[llvm-driver] Reinvoke clang as described by llvm driver extra args

[abrachet]

rebase

[abrachet]

Rebase

[phosek]

I see the following pattern being duplicated:

if (ToolContext.IsDriver)
  Path = ToolContext.ReinvokeArgs[0];
...
if (ToolContext.ReinvokeArgs.size() > 1)
  TheDriver.setLLVMDriverPrependArg(ToolContext.ReinvokeArgs.back());

Could we split ReinvokeArgs into Path and PrependArgs (which would be nullptr if there are none)?

It'd be great if we could avoid IsDriver altogether to minimize the difference in behavior between the driver and non-driver case.

[abrachet]

In D137800#4038012, @phosek wrote:

I see the following pattern being duplicated:

if (ToolContext.IsDriver)
  Path = ToolContext.ReinvokeArgs[0];
...
if (ToolContext.ReinvokeArgs.size() > 1)
  TheDriver.setLLVMDriverPrependArg(ToolContext.ReinvokeArgs.back());

Could we split ReinvokeArgs into Path and PrependArgs (which would be nullptr if there are none)?

Sure.

It'd be great if we could avoid IsDriver altogether to minimize the difference in behavior between the driver and non-driver case.

Unfortunately, I had to still keep one IsDriver check in the GetExecutablePath so that it doesn't get the path of the executable in the driver case

[phosek]

clang/include/clang/Driver/Driver.h

295

I'd call it just PrependArg since it might have uses outside of LLVM driver.

clang/tools/driver/driver.cpp

489

Is there a reason why we couldn't support -canonical-prefixes for the driver case?

[abrachet]

Support -canoncial-prefixes

[phosek]

llvm/include/llvm/Support/LLVMDriver.h

26–28

This only has one caller right now so I'd inline it.

30

This is unused so I'd just remove it.

[abrachet]

Use correct diff

[NoQ]

The ccc-analyzer part looks good to me, thanks for taking care of this!

[phosek]

clang/include/clang/Driver/Driver.h

295

This wasn't addressed yet.

llvm/tools/llvm-driver/llvm-driver.cpp

68

Can we set PrependArg to nullptr in this case? It seems like we only ever use PrependArg if NeedsPrependArg is true in which case we may not even need NeedsPrependArg and could just check if PrependArg != null.

[abrachet]

clang/include/clang/Driver/Driver.h

295

Whoops, sorry about that.

llvm/tools/llvm-driver/llvm-driver.cpp

68

It's necessary to support -canonical-prefixes, such that when it finds the realpath of the current executable, it can correctly invoke the llvm binary with the PrependArg

[phosek]

LGTM

llvm/include/llvm/Support/LLVMDriver.h

18–24

This is just a suggestion so feel free to ignore this, but I was thinking about these names a bit more and I think that Name and NeedsName might be more self-descriptive?

clang/include/clang/Driver/Driver.h

 #include "clang/Driver/DriverDiagnostic.h"
 #include "clang/Driver/InputInfo.h"
 #include "clang/Driver/Options.h"
 #include "clang/Driver/Phases.h"
 #include "clang/Driver/ToolChain.h"
 #include "clang/Driver/Types.h"
 #include "clang/Driver/Util.h"
 #include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/STLFunctionalExtras.h"
 #include "llvm/ADT/StringMap.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/Option/Arg.h"
 #include "llvm/Option/ArgList.h"
 #include "llvm/Support/StringSaver.h"
 
 #include <list>
 #include <map>
public:
   /// Set CC_PRINT_PROC_STAT mode, which causes the driver to dump
   /// performance report to CC_PRINT_PROC_STAT_FILE or to stdout.
   unsigned CCPrintProcessStats : 1;
 
   /// Pointer to the ExecuteCC1Tool function, if available.
   /// When the clangDriver lib is used through clang.exe, this provides a
   /// shortcut for executing the -cc1 command-line directly, in the same
   /// process.
-  typedef int (*CC1ToolFunc)(SmallVectorImpl<const char *> &ArgV);
+  using CC1ToolFunc =
+      llvm::function_ref<int(SmallVectorImpl<const char *> &ArgV)>;
   CC1ToolFunc CC1Main = nullptr;
 
 private:
   /// Raw target triple.
   std::string TargetTriple;
 
   /// Name to use when invoking gcc/g++.
   std::string CCCGenericGCCName;
 
   /// Paths to configuration files used.
   std::vector<std::string> ConfigFiles;
 
   /// Allocator for string saver.
   llvm::BumpPtrAllocator Alloc;
 
   /// Object that stores strings read from configuration file.
   llvm::StringSaver Saver;
 
   /// Arguments originated from configuration file.
   std::unique_ptr<llvm::opt::InputArgList> CfgOptions;
 
   /// Arguments originated from command line.
   std::unique_ptr<llvm::opt::InputArgList> CLOptions;
 
+  /// If this is non-null, the driver will prepend this argument before
+  /// reinvoking clang. This is useful for the llvm-driver where clang's
+  /// realpath will be to the llvm binary and not clang, so it must pass
+  /// "clang" as it's first argument.
+  const char *PrependArg;
+
   /// Whether to check that input files exist when constructing compilation
   /// jobs.
   unsigned CheckInputsExist : 1;
   /// Whether to probe for PCH files on disk, in order to upgrade
   /// -include foo.h to -include-pch foo.h.pch.
   unsigned ProbePrecompiled : 1;
 
 public:
public:
 
   bool getCheckInputsExist() const { return CheckInputsExist; }
 
   void setCheckInputsExist(bool Value) { CheckInputsExist = Value; }
 
   bool getProbePrecompiled() const { return ProbePrecompiled; }
   void setProbePrecompiled(bool Value) { ProbePrecompiled = Value; }
 
+  const char *getPrependArg() const { return PrependArg; }
+  void setPrependArg(const char *Value) { PrependArg = Value; }
+
   void setTargetAndMode(const ParsedClangName &TM) { ClangNameParts = TM; }
 
   const std::string &getTitle() { return DriverTitle; }
   void setTitle(std::string Value) { DriverTitle = std::move(Value); }
 
   std::string getTargetTriple() const { return TargetTriple; }
 
   /// Get the path to the main clang executable.

clang/include/clang/Driver/Job.h

class Command {
   const Tool &Creator;
 
   /// Whether and how to generate response files if the arguments are too long.
   ResponseFileSupport ResponseSupport;
 
   /// The executable to run.
   const char *Executable;
 
+  /// Optional argument to prepend.
+  const char *PrependArg;
+
   /// The list of program arguments (not including the implicit first
   /// argument, which will be the executable).
   llvm::opt::ArgStringList Arguments;
 
   /// The list of program inputs.
   std::vector<InputInfo> InputInfoList;
 
   /// The list of program arguments which are outputs. May be empty.
public:
   bool PrintInputFilenames = false;
 
   /// Whether the command will be executed in this process or not.
   bool InProcess = false;
 
   Command(const Action &Source, const Tool &Creator,
           ResponseFileSupport ResponseSupport, const char *Executable,
           const llvm::opt::ArgStringList &Arguments, ArrayRef<InputInfo> Inputs,
-          ArrayRef<InputInfo> Outputs = std::nullopt);
+          ArrayRef<InputInfo> Outputs = std::nullopt,
+          const char *PrependArg = nullptr);
   // FIXME: This really shouldn't be copyable, but is currently copied in some
   // error handling in Driver::generateCompilationDiagnostics.
   Command(const Command &) = default;
   virtual ~Command() = default;
 
   virtual void Print(llvm::raw_ostream &OS, const char *Terminator, bool Quote,
                      CrashReportInfo *CrashInfo = nullptr) const;
 

 
 /// Use the CC1 tool callback when available, to avoid creating a new process
 class CC1Command : public Command {
 public:
   CC1Command(const Action &Source, const Tool &Creator,
              ResponseFileSupport ResponseSupport, const char *Executable,
              const llvm::opt::ArgStringList &Arguments,
              ArrayRef<InputInfo> Inputs,
-             ArrayRef<InputInfo> Outputs = std::nullopt);
+             ArrayRef<InputInfo> Outputs = std::nullopt,
+             const char *PrependArg = nullptr);
 
   void Print(llvm::raw_ostream &OS, const char *Terminator, bool Quote,
              CrashReportInfo *CrashInfo = nullptr) const override;
 
   int Execute(ArrayRef<std::optional<StringRef>> Redirects, std::string *ErrMsg,
               bool *ExecutionFailed) const override;
 
   void setEnvironment(llvm::ArrayRef<const char *> NewEnvironment) override;

clang/lib/Driver/Driver.cpp

Driver::Driver(StringRef ClangExecutable, StringRef TargetTriple,
     : Diags(Diags), VFS(std::move(VFS)), Mode(GCCMode),
       SaveTemps(SaveTempsNone), BitcodeEmbed(EmbedNone),
       Offload(OffloadHostDevice), CXX20HeaderType(HeaderMode_None),
       ModulesModeCXX20(false), LTOMode(LTOK_None),
       ClangExecutable(ClangExecutable), SysRoot(DEFAULT_SYSROOT),
       DriverTitle(Title), CCCPrintBindings(false), CCPrintOptions(false),
       CCLogDiagnostics(false), CCGenDiagnostics(false),
       CCPrintProcessStats(false), TargetTriple(TargetTriple), Saver(Alloc),
-      CheckInputsExist(true), ProbePrecompiled(true),
+      PrependArg(nullptr), CheckInputsExist(true), ProbePrecompiled(true),
       SuppressMissingInputWarning(false) {
   // Provide a sane fallback if no VFS is specified.
   if (!this->VFS)
     this->VFS = llvm::vfs::getRealFileSystem();
 
   Name = std::string(llvm::sys::path::filename(ClangExecutable));
   Dir = std::string(llvm::sys::path::parent_path(ClangExecutable));
   InstalledDir = Dir; // Provide a sensible default installed dir.

clang/lib/Driver/Job.cpp

 #include <utility>
 
 using namespace clang;
 using namespace driver;
 
 Command::Command(const Action &Source, const Tool &Creator,
                  ResponseFileSupport ResponseSupport, const char *Executable,
                  const llvm::opt::ArgStringList &Arguments,
-                 ArrayRef<InputInfo> Inputs, ArrayRef<InputInfo> Outputs)
+                 ArrayRef<InputInfo> Inputs, ArrayRef<InputInfo> Outputs,
+                 const char *PrependArg)
     : Source(Source), Creator(Creator), ResponseSupport(ResponseSupport),
-      Executable(Executable), Arguments(Arguments) {
+      Executable(Executable), PrependArg(PrependArg), Arguments(Arguments) {
   for (const auto &II : Inputs)
     if (II.isFilename())
       InputInfoList.push_back(II);
   for (const auto &II : Outputs)
     if (II.isFilename())
       OutputFilenames.push_back(II.getFilename());
 }
 
if (ResponseSupport.ResponseKind != ResponseFileSupport::RF_FileList) {
     Out.push_back(ResponseFileFlag.c_str());
     return;
   }
 
   llvm::StringSet<> Inputs;
   for (const auto *InputName : InputFileList)
     Inputs.insert(InputName);
   Out.push_back(Executable);
+
+  if (PrependArg)
+    Out.push_back(PrependArg);
+
   // In a file list, build args vector ignoring parameters that will go in the
   // response file (elements of the InputFileList vector)
   bool FirstInput = true;
   for (const auto *Arg : Arguments) {
     if (Inputs.count(Arg) == 0) {
       Out.push_back(Arg);
     } else if (FirstInput) {
       FirstInput = false;
void Command::Print(raw_ostream &OS, const char *Terminator, bool Quote,
   OS << ' ';
   llvm::sys::printArg(OS, Executable, /*Quote=*/true);
 
   ArrayRef<const char *> Args = Arguments;
   SmallVector<const char *, 128> ArgsRespFile;
   if (ResponseFile != nullptr) {
     buildArgvForResponseFile(ArgsRespFile);
     Args = ArrayRef<const char *>(ArgsRespFile).slice(1); // no executable name
+  } else if (PrependArg) {
+    OS << ' ';
+    llvm::sys::printArg(OS, PrependArg, /*Quote=*/true);
   }
 
   bool HaveCrashVFS = CrashInfo && !CrashInfo->VFSPath.empty();
   for (size_t i = 0, e = Args.size(); i < e; ++i) {
     const char *const Arg = Args[i];
 
     if (CrashInfo) {
       int NumArgs = 0;

 
 int Command::Execute(ArrayRef<std::optional<StringRef>> Redirects,
                      std::string *ErrMsg, bool *ExecutionFailed) const {
   PrintFileNames();
 
   SmallVector<const char *, 128> Argv;
   if (ResponseFile == nullptr) {
     Argv.push_back(Executable);
+    if (PrependArg)
+      Argv.push_back(PrependArg);
     Argv.append(Arguments.begin(), Arguments.end());
     Argv.push_back(nullptr);
   } else {
     // If the command is too large, we need to put arguments in a response file.
     std::string RespContents;
     llvm::raw_string_ostream SS(RespContents);
 
     // Write file contents and build the Argv vector
return llvm::sys::ExecuteAndWait(Executable, Args, Env, Redirects,
                                    /*secondsToWait*/ 0, /*memoryLimit*/ 0,
                                    ErrMsg, ExecutionFailed, &ProcStat);
 }
 
 CC1Command::CC1Command(const Action &Source, const Tool &Creator,
                        ResponseFileSupport ResponseSupport,
                        const char *Executable,
                        const llvm::opt::ArgStringList &Arguments,
-                       ArrayRef<InputInfo> Inputs, ArrayRef<InputInfo> Outputs)
+                       ArrayRef<InputInfo> Inputs, ArrayRef<InputInfo> Outputs,
+                       const char *PrependArg)
     : Command(Source, Creator, ResponseSupport, Executable, Arguments, Inputs,
-              Outputs) {
+              Outputs, PrependArg) {
   InProcess = true;
 }
 
 void CC1Command::Print(raw_ostream &OS, const char *Terminator, bool Quote,
                        CrashReportInfo *CrashInfo) const {
   if (InProcess)
     OS << " (in-process)\n";
   Command::Print(OS, Terminator, Quote, CrashInfo);

clang/lib/Driver/ToolChains/Clang.cpp

for (const auto &II : Inputs) {
       if (II.isFilename())
         CmdArgs.push_back(II.getFilename());
       else
         II.getInputArg().renderAsInput(Args, CmdArgs);
     }
 
     C.addCommand(std::make_unique<Command>(
         JA, *this, ResponseFileSupport::AtFileUTF8(), D.getClangProgramPath(),
-        CmdArgs, Inputs, Output));
+        CmdArgs, Inputs, Output, D.getPrependArg()));
     return;
   }
 
   if (C.getDriver().embedBitcodeMarkerOnly() && !IsUsingLTO)
     CmdArgs.push_back("-fembed-bitcode=marker");
 
   // We normally speed up the clang process a bit by skipping destructors at
   // exit, but when we're generating diagnostics we can rely on some of the
for (const InputInfo &Input : FrontendInputs) {
     if (Input.isFilename())
       CmdArgs.push_back(Input.getFilename());
     else
       Input.getInputArg().renderAsInput(Args, CmdArgs);
   }
 
   if (D.CC1Main && !D.CCGenDiagnostics) {
     // Invoke the CC1 directly in this process
-    C.addCommand(std::make_unique<CC1Command>(JA, *this,
-                                              ResponseFileSupport::AtFileUTF8(),
-                                              Exec, CmdArgs, Inputs, Output));
-  } else {
-    C.addCommand(std::make_unique<Command>(JA, *this,
-                                           ResponseFileSupport::AtFileUTF8(),
-                                           Exec, CmdArgs, Inputs, Output));
+    C.addCommand(std::make_unique<CC1Command>(
+        JA, *this, ResponseFileSupport::AtFileUTF8(), Exec, CmdArgs, Inputs,
+        Output, D.getPrependArg()));
+  } else {
+    C.addCommand(std::make_unique<Command>(
+        JA, *this, ResponseFileSupport::AtFileUTF8(), Exec, CmdArgs, Inputs,
+        Output, D.getPrependArg()));
   }
 
   // Make the compile command echo its inputs for /showFilenames.
   if (Output.getType() == types::TY_Object &&
       Args.hasFlag(options::OPT__SLASH_showFilenames,
                    options::OPT__SLASH_showFilenames_, false)) {
     C.getJobs().getJobs().back()->PrintInputFilenames = true;
   }
if (Triple.isAMDGPU())
     handleAMDGPUCodeObjectVersionOptions(D, Args, CmdArgs, /*IsCC1As=*/true);
 
   assert(Input.isFilename() && "Invalid input.");
   CmdArgs.push_back(Input.getFilename());
 
   const char *Exec = getToolChain().getDriver().getClangProgramPath();
   if (D.CC1Main && !D.CCGenDiagnostics) {
     // Invoke cc1as directly in this process.
-    C.addCommand(std::make_unique<CC1Command>(JA, *this,
-                                              ResponseFileSupport::AtFileUTF8(),
-                                              Exec, CmdArgs, Inputs, Output));
-  } else {
-    C.addCommand(std::make_unique<Command>(JA, *this,
-                                           ResponseFileSupport::AtFileUTF8(),
-                                           Exec, CmdArgs, Inputs, Output));
+    C.addCommand(std::make_unique<CC1Command>(
+        JA, *this, ResponseFileSupport::AtFileUTF8(), Exec, CmdArgs, Inputs,
+        Output, D.getPrependArg()));
+  } else {
+    C.addCommand(std::make_unique<Command>(
+        JA, *this, ResponseFileSupport::AtFileUTF8(), Exec, CmdArgs, Inputs,
+        Output, D.getPrependArg()));
   }
 }
 
 // Begin OffloadBundler
 
 void OffloadBundler::ConstructJob(Compilation &C, const JobAction &JA,
                                   const InputInfo &Output,
                                   const InputInfoList &Inputs,

clang/test/Analysis/scan-build/lit.local.cfg

 # -*- Python -*-
 
 import lit.util
 import lit.formats
 import os
 
 use_lit_shell = os.environ.get("LIT_USE_INTERNAL_SHELL")
 config.test_format = lit.formats.ShTest(use_lit_shell == "0")
 
+clang_path = config.clang if config.have_llvm_driver else os.path.realpath(config.clang)
+
 config.substitutions.append(('%scan-build',
                              '\'%s\' --use-analyzer=%s ' % (
                                  lit.util.which('scan-build',
                                                 os.path.join(
                                                     config.clang_src_dir,
                                                     'tools',
                                                     'scan-build',
                                                     'bin')),
-                                 os.path.realpath(config.clang))))
+                                 clang_path)))

clang/test/CodeGen/debug-info-codeview-buildinfo.c

 
 int main(void) { return 42; }
 
 // CHECK:                       Types (.debug$T)
 // CHECK: ============================================================
 // CHECK: 0x[[PWD:.+]] | LF_STRING_ID [size = {{.+}}] ID: <no type>, String: [[PWDVAL:.+]]
 // CHECK: 0x[[FILEPATH:.+]] | LF_STRING_ID [size = {{.+}}] ID: <no type>, String: [[FILEPATHVAL:.+[\\/]debug-info-codeview-buildinfo.c]]
 // CHECK: 0x[[ZIPDB:.+]] | LF_STRING_ID [size = {{.+}}] ID: <no type>, String:
-// CHECK: 0x[[TOOL:.+]] | LF_STRING_ID [size = {{.+}}] ID: <no type>, String: [[TOOLVAL:.+[\\/]clang.*]]
+// CHECK: 0x[[TOOL:.+]] | LF_STRING_ID [size = {{.+}}] ID: <no type>, String: [[TOOLVAL:.+[\\/][clang|llvm].*]]
 // CHECK: 0x[[CMDLINE:.+]] | LF_STRING_ID [size = {{.+}}] ID: <no type>, String: "-cc1
 // CHECK: 0x{{.+}} | LF_BUILDINFO [size = {{.+}}]
 // CHECK-NEXT:          0x[[PWD]]: `[[PWDVAL]]`
 // CHECK-NEXT:          0x[[TOOL]]: `[[TOOLVAL]]`
 // CHECK-NEXT:          0x[[FILEPATH]]: `[[FILEPATHVAL]]`
 // CHECK-NEXT:          0x[[ZIPDB]]: ``
 // CHECK-NEXT:          0x[[CMDLINE]]: `"-cc1
 

clang/test/Driver/check-time-trace.cpp

 // CHECK:      "args":
 // CHECK:      "detail":
 // CHECK:      "dur":
 // CHECK:      "name":
 // CHECK-NEXT: "ph":
 // CHECK-NEXT: "pid":
 // CHECK-NEXT: "tid":
 // CHECK-NEXT: "ts":
-// CHECK:      "name": "clang{{.*}}"
+// CHECK:      "name": "{{clang|llvm}}{{.*}}"
 // CHECK:      "name": "process_name"
 // CHECK:      "name": "thread_name"
 
 template <typename T>
 struct Struct {
   T Num;
 };
 
 int main() {
   Struct<int> S;
 
   return 0;
 }

clang/tools/driver/cc1gen_reproducer_main.cpp

 
 #include "clang/Basic/Diagnostic.h"
 #include "clang/Basic/LLVM.h"
 #include "clang/Driver/Compilation.h"
 #include "clang/Driver/Driver.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/Support/FileSystem.h"
+#include "llvm/Support/LLVMDriver.h"
 #include "llvm/Support/TargetSelect.h"
 #include "llvm/Support/VirtualFileSystem.h"
 #include "llvm/Support/YAMLTraits.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/TargetParser/Host.h"
 #include <optional>
 
 using namespace clang;
static std::string generateReproducerMetaInfo(const ClangInvocationInfo &Info) {
   if (Info.Dump)
     llvm::outs() << "REPRODUCER METAINFO: " << OS.str() << "\n";
   return std::move(OS.str());
 }
 
 /// Generates a reproducer for a set of arguments from a specific invocation.
 static std::optional<driver::Driver::CompilationDiagnosticReport>
 generateReproducerForInvocationArguments(ArrayRef<const char *> Argv,
-                                         const ClangInvocationInfo &Info) {
+                                         const ClangInvocationInfo &Info,
+                                         const llvm::ToolContext &ToolContext) {
   using namespace driver;
   auto TargetAndMode = ToolChain::getTargetAndModeFromProgramName(Argv[0]);
 
   IntrusiveRefCntPtr<DiagnosticOptions> DiagOpts = new DiagnosticOptions;
 
   IntrusiveRefCntPtr<DiagnosticIDs> DiagID(new DiagnosticIDs());
   DiagnosticsEngine Diags(DiagID, &*DiagOpts, new IgnoringDiagConsumer());
   ProcessWarningOptions(Diags, *DiagOpts, /*ReportDiags=*/false);
-  Driver TheDriver(Argv[0], llvm::sys::getDefaultTargetTriple(), Diags);
+  Driver TheDriver(ToolContext.Path, llvm::sys::getDefaultTargetTriple(),
+                   Diags);
   TheDriver.setTargetAndMode(TargetAndMode);
+  if (ToolContext.NeedsPrependArg)
+    TheDriver.setPrependArg(ToolContext.PrependArg);
 
   std::unique_ptr<Compilation> C(TheDriver.BuildCompilation(Argv));
   if (C && !C->containsError()) {
     for (const auto &J : C->getJobs()) {
       if (const Command *Cmd = dyn_cast<Command>(&J)) {
         Driver::CompilationDiagnosticReport Report;
         TheDriver.generateCompilationDiagnostics(
             *C, *Cmd, generateReproducerMetaInfo(Info), &Report);
for (const auto &File : llvm::enumerate(Report.TemporaryFiles)) {
     if (File.index())
       OS << ',';
     OS << '"' << File.value() << '"';
   }
   OS << "]\n}\n";
 }
 
 int cc1gen_reproducer_main(ArrayRef<const char *> Argv, const char *Argv0,
-                           void *MainAddr) {
+                           void *MainAddr,
+                           const llvm::ToolContext &ToolContext) {
   if (Argv.size() < 1) {
     llvm::errs() << "error: missing invocation file\n";
     return 1;
   }
   // Parse the invocation descriptor.
   StringRef Input = Argv[0];
   llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> Buffer =
       llvm::MemoryBuffer::getFile(Input, /*IsText=*/true);
int cc1gen_reproducer_main(ArrayRef<const char *> Argv, const char *Argv0,
 
   // Create an invocation that will produce the reproducer.
   std::vector<const char *> DriverArgs;
   for (const auto &Arg : InvocationInfo.Arguments)
     DriverArgs.push_back(Arg.c_str());
   std::string Path = GetExecutablePath(Argv0, /*CanonicalPrefixes=*/true);
   DriverArgs[0] = Path.c_str();
   std::optional<driver::Driver::CompilationDiagnosticReport> Report =
-      generateReproducerForInvocationArguments(DriverArgs, InvocationInfo);
+      generateReproducerForInvocationArguments(DriverArgs, InvocationInfo,
+                                               ToolContext);
 
   // Emit the information about the reproduce files to stdout.
   int Result = 1;
   if (Report) {
     printReproducerInformation(llvm::outs(), InvocationInfo, *Report);
     Result = 0;
   }
 
   // Remove the input file.
   llvm::sys::fs::remove(Input);
   return Result;
 }

clang/tools/driver/driver.cpp

static void ApplyQAOverride(SmallVectorImpl<const char*> &Args,
   }
 }
 
 extern int cc1_main(ArrayRef<const char *> Argv, const char *Argv0,
                     void *MainAddr);
 extern int cc1as_main(ArrayRef<const char *> Argv, const char *Argv0,
                       void *MainAddr);
 extern int cc1gen_reproducer_main(ArrayRef<const char *> Argv,
-                                  const char *Argv0, void *MainAddr);
+                                  const char *Argv0, void *MainAddr,
+                                  const llvm::ToolContext &);
 
 static void insertTargetAndModeArgs(const ParsedClangName &NameParts,
                                     SmallVectorImpl<const char *> &ArgVector,
                                     std::set<std::string> &SavedStrings) {
   // Put target and mode arguments at the start of argument list so that
   // arguments specified in command line could override them. Avoid putting
   // them at index 0, as an option like '-cc1' must remain the first.
   int InsertionPoint = 0;
static void SetInstallDir(SmallVectorImpl<const char *> &argv,
   if (CanonicalPrefixes)
     llvm::sys::fs::make_absolute(InstalledPath);
 
   StringRef InstalledPathParent(llvm::sys::path::parent_path(InstalledPath));
   if (llvm::sys::fs::exists(InstalledPathParent))
     TheDriver.setInstalledDir(InstalledPathParent);
 }
 
-static int ExecuteCC1Tool(SmallVectorImpl<const char *> &ArgV) {
+static int ExecuteCC1Tool(SmallVectorImpl<const char *> &ArgV,
+                          const llvm::ToolContext &ToolContext) {
   // If we call the cc1 tool from the clangDriver library (through
   // Driver::CC1Main), we need to clean up the options usage count. The options
   // are currently global, and they might have been used previously by the
   // driver.
   llvm::cl::ResetAllOptionOccurrences();
 
   llvm::BumpPtrAllocator A;
   llvm::cl::ExpansionContext ECtx(A, llvm::cl::TokenizeGNUCommandLine);
   if (llvm::Error Err = ECtx.expandResponseFiles(ArgV)) {
     llvm::errs() << toString(std::move(Err)) << '\n';
     return 1;
   }
   StringRef Tool = ArgV[1];
   void *GetExecutablePathVP = (void *)(intptr_t)GetExecutablePath;
   if (Tool == "-cc1")
     return cc1_main(ArrayRef(ArgV).slice(1), ArgV[0], GetExecutablePathVP);
   if (Tool == "-cc1as")
     return cc1as_main(ArrayRef(ArgV).slice(2), ArgV[0], GetExecutablePathVP);
   if (Tool == "-cc1gen-reproducer")
     return cc1gen_reproducer_main(ArrayRef(ArgV).slice(2), ArgV[0],
-                                  GetExecutablePathVP);
+                                  GetExecutablePathVP, ToolContext);
   // Reject unknown tools.
   llvm::errs() << "error: unknown integrated tool '" << Tool << "'. "
                << "Valid tools include '-cc1' and '-cc1as'.\n";
   return 1;
 }
 
-int clang_main(int Argc, char **Argv, const llvm::ToolContext &) {
+int clang_main(int Argc, char **Argv, const llvm::ToolContext &ToolContext) {
   noteBottomOfStack();
   llvm::InitLLVM X(Argc, Argv);
   llvm::setBugReportMsg("PLEASE submit a bug report to " BUG_REPORT_URL
                         " and include the crash backtrace, preprocessed "
                         "source, and associated run script.\n");
   SmallVector<const char *, 256> Args(Argv, Argv + Argc);
 
   if (llvm::sys::Process::FixupStandardFileDescriptors())
     return 1;
 
   llvm::InitializeAllTargets();
 
   llvm::BumpPtrAllocator A;
   llvm::StringSaver Saver(A);
 
+  const char *ProgName =
+      ToolContext.NeedsPrependArg ? ToolContext.PrependArg : ToolContext.Path;
+
   // Parse response files using the GNU syntax, unless we're in CL mode. There
-  // are two ways to put clang in CL compatibility mode: Args[0] is either
+  // are two ways to put clang in CL compatibility mode: ProgName is either
   // clang-cl or cl, or --driver-mode=cl is on the command line. The normal
   // command line parsing can't happen until after response file parsing, so we
   // have to manually search for a --driver-mode=cl argument the hard way.
   // Finally, our -cc1 tools don't care which tokenization mode we use because
   // response files written by clang will tokenize the same way in either mode.
   bool ClangCLMode =
-      IsClangCL(getDriverMode(Args[0], llvm::ArrayRef(Args).slice(1)));
+      IsClangCL(getDriverMode(ProgName, llvm::ArrayRef(Args).slice(1)));
   enum { Default, POSIX, Windows } RSPQuoting = Default;
   for (const char *F : Args) {
     if (strcmp(F, "--rsp-quoting=posix") == 0)
       RSPQuoting = POSIX;
     else if (strcmp(F, "--rsp-quoting=windows") == 0)
       RSPQuoting = Windows;
   }
 
int clang_main(int Argc, char **Argv, const llvm::ToolContext &ToolContext) {
   auto FirstArg = llvm::find_if(llvm::drop_begin(Args),
                                 [](const char *A) { return A != nullptr; });
   if (FirstArg != Args.end() && StringRef(*FirstArg).startswith("-cc1")) {
     // If -cc1 came from a response file, remove the EOL sentinels.
     if (MarkEOLs) {
       auto newEnd = std::remove(Args.begin(), Args.end(), nullptr);
       Args.resize(newEnd - Args.begin());
     }
-    return ExecuteCC1Tool(Args);
+    return ExecuteCC1Tool(Args, ToolContext);
   }
 
   // Handle options that need handling before the real command line parsing in
   // Driver::BuildCompilation()
   bool CanonicalPrefixes = true;
   for (int i = 1, size = Args.size(); i < size; ++i) {
     // Skip end-of-line response file markers
     if (Args[i] == nullptr)
int clang_main(int Argc, char **Argv, const llvm::ToolContext &ToolContext) {
   std::set<std::string> SavedStrings;
   // Handle CCC_OVERRIDE_OPTIONS, used for editing a command line behind the
   // scenes.
   if (const char *OverrideStr = ::getenv("CCC_OVERRIDE_OPTIONS")) {
     // FIXME: Driver shouldn't take extra initial argument.
     ApplyQAOverride(Args, OverrideStr, SavedStrings);
   }
 
-  std::string Path = GetExecutablePath(Args[0], CanonicalPrefixes);
+  std::string Path = GetExecutablePath(ToolContext.Path, CanonicalPrefixes);
 
   // Whether the cc1 tool should be called inside the current process, or if we
   // should spawn a new clang subprocess (old behavior).
   // Not having an additional process saves some execution time of Windows,
   // and makes debugging and profiling easier.
   bool UseNewCC1Process = CLANG_SPAWN_CC1;
   for (const char *Arg : Args)
     UseNewCC1Process = llvm::StringSwitch<bool>(Arg)
                            .Case("-fno-integrated-cc1", true)
                            .Case("-fintegrated-cc1", false)
                            .Default(UseNewCC1Process);
 
   IntrusiveRefCntPtr<DiagnosticOptions> DiagOpts =
       CreateAndPopulateDiagOpts(Args);
 
   TextDiagnosticPrinter *DiagClient
     = new TextDiagnosticPrinter(llvm::errs(), &*DiagOpts);
-  FixupDiagPrefixExeName(DiagClient, Path);
+  FixupDiagPrefixExeName(DiagClient, ProgName);
 
   IntrusiveRefCntPtr<DiagnosticIDs> DiagID(new DiagnosticIDs());
 
   DiagnosticsEngine Diags(DiagID, &*DiagOpts, DiagClient);
 
   if (!DiagOpts->DiagnosticSerializationFile.empty()) {
     auto SerializedConsumer =
         clang::serialized_diags::create(DiagOpts->DiagnosticSerializationFile,
                                         &*DiagOpts, /*MergeChildRecords=*/true);
     Diags.setClient(new ChainedDiagnosticConsumer(
         Diags.takeClient(), std::move(SerializedConsumer)));
   }
 
   ProcessWarningOptions(Diags, *DiagOpts, /*ReportDiags=*/false);
 
   Driver TheDriver(Path, llvm::sys::getDefaultTargetTriple(), Diags);
   SetInstallDir(Args, TheDriver, CanonicalPrefixes);
-  auto TargetAndMode = ToolChain::getTargetAndModeFromProgramName(Args[0]);
+  auto TargetAndMode = ToolChain::getTargetAndModeFromProgramName(ProgName);
   TheDriver.setTargetAndMode(TargetAndMode);
+  // If -canonical-prefixes is set, GetExecutablePath will have resolved Path
+  // to the llvm driver binary, not clang. In this case, we need to use
+  // PrependArg which should be clang-*. Checking just CanonicalPrefixes is
+  // safe even in the normal case because PrependArg will be null so
+  // setPrependArg will be a no-op.
+  if (ToolContext.NeedsPrependArg || CanonicalPrefixes)
+    TheDriver.setPrependArg(ToolContext.PrependArg);
 
   insertTargetAndModeArgs(TargetAndMode, Args, SavedStrings);
 
   if (!SetBackdoorDriverOutputsFromEnvVars(TheDriver))
     return 1;
 
   if (!UseNewCC1Process) {
-    TheDriver.CC1Main = &ExecuteCC1Tool;
+    TheDriver.CC1Main = [&ToolContext](SmallVectorImpl<const char *> &ArgV) {
+      return ExecuteCC1Tool(ArgV, ToolContext);
+    };
     // Ensure the CC1Command actually catches cc1 crashes
     llvm::CrashRecoveryContext::Enable();
   }
 
   std::unique_ptr<Compilation> C(TheDriver.BuildCompilation(Args));
 
   Driver::ReproLevel ReproLevel = Driver::ReproLevel::OnCrash;
   if (Arg *A = C->getArgs().getLastArg(options::OPT_gen_reproducer_eq)) {

clang/tools/scan-build/libexec/ccc-analyzer

while (<$OutputStream>) {
     next if (!/\s"?-cc1"?\s/);
     $line = $_;
   }
   die "could not find clang line\n" if (!defined $line);
   # Strip leading and trailing whitespace characters.
   $line =~ s/^\s+|\s+$//g;
   my @items = quotewords('\s+', 0, $line);
   my $cmd = shift @items;
-  die "cannot find 'clang' in 'clang' command\n" if (!($cmd =~ /clang/));
+  die "cannot find 'clang' in 'clang' command\n" if (!($cmd =~ /clang/ || basename($cmd) =~ /llvm/));
+  # If this is the llvm-driver the internal command will look like "llvm clang ...".
+  # Later this will be invoked like "clang clang ...", so skip over it.
+  if (basename($cmd) =~ /llvm/) {
+    die "Expected first arg to llvm driver to be 'clang'" if $items[0] ne "clang";
+    shift @items;
+  }
   return \@items;
 }
 
 sub Analyze {
   my ($Clang, $OriginalArgs, $AnalyzeArgs, $Lang, $Output, $Verbose, $HtmlDir,
       $file) = @_;
 
   my @Args = @$OriginalArgs;

llvm/cmake/modules/llvm-driver-template.cpp.in

 //===-- driver-template.cpp -----------------------------------------------===//
 //
 // 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
 //
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Support/LLVMDriver.h"
 #include "llvm/ADT/ArrayRef.h"
 
 int @TOOL_NAME@_main(int argc, char **, const llvm::ToolContext &);
 
-int main(int argc, char **argv) { return @TOOL_NAME@_main(argc, argv, {}); }
+int main(int argc, char **argv) {
+  return @TOOL_NAME@_main(argc, argv, {argv[0], nullptr, false});
+}

llvm/include/llvm/Support/LLVMDriver.h

 //===- LLVMDriver.h ---------------------------------------------*- 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
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_SUPPORT_LLVMDRIVER_H
 #define LLVM_SUPPORT_LLVMDRIVER_H
 
+#include "llvm/ADT/SmallVector.h"
+
 namespace llvm {
 
-struct ToolContext {};
+struct ToolContext {
+  const char *Path;
+  const char *PrependArg;
+  // PrependArg will be added unconditionally by the llvm-driver, but
+  // NeedsPrependArg will be false if Path is adequate to reinvoke the tool.
+  // This is useful if realpath is ever called on Path, in which case it will
+  // point to the llvm-driver executable, where PrependArg will be needed to
+  // invoke the correct tool.
+  bool NeedsPrependArg;
+};
 
 } // namespace llvm
 
 #endif

llvm/lib/Support/Path.cpp

 // Include the truly platform-specific parts.
 #if defined(LLVM_ON_UNIX)
 #include "Unix/Path.inc"
 #endif
 #if defined(_WIN32)
 #include "Windows/Path.inc"
 #endif
 
-bool IsLLVMDriver = false;
-
 namespace llvm {
 namespace sys {
 namespace fs {
 
-std::string getMainExecutable(const char *Argv0, void *MainAddr) {
-  if (IsLLVMDriver)
-    return sys::path::stem(Argv0).str();
-  return getMainExecutableImpl(Argv0, MainAddr);
-}
-
 TempFile::TempFile(StringRef Name, int FD)
     : TmpName(std::string(Name)), FD(FD) {}
 TempFile::TempFile(TempFile &&Other) { *this = std::move(Other); }
 TempFile &TempFile::operator=(TempFile &&Other) {
   TmpName = std::move(Other.TmpName);
   FD = Other.FD;
   Other.Done = true;
   Other.FD = -1;

llvm/lib/Support/Unix/Path.inc

static char *getprogpath(char ret[PATH_MAX], const char *bin) {
   }
   free(s);
   return nullptr;
 }
 #endif // __FreeBSD__ || __NetBSD__ || __FreeBSD_kernel__
 
 /// GetMainExecutable - Return the path to the main executable, given the
 /// value of argv[0] from program startup.
-std::string getMainExecutableImpl(const char *argv0, void *MainAddr) {
+std::string getMainExecutable(const char *argv0, void *MainAddr) {
 #if defined(__APPLE__)
   // On OS X the executable path is saved to the stack by dyld. Reading it
   // from there is much faster than calling dladdr, especially for large
   // binaries with symbols.
   char exe_path[PATH_MAX];
   uint32_t size = sizeof(exe_path);
   if (_NSGetExecutablePath(exe_path, &size) == 0) {
     char link_path[PATH_MAX];

llvm/lib/Support/Windows/Path.inc

 }
 
 } // end namespace windows
 
 namespace fs {
 
 const file_t kInvalidFile = INVALID_HANDLE_VALUE;
 
-std::string getMainExecutableImpl(const char *argv0, void *MainExecAddr) {
+std::string getMainExecutable(const char *argv0, void *MainExecAddr) {
   SmallVector<wchar_t, MAX_PATH> PathName;
   PathName.resize_for_overwrite(PathName.capacity());
   DWORD Size = ::GetModuleFileNameW(NULL, PathName.data(), PathName.size());
 
   // A zero return value indicates a failure other than insufficient space.
   if (Size == 0)
     return "";
 

llvm/tools/llvm-driver/llvm-driver.cpp

llvm::outs() << "OVERVIEW: llvm toolchain driver\n\n"
                << "USAGE: llvm [subcommand] [options]\n\n"
                << "SUBCOMMANDS:\n\n"
                << subcommands
                << "\n  Type \"llvm <subcommand> --help\" to get more help on a "
                   "specific subcommand\n\n"
                << "OPTIONS:\n\n  --help - Display this message";
 }
 
-static int findTool(int Argc, char **Argv) {
+static int findTool(int Argc, char **Argv, const char *Argv0) {
   if (!Argc) {
     printHelpMessage();
     return 1;
   }
 
   StringRef ToolName = Argv[0];
 
   if (ToolName == "--help") {
auto IsImpl = [=](StringRef Stem) {
                                       !llvm::isAlnum(Stem[I + Tool.size()]));
     };
     for (StringRef S : {Stem, sys::path::filename(ToolName)})
       if (IsImpl(S))
         return true;
     return false;
   };
 
+  auto MakeDriverArgs = [=]() -> llvm::ToolContext {
+    if (ToolName != Argv0)
+      return {Argv0, ToolName.data(), true};
+    return {Argv0, sys::path::filename(Argv0).data(), false};
+  };
+
 #define LLVM_DRIVER_TOOL(tool, entry)                                          \
   if (Is(tool))                                                                \
-    return entry##_main(Argc, Argv, {});
+    return entry##_main(Argc, Argv, MakeDriverArgs());
 #include "LLVMDriverTools.def"
 
   if (Is("llvm"))
-    return findTool(Argc - 1, Argv + 1);
+    return findTool(Argc - 1, Argv + 1, Argv0);
 
   printHelpMessage();
   return 1;
 }
 
-extern bool IsLLVMDriver;
-
-int main(int Argc, char **Argv) {
-  IsLLVMDriver = true;
-  return findTool(Argc, Argv);
-}
+int main(int Argc, char **Argv) { return findTool(Argc, Argv, Argv[0]); }