PPDirectives.cpp revision cf29e0716bb3ecbbc15b74cd648367d6b075fdf0
1//===--- PPDirectives.cpp - Directive Handling for Preprocessor -----------===//
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 file implements # directive processing for the Preprocessor.
11//
12//===----------------------------------------------------------------------===//
13
14#include "clang/Lex/Preprocessor.h"
15#include "clang/Lex/LiteralSupport.h"
16#include "clang/Lex/HeaderSearch.h"
17#include "clang/Lex/MacroInfo.h"
18#include "clang/Lex/LexDiagnostic.h"
19#include "clang/Basic/SourceManager.h"
20#include "llvm/ADT/APInt.h"
21using namespace clang;
22
23//===----------------------------------------------------------------------===//
24// Utility Methods for Preprocessor Directive Handling.
25//===----------------------------------------------------------------------===//
26
27MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) {
28  MacroInfo *MI;
29
30  if (!MICache.empty()) {
31    MI = MICache.back();
32    MICache.pop_back();
33  } else
34    MI = (MacroInfo*) BP.Allocate<MacroInfo>();
35  new (MI) MacroInfo(L);
36  return MI;
37}
38
39/// ReleaseMacroInfo - Release the specified MacroInfo.  This memory will
40///  be reused for allocating new MacroInfo objects.
41void Preprocessor::ReleaseMacroInfo(MacroInfo* MI) {
42  MICache.push_back(MI);
43  MI->FreeArgumentList();
44}
45
46
47/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the
48/// current line until the tok::eom token is found.
49void Preprocessor::DiscardUntilEndOfDirective() {
50  Token Tmp;
51  do {
52    LexUnexpandedToken(Tmp);
53  } while (Tmp.isNot(tok::eom));
54}
55
56/// ReadMacroName - Lex and validate a macro name, which occurs after a
57/// #define or #undef.  This sets the token kind to eom and discards the rest
58/// of the macro line if the macro name is invalid.  isDefineUndef is 1 if
59/// this is due to a a #define, 2 if #undef directive, 0 if it is something
60/// else (e.g. #ifdef).
61void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) {
62  // Read the token, don't allow macro expansion on it.
63  LexUnexpandedToken(MacroNameTok);
64
65  // Missing macro name?
66  if (MacroNameTok.is(tok::eom)) {
67    Diag(MacroNameTok, diag::err_pp_missing_macro_name);
68    return;
69  }
70
71  IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
72  if (II == 0) {
73    std::string Spelling = getSpelling(MacroNameTok);
74    const IdentifierInfo &Info = Identifiers.get(Spelling);
75    if (Info.isCPlusPlusOperatorKeyword())
76      // C++ 2.5p2: Alternative tokens behave the same as its primary token
77      // except for their spellings.
78      Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling;
79    else
80      Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
81    // Fall through on error.
82  } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) {
83    // Error if defining "defined": C99 6.10.8.4.
84    Diag(MacroNameTok, diag::err_defined_macro_name);
85  } else if (isDefineUndef && II->hasMacroDefinition() &&
86             getMacroInfo(II)->isBuiltinMacro()) {
87    // Error if defining "__LINE__" and other builtins: C99 6.10.8.4.
88    if (isDefineUndef == 1)
89      Diag(MacroNameTok, diag::pp_redef_builtin_macro);
90    else
91      Diag(MacroNameTok, diag::pp_undef_builtin_macro);
92  } else {
93    // Okay, we got a good identifier node.  Return it.
94    return;
95  }
96
97  // Invalid macro name, read and discard the rest of the line.  Then set the
98  // token kind to tok::eom.
99  MacroNameTok.setKind(tok::eom);
100  return DiscardUntilEndOfDirective();
101}
102
103/// CheckEndOfDirective - Ensure that the next token is a tok::eom token.  If
104/// not, emit a diagnostic and consume up until the eom.
105void Preprocessor::CheckEndOfDirective(const char *DirType) {
106  Token Tmp;
107  // Lex unexpanded tokens: macros might expand to zero tokens, causing us to
108  // miss diagnosing invalid lines.
109  LexUnexpandedToken(Tmp);
110
111  // There should be no tokens after the directive, but we allow them as an
112  // extension.
113  while (Tmp.is(tok::comment))  // Skip comments in -C mode.
114    LexUnexpandedToken(Tmp);
115
116  if (Tmp.isNot(tok::eom)) {
117    Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType;
118    DiscardUntilEndOfDirective();
119  }
120}
121
122
123
124/// SkipExcludedConditionalBlock - We just read a #if or related directive and
125/// decided that the subsequent tokens are in the #if'd out portion of the
126/// file.  Lex the rest of the file, until we see an #endif.  If
127/// FoundNonSkipPortion is true, then we have already emitted code for part of
128/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk
129/// is true, then #else directives are ok, if not, then we have already seen one
130/// so a #else directive is a duplicate.  When this returns, the caller can lex
131/// the first valid token.
132void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
133                                                bool FoundNonSkipPortion,
134                                                bool FoundElse) {
135  ++NumSkipped;
136  assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?");
137
138  CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false,
139                                 FoundNonSkipPortion, FoundElse);
140
141  if (CurPTHLexer) {
142    PTHSkipExcludedConditionalBlock();
143    return;
144  }
145
146  // Enter raw mode to disable identifier lookup (and thus macro expansion),
147  // disabling warnings, etc.
148  CurPPLexer->LexingRawMode = true;
149  Token Tok;
150  while (1) {
151    if (CurLexer)
152      CurLexer->Lex(Tok);
153    else
154      CurPTHLexer->Lex(Tok);
155
156    // If this is the end of the buffer, we have an error.
157    if (Tok.is(tok::eof)) {
158      // Emit errors for each unterminated conditional on the stack, including
159      // the current one.
160      while (!CurPPLexer->ConditionalStack.empty()) {
161        Diag(CurPPLexer->ConditionalStack.back().IfLoc,
162             diag::err_pp_unterminated_conditional);
163        CurPPLexer->ConditionalStack.pop_back();
164      }
165
166      // Just return and let the caller lex after this #include.
167      break;
168    }
169
170    // If this token is not a preprocessor directive, just skip it.
171    if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine())
172      continue;
173
174    // We just parsed a # character at the start of a line, so we're in
175    // directive mode.  Tell the lexer this so any newlines we see will be
176    // converted into an EOM token (this terminates the macro).
177    CurPPLexer->ParsingPreprocessorDirective = true;
178    if (CurLexer) CurLexer->SetCommentRetentionState(false);
179
180
181    // Read the next token, the directive flavor.
182    LexUnexpandedToken(Tok);
183
184    // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
185    // something bogus), skip it.
186    if (Tok.isNot(tok::identifier)) {
187      CurPPLexer->ParsingPreprocessorDirective = false;
188      // Restore comment saving mode.
189      if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
190      continue;
191    }
192
193    // If the first letter isn't i or e, it isn't intesting to us.  We know that
194    // this is safe in the face of spelling differences, because there is no way
195    // to spell an i/e in a strange way that is another letter.  Skipping this
196    // allows us to avoid looking up the identifier info for #define/#undef and
197    // other common directives.
198    const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation());
199    char FirstChar = RawCharData[0];
200    if (FirstChar >= 'a' && FirstChar <= 'z' &&
201        FirstChar != 'i' && FirstChar != 'e') {
202      CurPPLexer->ParsingPreprocessorDirective = false;
203      // Restore comment saving mode.
204      if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
205      continue;
206    }
207
208    // Get the identifier name without trigraphs or embedded newlines.  Note
209    // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
210    // when skipping.
211    // TODO: could do this with zero copies in the no-clean case by using
212    // strncmp below.
213    char Directive[20];
214    unsigned IdLen;
215    if (!Tok.needsCleaning() && Tok.getLength() < 20) {
216      IdLen = Tok.getLength();
217      memcpy(Directive, RawCharData, IdLen);
218      Directive[IdLen] = 0;
219    } else {
220      std::string DirectiveStr = getSpelling(Tok);
221      IdLen = DirectiveStr.size();
222      if (IdLen >= 20) {
223        CurPPLexer->ParsingPreprocessorDirective = false;
224        // Restore comment saving mode.
225        if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
226        continue;
227      }
228      memcpy(Directive, &DirectiveStr[0], IdLen);
229      Directive[IdLen] = 0;
230      FirstChar = Directive[0];
231    }
232
233    if (FirstChar == 'i' && Directive[1] == 'f') {
234      if ((IdLen == 2) ||   // "if"
235          (IdLen == 5 && !strcmp(Directive+2, "def")) ||   // "ifdef"
236          (IdLen == 6 && !strcmp(Directive+2, "ndef"))) {  // "ifndef"
237        // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
238        // bother parsing the condition.
239        DiscardUntilEndOfDirective();
240        CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true,
241                                       /*foundnonskip*/false,
242                                       /*fnddelse*/false);
243      }
244    } else if (FirstChar == 'e') {
245      if (IdLen == 5 && !strcmp(Directive+1, "ndif")) {  // "endif"
246        CheckEndOfDirective("#endif");
247        PPConditionalInfo CondInfo;
248        CondInfo.WasSkipping = true; // Silence bogus warning.
249        bool InCond = CurPPLexer->popConditionalLevel(CondInfo);
250        InCond = InCond;  // Silence warning in no-asserts mode.
251        assert(!InCond && "Can't be skipping if not in a conditional!");
252
253        // If we popped the outermost skipping block, we're done skipping!
254        if (!CondInfo.WasSkipping)
255          break;
256      } else if (IdLen == 4 && !strcmp(Directive+1, "lse")) { // "else".
257        // #else directive in a skipping conditional.  If not in some other
258        // skipping conditional, and if #else hasn't already been seen, enter it
259        // as a non-skipping conditional.
260        CheckEndOfDirective("#else");
261        PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
262
263        // If this is a #else with a #else before it, report the error.
264        if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else);
265
266        // Note that we've seen a #else in this conditional.
267        CondInfo.FoundElse = true;
268
269        // If the conditional is at the top level, and the #if block wasn't
270        // entered, enter the #else block now.
271        if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
272          CondInfo.FoundNonSkip = true;
273          break;
274        }
275      } else if (IdLen == 4 && !strcmp(Directive+1, "lif")) {  // "elif".
276        PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
277
278        bool ShouldEnter;
279        // If this is in a skipping block or if we're already handled this #if
280        // block, don't bother parsing the condition.
281        if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
282          DiscardUntilEndOfDirective();
283          ShouldEnter = false;
284        } else {
285          // Restore the value of LexingRawMode so that identifiers are
286          // looked up, etc, inside the #elif expression.
287          assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
288          CurPPLexer->LexingRawMode = false;
289          IdentifierInfo *IfNDefMacro = 0;
290          ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
291          CurPPLexer->LexingRawMode = true;
292        }
293
294        // If this is a #elif with a #else before it, report the error.
295        if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else);
296
297        // If this condition is true, enter it!
298        if (ShouldEnter) {
299          CondInfo.FoundNonSkip = true;
300          break;
301        }
302      }
303    }
304
305    CurPPLexer->ParsingPreprocessorDirective = false;
306    // Restore comment saving mode.
307    if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
308  }
309
310  // Finally, if we are out of the conditional (saw an #endif or ran off the end
311  // of the file, just stop skipping and return to lexing whatever came after
312  // the #if block.
313  CurPPLexer->LexingRawMode = false;
314}
315
316void Preprocessor::PTHSkipExcludedConditionalBlock() {
317
318  while(1) {
319    assert(CurPTHLexer);
320    assert(CurPTHLexer->LexingRawMode == false);
321
322    // Skip to the next '#else', '#elif', or #endif.
323    if (CurPTHLexer->SkipBlock()) {
324      // We have reached an #endif.  Both the '#' and 'endif' tokens
325      // have been consumed by the PTHLexer.  Just pop off the condition level.
326      PPConditionalInfo CondInfo;
327      bool InCond = CurPTHLexer->popConditionalLevel(CondInfo);
328      InCond = InCond;  // Silence warning in no-asserts mode.
329      assert(!InCond && "Can't be skipping if not in a conditional!");
330      break;
331    }
332
333    // We have reached a '#else' or '#elif'.  Lex the next token to get
334    // the directive flavor.
335    Token Tok;
336    LexUnexpandedToken(Tok);
337
338    // We can actually look up the IdentifierInfo here since we aren't in
339    // raw mode.
340    tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID();
341
342    if (K == tok::pp_else) {
343      // #else: Enter the else condition.  We aren't in a nested condition
344      //  since we skip those. We're always in the one matching the last
345      //  blocked we skipped.
346      PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
347      // Note that we've seen a #else in this conditional.
348      CondInfo.FoundElse = true;
349
350      // If the #if block wasn't entered then enter the #else block now.
351      if (!CondInfo.FoundNonSkip) {
352        CondInfo.FoundNonSkip = true;
353
354        // Consume the eom token.
355        CurPTHLexer->ParsingPreprocessorDirective = true;
356        LexUnexpandedToken(Tok);
357        assert(Tok.is(tok::eom));
358        CurPTHLexer->ParsingPreprocessorDirective = false;
359
360        break;
361      }
362
363      // Otherwise skip this block.
364      continue;
365    }
366
367    assert(K == tok::pp_elif);
368    PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
369
370    // If this is a #elif with a #else before it, report the error.
371    if (CondInfo.FoundElse)
372      Diag(Tok, diag::pp_err_elif_after_else);
373
374    // If this is in a skipping block or if we're already handled this #if
375    // block, don't bother parsing the condition.  We just skip this block.
376    if (CondInfo.FoundNonSkip)
377      continue;
378
379    // Evaluate the condition of the #elif.
380    IdentifierInfo *IfNDefMacro = 0;
381    CurPTHLexer->ParsingPreprocessorDirective = true;
382    bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
383    CurPTHLexer->ParsingPreprocessorDirective = false;
384
385    // If this condition is true, enter it!
386    if (ShouldEnter) {
387      CondInfo.FoundNonSkip = true;
388      break;
389    }
390
391    // Otherwise, skip this block and go to the next one.
392    continue;
393  }
394}
395
396/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file,
397/// return null on failure.  isAngled indicates whether the file reference is
398/// for system #include's or not (i.e. using <> instead of "").
399const FileEntry *Preprocessor::LookupFile(const char *FilenameStart,
400                                          const char *FilenameEnd,
401                                          bool isAngled,
402                                          const DirectoryLookup *FromDir,
403                                          const DirectoryLookup *&CurDir) {
404  // If the header lookup mechanism may be relative to the current file, pass in
405  // info about where the current file is.
406  const FileEntry *CurFileEnt = 0;
407  if (!FromDir) {
408    FileID FID = getCurrentFileLexer()->getFileID();
409    CurFileEnt = SourceMgr.getFileEntryForID(FID);
410
411    // If there is no file entry associated with this file, it must be the
412    // predefines buffer.  Any other file is not lexed with a normal lexer, so
413    // it won't be scanned for preprocessor directives.   If we have the
414    // predefines buffer, resolve #include references (which come from the
415    // -include command line argument) as if they came from the main file, this
416    // affects file lookup etc.
417    if (CurFileEnt == 0) {
418      FID = SourceMgr.getMainFileID();
419      CurFileEnt = SourceMgr.getFileEntryForID(FID);
420    }
421  }
422
423  // Do a standard file entry lookup.
424  CurDir = CurDirLookup;
425  const FileEntry *FE =
426    HeaderInfo.LookupFile(FilenameStart, FilenameEnd,
427                          isAngled, FromDir, CurDir, CurFileEnt);
428  if (FE) return FE;
429
430  // Otherwise, see if this is a subframework header.  If so, this is relative
431  // to one of the headers on the #include stack.  Walk the list of the current
432  // headers on the #include stack and pass them to HeaderInfo.
433  if (IsFileLexer()) {
434    if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID())))
435      if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, FilenameEnd,
436                                                    CurFileEnt)))
437        return FE;
438  }
439
440  for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
441    IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1];
442    if (IsFileLexer(ISEntry)) {
443      if ((CurFileEnt =
444           SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID())))
445        if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart,
446                                                      FilenameEnd, CurFileEnt)))
447          return FE;
448    }
449  }
450
451  // Otherwise, we really couldn't find the file.
452  return 0;
453}
454
455
456//===----------------------------------------------------------------------===//
457// Preprocessor Directive Handling.
458//===----------------------------------------------------------------------===//
459
460/// HandleDirective - This callback is invoked when the lexer sees a # token
461/// at the start of a line.  This consumes the directive, modifies the
462/// lexer/preprocessor state, and advances the lexer(s) so that the next token
463/// read is the correct one.
464void Preprocessor::HandleDirective(Token &Result) {
465  // FIXME: Traditional: # with whitespace before it not recognized by K&R?
466
467  // We just parsed a # character at the start of a line, so we're in directive
468  // mode.  Tell the lexer this so any newlines we see will be converted into an
469  // EOM token (which terminates the directive).
470  CurPPLexer->ParsingPreprocessorDirective = true;
471
472  ++NumDirectives;
473
474  // We are about to read a token.  For the multiple-include optimization FA to
475  // work, we have to remember if we had read any tokens *before* this
476  // pp-directive.
477  bool ReadAnyTokensBeforeDirective = CurPPLexer->MIOpt.getHasReadAnyTokensVal();
478
479  // Read the next token, the directive flavor.  This isn't expanded due to
480  // C99 6.10.3p8.
481  LexUnexpandedToken(Result);
482
483  // C99 6.10.3p11: Is this preprocessor directive in macro invocation?  e.g.:
484  //   #define A(x) #x
485  //   A(abc
486  //     #warning blah
487  //   def)
488  // If so, the user is relying on non-portable behavior, emit a diagnostic.
489  if (InMacroArgs)
490    Diag(Result, diag::ext_embedded_directive);
491
492TryAgain:
493  switch (Result.getKind()) {
494  case tok::eom:
495    return;   // null directive.
496  case tok::comment:
497    // Handle stuff like "# /*foo*/ define X" in -E -C mode.
498    LexUnexpandedToken(Result);
499    goto TryAgain;
500
501  case tok::numeric_constant:  // # 7  GNU line marker directive.
502    return HandleDigitDirective(Result);
503  default:
504    IdentifierInfo *II = Result.getIdentifierInfo();
505    if (II == 0) break;  // Not an identifier.
506
507    // Ask what the preprocessor keyword ID is.
508    switch (II->getPPKeywordID()) {
509    default: break;
510    // C99 6.10.1 - Conditional Inclusion.
511    case tok::pp_if:
512      return HandleIfDirective(Result, ReadAnyTokensBeforeDirective);
513    case tok::pp_ifdef:
514      return HandleIfdefDirective(Result, false, true/*not valid for miopt*/);
515    case tok::pp_ifndef:
516      return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective);
517    case tok::pp_elif:
518      return HandleElifDirective(Result);
519    case tok::pp_else:
520      return HandleElseDirective(Result);
521    case tok::pp_endif:
522      return HandleEndifDirective(Result);
523
524    // C99 6.10.2 - Source File Inclusion.
525    case tok::pp_include:
526      return HandleIncludeDirective(Result);            // Handle #include.
527
528    // C99 6.10.3 - Macro Replacement.
529    case tok::pp_define:
530      return HandleDefineDirective(Result);
531    case tok::pp_undef:
532      return HandleUndefDirective(Result);
533
534    // C99 6.10.4 - Line Control.
535    case tok::pp_line:
536      return HandleLineDirective(Result);
537
538    // C99 6.10.5 - Error Directive.
539    case tok::pp_error:
540      return HandleUserDiagnosticDirective(Result, false);
541
542    // C99 6.10.6 - Pragma Directive.
543    case tok::pp_pragma:
544      return HandlePragmaDirective();
545
546    // GNU Extensions.
547    case tok::pp_import:
548      return HandleImportDirective(Result);
549    case tok::pp_include_next:
550      return HandleIncludeNextDirective(Result);
551
552    case tok::pp_warning:
553      Diag(Result, diag::ext_pp_warning_directive);
554      return HandleUserDiagnosticDirective(Result, true);
555    case tok::pp_ident:
556      return HandleIdentSCCSDirective(Result);
557    case tok::pp_sccs:
558      return HandleIdentSCCSDirective(Result);
559    case tok::pp_assert:
560      //isExtension = true;  // FIXME: implement #assert
561      break;
562    case tok::pp_unassert:
563      //isExtension = true;  // FIXME: implement #unassert
564      break;
565    }
566    break;
567  }
568
569  // If we reached here, the preprocessing token is not valid!
570  Diag(Result, diag::err_pp_invalid_directive);
571
572  // Read the rest of the PP line.
573  DiscardUntilEndOfDirective();
574
575  // Okay, we're done parsing the directive.
576}
577
578/// GetLineValue - Convert a numeric token into an unsigned value, emitting
579/// Diagnostic DiagID if it is invalid, and returning the value in Val.
580static bool GetLineValue(Token &DigitTok, unsigned &Val,
581                         unsigned DiagID, Preprocessor &PP) {
582  if (DigitTok.isNot(tok::numeric_constant)) {
583    PP.Diag(DigitTok, DiagID);
584
585    if (DigitTok.isNot(tok::eom))
586      PP.DiscardUntilEndOfDirective();
587    return true;
588  }
589
590  llvm::SmallString<64> IntegerBuffer;
591  IntegerBuffer.resize(DigitTok.getLength());
592  const char *DigitTokBegin = &IntegerBuffer[0];
593  unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin);
594  NumericLiteralParser Literal(DigitTokBegin, DigitTokBegin+ActualLength,
595                               DigitTok.getLocation(), PP);
596  if (Literal.hadError)
597    return true;   // Error already emitted.
598
599  if (Literal.isFloatingLiteral() || Literal.isImaginary) {
600    PP.Diag(DigitTok, DiagID);
601    return true;
602  }
603
604  // Parse the integer literal into Result.
605  llvm::APInt APVal(32, 0);
606  if (Literal.GetIntegerValue(APVal)) {
607    // Overflow parsing integer literal.
608    PP.Diag(DigitTok, DiagID);
609    return true;
610  }
611  Val = APVal.getZExtValue();
612
613  // Reject 0, this is needed both by #line numbers and flags.
614  if (Val == 0) {
615    PP.Diag(DigitTok, DiagID);
616    PP.DiscardUntilEndOfDirective();
617    return true;
618  }
619
620  return false;
621}
622
623/// HandleLineDirective - Handle #line directive: C99 6.10.4.  The two
624/// acceptable forms are:
625///   # line digit-sequence
626///   # line digit-sequence "s-char-sequence"
627void Preprocessor::HandleLineDirective(Token &Tok) {
628  // Read the line # and string argument.  Per C99 6.10.4p5, these tokens are
629  // expanded.
630  Token DigitTok;
631  Lex(DigitTok);
632
633  // Validate the number and convert it to an unsigned.
634  unsigned LineNo;
635  if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer, *this))
636    return;
637
638  // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
639  // number greater than 2147483647".  C90 requires that the line # be <= 32767.
640  unsigned LineLimit = Features.C99 ? 2147483648U : 32768U;
641  if (LineNo >= LineLimit)
642    Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit;
643
644  int FilenameID = -1;
645  Token StrTok;
646  Lex(StrTok);
647
648  // If the StrTok is "eom", then it wasn't present.  Otherwise, it must be a
649  // string followed by eom.
650  if (StrTok.is(tok::eom))
651    ; // ok
652  else if (StrTok.isNot(tok::string_literal)) {
653    Diag(StrTok, diag::err_pp_line_invalid_filename);
654    DiscardUntilEndOfDirective();
655    return;
656  } else {
657    // Parse and validate the string, converting it into a unique ID.
658    StringLiteralParser Literal(&StrTok, 1, *this);
659    assert(!Literal.AnyWide && "Didn't allow wide strings in");
660    if (Literal.hadError)
661      return DiscardUntilEndOfDirective();
662    if (Literal.Pascal) {
663      Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
664      return DiscardUntilEndOfDirective();
665    }
666    FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(),
667                                                  Literal.GetStringLength());
668
669    // Verify that there is nothing after the string, other than EOM.
670    CheckEndOfDirective("#line");
671  }
672
673  SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID);
674}
675
676/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
677/// marker directive.
678static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
679                                bool &IsSystemHeader, bool &IsExternCHeader,
680                                Preprocessor &PP) {
681  unsigned FlagVal;
682  Token FlagTok;
683  PP.Lex(FlagTok);
684  if (FlagTok.is(tok::eom)) return false;
685  if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
686    return true;
687
688  if (FlagVal == 1) {
689    IsFileEntry = true;
690
691    PP.Lex(FlagTok);
692    if (FlagTok.is(tok::eom)) return false;
693    if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
694      return true;
695  } else if (FlagVal == 2) {
696    IsFileExit = true;
697
698    SourceManager &SM = PP.getSourceManager();
699    // If we are leaving the current presumed file, check to make sure the
700    // presumed include stack isn't empty!
701    FileID CurFileID =
702      SM.getDecomposedInstantiationLoc(FlagTok.getLocation()).first;
703    PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation());
704
705    // If there is no include loc (main file) or if the include loc is in a
706    // different physical file, then we aren't in a "1" line marker flag region.
707    SourceLocation IncLoc = PLoc.getIncludeLoc();
708    if (IncLoc.isInvalid() ||
709        SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) {
710      PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop);
711      PP.DiscardUntilEndOfDirective();
712      return true;
713    }
714
715    PP.Lex(FlagTok);
716    if (FlagTok.is(tok::eom)) return false;
717    if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
718      return true;
719  }
720
721  // We must have 3 if there are still flags.
722  if (FlagVal != 3) {
723    PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
724    PP.DiscardUntilEndOfDirective();
725    return true;
726  }
727
728  IsSystemHeader = true;
729
730  PP.Lex(FlagTok);
731  if (FlagTok.is(tok::eom)) return false;
732  if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
733    return true;
734
735  // We must have 4 if there is yet another flag.
736  if (FlagVal != 4) {
737    PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
738    PP.DiscardUntilEndOfDirective();
739    return true;
740  }
741
742  IsExternCHeader = true;
743
744  PP.Lex(FlagTok);
745  if (FlagTok.is(tok::eom)) return false;
746
747  // There are no more valid flags here.
748  PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
749  PP.DiscardUntilEndOfDirective();
750  return true;
751}
752
753/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
754/// one of the following forms:
755///
756///     # 42
757///     # 42 "file" ('1' | '2')?
758///     # 42 "file" ('1' | '2')? '3' '4'?
759///
760void Preprocessor::HandleDigitDirective(Token &DigitTok) {
761  // Validate the number and convert it to an unsigned.  GNU does not have a
762  // line # limit other than it fit in 32-bits.
763  unsigned LineNo;
764  if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer,
765                   *this))
766    return;
767
768  Token StrTok;
769  Lex(StrTok);
770
771  bool IsFileEntry = false, IsFileExit = false;
772  bool IsSystemHeader = false, IsExternCHeader = false;
773  int FilenameID = -1;
774
775  // If the StrTok is "eom", then it wasn't present.  Otherwise, it must be a
776  // string followed by eom.
777  if (StrTok.is(tok::eom))
778    ; // ok
779  else if (StrTok.isNot(tok::string_literal)) {
780    Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
781    return DiscardUntilEndOfDirective();
782  } else {
783    // Parse and validate the string, converting it into a unique ID.
784    StringLiteralParser Literal(&StrTok, 1, *this);
785    assert(!Literal.AnyWide && "Didn't allow wide strings in");
786    if (Literal.hadError)
787      return DiscardUntilEndOfDirective();
788    if (Literal.Pascal) {
789      Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
790      return DiscardUntilEndOfDirective();
791    }
792    FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(),
793                                                  Literal.GetStringLength());
794
795    // If a filename was present, read any flags that are present.
796    if (ReadLineMarkerFlags(IsFileEntry, IsFileExit,
797                            IsSystemHeader, IsExternCHeader, *this))
798      return;
799  }
800
801  // Create a line note with this information.
802  SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID,
803                        IsFileEntry, IsFileExit,
804                        IsSystemHeader, IsExternCHeader);
805}
806
807
808/// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
809///
810void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
811                                                 bool isWarning) {
812  // PTH doesn't emit #warning or #error directives.
813  if (CurPTHLexer)
814    return CurPTHLexer->DiscardToEndOfLine();
815
816  // Read the rest of the line raw.  We do this because we don't want macros
817  // to be expanded and we don't require that the tokens be valid preprocessing
818  // tokens.  For example, this is allowed: "#warning `   'foo".  GCC does
819  // collapse multiple consequtive white space between tokens, but this isn't
820  // specified by the standard.
821  std::string Message = CurLexer->ReadToEndOfLine();
822  if (isWarning)
823    Diag(Tok, diag::pp_hash_warning) << Message;
824  else
825    Diag(Tok, diag::err_pp_hash_error) << Message;
826}
827
828/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
829///
830void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
831  // Yes, this directive is an extension.
832  Diag(Tok, diag::ext_pp_ident_directive);
833
834  // Read the string argument.
835  Token StrTok;
836  Lex(StrTok);
837
838  // If the token kind isn't a string, it's a malformed directive.
839  if (StrTok.isNot(tok::string_literal) &&
840      StrTok.isNot(tok::wide_string_literal)) {
841    Diag(StrTok, diag::err_pp_malformed_ident);
842    if (StrTok.isNot(tok::eom))
843      DiscardUntilEndOfDirective();
844    return;
845  }
846
847  // Verify that there is nothing after the string, other than EOM.
848  CheckEndOfDirective("#ident");
849
850  if (Callbacks)
851    Callbacks->Ident(Tok.getLocation(), getSpelling(StrTok));
852}
853
854//===----------------------------------------------------------------------===//
855// Preprocessor Include Directive Handling.
856//===----------------------------------------------------------------------===//
857
858/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
859/// checked and spelled filename, e.g. as an operand of #include. This returns
860/// true if the input filename was in <>'s or false if it were in ""'s.  The
861/// caller is expected to provide a buffer that is large enough to hold the
862/// spelling of the filename, but is also expected to handle the case when
863/// this method decides to use a different buffer.
864bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
865                                              const char *&BufStart,
866                                              const char *&BufEnd) {
867  // Get the text form of the filename.
868  assert(BufStart != BufEnd && "Can't have tokens with empty spellings!");
869
870  // Make sure the filename is <x> or "x".
871  bool isAngled;
872  if (BufStart[0] == '<') {
873    if (BufEnd[-1] != '>') {
874      Diag(Loc, diag::err_pp_expects_filename);
875      BufStart = 0;
876      return true;
877    }
878    isAngled = true;
879  } else if (BufStart[0] == '"') {
880    if (BufEnd[-1] != '"') {
881      Diag(Loc, diag::err_pp_expects_filename);
882      BufStart = 0;
883      return true;
884    }
885    isAngled = false;
886  } else {
887    Diag(Loc, diag::err_pp_expects_filename);
888    BufStart = 0;
889    return true;
890  }
891
892  // Diagnose #include "" as invalid.
893  if (BufEnd-BufStart <= 2) {
894    Diag(Loc, diag::err_pp_empty_filename);
895    BufStart = 0;
896    return "";
897  }
898
899  // Skip the brackets.
900  ++BufStart;
901  --BufEnd;
902  return isAngled;
903}
904
905/// ConcatenateIncludeName - Handle cases where the #include name is expanded
906/// from a macro as multiple tokens, which need to be glued together.  This
907/// occurs for code like:
908///    #define FOO <a/b.h>
909///    #include FOO
910/// because in this case, "<a/b.h>" is returned as 7 tokens, not one.
911///
912/// This code concatenates and consumes tokens up to the '>' token.  It returns
913/// false if the > was found, otherwise it returns true if it finds and consumes
914/// the EOM marker.
915static bool ConcatenateIncludeName(llvm::SmallVector<char, 128> &FilenameBuffer,
916                                   Preprocessor &PP) {
917  Token CurTok;
918
919  PP.Lex(CurTok);
920  while (CurTok.isNot(tok::eom)) {
921    // Append the spelling of this token to the buffer. If there was a space
922    // before it, add it now.
923    if (CurTok.hasLeadingSpace())
924      FilenameBuffer.push_back(' ');
925
926    // Get the spelling of the token, directly into FilenameBuffer if possible.
927    unsigned PreAppendSize = FilenameBuffer.size();
928    FilenameBuffer.resize(PreAppendSize+CurTok.getLength());
929
930    const char *BufPtr = &FilenameBuffer[PreAppendSize];
931    unsigned ActualLen = PP.getSpelling(CurTok, BufPtr);
932
933    // If the token was spelled somewhere else, copy it into FilenameBuffer.
934    if (BufPtr != &FilenameBuffer[PreAppendSize])
935      memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
936
937    // Resize FilenameBuffer to the correct size.
938    if (CurTok.getLength() != ActualLen)
939      FilenameBuffer.resize(PreAppendSize+ActualLen);
940
941    // If we found the '>' marker, return success.
942    if (CurTok.is(tok::greater))
943      return false;
944
945    PP.Lex(CurTok);
946  }
947
948  // If we hit the eom marker, emit an error and return true so that the caller
949  // knows the EOM has been read.
950  PP.Diag(CurTok.getLocation(), diag::err_pp_expects_filename);
951  return true;
952}
953
954/// HandleIncludeDirective - The "#include" tokens have just been read, read the
955/// file to be included from the lexer, then include it!  This is a common
956/// routine with functionality shared between #include, #include_next and
957/// #import.  LookupFrom is set when this is a #include_next directive, it
958/// specifies the file to start searching from.
959void Preprocessor::HandleIncludeDirective(Token &IncludeTok,
960                                          const DirectoryLookup *LookupFrom,
961                                          bool isImport) {
962
963  Token FilenameTok;
964  CurPPLexer->LexIncludeFilename(FilenameTok);
965
966  // Reserve a buffer to get the spelling.
967  llvm::SmallVector<char, 128> FilenameBuffer;
968  const char *FilenameStart, *FilenameEnd;
969
970  switch (FilenameTok.getKind()) {
971  case tok::eom:
972    // If the token kind is EOM, the error has already been diagnosed.
973    return;
974
975  case tok::angle_string_literal:
976  case tok::string_literal: {
977    FilenameBuffer.resize(FilenameTok.getLength());
978    FilenameStart = &FilenameBuffer[0];
979    unsigned Len = getSpelling(FilenameTok, FilenameStart);
980    FilenameEnd = FilenameStart+Len;
981    break;
982  }
983
984  case tok::less:
985    // This could be a <foo/bar.h> file coming from a macro expansion.  In this
986    // case, glue the tokens together into FilenameBuffer and interpret those.
987    FilenameBuffer.push_back('<');
988    if (ConcatenateIncludeName(FilenameBuffer, *this))
989      return;   // Found <eom> but no ">"?  Diagnostic already emitted.
990    FilenameStart = &FilenameBuffer[0];
991    FilenameEnd = &FilenameBuffer[FilenameBuffer.size()];
992    break;
993  default:
994    Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
995    DiscardUntilEndOfDirective();
996    return;
997  }
998
999  bool isAngled = GetIncludeFilenameSpelling(FilenameTok.getLocation(),
1000                                             FilenameStart, FilenameEnd);
1001  // If GetIncludeFilenameSpelling set the start ptr to null, there was an
1002  // error.
1003  if (FilenameStart == 0) {
1004    DiscardUntilEndOfDirective();
1005    return;
1006  }
1007
1008  // Verify that there is nothing after the filename, other than EOM.  Use the
1009  // preprocessor to lex this in case lexing the filename entered a macro.
1010  CheckEndOfDirective("#include");
1011
1012  // Check that we don't have infinite #include recursion.
1013  if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
1014    Diag(FilenameTok, diag::err_pp_include_too_deep);
1015    return;
1016  }
1017
1018  // Search include directories.
1019  const DirectoryLookup *CurDir;
1020  const FileEntry *File = LookupFile(FilenameStart, FilenameEnd,
1021                                     isAngled, LookupFrom, CurDir);
1022  if (File == 0) {
1023    Diag(FilenameTok, diag::err_pp_file_not_found)
1024       << std::string(FilenameStart, FilenameEnd);
1025    return;
1026  }
1027
1028  // Ask HeaderInfo if we should enter this #include file.  If not, #including
1029  // this file will have no effect.
1030  if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport))
1031    return;
1032
1033  // The #included file will be considered to be a system header if either it is
1034  // in a system include directory, or if the #includer is a system include
1035  // header.
1036  SrcMgr::CharacteristicKind FileCharacter =
1037    std::max(HeaderInfo.getFileDirFlavor(File),
1038             SourceMgr.getFileCharacteristic(FilenameTok.getLocation()));
1039
1040  // Look up the file, create a File ID for it.
1041  FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(),
1042                                      FileCharacter);
1043  if (FID.isInvalid()) {
1044    Diag(FilenameTok, diag::err_pp_file_not_found)
1045      << std::string(FilenameStart, FilenameEnd);
1046    return;
1047  }
1048
1049  // Finally, if all is good, enter the new file!
1050  EnterSourceFile(FID, CurDir);
1051}
1052
1053/// HandleIncludeNextDirective - Implements #include_next.
1054///
1055void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) {
1056  Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
1057
1058  // #include_next is like #include, except that we start searching after
1059  // the current found directory.  If we can't do this, issue a
1060  // diagnostic.
1061  const DirectoryLookup *Lookup = CurDirLookup;
1062  if (isInPrimaryFile()) {
1063    Lookup = 0;
1064    Diag(IncludeNextTok, diag::pp_include_next_in_primary);
1065  } else if (Lookup == 0) {
1066    Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
1067  } else {
1068    // Start looking up in the next directory.
1069    ++Lookup;
1070  }
1071
1072  return HandleIncludeDirective(IncludeNextTok, Lookup);
1073}
1074
1075/// HandleImportDirective - Implements #import.
1076///
1077void Preprocessor::HandleImportDirective(Token &ImportTok) {
1078  Diag(ImportTok, diag::ext_pp_import_directive);
1079
1080  return HandleIncludeDirective(ImportTok, 0, true);
1081}
1082
1083//===----------------------------------------------------------------------===//
1084// Preprocessor Macro Directive Handling.
1085//===----------------------------------------------------------------------===//
1086
1087/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1088/// definition has just been read.  Lex the rest of the arguments and the
1089/// closing ), updating MI with what we learn.  Return true if an error occurs
1090/// parsing the arg list.
1091bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) {
1092  llvm::SmallVector<IdentifierInfo*, 32> Arguments;
1093
1094  Token Tok;
1095  while (1) {
1096    LexUnexpandedToken(Tok);
1097    switch (Tok.getKind()) {
1098    case tok::r_paren:
1099      // Found the end of the argument list.
1100      if (Arguments.empty())  // #define FOO()
1101        return false;
1102      // Otherwise we have #define FOO(A,)
1103      Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
1104      return true;
1105    case tok::ellipsis:  // #define X(... -> C99 varargs
1106      // Warn if use of C99 feature in non-C99 mode.
1107      if (!Features.C99) Diag(Tok, diag::ext_variadic_macro);
1108
1109      // Lex the token after the identifier.
1110      LexUnexpandedToken(Tok);
1111      if (Tok.isNot(tok::r_paren)) {
1112        Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1113        return true;
1114      }
1115      // Add the __VA_ARGS__ identifier as an argument.
1116      Arguments.push_back(Ident__VA_ARGS__);
1117      MI->setIsC99Varargs();
1118      MI->setArgumentList(&Arguments[0], Arguments.size());
1119      return false;
1120    case tok::eom:  // #define X(
1121      Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1122      return true;
1123    default:
1124      // Handle keywords and identifiers here to accept things like
1125      // #define Foo(for) for.
1126      IdentifierInfo *II = Tok.getIdentifierInfo();
1127      if (II == 0) {
1128        // #define X(1
1129        Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
1130        return true;
1131      }
1132
1133      // If this is already used as an argument, it is used multiple times (e.g.
1134      // #define X(A,A.
1135      if (std::find(Arguments.begin(), Arguments.end(), II) !=
1136          Arguments.end()) {  // C99 6.10.3p6
1137        Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II;
1138        return true;
1139      }
1140
1141      // Add the argument to the macro info.
1142      Arguments.push_back(II);
1143
1144      // Lex the token after the identifier.
1145      LexUnexpandedToken(Tok);
1146
1147      switch (Tok.getKind()) {
1148      default:          // #define X(A B
1149        Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
1150        return true;
1151      case tok::r_paren: // #define X(A)
1152        MI->setArgumentList(&Arguments[0], Arguments.size());
1153        return false;
1154      case tok::comma:  // #define X(A,
1155        break;
1156      case tok::ellipsis:  // #define X(A... -> GCC extension
1157        // Diagnose extension.
1158        Diag(Tok, diag::ext_named_variadic_macro);
1159
1160        // Lex the token after the identifier.
1161        LexUnexpandedToken(Tok);
1162        if (Tok.isNot(tok::r_paren)) {
1163          Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1164          return true;
1165        }
1166
1167        MI->setIsGNUVarargs();
1168        MI->setArgumentList(&Arguments[0], Arguments.size());
1169        return false;
1170      }
1171    }
1172  }
1173}
1174
1175/// HandleDefineDirective - Implements #define.  This consumes the entire macro
1176/// line then lets the caller lex the next real token.
1177void Preprocessor::HandleDefineDirective(Token &DefineTok) {
1178  ++NumDefined;
1179
1180  Token MacroNameTok;
1181  ReadMacroName(MacroNameTok, 1);
1182
1183  // Error reading macro name?  If so, diagnostic already issued.
1184  if (MacroNameTok.is(tok::eom))
1185    return;
1186
1187  // If we are supposed to keep comments in #defines, reenable comment saving
1188  // mode.
1189  if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
1190
1191  // Create the new macro.
1192  MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation());
1193
1194  Token Tok;
1195  LexUnexpandedToken(Tok);
1196
1197  // If this is a function-like macro definition, parse the argument list,
1198  // marking each of the identifiers as being used as macro arguments.  Also,
1199  // check other constraints on the first token of the macro body.
1200  if (Tok.is(tok::eom)) {
1201    // If there is no body to this macro, we have no special handling here.
1202  } else if (Tok.is(tok::l_paren) && !Tok.hasLeadingSpace()) {
1203    // This is a function-like macro definition.  Read the argument list.
1204    MI->setIsFunctionLike();
1205    if (ReadMacroDefinitionArgList(MI)) {
1206      // Forget about MI.
1207      ReleaseMacroInfo(MI);
1208      // Throw away the rest of the line.
1209      if (CurPPLexer->ParsingPreprocessorDirective)
1210        DiscardUntilEndOfDirective();
1211      return;
1212    }
1213
1214    // Read the first token after the arg list for down below.
1215    LexUnexpandedToken(Tok);
1216  } else if (!Tok.hasLeadingSpace()) {
1217    // C99 requires whitespace between the macro definition and the body.  Emit
1218    // a diagnostic for something like "#define X+".
1219    if (Features.C99) {
1220      Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
1221    } else {
1222      // FIXME: C90/C++ do not get this diagnostic, but it does get a similar
1223      // one in some cases!
1224    }
1225  } else {
1226    // This is a normal token with leading space.  Clear the leading space
1227    // marker on the first token to get proper expansion.
1228    Tok.clearFlag(Token::LeadingSpace);
1229  }
1230
1231  // If this is a definition of a variadic C99 function-like macro, not using
1232  // the GNU named varargs extension, enabled __VA_ARGS__.
1233
1234  // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
1235  // This gets unpoisoned where it is allowed.
1236  assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!");
1237  if (MI->isC99Varargs())
1238    Ident__VA_ARGS__->setIsPoisoned(false);
1239
1240  // Read the rest of the macro body.
1241  if (MI->isObjectLike()) {
1242    // Object-like macros are very simple, just read their body.
1243    while (Tok.isNot(tok::eom)) {
1244      MI->AddTokenToBody(Tok);
1245      // Get the next token of the macro.
1246      LexUnexpandedToken(Tok);
1247    }
1248
1249  } else {
1250    // Otherwise, read the body of a function-like macro.  This has to validate
1251    // the # (stringize) operator.
1252    while (Tok.isNot(tok::eom)) {
1253      MI->AddTokenToBody(Tok);
1254
1255      // Check C99 6.10.3.2p1: ensure that # operators are followed by macro
1256      // parameters in function-like macro expansions.
1257      if (Tok.isNot(tok::hash)) {
1258        // Get the next token of the macro.
1259        LexUnexpandedToken(Tok);
1260        continue;
1261      }
1262
1263      // Get the next token of the macro.
1264      LexUnexpandedToken(Tok);
1265
1266      // Not a macro arg identifier?
1267      if (!Tok.getIdentifierInfo() ||
1268          MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) {
1269        Diag(Tok, diag::err_pp_stringize_not_parameter);
1270        ReleaseMacroInfo(MI);
1271
1272        // Disable __VA_ARGS__ again.
1273        Ident__VA_ARGS__->setIsPoisoned(true);
1274        return;
1275      }
1276
1277      // Things look ok, add the param name token to the macro.
1278      MI->AddTokenToBody(Tok);
1279
1280      // Get the next token of the macro.
1281      LexUnexpandedToken(Tok);
1282    }
1283  }
1284
1285
1286  // Disable __VA_ARGS__ again.
1287  Ident__VA_ARGS__->setIsPoisoned(true);
1288
1289  // Check that there is no paste (##) operator at the begining or end of the
1290  // replacement list.
1291  unsigned NumTokens = MI->getNumTokens();
1292  if (NumTokens != 0) {
1293    if (MI->getReplacementToken(0).is(tok::hashhash)) {
1294      Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
1295      ReleaseMacroInfo(MI);
1296      return;
1297    }
1298    if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
1299      Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
1300      ReleaseMacroInfo(MI);
1301      return;
1302    }
1303  }
1304
1305  // If this is the primary source file, remember that this macro hasn't been
1306  // used yet.
1307  if (isInPrimaryFile())
1308    MI->setIsUsed(false);
1309
1310  // Finally, if this identifier already had a macro defined for it, verify that
1311  // the macro bodies are identical and free the old definition.
1312  if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) {
1313    // It is very common for system headers to have tons of macro redefinitions
1314    // and for warnings to be disabled in system headers.  If this is the case,
1315    // then don't bother calling MacroInfo::isIdenticalTo.
1316    if (!Diags.getSuppressSystemWarnings() ||
1317        !SourceMgr.isInSystemHeader(DefineTok.getLocation())) {
1318      if (!OtherMI->isUsed())
1319        Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
1320
1321      // Macros must be identical.  This means all tokes and whitespace
1322      // separation must be the same.  C99 6.10.3.2.
1323      if (!MI->isIdenticalTo(*OtherMI, *this)) {
1324        Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
1325          << MacroNameTok.getIdentifierInfo();
1326        Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition);
1327      }
1328    }
1329
1330    ReleaseMacroInfo(OtherMI);
1331  }
1332
1333  setMacroInfo(MacroNameTok.getIdentifierInfo(), MI);
1334}
1335
1336/// HandleUndefDirective - Implements #undef.
1337///
1338void Preprocessor::HandleUndefDirective(Token &UndefTok) {
1339  ++NumUndefined;
1340
1341  Token MacroNameTok;
1342  ReadMacroName(MacroNameTok, 2);
1343
1344  // Error reading macro name?  If so, diagnostic already issued.
1345  if (MacroNameTok.is(tok::eom))
1346    return;
1347
1348  // Check to see if this is the last token on the #undef line.
1349  CheckEndOfDirective("#undef");
1350
1351  // Okay, we finally have a valid identifier to undef.
1352  MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo());
1353
1354  // If the macro is not defined, this is a noop undef, just return.
1355  if (MI == 0) return;
1356
1357  if (!MI->isUsed())
1358    Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
1359
1360  // Free macro definition.
1361  ReleaseMacroInfo(MI);
1362  setMacroInfo(MacroNameTok.getIdentifierInfo(), 0);
1363}
1364
1365
1366//===----------------------------------------------------------------------===//
1367// Preprocessor Conditional Directive Handling.
1368//===----------------------------------------------------------------------===//
1369
1370/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive.  isIfndef is
1371/// true when this is a #ifndef directive.  ReadAnyTokensBeforeDirective is true
1372/// if any tokens have been returned or pp-directives activated before this
1373/// #ifndef has been lexed.
1374///
1375void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef,
1376                                        bool ReadAnyTokensBeforeDirective) {
1377  ++NumIf;
1378  Token DirectiveTok = Result;
1379
1380  Token MacroNameTok;
1381  ReadMacroName(MacroNameTok);
1382
1383  // Error reading macro name?  If so, diagnostic already issued.
1384  if (MacroNameTok.is(tok::eom)) {
1385    // Skip code until we get to #endif.  This helps with recovery by not
1386    // emitting an error when the #endif is reached.
1387    SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
1388                                 /*Foundnonskip*/false, /*FoundElse*/false);
1389    return;
1390  }
1391
1392  // Check to see if this is the last token on the #if[n]def line.
1393  CheckEndOfDirective(isIfndef ? "#ifndef" : "#ifdef");
1394
1395  if (CurPPLexer->getConditionalStackDepth() == 0) {
1396    // If the start of a top-level #ifdef, inform MIOpt.
1397    if (!ReadAnyTokensBeforeDirective) {
1398      assert(isIfndef && "#ifdef shouldn't reach here");
1399      CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MacroNameTok.getIdentifierInfo());
1400    } else
1401      CurPPLexer->MIOpt.EnterTopLevelConditional();
1402  }
1403
1404  IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
1405  MacroInfo *MI = getMacroInfo(MII);
1406
1407  // If there is a macro, process it.
1408  if (MI)  // Mark it used.
1409    MI->setIsUsed(true);
1410
1411  // Should we include the stuff contained by this directive?
1412  if (!MI == isIfndef) {
1413    // Yes, remember that we are inside a conditional, then lex the next token.
1414    CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), /*wasskip*/false,
1415                                   /*foundnonskip*/true, /*foundelse*/false);
1416  } else {
1417    // No, skip the contents of this block and return the first token after it.
1418    SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
1419                                 /*Foundnonskip*/false,
1420                                 /*FoundElse*/false);
1421  }
1422}
1423
1424/// HandleIfDirective - Implements the #if directive.
1425///
1426void Preprocessor::HandleIfDirective(Token &IfToken,
1427                                     bool ReadAnyTokensBeforeDirective) {
1428  ++NumIf;
1429
1430  // Parse and evaluation the conditional expression.
1431  IdentifierInfo *IfNDefMacro = 0;
1432  bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro);
1433
1434
1435  // If this condition is equivalent to #ifndef X, and if this is the first
1436  // directive seen, handle it for the multiple-include optimization.
1437  if (CurPPLexer->getConditionalStackDepth() == 0) {
1438    if (!ReadAnyTokensBeforeDirective && IfNDefMacro)
1439      CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro);
1440    else
1441      CurPPLexer->MIOpt.EnterTopLevelConditional();
1442  }
1443
1444  // Should we include the stuff contained by this directive?
1445  if (ConditionalTrue) {
1446    // Yes, remember that we are inside a conditional, then lex the next token.
1447    CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
1448                                   /*foundnonskip*/true, /*foundelse*/false);
1449  } else {
1450    // No, skip the contents of this block and return the first token after it.
1451    SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false,
1452                                 /*FoundElse*/false);
1453  }
1454}
1455
1456/// HandleEndifDirective - Implements the #endif directive.
1457///
1458void Preprocessor::HandleEndifDirective(Token &EndifToken) {
1459  ++NumEndif;
1460
1461  // Check that this is the whole directive.
1462  CheckEndOfDirective("#endif");
1463
1464  PPConditionalInfo CondInfo;
1465  if (CurPPLexer->popConditionalLevel(CondInfo)) {
1466    // No conditionals on the stack: this is an #endif without an #if.
1467    Diag(EndifToken, diag::err_pp_endif_without_if);
1468    return;
1469  }
1470
1471  // If this the end of a top-level #endif, inform MIOpt.
1472  if (CurPPLexer->getConditionalStackDepth() == 0)
1473    CurPPLexer->MIOpt.ExitTopLevelConditional();
1474
1475  assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
1476         "This code should only be reachable in the non-skipping case!");
1477}
1478
1479
1480void Preprocessor::HandleElseDirective(Token &Result) {
1481  ++NumElse;
1482
1483  // #else directive in a non-skipping conditional... start skipping.
1484  CheckEndOfDirective("#else");
1485
1486  PPConditionalInfo CI;
1487  if (CurPPLexer->popConditionalLevel(CI)) {
1488    Diag(Result, diag::pp_err_else_without_if);
1489    return;
1490  }
1491
1492  // If this is a top-level #else, inform the MIOpt.
1493  if (CurPPLexer->getConditionalStackDepth() == 0)
1494    CurPPLexer->MIOpt.EnterTopLevelConditional();
1495
1496  // If this is a #else with a #else before it, report the error.
1497  if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
1498
1499  // Finally, skip the rest of the contents of this block and return the first
1500  // token after it.
1501  return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
1502                                      /*FoundElse*/true);
1503}
1504
1505void Preprocessor::HandleElifDirective(Token &ElifToken) {
1506  ++NumElse;
1507
1508  // #elif directive in a non-skipping conditional... start skipping.
1509  // We don't care what the condition is, because we will always skip it (since
1510  // the block immediately before it was included).
1511  DiscardUntilEndOfDirective();
1512
1513  PPConditionalInfo CI;
1514  if (CurPPLexer->popConditionalLevel(CI)) {
1515    Diag(ElifToken, diag::pp_err_elif_without_if);
1516    return;
1517  }
1518
1519  // If this is a top-level #elif, inform the MIOpt.
1520  if (CurPPLexer->getConditionalStackDepth() == 0)
1521    CurPPLexer->MIOpt.EnterTopLevelConditional();
1522
1523  // If this is a #elif with a #else before it, report the error.
1524  if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
1525
1526  // Finally, skip the rest of the contents of this block and return the first
1527  // token after it.
1528  return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
1529                                      /*FoundElse*/CI.FoundElse);
1530}
1531
1532