[llvm-rc] Add .rc scripts tokenizer.

[mnbvmar]

Summary

This extends the shell of llvm-rc tool with the ability of tokenization of the input files. Currently, ASCII and ASCII-compatible UTF-8 files are supported.

This is an LLVM reimplementation of PoC tool by Nico Weber.

[zturner]

llvm/test/tools/llvm-rc/Inputs/tokens.rc

2

What about comment markers and preprocessor directives, pragma directives, and preprocessor operators?

llvm/tools/llvm-rc/ResourceScriptToken.cpp

27

No reason to past StringRef by const reference, since it is basically a const reference by design. Just pass by value.

35

You can use TokenValue.getAsInteger()

44

llvm style guide does not allow functions in anonymous namespace, only types and classes. This is because making functions global static do the same thing only a little bit "better". So instead, please make all these functions file static global functions.

llvm/tools/llvm-rc/ResourceScriptToken.h

56

Do we need to support 64-bit integer literals? I would imagine so, but I'm not sure.

58–59

StringRef is an immutable data structure. So there is no reason to return a reference to one, because even with a reference you can't modify it.

64

Can you put this file inside of the llvm namespace? Then you don't need these explicit namespace qualifications.

76

I would suggest returning an Expected<std::vector<RCToken>> here, and dropping the Error parameter.

[ecbeckmann]

llvm/test/tools/llvm-rc/Inputs/tokens.rc

2

We decided to get the rest of rc.exe working before worrying about preprocessing, and assume we are only dealing with .rc files that don't need preprocessing. Once this part is finished we can integrate into the chromium build system using clang preprocessor piped into llvm-rc. Finally we will see about busy-boxing a clang invokation (like clang-rc) that handles preprocessing.

[mnbvmar]

Fixed style/nits; tokenizer starts using Error/Expected.

[mnbvmar]

Forgot to remove llvm::'s.

[mnbvmar]

llvm/tools/llvm-rc/ResourceScriptToken.h

56

Looking through the MS binary resource files (.res) documentation, I couldn't find any QWORD fields. Nico might know more about this.

[ecbeckmann]

llvm/tools/llvm-rc/ResourceScriptToken.h

56

It looks like when integer IDs/Types are given that overflow 32bits, rc.exe quietly ignores the most significant portions of the number, choosing the LS 32 bits.

[ecbeckmann]

llvm/tools/llvm-rc/ResourceScriptToken.h

56

Also by experimenting with dimensions given to DIALOG, it seems that integers occurring in resource definition statements are treated the same way. In fact, rc only uses 16 bit numbers, both for type/name ids and also for resource definition params.

[mnbvmar]

llvm/tools/llvm-rc/ResourceScriptToken.h

56

What's even more weird, when a number divisible by 2^16 (65536, 131072, ...) is given as a type/name id, a Unicode string representing the number is produced by original rc.exe. However, it's not always the case - in some other cases, where a 16-bit number is to be given, the tool just outputs LS 16 bits (i.e., 0).

However, the correct behavior is to reject the input as these numbers should fit in WORD.

[rnk]

This seems basically done. @zturner do you have any unaddressed comments?

llvm/tools/llvm-rc/ResourceScriptToken.cpp

46

I think this namespace is unnecessary

56

assert will be compiled away in release builds, so this will just return uninitialized data. :( Unfortunately, the accepted pattern is:

uint32_t Result;
bool S = rcGetAsInteger(TokenValue, Result);
(void)S; // silence -Wunused-variable in NDEBUG builds
assert(S);

Compilers are dumb.

[mnbvmar]

Remove unnecessary namespace annotation; fix the non-evaluated assert() bug.

[rnk]

lgtm

llvm/test/tools/llvm-rc/Inputs/tokens.rc

+1 + 2 - 3214L & 0x120894 032173 2|&~+(-7){0xabcdef 0xABCDEFl} Begin End
+He11o LLVM
+
+"RC string test.",L"Another RC string test.'&{",42,100
+
+
+
+         ":))"

llvm/test/tools/llvm-rc/tokenizer.test

+; RUN: llvm-rc /V %p/Inputs/tokens.rc | FileCheck %s
+
+; CHECK:  Int: 1; int value = 1
+; CHECK-NEXT:  Plus: +
+; CHECK-NEXT:  Int: 2; int value = 2
+; CHECK-NEXT:  Minus: -
+; CHECK-NEXT:  Int: 3214L; int value = 3214
+; CHECK-NEXT:  Amp: &
+; CHECK-NEXT:  Int: 0x120894; int value = 1181844
+; CHECK-NEXT:  Int: 032173; int value = 13435
+; CHECK-NEXT:  Int: 2; int value = 2
+; CHECK-NEXT:  Pipe: |
+; CHECK-NEXT:  Amp: &
+; CHECK-NEXT:  Tilde: ~
+; CHECK-NEXT:  Plus: +
+; CHECK-NEXT:  LeftParen: (
+; CHECK-NEXT:  Minus: -
+; CHECK-NEXT:  Int: 7; int value = 7
+; CHECK-NEXT:  RightParen: )
+; CHECK-NEXT:  BlockBegin: {
+; CHECK-NEXT:  Int: 0xabcdef; int value = 11259375
+; CHECK-NEXT:  Int: 0xABCDEFl; int value = 11259375
+; CHECK-NEXT:  BlockEnd: }
+; CHECK-NEXT:  BlockBegin: Begin
+; CHECK-NEXT:  BlockEnd: End
+; CHECK-NEXT:  Identifier: He11o
+; CHECK-NEXT:  Identifier: LLVM
+; CHECK-NEXT:  String: "RC string test."
+; CHECK-NEXT:  Comma: ,
+; CHECK-NEXT:  String: L"Another RC string test.'&{"
+; CHECK-NEXT:  Comma: ,
+; CHECK-NEXT:  Int: 42; int value = 42
+; CHECK-NEXT:  Comma: ,
+; CHECK-NEXT:  Int: 100; int value = 100
+; CHECK-NEXT:  String: ":))"

llvm/tools/llvm-rc/CMakeLists.txt

 set(LLVM_LINK_COMPONENTS
   Option
   Support
   )
 
 set(LLVM_TARGET_DEFINITIONS Opts.td)
 
 tablegen(LLVM Opts.inc -gen-opt-parser-defs)
 add_public_tablegen_target(RcTableGen)
 
 add_llvm_tool(llvm-rc
   llvm-rc.cpp
+  ResourceScriptToken.cpp
   )

llvm/tools/llvm-rc/ResourceScriptToken.h

+//===-- ResourceScriptToken.h -----------------------------------*- C++-*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===---------------------------------------------------------------------===//
+//
+// This declares the .rc script tokens and defines an interface for tokenizing
+// the input data. The list of available tokens is located at
+// ResourceScriptTokenList.h.
+//
+// Note that the tokenizer does not support comments or preprocessor
+// directives. The preprocessor should do its work on the .rc file before
+// running llvm-rc.
+//
+// As for now, it is possible to parse ASCII files only (the behavior on
+// UTF files might be undefined). However, it already consumes UTF-8 BOM, if
+// there is any. Thus, ASCII-compatible UTF-8 files are tokenized correctly.
+//
+// Ref: msdn.microsoft.com/en-us/library/windows/desktop/aa380599(v=vs.85).aspx
+//
+//===---------------------------------------------------------------------===//
+
+#ifndef LLVM_TOOLS_LLVMRC_RESOURCESCRIPTTOKEN_H
+#define LLVM_TOOLS_LLVMRC_RESOURCESCRIPTTOKEN_H
+
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/Error.h"
+
+#include <cstdint>
+#include <map>
+#include <string>
+#include <vector>
+
+namespace llvm {
+
+// A definition of a single resource script token. Each token has its kind
+// (declared in ResourceScriptTokenList) and holds a value - a reference
+// representation of the token.
+// RCToken does not claim ownership on its value. A memory buffer containing
+// the token value should be stored in a safe place and cannot be freed
+// nor reallocated.
+class RCToken {
+public:
+  enum class Kind {
+#define TOKEN(Name) Name,
+#define SHORT_TOKEN(Name, Ch) Name,
+#include "ResourceScriptTokenList.h"
+#undef TOKEN
+#undef SHORT_TOKEN
+  };
+
+  RCToken(RCToken::Kind RCTokenKind, StringRef Value);
+
+  // Get an integer value of the integer token.
+  uint32_t intValue() const;
+
+  StringRef value() const;
+  Kind kind() const;
+
+private:
+  Kind TokenKind;
+  StringRef TokenValue;
+};
+
+// Tokenize Input.
+// In case no error occured, the return value contains
+//   tokens in order they were in the input file.
+// In case of any error, the return value contains
+//   a textual representation of error.
+//
+// Tokens returned by this function hold only references to the parts
+// of the Input. Memory buffer containing Input cannot be freed,
+// modified or reallocated.
+Expected<std::vector<RCToken>> tokenizeRC(StringRef Input);
+
+} // namespace llvm
+
+#endif

llvm/tools/llvm-rc/ResourceScriptToken.cpp

+//===-- ResourceScriptToken.cpp ---------------------------------*- C++-*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===---------------------------------------------------------------------===//
+//
+// This file implements an interface defined in ResourceScriptToken.h.
+// In particular, it defines an .rc script tokenizer.
+//
+//===---------------------------------------------------------------------===//
+
+#include "ResourceScriptToken.h"
+#include "llvm/Support/raw_ostream.h"
+
+#include <algorithm>
+#include <cassert>
+#include <cctype>
+#include <cstdlib>
+#include <utility>
+
+using namespace llvm;
+
+using Kind = RCToken::Kind;
+
+// Checks if Representation is a correct description of an RC integer.
+// It should be a 32-bit unsigned integer, either decimal, octal (0[0-7]+),
+// or hexadecimal (0x[0-9a-f]+). It might be followed by a single 'L'
+// character (that is the difference between our representation and
+// StringRef's one). If Representation is correct, 'true' is returned and
+// the return value is put back in Num.
+static bool rcGetAsInteger(StringRef Representation, uint32_t &Num) {
+  size_t Length = Representation.size();
+  if (Length == 0)
+    return false;
+  // Strip the last 'L' if unnecessary.
+  if (std::toupper(Representation.back()) == 'L')
+    Representation = Representation.drop_back(1);
+
+  return !Representation.getAsInteger<uint32_t>(0, Num);
+}
+
+RCToken::RCToken(RCToken::Kind RCTokenKind, StringRef Value)
+    : TokenKind(RCTokenKind), TokenValue(Value) {}
+
+uint32_t RCToken::intValue() const {
+  assert(TokenKind == Kind::Int);
+  // We assume that the token already is a correct integer (checked by
+  // rcGetAsInteger).
+  uint32_t Result;
+  bool IsSuccess = rcGetAsInteger(TokenValue, Result);
+  assert(IsSuccess);
+  (void)IsSuccess;  // Silence the compiler warning when -DNDEBUG flag is on.
+  return Result;
+}
+
+StringRef RCToken::value() const { return TokenValue; }
+
+Kind RCToken::kind() const { return TokenKind; }
+
+static Error getStringError(const Twine &message) {
+  return make_error<StringError>("Error parsing file: " + message,
+                                 inconvertibleErrorCode());
+}
+
+namespace {
+
+class Tokenizer {
+public:
+  Tokenizer(StringRef Input) : Data(Input), DataLength(Input.size()) {}
+
+  Expected<std::vector<RCToken>> run();
+
+private:
+  // All 'advancing' methods return boolean values; if they're equal to false,
+  // the stream has ended or failed.
+  bool advance(size_t Amount = 1);
+  bool skipWhitespaces();
+
+  // Consumes a token. If any problem occurred, a non-empty Error is returned.
+  Error consumeToken(const Kind TokenKind);
+
+  // Check if tokenizer is about to read FollowingChars.
+  bool willNowRead(StringRef FollowingChars) const;
+
+  // Check if tokenizer can start reading an identifier at current position.
+  // The original tool did non specify the rules to determine what is a correct
+  // identifier. We assume they should follow the C convention:
+  // [a-zA-z_][a-zA-Z0-9_]*.
+  bool canStartIdentifier() const;
+  // Check if tokenizer can continue reading an identifier.
+  bool canContinueIdentifier() const;
+
+  // Check if tokenizer can start reading an integer.
+  // A correct integer always starts with a 0-9 digit,
+  // can contain characters 0-9A-Fa-f (digits),
+  // Ll (marking the integer is 32-bit), Xx (marking the representation
+  // is hexadecimal). As some kind of separator should come after the
+  // integer, we can consume the integer until a non-alphanumeric
+  // character.
+  bool canStartInt() const;
+  bool canContinueInt() const;
+
+  bool canStartString() const;
+
+  bool streamEof() const;
+
+  // Classify the token that is about to be read from the current position.
+  Kind classifyCurrentToken() const;
+
+  // Process the Kind::Identifier token - check if it is
+  // an identifier describing a block start or end.
+  void processIdentifier(RCToken &token) const;
+
+  StringRef Data;
+  size_t DataLength, Pos;
+};
+
+Expected<std::vector<RCToken>> Tokenizer::run() {
+  Pos = 0;
+  std::vector<RCToken> Result;
+
+  // Consume an optional UTF-8 Byte Order Mark.
+  if (willNowRead("\xef\xbb\xbf"))
+    advance(3);
+
+  while (!streamEof()) {
+    if (!skipWhitespaces())
+      break;
+
+    Kind TokenKind = classifyCurrentToken();
+    if (TokenKind == Kind::Invalid)
+      return getStringError("Invalid token found at position " + Twine(Pos));
+
+    const size_t TokenStart = Pos;
+    if (Error TokenError = consumeToken(TokenKind))
+      return std::move(TokenError);
+
+    RCToken Token(TokenKind, Data.take_front(Pos).drop_front(TokenStart));
+    if (TokenKind == Kind::Identifier) {
+      processIdentifier(Token);
+    } else if (TokenKind == Kind::Int) {
+      uint32_t TokenInt;
+      if (!rcGetAsInteger(Token.value(), TokenInt)) {
+        // The integer has incorrect format or cannot be represented in
+        // a 32-bit integer.
+        return getStringError("Integer invalid or too large: " +
+                              Token.value().str());
+      }
+    }
+
+    Result.push_back(Token);
+  }
+
+  return Result;
+}
+
+bool Tokenizer::advance(size_t Amount) {
+  Pos += Amount;
+  return !streamEof();
+}
+
+bool Tokenizer::skipWhitespaces() {
+  while (!streamEof() && std::isspace(Data[Pos]))
+    advance();
+  return !streamEof();
+}
+
+Error Tokenizer::consumeToken(const Kind TokenKind) {
+  switch (TokenKind) {
+  // One-character token consumption.
+#define TOKEN(Name)
+#define SHORT_TOKEN(Name, Ch) case Kind::Name:
+#include "ResourceScriptTokenList.h"
+#undef TOKEN
+#undef SHORT_TOKEN
+    advance();
+    return Error::success();
+
+  case Kind::Identifier:
+    while (!streamEof() && canContinueIdentifier())
+      advance();
+    return Error::success();
+
+  case Kind::Int:
+    while (!streamEof() && canContinueInt())
+      advance();
+    return Error::success();
+
+  case Kind::String:
+    // Consume the preceding 'L', if there is any.
+    if (std::toupper(Data[Pos]) == 'L')
+      advance();
+    // Consume the double-quote.
+    advance();
+
+    // Consume the characters until the end of the file, line or string.
+    while (true) {
+      if (streamEof()) {
+        return getStringError("Unterminated string literal.");
+      } else if (Data[Pos] == '"') {
+        // Consume the ending double-quote.
+        advance();
+        return Error::success();
+      } else if (Data[Pos] == '\n') {
+        return getStringError("String literal not terminated in the line.");
+      }
+
+      advance();
+    }
+
+  case Kind::Invalid:
+    assert(false && "Cannot consume an invalid token.");
+  }
+}
+
+bool Tokenizer::willNowRead(StringRef FollowingChars) const {
+  return Data.drop_front(Pos).startswith(FollowingChars);
+}
+
+bool Tokenizer::canStartIdentifier() const {
+  assert(!streamEof());
+
+  const char CurChar = Data[Pos];
+  return std::isalpha(CurChar) || CurChar == '_';
+}
+
+bool Tokenizer::canContinueIdentifier() const {
+  assert(!streamEof());
+  const char CurChar = Data[Pos];
+  return std::isalnum(CurChar) || CurChar == '_';
+}
+
+bool Tokenizer::canStartInt() const {
+  assert(!streamEof());
+  return std::isdigit(Data[Pos]);
+}
+
+bool Tokenizer::canContinueInt() const {
+  assert(!streamEof());
+  return std::isalnum(Data[Pos]);
+}
+
+bool Tokenizer::canStartString() const {
+  return willNowRead("\"") || willNowRead("L\"") || willNowRead("l\"");
+}
+
+bool Tokenizer::streamEof() const { return Pos == DataLength; }
+
+Kind Tokenizer::classifyCurrentToken() const {
+  if (canStartInt())
+    return Kind::Int;
+  if (canStartString())
+    return Kind::String;
+  // BEGIN and END are at this point of lexing recognized as identifiers.
+  if (canStartIdentifier())
+    return Kind::Identifier;
+
+  const char CurChar = Data[Pos];
+
+  switch (CurChar) {
+  // One-character token classification.
+#define TOKEN(Name)
+#define SHORT_TOKEN(Name, Ch)                                                  \
+  case Ch:                                                                     \
+    return Kind::Name;
+#include "ResourceScriptTokenList.h"
+#undef TOKEN
+#undef SHORT_TOKEN
+
+  default:
+    return Kind::Invalid;
+  }
+}
+
+void Tokenizer::processIdentifier(RCToken &Token) const {
+  assert(Token.kind() == Kind::Identifier);
+  StringRef Name = Token.value();
+
+  if (Name.equals_lower("begin"))
+    Token = RCToken(Kind::BlockBegin, Name);
+  else if (Name.equals_lower("end"))
+    Token = RCToken(Kind::BlockEnd, Name);
+}
+
+} // anonymous namespace
+
+namespace llvm {
+
+Expected<std::vector<RCToken>> tokenizeRC(StringRef Input) {
+  return Tokenizer(Input).run();
+}
+
+} // namespace llvm

llvm/tools/llvm-rc/ResourceScriptTokenList.h

+//===-- ResourceScriptTokenList.h -------------------------------*- C++-*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===---------------------------------------------------------------------===//
+//
+// This is a part of llvm-rc tokenizer. It lists all the possible tokens
+// that might occur in a correct .rc script.
+//
+//===---------------------------------------------------------------------===//
+
+
+// Long tokens. They might consist of more than one character.
+TOKEN(Invalid)      // Invalid token. Should not occur in a valid script.
+TOKEN(Int)          // Integer (decimal, octal or hexadecimal).
+TOKEN(String)       // String value.
+TOKEN(Identifier)   // Script identifier (resource name or type).
+
+// Short tokens. They usually consist of exactly one character.
+// The definitions are of the form SHORT_TOKEN(TokenName, TokenChar).
+// TokenChar is the one-character token representation occuring in the correct
+// .rc scripts.
+SHORT_TOKEN(BlockBegin, '{')   // Start of the script block; can also be BEGIN.
+SHORT_TOKEN(BlockEnd, '}')     // End of the block; can also be END.
+SHORT_TOKEN(Comma, ',')        // Comma - resource arguments separator.
+SHORT_TOKEN(Plus, '+')         // Addition operator.
+SHORT_TOKEN(Minus, '-')        // Subtraction operator.
+SHORT_TOKEN(Pipe, '|')         // Bitwise-OR operator.
+SHORT_TOKEN(Amp, '&')          // Bitwise-AND operator.
+SHORT_TOKEN(Tilde, '~')        // Bitwise-NOT operator.
+SHORT_TOKEN(LeftParen, '(')    // Left parenthesis in the script expressions.
+SHORT_TOKEN(RightParen, ')')   // Right parenthesis.

llvm/tools/llvm-rc/llvm-rc.cpp

-//===- llvm-rc.cpp - Compile .rc scripts into .res -------------*- C++ -*--===//
+//===-- llvm-rc.cpp - Compile .rc scripts into .res -------------*- C++ -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 //
 // Compile .rc scripts into .res files. This is intended to be a
 // platform-independent port of Microsoft's rc.exe tool.
 //
 //===----------------------------------------------------------------------===//
 
+#include "ResourceScriptToken.h"
+
 #include "llvm/Option/Arg.h"
 #include "llvm/Option/ArgList.h"
 #include "llvm/Support/Error.h"
 #include "llvm/Support/ManagedStatic.h"
 #include "llvm/Support/PrettyStackTrace.h"
 #include "llvm/Support/Process.h"
 #include "llvm/Support/Signals.h"
 #include "llvm/Support/raw_ostream.h"

 };
 
 class RcOptTable : public opt::OptTable {
 public:
   RcOptTable() : OptTable(InfoTable, /* IgnoreCase = */ true) {}
 };
 
 static ExitOnError ExitOnErr;
+
+LLVM_ATTRIBUTE_NORETURN static void fatalError(Twine Message) {
+  errs() << Message << "\n";
+  exit(1);
+}
+
 } // anonymous namespace
 
 int main(int argc_, const char *argv_[]) {
   sys::PrintStackTraceOnErrorSignal(argv_[0]);
   PrettyStackTraceProgram X(argc_, argv_);
 
   ExitOnErr.setBanner("llvm-rc: ");
 
   SmallVector<const char *, 256> argv;
   SpecificBumpPtrAllocator<char> ArgAllocator;
   ExitOnErr(errorCodeToError(sys::Process::GetArgumentVector(
       argv, makeArrayRef(argv_, argc_), ArgAllocator)));
 
   llvm_shutdown_obj Y;
 
   RcOptTable T;
   unsigned MAI, MAC;
   ArrayRef<const char *> ArgsArr = makeArrayRef(argv_ + 1, argc_);
   opt::InputArgList InputArgs = T.ParseArgs(ArgsArr, MAI, MAC);
 
   // The tool prints nothing when invoked with no command-line arguments.
-  if (InputArgs.hasArg(OPT_HELP))
+  if (InputArgs.hasArg(OPT_HELP)) {
     T.PrintHelp(outs(), "rc", "Resource Converter", false);
+    return 0;
+  }
+
+  const bool BeVerbose = InputArgs.hasArg(OPT_VERBOSE);
+
+  std::vector<std::string> InArgsInfo = InputArgs.getAllArgValues(OPT_INPUT);
+  if (InArgsInfo.size() != 1) {
+    fatalError("Exactly one input file should be provided.");
+  }
+
+  // Read and tokenize the input file.
+  const Twine &Filename = InArgsInfo[0];
+  ErrorOr<std::unique_ptr<MemoryBuffer>> File = MemoryBuffer::getFile(Filename);
+  if (!File) {
+    fatalError("Error opening file '" + Filename +
+               "': " + File.getError().message());
+  }
+
+  std::unique_ptr<MemoryBuffer> FileContents = std::move(*File);
+  StringRef Contents = FileContents->getBuffer();
+
+  std::vector<RCToken> Tokens = ExitOnErr(tokenizeRC(Contents));
+
+  if (BeVerbose) {
+    const Twine TokenNames[] = {
+#define TOKEN(Name) #Name,
+#define SHORT_TOKEN(Name, Ch) #Name,
+#include "ResourceScriptTokenList.h"
+#undef TOKEN
+#undef SHORT_TOKEN
+    };
+
+    for (const RCToken &Token : Tokens) {
+      outs() << TokenNames[static_cast<int>(Token.kind())] << ": "
+             << Token.value();
+      if (Token.kind() == RCToken::Kind::Int)
+        outs() << "; int value = " << Token.intValue();
+
+      outs() << "\n";
+    }
+  }
 
   return 0;
 }