Hitting unreachable arm in on_ready.rs

[nocommentsgood]

When using OnReady I'm hitting a match arm with unreachable!(). This occurs when, in the example below, the TestUnreachable struct attempts to get a node in the scene tree using an incorrect path.

I'm still on version 0.3.5.

ERROR: [panic /home/me/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/godot-core-0.3.5/src/obj/on_ready.rs:270]
internal error: entered unreachable code
Context: IDKAnymore::process
at: godot_core::private::set_gdext_hook::{{closure}} (/home/me/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/godot-core-0.3.5/src/private.rs:310)

Reproduction:
Scene with root node of type SomeType. Add a child Timer node (I'm assuming this occurs with any node type).

struct TestUnreachable {
    timer: Gd<Timer>,
}

impl TestUnreachable {
    fn new(some_node: &Gd<Node>) -> Self {
        Self {
            timer: some_node.get_node_as::<Timer>("WrongName"),
        }
    }
}

#[derive(GodotClass)]
#[class(init, base = Node)]
struct SomeType {
    #[init(val = OnReady::from_base_fn(|base| TestUnreachable::new(base)))]
    test: OnReady<TestUnreachable>,
    base: Base<Node>,
}

#[godot_api]
impl INode for SomeType {
    fn process(&mut self, delta: f32) {
        dbg!(&self.test.timer);
    }
}

The first time the program attempts to read/write the value of the timer, it correctly panics with "There is no node of type Timer at path ...". Any subsequent attempts, such as trying to debug it in process() hit the unreachable() arm.

I understand that this should panic, as I am providing the incorrect name of the node, but I'm pretty sure an unreachable arm should not be reachable.

[TitanNano]

gdext/godot-core/src/obj/on_ready.rs

Lines 247 to 249 in 8f47818

	self.state = InitState::Initialized {
	value: initializer(base),
	};

When the initializer panics, the InitState::AutoInitializing remains, and the expected ephemeralness of AutoInitializing is violated.

InitState is not unwind safe, but we are treating it as such. The easiest solution would probably be to treat InitState as poisoned when in AutoInitializing and properly panic again.

[TitanNano]

@nocommentsgood on the latest master. This should now correctly cause a descriptive panic each time the field is accessed.