lib/MC/MCParser/AsmLexer.cpp

   1 //===- AsmLexer.cpp - Lexer for Assembly Files ----------------------------===//
   2 //
   3 //                     The LLVM Compiler Infrastructure
   4 //
   5 // This file is distributed under the University of Illinois Open Source
   6 // License. See LICENSE.TXT for details.
   7 //
   8 //===----------------------------------------------------------------------===//
   9 //
  10 // This class implements the lexer for assembly files.
  11 //
  12 //===----------------------------------------------------------------------===//
  13
  14 #include "llvm/MC/MCParser/AsmLexer.h"
  15 #include "llvm/Support/SMLoc.h"
  16 #include "llvm/Support/MemoryBuffer.h"
  17 #include "llvm/MC/MCAsmInfo.h"
  18 #include <cerrno>
  19 #include <cstdio>
  20 #include <cstdlib>
  21 using namespace llvm;
  22
  23 AsmLexer::AsmLexer(const MCAsmInfo &_MAI) : MAI(_MAI)  {
  24   CurBuf = NULL;
  25   CurPtr = NULL;
  26 }
  27
  28 AsmLexer::~AsmLexer() {
  29 }
  30
  31 void AsmLexer::setBuffer(const MemoryBuffer *buf, const char *ptr) {
  32   CurBuf = buf;
  33
  34   if (ptr)
  35     CurPtr = ptr;
  36   else
  37     CurPtr = CurBuf->getBufferStart();
  38
  39   TokStart = 0;
  40 }
  41
  42 /// ReturnError - Set the error to the specified string at the specified
  43 /// location.  This is defined to always return AsmToken::Error.
  44 AsmToken AsmLexer::ReturnError(const char *Loc, const std::string &Msg) {
  45   SetError(SMLoc::getFromPointer(Loc), Msg);
  46
  47   return AsmToken(AsmToken::Error, StringRef(Loc, 0));
  48 }
  49
  50 int AsmLexer::getNextChar() {
  51   char CurChar = *CurPtr++;
  52   switch (CurChar) {
  53   default:
  54     return (unsigned char)CurChar;
  55   case 0:
  56     // A nul character in the stream is either the end of the current buffer or
  57     // a random nul in the file.  Disambiguate that here.
  58     if (CurPtr-1 != CurBuf->getBufferEnd())
  59       return 0;  // Just whitespace.
  60
  61     // Otherwise, return end of file.
  62     --CurPtr;  // Another call to lex will return EOF again.
  63     return EOF;
  64   }
  65 }
  66
  67 /// LexFloatLiteral: [0-9]*[.][0-9]*([eE][+-]?[0-9]*)?
  68 ///
  69 /// The leading integral digit sequence and dot should have already been
  70 /// consumed, some or all of the fractional digit sequence *can* have been
  71 /// consumed.
  72 AsmToken AsmLexer::LexFloatLiteral() {
  73   // Skip the fractional digit sequence.
  74   while (isdigit(*CurPtr))
  75     ++CurPtr;
  76
  77   // Check for exponent; we intentionally accept a slighlty wider set of
  78   // literals here and rely on the upstream client to reject invalid ones (e.g.,
  79   // "1e+").
  80   if (*CurPtr == 'e' || *CurPtr == 'E') {
  81     ++CurPtr;
  82     if (*CurPtr == '-' || *CurPtr == '+')
  83       ++CurPtr;
  84     while (isdigit(*CurPtr))
  85       ++CurPtr;
  86   }
  87
  88   return AsmToken(AsmToken::Real,
  89                   StringRef(TokStart, CurPtr - TokStart));
  90 }
  91
  92 /// LexIdentifier: [a-zA-Z_.][a-zA-Z0-9_$.@]*
  93 static bool IsIdentifierChar(char c) {
  94   return isalnum(c) || c == '_' || c == '$' || c == '.' || c == '@';
  95 }
  96 AsmToken AsmLexer::LexIdentifier() {
  97   // Check for floating point literals.
  98   if (CurPtr[-1] == '.' && isdigit(*CurPtr)) {
  99     // Disambiguate a .1243foo identifier from a floating literal.
 100     while (isdigit(*CurPtr))
 101       ++CurPtr;
 102     if (*CurPtr == 'e' || *CurPtr == 'E' || !IsIdentifierChar(*CurPtr))
 103       return LexFloatLiteral();
 104   }
 105
 106   while (IsIdentifierChar(*CurPtr))
 107     ++CurPtr;
 108
 109   // Handle . as a special case.
 110   if (CurPtr == TokStart+1 && TokStart[0] == '.')
 111     return AsmToken(AsmToken::Dot, StringRef(TokStart, 1));
 112
 113   return AsmToken(AsmToken::Identifier, StringRef(TokStart, CurPtr - TokStart));
 114 }
 115
 116 /// LexSlash: Slash: /
 117 ///           C-Style Comment: /* ... */
 118 AsmToken AsmLexer::LexSlash() {
 119   switch (*CurPtr) {
 120   case '*': break; // C style comment.
 121   case '/': return ++CurPtr, LexLineComment();
 122   default:  return AsmToken(AsmToken::Slash, StringRef(CurPtr-1, 1));
 123   }
 124
 125   // C Style comment.
 126   ++CurPtr;  // skip the star.
 127   while (1) {
 128     int CurChar = getNextChar();
 129     switch (CurChar) {
 130     case EOF:
 131       return ReturnError(TokStart, "unterminated comment");
 132     case '*':
 133       // End of the comment?
 134       if (CurPtr[0] != '/') break;
 135
 136       ++CurPtr;   // End the */.
 137       return LexToken();
 138     }
 139   }
 140 }
 141
 142 /// LexLineComment: Comment: #[^\n]*
 143 ///                        : //[^\n]*
 144 AsmToken AsmLexer::LexLineComment() {
 145   // FIXME: This is broken if we happen to a comment at the end of a file, which
 146   // was .included, and which doesn't end with a newline.
 147   int CurChar = getNextChar();
 148   while (CurChar != '\n' && CurChar != '\n' && CurChar != EOF)
 149     CurChar = getNextChar();
 150
 151   if (CurChar == EOF)
 152     return AsmToken(AsmToken::Eof, StringRef(CurPtr, 0));
 153   return AsmToken(AsmToken::EndOfStatement, StringRef(CurPtr, 0));
 154 }
 155
 156 static void SkipIgnoredIntegerSuffix(const char *&CurPtr) {
 157   if (CurPtr[0] == 'L' && CurPtr[1] == 'L')
 158     CurPtr += 2;
 159   if (CurPtr[0] == 'U' && CurPtr[1] == 'L' && CurPtr[2] == 'L')
 160     CurPtr += 3;
 161 }
 162
 163 /// LexDigit: First character is [0-9].
 164 ///   Local Label: [0-9][:]
 165 ///   Forward/Backward Label: [0-9][fb]
 166 ///   Binary integer: 0b[01]+
 167 ///   Octal integer: 0[0-7]+
 168 ///   Hex integer: 0x[0-9a-fA-F]+
 169 ///   Decimal integer: [1-9][0-9]*
 170 AsmToken AsmLexer::LexDigit() {
 171   // Decimal integer: [1-9][0-9]*
 172   if (CurPtr[-1] != '0' || CurPtr[0] == '.') {
 173     while (isdigit(*CurPtr))
 174       ++CurPtr;
 175
 176     // Check for floating point literals.
 177     if (*CurPtr == '.' || *CurPtr == 'e') {
 178       ++CurPtr;
 179       return LexFloatLiteral();
 180     }
 181
 182     StringRef Result(TokStart, CurPtr - TokStart);
 183
 184     long long Value;
 185     if (Result.getAsInteger(10, Value)) {
 186       // We have to handle minint_as_a_positive_value specially, because
 187       // - minint_as_a_positive_value = minint and it is valid.
 188       if (Result == "9223372036854775808")
 189         Value = -9223372036854775808ULL;
 190       else
 191         return ReturnError(TokStart, "Invalid decimal number");
 192     }
 193
 194     // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
 195     // suffixes on integer literals.
 196     SkipIgnoredIntegerSuffix(CurPtr);
 197
 198     return AsmToken(AsmToken::Integer, Result, Value);
 199   }
 200
 201   if (*CurPtr == 'b') {
 202     ++CurPtr;
 203     // See if we actually have "0b" as part of something like "jmp 0b\n"
 204     if (!isdigit(CurPtr[0])) {
 205       --CurPtr;
 206       StringRef Result(TokStart, CurPtr - TokStart);
 207       return AsmToken(AsmToken::Integer, Result, 0);
 208     }
 209     const char *NumStart = CurPtr;
 210     while (CurPtr[0] == '0' || CurPtr[0] == '1')
 211       ++CurPtr;
 212
 213     // Requires at least one binary digit.
 214     if (CurPtr == NumStart)
 215       return ReturnError(TokStart, "Invalid binary number");
 216
 217     StringRef Result(TokStart, CurPtr - TokStart);
 218
 219     long long Value;
 220     if (Result.substr(2).getAsInteger(2, Value))
 221       return ReturnError(TokStart, "Invalid binary number");
 222
 223     // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
 224     // suffixes on integer literals.
 225     SkipIgnoredIntegerSuffix(CurPtr);
 226
 227     return AsmToken(AsmToken::Integer, Result, Value);
 228   }
 229
 230   if (*CurPtr == 'x') {
 231     ++CurPtr;
 232     const char *NumStart = CurPtr;
 233     while (isxdigit(CurPtr[0]))
 234       ++CurPtr;
 235
 236     // Requires at least one hex digit.
 237     if (CurPtr == NumStart)
 238       return ReturnError(CurPtr-2, "Invalid hexadecimal number");
 239
 240     unsigned long long Result;
 241     if (StringRef(TokStart, CurPtr - TokStart).getAsInteger(0, Result))
 242       return ReturnError(TokStart, "Invalid hexadecimal number");
 243
 244     // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
 245     // suffixes on integer literals.
 246     SkipIgnoredIntegerSuffix(CurPtr);
 247
 248     return AsmToken(AsmToken::Integer, StringRef(TokStart, CurPtr - TokStart),
 249                     (int64_t)Result);
 250   }
 251
 252   // Must be an octal number, it starts with 0.
 253   while (*CurPtr >= '0' && *CurPtr <= '7')
 254     ++CurPtr;
 255
 256   StringRef Result(TokStart, CurPtr - TokStart);
 257   long long Value;
 258   if (Result.getAsInteger(8, Value))
 259     return ReturnError(TokStart, "Invalid octal number");
 260
 261   // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
 262   // suffixes on integer literals.
 263   SkipIgnoredIntegerSuffix(CurPtr);
 264
 265   return AsmToken(AsmToken::Integer, Result, Value);
 266 }
 267
 268 /// LexQuote: String: "..."
 269 AsmToken AsmLexer::LexQuote() {
 270   int CurChar = getNextChar();
 271   // TODO: does gas allow multiline string constants?
 272   while (CurChar != '"') {
 273     if (CurChar == '\\') {
 274       // Allow \", etc.
 275       CurChar = getNextChar();
 276     }
 277
 278     if (CurChar == EOF)
 279       return ReturnError(TokStart, "unterminated string constant");
 280
 281     CurChar = getNextChar();
 282   }
 283
 284   return AsmToken(AsmToken::String, StringRef(TokStart, CurPtr - TokStart));
 285 }
 286
 287 StringRef AsmLexer::LexUntilEndOfStatement() {
 288   TokStart = CurPtr;
 289
 290   while (!isAtStartOfComment(*CurPtr) && // Start of line comment.
 291           *CurPtr != ';' &&  // End of statement marker.
 292          *CurPtr != '\n' &&
 293          *CurPtr != '\r' &&
 294          (*CurPtr != 0 || CurPtr != CurBuf->getBufferEnd())) {
 295     ++CurPtr;
 296   }
 297   return StringRef(TokStart, CurPtr-TokStart);
 298 }
 299
 300 bool AsmLexer::isAtStartOfComment(char Char) {
 301   // FIXME: This won't work for multi-character comment indicators like "//".
 302   return Char == *MAI.getCommentString();
 303 }
 304
 305 AsmToken AsmLexer::LexToken() {
 306   TokStart = CurPtr;
 307   // This always consumes at least one character.
 308   int CurChar = getNextChar();
 309
 310   if (isAtStartOfComment(CurChar))
 311     return LexLineComment();
 312
 313   switch (CurChar) {
 314   default:
 315     // Handle identifier: [a-zA-Z_.][a-zA-Z0-9_$.@]*
 316     if (isalpha(CurChar) || CurChar == '_' || CurChar == '.')
 317       return LexIdentifier();
 318
 319     // Unknown character, emit an error.
 320     return ReturnError(TokStart, "invalid character in input");
 321   case EOF: return AsmToken(AsmToken::Eof, StringRef(TokStart, 0));
 322   case 0:
 323   case ' ':
 324   case '\t':
 325     // Ignore whitespace.
 326     return LexToken();
 327   case '\n': // FALL THROUGH.
 328   case '\r': // FALL THROUGH.
 329   case ';': return AsmToken(AsmToken::EndOfStatement, StringRef(TokStart, 1));
 330   case ':': return AsmToken(AsmToken::Colon, StringRef(TokStart, 1));
 331   case '+': return AsmToken(AsmToken::Plus, StringRef(TokStart, 1));
 332   case '-': return AsmToken(AsmToken::Minus, StringRef(TokStart, 1));
 333   case '~': return AsmToken(AsmToken::Tilde, StringRef(TokStart, 1));
 334   case '(': return AsmToken(AsmToken::LParen, StringRef(TokStart, 1));
 335   case ')': return AsmToken(AsmToken::RParen, StringRef(TokStart, 1));
 336   case '[': return AsmToken(AsmToken::LBrac, StringRef(TokStart, 1));
 337   case ']': return AsmToken(AsmToken::RBrac, StringRef(TokStart, 1));
 338   case '{': return AsmToken(AsmToken::LCurly, StringRef(TokStart, 1));
 339   case '}': return AsmToken(AsmToken::RCurly, StringRef(TokStart, 1));
 340   case '*': return AsmToken(AsmToken::Star, StringRef(TokStart, 1));
 341   case ',': return AsmToken(AsmToken::Comma, StringRef(TokStart, 1));
 342   case '$': return AsmToken(AsmToken::Dollar, StringRef(TokStart, 1));
 343   case '@': return AsmToken(AsmToken::At, StringRef(TokStart, 1));
 344   case '=':
 345     if (*CurPtr == '=')
 346       return ++CurPtr, AsmToken(AsmToken::EqualEqual, StringRef(TokStart, 2));
 347     return AsmToken(AsmToken::Equal, StringRef(TokStart, 1));
 348   case '|':
 349     if (*CurPtr == '|')
 350       return ++CurPtr, AsmToken(AsmToken::PipePipe, StringRef(TokStart, 2));
 351     return AsmToken(AsmToken::Pipe, StringRef(TokStart, 1));
 352   case '^': return AsmToken(AsmToken::Caret, StringRef(TokStart, 1));
 353   case '&':
 354     if (*CurPtr == '&')
 355       return ++CurPtr, AsmToken(AsmToken::AmpAmp, StringRef(TokStart, 2));
 356     return AsmToken(AsmToken::Amp, StringRef(TokStart, 1));
 357   case '!':
 358     if (*CurPtr == '=')
 359       return ++CurPtr, AsmToken(AsmToken::ExclaimEqual, StringRef(TokStart, 2));
 360     return AsmToken(AsmToken::Exclaim, StringRef(TokStart, 1));
 361   case '%': return AsmToken(AsmToken::Percent, StringRef(TokStart, 1));
 362   case '/': return LexSlash();
 363   case '#': return AsmToken(AsmToken::Hash, StringRef(TokStart, 1));
 364   case '"': return LexQuote();
 365   case '0': case '1': case '2': case '3': case '4':
 366   case '5': case '6': case '7': case '8': case '9':
 367     return LexDigit();
 368   case '<':
 369     switch (*CurPtr) {
 370     case '<': return ++CurPtr, AsmToken(AsmToken::LessLess,
 371                                         StringRef(TokStart, 2));
 372     case '=': return ++CurPtr, AsmToken(AsmToken::LessEqual,
 373                                         StringRef(TokStart, 2));
 374     case '>': return ++CurPtr, AsmToken(AsmToken::LessGreater,
 375                                         StringRef(TokStart, 2));
 376     default: return AsmToken(AsmToken::Less, StringRef(TokStart, 1));
 377     }
 378   case '>':
 379     switch (*CurPtr) {
 380     case '>': return ++CurPtr, AsmToken(AsmToken::GreaterGreater,
 381                                         StringRef(TokStart, 2));
 382     case '=': return ++CurPtr, AsmToken(AsmToken::GreaterEqual,
 383                                         StringRef(TokStart, 2));
 384     default: return AsmToken(AsmToken::Greater, StringRef(TokStart, 1));
 385     }
 386
 387   // TODO: Quoted identifiers (objc methods etc)
 388   // local labels: [0-9][:]
 389   // Forward/backward labels: [0-9][fb]
 390   // Integers, fp constants, character constants.
 391   }
 392 }