Graham Markall

Ah, even if I permit the array return type (by commenting out the check in `BaseTypeInference.run_pass.legalize_return_type()`) I hit an error resolving the `real` atrtibute: ``` Traceback (most recent call last):...

Progress towards resolving this issue can be observed in PR #8310.

One potential fix is now ready for consideration in #8310. Whether this is the right fix and the general path forward is a complex decision, so I've added it to...

@PABannier Thanks for identifying the issue! One thing worth trying in general when Numba-compiled code segfaults is to run without the JIT and check for any Python examples. This can...

gpuci run tests

> I understand all that, my intent here is asking for additional input on whether we want an RMM-specific variable (and consequently specific variables for other libraries, e.g., `CUPY_CUDA_PTDS`) or...

> Perhaps @jakirkham and @gmarkall who have more experience than I do have a better suggestion to handle this issue though. I only have a basic working experience of Cython...

For reference, I implemented something similar in Numba ( see https://github.com/numba/numba/pull/6936) and went with `NUMBA_CUDA_PER_THREAD_DEFAULT_STREAM`, for the reasons outlined previously: > In Numba we'd probably want to have an env...

> 1. @gmarkall's original reason for wanting a separate way to disable submodules was because when we mirror a repo, we don't also mirror the submodules (I think, correct me...

@apmasell Many thanks for the updates - could you please push changes as new commits on top of existing history? Otherwise Github starts losing track of review comments and commits...