Type instability when specifying solver options

[densmojd]

Solving an ODE with default options is fine, specifying an option results in a type instability. Perhaps related to #2530.

using OrdinaryDiffEq

prob = ODEProblem{false}((u, p, t) -> -u, 1.0, (0.0, 1.0))

@code_warntype solve(prob) # type stable
@code_warntype solve(prob, reltol = 1E-3) # type unstable

I see similar behavior when specifying abstol and dt.

I'm using Julia 1.11.2 and OrdinaryDiffEq.jl v6.90.1

[oscardssmith]

The problem here is that the automatic algorithm detection isn't written in the best way. To fix it, we likely want to move the OrdinaryDiffEq automatic alg detection to use the same mechanism NonlinearSolve uses.