[f-o-o-g-s]

3e6be23's asleep() calls get_power() on the device find_device() hands back, which isn't open yet. On a wireless mouse that raises AttributeError: 'NoneType' object has no attribute 'ctrl_transfer' whether it's awake or not, so the script refuses every time. It now opens and closes the device around the read, the same way the window does for each of its own reads.

Checked on the X3 on the 1K dongle both ways:

• awake: 8 of 8, and no warning from the end-of-run check
• asleep (50 s untouched): stopped before the first shot with the driver's "no reply from the mouse ... within 1000 ms", and wrote nothing

Small note on the docstring: on the 1K dongle a sleeping mouse gives that timeout, not a 0xFF reply. It still raises, so the check works either way.

The second commit retakes the set on main with the script as it ships plus this fix, no env tricks. Five pages change: the Customize icon in the sidebar (it was the missing-image box in all of them) and the title bar, now just the close button. The three dialogs came out pixel-identical, so I left them alone. Readings are real again: 1000 Hz, 1600 DPI, 96% signal, battery 85% this time.

[packerlschupfer]

Merged, both commits.

Good catch, and a fair one — find_device() handing back a closed device only shows on hardware I don't have. My X2A takes the hasattr(device, 'get_power') exit, so the entire path shipped untested here. A day between writing that check and you finding it broken is a good trade.

One follow-up on top (f110422): an open() that fails now gets its own message. "Move the mouse to wake it, or pass --ignore-sleep" is the wrong advice when the real answer is the udev rules, and --ignore-sleep would then quietly take a full set of empty screenshots — so the unreachable case stops the run whatever the flags say. Your docstring note went in with it: through the dongle it times out rather than answering 0xFF, and either way the driver raises, which is all the check needs.

The retake was the right call too — the script as it ships, no env tricks, is exactly what someone regenerating from scratch should get.

And stepping back: thank you for this whole run. The 1K dongle's battery, charging, link quality and paired-mouse identity; hidraw matched on PID so a dongle can't be mistaken for the mouse next to it; the listener regression I introduced, found and fixed with a CI check that would have caught me; the DPI stage numbering settled on hardware nobody here owns; and the first screenshot set that actually shows the wireless pages. That is a lot of this project's recent surface, and all of it came with the hardware evidence attached. Much appreciated.