt\main_window.py", line 1797, in toggle_search
self.search_box.setFocus(1)
TypeError: arguments did not match any overloaded call:
setFocus(self): too many arguments
setFocus(self, reason: Qt.FocusReason): argument 1 has unexpected type 'int'
The wayland plugin would require at least debian 12 (or ubuntu 22.04) at runtime.
see https://github.com/spesmilo/electrum/pull/9189#pullrequestreview-2309983943 :
> I've now tried running the appimage on debian 10 (oldoldstable), and am getting an error with wayland.
>
> ```
> 4.16 | D | util.profiler | Plugins.__init__ 0.0422 sec
> 4.16 | I | daemon.Daemon | launching GUI: qt
> 4.76 | I | gui.qt.ElectrumGui | Qt GUI starting up... Qt=6.7.1, PyQt=6.7.1
> /tmp/.mount_electrFlGFOt/usr/bin/python3: symbol lookup error: /tmp/.mount_electrFlGFOt/usr/lib/python3.11/site-packages/PyQt6/Qt6/plugins/platforms/../../lib/libQt6WaylandClient.so.6: undefined symbol: wl_proxy_marshal_flags
> ```
>
> If I explicitly specify `QT_QPA_PLATFORM=xcb`, it starts and works as expected. But it picks wayland by default.
> I found https://bugreports.qt.io/browse/QTBUG-114635 and it looks like even debian 11 might be affected.
Some checkboxes, e.g. main_window.warn_if_testnet became buggy with pyqt6:
looks like the stateChanged signal passes an int, not a Qt.CheckState.
(and note that Qt.CheckState is an Enum, not an IntEnum).
So `x == Qt.CheckState.Checked` would always evaluate to False.
```
def on_cb(_x):
print(f"heyheyhey. {_x=!r}, {Qt.CheckState.Checked=!r}, {cb.checkState()=!r}, {cb.isChecked()=!r}")
cb = QCheckBox("")
cb.stateChanged.connect(on_cb)
```
heyheyhey. x=2, Qt.CheckState.Checked=<CheckState.Checked: 2>, cb.checkState()=<CheckState.Checked: 2>, cb.isChecked()=True
heyheyhey. x=0, Qt.CheckState.Checked=<CheckState.Checked: 2>, cb.checkState()=<CheckState.Unchecked: 0>, cb.isChecked()=False
macOS reserves the "About" menu item name, similarly to "Preferences" (see a few lines above).
The "About" keyword seems even more strictly locked down:
not allowed as either a prefix or a suffix.
- With Qt5, a matching menu item is simply auto-recognised as the special "About" item,
- but with Qt6, it seems we explicitly have to do this dance, as directly adding
a menu item with the "About" name results in a segfault...
I noticed that when running from source on macOS, the OS menubar has an "About Python" menu item,
instead of an "About Electrum" menu item. I tried to fix that by this, but actually it is not working :P
Nevertheless, this looks useful at least on Linux and Windows. E.g. when instantiating a new dialog
without an explicit title, the qt application name is used as default.
The application name, without this change AFAICS was already "Electrum" when running any of the binaries.
However when running from source, it was in some cases "python" or "run_electrum" or even "electrum-4"
(depending on OS and how the main script is started).
Now it is consistent -- except on macOS it still is not, as there it really wants to look for a .plist...
In some cases this makes it much easier to successfully scan a QR code.
I was trying to scan a PSBT using a laptop camera from my phone screen for 2 minutes, until I realised the screen brightness was the issue. o.O
- between the Back btn and the Next btn, the latter should have priority for focus
- but if the user needs to enter the wallet password, that textedit should have focus
closes https://github.com/spesmilo/electrum/issues/9157
related c6802adbae
The history tab would show an incorrect feerate for partial/unsigned (local) txs,
if they had any p2sh/p2wsh txins. We would just guess the script is p2wpkh, and
use that for the size calc. Now with calling add_info_from_wallet, the correct
size is used to calculate the feerate.
(The gui tx dialogs call add_info_from_wallet independently, so the size/feerate
shown there were already correct.)
The messages are sometimes logged and sometimes shown to the user,
- for logging we might not want to truncate or have higher limits,
- but when shown to the user, we definitely want to truncate the error text.
It is simplest to just do the truncation here, at the lowest level.
Note that we usually prepend the error text with a header e.g. "[DO NOT TRUST THIS MESSAGE]"
and if the error text is too long, this header at the beginning might get "lost" in some way.
Hence we should truncate the error text.
SomberNight