feat: Add initial implementation for the generation of an executable file (#92)

* refactor: Fix llvm usage with memory leak

* Add initial option for generating executable

Author: xmnlab

.gitignore

@@ -45,6 +45,8 @@ bin/*
 
 # arx
 arxc
+*.arxc
+*.arx.o
 
 # IDE
 .vscode

.makim.yaml

@@ -78,7 +78,7 @@ groups:
           - target: build.release
             args:
               build-type: "debug"
-              extras: "-Db_coverage=true -Doptimization=0"  # -Db_sanitize=address
+              extras: "-Db_coverage=true -Doptimization=0"  #  -Db_sanitize=address
               clean: {{ args.clean }}
               asan-options: "fast_unwind_on_malloc=0:verbosity=2:log_threads=1"
               lsan-options: "fast_unwind_on_malloc=0:verbosity=2:log_threads=1"
@@ -204,7 +204,7 @@ groups:
           # another options: halt_on_error=true
           # it could be a useful flag: -fsanitize-recover
           export LSAN_OPTIONS=verbosity=2:log_threads=1
-          export ASAN_OPTIONS=verbosity=2:check_initialization_order=1:detect_leaks=1
+          export ASAN_OPTIONS=verbosity=2:check_initialization_order=1:detect_leaks=1:fast_unwind_on_malloc=0
 
           if [[ "{{ args.debug  }}" == "True" ]]; then
             GDB="gdb --args"
@@ -216,13 +216,14 @@ groups:
           # load utils functions
           . "${TEST_DIR_PATH}/scripts/utils.sh"
 
-          ARX="${GDB} ./build/arx"
+          ARX="${GDB} ./build/arx --build-lib"
           MAIN_EXE="${TMP_DIR}/main"
 
           LD_LIBRARY_PATH=${CONDA_PREFIX}/lib:$LD_LIBRARY_PATH
           export LD_LIBRARY_PATH
 
-          CLANG_EXTRAS="$(llvm-config --cxxflags)"
+          CLANG_EXTRAS="-fno-omit-frame-pointer"
+          CLANG_EXTRAS="${CLANG_EXTRAS} $(llvm-config --cxxflags)"
           CLANG_EXTRAS="${CLANG_EXTRAS} $(llvm-config --ldflags --libs core executionengine interpreter analysis native bitwriter)"
           CLANG_EXTRAS="${CLANG_EXTRAS} $(pkg-config --cflags --libs arrow-glib)"
           CLANG_EXTRAS=$(echo "${CLANG_EXTRAS}" | sed s/-std=c++14/-std=c++20/g)
@@ -248,7 +249,6 @@ groups:
             echo ">>> RUN THE PROGRAM:"
             chmod +x ${MAIN_EXE}
             ${GDB} ${MAIN_EXE}
-            break
           done
 
       gen-ast:

meson.build

@@ -9,7 +9,7 @@ project('arx', 'cpp', 'c',
 
 add_global_arguments(
   '-Wpedantic',
-  '-Wno-c++98-compat-pedantic',
+  # '-Wno-c++98-compat-pedantic',
   '-Wno-padded',
   '-Wno-missing-prototypes',
   '-Wshadow',
@@ -33,10 +33,19 @@ PROJECT_PATH = meson.source_root()
 
 cxx = meson.get_compiler('cpp')
 
+llvm_modules = [
+  'core',
+  'executionengine',
+  'object',
+  'orcjit',
+  'support',
+  'native',
+]
+
 deps = [
   dependency('arrow'),
   dependency('arrow-glib'),
-  dependency('llvm', version : '>=14.0.0'),
+  dependency('llvm', version : '>=15.0.0', modules : llvm_modules),
   dependency('CLI11'),
   dependency('threads'),
   dependency('glog'),

src/codegen/arx-llvm.cpp

@@ -22,3 +22,5 @@ llvm::Type* ArxLLVM::DOUBLE_TYPE;
 llvm::Type* ArxLLVM::FLOAT_TYPE;
 llvm::Type* ArxLLVM::INT8_TYPE;
 llvm::Type* ArxLLVM::INT32_TYPE;
+
+extern bool IS_BUILD_LIB = false;  // default value

src/codegen/arx-llvm.h

@@ -25,3 +25,5 @@ class ArxLLVM {
   static llvm::Type* INT8_TYPE;
   static llvm::Type* INT32_TYPE;
 };
+
+extern bool IS_BUILD_LIB;

src/codegen/ast-to-llvm-ir.cpp

@@ -297,7 +297,7 @@ auto ASTToLLVMIRVisitor::Initialize() -> void {
  *
  * @param ast The AST tree object.
  */
-auto compile_llvm_ir(TreeAST& ast) -> void {
+auto compile_llvm_ir(TreeAST& ast) -> int {
   auto codegen = std::make_unique<ASTToLLVMIRVisitor>(ASTToLLVMIRVisitor());
 
   Lexer::getNextToken();
@@ -356,19 +356,19 @@ auto compile_llvm_ir(TreeAST& ast) -> void {
 
   // Print out all of the generated code.
   ArxLLVM::module->print(llvm::errs(), nullptr);
+
+  return 0;
 }
 
 /**
  * @brief Open the Arx shell.
  *
  */
-auto open_shell_llvm_ir() -> void {
+auto open_shell_llvm_ir() -> int {
   // Prime the first token.
   fprintf(stderr, "Arx %s \n", ARX_VERSION.c_str());
   fprintf(stderr, ">>> ");
 
   auto ast = std::make_unique<TreeAST>(TreeAST());
-  compile_llvm_ir(*ast);
-
-  exit(0);
+  return compile_llvm_ir(*ast);
 }

src/codegen/ast-to-llvm-ir.h

@@ -31,8 +31,8 @@
 #include "codegen/ast-to-object.h"
 #include "parser.h"
 
-auto compile_llvm_ir(TreeAST&) -> void;
-auto open_shell_llvm_ir() -> void;
+auto compile_llvm_ir(TreeAST&) -> int;
+auto open_shell_llvm_ir() -> int;
 
 class ASTToLLVMIRVisitor : public ASTToObjectVisitor {
  public:

src/codegen/ast-to-object.cpp

@@ -1,4 +1,17 @@
-#include "codegen/ast-to-object.h"      // for ASTToObjectVisitor, compile_o...
+#include <cassert>  // for assert
+#include <cstdio>   // for fprintf, stderr, fputc
+#include <cstdlib>  // for exit
+#include <fstream>  // for operator<<
+#include <iostream>
+#include <map>      // for map, operator==, _Rb_tree_ite...
+#include <memory>   // for unique_ptr, allocator, make_u...
+#include <numeric>  // for accumulate
+#include <sstream>
+#include <string>        // for string, operator<=>, operator+
+#include <system_error>  // for error_code
+#include <utility>       // for pair, move
+#include <vector>        // for vector
+
 #include <glog/logging.h>               // for COMPACT_GOOGLE_LOG_INFO, LOG
 #include <llvm/ADT/APFloat.h>           // for APFloat
 #include <llvm/ADT/iterator_range.h>    // for iterator_range
@@ -26,25 +39,32 @@
 #include <llvm/Support/TargetSelect.h>  // for InitializeAllAsmParsers, Init...
 #include <llvm/Target/TargetMachine.h>  // for TargetMachine
 #include <llvm/Target/TargetOptions.h>  // for TargetOptions
-#include <cassert>                      // for assert
-#include <cstdio>                       // for fprintf, stderr, fputc
-#include <cstdlib>                      // for exit
-#include <fstream>                      // for operator<<
-#include <map>                          // for map, operator==, _Rb_tree_ite...
-#include <memory>                       // for unique_ptr, allocator, make_u...
-#include <string>                       // for string, operator<=>, operator+
-#include <system_error>                 // for error_code
-#include <utility>                      // for pair, move
-#include <vector>                       // for vector
-#include "codegen/arx-llvm.h"           // for ArxLLVM
-#include "error.h"                      // for LogErrorV
-#include "lexer.h"                      // for Lexer
-#include "parser.h"                     // for PrototypeAST, ExprAST, ForExp...
+
+#include "codegen/arx-llvm.h"       // for ArxLLVM
+#include "codegen/ast-to-object.h"  // for ASTToObjectVisitor, compile_o...
+#include "error.h"                  // for LogErrorV
+#include "io.h"                     // for ArxFile
+#include "lexer.h"                  // for Lexer
+#include "parser.h"                 // for PrototypeAST, ExprAST, ForExp...
 
 namespace llvm {
   class Value;
 }
 
+std::string string_join(
+  const std::vector<std::string>& elements, const std::string& delimiter) {
+  if (elements.empty()) {
+    return "";
+  }
+
+  std::string str;
+  for (auto v : elements) {
+    str += v + delimiter;
+  }
+  str = str.substr(0, str.size() - delimiter.size());
+  return str;
+}
+
 extern std::string INPUT_FILE;
 extern std::string OUTPUT_FILE;
 extern std::string ARX_VERSION;
@@ -635,7 +655,7 @@ extern "C" DLLEXPORT auto printd(double X) -> double {
  *
  * @param tree_ast The AST tree object.
  */
-auto compile_object(TreeAST& tree_ast) -> void {
+auto compile_object(TreeAST& tree_ast) -> int {
   auto codegen = std::make_unique<ASTToObjectVisitor>(ASTToObjectVisitor());
 
   Lexer::getNextToken();
@@ -669,7 +689,7 @@ auto compile_object(TreeAST& tree_ast) -> void {
   // TargetRegistry or we have a bogus target triple.
   if (!Target) {
     llvm::errs() << Error;
-    exit(1);
+    return 1;
   }
 
   auto CPU = "generic";
@@ -692,40 +712,95 @@ auto compile_object(TreeAST& tree_ast) -> void {
   std::error_code EC;
 
   if (OUTPUT_FILE == "") {
-    OUTPUT_FILE = "./output.o";
+    OUTPUT_FILE = INPUT_FILE + ".o";
   }
 
   llvm::raw_fd_ostream dest(OUTPUT_FILE, EC, llvm::sys::fs::OF_None);
 
   if (EC) {
     llvm::errs() << "Could not open file: " << EC.message();
-    exit(1);
+    return 1;
   }
 
   llvm::legacy::PassManager pass;
+
   auto FileType = llvm::CGFT_ObjectFile;
 
   if (TheTargetMachine->addPassesToEmitFile(pass, dest, nullptr, FileType)) {
     llvm::errs() << "TheTargetMachine can't emit a file of this type";
-    exit(1);
+    return 1;
   }
 
   pass.run(*ArxLLVM::module);
   dest.flush();
+
+  if (IS_BUILD_LIB) {
+    return 0;
+  }
+
+  // generate an executable file
+
+  std::string linker_path = "clang++";
+  std::string executable_path = INPUT_FILE + "c";
+  // note: it just have a purpose to demonstrate an initial implementation
+  //       it will be improved in a follow-up PR
+  std::string content =
+    "#include <iostream>\n"
+    "int main() {\n"
+    "  std::cout << \"ARX[WARNING]: "
+    "This is an empty executable file\" << std::endl;\n"
+    "}\n";
+
+  std::string main_cpp_path = ArxFile::create_tmp_file(content);
+
+  if (main_cpp_path == "") {
+    llvm::errs() << "ARX[FAIL]: Executable file was not created.";
+    return 1;
+  }
+
+  /* Example (running it from a shell prompt):
+     clang++ \
+       ${CLANG_EXTRAS} \
+       ${DEBUG_FLAGS} \
+       -fPIC \
+       -std=c++20 \
+       "${TEST_DIR_PATH}/main-objects/${test_name}.cpp" \
+       ${OBJECT_FILE} \
+       -o "${TMP_DIR}/main"
+  */
+
+  std::vector<std::string> compiler_args{
+    "-fPIC", "-std=c++20", main_cpp_path, OUTPUT_FILE, "-o", executable_path};
+
+  // Add any additional compiler flags or include paths as needed
+  // compiler_args.push_back("-I/path/to/include");
+
+  std::string compiler_cmd =
+    linker_path + " " + string_join(compiler_args, " ");
+
+  std::cout << "ARX[INFO]: " << compiler_cmd << std::endl;
+  int compile_result = system(compiler_cmd.c_str());
+
+  // ArxFile::delete_file(main_cpp_path);
+
+  if (compile_result != 0) {
+    llvm::errs() << "failed to compile and link object file";
+    exit(1);
+  }
+
+  return 0;
 }
 
 /**
  * @brief Open the Arx shell.
  *
  */
-auto open_shell_object() -> void {
+auto open_shell_object() -> int {
   // Prime the first token.
   fprintf(stderr, "Arx %s \n", ARX_VERSION.c_str());
   fprintf(stderr, ">>> ");
 
   auto ast = std::make_unique<TreeAST>(TreeAST());
 
-  compile_object(*ast);
-
-  exit(0);
+  return compile_object(*ast);
 }

src/codegen/ast-to-object.h

@@ -22,8 +22,8 @@ namespace llvm {
   class Value;
 }
 
-auto compile_object(TreeAST&) -> void;
-auto open_shell_object() -> void;
+auto compile_object(TreeAST&) -> int;
+auto open_shell_object() -> int;
 
 class ASTToObjectVisitor : public Visitor {
  public:

src/codegen/ast-to-stdout.cpp

@@ -233,7 +233,7 @@ void ASTToOutputVisitor::visit(FunctionAST& expr) {
   std::cout << this->indentation() << ")" << std::endl;
 }
 
-auto print_ast(TreeAST& ast) -> void {
+auto print_ast(TreeAST& ast) -> int {
   auto visitor_print =
     std::make_unique<ASTToOutputVisitor>(ASTToOutputVisitor());
 
@@ -246,4 +246,5 @@ auto print_ast(TreeAST& ast) -> void {
   }
 
   std::cout << "]" << std::endl;
+  return 0;
 }