Crashes while unhooking

[OneXDeveloper]

Hi, me again! Been awhile!

I'm still facing crashes frequently while ejecting for DX12 game D2R. I also sometimes get crashes while injecting. I think there might still be a sync issue somewhere. Could you help me debug? What info can I give you that would help?

Here's the last log I get when it crashes during eject.

DEBUG hudhook::hooks::dx12: WndProc called before hook was set

If it would be easier to chat over discord, I'd be down to help track these down. They're driving me crazy :P

[veeenu]

Hi! Thank you for reporting this!

Unfortunately it's not a great moment for me, between commitments and other personal issues, so I've paused work on hudhook for a bit. I'll still review pull requests though, so if you get to this before I have a chance to, feel free to open one!

The culprit is most likely around here:

hudhook/src/hooks/dx12.rs

Lines 782 to 809 in bebcddc

	unsafe fn unhook(&mut self) {
	trace!("Disabling hooks...");
	CQECL_RUNNING.wait();
	PRESENT_RUNNING.wait();
	RBUF_RUNNING.wait();
	trace!("Cleaning up renderer...");
	if let Some(renderer) = IMGUI_RENDERER.take() {
	let mut renderer = renderer.lock();
	// XXX
	// This is a hack for solving this concurrency issue:
	// https://github.com/veeenu/hudhook/issues/34
	// We should investigate deeper into this and find a way of synchronizing with
	// the moment the actual resources involved in the rendering are
	// dropped. Using a condvar like above does not work, and still
	// leads clients to crash.
	//
	// The 34ms value was chosen because it's a bit more than 1 frame @ 30fps.
	thread::sleep(Duration::from_millis(34));
	renderer.cleanup(None);
	}
	drop(IMGUI_RENDER_LOOP.take());
	COMMAND_QUEUE_GUARD.take();
	DXGI_DEBUG_ENABLED.store(false, Ordering::SeqCst);
	}

The synchronization is already a hack, and a potential solution is included in the comment, but I (personally) need a better understanding of the DX12 pipeline to debug that. Attaching a debugger to inspect the crash should provide some insight.