[clang][flang][mlir] Support -frecord-command-line option (#102975)

Add support for the -frecord-command-line option that will produce the llvm.commandline metadata which will eventually be saved in the object file. This behavior is also supported in clang. Some refactoring of the code in flang to handle these command line options was carried out. The corresponding -grecord-command-line option which saves the command line in the debug information has not yet been enabled for flang.
llvm · Sep 20, 2024 · b3533a1 · b3533a1
1 parent 594efd2
commit b3533a1
Show file tree

Hide file tree

Showing 23 changed files with 245 additions and 59 deletions.
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
@@ -1994,16 +1994,18 @@ def fparse_all_comments : Flag<["-"], "fparse-all-comments">, Group<f_clang_Grou
   Visibility<[ClangOption, CC1Option]>,
   MarshallingInfoFlag<LangOpts<"CommentOpts.ParseAllComments">>;
 def frecord_command_line : Flag<["-"], "frecord-command-line">,
-  DocBrief<[{Generate a section named ".GCC.command.line" containing the clang
+  DocBrief<[{Generate a section named ".GCC.command.line" containing the
 driver command-line. After linking, the section may contain multiple command
 lines, which will be individually terminated by null bytes. Separate arguments
 within a command line are combined with spaces; spaces and backslashes within an
 argument are escaped with backslashes. This format differs from the format of
 the equivalent section produced by GCC with the -frecord-gcc-switches flag.
 This option is currently only supported on ELF targets.}]>,
-  Group<f_clang_Group>;
+  Group<f_Group>,
+  Visibility<[ClangOption, CLOption, DXCOption, FlangOption]>;
 def fno_record_command_line : Flag<["-"], "fno-record-command-line">,
-  Group<f_clang_Group>;
+  Group<f_Group>,
+  Visibility<[ClangOption, CLOption, DXCOption, FlangOption]>;
 def : Flag<["-"], "frecord-gcc-switches">, Alias<frecord_command_line>;
 def : Flag<["-"], "fno-record-gcc-switches">, Alias<fno_record_command_line>;
 def fcommon : Flag<["-"], "fcommon">, Group<f_Group>,
@@ -7141,6 +7143,9 @@ def mrelocation_model : Separate<["-"], "mrelocation-model">,
   NormalizedValues<["Static", "PIC_", "ROPI", "RWPI", "ROPI_RWPI", "DynamicNoPIC"]>,
   MarshallingInfoEnum<CodeGenOpts<"RelocationModel">, "PIC_">;
 def debug_info_kind_EQ : Joined<["-"], "debug-info-kind=">;
+def record_command_line : Separate<["-"], "record-command-line">,
+  HelpText<"The string to embed in the .LLVM.command.line section.">,
+  MarshallingInfoString<CodeGenOpts<"RecordCommandLine">>;
 
 } // let Visibility = [CC1Option, CC1AsOption, FC1Option]
 
@@ -7161,9 +7166,6 @@ def debugger_tuning_EQ : Joined<["-"], "debugger-tuning=">,
 def dwarf_debug_flags : Separate<["-"], "dwarf-debug-flags">,
   HelpText<"The string to embed in the Dwarf debug flags record.">,
   MarshallingInfoString<CodeGenOpts<"DwarfDebugFlags">>;
-def record_command_line : Separate<["-"], "record-command-line">,
-  HelpText<"The string to embed in the .LLVM.command.line section.">,
-  MarshallingInfoString<CodeGenOpts<"RecordCommandLine">>;
 def compress_debug_sections_EQ : Joined<["-", "--"], "compress-debug-sections=">,
     HelpText<"DWARF debug sections compression type">, Values<"none,zlib,zstd">,
     NormalizedValuesScope<"llvm::DebugCompressionType">, NormalizedValues<["None", "Zlib", "Zstd"]>,

diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -94,24 +94,6 @@ static void CheckCodeGenerationOptions(const Driver &D, const ArgList &Args) {
                                                       << "-static";
 }
 
-// Add backslashes to escape spaces and other backslashes.
-// This is used for the space-separated argument list specified with
-// the -dwarf-debug-flags option.
-static void EscapeSpacesAndBackslashes(const char *Arg,
-                                       SmallVectorImpl<char> &Res) {
-  for (; *Arg; ++Arg) {
-    switch (*Arg) {
-    default:
-      break;
-    case ' ':
-    case '\\':
-      Res.push_back('\\');
-      break;
-    }
-    Res.push_back(*Arg);
-  }
-}
-
 /// Apply \a Work on the current tool chain \a RegularToolChain and any other
 /// offloading tool chain that is associated with the current action \a JA.
 static void
@@ -7702,31 +7684,10 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
   // Also record command line arguments into the debug info if
   // -grecord-gcc-switches options is set on.
   // By default, -gno-record-gcc-switches is set on and no recording.
-  auto GRecordSwitches =
-      Args.hasFlag(options::OPT_grecord_command_line,
-                   options::OPT_gno_record_command_line, false);
-  auto FRecordSwitches =
-      Args.hasFlag(options::OPT_frecord_command_line,
-                   options::OPT_fno_record_command_line, false);
-  if (FRecordSwitches && !Triple.isOSBinFormatELF() &&
-      !Triple.isOSBinFormatXCOFF() && !Triple.isOSBinFormatMachO())
-    D.Diag(diag::err_drv_unsupported_opt_for_target)
-        << Args.getLastArg(options::OPT_frecord_command_line)->getAsString(Args)
-        << TripleStr;
-  if (TC.UseDwarfDebugFlags() || GRecordSwitches || FRecordSwitches) {
-    ArgStringList OriginalArgs;
-    for (const auto &Arg : Args)
-      Arg->render(Args, OriginalArgs);
-
-    SmallString<256> Flags;
-    EscapeSpacesAndBackslashes(Exec, Flags);
-    for (const char *OriginalArg : OriginalArgs) {
-      SmallString<128> EscapedArg;
-      EscapeSpacesAndBackslashes(OriginalArg, EscapedArg);
-      Flags += " ";
-      Flags += EscapedArg;
-    }
-    auto FlagsArgString = Args.MakeArgString(Flags);
+  auto GRecordSwitches = false;
+  auto FRecordSwitches = false;
+  if (shouldRecordCommandLine(TC, Args, FRecordSwitches, GRecordSwitches)) {
+    auto FlagsArgString = renderEscapedCommandLine(TC, Args);
     if (TC.UseDwarfDebugFlags() || GRecordSwitches) {
       CmdArgs.push_back("-dwarf-debug-flags");
       CmdArgs.push_back(FlagsArgString);
@@ -8726,10 +8687,10 @@ void ClangAs::ConstructJob(Compilation &C, const JobAction &JA,
 
     SmallString<256> Flags;
     const char *Exec = getToolChain().getDriver().getClangProgramPath();
-    EscapeSpacesAndBackslashes(Exec, Flags);
+    escapeSpacesAndBackslashes(Exec, Flags);
     for (const char *OriginalArg : OriginalArgs) {
       SmallString<128> EscapedArg;
-      EscapeSpacesAndBackslashes(OriginalArg, EscapedArg);
+      escapeSpacesAndBackslashes(OriginalArg, EscapedArg);
       Flags += " ";
       Flags += EscapedArg;
     }

diff --git a/clang/lib/Driver/ToolChains/CommonArgs.cpp b/clang/lib/Driver/ToolChains/CommonArgs.cpp
@@ -2960,3 +2960,62 @@ void tools::addMCModel(const Driver &D, const llvm::opt::ArgList &Args,
     }
   }
 }
+
+void tools::escapeSpacesAndBackslashes(const char *Arg,
+                                       llvm::SmallVectorImpl<char> &Res) {
+  for (; *Arg; ++Arg) {
+    switch (*Arg) {
+    default:
+      break;
+    case ' ':
+    case '\\':
+      Res.push_back('\\');
+      break;
+    }
+    Res.push_back(*Arg);
+  }
+}
+
+const char *tools::renderEscapedCommandLine(const ToolChain &TC,
+                                            const llvm::opt::ArgList &Args) {
+  const Driver &D = TC.getDriver();
+  const char *Exec = D.getClangProgramPath();
+
+  llvm::opt::ArgStringList OriginalArgs;
+  for (const auto &Arg : Args)
+    Arg->render(Args, OriginalArgs);
+
+  llvm::SmallString<256> Flags;
+  escapeSpacesAndBackslashes(Exec, Flags);
+  for (const char *OriginalArg : OriginalArgs) {
+    llvm::SmallString<128> EscapedArg;
+    escapeSpacesAndBackslashes(OriginalArg, EscapedArg);
+    Flags += " ";
+    Flags += EscapedArg;
+  }
+
+  return Args.MakeArgString(Flags);
+}
+
+bool tools::shouldRecordCommandLine(const ToolChain &TC,
+                                    const llvm::opt::ArgList &Args,
+                                    bool &FRecordCommandLine,
+                                    bool &GRecordCommandLine) {
+  const Driver &D = TC.getDriver();
+  const llvm::Triple &Triple = TC.getEffectiveTriple();
+  const std::string &TripleStr = Triple.getTriple();
+
+  FRecordCommandLine =
+      Args.hasFlag(options::OPT_frecord_command_line,
+                   options::OPT_fno_record_command_line, false);
+  GRecordCommandLine =
+      Args.hasFlag(options::OPT_grecord_command_line,
+                   options::OPT_gno_record_command_line, false);
+  if (FRecordCommandLine && !Triple.isOSBinFormatELF() &&
+      !Triple.isOSBinFormatXCOFF() && !Triple.isOSBinFormatMachO())
+    D.Diag(diag::err_drv_unsupported_opt_for_target)
+        << Args.getLastArg(options::OPT_frecord_command_line)->getAsString(Args)
+        << TripleStr;
+
+  return FRecordCommandLine || TC.UseDwarfDebugFlags() || GRecordCommandLine;
+}
diff --git a/clang/lib/Driver/ToolChains/CommonArgs.h b/clang/lib/Driver/ToolChains/CommonArgs.h
@@ -233,6 +233,31 @@ void addMCModel(const Driver &D, const llvm::opt::ArgList &Args,
                 const llvm::Reloc::Model &RelocationModel,
                 llvm::opt::ArgStringList &CmdArgs);
 
+/// Add backslashes to escape spaces and other backslashes.
+/// This is used for the space-separated argument list specified with
+/// the -dwarf-debug-flags option.
+void escapeSpacesAndBackslashes(const char *Arg,
+                                llvm::SmallVectorImpl<char> &Res);
+
+/// Join the args in the given ArgList, escape spaces and backslashes and
+/// return the joined string. This is used when saving the command line as a
+/// result of using either the -frecord-command-line or -grecord-command-line
+/// options. The lifetime of the returned c-string will match that of the Args
+/// argument.
+const char *renderEscapedCommandLine(const ToolChain &TC,
+                                     const llvm::opt::ArgList &Args);
+
+/// Check if the command line should be recorded in the object file. This is
+/// done if either -frecord-command-line or -grecord-command-line options have
+/// been passed. This also does some error checking since -frecord-command-line
+/// is currently only supported on ELF platforms. The last two boolean
+/// arguments are out parameters and will be set depending on the command
+/// line options that were passed.
+bool shouldRecordCommandLine(const ToolChain &TC,
+                             const llvm::opt::ArgList &Args,
+                             bool &FRecordCommandLine,
+                             bool &GRecordCommandLine);
+
 } // end namespace tools
 } // end namespace driver
 } // end namespace clang

diff --git a/clang/lib/Driver/ToolChains/Flang.cpp b/clang/lib/Driver/ToolChains/Flang.cpp
@@ -885,6 +885,20 @@ void Flang::ConstructJob(Compilation &C, const JobAction &JA,
 
   addDashXForInput(Args, Input, CmdArgs);
 
+  bool FRecordCmdLine = false;
+  bool GRecordCmdLine = false;
+  if (shouldRecordCommandLine(TC, Args, FRecordCmdLine, GRecordCmdLine)) {
+    const char *CmdLine = renderEscapedCommandLine(TC, Args);
+    if (FRecordCmdLine) {
+      CmdArgs.push_back("-record-command-line");
+      CmdArgs.push_back(CmdLine);
+    }
+    if (TC.UseDwarfDebugFlags() || GRecordCmdLine) {
+      CmdArgs.push_back("-dwarf-debug-flags");
+      CmdArgs.push_back(CmdLine);
+    }
+  }
+
   CmdArgs.push_back(Input.getFilename());
 
   // TODO: Replace flang-new with flang once the new driver replaces the

diff --git a/flang/include/flang/Frontend/CodeGenOptions.h b/flang/include/flang/Frontend/CodeGenOptions.h
@@ -63,6 +63,9 @@ class CodeGenOptions : public CodeGenOptionsBase {
   /// The directory where temp files are stored if specified by -save-temps
   std::optional<std::string> SaveTempsDir;
 
+  /// The string containing the commandline for the llvm.commandline metadata.
+  std::optional<std::string> RecordCommandLine;
+
   /// The name of the file to which the backend should save YAML optimization
   /// records.
   std::string OptRecordFile;

diff --git a/flang/include/flang/Lower/Bridge.h b/flang/include/flang/Lower/Bridge.h
@@ -14,6 +14,8 @@
 #define FORTRAN_LOWER_BRIDGE_H
 
 #include "flang/Common/Fortran.h"
+#include "flang/Frontend/CodeGenOptions.h"
+#include "flang/Frontend/TargetOptions.h"
 #include "flang/Lower/AbstractConverter.h"
 #include "flang/Lower/EnvironmentDefault.h"
 #include "flang/Lower/LoweringOptions.h"
@@ -65,11 +67,13 @@ class LoweringBridge {
          const Fortran::lower::LoweringOptions &loweringOptions,
          const std::vector<Fortran::lower::EnvironmentDefault> &envDefaults,
          const Fortran::common::LanguageFeatureControl &languageFeatures,
-         const llvm::TargetMachine &targetMachine, llvm::StringRef tuneCPU) {
+         const llvm::TargetMachine &targetMachine,
+         const Fortran::frontend::TargetOptions &targetOptions,
+         const Fortran::frontend::CodeGenOptions &codeGenOptions) {
     return LoweringBridge(ctx, semanticsContext, defaultKinds, intrinsics,
                           targetCharacteristics, allCooked, triple, kindMap,
                           loweringOptions, envDefaults, languageFeatures,
-                          targetMachine, tuneCPU);
+                          targetMachine, targetOptions, codeGenOptions);
   }
 
   //===--------------------------------------------------------------------===//
@@ -148,7 +152,9 @@ class LoweringBridge {
       const Fortran::lower::LoweringOptions &loweringOptions,
       const std::vector<Fortran::lower::EnvironmentDefault> &envDefaults,
       const Fortran::common::LanguageFeatureControl &languageFeatures,
-      const llvm::TargetMachine &targetMachine, const llvm::StringRef tuneCPU);
+      const llvm::TargetMachine &targetMachine,
+      const Fortran::frontend::TargetOptions &targetOptions,
+      const Fortran::frontend::CodeGenOptions &codeGenOptions);
   LoweringBridge() = delete;
   LoweringBridge(const LoweringBridge &) = delete;
 

diff --git a/flang/include/flang/Optimizer/Dialect/Support/FIRContext.h b/flang/include/flang/Optimizer/Dialect/Support/FIRContext.h
@@ -77,6 +77,12 @@ void setIdent(mlir::ModuleOp mod, llvm::StringRef ident);
 /// Get the compiler identifier from the Module.
 llvm::StringRef getIdent(mlir::ModuleOp mod);
 
+/// Set the command line used in this invocation.
+void setCommandline(mlir::ModuleOp mod, llvm::StringRef cmdLine);
+
+/// Get the command line used in this invocation.
+llvm::StringRef getCommandline(mlir::ModuleOp mod);
+
 /// Helper for determining the target from the host, etc. Tools may use this
 /// function to provide a consistent interpretation of the `--target=<string>`
 /// command-line option.

diff --git a/flang/lib/Frontend/CompilerInvocation.cpp b/flang/lib/Frontend/CompilerInvocation.cpp
@@ -348,6 +348,12 @@ static void parseCodeGenArgs(Fortran::frontend::CodeGenOptions &opts,
   if (auto *a = args.getLastArg(clang::driver::options::OPT_save_temps_EQ))
     opts.SaveTempsDir = a->getValue();
 
+  // -record-command-line option.
+  if (const llvm::opt::Arg *a =
+          args.getLastArg(clang::driver::options::OPT_record_command_line)) {
+    opts.RecordCommandLine = a->getValue();
+  }
+
   // -mlink-builtin-bitcode
   for (auto *a :
        args.filtered(clang::driver::options::OPT_mlink_builtin_bitcode))

diff --git a/flang/lib/Frontend/FrontendActions.cpp b/flang/lib/Frontend/FrontendActions.cpp
@@ -302,7 +302,7 @@ bool CodeGenAction::beginSourceFileAction() {
       kindMap, ci.getInvocation().getLoweringOpts(),
       ci.getInvocation().getFrontendOpts().envDefaults,
       ci.getInvocation().getFrontendOpts().features, targetMachine,
-      ci.getInvocation().getTargetOpts().cpuToTuneFor);
+      ci.getInvocation().getTargetOpts(), ci.getInvocation().getCodeGenOpts());
 
   // Fetch module from lb, so we can set
   mlirModule = std::make_unique<mlir::ModuleOp>(lb.getModule());

diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
@@ -6064,7 +6064,9 @@ Fortran::lower::LoweringBridge::LoweringBridge(
     const Fortran::lower::LoweringOptions &loweringOptions,
     const std::vector<Fortran::lower::EnvironmentDefault> &envDefaults,
     const Fortran::common::LanguageFeatureControl &languageFeatures,
-    const llvm::TargetMachine &targetMachine, const llvm::StringRef tuneCPU)
+    const llvm::TargetMachine &targetMachine,
+    const Fortran::frontend::TargetOptions &targetOpts,
+    const Fortran::frontend::CodeGenOptions &cgOpts)
     : semanticsContext{semanticsContext}, defaultKinds{defaultKinds},
       intrinsics{intrinsics}, targetCharacteristics{targetCharacteristics},
       cooked{&cooked}, context{context}, kindMap{kindMap},
@@ -6121,11 +6123,13 @@ Fortran::lower::LoweringBridge::LoweringBridge(
   fir::setTargetTriple(*module.get(), triple);
   fir::setKindMapping(*module.get(), kindMap);
   fir::setTargetCPU(*module.get(), targetMachine.getTargetCPU());
-  fir::setTuneCPU(*module.get(), tuneCPU);
+  fir::setTuneCPU(*module.get(), targetOpts.cpuToTuneFor);
   fir::setTargetFeatures(*module.get(), targetMachine.getTargetFeatureString());
   fir::support::setMLIRDataLayout(*module.get(),
                                   targetMachine.createDataLayout());
   fir::setIdent(*module.get(), Fortran::common::getFlangFullVersion());
+  if (cgOpts.RecordCommandLine)
+    fir::setCommandline(*module.get(), *cgOpts.RecordCommandLine);
 }
 
 void Fortran::lower::genCleanUpInRegionIfAny(

diff --git a/flang/lib/Optimizer/Dialect/Support/FIRContext.cpp b/flang/lib/Optimizer/Dialect/Support/FIRContext.cpp
@@ -130,6 +130,22 @@ llvm::StringRef fir::getIdent(mlir::ModuleOp mod) {
   return {};
 }
 
+void fir::setCommandline(mlir::ModuleOp mod, llvm::StringRef cmdLine) {
+  if (cmdLine.empty())
+    return;
+
+  mlir::MLIRContext *ctx = mod.getContext();
+  mod->setAttr(mlir::LLVM::LLVMDialect::getCommandlineAttrName(),
+               mlir::StringAttr::get(ctx, cmdLine));
+}
+
+llvm::StringRef fir::getCommandline(mlir::ModuleOp mod) {
+  if (auto attr = mod->getAttrOfType<mlir::StringAttr>(
+          mlir::LLVM::LLVMDialect::getCommandlineAttrName()))
+    return attr;
+  return {};
+}
+
 std::string fir::determineTargetTriple(llvm::StringRef triple) {
   // Treat "" or "default" as stand-ins for the default machine.
   if (triple.empty() || triple == "default")

diff --git a/flang/test/Driver/frecord-command-line.f90 b/flang/test/Driver/frecord-command-line.f90
@@ -0,0 +1,16 @@
+! This only checks that the command line is correctly passed on to the
+! -record-command-line option FC1 option and that the latter does not complain
+! about anything. The correct lowering to a module attribute and beyond will
+! be checked in other tests.
+!
+! RUN: %flang -### -target x86_64-unknown-linux -frecord-command-line %s 2>&1 | FileCheck --check-prefix=CHECK-RECORD %s
+! RUN: %flang -### -target x86_64-unknown-macosx -frecord-command-line %s 2>&1 | FileCheck --check-prefix=CHECK-RECORD %s
+! RUN: not %flang -### -target x86_64-unknown-windows -frecord-command-line %s 2>&1 | FileCheck --check-prefix=CHECK-RECORD-ERROR %s
+
+! RUN: %flang -### -target x86_64-unknown-linux -fno-record-command-line %s 2>&1 | FileCheck --check-prefix=CHECK-NO-RECORD %s
+! RUN: %flang -### -target x86_64-unknown-macosx -fno-record-command-line %s 2>&1 | FileCheck --check-prefix=CHECK-NO-RECORD %s
+! RUN: %flang -### -target x86_64-unknown-windows -fno-record-command-line %s 2>&1 | FileCheck --check-prefix=CHECK-NO-RECORD %s
+
+! CHECK-RECORD: "-record-command-line"
+! CHECK-NO-RECORD-NOT: "-record-command-line"
+! CHECK-RECORD-ERROR: error: unsupported option '-frecord-command-line' for target
diff --git a/flang/test/Lower/record-command-line.f90 b/flang/test/Lower/record-command-line.f90
@@ -0,0 +1,9 @@
+! The actual command line is recorded by the frontend and passed on to FC1 as
+! the argument to -record-command-line, so in this test, we just match against
+! some string with spaces that mimics what a hypothetical command line.
+
+! RUN: %flang_fc1 -record-command-line "exec -o infile" %s -emit-fir -o - | FileCheck %s
+
+! CHECK: module attributes {
+! CHECK-SAME: llvm.commandline = "exec -o infile"
+
diff --git a/flang/tools/bbc/CMakeLists.txt b/flang/tools/bbc/CMakeLists.txt
@@ -35,4 +35,5 @@ FortranParser
 FortranEvaluate
 FortranSemantics
 FortranLower
+flangFrontend
 )