less jl_get_global reflection

[vtjnash]

The remaining ones I see are

• Base.stdout
• Base.DL_LOAD_PATH
• Core._parse
• Base._atexit
• Base._postoutput
• Core.ARGS
• Base._start
• Base.Libc.Libdl.dlopen
• Base.task_done_hook
• Base.wait
• Base.CoreLogging.logmsg_shim
• __init__
• Base.register_root_module
• Base.loaded_modules_array
• Base.__toplevel__
• Base.require
• aspects of staticdata.c/staticdata_utils.c/precompile.c

(of these, the Base.register_root_module / Base.loaded_modules_array should probably be moved to C still, since it is a thread-safety issue / data-race / deadlock risk right now)

[inkydragon]

This pr removes some export functions from julia.h, maybe it's worth adding a NEWS?

[vtjnash]

We don't have a stable API promise for C code, including exports, so I don't think there is anything NEWS-worthy here; we change exports on every release and that is just what should be expected.

[gbaraldi]

Isn't the yeild thing important if someone is embedding julia?

[vtjnash]

They probably shouldn't call yield either, since that is highly sus even if you aren't embedding Julia and just writing "regular" code

[gbaraldi]

Makes sense, that question aside LGTM

[PallHaraldsson]

@vtjnash, @KristofferC, Can even just jl_stdout_obj be added back in for 1.11.2 (I think it would be enough, but didn't take a good look if it's a first in a series of what would fail, seemingly nothing more is needed to add back)? It broke an R interop package (why PkgEval didn't catch it):

JuliaInterop/JuliaCall#237 (comment)

It's "used" there, but not even in reality, so the easy fix is there, but requires a new merge, and maintainer (@Non-Contradiction) absent...

I would be willing to do some maintenance on that package, at least merge that trivial PR, but I just can't, since I do not have privileges. Long-term this may need to go here in Julia 1.12.

https://github.com/JuliaLang/julia/pull/53250/files#diff-5fd26b7e1f502ffc4d75b7bcf48c484da73800684ecc04951b7765a3f68f3fdbL549

I see jl_stdout_obj(void) is almost identical to the jl_stderr_obj that was kept there. And both are reference in that package.

Are both actually wanted as exported API (or neither, and was it never promised, i.e. experimental?).