Type Collector meta-issue

[gtrepta]

stable-mir-json/src/printer.rs

Lines 490 to 494 in 8c178ef

	struct TyCollector<'tcx> {
	tcx: TyCtxt<'tcx>,
	types: TyMap,
	resolved: HashSet<stable_mir::ty::Ty>,
	}

TODO

• Emit signatures for monomorphized FnDefs
  

  stable-mir-json/src/printer.rs

  Lines 534 to 538 in 8c178ef

  	TyKind::RigidTy(RigidTy::FnDef(def, ref args)) => {
  	self.resolved.insert(*ty);
  	let instance = Instance::resolve(def, args).unwrap();
  	self.visit_instance(instance)
  	}

  
  

  stable-mir-json/src/printer.rs

  Lines 508 to 514 in 8c178ef

  	fn visit_instance(&mut self, instance: Instance) -> ControlFlow<<Self as Visitor>::Break> {
  	let fn_abi = instance.fn_abi().unwrap();
  	let mut inputs_outputs: Vec<stable_mir::ty::Ty> =
  	fn_abi.args.iter().map(|arg_abi| arg_abi.ty).collect();
  	inputs_outputs.push(fn_abi.ret.ty);
  	inputs_outputs.super_visit(self)
  	}

  
  The resolved argument and return types get added to the collector's TyMap, but should we also add the type for the resolved function signature?

• Investigate CoroutineWitness
  

  stable-mir-json/src/printer.rs

  Lines 532 to 533 in 8c178ef

  	// Break on CoroutineWitnesses, because they aren't expected when getting the layout
  	TyKind::RigidTy(RigidTy::CoroutineWitness(..)) => ControlFlow::Break(()),

• Investigate whether or not we're adding unmonomorphized/mir to the type map.
  Look into anything offered by the compiler interface like predicates, or match/print statements we can come up with to catch anything.