godotengine
TabContainer not set up properly when its children are ready
Tested versions
v4.3.stable.official [77dcf97d8] Works fine in v4.2.2 (and before)
System information
Windows 10
Issue description
When instantiating TabContainer children via code the TabContainer is not set up properly when one of the children is ready. Compare these two outputs:
v4.2.2
v4.3
This forces users to
call_deferred any logic that relies on these properties, which is pretty inconvenient.
Steps to reproduce
- Instantiate a TabContainer
- Instantiate a Control node as a child of the TabContainer
- Connect the
readysignal of the child - Check state of the TabContainer within the connected function
Minimal reproduction project (MRP)
TabContainerCurrentTab.zip This is a 4.2.2 project which can easily be converted to 4.3
This is just because the default changed in 4.3, from 0 to -1.
The child needs to become ready before the TabContainer can handle it, this is true in 4.2 as well. The child's ready notification is not a good place to use its parent TabContainer's methods. (get_current_tab_control gets the control on demand which is why it works, but stuff like get_tab_count will not see this newest child).
If you print immediately after the add_child call, you can see that the current tab is set immediately and doesn't need to wait a frame.
@kitbdev You are right, the state of the TabContainer in 4.3 and 4.2.2 is not really different. current_tab as you said just defaults to -1 now instead of 0. If I print get_tab_count() both in 4.2.2 and 4.3 0 is being printed, so 4.2.2 is actually the inconsistent one with get_current_tab_control() not being null while get_tab_count() is 0.
The problem is from the perspective of the user I think it is a fair assumption that a TabContainer should be useable when its children are ready. Else you run into this dilemma where you have to filter out the case where a child emits a signal you want to react to, but cant because the TabContainer is not set up properly yet. Workarounds are possible, but not very convenient.
I'll leave this issue open for the responsible people to decide whether the current behaviour has to be changed/can be changed or if this is how it should be.
Probably caused by call_deferred in the Godot code. I really dont understand why this method is used that often.
Deferring something should only be used when needed, e.g. there must be a good reason (queueing something in, e.g. drawing, focus something later, ...).
Having said that, I agree that this is probably not intended to work.
I think the node is added first, ready is emitted and after that the tab_container will react to the change (add_child_notify).
In GDScript, it's easier to deal with this, since you can use await get_tree().process_frame to solve many issues. However, sometimes you need to use it twice in a row, possibly because a deferred method is called somewhere after using call_deferred(), which makes it harder to solve this issue in C++.
