Type instability in jacobian! filling JacobianResult when using StaticArrays

[mapclyps]

Running the following code

using ForwardDiff
using StaticArrays

f(x)=SA[x[1]^2+x[2]^2, x[2]^2+x[3]^2]

x=SA[1.0,2,3]

y=f(x)

imdr=DiffResults.JacobianResult(y,x)

@code_warntype ForwardDiff.jacobian!(imdr,f,x) 

shows up as not type stable for me in Julia 1.8.5.
Maybe also related is this older topic on discourse https://discourse.julialang.org/t/am-i-using-diffresults-jl-correctly/35894

I think it is coming from

ForwardDiff.jl/ext/ForwardDiffStaticArraysExt.jl

Line 108 in 5bd7116

result = DiffResults.value!(d -> value(T,d), result, ydual)

Replacing the anonymous function d -> value(T,d) with Base.Fix1(value,T) fixes it for me.
I can make a PR in case this is the desired solution.

[mcabbott]

Base.Fix1(value,T) fixes it for me.

This seems completely safe, a PR would be great.

[ojwoodford]

Possibly related.

[mapclyps]

I was trying to write a test for my proposed fix, but it did not have an inference issue in the tests. It seems like it depends on whether something else is executed before.

using ForwardDiff
using StaticArrays
using Test

sx = @SArray rand(3,3)
addone(A) = A.+one(eltype(A))
sy = addone(sx)

sresult = DiffResults.JacobianResult(sy, sx)
ff(A)=sin.(A)

@inferred ForwardDiff.jacobian!(sresult, ff, sx) #throws the error (type instability) as reported
@inferred ForwardDiff.jacobian!(sresult, addone, sx) #no error 

Swapping the last two lines, interestingly, removes the error entirely.
It seems that there is some hidden state or something?

[KristofferC]

All of this seems to work now