Merge branch 'main' into semantics-preserving-code-coverage

.github/workflows/build.sh

@@ -148,7 +148,6 @@ case "$(uname)" in
   # The C++ code generated by Dredd may require C++20.
   export DREDD_EXTRA_CXX_ARGS="-std=c++20"
   export DREDD_EXTRA_C_ARGS="-std=c17"
-  DREDD_SKIP_COPY_EXECUTABLE=1 ./scripts/check_single_file_tests.sh
   ;;
 
 "Darwin"*)
@@ -179,11 +178,10 @@ case "$(uname)" in
   rm test/single_file/structured_binding.cc
   rm test/single_file/structured_binding.cc.expected
   rm test/single_file/structured_binding.cc.noopt.expected
-  
+
   # The C++ code generated by Dredd may require recent C/C++ support.
   export DREDD_EXTRA_CXX_ARGS="-std=c++20"
   export DREDD_EXTRA_C_ARGS="-std=c17"
-  DREDD_SKIP_COPY_EXECUTABLE=1 ./scripts/check_single_file_tests.sh
   ;;
 
 "MINGW"*|"MSYS_NT"*)
@@ -239,8 +237,6 @@ case "$(uname)" in
   # The C code generated by Dredd may require recent C/C++ support.
   export DREDD_EXTRA_C_ARGS="/std:c17"
   export DREDD_EXTRA_CXX_ARGS="/std:c++20"
-
-  DREDD_SKIP_COPY_EXECUTABLE=1 ./scripts/check_single_file_tests.sh
   ;;
 
 *)
@@ -249,4 +245,5 @@ case "$(uname)" in
   ;;
 esac
 
+DREDD_SKIP_COPY_EXECUTABLE=1 ./scripts/check_single_file_tests.sh
 DREDD_SKIP_COPY_EXECUTABLE=1 ./scripts/check_execute_tests.py

.github/workflows/c_apps.sh

@@ -104,7 +104,7 @@ pushd curl
       FILES+=("${f}")
   done
 
-  "${DREDD_EXECUTABLE}" $DREDD_SEMANTICS_PRESERVING --mutation-info-file temp.json -p "build/compile_commands.json" "${FILES[@]}"
+  "${DREDD_EXECUTABLE}" $DREDD_SEMANTICS_PRESERVING --mutation-info-file temp.json -p "build" "${FILES[@]}"
   pushd build
     ninja
     # TODO: run some tests
@@ -133,7 +133,7 @@ pushd zstd
   do
     FILES+=("${f}")
   done
-  "${DREDD_EXECUTABLE}" $DREDD_SEMANTICS_PRESERVING --mutation-info-file temp.json -p "temp/compile_commands.json" "${FILES[@]}"
+  "${DREDD_EXECUTABLE}" $DREDD_SEMANTICS_PRESERVING --mutation-info-file temp.json -p "temp" "${FILES[@]}"
   # Build mutated zstd
   make clean
   CFLAGS=-O0 make zstd-release

.github/workflows/cxx_apps.sh

@@ -96,7 +96,7 @@ pushd examples/math
   cp -r math math-original
   cmake -S . -B build -G Ninja -DCMAKE_EXPORT_COMPILE_COMMANDS=ON
   cmake --build build
-  ${DREDD_EXECUTABLE} $DREDD_SEMANTICS_PRESERVING -p build/compile_commands.json --mutation-info-file mutation-info.json math/src/*.cc
+  ${DREDD_EXECUTABLE} $DREDD_SEMANTICS_PRESERVING -p build --mutation-info-file mutation-info.json math/src/*.cc
   ./build/mathtest/mathtest
   NUM_MUTANTS=`python3 ${DREDD_ROOT}/scripts/query_mutant_info.py mutation-info.json --largest-mutant-id`
   EXPECTED_NUM_MUTANTS=1132
@@ -125,7 +125,7 @@ pushd examples/threaded
     [[ -e "$f" ]] || break
     FILES+=("${DREDD_ROOT}/examples/threaded/${f}")
   done
-  ${DREDD_EXECUTABLE} $DREDD_SEMANTICS_PRESERVING --mutation-info-file temp.json -p "${DREDD_ROOT}/examples/threaded/build/compile_commands.json" "${FILES[@]}"
+  ${DREDD_EXECUTABLE} $DREDD_SEMANTICS_PRESERVING --mutation-info-file temp.json -p "${DREDD_ROOT}/examples/threaded/build" "${FILES[@]}"
   cmake --build build
   # Check that the application runs correctly and that there are no data races.
   TSAN_OPTIONS=halt_on_error=1 ./build/threaded > threaded_output.txt
@@ -155,14 +155,14 @@ pushd SPIRV-Tools
     [[ -e "$f" ]] || break
     FILES+=("${DREDD_ROOT}/SPIRV-Tools/${f}")
   done
-  ${DREDD_EXECUTABLE} $DREDD_SEMANTICS_PRESERVING --mutation-info-file mutation-info.json -p "${DREDD_ROOT}/SPIRV-Tools/build/compile_commands.json" "${FILES[@]}"
+  ${DREDD_EXECUTABLE} $DREDD_SEMANTICS_PRESERVING --mutation-info-file mutation-info.json -p "${DREDD_ROOT}/SPIRV-Tools/build" "${FILES[@]}"
   cmake --build build --target test_val_abcde test_val_capability test_val_fghijklmnop test_val_limits test_val_rstuvw
   ./build/test/val/test_val_abcde
   ./build/test/val/test_val_capability
   ./build/test/val/test_val_fghijklmnop
   ./build/test/val/test_val_rstuvw
   NUM_MUTANTS=`python3 ${DREDD_ROOT}/scripts/query_mutant_info.py mutation-info.json --largest-mutant-id`
-  EXPECTED_NUM_MUTANTS=67998
+  EXPECTED_NUM_MUTANTS=68841
   if [ ${NUM_MUTANTS} -ne ${EXPECTED_NUM_MUTANTS} ]
   then
      echo "Found ${NUM_MUTANTS} mutants when mutating the SPIR-V validator source code. Expected ${EXPECTED_NUM_MUTANTS}. If Dredd changed recently, the expected value may just need to be updated, if it still looks sensible. Otherwise, there is likely a problem."
@@ -185,10 +185,10 @@ pushd llvm-project
     [[ -e "$f" ]] || break
     FILES+=("${DREDD_ROOT}/llvm-project/${f}")
   done
-  ${DREDD_EXECUTABLE} $DREDD_SEMANTICS_PRESERVING --mutation-info-file mutation-info.json -p "${DREDD_ROOT}/llvm-project/build/compile_commands.json" "${FILES[@]}"
+  ${DREDD_EXECUTABLE} $DREDD_SEMANTICS_PRESERVING --mutation-info-file mutation-info.json -p "${DREDD_ROOT}/llvm-project/build" "${FILES[@]}"
   cmake --build build --target LLVMInstCombine
   NUM_MUTANTS=`python3 ${DREDD_ROOT}/scripts/query_mutant_info.py mutation-info.json --largest-mutant-id`
-  EXPECTED_NUM_MUTANTS=97675
+  EXPECTED_NUM_MUTANTS=98042
   if [ ${NUM_MUTANTS} -ne ${EXPECTED_NUM_MUTANTS} ]
   then
      echo "Found ${NUM_MUTANTS} mutants when mutating the LLVM source code. Expected ${EXPECTED_NUM_MUTANTS}. If Dredd changed recently, the expected value may just need to be updated, if it still looks sensible. Otherwise, there is likely a problem."

README.md

@@ -94,7 +94,6 @@ You can also enable multiple mutants by setting the environment variable to a co
 To clean up and restore the file `pi.cc` to it's initial state, run
 ```
 rm examples/simple/pi
-rm temp.json
 git checkout HEAD examples/simple/pi.cc
 ```
 
@@ -131,12 +130,16 @@ All tests should pass.
 To mutate all of the `.cc` files in the library use the following command:
 
 ```
-${DREDD_EXECUTABLE} -p build/compile_commands.json math/src/*.cc --mutation-info-file mutant-info.json
+${DREDD_EXECUTABLE} -p build math/src/*.cc --mutation-info-file mutant-info.json
 ```
 
-The `-p` option allows the compilation database generated by CMake above to be passed to Dredd.
+The `-p` option allows the compilation database generated by CMake above to be passed to Dredd: the `compile_commands.json` file in `build` will be used as a compilation database.
 This is so that Dredd knows the correct compiler options to use when processing each source file.
 
+The optional `--mutation-info-file` argument is used to specify a JSON file to which Dredd will output
+machine-readable information about the mutations it applied. We explain how the `mutant-info.json` file created
+via the `--mutation-info-file` argument can be used to query the mutants that Dredd has introduced.
+
 You can run `git status` to see which files have changed, and `git diff` to see
 the effect that Dredd has had on these files. These changes will be hard to understand as they are not intended to be
 readable by humans.
@@ -210,7 +213,7 @@ rm -rf examples/math-original
 
 ## Building Dredd from source
 
-The following instructions have been tested on Ubuntu 20.04.
+The following instructions have been tested on Ubuntu 22.04.
 
 ### Prerequisites
 

scripts/check_one_execute_test.py

@@ -57,23 +57,29 @@
 dredd_result = subprocess.run(cmd)
 if dredd_result.returncode != 0:
     print("Dredd failed.")
+    print(dredd_result.stdout.decode('utf-8'))
     print(dredd_result.stderr.decode('utf-8'))
     sys.exit(1)
 
 # Compile the mutated program.
 COMPILER = os.environ['CXX'] if test_is_cxx else os.environ['CC']
-cmd = [COMPILER, f'harness.{extension}', f'tomutate.{extension}', '-o', 'test_executable']
+EXTRA_COMPILER_ARGS = (os.environ.get('DREDD_EXTRA_CXX_ARGS', '') if test_is_cxx else os.environ.get('DREDD_EXTRA_C_ARGS', '')).split()
+cmd = [COMPILER] + EXTRA_COMPILER_ARGS + [f'harness.{extension}', f'tomutate.{extension}', '-o', 'test_executable']
 compile_result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
 if compile_result.returncode != 0:
     print("Error compiling mutated file.")
+    print(compile_result.stdout.decode('utf-8'))
     print(compile_result.stderr.decode('utf-8'))
+    print("Content of mutated file:")
+    print(open(f'tomutate.{extension}', 'r').read())
     sys.exit(2)
 
 # Run the compiled mutated program with no mutants enabled, and check the result is as expected.
 cmd = [os.getcwd() + os.sep + 'test_executable']
 original_result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
 if original_result.returncode != 0:
     print("Error running non-mutated executable")
+    print(original_result.stdout.decode('utf-8'))
     print(original_result.stderr.decode('utf-8'))
     sys.exit(3)
 actual_original_output = original_result.stdout.decode('utf-8').strip()
@@ -88,6 +94,7 @@
 largest_mutant_id_result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
 if largest_mutant_id_result.returncode != 0:
     print("Error finding largest mutant id.")
+    print(largest_mutant_id_result.stdout.decode('utf-8'))
     print(largest_mutant_id_result.stderr.decode('utf-8'))
     sys.exit(5)
 largest_mutant_id = int(largest_mutant_id_result.stdout.decode('utf-8'))

scripts/query_mutant_info.py

@@ -33,24 +33,19 @@ class MutantInfo:
     instance: Optional[Dict]
 
 
-def build_mapping_for_node(mutation_tree_node: Dict, file_info: Dict, result: Dict[int, MutantInfo]) -> None:
-    for mutation_group in mutation_tree_node["mutationGroups"]:
-        assert len(mutation_group) == 1
-        key: str = next(iter(mutation_group))
-        if key in ["replaceExpr", "replaceUnaryOperator", "replaceBinaryOperator"]:
-            for instance in mutation_group[key]["instances"]:
-                result[instance["mutationId"]] = MutantInfo(key, file_info, mutation_group, instance)
-        else:
-            assert key == "removeStmt"
-            result[mutation_group[key]["mutationId"]] = MutantInfo(key, file_info, mutation_group, None)
-    for child in mutation_tree_node["children"]:
-        build_mapping_for_node(child, file_info, result)
-
-
 def build_mutant_to_node_mapping(json_info: Dict) -> Dict[int, MutantInfo]:
     result: Dict[int, MutantInfo] = {}
     for file_info in json_info["infoForFiles"]:
-        build_mapping_for_node(file_info["mutationTreeRoot"], file_info, result)
+        for mutation_tree_node in file_info["mutationTree"]:
+            for mutation_group in mutation_tree_node["mutationGroups"]:
+                assert len(mutation_group) == 1
+                key: str = next(iter(mutation_group))
+                if key in ["replaceExpr", "replaceUnaryOperator", "replaceBinaryOperator"]:
+                    for instance in mutation_group[key]["instances"]:
+                        result[instance["mutationId"]] = MutantInfo(key, file_info, mutation_group, instance)
+                else:
+                    assert key == "removeStmt"
+                    result[mutation_group[key]["mutationId"]] = MutantInfo(key, file_info, mutation_group, None)
     return result
 
 
@@ -311,7 +306,8 @@ def main() -> int:
                         type=Path)
     parser.add_argument("--largest-mutant-id",
                         help="Show the largest id among all mutants that are present. Mutants are normally numbered "
-                             "contiguously starting from 0, so this is related to the total number of mutants.",
+                             "contiguously starting from 0, so this is related to the total number of mutants. "
+                             "Prints -1 if there are no mutants. ",
                         action='store_true')
     parser.add_argument("--show-info-for-mutant",
                         help="Show information about a given mutant",
@@ -323,7 +319,8 @@ def main() -> int:
     mapping: Dict[int, MutantInfo] = build_mutant_to_node_mapping(json_info)
 
     if args.largest_mutant_id:
-        print(max(list(mapping)))
+        mutant_ids: List[int] = list(mapping.keys())
+        print(-1 if not mutant_ids else max(mutant_ids))
         return 0
 
     if args.show_info_for_mutant is not None:

src/libdredd/include/libdredd/protobufs/dredd.proto

@@ -19,12 +19,18 @@ message MutationInfo { repeated MutationInfoForFile info_for_files = 1; }
 
 message MutationInfoForFile {
   string filename = 1;
-  MutationTreeNode mutation_tree_root = 2;
+
+  // A mutation tree is represented in a flat manner as a list of nodes.
+  // Children of a given node are identified via indices into this list.
+  repeated MutationTreeNode mutation_tree = 2;
 }
 
 message MutationTreeNode {
   repeated MutationGroup mutation_groups = 1;
-  repeated MutationTreeNode children = 2;
+
+  // The children of a node are identified via indices into the flat list
+  // of nodes that represents the mutation tree.
+  repeated uint32 children = 2;
 }
 
 message MutationGroup {

src/libdredd/include/libdredd/util.h

@@ -19,6 +19,8 @@
 
 #include "clang/AST/ASTContext.h"
 #include "clang/AST/Expr.h"
+#include "clang/AST/ParentMapContext.h"
+#include "clang/AST/Stmt.h"
 #include "clang/AST/Type.h"
 #include "clang/Basic/SourceLocation.h"
 #include "clang/Basic/SourceManager.h"
@@ -146,6 +148,24 @@ std::string TypeToUpperLimit(const clang::BuiltinType* type,
 
 std::string TypeToLowerLimit(const clang::BuiltinType* type,
                              const clang::ASTContext& ast_context);
+
+// It is often necessary to ask whether a given statement (which includes
+// expressions) has a parent of a given type. This helper returns nullptr if
+// the given statement has no parent of the template parameter type, and
+// otherwise returns the first parent that does have the template parameter
+// type.
+template <typename RequiredParentT>
+const RequiredParentT* GetFirstParentOfType(const clang::Stmt& stmt,
+                                            clang::ASTContext& ast_context) {
+  for (const auto& parent : ast_context.getParents(stmt)) {
+    const auto* candidate_result = parent.template get<RequiredParentT>();
+    if (candidate_result != nullptr) {
+      return candidate_result;
+    }
+  }
+  return nullptr;
+}
+
 }  // namespace dredd
 
 #endif  // LIBDREDD_UTIL_H

src/libdredd/include_private/include/libdredd/mutate_ast_consumer.h

@@ -67,9 +67,11 @@ class MutateAstConsumer : public clang::ASTConsumer {
   [[nodiscard]] std::string GetMutantTrackingDreddPreludeC(
       int initial_mutation_id) const;
 
-  std::optional<protobufs::MutationTreeNode> ApplyMutations(
-      const MutationTreeNode& mutation_tree_node, int initial_mutation_id,
+  void ApplyMutations(
+      const MutationTreeNode& dredd_mutation_tree_node, int initial_mutation_id,
       clang::ASTContext& context,
+      protobufs::MutationInfoForFile& protobufs_mutation_info_for_file,
+      protobufs::MutationTreeNode& protobufs_mutation_tree_node,
       std::unordered_set<std::string>& dredd_declarations,
       std::unordered_set<std::string>& dredd_macros, bool build_tree);
 

src/libdredd/include_private/include/libdredd/mutate_visitor.h

@@ -106,6 +106,13 @@ class MutateVisitor : public clang::RecursiveASTVisitor<MutateVisitor> {
     return start_location_of_first_function_in_source_file_;
   }
 
+  // Yields the C++ constant-sized arrays, whose size expressions need to be
+  // rewritten.
+  [[nodiscard]] const std::vector<const clang::VarDecl*>&
+  GetConstantSizedArraysToRewrite() const {
+    return constant_sized_arrays_to_rewrite_;
+  }
+
  private:
   // Helper class that uses the RAII pattern to support pushing a new mutation
   // tree node on to the stack of mutation tree nodes used during visitation,
@@ -169,13 +176,10 @@ class MutateVisitor : public clang::RecursiveASTVisitor<MutateVisitor> {
 
   void UpdateStartLocationOfFirstFunctionInSourceFile();
 
-  // It is often necessary to ask whether a given statement (which includes
-  // expressions) has a parent of a given type. This helper returns nullptr if
-  // the given statement has no parent of the template parameter type, and
-  // otherwise returns the first parent that does have the template parameter
-  // type.
-  template <typename RequiredParentT>
-  const RequiredParentT* GetFirstParentOfType(const clang::Stmt& stmt) const;
+  // Mutating an enum constant can be problematic when the enum constant is used
+  // to implicitly construct a C++ object. This helper method allows detecting
+  // this special case, so that it can be ignored.
+  bool IsConversionOfEnumToConstructor(const clang::Expr& expr) const;
 
   const clang::CompilerInstance* compiler_instance_;
   bool optimise_mutations_;
@@ -222,6 +226,10 @@ class MutateVisitor : public clang::RecursiveASTVisitor<MutateVisitor> {
   // tracked, and mutations are not applied to expression nodes whose start
   // location is one of these locations.
   std::set<clang::SourceLocation> var_decl_source_locations_;
+
+  // This records C++ constant-sized array declarations, so that size
+  // expressions can be rewritten with the integers to which they evaluate.
+  std::vector<const clang::VarDecl*> constant_sized_arrays_to_rewrite_;
 };
 
 }  // namespace dredd