From 4f1903aa4582f5cd96a91f168e4e9c5fb42a6136 Mon Sep 17 00:00:00 2001
From: Per Lundberg <perlun@gmail.com>
Date: Thu, 11 Apr 2024 21:52:38 +0300
Subject: [PATCH] (compiler) Add `-c` flag for "compile and assemble only"

This commit is a small step towards #454. By adding this flag, it's
possible to "use Perlang as a transpiler" only. Okay, a _little_ more
than a transpiler, since it also emits ELF object files. :) But still,
it helps us towards a state where we can gradually, incrementally
rewrite parts of the Perlang compiler in Perlang. Oh, the joy! Oh, the
glory!
---
 Perlang.sln                                   |  2 +-
 release-notes/v0.5.0.md                       |  3 +
 src/Perlang.ConsoleApp/Program.cs             | 69 ++++++++++++++++++-
 .../Compiler/PerlangCompiler.cs               | 62 +++++++++++++++--
 4 files changed, 129 insertions(+), 7 deletions(-)

diff --git a/Perlang.sln b/Perlang.sln
index d68cb03a..50a2704f 100644
--- a/Perlang.sln
+++ b/Perlang.sln
@@ -29,7 +29,7 @@ Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "Perlang.Tests.Architecture"
 EndProject
 Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "Perlang.Compiler", "src\Perlang.Compiler\Perlang.Compiler.csproj", "{7E7CF7BF-F792-49C9-BB16-67FE856D8D51}"
 EndProject
-Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "Solution Items", "Solution Items", "{47B4A37F-6999-4510-9815-418E03D7B39E}"
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "_Solution Items", "_Solution Items", "{47B4A37F-6999-4510-9815-418E03D7B39E}"
 	ProjectSection(SolutionItems) = preProject
 		README.md = README.md
 	EndProjectSection
diff --git a/release-notes/v0.5.0.md b/release-notes/v0.5.0.md
index c069787e..1af3ccd3 100644
--- a/release-notes/v0.5.0.md
+++ b/release-notes/v0.5.0.md
@@ -7,6 +7,7 @@
 - `BigInt::pow`: Detect negative exponents and throw an exception [[#450][450]]
 - Implement initial native `ASCIIString` class [[#451][451]]
 - Wrap `ASCIIString` in `std::shared_ptr<T>` [[#453][453]]
+- Add `-c` flag for "compile and assemble only" [[#455][455]]
 
 ### Changed
 #### Data types
@@ -27,3 +28,5 @@
 [449]: https://github.com/perlang-org/perlang/pull/449
 [450]: https://github.com/perlang-org/perlang/pull/450
 [451]: https://github.com/perlang-org/perlang/pull/451
+[453]: https://github.com/perlang-org/perlang/pull/453
+[455]: https://github.com/perlang-org/perlang/pull/455
diff --git a/src/Perlang.ConsoleApp/Program.cs b/src/Perlang.ConsoleApp/Program.cs
index 0409a157..f4d32e4f 100644
--- a/src/Perlang.ConsoleApp/Program.cs
+++ b/src/Perlang.ConsoleApp/Program.cs
@@ -121,6 +121,7 @@ public static int MainWithCustomConsole(string[] args, IPerlangConsole console)
         {
             var versionOption = new Option<bool>(new[] { "--version", "-v" }, "Show version information");
             var detailedVersionOption = new Option<bool>("-V", "Show detailed version information");
+            var compileAndAssembleOnlyOption = new Option<bool>("-c", "Compile and assemble to a .o file, but do not produce and execute an executable");
             var printOption = new Option<string>("-p", "Parse a single-line script and output a human-readable version of the AST") { ArgumentHelpName = "script" };
             var noWarnAsErrorOption = new Option<string>("-Wno-error", "Treats specified warning as a warning instead of an error.") { ArgumentHelpName = "error" };
 
@@ -143,6 +144,7 @@ public static int MainWithCustomConsole(string[] args, IPerlangConsole console)
             {
                 versionOption,
                 detailedVersionOption,
+                compileAndAssembleOnlyOption,
                 printOption,
                 noWarnAsErrorOption
             };
@@ -200,9 +202,32 @@ public static int MainWithCustomConsole(string[] args, IPerlangConsole console)
                         // TODO: Tried to fix this using some logic in the rootCommand.AddValidator() lambda, but I
                         // TODO: couldn't get it working. Since this is going to be rewritten in Perlang at some point
                         // TODO: anyway, let's not spend too much time thinking about it.
-                        Console.Error.WriteLine("One of the -p <script> or <script-name> arguments must be provided");
+                        Console.Error.WriteLine("ERROR: One of the -p <script> or <script-name> arguments must be provided");
                         return Task.FromResult(1);
                     }
+                    else if (parseResult.HasOption(compileAndAssembleOnlyOption))
+                    {
+                        string scriptName = parseResult.GetValueForArgument(scriptNameArgument);
+
+                        // Console.Out.WriteLine(parseResult.Tokens);
+                        // Console.Out.WriteLine(parseResult.GetValueForArgument(scriptArguments));
+                        //
+                        // if (parseResult.Tokens.Count != 1)
+                        // {
+                        //     Console.Error.WriteLine("ERROR: When the -c option is used, no script arguments can be provided since the Perlang script/program will not get executed");
+                        //     return Task.FromResult(1);
+                        // }
+
+                        var program = new Program(
+                            replMode: false,
+                            standardOutputHandler: console.WriteStdoutLine,
+                            disabledWarningsAsErrors: disabledWarningsAsErrorsList,
+                            experimentalCompilation: PerlangMode.ExperimentalCompilation
+                        );
+
+                        int result = program.CompileAndAssembleFile(scriptName);
+                        return Task.FromResult(result);
+                    }
                     else
                     {
                         string scriptName = parseResult.GetValueForArgument(scriptNameArgument);
@@ -239,6 +264,14 @@ public static int MainWithCustomConsole(string[] args, IPerlangConsole console)
                 })
             };
 
+            rootCommand.AddValidator(result =>
+            {
+                if (result.HasOption(compileAndAssembleOnlyOption) && result.HasOption(printOption))
+                {
+                    result.ErrorMessage = "Error: the -c and -p options are mutually exclusive";
+                }
+            });
+
             rootCommand.AddArgument(scriptNameArgument);
             rootCommand.AddArgument(scriptArguments);
 
@@ -320,6 +353,35 @@ private int RunFile(string path)
             return (int)ExitCodes.SUCCESS;
         }
 
+        private int CompileAndAssembleFile(string path)
+        {
+            if (!File.Exists(path))
+            {
+                standardErrorHandler(Lang.String.from($"Error: File {path} not found"));
+                return (int)ExitCodes.FILE_NOT_FOUND;
+            }
+
+            var bytes = File.ReadAllBytes(path);
+            string source = Encoding.UTF8.GetString(bytes);
+
+            CompileAndAssemble(source, path, CompilerWarning);
+
+            // Indicate an error in the exit code.
+            if (hadError)
+            {
+                return (int)ExitCodes.ERROR;
+            }
+
+            // Should never happen, but _if_ it does, it's surely better that we return with non-zero rather than falsely
+            // imply success.
+            if (hadRuntimeError)
+            {
+                return (int)ExitCodes.RUNTIME_ERROR;
+            }
+
+            return (int)ExitCodes.SUCCESS;
+        }
+
         internal int Run(string source, CompilerWarningHandler compilerWarningHandler)
         {
             object? result = interpreter.Eval(source, ScanError, ParseError, NameResolutionError, ValidationError, ValidationError, compilerWarningHandler);
@@ -337,6 +399,11 @@ private void CompileAndRun(string source, string path, CompilerWarningHandler co
             compiler.CompileAndRun(source, path, CompilerFlags.None, ScanError, ParseError, NameResolutionError, ValidationError, ValidationError, compilerWarningHandler);
         }
 
+        private void CompileAndAssemble(string source, string path, CompilerWarningHandler compilerWarningHandler)
+        {
+            compiler.CompileAndAssemble(source, path, CompilerFlags.None, ScanError, ParseError, NameResolutionError, ValidationError, ValidationError, compilerWarningHandler);
+        }
+
         private void ParseAndPrint(string source)
         {
             string? result = interpreter.Parse(source, ScanError, ParseError);
diff --git a/src/Perlang.Interpreter/Compiler/PerlangCompiler.cs b/src/Perlang.Interpreter/Compiler/PerlangCompiler.cs
index 4e13c7ed..f6bff698 100644
--- a/src/Perlang.Interpreter/Compiler/PerlangCompiler.cs
+++ b/src/Perlang.Interpreter/Compiler/PerlangCompiler.cs
@@ -272,6 +272,47 @@ private void RegisterGlobalClasses()
         return executablePath;
     }
 
+    /// <summary>
+    /// Compiles and assembles the given Perlang program to a .o file (ELF object file).
+    /// </summary>
+    /// <param name="source">The Perlang program to compile.</param>
+    /// <param name="path">The path to the source file (used for generating error messages).</param>
+    /// <param name="compilerFlags">One or more <see cref="CompilerFlags"/> to use.</param>
+    /// <param name="scanErrorHandler">A handler for scanner errors.</param>
+    /// <param name="parseErrorHandler">A handler for parse errors.</param>
+    /// <param name="nameResolutionErrorHandler">A handler for resolve errors.</param>
+    /// <param name="typeValidationErrorHandler">A handler for type validation errors.</param>
+    /// <param name="immutabilityValidationErrorHandler">A handler for immutability validation errors.</param>
+    /// <param name="compilerWarningHandler">A handler for compiler warnings.</param>
+    /// <returns>The path to the compiled .o file. Note that this will be non-null even on unsuccessful
+    /// compilation.</returns>
+    public string? CompileAndAssemble(
+        string source,
+        string path,
+        CompilerFlags compilerFlags,
+        ScanErrorHandler scanErrorHandler,
+        ParseErrorHandler parseErrorHandler,
+        NameResolutionErrorHandler nameResolutionErrorHandler,
+        ValidationErrorHandler typeValidationErrorHandler,
+        ValidationErrorHandler immutabilityValidationErrorHandler,
+        CompilerWarningHandler compilerWarningHandler)
+    {
+        string? executablePath = Compile(
+            source,
+            path,
+            compilerFlags,
+            scanErrorHandler,
+            parseErrorHandler,
+            nameResolutionErrorHandler,
+            typeValidationErrorHandler,
+            immutabilityValidationErrorHandler,
+            compilerWarningHandler,
+            compileAndAssembleOnly: true
+        );
+
+        return executablePath;
+    }
+
     /// <summary>
     /// Compiles the given Perlang program to an executable.
     /// </summary>
@@ -284,6 +325,8 @@ private void RegisterGlobalClasses()
     /// <param name="typeValidationErrorHandler">A handler for type validation errors.</param>
     /// <param name="immutabilityValidationErrorHandler">A handler for immutability validation errors.</param>
     /// <param name="compilerWarningHandler">A handler for compiler warnings.</param>
+    /// <param name="compileAndAssembleOnly">A flag which enables functionality similar to `gcc -c`; compile and assemble
+    /// the given program, but do not generate an executable (and inherently do not execute it).</param>
     /// <returns>The path to the generated executable file, or `null` if compilation failed.</returns>
     private string? Compile(
         string source,
@@ -294,15 +337,16 @@ private void RegisterGlobalClasses()
         NameResolutionErrorHandler nameResolutionErrorHandler,
         ValidationErrorHandler typeValidationErrorHandler,
         ValidationErrorHandler immutabilityValidationErrorHandler,
-        CompilerWarningHandler compilerWarningHandler)
+        CompilerWarningHandler compilerWarningHandler,
+        bool compileAndAssembleOnly = false)
     {
         string targetCppFile = Path.ChangeExtension(path, ".cc");
 
 #if _WINDOWS
         // clang is very unlikely to have been available on Windows anyway, but why not...
-        string targetExecutable = Path.ChangeExtension(targetCppFile, "exe");
+        string targetExecutable = Path.ChangeExtension(targetCppFile, compileAndAssembleOnly ? "obj" : "exe");
 #else
-        string targetExecutable = Path.ChangeExtension(targetCppFile, null);
+        string targetExecutable = Path.ChangeExtension(targetCppFile, compileAndAssembleOnly ? "o" : null);
 #endif
 
         // TODO: Check the creation time of *all* dependencies here, including the stdlib (both .so/.dll and .h files
@@ -453,6 +497,13 @@ private void RegisterGlobalClasses()
 
         try
         {
+            if (compileAndAssembleOnly && methods["main"].MethodBody.Length == 0)
+            {
+                // We generate a default "main" method elsewhere. In -c mode, we want to remove this method if it's empty,
+                // to avoid conflicts with the "real" main method of the program the .o file is being linked into.
+                methods.Remove("main");
+            }
+
             // The AST traverser writes its output to the transpiledSource field.
             Compile(statements);
 
@@ -539,6 +590,7 @@ private void RegisterGlobalClasses()
                     "--std=c++17",
 
                     "-I", Path.Combine(stdlibPath, "include"),
+                    compileAndAssembleOnly ? "-c" : "",
                     "-o", targetExecutable,
 
                     // Useful while debugging
@@ -604,10 +656,10 @@ private void RegisterGlobalClasses()
                     targetCppFile,
 
                     // TODO: Support Windows static libraries as well
-                    Path.Combine(stdlibPath, "lib/libstdlib.a"),
+                    compileAndAssembleOnly ? "" : Path.Combine(stdlibPath, "lib/libstdlib.a"),
 
                     // Needed by the Perlang stdlib
-                    "-lm"
+                    compileAndAssembleOnly ? "" : "-lm"
                 },
                 RedirectStandardOutput = true,
                 RedirectStandardError = true