Switch lto codegen to using diagnostic handlers

[ygao]

Summary

With r252791, the C api will install a default diagnostic handler to print to sLastErrorString, therefore
we no longer need to pass around this sLastErrorString to any the LTOCodeGenerator methods.

This patch removes the std::string& argument from a number of C++ LTO API calls and instead
makes them use the installed diagnostic handler. This would also improve consistency of diagnostic
handling infrastructure: if an LTO client used lto_codegen_set_diagnostic_handler() to install a
custom error handler, we do not want some error messages to go through the custom error handler,
and some other error messages to go into sLastErrorString.

The part that I am not entirely sure about is what should happen when there is no installed
diagnostic handler at all. In this patch, the fall-back is to print the message to stderr. For this case
to happen, if the client is using the C api, the client must have explicitly called
lto_codegen_set_diagnostic_handler(nullptr), and maybe the intention is to not print any error at all?
I am not sure.

[mehdi_amini]

Can you upload your patch with context please? (git diff -U9999, or using arc diff)

[• rafael]

This is awesome.

Please just wait for a second opinion from someone at Apple.

[ygao]

Addressing Duncan's comment: when no custom diagnostic handler is installed, use the default
diagnostic handler, which is LLVMContext.diagnose().
Addressing Mehdi's comment. Uploading the patch with more context.
Rebased the patch to trunk r253253.

[mehdi_amini]

Thanks for the cleanup!

[ygao]

Thanks for reviewing. Closed by r253367.

include/llvm/LTO/LTOCodeGenerator.h

struct LTOCodeGenerator {
   ///
   /// Like \a setCodeGenDebugOptions(), this must be called before
   /// LTOCodeGenerator::compilexxx() and
   /// LTOCodeGenerator::writeMergedModules().
   void parseCodeGenDebugOptions();
 
   /// Write the merged module to the file specified by the given path.  Return
   /// true on success.
-  bool writeMergedModules(const char *Path, std::string &ErrMsg);
+  bool writeMergedModules(const char *Path);
 
   /// Compile the merged module into a *single* object file; the path to object
   /// file is returned to the caller via argument "name". Return true on
   /// success.
   ///
   /// \note It is up to the linker to remove the intermediate object file.  Do
   /// not try to remove the object file in LTOCodeGenerator's destructor as we
   /// don't who (LTOCodeGenerator or the obj file) will last longer.
   bool compile_to_file(const char **Name, bool DisableVerify,
                        bool DisableInline, bool DisableGVNLoadPRE,
-                       bool DisableVectorization, std::string &ErrMsg);
+                       bool DisableVectorization);
 
   /// As with compile_to_file(), this function compiles the merged module into
   /// single object file. Instead of returning the object-file-path to the
   /// caller (linker), it brings the object to a buffer, and return the buffer
   /// to the caller. This function should delete intermediate object file once
   /// its content is brought to memory. Return NULL if the compilation was not
   /// successful.
   std::unique_ptr<MemoryBuffer> compile(bool DisableVerify, bool DisableInline,
                                         bool DisableGVNLoadPRE,
-                                        bool DisableVectorization,
-                                        std::string &errMsg);
+                                        bool DisableVectorization);
 
   /// Optimizes the merged module.  Returns true on success.
   bool optimize(bool DisableVerify, bool DisableInline, bool DisableGVNLoadPRE,
-                bool DisableVectorization, std::string &ErrMsg);
+                bool DisableVectorization);
 
   /// Compiles the merged optimized module into a single object file. It brings
   /// the object to a buffer, and returns the buffer to the caller. Return NULL
   /// if the compilation was not successful.
-  std::unique_ptr<MemoryBuffer> compileOptimized(std::string &ErrMsg);
+  std::unique_ptr<MemoryBuffer> compileOptimized();
 
   /// Compile the merged optimized module into out.size() object files each
   /// representing a linkable partition of the module. If out contains more
   /// than one element, code generation is done in parallel with out.size()
   /// threads.  Object files will be written to members of out. Returns true on
   /// success.
-  bool compileOptimized(ArrayRef<raw_pwrite_stream *> Out, std::string &ErrMsg);
+  bool compileOptimized(ArrayRef<raw_pwrite_stream *> Out);
 
   void setDiagnosticHandler(lto_diagnostic_handler_t, void *);
 
   LLVMContext &getContext() { return Context; }
 
 private:
   void initializeLTOPasses();
 
-  bool compileOptimizedToFile(const char **Name, std::string &ErrMsg);
+  bool compileOptimizedToFile(const char **Name);
   void applyScopeRestrictions();
   void applyRestriction(GlobalValue &GV, ArrayRef<StringRef> Libcalls,
                         std::vector<const char *> &MustPreserveList,
                         SmallPtrSetImpl<GlobalValue *> &AsmUsed,
                         Mangler &Mangler);
-  bool determineTarget(std::string &ErrMsg);
+  bool determineTarget();
 
   static void DiagnosticHandler(const DiagnosticInfo &DI, void *Context);
 
   void DiagnosticHandler2(const DiagnosticInfo &DI);
 
+  void emitError(const std::string &ErrMsg);
+
   typedef StringMap<uint8_t> StringSet;
 
   std::unique_ptr<LLVMContext> OwnedContext;
   LLVMContext &Context;
   std::unique_ptr<Module> MergedModule;
   Linker IRLinker;
   std::unique_ptr<TargetMachine> TargetMach;
   bool EmitDwarfDebugInfo = false;

lib/LTO/LTOCodeGenerator.cpp

case 2:
     CGOptLevel = CodeGenOpt::Default;
     break;
   case 3:
     CGOptLevel = CodeGenOpt::Aggressive;
     break;
   }
 }
 
-bool LTOCodeGenerator::writeMergedModules(const char *Path,
-                                          std::string &ErrMsg) {
-  if (!determineTarget(ErrMsg))
+bool LTOCodeGenerator::writeMergedModules(const char *Path) {
+  if (!determineTarget())
     return false;
 
   // mark which symbols can not be internalized
   applyScopeRestrictions();
 
   // create output file
   std::error_code EC;
   tool_output_file Out(Path, EC, sys::fs::F_None);
   if (EC) {
-    ErrMsg = "could not open bitcode file for writing: ";
+    std::string ErrMsg = "could not open bitcode file for writing: ";
     ErrMsg += Path;
+    emitError(ErrMsg);
     return false;
   }
 
   // write bitcode to it
   WriteBitcodeToFile(MergedModule.get(), Out.os(), ShouldEmbedUselists);
   Out.os().close();
 
   if (Out.os().has_error()) {
-    ErrMsg = "could not write bitcode file: ";
+    std::string ErrMsg = "could not write bitcode file: ";
     ErrMsg += Path;
+    emitError(ErrMsg);
     Out.os().clear_error();
     return false;
   }
 
   Out.keep();
   return true;
 }
 
-bool LTOCodeGenerator::compileOptimizedToFile(const char **Name,
-                                              std::string &ErrMsg) {
+bool LTOCodeGenerator::compileOptimizedToFile(const char **Name) {
   // make unique temp .o file to put generated object file
   SmallString<128> Filename;
   int FD;
   std::error_code EC =
       sys::fs::createTemporaryFile("lto-llvm", "o", FD, Filename);
   if (EC) {
-    ErrMsg = EC.message();
+    emitError(EC.message());
     return false;
   }
 
   // generate object file
   tool_output_file objFile(Filename.c_str(), FD);
 
-  bool genResult = compileOptimized(&objFile.os(), ErrMsg);
+  bool genResult = compileOptimized(&objFile.os());
   objFile.os().close();
   if (objFile.os().has_error()) {
     objFile.os().clear_error();
     sys::fs::remove(Twine(Filename));
     return false;
   }
 
   objFile.keep();
   if (!genResult) {
     sys::fs::remove(Twine(Filename));
     return false;
   }
 
   NativeObjectPath = Filename.c_str();
   *Name = NativeObjectPath.c_str();
   return true;
 }
 
 std::unique_ptr<MemoryBuffer>
-LTOCodeGenerator::compileOptimized(std::string &ErrMsg) {
+LTOCodeGenerator::compileOptimized() {
   const char *name;
-  if (!compileOptimizedToFile(&name, ErrMsg))
+  if (!compileOptimizedToFile(&name))
     return nullptr;
 
   // read .o file into memory buffer
   ErrorOr<std::unique_ptr<MemoryBuffer>> BufferOrErr =
       MemoryBuffer::getFile(name, -1, false);
   if (std::error_code EC = BufferOrErr.getError()) {
-    ErrMsg = EC.message();
+    emitError(EC.message());
     sys::fs::remove(NativeObjectPath);
     return nullptr;
   }
 
   // remove temp files
   sys::fs::remove(NativeObjectPath);
 
   return std::move(*BufferOrErr);
 }
 
 bool LTOCodeGenerator::compile_to_file(const char **Name, bool DisableVerify,
                                        bool DisableInline,
                                        bool DisableGVNLoadPRE,
-                                       bool DisableVectorization,
-                                       std::string &ErrMsg) {
+                                       bool DisableVectorization) {
   if (!optimize(DisableVerify, DisableInline, DisableGVNLoadPRE,
-                DisableVectorization, ErrMsg))
+                DisableVectorization))
     return false;
 
-  return compileOptimizedToFile(Name, ErrMsg);
+  return compileOptimizedToFile(Name);
 }
 
 std::unique_ptr<MemoryBuffer>
 LTOCodeGenerator::compile(bool DisableVerify, bool DisableInline,
-                          bool DisableGVNLoadPRE, bool DisableVectorization,
-                          std::string &ErrMsg) {
+                          bool DisableGVNLoadPRE, bool DisableVectorization) {
   if (!optimize(DisableVerify, DisableInline, DisableGVNLoadPRE,
-                DisableVectorization, ErrMsg))
+                DisableVectorization))
     return nullptr;
 
-  return compileOptimized(ErrMsg);
+  return compileOptimized();
 }
 
-bool LTOCodeGenerator::determineTarget(std::string &ErrMsg) {
+bool LTOCodeGenerator::determineTarget() {
   if (TargetMach)
     return true;
 
   std::string TripleStr = MergedModule->getTargetTriple();
   if (TripleStr.empty()) {
     TripleStr = sys::getDefaultTargetTriple();
     MergedModule->setTargetTriple(TripleStr);
   }
   llvm::Triple Triple(TripleStr);
 
   // create target machine from info for merged modules
+  std::string ErrMsg;
   const Target *march = TargetRegistry::lookupTarget(TripleStr, ErrMsg);
-  if (!march)
+  if (!march) {
+    emitError(ErrMsg);
     return false;
+  }
 
   // Construct LTOModule, hand over ownership of module and target. Use MAttr as
   // the default set of features.
   SubtargetFeatures Features(MAttr);
   Features.getDefaultSubtargetFeatures(Triple);
   FeatureStr = Features.getString();
   // Set a default CPU for Darwin triples.
   if (MCpu.empty() && Triple.isOSDarwin()) {
void LTOCodeGenerator::applyScopeRestrictions() {
   passes.run(*MergedModule);
 
   ScopeRestrictionsDone = true;
 }
 
 /// Optimize merged modules using various IPO passes
 bool LTOCodeGenerator::optimize(bool DisableVerify, bool DisableInline,
                                 bool DisableGVNLoadPRE,
-                                bool DisableVectorization,
-                                std::string &ErrMsg) {
-  if (!this->determineTarget(ErrMsg))
+                                bool DisableVectorization) {
+  if (!this->determineTarget())
     return false;
 
   // Mark which symbols can not be internalized
   this->applyScopeRestrictions();
 
   // Instantiate the pass manager to organize the passes.
   legacy::PassManager passes;
 
bool LTOCodeGenerator::optimize(bool DisableVerify, bool DisableInline,
   PMB.populateLTOPassManager(passes);
 
   // Run our queue of passes all at once now, efficiently.
   passes.run(*MergedModule);
 
   return true;
 }
 
-bool LTOCodeGenerator::compileOptimized(ArrayRef<raw_pwrite_stream *> Out,
-                                        std::string &ErrMsg) {
-  if (!this->determineTarget(ErrMsg))
+bool LTOCodeGenerator::compileOptimized(ArrayRef<raw_pwrite_stream *> Out) {
+  if (!this->determineTarget())
     return false;
 
   legacy::PassManager preCodeGenPasses;
 
   // If the bitcode files contain ARC code and were compiled with optimization,
   // the ObjCARCContractPass must be run, so do it unconditionally here.
   preCodeGenPasses.add(createObjCARCContractPass());
   preCodeGenPasses.run(*MergedModule);
LTOCodeGenerator::setDiagnosticHandler(lto_diagnostic_handler_t DiagHandler,
   this->DiagContext = Ctxt;
   if (!DiagHandler)
     return Context.setDiagnosticHandler(nullptr, nullptr);
   // Register the LTOCodeGenerator stub in the LLVMContext to forward the
   // diagnostic to the external DiagHandler.
   Context.setDiagnosticHandler(LTOCodeGenerator::DiagnosticHandler, this,
                                /* RespectFilters */ true);
 }
+
+namespace {
+class LTODiagnosticInfo : public DiagnosticInfo {
+  const Twine &Msg;
+public:
+  LTODiagnosticInfo(const Twine &DiagMsg, DiagnosticSeverity Severity=DS_Error)
+      : DiagnosticInfo(DK_Linker, Severity), Msg(DiagMsg) {}
+  void print(DiagnosticPrinter &DP) const override { DP << Msg; }
+};
+}
+
+void LTOCodeGenerator::emitError(const std::string &ErrMsg) {
+  if (DiagHandler)
+    (*DiagHandler)(LTO_DS_ERROR, ErrMsg.c_str(), DiagContext);
+  else
+    Context.diagnose(LTODiagnosticInfo(ErrMsg));
+}

tools/llvm-lto/llvm-lto.cpp

if (i > 0)
       attrs.append(",");
     attrs.append(MAttrs[i]);
   }
 
   if (!attrs.empty())
     CodeGen.setAttr(attrs.c_str());
 
   if (!OutputFilename.empty()) {
-    std::string ErrorInfo;
     if (!CodeGen.optimize(DisableVerify, DisableInline, DisableGVNLoadPRE,
-                          DisableLTOVectorization, ErrorInfo)) {
-      errs() << argv[0] << ": error optimizing the code: " << ErrorInfo << "\n";
+                          DisableLTOVectorization)) {
+      // Diagnostic messages should have been printed by the handler.
+      errs() << argv[0] << ": error optimizing the code\n";
       return 1;
     }
 
     std::list<tool_output_file> OSs;
     std::vector<raw_pwrite_stream *> OSPtrs;
     for (unsigned I = 0; I != Parallelism; ++I) {
       std::string PartFilename = OutputFilename;
       if (Parallelism != 1)
         PartFilename += "." + utostr(I);
       std::error_code EC;
       OSs.emplace_back(PartFilename, EC, sys::fs::F_None);
       if (EC) {
         errs() << argv[0] << ": error opening the file '" << PartFilename
                << "': " << EC.message() << "\n";
         return 1;
       }
       OSPtrs.push_back(&OSs.back().os());
     }
 
-    if (!CodeGen.compileOptimized(OSPtrs, ErrorInfo)) {
-      errs() << argv[0] << ": error compiling the code: " << ErrorInfo << "\n";
+    if (!CodeGen.compileOptimized(OSPtrs)) {
+      // Diagnostic messages should have been printed by the handler.
+      errs() << argv[0] << ": error compiling the code\n";
       return 1;
     }
 
     for (tool_output_file &OS : OSs)
       OS.keep();
   } else {
     if (Parallelism != 1) {
       errs() << argv[0] << ": -j must be specified together with -o\n";
       return 1;
     }
 
-    std::string ErrorInfo;
     const char *OutputName = nullptr;
     if (!CodeGen.compile_to_file(&OutputName, DisableVerify, DisableInline,
-                                 DisableGVNLoadPRE, DisableLTOVectorization,
-                                 ErrorInfo)) {
-      errs() << argv[0]
-             << ": error compiling the code: " << ErrorInfo
-             << "\n";
+                                 DisableGVNLoadPRE, DisableLTOVectorization)) {
+      // Diagnostic messages should have been printed by the handler.
+      errs() << argv[0] << ": error compiling the code\n";
       return 1;
     }
 
     outs() << "Wrote native object file '" << OutputName << "'\n";
   }
 
   return 0;
 }

tools/lto/lto.cpp

if (!parsedOptions) {
     unwrap(cg)->parseCodeGenDebugOptions();
     lto_add_attrs(cg);
     parsedOptions = true;
   }
 }
 
 bool lto_codegen_write_merged_modules(lto_code_gen_t cg, const char *path) {
   maybeParseOptions(cg);
-  return !unwrap(cg)->writeMergedModules(path, sLastErrorString);
+  return !unwrap(cg)->writeMergedModules(path);
 }
 
 const void *lto_codegen_compile(lto_code_gen_t cg, size_t *length) {
   maybeParseOptions(cg);
   LibLTOCodeGenerator *CG = unwrap(cg);
   CG->NativeObjectFile =
       CG->compile(DisableVerify, DisableInline, DisableGVNLoadPRE,
-                  DisableLTOVectorization, sLastErrorString);
+                  DisableLTOVectorization);
   if (!CG->NativeObjectFile)
     return nullptr;
   *length = CG->NativeObjectFile->getBufferSize();
   return CG->NativeObjectFile->getBufferStart();
 }
 
 bool lto_codegen_optimize(lto_code_gen_t cg) {
   maybeParseOptions(cg);
   return !unwrap(cg)->optimize(DisableVerify, DisableInline, DisableGVNLoadPRE,
-                               DisableLTOVectorization, sLastErrorString);
+                               DisableLTOVectorization);
 }
 
 const void *lto_codegen_compile_optimized(lto_code_gen_t cg, size_t *length) {
   maybeParseOptions(cg);
   LibLTOCodeGenerator *CG = unwrap(cg);
-  CG->NativeObjectFile = CG->compileOptimized(sLastErrorString);
+  CG->NativeObjectFile = CG->compileOptimized();
   if (!CG->NativeObjectFile)
     return nullptr;
   *length = CG->NativeObjectFile->getBufferSize();
   return CG->NativeObjectFile->getBufferStart();
 }
 
 bool lto_codegen_compile_to_file(lto_code_gen_t cg, const char **name) {
   maybeParseOptions(cg);
   return !unwrap(cg)->compile_to_file(
       name, DisableVerify, DisableInline, DisableGVNLoadPRE,
-      DisableLTOVectorization, sLastErrorString);
+      DisableLTOVectorization);
 }
 
 void lto_codegen_debug_options(lto_code_gen_t cg, const char *opt) {
   unwrap(cg)->setCodeGenDebugOptions(opt);
 }
 
 unsigned int lto_api_version() { return LTO_API_VERSION; }