evdev device resources are not correctly released when kanata is killed by signals about kanata HOT 2 CLOSED

I'm not yet sure if what you're saying is correct for a terminating process.

Cleaning up using the functions you describe makes sense for a non-terminating process to not leak resources over time. However, for a terminating process, a non-buggy OS kernel must clean up after processes that die even if they did not do so themselves. Otherwise the kernel would run out of memory or file descriptors due to buggy programs or due to repeated killing via SIGKILL.

What specific resources are not cleaned up correctly? In the previous issue you mentioned a symlink, but if I understand correctly that was unrelated to evdev so that one is not relevant for this bug.

from kanata.

You are probably right about the cleaning stuff done by OS. And if you are right, there is no issue with current kanata.

What specific resources are not cleaned up correctly?

I have no idea what resources are needed by evdev devices.

In the previous issue you mentioned a symlink

Yeah, I come across this "issue" when I think about cleaning a symlink in the signal handler. I missed the point that OS should clean evdev resources for terminated kanata.

Sorry for the noise.

from kanata.