Ralf Gommers

> The reason it’s on 1.11.2 instead of 1.12 is because the docs were built before Pyodide released 0.26, and to enable the interactive examples we just unhid the buttons....

A full rebuild seems safer to me in that case.

Here's the [build.yml CI log](https://github.com/rgommers/python-xxhash/actions/runs/14959193468) for this commit. Note that the value of the environment variable is reviewable in the `env:` output of the job steps: 

This works for the `test` command but not the `ipython` or `bench` commands, because those commands now _extend_ the upstream `spin` commands with the same name, and within those upstream...

> Why does this now fail when the other ones don't? The Arch/PyPy failure looks unrelated and is happening on other PRs as well. Note that it's a different test...

> Yes there was: https://mail.python.org/pipermail/scipy-dev/2021-June/024904.html (do we have a search for the dev mailing list? I used the date here) No we don't. We'll fix that sometime in the next...

> Meson's install tags don't support arbitrary tags other than `runtime`, `python-runtime`, `devel`, and `tests`, That is supported just fine. On a clean build, try adding `install_tag: uniquetag'` to some...

> However, I we should think about how that might affect (Linux) distribution maintainers that run our tests while building their our package I believe you want to do this...

I can reproduce the problem with `spin` 0.12 for the `numpy.f2py` tests, as well as for SciPy in the in-progress work to move to `spin` (xref https://github.com/scientific-python/spin/pull/238#discussion_r1834428942). @lagru's explanation about...

I think the problem is now clearer to me (thanks @oleksandr-pavlyk and @leofang). The basic problem for the SYCL-based `dpctl` library is that the return value from [`__dlpack_device__`](https://data-apis.org/array-api/latest/API_specification/generated/array_api.array.__dlpack_device__.html#dlpack-device) is not...