Function calls made after awaiting on signal emitted from within a thread has a chance of failing abruptly without error.

[KnightNine]

Function calls made after awaiting on signal emitted from within a thread has a chance of failing abruptly without error.

[AThousandShips]

You are on a very outdated branch, please try on a supported branch and if you have your own code in your branch please remove it, bug reports are only for official code

[KnightNine]

After further testing I've found that this issue is only present in my project,
I cannot seem to replicate it in a fresh project.

The potential causes I can think of (as someone who knows little about the core engine stuff) are:

• Something could be interfering with threading across my Autoload scripts
• it's due to the size of my project.
• the import process from 3.X to 4.X has broke something in my project

(can someone with a large project and/or a project that was imported from 3.X try running a scene containing the above code?)

I synced my Godot build with the 4.3 master branch (v4.3.beta.custom_build [44ac302e3]), only preserving my modifications to astar so that my project wouldn't throw a bunch of errors due to making unsupported function calls. The result is that the issue persists.

I'll have to test removing stuff from my Autoload scripts to see if that changes anything and I'll add it here if I find anything.

[KnightNine]

yeah, I have no idea what causes it so far, I've disabled anything that could potentially run within my Autoload scripts and it hasn't resolved it. I'll note that in 4.3 the issue doesn't seem to appear as often when the project window is in focus.
(Also I still can't replicate it in a new project so I'm leaning towards this being somehow related to the project's size)

Really want to know if anyone can replicate it or if it's just me.

[AThousandShips]

Pretty hard to tell, I'd say the first step is to establish if it's independent of your own code, but with the vague contexts it's pretty impossible to tell or to replicate

[KnightNine]

I've further narrowed down the code above.

I can simplify the stability test function to:

func stability_test():
	label.text = ""
	#stability test:
	for i in range(1000):
		await reattach_to_thread()
		await get_tree().process_frame
		label.text += ", "+str(i)
		print(i)

with the issue persisting.

I also noticed that adding a delay or a print statement after the await get_tree().process_frame like so:

	await get_tree().process_frame
	OS.delay_usec(100)

fixes the issue. 🤔

This leads me to believe that this is some sort of deadlocking issue where 'await get_tree().process_frame' is being called at the exact same time as the "process_frame" signal is being emitted, the await doesn't know what to do and crashes. This would explain why it's so inconsistent.

Awaiting on a process frame from a thread isn't something that is restricted in the thread_safety_checks so maybe this possibility wasn't accounted for or seen to be possible.
Though I still can't explain why it only happens in my project...
I'm probably missing something.