Support inlining of calls where type arguments are passed to generic functions.

runtime/vm/flow_graph_type_propagator.cc

Index: runtime/vm/flow_graph_type_propagator.cc
diff --git a/runtime/vm/flow_graph_type_propagator.cc b/runtime/vm/flow_graph_type_propagator.cc
index fe3f74222035aec52e969c3e02d3af13226a1ca3..775d646b7ba7ffc3008fe80ae944c3c99dc87fdb 100644
--- a/runtime/vm/flow_graph_type_propagator.cc
+++ b/runtime/vm/flow_graph_type_propagator.cc
@@ -980,9 +980,9 @@ CompileType RelationalOpInstr::ComputeType() const {
 }
 
 
-CompileType CurrentContextInstr::ComputeType() const {
-  return CompileType(CompileType::kNonNullable, kContextCid,
-                     &Object::dynamic_type());
+CompileType SpecialParameterInstr::ComputeType() const {
+  intptr_t cid = kind() == kContext ? kContextCid : kTypeArgumentsCid;

[Vyacheslav Egorov (Google), 2017/07/03 16:15:43]

maybe make a switch here?

Then we would know which places to update if we add a new kind of special
parameter.

I think in general:

switch (kind()) {
 case A:
 case B:
}

is preferred to if (kind() == A) { } else { ASSERT(kind() == B); }

because in the first pattern compiler will issue a compile time error if switch
is not exhaustive. 

[regis, 2017/07/05 18:41:29]

Done.

+  return CompileType(CompileType::kNonNullable, cid, &Object::dynamic_type());
 }