docs 1.0 upgrade

[ArnoStrouwen]

No description provided.

[ArnoStrouwen]

@mortenpi is it expected behavior that Documenter complains about LinearAlgebra docstrings, while that module is not included in the modules keyword?
https://github.com/SciML/ExponentialUtilities.jl/actions/runs/6209285369/job/16856334243#step:5:28
https://github.com/SciML/ExponentialUtilities.jl/blob/master/docs/make.jl#L10

[mortenpi]

You're overloading that function and attaching a docstring to it here:

ExponentialUtilities.jl/src/StegrWork.jl

Lines 54 to 67 in f317b06

	"""
	stegr!(α, β, sw)
	Diagonalize the real-symmetric tridiagonal matrix with `α` on the
	diagonal and `β` on the super-/subdiagonal, using the workspaces
	allocated in `sw`.
	"""
	function stegr!(α::AbstractVector{T}, β::AbstractVector{T}, sw::StegrWork{T}) where {T}
	# @assert length(sw.dv) >= length(α)
	# @assert length(sw.ev) >= length(β)
	copyto!(sw.dv, α)
	copyto!(sw.ev, β)
	stegr!(BlasInt(length(α)), sw)
	end

So yes, it's expected. It's searching for docstrings (not functions etc) in your modules.

Edit: I think we can do a better job with the error here though: JuliaDocs/Documenter.jl#2265

[ChrisRackauckas]

Why is it not just the public functions? JuliaLang/julia#50105 defines the public semantics, why wouldn't Documenter follow them?

[mortenpi]

I mean.. that PR got merged like 2 days ago, and doesn't even exist in a Julia release. That said, yes, I do think we want to update Documenter's logic here eventually (JuliaDocs/Documenter.jl#600 (comment)).

[ChrisRackauckas]

Well you were saying you were going to throw a better error in the future. I was just saying that if anything should happen in the future, this case shouldn't error unless it's a public function.

[mortenpi]

Ah. Well, you can still run into the error if you're legitimately extending a function from somewhere else. The confusion here, from my perspective, is that it looks like Documenter is complaining about someone else's docstring. But it's actually complaining about one of yours, which you do want to include in the manual (probably; if it's public stuff that is). So making this error better would be useful regardless.

That said.. LAPACK.stegr! is a public function though, no...? Or if it isn't, then why is ExponentialUtilities mucking with it? Or.. if it's adding like a private method, then the public keyword won't really help here, because you're running into the private method edge case that the new keyword doesn't really handle?