API changes tracking issue

[mischnic]

libui-node	libui-napi
~libui.Ui.onShouldQuit~	~libui.onShouldQuit~
~UiSlider::constructor() = constructor(0, 100)~	done
~the free methods~	~missing (tolerable?!)~
~the atomic bool in Windows event-loop~	not needed anymore

Missing

• ~Dialogs~
• ~Font~
• ~main/quit~

• ~uiTimer~ done (does setTimeout/Interval need patching? NO)

Otherwise, proton-native seems to work.

[parro-it]

Otherwise, proton-native seems to work.

🎉 we almost complete everything in two weeks first commit: https://github.com/parro-it/libui-napi/commit/f90ec051d992ca7da58fbabcf095676db87191d0

I'd never have said we can did it at such pace... we should open a start-up and conquer the world 😆

[parro-it]

the free methods

Honestly I don't see how they can be useful in a GC language... We can implements them as no-op just to avoid breaking something in the wild? But I doubt there's something in the wild using these...

[parro-it]

libui.Ui.onShouldQuit

Can we put that method on both places, like we did for enums?

[mischnic]

I'd never have said we can did it at such pace... we should open a start-up and conquer the world 😆

😆

But I doubt there's something in the wild using these...

👍

Can we put that method on both places, like we did for enums?

👍

[parro-it]

I think to avoid to implement quit and main method: they cause node asynchronous functions to stop working

[parro-it]

I think libui.Ui.[dialogs] was libui.UiDialogs, as it is now

[mischnic]

I think libui.Ui.[dialogs] was libui.UiDialogs, as it is now

You're right.

I think to avoid to implement quit and main method: they cause node asynchronous functions to stop working

So the only way to quit an app is to call stopLoop (is that equivalent) ?

[parro-it]

So the only way to quit an app is to call stopLoop ?

Yes, it internally calls quit(). Calling quit() after startLoop does not stop the app, because there is a libuv handler created just for the purpose of keeping the node event loop running... stopLoop close that handle before calling quit()

[mischnic]

I think to avoid to implement quit and main method: they cause node asynchronous functions to stop working

They were also never documented, so it's fine.

[mischnic]

Could the eventloop start/stop promise behaviour cause any difference?

[parro-it]

Could the eventloop start/stop promise behaviour cause any difference?

Yes, in the previous version, calling two startLoop sequentially, or two stopLoop sequentially, or stopLoop before startLoop is completed, doesn't throw an error.

But what really happened on such pattern greatly depends on the machine speed, current load etc. Probably a segfault or a thread deadlock or other weird things.

In the correct usage pattern, the new implementation should behave like the previous version, and the user is not required to await the promise, if he don't plan to start a loop immediately after the stop promises is settled (or viceversa)

So, I don't know if this could be considered a breaking change. It's basically a breaking change of behaviour in case of an error condition... I guess this could be fine