Add clang plugin for class usage replacing (::std::atomic<T> to LTestAtomic<T>)

.github/workflows/clang-format.yaml

@@ -10,5 +10,6 @@ jobs:
       - name: Run clang-format style check
         uses: jidicula/clang-format-action@v4.13.0
         with:
-          clang-format-version: '14'
+          clang-format-version: '19'
           check-path: '.'
+          fallback-style: 'Google'

.github/workflows/run-tests.yaml

@@ -3,18 +3,19 @@ on:
   push:
   pull_request:
     branches: [ "master" ]
+
 jobs:
   unit-tests:
     runs-on: ubuntu-latest
     defaults:
       run:
         shell: bash
     container: 
-      image: silkeh/clang:18
+      image: silkeh/clang:19
       options: --user root
     steps:
       - name: Install deps
-        run: apt update && apt install -y git ninja-build valgrind libboost-context-dev libgflags-dev
+        run: apt update && apt install -y git ninja-build valgrind libboost-context-dev libgflags-dev libclang-19-dev
       - name: Check out repository code
         uses: actions/checkout@v4
       - name: Build
@@ -29,12 +30,12 @@ jobs:
       run:
         shell: bash
     container: 
-      image: silkeh/clang:18
+      image: silkeh/clang:19
       options: --user root
     steps:
       - name: Install deps
         run: |
-          apt update && apt install -y git ninja-build valgrind libgoogle-glog-dev libsnappy-dev protobuf-compiler libboost-context-dev pkg-config libcapstone-dev && \
+          apt update && apt install -y git ninja-build valgrind libgoogle-glog-dev libsnappy-dev protobuf-compiler libboost-context-dev pkg-config libcapstone-dev libclang-19-dev && \
           git clone https://github.com/Kirillog/syscall_intercept.git &&  \
           cmake syscall_intercept -G Ninja -B syscall_intercept/build -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=clang && \
           cmake --build syscall_intercept/build --target install
@@ -100,12 +101,12 @@ jobs:
       run:
         shell: bash
     container: 
-      image: silkeh/clang:18
+      image: silkeh/clang:19
       options: --user root
     steps:
       - name: Install deps
         run: |
-          apt update && apt install -y git ninja-build valgrind libgoogle-glog-dev libsnappy-dev protobuf-compiler libboost-context-dev pkg-config libcapstone-dev && \
+          apt update && apt install -y git ninja-build valgrind libgoogle-glog-dev libsnappy-dev protobuf-compiler libboost-context-dev pkg-config libcapstone-dev libclang-19-dev && \
           git clone https://github.com/Kirillog/syscall_intercept.git &&  \
           cmake syscall_intercept -G Ninja -B syscall_intercept/build -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=clang && \
           cmake --build syscall_intercept/build --target install

CMakeLists.txt

@@ -5,8 +5,13 @@ project(lintest)
 
 set(CMAKE_CXX_STANDARD 20)
 
+# in order for clang tool to work, this must be ON (it will generate the compilation database)
+# see https://clang.llvm.org/docs/JSONCompilationDatabase.html
 set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
+# set this to ON in order to substitute std::atomic<T> with custom implementation
+set(APPLY_CLANG_TOOL OFF)
+
 # TODO(kmitkin): require to understand, what is it considered to be "optimized" build
 # set(CMAKE_CXX_FLAGS_RELEASE "???")
 set(CMAKE_CXX_FLAGS_DEBUG "-g -ggdb3 -O0 -fno-omit-frame-pointer")
@@ -24,6 +29,7 @@ include(GoogleTest)
 fuzztest_setup_fuzzing_flags()
 enable_testing()
 
+add_subdirectory(clangpass)
 add_subdirectory(codegen)
 add_subdirectory(runtime)
 

Dockerfile

@@ -1,6 +1,6 @@
 FROM silkeh/clang:19 AS ltest
 
-RUN apt update && apt install -y git ninja-build valgrind libboost-context-dev libgflags-dev libstdc++-11-dev
+RUN apt update && apt install -y git ninja-build valgrind libboost-context-dev libgflags-dev libstdc++-11-dev libclang-19-dev
 RUN mv /usr/lib/gcc/x86_64-linux-gnu/12 /usr/lib/gcc/x86_64-linux-gnu/_12
 
 FROM ltest as blocking

clangpass/CMakeLists.txt

@@ -0,0 +1,64 @@
+#===============================================================================
+# SETUP CLANG PLUGIN
+#===============================================================================
+find_package(Clang REQUIRED CONFIG)
+if("${LLVM_VERSION_MAJOR}" VERSION_LESS 19)
+  message(FATAL_ERROR "Found LLVM ${LLVM_VERSION_MAJOR}, but need LLVM 19 or above")
+endif()
+
+include_directories(SYSTEM "${LLVM_INCLUDE_DIRS};${CLANG_INCLUDE_DIRS}")
+
+if(NOT LLVM_ENABLE_RTTI)
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fno-rtti")
+endif()
+
+# -fvisibility-inlines-hidden is set when building LLVM and on Darwin warnings
+# are triggered if llvm-tutor is built without this flag (though otherwise it
+# builds fine). For consistency, add it here too.
+include(CheckCXXCompilerFlag)
+check_cxx_compiler_flag("-fvisibility-inlines-hidden"
+  SUPPORTS_FVISIBILITY_INLINES_HIDDEN_FLAG)
+if(${SUPPORTS_FVISIBILITY_INLINES_HIDDEN_FLAG} EQUAL "1")
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fvisibility-inlines-hidden")
+endif()
+
+add_library(ClangPass SHARED "${CMAKE_CURRENT_SOURCE_DIR}/clangpass.cpp")
+
+target_include_directories(
+  ClangPass
+  PRIVATE
+  "${CMAKE_CURRENT_SOURCE_DIR}/include"
+)
+
+# Allow undefined symbols in shared objects on Darwin (this is the default
+# behaviour on Linux)
+target_link_libraries(ClangPass
+  "$<$<PLATFORM_ID:Darwin>:-undefined dynamic_lookup>")
+
+
+#===============================================================================
+# SETUP CLANG TOOL
+#===============================================================================
+set(CLANG_TOOL "ClangPassTool")
+set(CLANG_TOOL_SOURCES
+  "${CMAKE_CURRENT_SOURCE_DIR}/clangpass_tool.cpp"
+  "${CMAKE_CURRENT_SOURCE_DIR}/clangpass.cpp"
+)
+
+add_executable(
+  ${CLANG_TOOL}
+  ${CLANG_TOOL_SOURCES}
+)
+
+# Configure include directories for 'tool'
+target_include_directories(
+  ${CLANG_TOOL}
+  PRIVATE
+  "${CMAKE_CURRENT_SOURCE_DIR}/include"
+)
+
+# Link in the required libraries 
+target_link_libraries(
+  ${CLANG_TOOL}
+  clangTooling
+)

clangpass/clangpass.cpp

@@ -0,0 +1,218 @@
+//==============================================================================
+// FILE:
+//    clangpass.cpp
+//
+// DESCRIPTION:
+//    Substitutes all input files std::atomic<T> usages with LTestAtomic<T> with
+//    the same API.
+//
+// USAGE:
+// clang++ -Xclang -load -Xclang ./build/lib/libClangPass.so -Xclang -add-plugin -Xclang ClangPass ./AtomicsReplacer/test-project/main.cpp 
+//
+// License: The Unlicense
+//==============================================================================
+
+#include "clang/AST/ASTConsumer.h"
+#include "clang/ASTMatchers/ASTMatchFinder.h"
+#include "clang/Rewrite/Core/Rewriter.h"
+#include "clang/Rewrite/Frontend/FixItRewriter.h"
+#include "clang/Tooling/CommonOptionsParser.h"
+#include "clang/AST/Expr.h"
+#include "clang/AST/ExprCXX.h"
+#include "clang/AST/RecursiveASTVisitor.h"
+#include "clang/ASTMatchers/ASTMatchers.h"
+#include "clang/Frontend/CompilerInstance.h"
+#include "clang/Frontend/FrontendPluginRegistry.h"
+#include "clang/Tooling/Refactoring/Rename/RenamingAction.h"
+#include "llvm/ADT/StringMap.h"
+#include "llvm/Support/raw_ostream.h"
+#include "llvm/Support/raw_ostream.h"
+
+#include "clangpass.h"
+
+using namespace clang;
+using namespace ast_matchers;
+using namespace llvm;
+
+//-----------------------------------------------------------------------------
+// ASTFinder callback
+//-----------------------------------------------------------------------------
+CodeRefactorMatcher::CodeRefactorMatcher(
+  ASTContext& Context,
+  clang::Rewriter &RewriterForCodeRefactor,
+  std::string ClassNameToReplace,
+  std::string ClassNameToInsert
+) : Context(Context),
+    CodeRefactorRewriter(RewriterForCodeRefactor),
+    ClassNameToReplace(ClassNameToReplace),
+    ClassNameToInsert(ClassNameToInsert) {}
+
+void CodeRefactorMatcher::onEndOfTranslationUnit() {
+  const SourceManager& SM = CodeRefactorRewriter.getSourceMgr();    
+  FileID MainFileID = SM.getMainFileID();
+  const RewriteBuffer& Buffer = CodeRefactorRewriter.getEditBuffer(MainFileID);
+
+  // Output to stdout
+  llvm::outs() << "Transformed code:\n";
+  Buffer.write(llvm::outs());
+  llvm::outs() << "\n";
+
+  // Output to file
+  const FileEntry *Entry = SM.getFileEntryForID(MainFileID);
+  StringRef OriginalFilename = Entry->tryGetRealPathName();
+
+  size_t slashIndex = OriginalFilename.rfind("/");
+  // the path should be absolute, so in the worst case we will get '/' as index 0
+  assert(slashIndex != std::string::npos);
+  slashIndex += 1; // include the '/' itself
+
+  std::string Path = std::string(OriginalFilename.begin(), OriginalFilename.begin() + slashIndex);
+  std::string SourceFilename = std::string(OriginalFilename.begin() + slashIndex, OriginalFilename.end());
+
+  llvm::outs() << "Original filename: " << OriginalFilename << "\n";
+  std::string OutputFilename = Path + "__tmp_" + SourceFilename;
+
+  std::error_code EC;
+  llvm::raw_fd_ostream OS(OutputFilename, EC, llvm::sys::fs::OF_None);
+
+  if (EC) {
+    llvm::errs() << "Error: Could not open output file: " << EC.message() << "\n";
+    return;
+  }
+
+  llvm::outs() << "Writing to file: " << OutputFilename << "\n";
+  //OS << std::string(Buffer->begin(), Buffer->end());
+  Buffer.write(OS);
+  OS.close();
+}
+
+void CodeRefactorMatcher::run(const clang::ast_matchers::MatchFinder::MatchResult &Result) {
+  if (const auto* ETL = Result.Nodes.getNodeAs<ElaboratedTypeLoc>("ElaboratedTypeLoc")) {      
+    const auto* TemplType = ETL->getType()->getAs<TemplateSpecializationType>();
+    if (!TemplType) {
+      return;
+    }
+
+    CodeRefactorRewriter.ReplaceText(ETL->getSourceRange(), ClassNameToInsert + GetArgumentsFromTemplateType(TemplType));
+  }
+
+  if (const auto* QTL = Result.Nodes.getNodeAs<QualifiedTypeLoc>("QualifiedTypeLoc")) {
+    const auto* TemplType = QTL->getType()->getAs<TemplateSpecializationType>();
+    if (!TemplType) {
+      return;
+    }
+
+    CodeRefactorRewriter.ReplaceText(QTL->getSourceRange(), ClassNameToInsert + GetArgumentsFromTemplateType(TemplType));
+  }
+}
+
+std::string CodeRefactorMatcher::GetArgumentsFromTemplateType(const TemplateSpecializationType *TST) {
+  std::string args;
+  llvm::raw_string_ostream os(args);
+  printTemplateArgumentList(os, TST->template_arguments(), Context.getPrintingPolicy());
+  return args;
+}
+
+// Util function for debugging purposes
+std::string CodeRefactorMatcher::getSourceRangeAsString(const SourceRange& SR) const {
+  auto& sm = CodeRefactorRewriter.getSourceMgr();
+  auto& langOpts = CodeRefactorRewriter.getLangOpts();
+
+  clang::SourceLocation start = SR.getBegin();
+  clang::SourceLocation end = SR.getEnd();
+  end = clang::Lexer::getLocForEndOfToken(end, 0, sm, langOpts);
+
+  bool isInvalid = false;
+  const char *startData = sm.getCharacterData(start, &isInvalid);
+
+  if (isInvalid) {
+    return "<invalid begin>";
+    isInvalid = false;
+  }
+
+  const char *endData = sm.getCharacterData(end, &isInvalid);
+
+  if (isInvalid) {
+    return "<invalid end>";
+    isInvalid = false;
+  }
+  size_t length = endData - startData;
+
+  return std::string(startData, length);
+}
+
+
+//-----------------------------------------------------------------------------
+// ASTConsumer
+//-----------------------------------------------------------------------------
+CodeRefactorASTConsumer::CodeRefactorASTConsumer(
+  ASTContext& Context,
+  clang::Rewriter &R,
+  std::string ClassNameToReplace,
+  std::string ClassNameToInsert
+): CodeRefactorHandler(Context, R, ClassNameToReplace, ClassNameToInsert),
+    ClassNameToReplace(ClassNameToReplace),
+    ClassNameToInsert(ClassNameToInsert) {
+  // Does not support matching the parameters of the functions
+  const auto MatcherForFQTemplateTypes = elaboratedTypeLoc(
+    hasNamedTypeLoc(
+      loc(
+        templateSpecializationType(
+          hasDeclaration(
+            classTemplateSpecializationDecl(
+              hasName(ClassNameToReplace)
+            )
+          )
+        )
+      )
+    )
+  );
+
+  // Uses previous matcher inside, but returns a wrapping QualifiedTypeLoc node
+  // which is used in the function parameters
+  const auto MatcherForFQTemplateParams = qualifiedTypeLoc(
+    hasUnqualifiedLoc(
+      MatcherForFQTemplateTypes
+    )
+  );
+
+  Finder.addMatcher(MatcherForFQTemplateTypes.bind("ElaboratedTypeLoc"), &CodeRefactorHandler);
+  Finder.addMatcher(MatcherForFQTemplateParams.bind("QualifiedTypeLoc"), &CodeRefactorHandler);
+}
+
+void CodeRefactorASTConsumer::HandleTranslationUnit(clang::ASTContext &Ctx) {
+  Finder.matchAST(Ctx);
+}
+
+
+//-----------------------------------------------------------------------------
+// FrontendAction
+//-----------------------------------------------------------------------------
+class CodeRefactorAddPluginAction : public PluginASTAction {
+public:
+  bool ParseArgs(const CompilerInstance &CI, const std::vector<std::string> &Args) override {
+    return true;
+  }
+
+  // Returns our ASTConsumer per translation unit.
+  std::unique_ptr<ASTConsumer> CreateASTConsumer(CompilerInstance &CI,
+                                                  StringRef file) override {
+    RewriterForCodeRefactor.setSourceMgr(CI.getSourceManager(),
+                                          CI.getLangOpts());
+    return std::make_unique<CodeRefactorASTConsumer>(
+        CI.getASTContext(), RewriterForCodeRefactor, ClassNameToReplace, ClassNameToInsert);
+  }
+
+private:
+  Rewriter RewriterForCodeRefactor;
+  // The shared library has predefined replace- and insert-class names,
+  // users are intended to use clangpass_tool.cpp instead.
+  std::string ClassNameToReplace = "::std::atomic";
+  std::string ClassNameToInsert = "LTestAtomic";
+};
+
+
+static FrontendPluginRegistry::Add<CodeRefactorAddPluginAction> X(
+  "ClangPass",
+  "Replace all std::atomic usages with LTestAtomic"
+);