Compile vararg

README.md

@@ -0,0 +1,59 @@
+# Simple C-compiler
+
+This is a simple work-in-progress C-compiler (or C-like) written in C++ with the
+goal of learning to write C++ better in the future, and to provide reference for
+my knowledge about modern C++ programming.
+
+As of writing, a simple fibonacci sequence program is already possible to be
+compiled and executed, and can be viewed via [`test.c`](./test.c).
+
+As far as compiler-design goes, this project still falls behind my other
+project, [Reid-LLVM](https://git.teascade.net/teascade/reid-llvm), which is
+significantly more capable and more robust.
+
+## Structure of the program
+
+The program is structured into several different staged, all of which are
+orchestrated via main.cpp.
+
+Currently the stages are as follows:
+
+1. Firstly, the program is **tokenized**. This stage could also be called the
+   lexer, depending on your preference. In this stage, the source code for the
+   program is transformed into discrete tokens which can then be used during the
+   parsing phase easier than regular text. The code for this stage is mostly in
+   [`src/tokens.cpp`](src/tokens.cpp).
+2. **TODO:** Preprocessing stage hasn't yet been developed, but it will go here.
+3. Then the program is **parsed**. This is the stage where the tokens from the
+   previous stage(s) are converted into an Abstract Syntax Tree (AST), which is
+   a format that is easier for the computer to process. The AST itself lives in
+   [`src/ast.h`](src/ast.h), and the code for the parsing phase lives in
+   [`src/parsing.cpp`](src/parsing.cpp).
+4. **TODO:** Typechecking phase hasn't yet been developed, but it will go here.
+5. Finally the program is **compiled**, or in other words **code-generated**,
+   hence why this is the **codegen** stage. This is where the AST from the
+   previous stages is taken and LLVM Intermediate Representation is produced
+   using LLVM-bindings. 
+
+## Compiling and running the program
+
+In order to compile the program, you need the following:
+- CMake
+- C++20 (or newer) capable compiler
+- LLVM 21.1.0 or newer
+
+And in order to execute the program which is compiled you also need:
+- LLVM 21.1.0 or newer (as it is dynamically linked)
+- `whereis`-utility in `$PATH`
+- `ld`-utility in `$PATH`
+
+Then, to compile the program you run:
+```sh
+cmake -Bbuild
+make -C build
+```
+
+and to run the program, run simply `./build/llvm_c_compiler`. This will read a
+file called `test.c` from `$PWD`, and produce two files (`test.o` and `test`).
+An executable file called `test` is produced as a result, compiled from the
+original `test.c`.

src/codegen.cpp

@@ -187,7 +187,7 @@ namespace AST {
             param_ty_ptrs.push_back(std::move(param.second));
         }
 
-        auto fn_ty_ptr = std::shared_ptr<types::Type>{ new types::FunctionType{ ret_ty_ptr, param_ty_ptrs } };
+        auto fn_ty_ptr = std::shared_ptr<types::Type>{ new types::FunctionType{ ret_ty_ptr, param_ty_ptrs, this->m_is_vararg } };
 
         auto fn_ty = fn_ty_ptr->codegen(builder);
         auto function = llvm::Function::Create(
@@ -254,7 +254,7 @@ namespace types {
 
         auto ret_ty = this->m_ret_ty->codegen(builder);
 
-        return llvm::FunctionType::get(ret_ty, params, false);
+        return llvm::FunctionType::get(ret_ty, params, this->m_vararg);
     }
 
     llvm::Type* PointerType::codegen(codegen::Builder& builder) {

src/main.cpp

@@ -159,11 +159,18 @@ std::optional<CompileOutput> compile(std::string_view in_filename) {
     };
 }
 
+struct ClosePipeDeleter {
+    // Note 2: Consider adding noexcept.
+    void operator()(FILE* file) const {
+        pclose(file);
+    }
+};
+
 /// @brief Executes a given command and returns the output as std::string
 std::string exec(const char* cmd) {
     std::array<char, 128> buffer;
     std::string result;
-    std::unique_ptr<FILE, decltype(&pclose)> pipe(popen(cmd, "r"), pclose);
+    std::unique_ptr<FILE, ClosePipeDeleter> pipe(popen(cmd, "r"), ClosePipeDeleter{});
     if (!pipe) {
         throw std::runtime_error("popen() failed!");
     }

src/parsing.cpp

@@ -42,7 +42,7 @@ namespace parsing {
 
                 stream.m_position = inner.m_position;
 
-                return std::move(returned);
+                return returned;
             }
             catch (std::runtime_error& error) {
                 return std::string{ error.what() };

src/types.cpp

@@ -109,6 +109,12 @@ namespace types {
             out << param->formatted();
         }
 
+        if (this->m_vararg) {
+            if (counter > 0)
+                out << ", ";
+            out << "...";
+        }
+
         out << ") -> " << this->m_ret_ty->formatted();
         return out.str();
     }

src/types.h

@@ -59,9 +59,10 @@ namespace types {
     private:
         std::shared_ptr<Type> m_ret_ty;
         std::vector<std::shared_ptr<Type>> m_param_tys;
+        bool m_vararg;
     public:
-        FunctionType(std::shared_ptr<Type> ret_ty, std::vector<std::shared_ptr<Type>> param_tys)
-            : m_ret_ty{ std::move(ret_ty) }, m_param_tys{ std::move(param_tys) } {
+        FunctionType(std::shared_ptr<Type> ret_ty, std::vector<std::shared_ptr<Type>> param_tys, bool vararg)
+            : m_ret_ty{ std::move(ret_ty) }, m_param_tys{ std::move(param_tys) }, m_vararg{ vararg } {
         }
         virtual ~FunctionType() override = default;
         virtual std::string formatted() override;