Caching of numba functions leads to error

Question

Whenever I use caching on a function, ..

@jit(cache=True)
def test(x):
    return x

.. I get a RuntimeError similar to ..

RuntimeError: cannot cache function 'test.test': no locator available for file '/tmp/ipykernel_3392909/373547737.py'

.. and yes, /tmp should be writable (rwxrwxrwt).

Whats happening here?

Answer 1

It seems to have been caused by a mix-up of conda and pip. I thought conda would prevent that, but no. After having wiped pip and conda completely, it now works by using conda alone.

Additionally, there seems to be a bug involved.

Advice to the outside world: Don't use conda if you don't have a special reason to.