Failures are Distributed related (EAGAIN: Resource temporarily not available), seems like some tasks can't spawn theirsetenv(...) invocations. Unrelated to this PR.

What if we printed it likex -> sin(x...)? That way it'd tell you what it did? Or is that too fancy?

Well it'd be lying to you, as that's not how its implemented (edit: in this PR) 🤷 That used to be the old implementation though. Changing it was necessary to allow for easier printing.

Compared to the old way of just printing whatever number anonymous function the result ofsplat happened to be, this representation is already a big step up and allows you to do?Base.splat(sin) (we'd have to export it to make?splat work) and you'll get the docs for it, where it says that this is equivalent to the lambda version.

Since it's not exported, I don't think it's used very wildly right now. I'm working on a package (PropCheck.jl) where usingsplat to interact with it makes everything a bunch easier (see e.g. the tests inBigNums.jl). I got annoyed with not knowing which function was wrapped, hence this PR.

I'm talking about this PR 😅 In this PR, this is no longer a simple anonymous function, precisely because it makes printing easier. If it were not for the new struct, I'd have to callsplat first, to get the anonymous function to customize its printing.

I should note that this works the exact same way as the anonymous function, except doing the hoisting to astruct manually instead of implicitly by the compiler.

string can be overloaded, though it is uncommon, and usually just for performance to give the same result

I think you meant to answer in the review thread above ;)

I've incorporated the feedback, fixed printing, added the deprecation and replaced uses of the old syntax in the rest of Base. It's also rebased to the current master. Now let's hope nothing broke 😬

Rebase & rerun of CI - I've also moved the deprecation notice to the 1.9 deprecation block, since this definitely isn't going to make 1.8 anymore. I'll write NEWS.md once CI is happy.

Add export?

$ /etc/buildkite-agent/plugins/github-com-JuliaCI-external-buildkite-buildkite-plugin-v1/hooks/post-checkout
--
  | fatal: destination path '.buildkite' already exists and is not an empty directory.
  | 🚨 Error: The plugin external-buildkite post-checkout hook exited with status 128

Uuh.. Unrelated? Doesn't look good though..

In either case, I've also added a NEWS.md entry since CI was green last night. Should be good to go now.

The macos failure seems to stem fromLibgit2:

Test Failed at /Users/julia/buildbot/worker-tabularasa/tester_macos64/build/share/julia/stdlib/v1.9/LibGit2/test/libgit2.jl:1754
Expression: LibGit2.headname(repo) == master_branch
Evaluated: "main" == "master"

I don't know where this should fall in the release cycle, but it'd be nice to get this in a future 1.9 :)

Rebase onmaster!

Failure in BuildKite macOS iscmdlineargs, seems unrelated.