Fix junk detection: getting the filenames right is tricky

include/mull/JunkDetection/CXX/ASTStorage.h

@@ -28,6 +28,7 @@ class ThreadSafeASTUnit {
   bool isInSystemHeader(clang::SourceLocation &location);
 
   clang::Decl *getDecl(clang::SourceLocation &location);
+  bool hasAST() const;
 
 private:
   void recordDeclarations();

lib/Driver.cpp

@@ -397,6 +397,7 @@ void mull::mutateBitcode(llvm::Module &module) {
     diagnostics.info("Using configuration "s + configPath);
     configuration = Configuration::loadFromDisk(diagnostics, configPath);
   }
+  configuration.parallelization.normalize();
 
   if (configuration.debugEnabled) {
     diagnostics.enableDebugMode();

lib/JunkDetection/CXX/ASTStorage.cpp

@@ -81,7 +81,9 @@ const clang::FileEntry *ThreadSafeASTUnit::findFileEntry(const std::string &file
     if (!llvm::sys::path::is_absolute(currentSourceFilePath)) {
       currentSourceFilePath = it->first->tryGetRealPathName();
     }
-    if (currentSourceFilePath.equals(filePath)) {
+    llvm::SmallString<PATH_MAX> realFilePath;
+    llvm::sys::fs::real_path(filePath, realFilePath);
+    if (currentSourceFilePath.equals(filePath) || currentSourceFilePath.equals(realFilePath)) {
       file = it->first;
       break;
     }
@@ -181,6 +183,10 @@ clang::Decl *ThreadSafeASTUnit::getDecl(clang::SourceLocation &location) {
   return nullptr;
 }
 
+bool ThreadSafeASTUnit::hasAST() const {
+  return ast != nullptr;
+}
+
 ASTStorage::ASTStorage(Diagnostics &diagnostics, const std::string &cxxCompilationDatabasePath,
                        const std::string &cxxCompilationFlags,
                        const std::unordered_map<std::string, std::string> &bitcodeCompilationFlags)
@@ -214,7 +220,9 @@ ThreadSafeASTUnit *ASTStorage::findAST(const std::string &sourceFile) {
   for (auto &flag : compilationFlags) {
     args.push_back(flag.c_str());
   }
-  args.push_back(sourceFile.c_str());
+  if (args.size() == 1) {
+    args.push_back(sourceFile.c_str());
+  }
 
   clang::IntrusiveRefCntPtr<clang::DiagnosticsEngine> diagnosticsEngine(
       clang::CompilerInstance::createDiagnostics(new clang::DiagnosticOptions));

lib/JunkDetection/CXX/CXXJunkDetector.cpp

@@ -192,7 +192,7 @@ bool CXXJunkDetector::isJunk(MutationPoint *point) {
   }
 
   ThreadSafeASTUnit *ast = astStorage.findAST(point->getSourceLocation());
-  if (!ast) {
+  if (!ast->hasAST()) {
     return true;
   }
   clang::SourceLocation location = ast->getLocation(point->getSourceLocation());

lib/JunkDetection/CXX/CompilationDatabase.cpp

@@ -39,12 +39,6 @@ static CompilationDatabase::Flags flagsFromCommand(const clang::tooling::Compile
   CompilationDatabase::Flags flags(command.CommandLine);
   flags = filterFlags(flags, true);
 
-  // The compilation database produced from running
-  // clang ... -MJ <comp.database.json>
-  // contains a file name in the "arguments" array. Since the file name
-  // itself is not a compilation flag we filter it out.
-  flags.erase(std::remove(flags.begin(), flags.end(), command.Filename), flags.end());
-
   // append extraFlags
   std::copy(std::begin(extraFlags), std::end(extraFlags), std::back_inserter(flags));
 
@@ -78,9 +72,6 @@ createBitcodeFlags(Diagnostics &diagnostics,
 
     fileFlags = filterFlags(fileFlags, true);
 
-    /// Remove file name from the list of flags
-    fileFlags.erase(std::remove(fileFlags.begin(), fileFlags.end(), filename), fileFlags.end());
-
     for (const auto &extraFlag : extraFlags) {
       fileFlags.push_back(extraFlag);
     }

tests-lit/tests/junk_detection/02_junk_detection_using_extra_flags/mull.no_flag.yml

@@ -3,3 +3,4 @@ mutators:
   - cxx_remove_void_call
 compilerFlags:
   - -DWRONG_FLAG=1
+  - ./sample.cpp

tests-lit/tests/junk_detection/02_junk_detection_using_extra_flags/mull.with_flag.yml

@@ -3,3 +3,4 @@ mutators:
   - cxx_remove_void_call
 compilerFlags:
   - -DFLAG=1
+  - ./sample.cpp

tests-lit/tests/junk_detection/02_junk_detection_using_extra_flags/sample.cpp

@@ -21,8 +21,8 @@ RUN: %mull_cxx -linker=%clang_cxx -linker-flags="%sysroot" -disable-junk-detecti
 
 WITHOUT-JUNK-DETECTION:{{^.*}}sample.cpp:5:13: warning: Survived: Removed the call to the function [cxx_remove_void_call]{{$}}
 
-RUN: cd / && env MULL_CONFIG=%S/mull.no_flag.yml %clang_cxx %sysroot -O0 %pass_mull_ir_frontend -g -DFLAG=1 %s -o %s-ir-no-flag.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-NO-FLAG-MUTATE
-RUN: %mull_cxx -mutate-only -output=%s-no-flag.exe -linker=%clang_cxx -linker-flags="%sysroot" -mutators=cxx_add_to_sub -mutators=cxx_remove_void_call -reporters=IDE -ide-reporter-show-killed -compilation-flags '-DWRONG_FLAG=1' %s.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-NO-FLAG-MUTATE
+RUN: cd %S && env MULL_CONFIG=%S/mull.no_flag.yml %clang_cxx %sysroot -O0 %pass_mull_ir_frontend -g -DFLAG=1 %s -o %s-ir-no-flag.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-NO-FLAG-MUTATE
+RUN: %mull_cxx -mutate-only -output=%s-no-flag.exe -linker=%clang_cxx -linker-flags="%sysroot" -mutators=cxx_add_to_sub -mutators=cxx_remove_void_call -reporters=IDE -ide-reporter-show-killed -compilation-flags="-DWRONG_FLAG=1 %s" %s.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-NO-FLAG-MUTATE
 RUN: %mull_runner -ide-reporter-show-killed %s-no-flag.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-NO-FLAG
 RUN: %mull_runner -ide-reporter-show-killed %s-ir-no-flag.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-NO-FLAG
 
@@ -35,8 +35,8 @@ WITH-JUNK-DETECTION-NO-FLAG-MUTATE:Make sure that the flags provided to Mull are
 TODO: It is interesting why there is no junk even if we have the error above.
 WITH-JUNK-DETECTION-NO-FLAG:[info] Killed mutants (1/1):
 
-RUN: cd / && env MULL_CONFIG=%S/mull.with_flag.yml %clang_cxx %sysroot -O0 %pass_mull_ir_frontend -g -DFLAG=1 %s -o %s-ir-with-flag.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-WITH-FLAG-MUTATE
-RUN: %mull_cxx -mutate-only -output=%s-with-flag.exe -linker=%clang_cxx -linker-flags="%sysroot" -mutators=cxx_add_to_sub -mutators=cxx_remove_void_call -reporters=IDE -ide-reporter-show-killed -compilation-flags '-DFLAG=1' %s.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-WITH-FLAG-MUTATE
+RUN: cd %S && env MULL_CONFIG=%S/mull.with_flag.yml %clang_cxx %sysroot -O0 %pass_mull_ir_frontend -g -DFLAG=1 %s -o %s-ir-with-flag.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-WITH-FLAG-MUTATE
+RUN: %mull_cxx -mutate-only -output=%s-with-flag.exe -linker=%clang_cxx -linker-flags="%sysroot" -mutators=cxx_add_to_sub -mutators=cxx_remove_void_call -reporters=IDE -ide-reporter-show-killed -compilation-flags '-DFLAG=1 %s' %s.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-WITH-FLAG-MUTATE
 RUN: %mull_runner -reporters=IDE -ide-reporter-show-killed %s-with-flag.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-WITH-FLAG
 RUN: %mull_runner -reporters=IDE -ide-reporter-show-killed %s-ir-with-flag.exe 2>&1 | %filecheck %s --dump-input=fail --strict-whitespace --match-full-lines --check-prefix=WITH-JUNK-DETECTION-WITH-FLAG
 

tests-lit/tests/runner/mutants-from-dylib/mull.yml

@@ -1,4 +1,5 @@
 mutators:
   - cxx_add_to_sub
 compilerFlags:
-  - "-DSHARED_LIB"
+  - "-DSHARED_LIB"
+  - ./test.c

tests-lit/tests/runner/mutants-from-dylib/test.c

@@ -24,7 +24,7 @@ RUN: %clang_cc %sysroot -DSHARED_LIB -fPIC -shared %s %pass_mull_ir_frontend -g
 RUN: cd %S; %clang_cc %sysroot -DTEST_BIN ./test.c %S/shared-ir.lib -o ./test-ir.exe
 
 RUN: %clang_cc %sysroot -DSHARED_LIB -fPIC -shared %s -fembed-bitcode -g -o %S/shared.lib
-RUN: unset TERM; %mull_cxx -mutate-only --compilation-flags=-DSHARED_LIB -linker=%clang_cc -linker-flags="%sysroot -shared -fPIC" --output=%S/mutated.lib %S/shared.lib
+RUN: unset TERM; %mull_cxx -mutate-only --compilation-flags="-DSHARED_LIB %s" -linker=%clang_cc -linker-flags="%sysroot -shared -fPIC" --output=%S/mutated.lib %S/shared.lib
 RUN: cd %S; %clang_cc %sysroot -DTEST_BIN ./test.c ./mutated.lib -o ./test.exe
 RUN: cd /; unset TERM; %mull_runner -ld-search-path=%S -ide-reporter-show-killed %S/test.exe | %filecheck %s --dump-input=fail --match-full-lines --check-prefix=CHECK
 RUN: cd /; unset TERM; %mull_runner -ld-search-path=%S -ide-reporter-show-killed %S/test-ir.exe | %filecheck %s --dump-input=fail --match-full-lines --check-prefix=CHECK

tests/JunkDetection/CXXJunkDetectorTests.cpp

@@ -280,8 +280,9 @@ TEST(CXXJunkDetector, no_compdb) {
 
   ASSERT_EQ(points.size(), 8U);
 
-  std::string cxxCompilationFlags =
-      std::string("-I ") + fixtures::junk_detection_compdb_include__path();
+  std::string cxxCompilationFlags = std::string("-I ") +
+                                    fixtures::junk_detection_compdb_include__path() + " " +
+                                    bitcode->getModule()->getSourceFileName();
 
   ASTStorage astStorage(diagnostics, "", cxxCompilationFlags, {});
 

tests/JunkDetection/CompilationDatabaseTests.cpp

@@ -54,13 +54,14 @@ TEST(CompilationDatabaseFromFile, loadsFromValidFiles) {
       const CompilationDatabase database = CompilationDatabase::fromFile(diagnostics, path, "", {});
 
       auto compilationFlags = database.compilationFlagsForFile(file);
-      ASSERT_EQ(compilationFlags.size(), size_t(5));
+      ASSERT_EQ(compilationFlags.size(), size_t(6));
 
       ASSERT_EQ(compilationFlags.at(0), "-I"s);
       ASSERT_EQ(compilationFlags.at(1), "foo"s);
       ASSERT_EQ(compilationFlags.at(2), "-I"s);
       ASSERT_EQ(compilationFlags.at(3), "bar"s);
       ASSERT_EQ(compilationFlags.at(4), "-c"s);
+      ASSERT_EQ(compilationFlags.at(5), "foobar.cpp"s);
     }
   }
 }
@@ -73,17 +74,18 @@ TEST(CompilationDatabaseFromFile, includesCompilationFlagsPassedSeparately) {
 
   const std::string file("/foo/bar/foobar.cpp");
   auto compilationFlags = database.compilationFlagsForFile(file);
-  ASSERT_EQ(compilationFlags.size(), size_t(9));
+  ASSERT_EQ(compilationFlags.size(), size_t(10));
 
   ASSERT_EQ(compilationFlags.at(0), "-I"s);
   ASSERT_EQ(compilationFlags.at(1), "foo"s);
   ASSERT_EQ(compilationFlags.at(2), "-I"s);
   ASSERT_EQ(compilationFlags.at(3), "bar"s);
   ASSERT_EQ(compilationFlags.at(4), "-c"s);
-  ASSERT_EQ(compilationFlags.at(5), "-isystem"s);
-  ASSERT_EQ(compilationFlags.at(6), "/usr/local/include"s);
-  ASSERT_EQ(compilationFlags.at(7), "-isystem"s);
-  ASSERT_EQ(compilationFlags.at(8), "/usr/include"s);
+  ASSERT_EQ(compilationFlags.at(5), "foobar.cpp"s);
+  ASSERT_EQ(compilationFlags.at(6), "-isystem"s);
+  ASSERT_EQ(compilationFlags.at(7), "/usr/local/include"s);
+  ASSERT_EQ(compilationFlags.at(8), "-isystem"s);
+  ASSERT_EQ(compilationFlags.at(9), "/usr/include"s);
 }
 
 TEST(CompilationDatabaseFromFile, parsesCompilationFlagsFromClangMJCommandValid) {
@@ -95,12 +97,13 @@ TEST(CompilationDatabaseFromFile, parsesCompilationFlagsFromClangMJCommandValid)
   const std::string file("/tmp/main.cpp");
   auto compilationFlags = database.compilationFlagsForFile(file);
 
-  ASSERT_EQ(compilationFlags.size(), size_t(20));
+  ASSERT_EQ(compilationFlags.size(), size_t(21));
   ASSERT_EQ(compilationFlags.at(0), "-xc++"s);
-  ASSERT_EQ(compilationFlags.at(1), "-fembed-bitcode=all"s);
-  ASSERT_EQ(compilationFlags.at(2), "-g"s);
-  ASSERT_EQ(compilationFlags.at(18), "-isystem"s);
-  ASSERT_EQ(compilationFlags.at(19), "/usr/include"s);
+  ASSERT_EQ(compilationFlags.at(1), "main.cpp"s);
+  ASSERT_EQ(compilationFlags.at(2), "-fembed-bitcode=all"s);
+  ASSERT_EQ(compilationFlags.at(3), "-g"s);
+  ASSERT_EQ(compilationFlags.at(19), "-isystem"s);
+  ASSERT_EQ(compilationFlags.at(20), "/usr/include"s);
 }
 
 TEST(CompilationDatabaseFromFile, parsesCompilationDatabaseWithEscapedQuotes) {
@@ -111,7 +114,7 @@ TEST(CompilationDatabaseFromFile, parsesCompilationDatabaseWithEscapedQuotes) {
   const std::string file("/tmp/main.cpp");
   auto compilationFlags = database.compilationFlagsForFile(file);
 
-  ASSERT_EQ(compilationFlags.size(), size_t(5));
+  ASSERT_EQ(compilationFlags.size(), size_t(6));
   ASSERT_EQ(compilationFlags.at(0), "-DQT_CORE_LIB"s);
   ASSERT_EQ(compilationFlags.at(1), "-DQT_TESTCASE_BUILDDIR=\"/src/builds/amd64-linux-mull\""s);
   ASSERT_EQ(compilationFlags.at(2), "-o"s);

tests/fixtures/junk_detection/compdb/compile_commands.json.in

@@ -1,7 +1,7 @@
 [
 {
   "directory": "@DIR@",
-  "command": "clang -I @INC@/include",
+  "command": "clang -I @INC@/include @DIR@/main.cpp",
   "file": "main.cpp"
 }
 ]

tools/mull-cxx-ir-frontend/CMakeLists.txt

@@ -5,4 +5,7 @@ target_include_directories(mull-cxx-ir-frontend-${LLVM_VERSION_MAJOR} SYSTEM PRI
 set_target_properties(mull-cxx-ir-frontend-${LLVM_VERSION_MAJOR} PROPERTIES COMPILE_FLAGS ${MULL_CXX_FLAGS})
 set_target_properties(mull-cxx-ir-frontend-${LLVM_VERSION_MAJOR} PROPERTIES LIBRARY_OUTPUT_NAME mull-ir-frontend-${LLVM_VERSION_MAJOR})
 set_target_properties(mull-cxx-ir-frontend-${LLVM_VERSION_MAJOR} PROPERTIES PREFIX "")
-set_target_properties(mull-cxx-ir-frontend-${LLVM_VERSION_MAJOR} PROPERTIES SUFFIX "")
+set_target_properties(mull-cxx-ir-frontend-${LLVM_VERSION_MAJOR} PROPERTIES SUFFIX "")
+install(TARGETS mull-cxx-ir-frontend-${LLVM_VERSION_MAJOR}
+  RUNTIME DESTINATION lib
+)

tools/mull-cxx-ir-frontend/mull-cxx-ir-frontend.cpp

@@ -9,8 +9,9 @@ namespace {
 class MullIRFrontend : public llvm::PassInfoMixin<MullIRFrontend> {
 public:
   llvm::PreservedAnalyses run(llvm::Module &module, llvm::ModuleAnalysisManager &mam) {
+    //    module.print(llvm::errs(), nullptr);
     mull::mutateBitcode(module);
-    module.print(llvm::errs(), nullptr);
+    //    module.print(llvm::errs(), nullptr);
     return llvm::PreservedAnalyses::none();
   }
 };