Seemingly, the memory leaks are bugs in userspace -- Mesa, my driver -- rather than the kernelspace -- as we all know @lina wrote a perfect driver, since apparently bugs are impossible in Rust or something like that.

THAT BEING SAID, the memory leaks have exposed one or two actual kernel bugs. No matter what userspace does wrong, it should be impossible to hit kernel WARNs. That it took low memory conditions to hit means that my buggy C code is helping somehow! 😋​

So... I should keep writing buggy C code, right?

("I really don't think that's the message you should take away, Alyssa.")