Add confirmation dialog before installing or uninstalling

[bradrn]

Fixes #6.

Some nasty subtleties here… in particular, digging through the gi-gtk source reveals that the ?self implicit parameter gets disowned as soon as the callback returns. Thus, in the message box respond signal, properties must be set on the original Gtk.Switch object, rather than on the ?self pointer of the outer callback (even though they should in theory be the same), since that pointer gets disowned before the message box signal can run.

Another point is that setting the activity programmatically also triggers the stateSet signal. Blocking this is more difficult than it should be, since it requires the handler ID, which can’t be accessed until after the message box has been created! It’s easier to just wrap the whole mockInstall function in a conditional, to ensure that the dialog isn’t shown twice.

[Kleidukos]

@bradrn this is interesting. Definitely would be worth it to write that down somewhere to help guide others.

[bradrn]

Yes, I definitely think it should be documented somewhere… but I’m not at all sure where. Maybe I’ll just open an issue on https://github.com/haskell-gi/haskell-gi.

(Also, do I need to worry about the failed CI?)

[Kleidukos]

(Also, do I need to worry about the failed CI?)

This is where a nix setup would be useful… The runners are still running on an Ubuntu LTS that has an old version of libadwaita, if I'm not mistaken

[bradrn]

Further on ?self… thanks to @garetxe [ haskell-gi/haskell-gi#408 (comment) ], this issue is now fixed in haskell-gi-0.26.5. The freeze file pins us to haskell-gi-0.26.4, though, so probably it’s best to continue avoiding ?self for now.

[Kleidukos]

@bradrn you can update the freeze file, it's better if we bump the version 👍

Regarding the CI I'll merge the PR despite of the runner.