Bradley Dice

Thanks @davidwendt @wence- @PointKernel for the reviews!

The `hash_combine` step is a crucial part of our hasher. It ensures that hashing is sensitive to the order of input columns. Other implementations may choose to serialize all inputs...

@brandon-b-miller Is there a way to split this PR into pieces? Just trying to help revive it so it can get into 24.10.

@mroeschke I added the label `wontfix` -- please use that when closing issues for this reason.

RAPIDS has been updated to CCCL 2.5 (a pre-release commit for now, and eventually the 2.5.0 release).

Thanks for this @mroeschke! We see this a lot in RAPIDS CI. A fix would be really helpful.

Are there any ways in which CCCL's `cuda::memcpy_async` could help us here? https://nvidia.github.io/cccl/libcudacxx/extended_api/asynchronous_operations/memcpy_async.html I'm not super familiar with the facilities libcudacxx provides for this, but perhaps this could be in...

This is still an issue -- the wrong error is raised when comparing nested types. ```python >>> cudf.Series([[1, 2, 3], [4, 5]]) == cudf.Series([{"a": 1}, {"a": 2}]) ... TypeError: object...

I'm going to try and refactor https://github.com/rapidsai/cudf/blob/b8503bc000f19b983b19292b16f0048254f2b3a9/cpp/src/binaryop/compiled/binary_ops.cu#L399 into an equality/comparator operator for `cudf::binary_operation` that works with lists and structs. That should help unblock the lists/structs cases in this PR.

> Are these issues the best way to communicate? Issues (or comments on pull requests) are good for communication about specific topics in the code or coordinating work. Slack is...