Thanks! Launched.

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.96%. Comparing base (a2bdbe5) to head (d00be4b).
Report is 23 commits behind head on main.

@@           Coverage Diff           @@
##             main     #710   +/-   ##
=======================================
  Coverage   99.96%   99.96%           
=======================================
  Files          63       63           
  Lines        2754     2754           
=======================================
  Hits         2753     2753           
  Misses          1        1           

Did not find Miniforge3-Linux-arm64.sh latest in cache

Perhaps see conda-incubator/setup-miniconda#385

Thanks! That should be fixed with 6b05be5.

Also, I'm sorry – I didn't notice that there is an open PR for this: #315. I should have checked! I noticed a few prior attempts when scrolling through #288, but I didn't see that PR as it wasn't linked to the issue 🙈

I would be happy to go forward with whatever is suggested.

In the meantime, I added a CHANGELOG entry for this change via 82a4565. I am happy to preserve authorship from the previous PRs or close my PR; either option is acceptable to me.

Ah, thanks – I'll have to take a look at that; I don't immediately see why that happens. The only reference to pthread_atfork is in https://github.com/Blosc/c-blosc/blob/dcf6813d46f9fcbf7af4395b5536f886dfc76dd0/blosc/blosc.c#L2225, but https://anaconda.org/conda-forge/blosc/ and https://anaconda.org/conda-forge/python-blosc/ successfully ship AArch64, so maybe there's something to uncover there. The recommendation is to switch to blosc-2, though: Blosc/c-blosc#268, xref: #413, and that might require supporting either both or dropping blosc-1 because of the lack of forward compatibility.

I guess we aren't linking pthread properly, so the symbol isn't resolved.

@joshmoore, the tests should be fixed now. They ran on my fork and passed (see logs); the failure sign there is just because the Codecov token won't be set up in my fork's settings. Could you please relaunch the workflows here?

Launched.

Thanks! :) The Ubuntu-3.11 failure looks like a sporadic one.

Thanks @agriyakhetarpal, I am excited for this to land!

When launching the devcontainer of pipefunc building numcodecs takes like 99% of the time, even though I am installing ≈200 dependencies.

Hope this PR will have the momentum to get merged (unlike #315 which seems inactive).

Thanks, @basnijholt; I'm happy to address any review comments that might come up, though I think the PR is complete from my end. I intend to keep it up to date. 😃

(Relaunched the flaky 3.11 build)

@dstansby, thanks for the review! All good to go, now.

Relaunched.