Add interface bounds to Impls

[felinira]

What this does

• Require everyone to specify the @implements directive for all interfaces that are implemented by the object, even interfaces implemented by the parent. (positive check)

What this doesn't do

• Check whether the interface is actually implemented (negative check), you can still write whatever in the wrapper macro and nothing checks this. This is a hard problem, and if at all only solveable at runtime. (not sure how, but it could be possible by doing some assertions on type_init maybe, needs more research)
• I think the error messages could be better, which would be possible by adding #[diagnostic::on_unimplemented] for the IsA trait (it is stable since 1.78).

See also gtk-rs/gtk-rs-core#1551

[felinira]

For consistency with other implementations. This derives from glib::Object, but implements gtk::PrintOperationPreview

[bilelmoussaoui]

Does a print operation necessarily requires a print operation preview?

[felinira]

yes https://docs.gtk.org/gtk4/class.PrintOperation.html

It is surprising to me as well that it's set up that way, but it is how it is.

class Gtk.PrintOperation : GObject.Object
  implements Gtk.PrintOperationPreview {
  /* No available fields */
}

[felinira]

forgot doc tests, they should hopefully pass now

[bilelmoussaoui]

otherwise lgtm

[bilelmoussaoui]

Almost there!

[felinira]

Almost there!

CI driven development! 🙄

I could have spent this time to make a toolbox with gtk nightly so that I can use --all-features locally, but alas I spent it fighting CI instead.

[bilelmoussaoui]

Actually, you don't even need that. There is a trick, to build by passing the same rustc flags we pass for the docs. It won't try to link against gtk ;) I forgot the details though, can find that again if you ever need it