WIP: Experiments with scalable canvas

[Debilski]

Workable solution for #820. Still needs adjustments. Unclear, if the tk internal scaling for widgets(?) self.root.tk.call('tk', 'scaling', xxx) will be needed.

How does it work: We can use Tk to measure the pixel width of a text with a given font size. The result will be in raw pixels (on Linux). On macOS the number is resolution independent (which is good: on macOS it already works as it should). I could not test with Windows yet.

This number will then scale the geometry of the Window and then our own scaling mechanics will resize the canvas to make everything fit.

Screen is set to scale to 225%. Without the fix:

With the fix:

Note that the font and hard-coded numbers for the positioning of the header are currently not scaled, which is why they do look off. Work also needs to be done for the debug mode whose 1px grid lines look very tiny on higher scaling factors.

[Debilski]

Useful addition might be to add a few preset scaling factors (on the number keys?) that can be quickly selected.

[Debilski]

Very weird things going on here. Directly after a reboot, the measuring does not work. It only measures correctly the second time that I run pelita. Do you happen to have any idea @otizonaizit ?

Screencast.From.2025-08-18.16-31-18.mp4

[Debilski]

gitk is having the same problem (works fine on second open). If gitk is opened before Pelita, then all is good.

[Debilski]

Traced it down to being related to Wayland/Xwayland I think. If I run something like xclock before Pelita, all is good …