Reland [clang][modules-driver] Add scanner to detect C++20 module presence (llvm#153497)

This patch is part of a series to support driver managed module builds for C++ named modules and Clang modules. This introduces a scanner that detects C++ named module usage early in the driver with only negligible overhead. For now, it is enabled only with the `-fmodules-driver` flag and serves solely diagnostic purposes. In the future, the scanner will be enabled for any (modules-driver compatible) compilation with two or more inputs, and will help the driver determine whether to implicitly enable the modules driver. Since the scanner adds very little overhead, we are also exploring enabling it for compilations with only a single input. This approach could allow us to detect `import std` usage in a single-file compilation, which would then activate the modules driver. For performance measurements on this, see https://github.com/naveen-seth/llvm-dev-cxx-modules-check-benchmark. RFC for driver managed module builds: https://discourse.llvm.org/t/rfc-modules-support-simple-c-20-modules-use-from-the-clang-driver-without-a-build-system This patch relands the reland (2d31fc8) for commit ded1426. The earlier reland failed due to a missing link dependency on `clangLex`. This reland fixes the issue by adding the link dependency after discussing it in the following RFC: https://discourse.llvm.org/t/rfc-driver-link-the-driver-against-clangdependencyscanning-clangast-clangfrontend-clangserialization-and-clanglex

Author: naveen-seth

clang/include/clang/Basic/DiagnosticDriverKinds.td

@@ -581,6 +581,13 @@ def err_drv_reduced_module_output_overrided : Warning<
  "please consider use '-fmodule-output=' to specify the output file for reduced BMI explicitly">,
  InGroup<DiagGroup<"reduced-bmi-output-overrided">>;
 
+def remark_found_cxx20_module_usage : Remark<
+ "found C++20 module usage in file '%0'">,
+ InGroup<ModulesDriver>;
+def remark_performing_driver_managed_module_build : Remark<
+ "performing driver managed module build">,
+ InGroup<ModulesDriver>;
+
 def warn_drv_delayed_template_parsing_after_cxx20 : Warning<
  "-fdelayed-template-parsing is deprecated after C++20">,
  InGroup<DiagGroup<"delayed-template-parsing-in-cxx20">>;

clang/include/clang/Basic/DiagnosticGroups.td

@@ -635,6 +635,7 @@ def ModuleConflict : DiagGroup<"module-conflict">;
 def ModuleFileExtension : DiagGroup<"module-file-extension">;
 def ModuleIncludeDirectiveTranslation : DiagGroup<"module-include-translation">;
 def ModuleMap : DiagGroup<"module-map">;
+def ModulesDriver : DiagGroup<"modules-driver">;
 def RoundTripCC1Args : DiagGroup<"round-trip-cc1-args">;
 def NewlineEOF : DiagGroup<"newline-eof">;
 def Nullability : DiagGroup<"nullability">;

clang/include/clang/Driver/Driver.h

@@ -512,6 +512,9 @@ class Driver {
 
  /// BuildActions - Construct the list of actions to perform for the
  /// given arguments, which are only done for a single architecture.
+ /// If the compilation is an explicit module build, delegates to
+ /// BuildDriverManagedModuleBuildActions. Otherwise, BuildDefaultActions is
+ /// used.
  ///
  /// \param C - The compilation that is being built.
  /// \param Args - The input arguments.
@@ -796,6 +799,35 @@ class Driver {
  /// compilation based on which -f(no-)?lto(=.*)? option occurs last.
  void setLTOMode(const llvm::opt::ArgList &Args);
 
+ /// BuildDefaultActions - Constructs the list of actions to perform
+ /// for the provided arguments, which are only done for a single architecture.
+ ///
+ /// \param C - The compilation that is being built.
+ /// \param Args - The input arguments.
+ /// \param Actions - The list to store the resulting actions onto.
+ void BuildDefaultActions(Compilation &C, llvm::opt::DerivedArgList &Args,
+ const InputList &Inputs, ActionList &Actions) const;
+
+ /// BuildDriverManagedModuleBuildActions - Performs a dependency
+ /// scan and constructs the list of actions to perform for dependency order
+ /// and the provided arguments. This is only done for a single a architecture.
+ ///
+ /// \param C - The compilation that is being built.
+ /// \param Args - The input arguments.
+ /// \param Actions - The list to store the resulting actions onto.
+ void BuildDriverManagedModuleBuildActions(Compilation &C,
+ llvm::opt::DerivedArgList &Args,
+ const InputList &Inputs,
+ ActionList &Actions) const;
+
+ /// Scans the leading lines of the C++ source inputs to detect C++20 module
+ /// usage.
+ ///
+ /// \returns True if module usage is detected, false otherwise, or an error on
+ /// read failure.
+ llvm::ErrorOr<bool>
+ ScanInputsForCXX20ModulesUsage(const InputList &Inputs) const;
+
  /// Retrieves a ToolChain for a particular \p Target triple.
  ///
  /// Will cache ToolChains for the life of the driver object, and create them

clang/include/clang/Driver/Options.td

@@ -3296,6 +3296,13 @@ defm modules_reduced_bmi : BoolOption<"f", "modules-reduced-bmi",
  PosFlag<SetTrue, [], [ClangOption, CC1Option],
  "Generate the reduced BMI">>;
 
+def fmodules_driver : Flag<["-"], "fmodules-driver">,
+ Group<f_Group>, Visibility<[ClangOption]>,
+ HelpText<"Enable support for driver managed module builds (experimental)">;
+def fno_modules_driver : Flag<["-"], "fno-modules-driver">,
+ Group<f_Group>, Visibility<[ClangOption]>,
+ HelpText<"Disable support for driver managed module builds (experimental)">;
+
 def experimental_modules_reduced_bmi : Flag<["-"], "fexperimental-modules-reduced-bmi">,
  Group<f_Group>, Visibility<[ClangOption, CC1Option]>, Alias<fmodules_reduced_bmi>;
 

clang/include/clang/Lex/DependencyDirectivesScanner.h

@@ -135,6 +135,13 @@ void printDependencyDirectivesAsSource(
  ArrayRef<dependency_directives_scan::Directive> Directives,
  llvm::raw_ostream &OS);
 
+/// Scan an input source buffer for C++20 named module usage.
+///
+/// \param Source The input source buffer.
+///
+/// \returns true if any C++20 named modules related directive was found.
+bool scanInputForCXX20ModulesUsage(StringRef Source);
+
 /// Functor that returns the dependency directives for a given file.
 class DependencyDirectivesGetter {
 public:

clang/lib/Driver/CMakeLists.txt

@@ -98,5 +98,6 @@ add_clang_library(clangDriver
 
  LINK_LIBS
  clangBasic
+ clangLex
  ${system_libs}
  )

clang/lib/Driver/Driver.cpp

@@ -66,6 +66,7 @@
 #include "clang/Driver/Tool.h"
 #include "clang/Driver/ToolChain.h"
 #include "clang/Driver/Types.h"
+#include "clang/Lex/DependencyDirectivesScanner.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallSet.h"
@@ -4188,6 +4189,11 @@ void Driver::handleArguments(Compilation &C, DerivedArgList &Args,
  YcArg = nullptr;
  }
 
+ if (Args.hasArgNoClaim(options::OPT_fmodules_driver))
+ // TODO: Check against all incompatible -fmodules-driver arguments
+ if (!ModulesModeCXX20 && !Args.hasArgNoClaim(options::OPT_fmodules))
+ Args.eraseArg(options::OPT_fmodules_driver);
+
  Arg *FinalPhaseArg;
  phases::ID FinalPhase = getFinalPhase(Args, &FinalPhaseArg);
 
@@ -4314,6 +4320,33 @@ void Driver::handleArguments(Compilation &C, DerivedArgList &Args,
  }
 }
 
+static bool hasCXXModuleInputType(const Driver::InputList &Inputs) {
+ const auto IsTypeCXXModule = [](const auto &Input) -> bool {
+ const auto TypeID = Input.first;
+ return (TypeID == types::TY_CXXModule);
+ };
+ return llvm::any_of(Inputs, IsTypeCXXModule);
+}
+
+llvm::ErrorOr<bool>
+Driver::ScanInputsForCXX20ModulesUsage(const InputList &Inputs) const {
+ const auto CXXInputs = llvm::make_filter_range(
+ Inputs, [](const auto &Input) { return types::isCXX(Input.first); });
+ for (const auto &Input : CXXInputs) {
+ StringRef Filename = Input.second->getSpelling();
+ auto ErrOrBuffer = VFS->getBufferForFile(Filename);
+ if (!ErrOrBuffer)
+ return ErrOrBuffer.getError();
+ const auto Buffer = std::move(*ErrOrBuffer);
+
+ if (scanInputForCXX20ModulesUsage(Buffer->getBuffer())) {
+ Diags.Report(diag::remark_found_cxx20_module_usage) << Filename;
+ return true;
+ }
+ }
+ return false;
+}
+
 void Driver::BuildActions(Compilation &C, DerivedArgList &Args,
  const InputList &Inputs, ActionList &Actions) const {
  llvm::PrettyStackTraceString CrashInfo("Building compilation actions");
@@ -4325,6 +4358,33 @@ void Driver::BuildActions(Compilation &C, DerivedArgList &Args,
 
  handleArguments(C, Args, Inputs, Actions);
 
+ if (Args.hasFlag(options::OPT_fmodules_driver,
+ options::OPT_fno_modules_driver, false)) {
+ // TODO: Move the logic for implicitly enabling explicit-module-builds out
+ // of -fmodules-driver once it is no longer experimental.
+ // Currently, this serves diagnostic purposes only.
+ bool UsesCXXModules = hasCXXModuleInputType(Inputs);
+ if (!UsesCXXModules) {
+ const auto ErrOrScanResult = ScanInputsForCXX20ModulesUsage(Inputs);
+ if (!ErrOrScanResult) {
+ Diags.Report(diag::err_cannot_open_file)
+ << ErrOrScanResult.getError().message();
+ return;
+ }
+ UsesCXXModules = *ErrOrScanResult;
+ }
+ if (UsesCXXModules || Args.hasArg(options::OPT_fmodules))
+ BuildDriverManagedModuleBuildActions(C, Args, Inputs, Actions);
+ return;
+ }
+
+ BuildDefaultActions(C, Args, Inputs, Actions);
+}
+
+void Driver::BuildDefaultActions(Compilation &C, DerivedArgList &Args,
+ const InputList &Inputs,
+ ActionList &Actions) const {
+
  bool UseNewOffloadingDriver =
  C.isOffloadingHostKind(Action::OFK_OpenMP) ||
  C.isOffloadingHostKind(Action::OFK_SYCL) ||
@@ -4608,6 +4668,13 @@ void Driver::BuildActions(Compilation &C, DerivedArgList &Args,
  Args.ClaimAllArgs(options::OPT_cl_ignored_Group);
 }
 
+void Driver::BuildDriverManagedModuleBuildActions(
+ Compilation &C, llvm::opt::DerivedArgList &Args, const InputList &Inputs,
+ ActionList &Actions) const {
+ Diags.Report(diag::remark_performing_driver_managed_module_build);
+ return;
+}
+
 /// Returns the canonical name for the offloading architecture when using a HIP
 /// or CUDA architecture.
 static StringRef getCanonicalArchString(Compilation &C,

clang/lib/Lex/DependencyDirectivesScanner.cpp

@@ -83,6 +83,8 @@ struct Scanner {
  /// \returns True on error.
  bool scan(SmallVectorImpl<Directive> &Directives);
 
+ friend bool clang::scanInputForCXX20ModulesUsage(StringRef Source);
+
 private:
  /// Lexes next token and advances \p First and the \p Lexer.
  [[nodiscard]] dependency_directives_scan::Token &
@@ -1075,3 +1077,51 @@ void clang::printDependencyDirectivesAsSource(
  }
  }
 }
+
+static void skipUntilMaybeCXX20ModuleDirective(const char *&First,
+ const char *const End) {
+ assert(First <= End);
+ while (First != End) {
+ if (*First == '#') {
+ ++First;
+ skipToNewlineRaw(First, End);
+ }
+ skipWhitespace(First, End);
+ if (const auto Len = isEOL(First, End)) {
+ First += Len;
+ continue;
+ }
+ break;
+ }
+}
+
+bool clang::scanInputForCXX20ModulesUsage(StringRef Source) {
+ const char *First = Source.begin();
+ const char *const End = Source.end();
+ skipUntilMaybeCXX20ModuleDirective(First, End);
+ if (First == End)
+ return false;
+
+ // Check if the next token can even be a module directive before creating a
+ // full lexer.
+ if (!(*First == 'i' || *First == 'e' || *First == 'm'))
+ return false;
+
+ llvm::SmallVector<dependency_directives_scan::Token> Tokens;
+ Scanner S(StringRef(First, End - First), Tokens, nullptr, SourceLocation());
+ S.TheLexer.setParsingPreprocessorDirective(true);
+ if (S.lexModule(First, End))
+ return false;
+ auto IsCXXNamedModuleDirective = [](const DirectiveWithTokens &D) {
+ switch (D.Kind) {
+ case dependency_directives_scan::cxx_module_decl:
+ case dependency_directives_scan::cxx_import_decl:
+ case dependency_directives_scan::cxx_export_module_decl:
+ case dependency_directives_scan::cxx_export_import_decl:
+ return true;
+ default:
+ return false;
+ }
+ };
+ return llvm::any_of(S.DirsWithToks, IsCXXNamedModuleDirective);
+}