Fix a small bug in message origin for warnings from static field acce…

…ss on generic type (#3186) The warning needs to be generated from the origin where the field reference comes from - so if it's a method body, then from the method body. This is because the instantiation is determined by that place, not the field itself (the field is defined on an open generic definition of the type). This syncs the relevants tests with a NativeAOT work-in-progress change, but in generately it adds a lot more cases and some better comments. Co-authored-by: Tlakaelel Axayakatl Ceja <[email protected]>
dotnet · Jan 17, 2023 · c790896 · c790896
1 parent 4452038
commit c790896
Show file tree

Hide file tree

Showing 2 changed files with 320 additions and 22 deletions.
diff --git a/src/linker/Linker.Steps/MarkStep.cs b/src/linker/Linker.Steps/MarkStep.cs
@@ -1610,7 +1610,7 @@ protected void MarkField (FieldReference reference, DependencyInfo reason, in Me
 				Debug.Assert (reason.Kind == DependencyKind.FieldAccess || reason.Kind == DependencyKind.Ldtoken);
 				// Blame the field reference (without actually marking) on the original reason.
 				Tracer.AddDirectDependency (reference, reason, marked: false);
-				MarkType (reference.DeclaringType, new DependencyInfo (DependencyKind.DeclaringType, reference), new MessageOrigin (Context.TryResolve (reference)));
+				MarkType (reference.DeclaringType, new DependencyInfo (DependencyKind.DeclaringType, reference));
 
 				// Blame the field definition that we will resolve on the field reference.
 				reason = new DependencyInfo (DependencyKind.FieldOnGenericInstance, reference);