It seems there are many generated shims during the codegen process, and these shims do have MIR bodies and we are also able to fetch the bodies. my question is, when doing codegen, do we distinguish shims, or we just do codegen for all instances? I am asking because I am looking into codegening call_once
of Fn traits but it seems the parameters etc does not look quite the same.
Shims are compiled like normal functions. They just have special InstanceDef
variants.