[PATCH] D155818: [CloneFunction][DebugInfo] Clone DISubprogram's local types

[David Blaikie via Phabricator via llvm-commits]

dblaikie added a comment.

> Additionally, when tracking function-local types in the DICompileUnit (via retainedTypes or enums fields), it seems reasonable to track the clones as well.

Hmm, when do we have function-local types in the DICompileUnit list? I thought we always put those on the subprogram now?



================
Comment at: llvm/lib/Transforms/Utils/CloneFunction.cpp:299-307
+          auto It = VMap.MD().find(Type);
+
+          if (It != VMap.MD().end() && It->first != It->second)
+            TypesToAdd.push_back(It->second.get());
+          else if (It == VMap.MD().end()) {
+            Metadata *Cloned = MDNode::replaceWithDistinct(Type->clone());
+            TypesToAdd.push_back(Cloned);
----------------
maybe turn this into a lambda or helper function - since it's 10 lines that's repeated twice, by the looks of it? (can pass in the list to push back to so one pushes back to the enums, and the other pushes back to the types)


================
Comment at: llvm/lib/Transforms/Utils/CloneFunction.cpp:311-318
+      SmallVector<Metadata *> NewEnums(CU->getEnumTypes().begin(),
+                                       CU->getEnumTypes().end()),
+          NewTypes(CU->getRetainedTypes().begin(),
+                   CU->getRetainedTypes().end());
+      NewEnums.append(EnumsToAdd);
+      NewTypes.append(TypesToAdd);
+      CU->replaceEnumTypes(MDNode::get(CU->getContext(), NewEnums));
----------------
Maybe this could be streamlined a bit too - maybe doesn't saves lines, but makes it easier to read?



Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D155818/new/

https://reviews.llvm.org/D155818