Unable to pass config to call back into AD

[rcalxrc08]

I have the following:

function my_func(x)
	return sin(x)
end
using ChainRulesCore
using AbstractDifferentiation: derivative, jacobian
using Diffractor: DiffractorForwardBackend, DiffractorRuleConfig
function ChainRulesCore.frule(config::RuleConfig{>:HasForwardsMode}, (_, dx), ::typeof(my_func), x)
    res=my_func(x)
    _, der_fwd = ChainRulesCore.frule_via_ad(config, (ChainRulesCore.NoTangent(), dx), sin, x)
    @show "hello from fd with right config with calling back"
    return res, der_fwd
end

function my_func2(x)
	return sin(x)
end

function ChainRulesCore.frule((_, dx), ::typeof(my_func2), x)
    res=my_func2(x)
    _, der_fwd = ChainRulesCore.frule_via_ad(DiffractorRuleConfig(), (ChainRulesCore.NoTangent(), dx), sin, x)
    @show "hello from fd with wrong config with calling back"
    return res, der_fwd
end

derivative(DiffractorForwardBackend(), my_func, 1.0)
derivative(DiffractorForwardBackend(), my_func2, 1.0)

and I get the output just for the second call, in the first case the frule has something wrong. Is it supported this way of defining the rules for Diffractor.jl? Am I doing anything wrong?

[oxinabox]

You are not, you are doing everything correct.
Its just that right now diffractor forward mode isn't using its config.
I am not sure why it was commented out.
See here: https://github.com/JuliaDiff/Diffractor.jl/blob/b3e4ee058ef0ae59bc5c7623ed6770b52b6324f2/src/stage1/forward.jl#L110C17-L110C37
Could be we can just uncomment it out and add a test and be good to go.
It might have just been commented out while investigating something and forgotten to undone.
Or it might be something more annoying like inference issues.