Repository navigation
[Solved ?] Memory leak when updating gamepad detection during Runtime #78
Description
In order to update gamepad detection during runtime (as Jinput doesn't), I created a "class patch" to periodicaly nullify registered controllers in DefaultControllerEnvironment class, using reflection.
The code is quite effecive : if I unplug - replug the gamepad, the game correctly detect it !
Problem is, the more time is spent during runtime, the bigger the gap between an "actual gamepad input" and its detection, is... After 2 minutes, it's barely playable ... After 4, the game just doesn't respond to any gamepad input.
As said, the "patch" is quite good at its job ! And would be perfect without this problem... Any suggestion ?
Activity
Valkryst commentedon Oct 24, 2023
I also wrote a hacky solution for detecting new controllers, using a different method. I'm not sure, but maybe this works without the memory leak that you're running into?
sunsigne commentedon Oct 25, 2023
Ty ! I'll give a close look, try it, and give feedback soon :)
sunsigne commentedon Oct 25, 2023
I just tried your "hack", pour over your code to understand what you were doing, slightly adujst it to works with my system, etc.
You did a pretty job there ! It works perfectly ! It's even easier to use that what i was doing.
To clarify for people reading this and willing to have a clarification on what our codes are about :
ControllerEnvironment.getDefaultEnvironment() -> "Env"
[-]Memory leak when updating gamepad detection during Runtime[/-][+]|Solved] Memory leak when updating gamepad detection during Runtime[/+][-]|Solved] Memory leak when updating gamepad detection during Runtime[/-][+][Solved] Memory leak when updating gamepad detection during Runtime[/+]Valkryst commentedon Oct 25, 2023
@sunsigne
Someone just found a memory leak in my solution as well. Currently working on a fix
Endolf commentedon Oct 25, 2023
Looks like there is some momentum behind this version, good to see.
The reason for the RawInputEventQueue creating a thread is that direct input (at least on some older version of windows/directx?) required using the same thread to read the data as opened the device, otherwise no data was being read. It could be modified to only create that thread in certain circumstances...
Linux has a similar issue with the event devices needing a single thread if memory serves.
sunsigne commentedon Oct 25, 2023
Ho... well, i didn't noticed it because mine was visible after 5~10 minutes, as yours is 'only' visible after 40 minutes i didn't see it. I'll give a look on the topic and study the problem... I've no doubt we will eventually find a solution :)
[-][Solved] Memory leak when updating gamepad detection during Runtime[/-][+][Solved ?] Memory leak when updating gamepad detection during Runtime[/+]