Michael Abbott

That sounds a bit like piracy, which is bad. Does there seem to be a fix involving sending `()` or `(;)` to `nothing`, maybe using `Zygote._project` as above? That would...

This is from inside the `foldl` rule, on an empty tuple. Which comes from this `foreach` on a leaf node: ``` reset!(m::Recur) = (m.state = m.cell.state0) reset!(m) = foreach(reset!, functor(m)[1])...

I had a go in #1932 . But right now I think a supertype like #2028 might be cleaner than having an ever-expanding the zoo of things you ought to...

Shall we try this out? Seems like a step forwards. I'm not too sure how @Saransh-cpp made this, but I guess replacing a colour is easier than adding new borders...

I actually think this looks fine. Maybe there's a design where the text & arrow have boundaries like the "layers" do, but this sounds like a lot of work, and...

With these two PRs: ```julia julia> @time_imports using ForwardDiff 6.3 ms IrrationalConstants 0.9 ms DiffRules 2.7 ms StaticArraysCore 2.3 ms DiffResults 19.3 ms Preferences 1.2 ms OpenLibm_jll 0.3 ms NaNMath...

CI gives errors like this, which don't show up when running simple examples locally: ``` ...testing specialized StaticArray codepaths Hessians: Error During Test at /home/runner/work/ForwardDiff.jl/ForwardDiff.jl/test/HessianTest.jl:106 Test threw exception Expression: ForwardDiff.hessian(prod,...

Yes that's my guess too. I did not try very hard to replicate locally, but I did use ForwardDiff first, before loading StaticArrays. Scrolling through it looks like all are...

> seems that one can't use Base functionality of static arrays inside of `@generated` functions It might be narrower than that. Functions which act on the instance can't be called...

Unfortunately this is a known problem, e.g. https://github.com/mcabbott/Tullio.jl/issues/30 . I have not worked much on making this faster since then.