Adrian Seyboldt

There are still some test failures, but maybe they are unrelated? The workflow fails even before the tests are started.

We simplified it from a function where the literal_unroll actually did something useful, this was just the shortest example we could find that still resulted in a segfault. And you...

@stuartarchibald We'll go through things with a debugger next week, hopefully we can locate the problem a bit better.

That would still segfault for `axis = -3`. Maybe there should be a helper function that checks axis arguments and raises errors? ```python def normalize_axis(ndim, axis): if axis < 0:...

Ah, maybe this one looks right? https://github.com/numba/numba/blob/main/numba/np/arrayobj.py#L4990 I'm not really sure about how this should be used in the split implementation.

I would also like to see a way to get the transformed variables in the trace. The fact that this isn't possible right now is pretty much the only reason...

I'm seeing the same thing locally as well. The rerender is stuck after the log-line `Adding in variants from argument_variants` with constant high CPU usage.

Can you share an example that produces this error? That would make it easier to debug.

Oh, sorry you said it's the example from the readme, my mistake...

Thank you for the report, I can reproduce this issue. I think this was introduced by a small change in the behavior of the Op.make_node function in aesara, I'll have...