nora

Or, alternatively: `strip = true_but_not_user_backtrace` exists, it's called `strip = false`. Everything more will affect backtraces.

People like storing type IDs, so increasing the size again from 128 but to 256 bit would be an increase in memory usage and decrease in performance, which is a...

I think the point of this issue is that as-is, I think *`downcast` is sound today*. If you write safe code, no matter how bad, you will not hit any...

Why exactly are we spending precious time arguing about whether SipHash is more secure than MD5? It's clear that neither of the two are cryptographically secure, and that's not the...

I think even if TypeId was MD5 (which it obviously will never be), my point still stands. 2^18 is, as the name implies, *security*. Against adversarial inputs. Which I did...

You mentioned that you discussed this in the lang meeting, so I removed the label.

Note that `#[used(linker)]` is `__attribute__((used,retain))`, not `__attribute__((used))`, which is just `#[used(compiler)]`. | | Compiler | Linker | |------|-------------------------|-----------------------------------------------------| | GCC C | `__attribute__((used))` | `__attribute__((retain))` (note that you need both)...

The lint name is very long, it would be nice if we could come up with a shorter name. I don't have a suggestion though.

because that makes no sense, the unsafety is at a meta-level and there are no safety requirements on the macro function itself

Yes, we rolled it back for now to fix the warnings on our side and in places like here before enabling it again to reduce fallout. I see that this...