x' * SMatrix fails for SVector inputs #308

using ReverseDiff, StaticArrays
M = SMatrix{2,2}(1.0, 2.0, 3.0, 4.0)
ReverseDiff.gradient(x -> sum(abs2, x' * M), SVector(3.0, 4.0))

MethodError: no method matching _mul!(::StaticArrays.TSize{(1, 2), :adjoint}, ::MVector{2, Float64}, ::Size{(1, 2)}, ::Size{(2, 2)}, ::Adjoint{Float64, MVector{2, Float64}}, ::SMatrix{2, 2, Float64, 4}, ::StaticArrays.NoMulAdd{Float64, Float64})

transpose(x) * M fails in the same way. With a Vector input, both work.

The tracked input's value is an MVector, so the product's value is an Adjoint of an MVector. mul! then writes the result into the MVector under that Adjoint, and StaticArrays has no _mul! for writing a row vector times a matrix into a plain vector. This happens on master, both before and after #307.

🤖 Generated with Claude Code