Can't set variable on nested callable program

I have a light sampling program like so:

rtDeclareVariable(float3, L_e, , );
rtDeclareVariable(rtCallableProgramId<float3()>, get_color, , );

RT_CALLABLE_PROGRAM LightSample light_sample(const float3 p, const Onb onb, const float u1, const float u2) {
    LightSample ls;
    float3 omega_l;
    cosine_sample_hemisphere(u1, u2, omega_l);
    ls.omega_i = normalize(omega_l.z * onb.m_normal + omega_l.x * onb.m_tangent + omega_l.y * onb.m_binormal);
    /*ls.L_i = L_e;*/ //< this works fine
    ls.L_i = get_color(); //< this also works fine, but only if get_color() returns a hard-coded value
    ls.pdf = 1.0f;
    return ls;

the callable program get_color is defined like so:

rtDeclareVariable(float3, in_color, , );
RT_CALLABLE_PROGRAM float3 get_color() {
    /* returning the float3 directly here works, returning the variable doesn't */
    /*return make_float3(0.2f, 0.5f, 1.0f);*/
    return in_color;

I can set the variable L_e in the light_sample program and everything works, and if construct and return a hard-coded value in get_color(), that works too, but trying to read from the in_color variable doesn’t work - the function just returns (0,0,0). Both L_e and in_color are set in the exact same way on their respective programs.

Everything validates correctly and no exceptions fire. I’m using OptiX 4.0.2 with cuda 8 on a sm_30 card on OS X. Any ideas?

Never mind, I think I figured it out for myself. If I assign in_color to a local variable in the function, then return that, it works:

rtDeclareVariable(float3, in_color, , );
RT_CALLABLE_PROGRAM float3 get_color() {
    float3 ret = in_color;
    return ret;

Ok, thanks for the report.
That sounds like a bug in OptiX. I’ll file one for investigation.