[PowerPC][AIX] Support #pragma comment copyright for AIX

Author: Tony Varghese

clang/include/clang/Basic/DiagnosticParseKinds.td

@@ -1365,6 +1365,9 @@ def err_pragma_comment_unknown_kind : Error<"unknown kind of pragma comment">;
 // PS4 recognizes only #pragma comment(lib)
 def warn_pragma_comment_ignored : Warning<"'#pragma comment %0' ignored">,
   InGroup<IgnoredPragmas>;
+def warn_pragma_comment_once : Warning<"'#pragma comment %0' "
+  "can be specified only once per source file - ignored">,
+  InGroup<IgnoredPragmas>;
 // - #pragma detect_mismatch
 def err_pragma_detect_mismatch_malformed : Error<
   "pragma detect_mismatch is malformed; it requires two comma-separated "

clang/include/clang/Basic/PragmaKinds.h

@@ -17,7 +17,8 @@ enum PragmaMSCommentKind {
   PCK_Lib,      // #pragma comment(lib, ...)
   PCK_Compiler, // #pragma comment(compiler, ...)
   PCK_ExeStr,   // #pragma comment(exestr, ...)
-  PCK_User      // #pragma comment(user, ...)
+  PCK_User,     // #pragma comment(user, ...)
+  PCK_Copyright // #pragma comment(copyright, ...)
 };
 
 enum PragmaMSStructKind {

clang/lib/AST/TextNodeDumper.cpp

@@ -2523,6 +2523,9 @@ void TextNodeDumper::VisitPragmaCommentDecl(const PragmaCommentDecl *D) {
   case PCK_User:
     OS << "user";
     break;
+  case PCK_Copyright:
+    OS << "copyright";
+    break;
   }
   StringRef Arg = D->getArg();
   if (!Arg.empty())

clang/lib/CodeGen/CodeGenModule.cpp

@@ -1555,6 +1555,14 @@ void CodeGenModule::Release() {
 
   EmitBackendOptionsMetadata(getCodeGenOpts());
 
+  // Emit copyright metadata for AIX
+  if (!AIXCopyrightComment.empty()) {
+    auto *NMD =
+        getModule().getOrInsertNamedMetadata("aix.copyright.comment");
+    for (auto *MD : AIXCopyrightComment)
+      NMD->addOperand(MD);
+  }
+
   // If there is device offloading code embed it in the host now.
   EmbedObject(&getModule(), CodeGenOpts, *getFileSystem(), getDiags());
 
@@ -3317,6 +3325,23 @@ void CodeGenModule::AddDependentLib(StringRef Lib) {
   LinkerOptionsMetadata.push_back(llvm::MDNode::get(C, MDOpts));
 }
 
+/// Process the #pragma comment(copyright, " copy right string ")
+/// and create llvm metadata for the copyrgiht
+void CodeGenModule::ProcessPragmaCommentCopyright(StringRef Comment) {
+
+  // Pragma Comment Copyright is enabled only when:
+  //  - OS is AIX
+  //  - Comment is non empty
+  if (!getTriple().isOSAIX() || Comment.empty())
+    return;
+
+  // Create llvm metadata with the comment string
+  auto &C = getLLVMContext();
+  llvm::Metadata *Ops[] = {llvm::MDString::get(C, Comment.str())};
+  auto *Node = llvm::MDNode::get(C, Ops);
+  AIXCopyrightComment.push_back(Node); // This should be available during the runtime
+}
+
 /// Add link options implied by the given module, including modules
 /// it depends on, using a postorder walk.
 static void addLinkOptionsPostorder(CodeGenModule &CGM, Module *Mod,
@@ -7453,6 +7478,9 @@ void CodeGenModule::EmitTopLevelDecl(Decl *D) {
     case PCK_Lib:
         AddDependentLib(PCD->getArg());
       break;
+    case PCK_Copyright:      
+      ProcessPragmaCommentCopyright(PCD->getArg());
+      break;
     case PCK_Compiler:
     case PCK_ExeStr:
     case PCK_User:

clang/lib/CodeGen/CodeGenModule.h

@@ -587,6 +587,9 @@ class CodeGenModule : public CodeGenTypeCache {
   /// A vector of metadata strings for dependent libraries for ELF.
   SmallVector<llvm::MDNode *, 16> ELFDependentLibraries;
 
+  /// A vector of metadata strings for copyright comment for AIX
+  SmallVector<llvm::MDNode *, 16> AIXCopyrightComment;
+
   /// @name Cache for Objective-C runtime types
   /// @{
 
@@ -1458,6 +1461,8 @@ class CodeGenModule : public CodeGenTypeCache {
   /// Appends a dependent lib to the appropriate metadata value.
   void AddDependentLib(StringRef Lib);
 
+  /// Append AIX copyright comment to the module-level metadata.
+  void ProcessPragmaCommentCopyright(StringRef Comment); 
 
   llvm::GlobalVariable::LinkageTypes getFunctionLinkage(GlobalDecl GD);
 

clang/lib/Parse/ParsePragma.cpp

@@ -473,7 +473,8 @@ void Parser::initializePragmaHandlers() {
   PP.AddPragmaHandler(OpenACCHandler.get());
 
   if (getLangOpts().MicrosoftExt ||
-      getTargetInfo().getTriple().isOSBinFormatELF()) {
+      getTargetInfo().getTriple().isOSBinFormatELF() ||
+      getTargetInfo().getTriple().isOSAIX()) {
     MSCommentHandler = std::make_unique<PragmaCommentHandler>(Actions);
     PP.AddPragmaHandler(MSCommentHandler.get());
   }
@@ -595,7 +596,8 @@ void Parser::resetPragmaHandlers() {
   OpenACCHandler.reset();
 
   if (getLangOpts().MicrosoftExt ||
-      getTargetInfo().getTriple().isOSBinFormatELF()) {
+      getTargetInfo().getTriple().isOSBinFormatELF() ||
+      getTargetInfo().getTriple().isOSAIX()) {
     PP.RemovePragmaHandler(MSCommentHandler.get());
     MSCommentHandler.reset();
   }
@@ -3207,6 +3209,7 @@ void PragmaCommentHandler::HandlePragma(Preprocessor &PP,
     .Case("compiler", PCK_Compiler)
     .Case("exestr",   PCK_ExeStr)
     .Case("user",     PCK_User)
+    .Case("copyright", PCK_Copyright)
     .Default(PCK_Unknown);
   if (Kind == PCK_Unknown) {
     PP.Diag(Tok.getLocation(), diag::err_pragma_comment_unknown_kind);
@@ -3219,6 +3222,19 @@ void PragmaCommentHandler::HandlePragma(Preprocessor &PP,
     return;
   }
 
+  // pragma comment copyright can each appear only once in a TU.
+  if (PP.getTargetInfo().getTriple().isOSAIX()) {
+    static bool SeenAIXCopyright = false;
+    if (Kind == PCK_Copyright) {
+      if (SeenAIXCopyright) {
+        PP.Diag(Tok.getLocation(), diag::warn_pragma_comment_once)
+            << II->getName();
+        return;
+      }
+      SeenAIXCopyright = true;
+    }
+  }
+
   // Read the optional string if present.
   PP.Lex(Tok);
   std::string ArgumentString;
@@ -3245,6 +3261,12 @@ void PragmaCommentHandler::HandlePragma(Preprocessor &PP,
     return;
   }
 
+  if (PP.getTargetInfo().getTriple().isOSAIX()) {
+    // Accept and ignore well-formed copyright with empty string.
+    if(Kind == PCK_Copyright && ArgumentString.empty())
+      return;
+  }
+
   // If the pragma is lexically sound, notify any interested PPCallbacks.
   if (PP.getPPCallbacks())
     PP.getPPCallbacks()->PragmaComment(CommentLoc, II, ArgumentString);

clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix-multi-lto.c

@@ -0,0 +1,110 @@
+// REQUIRES: powerpc-registered-target, system-aix, clang
+//
+// This test verifies correct handling of `#pragma comment(copyright, ...)`
+// on AIX for multiple Translation Units (TUs).
+//
+// Each TU defines one `#pragma comment(copyright, "...")` which should:
+//   - Generate a unique read-only `__llvm_copyright` csect containing the string.
+//   - Create a `.ref` directive from at least one function in that TU to the
+//     corresponding copyright symbol.
+//   - Preserve these copyright strings across LTO and ThinLTO linking.
+//
+// -----------------------------------------------------------------------------
+//  Build WITHOUT LTO
+// -----------------------------------------------------------------------------
+// RUN: split-file %s %t
+//
+// RUN: %clang_cc1 -triple powerpc-ibm-aix -emit-llvm-bc %t/file1.c -o %t/file1.bc
+// RUN: %clang_cc1 -triple powerpc-ibm-aix -emit-llvm-bc %t/file2.c -o %t/file2.bc
+// RUN: %clang_cc1 -triple powerpc-ibm-aix -emit-llvm-bc %t/file3.c -o %t/file3.bc
+// RUN: %clang_cc1 -triple powerpc-ibm-aix -emit-llvm-bc %t/main.c  -o %t/main.bc
+//
+// Compile each bitcode file to XCOFF object and link them together:
+// RUN: %clang -c %t/file1.bc -o %t/file1.o
+// RUN: %clang -c %t/file2.bc -o %t/file2.o
+// RUN: %clang -c %t/file3.bc -o %t/file3.o
+// RUN: %clang -c %t/main.bc  -o %t/main.o
+// RUN: %clang %t/file1.o %t/file2.o %t/file3.o %t/main.o -o %t/nonlto.exe
+//
+// Verify assembly emission and linked outputs:
+// RUN: llc -mtriple=powerpc-ibm-aix -filetype=asm %t/file1.bc -o - | FileCheck %s --check-prefix=CHECK-ASM
+// RUN: /bin/strings -a %t/nonlto.exe | FileCheck %s --check-prefix=CHECK-STRINGS
+// RUN: llvm-nm %t/nonlto.exe | FileCheck %s --check-prefix=CHECK-NM
+// RUN: llvm-objdump -r %t/nonlto.exe | FileCheck %s --check-prefix=CHECK-OBJDUMP
+//
+// -----------------------------------------------------------------------------
+//  Build WITH Full LTO
+// -----------------------------------------------------------------------------
+// RUN: %clang -flto %t/file1.bc %t/file2.bc %t/file3.bc %t/main.bc -o %t/lto.exe
+// RUN: /bin/strings -a %t/lto.exe | FileCheck %s --check-prefix=CHECK-STRINGS
+// RUN: llvm-nm %t/lto.exe | FileCheck %s --check-prefix=CHECK-NM
+// RUN: llvm-objdump -r %t/lto.exe | FileCheck %s --check-prefix=CHECK-OBJDUMP
+//
+// -----------------------------------------------------------------------------
+//  Build WITH ThinLTO
+// -----------------------------------------------------------------------------
+// RUN: %clang -flto=thin %t/file1.bc %t/file2.bc %t/file3.bc %t/main.bc -o %t/lto-thin.exe
+// RUN: /bin/strings -a %t/lto-thin.exe | FileCheck %s --check-prefix=CHECK-STRINGS
+// RUN: llvm-nm %t/lto-thin.exe | FileCheck %s --check-prefix=CHECK-NM
+// RUN: llvm-objdump -r %t/lto-thin.exe | FileCheck %s --check-prefix=CHECK-OBJDUMP
+//
+// -----------------------------------------------------------------------------
+//  Assembly Checks (for a single TU)
+// -----------------------------------------------------------------------------
+//
+// Verify that the backend:
+//   - Emits a `.ref` directive to tie the string to the TU
+//   - Emits the string in a dedicated read-only csect
+//
+// CHECK-ASM:       .ref __aix_copyright_str
+// CHECK-ASM:       .csect __llvm_copyright[RO],2
+// CHECK-ASM-NEXT:  .lglobl __aix_copyright_str
+// CHECK-ASM:       __aix_copyright_str
+// CHECK-ASM:       .string "Copyright 2025 TU A"
+//
+// -----------------------------------------------------------------------------
+//  Final Binary Checks
+// -----------------------------------------------------------------------------
+//
+// Ensure all TUs’ copyright strings are preserved.
+// CHECK-STRINGS-DAG: Copyright 2025 TU A
+// CHECK-STRINGS-DAG: Copyright 2025 TU B
+// CHECK-STRINGS-DAG: Copyright 2025 TU C
+// CHECK-STRINGS-DAG: Copyright 2025 Main Program
+//
+// Check that the symbols are visible in the binary symbol table.
+// CHECK-NM: t __aix_copyright_str
+// CHECK-NM: t __llvm_copyright
+//
+// Ensure there’s a relocation record referencing the copyright symbol.
+// CHECK-OBJDUMP-LABEL: RELOCATION RECORDS FOR [.text]
+// CHECK-OBJDUMP: R_REF  __aix_copyright_str
+//
+
+//=== file1.c ===
+//--- file1.c
+#pragma comment(copyright, "Copyright 2025 TU A")
+void func1(void) {}
+
+//=== file2.c ===
+//--- file2.c
+#pragma comment(copyright, "Copyright 2025 TU B")
+void func2(void) {}
+
+//=== file3.c ===
+//--- file3.c
+#pragma comment(copyright, "Copyright 2025 TU C")
+void func3(void) {}
+
+//=== main.c ===
+//--- main.c
+#pragma comment(copyright, "Copyright 2025 Main Program")
+void func1(void);
+void func2(void);
+void func3(void);
+int main(void) {
+  func1();
+  func2();
+  func3();
+  return 0;
+}

clang/test/CodeGen/PowerPC/pragma-comment-copyright-aix.c

@@ -0,0 +1,35 @@
+// REQUIRES: powerpc-registered-target, system-aix
+// RUN: %clang_cc1 %s -triple powerpc-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -triple powerpc64-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -triple powerpc-ibm-aix -verify
+// RUN: %clang_cc1 %s -triple powerpc64-ibm-aix -verify
+// RUN: %clang_cc1 %s -DTEST_EMPTY_COPYRIGHT -triple powerpc-ibm-aix -verify
+
+// RUN: %clang_cc1 %s -x c++ -triple powerpc-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -x c++ -triple powerpc64-ibm-aix -O0 -disable-llvm-passes -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -x c++ -triple powerpc-ibm-aix -verify
+// RUN: %clang_cc1 %s -x c++ -triple powerpc64-ibm-aix -verify
+// RUN: %clang_cc1 %s -x c++ -DTEST_EMPTY_COPYRIGHT -triple powerpc-ibm-aix -verify
+
+#ifndef TEST_EMPTY_COPYRIGHT
+// Test basic pragma comment types
+#pragma comment(copyright, "@(#) Copyright")
+
+// Test duplicate copyright - should warn and ignore
+#pragma comment(copyright, "Duplicate Copyright") // expected-warning {{'#pragma comment copyright' can be specified only once per source file - ignored}}
+
+int main() { return 0; }
+
+// Check that both metadata sections are present
+// CHECK: !aix.copyright.comment = !{![[copyright:[0-9]+]]}
+
+// Check individual metadata content
+// CHECK: ![[copyright]] = !{!"@(#) Copyright"}
+
+#else
+// Test empty copyright string - valid with no warning
+#pragma comment(copyright, "") // expected-no-diagnostics
+
+int main() { return 0; }
+
+#endif

llvm/include/llvm/Transforms/Utils/CopyrightMetadataPass.h

@@ -0,0 +1,43 @@
+//===-- CopyrightMetadataPass.h - Lower AIX copyright metadata -*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// The CopyrightMetadataPass lowers the module-level metadata emitted by Clang
+// for `#pragma comment(copyright, "...")` on AIX:
+//
+//     !aix.copyright.comment = !{!"Copyright ..."}
+//
+// into an internal constant string global that is preserved across all compiler
+// and linker stages. Each translation unit produces one TU-local string symbol
+// (`__aix_copyright_str`), and the pass attaches `!implicit.ref` metadata to
+// defined functions referencing this symbol. The PowerPC AIX backend recognizes
+// this metadata and emits `.ref` directives in the XCOFF assembly, ensuring the
+// copyright strings:
+//
+//   • survive optimization and LTO,
+//   • are not removed by linker garbage collection, and
+//   • remain visible in the final binary.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_TRANSFORMS_UTILS_COPYRIGHTMETADATAPASS_H
+#define LLVM_TRANSFORMS_UTILS_COPYRIGHTMETADATAPASS_H
+
+#include "llvm/IR/PassManager.h"
+
+namespace llvm {
+class CopyrightMetadataPass : public PassInfoMixin<CopyrightMetadataPass> {
+public:
+  PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM);
+
+  static bool isRequired() { return true; }
+};
+
+
+} // namespace llvm
+
+#endif // LLVM_TRANSFORMS_UTILS_WYVERN_COPYRIGHTMETADATAPASS_H

llvm/lib/Passes/PassBuilder.cpp

@@ -346,6 +346,7 @@
 #include "llvm/Transforms/Utils/BreakCriticalEdges.h"
 #include "llvm/Transforms/Utils/CanonicalizeAliases.h"
 #include "llvm/Transforms/Utils/CanonicalizeFreezeInLoops.h"
+#include "llvm/Transforms/Utils/CopyrightMetadataPass.h"
 #include "llvm/Transforms/Utils/CountVisits.h"
 #include "llvm/Transforms/Utils/DXILUpgrade.h"
 #include "llvm/Transforms/Utils/Debugify.h"