RFC: Introduce DW_OP_LLVM_memory to describe variables in memory with dbg.value

I wouldn’t go as far to call it inconsistent, it is just that an IR dbg.value describing an SSA value has different semantics than a MIR DBG_VALUE (because in MIR we can actually distinguish between it describing a register or a register-indirect memory locations).

I agree that we are too fuzzy about the semantics of DW_OP_stack_value in DIExpressions.

– adrian