[llvm-branch-commits] [clang] [clang] callee_type metadata for indirect calls (PR #117036)

[Matt Arsenault via llvm-branch-commits]

================
@@ -2869,9 +2870,23 @@ static void setLinkageForGV(llvm::GlobalValue *GV, const NamedDecl *ND) {
     GV->setLinkage(llvm::GlobalValue::ExternalWeakLinkage);
 }
 
+static bool hasExistingGeneralizedTypeMD(llvm::Function *F) {
+  llvm::MDNode *MD = F->getMetadata(llvm::LLVMContext::MD_type);
+  if (!MD)
+    return false;
+  return MD->hasGeneralizedMDString();
+}
+
 void CodeGenModule::createFunctionTypeMetadataForIcall(const FunctionDecl *FD,
                                                        llvm::Function *F) {
-  // Only if we are checking indirect calls.
+  if (CodeGenOpts.CallGraphSection && !hasExistingGeneralizedTypeMD(F) &&
+      (!F->hasLocalLinkage() ||
+       F->getFunction().hasAddressTaken(nullptr, /*IgnoreCallbackUses=*/true,
+                                        /*IgnoreAssumeLikeCalls=*/true,
+                                        /*IgnoreLLVMUsed=*/false)))
+    F->addTypeMetadata(0, CreateMetadataIdentifierGeneralized(FD->getType()));
----------------
arsenm wrote:

If clang is emitting the initial IR, then won't any IR based use query be unreliable? Later uses could yet to be determined. But also it shouldn't be fundamentally problematic to be conservatively correct and emit the annotation?

https://github.com/llvm/llvm-project/pull/117036