hessian: he(...,atomic = usingAtomics())

[Alexios Galanos]

I've noticed a strange behavior when checking my package against some O/S on rhub.

Assume I set up an object as:

    fun <- MakeADFun(data = data, parameters = par_list, DLL = "mydll_TMBExports",
                         map = map, trace = FALSE, silent = TRUE)

When I try to extract the hessian using:

fun$he(optimal_pars)

I "sometimes" get:
Error in `EvalADFunObject()`:
! Atomic 'set_dependent' order not implemented.

When I debug and check usingAtomics(), it does not exist. However, using fun$env$usingAtomics() is ok  i.e. fun$he(optimal_pars, atomic = fun$env$usingAtomics())

This happens on the linux flavors of rhub (for checking), but not on macos.

I'm just curious whether anyone has since this behavior or am I doing something wrong?

Thanks,

Alexios