Siu Kwan Lam

I will be happy to see improvement to add contextual info for new style error to aid debugging. Python3.11 introduced `Exception.add_note` (https://docs.python.org/3/library/exceptions.html#BaseException.add_note) which can be useful to add context. That...

Triage meeting discussion note: - better to wait until old_style error capturing is fully removed. Then, we do a clean up of those logic before we can implement the `Exception.add_note`....

llvmlite do have `zeroext` and `signext`. Numba as the frontend will need to add the logic to insert those attributes when needed. It seems there's no harm just naively adding...

Actually, for this case of `tuple_new`, it's just a bug. The function should take `Py_ssize_t` type which is `i64` on 64-bit platforms.

For the general LLVM-IR problem, we have been lucky that none of the supported platform cared. I wonder if this is only a problem when dealing functions written in other...

Can you provide a minimal reproducer, i.e. a script that calls `func1`? Numba will specialize the function with respect to the argument types. Thus, the error you see can be...

> > Actually, for this case of `tuple_new`, it's just a bug. The function should take `Py_ssize_t` type which is `i64` on 64-bit platforms. > > Indeed, good catch! I'll...

~But, NVVM can never use a dynamic global. Do you have a situation where Numba is generating that?~ nevermind, just saw the issue comments

I don't want to encourage users to annotate function types. I think the current numba type system is very limiting and it should be replaced. With type annotation, there is...

For python 3.6 and 3.7, class-level annotation shows up in the `cls.__annotations__` ```python In [1]: class A: ...: bar: float ...: In [2]: A.__annotations__ Out[2]: {'bar': float} ```