Siu Kwan Lam

I confirmed that the patch fixes double-free segfault for `test_issue_8800`, `test_issue_9735`

Need to check what happens when the extra incref happened on a value that is never box; for example, the generator is used by another jit function.

Thank you for reporting the issue. I can replicate it on mainline (84f2b5a1148b2266f5d3988d43e0983bea9616a3).

@OutSquareCapital, we discussed this at the developer meeting. We recognize that the current API designs are making things difficult for typing annotations, but it is more important for us to...

https://github.com/numba/numba/pull/5938 is helping in this direction

@oleksii-kulyk, The `@overload` has unusual semantic comparing to regular python code. There's a "typing layer" and a "execution layer". The reason you are seeing the type mismatch is because you...

This is a known issue. For now, most type expression has to be done outside of compiled code.

There are a lot of redundant CFG computation in SSA and other parts. We need to find a way to get rid of them. (EDIT: this comment is about the...

I can confirm this issue. I have tracked it down to this change 96981f25dfc61e44101c171db4f52b41d22c4ba3. @DrTodd13 , why do we always need `force_tuple=True`? The user's script works when it is set...

OP issue confirmed. It highlights the problem in the current jitclass meta-class wiring. The decorator is returning a [JitClassType](https://github.com/numba/numba/blob/289f411613ad995866ad08549b1dfe766c758fa3/numba/experimental/jitclass/base.py#L82) but `instance.__class__` is a [subclass of `_box.Box`](https://github.com/numba/numba/blob/289f411613ad995866ad08549b1dfe766c758fa3/numba/experimental/jitclass/boxing.py#L179). The former wires in...