PPDirectives.cpp revision 2df37b8eaecc5382e0e511a738f88918f48d9454
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/HeaderSearch.h"
16#include "clang/Lex/MacroInfo.h"
17#include "clang/Basic/Diagnostic.h"
18#include "clang/Basic/SourceManager.h"
19using namespace clang;
20
21//===----------------------------------------------------------------------===//
22// Utility Methods for Preprocessor Directive Handling.
23//===----------------------------------------------------------------------===//
24
25/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the
26/// current line until the tok::eom token is found.
27void Preprocessor::DiscardUntilEndOfDirective() {
28  Token Tmp;
29  do {
30    LexUnexpandedToken(Tmp);
31  } while (Tmp.isNot(tok::eom));
32}
33
34/// isCXXNamedOperator - Returns "true" if the token is a named operator in C++.
35static bool isCXXNamedOperator(const std::string &Spelling) {
36  return Spelling == "and" || Spelling == "bitand" || Spelling == "bitor" ||
37    Spelling == "compl" || Spelling == "not" || Spelling == "not_eq" ||
38    Spelling == "or" || Spelling == "xor";
39}
40
41/// ReadMacroName - Lex and validate a macro name, which occurs after a
42/// #define or #undef.  This sets the token kind to eom and discards the rest
43/// of the macro line if the macro name is invalid.  isDefineUndef is 1 if
44/// this is due to a a #define, 2 if #undef directive, 0 if it is something
45/// else (e.g. #ifdef).
46void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) {
47  // Read the token, don't allow macro expansion on it.
48  LexUnexpandedToken(MacroNameTok);
49
50  // Missing macro name?
51  if (MacroNameTok.is(tok::eom)) {
52    Diag(MacroNameTok, diag::err_pp_missing_macro_name);
53    return;
54  }
55
56  IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
57  if (II == 0) {
58    std::string Spelling = getSpelling(MacroNameTok);
59    if (isCXXNamedOperator(Spelling))
60      // C++ 2.5p2: Alternative tokens behave the same as its primary token
61      // except for their spellings.
62      Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling;
63    else
64      Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
65    // Fall through on error.
66  } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) {
67    // Error if defining "defined": C99 6.10.8.4.
68    Diag(MacroNameTok, diag::err_defined_macro_name);
69  } else if (isDefineUndef && II->hasMacroDefinition() &&
70             getMacroInfo(II)->isBuiltinMacro()) {
71    // Error if defining "__LINE__" and other builtins: C99 6.10.8.4.
72    if (isDefineUndef == 1)
73      Diag(MacroNameTok, diag::pp_redef_builtin_macro);
74    else
75      Diag(MacroNameTok, diag::pp_undef_builtin_macro);
76  } else {
77    // Okay, we got a good identifier node.  Return it.
78    return;
79  }
80
81  // Invalid macro name, read and discard the rest of the line.  Then set the
82  // token kind to tok::eom.
83  MacroNameTok.setKind(tok::eom);
84  return DiscardUntilEndOfDirective();
85}
86
87/// CheckEndOfDirective - Ensure that the next token is a tok::eom token.  If
88/// not, emit a diagnostic and consume up until the eom.
89void Preprocessor::CheckEndOfDirective(const char *DirType) {
90  Token Tmp;
91  // Lex unexpanded tokens: macros might expand to zero tokens, causing us to
92  // miss diagnosing invalid lines.
93  LexUnexpandedToken(Tmp);
94
95  // There should be no tokens after the directive, but we allow them as an
96  // extension.
97  while (Tmp.is(tok::comment))  // Skip comments in -C mode.
98    LexUnexpandedToken(Tmp);
99
100  if (Tmp.isNot(tok::eom)) {
101    Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType;
102    DiscardUntilEndOfDirective();
103  }
104}
105
106
107
108/// SkipExcludedConditionalBlock - We just read a #if or related directive and
109/// decided that the subsequent tokens are in the #if'd out portion of the
110/// file.  Lex the rest of the file, until we see an #endif.  If
111/// FoundNonSkipPortion is true, then we have already emitted code for part of
112/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk
113/// is true, then #else directives are ok, if not, then we have already seen one
114/// so a #else directive is a duplicate.  When this returns, the caller can lex
115/// the first valid token.
116void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
117                                                bool FoundNonSkipPortion,
118                                                bool FoundElse) {
119  ++NumSkipped;
120  assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?");
121
122  CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false,
123                                 FoundNonSkipPortion, FoundElse);
124
125  // Enter raw mode to disable identifier lookup (and thus macro expansion),
126  // disabling warnings, etc.
127  CurPPLexer->LexingRawMode = true;
128  Token Tok;
129  while (1) {
130    if (CurLexer)
131      CurLexer->Lex(Tok);
132    else
133      CurPTHLexer->Lex(Tok);
134
135    // If this is the end of the buffer, we have an error.
136    if (Tok.is(tok::eof)) {
137      // Emit errors for each unterminated conditional on the stack, including
138      // the current one.
139      while (!CurPPLexer->ConditionalStack.empty()) {
140        Diag(CurPPLexer->ConditionalStack.back().IfLoc,
141             diag::err_pp_unterminated_conditional);
142        CurPPLexer->ConditionalStack.pop_back();
143      }
144
145      // Just return and let the caller lex after this #include.
146      break;
147    }
148
149    // If this token is not a preprocessor directive, just skip it.
150    if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine())
151      continue;
152
153    // We just parsed a # character at the start of a line, so we're in
154    // directive mode.  Tell the lexer this so any newlines we see will be
155    // converted into an EOM token (this terminates the macro).
156    CurPPLexer->ParsingPreprocessorDirective = true;
157    if (CurLexer) CurLexer->SetCommentRetentionState(false);
158
159
160    // Read the next token, the directive flavor.
161    LexUnexpandedToken(Tok);
162
163    // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
164    // something bogus), skip it.
165    if (Tok.isNot(tok::identifier)) {
166      CurPPLexer->ParsingPreprocessorDirective = false;
167      // Restore comment saving mode.
168      if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
169      continue;
170    }
171
172    // If the first letter isn't i or e, it isn't intesting to us.  We know that
173    // this is safe in the face of spelling differences, because there is no way
174    // to spell an i/e in a strange way that is another letter.  Skipping this
175    // allows us to avoid looking up the identifier info for #define/#undef and
176    // other common directives.
177    const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation());
178    char FirstChar = RawCharData[0];
179    if (FirstChar >= 'a' && FirstChar <= 'z' &&
180        FirstChar != 'i' && FirstChar != 'e') {
181      CurPPLexer->ParsingPreprocessorDirective = false;
182      // Restore comment saving mode.
183      if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
184      continue;
185    }
186
187    // Get the identifier name without trigraphs or embedded newlines.  Note
188    // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
189    // when skipping.
190    // TODO: could do this with zero copies in the no-clean case by using
191    // strncmp below.
192    char Directive[20];
193    unsigned IdLen;
194    if (!Tok.needsCleaning() && Tok.getLength() < 20) {
195      IdLen = Tok.getLength();
196      memcpy(Directive, RawCharData, IdLen);
197      Directive[IdLen] = 0;
198    } else {
199      std::string DirectiveStr = getSpelling(Tok);
200      IdLen = DirectiveStr.size();
201      if (IdLen >= 20) {
202        CurPPLexer->ParsingPreprocessorDirective = false;
203        // Restore comment saving mode.
204        if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
205        continue;
206      }
207      memcpy(Directive, &DirectiveStr[0], IdLen);
208      Directive[IdLen] = 0;
209    }
210
211    if (FirstChar == 'i' && Directive[1] == 'f') {
212      if ((IdLen == 2) ||   // "if"
213          (IdLen == 5 && !strcmp(Directive+2, "def")) ||   // "ifdef"
214          (IdLen == 6 && !strcmp(Directive+2, "ndef"))) {  // "ifndef"
215        // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
216        // bother parsing the condition.
217        DiscardUntilEndOfDirective();
218        CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true,
219                                       /*foundnonskip*/false,
220                                       /*fnddelse*/false);
221      }
222    } else if (FirstChar == 'e') {
223      if (IdLen == 5 && !strcmp(Directive+1, "ndif")) {  // "endif"
224        CheckEndOfDirective("#endif");
225        PPConditionalInfo CondInfo;
226        CondInfo.WasSkipping = true; // Silence bogus warning.
227        bool InCond = CurPPLexer->popConditionalLevel(CondInfo);
228        InCond = InCond;  // Silence warning in no-asserts mode.
229        assert(!InCond && "Can't be skipping if not in a conditional!");
230
231        // If we popped the outermost skipping block, we're done skipping!
232        if (!CondInfo.WasSkipping)
233          break;
234      } else if (IdLen == 4 && !strcmp(Directive+1, "lse")) { // "else".
235        // #else directive in a skipping conditional.  If not in some other
236        // skipping conditional, and if #else hasn't already been seen, enter it
237        // as a non-skipping conditional.
238        CheckEndOfDirective("#else");
239        PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
240
241        // If this is a #else with a #else before it, report the error.
242        if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else);
243
244        // Note that we've seen a #else in this conditional.
245        CondInfo.FoundElse = true;
246
247        // If the conditional is at the top level, and the #if block wasn't
248        // entered, enter the #else block now.
249        if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
250          CondInfo.FoundNonSkip = true;
251          break;
252        }
253      } else if (IdLen == 4 && !strcmp(Directive+1, "lif")) {  // "elif".
254        PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
255
256        bool ShouldEnter;
257        // If this is in a skipping block or if we're already handled this #if
258        // block, don't bother parsing the condition.
259        if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
260          DiscardUntilEndOfDirective();
261          ShouldEnter = false;
262        } else {
263          // Restore the value of LexingRawMode so that identifiers are
264          // looked up, etc, inside the #elif expression.
265          assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
266          CurPPLexer->LexingRawMode = false;
267          IdentifierInfo *IfNDefMacro = 0;
268          ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
269          CurPPLexer->LexingRawMode = true;
270        }
271
272        // If this is a #elif with a #else before it, report the error.
273        if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else);
274
275        // If this condition is true, enter it!
276        if (ShouldEnter) {
277          CondInfo.FoundNonSkip = true;
278          break;
279        }
280      }
281    }
282
283    CurPPLexer->ParsingPreprocessorDirective = false;
284    // Restore comment saving mode.
285    if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
286  }
287
288  // Finally, if we are out of the conditional (saw an #endif or ran off the end
289  // of the file, just stop skipping and return to lexing whatever came after
290  // the #if block.
291  CurPPLexer->LexingRawMode = false;
292}
293
294/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file,
295/// return null on failure.  isAngled indicates whether the file reference is
296/// for system #include's or not (i.e. using <> instead of "").
297const FileEntry *Preprocessor::LookupFile(const char *FilenameStart,
298                                          const char *FilenameEnd,
299                                          bool isAngled,
300                                          const DirectoryLookup *FromDir,
301                                          const DirectoryLookup *&CurDir) {
302  // If the header lookup mechanism may be relative to the current file, pass in
303  // info about where the current file is.
304  const FileEntry *CurFileEnt = 0;
305  if (!FromDir) {
306    SourceLocation FileLoc = getCurrentFileLexer()->getFileLoc();
307    CurFileEnt = SourceMgr.getFileEntryForLoc(FileLoc);
308  }
309
310  // Do a standard file entry lookup.
311  CurDir = CurDirLookup;
312  const FileEntry *FE =
313  HeaderInfo.LookupFile(FilenameStart, FilenameEnd,
314                        isAngled, FromDir, CurDir, CurFileEnt);
315  if (FE) return FE;
316
317  // Otherwise, see if this is a subframework header.  If so, this is relative
318  // to one of the headers on the #include stack.  Walk the list of the current
319  // headers on the #include stack and pass them to HeaderInfo.
320  if (CurLexer && !CurLexer->Is_PragmaLexer) {
321    if ((CurFileEnt = SourceMgr.getFileEntryForLoc(CurLexer->getFileLoc())))
322      if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, FilenameEnd,
323                                                    CurFileEnt)))
324        return FE;
325  }
326
327  for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
328    IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1];
329    if (ISEntry.TheLexer && !ISEntry.TheLexer->Is_PragmaLexer) {
330      if ((CurFileEnt =
331           SourceMgr.getFileEntryForLoc(ISEntry.TheLexer->getFileLoc())))
332        if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart,
333                                                      FilenameEnd, CurFileEnt)))
334          return FE;
335    }
336  }
337
338  // Otherwise, we really couldn't find the file.
339  return 0;
340}
341
342
343//===----------------------------------------------------------------------===//
344// Preprocessor Directive Handling.
345//===----------------------------------------------------------------------===//
346
347/// HandleDirective - This callback is invoked when the lexer sees a # token
348/// at the start of a line.  This consumes the directive, modifies the
349/// lexer/preprocessor state, and advances the lexer(s) so that the next token
350/// read is the correct one.
351void Preprocessor::HandleDirective(Token &Result) {
352  // FIXME: Traditional: # with whitespace before it not recognized by K&R?
353
354  // We just parsed a # character at the start of a line, so we're in directive
355  // mode.  Tell the lexer this so any newlines we see will be converted into an
356  // EOM token (which terminates the directive).
357  CurPPLexer->ParsingPreprocessorDirective = true;
358
359  ++NumDirectives;
360
361  // We are about to read a token.  For the multiple-include optimization FA to
362  // work, we have to remember if we had read any tokens *before* this
363  // pp-directive.
364  bool ReadAnyTokensBeforeDirective = CurPPLexer->MIOpt.getHasReadAnyTokensVal();
365
366  // Read the next token, the directive flavor.  This isn't expanded due to
367  // C99 6.10.3p8.
368  LexUnexpandedToken(Result);
369
370  // C99 6.10.3p11: Is this preprocessor directive in macro invocation?  e.g.:
371  //   #define A(x) #x
372  //   A(abc
373  //     #warning blah
374  //   def)
375  // If so, the user is relying on non-portable behavior, emit a diagnostic.
376  if (InMacroArgs)
377    Diag(Result, diag::ext_embedded_directive);
378
379TryAgain:
380  switch (Result.getKind()) {
381  case tok::eom:
382    return;   // null directive.
383  case tok::comment:
384    // Handle stuff like "# /*foo*/ define X" in -E -C mode.
385    LexUnexpandedToken(Result);
386    goto TryAgain;
387
388  case tok::numeric_constant:
389    // FIXME: implement # 7 line numbers!
390    DiscardUntilEndOfDirective();
391    return;
392  default:
393    IdentifierInfo *II = Result.getIdentifierInfo();
394    if (II == 0) break;  // Not an identifier.
395
396    // Ask what the preprocessor keyword ID is.
397    switch (II->getPPKeywordID()) {
398    default: break;
399    // C99 6.10.1 - Conditional Inclusion.
400    case tok::pp_if:
401      return HandleIfDirective(Result, ReadAnyTokensBeforeDirective);
402    case tok::pp_ifdef:
403      return HandleIfdefDirective(Result, false, true/*not valid for miopt*/);
404    case tok::pp_ifndef:
405      return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective);
406    case tok::pp_elif:
407      return HandleElifDirective(Result);
408    case tok::pp_else:
409      return HandleElseDirective(Result);
410    case tok::pp_endif:
411      return HandleEndifDirective(Result);
412
413    // C99 6.10.2 - Source File Inclusion.
414    case tok::pp_include:
415      return HandleIncludeDirective(Result);            // Handle #include.
416
417    // C99 6.10.3 - Macro Replacement.
418    case tok::pp_define:
419      return HandleDefineDirective(Result);
420    case tok::pp_undef:
421      return HandleUndefDirective(Result);
422
423    // C99 6.10.4 - Line Control.
424    case tok::pp_line:
425      // FIXME: implement #line
426      DiscardUntilEndOfDirective();
427      return;
428
429    // C99 6.10.5 - Error Directive.
430    case tok::pp_error:
431      return HandleUserDiagnosticDirective(Result, false);
432
433    // C99 6.10.6 - Pragma Directive.
434    case tok::pp_pragma:
435      return HandlePragmaDirective();
436
437    // GNU Extensions.
438    case tok::pp_import:
439      return HandleImportDirective(Result);
440    case tok::pp_include_next:
441      return HandleIncludeNextDirective(Result);
442
443    case tok::pp_warning:
444      Diag(Result, diag::ext_pp_warning_directive);
445      return HandleUserDiagnosticDirective(Result, true);
446    case tok::pp_ident:
447      return HandleIdentSCCSDirective(Result);
448    case tok::pp_sccs:
449      return HandleIdentSCCSDirective(Result);
450    case tok::pp_assert:
451      //isExtension = true;  // FIXME: implement #assert
452      break;
453    case tok::pp_unassert:
454      //isExtension = true;  // FIXME: implement #unassert
455      break;
456    }
457    break;
458  }
459
460  // If we reached here, the preprocessing token is not valid!
461  Diag(Result, diag::err_pp_invalid_directive);
462
463  // Read the rest of the PP line.
464  DiscardUntilEndOfDirective();
465
466  // Okay, we're done parsing the directive.
467}
468
469void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
470                                                 bool isWarning) {
471  // Read the rest of the line raw.  We do this because we don't want macros
472  // to be expanded and we don't require that the tokens be valid preprocessing
473  // tokens.  For example, this is allowed: "#warning `   'foo".  GCC does
474  // collapse multiple consequtive white space between tokens, but this isn't
475  // specified by the standard.
476  std::string Message = CurLexer->ReadToEndOfLine();
477
478  unsigned DiagID = isWarning ? diag::pp_hash_warning : diag::err_pp_hash_error;
479  Diag(Tok, DiagID) << Message;
480}
481
482/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
483///
484void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
485  // Yes, this directive is an extension.
486  Diag(Tok, diag::ext_pp_ident_directive);
487
488  // Read the string argument.
489  Token StrTok;
490  Lex(StrTok);
491
492  // If the token kind isn't a string, it's a malformed directive.
493  if (StrTok.isNot(tok::string_literal) &&
494      StrTok.isNot(tok::wide_string_literal)) {
495    Diag(StrTok, diag::err_pp_malformed_ident);
496    return;
497  }
498
499  // Verify that there is nothing after the string, other than EOM.
500  CheckEndOfDirective("#ident");
501
502  if (Callbacks)
503    Callbacks->Ident(Tok.getLocation(), getSpelling(StrTok));
504}
505
506//===----------------------------------------------------------------------===//
507// Preprocessor Include Directive Handling.
508//===----------------------------------------------------------------------===//
509
510/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
511/// checked and spelled filename, e.g. as an operand of #include. This returns
512/// true if the input filename was in <>'s or false if it were in ""'s.  The
513/// caller is expected to provide a buffer that is large enough to hold the
514/// spelling of the filename, but is also expected to handle the case when
515/// this method decides to use a different buffer.
516bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
517                                              const char *&BufStart,
518                                              const char *&BufEnd) {
519  // Get the text form of the filename.
520  assert(BufStart != BufEnd && "Can't have tokens with empty spellings!");
521
522  // Make sure the filename is <x> or "x".
523  bool isAngled;
524  if (BufStart[0] == '<') {
525    if (BufEnd[-1] != '>') {
526      Diag(Loc, diag::err_pp_expects_filename);
527      BufStart = 0;
528      return true;
529    }
530    isAngled = true;
531  } else if (BufStart[0] == '"') {
532    if (BufEnd[-1] != '"') {
533      Diag(Loc, diag::err_pp_expects_filename);
534      BufStart = 0;
535      return true;
536    }
537    isAngled = false;
538  } else {
539    Diag(Loc, diag::err_pp_expects_filename);
540    BufStart = 0;
541    return true;
542  }
543
544  // Diagnose #include "" as invalid.
545  if (BufEnd-BufStart <= 2) {
546    Diag(Loc, diag::err_pp_empty_filename);
547    BufStart = 0;
548    return "";
549  }
550
551  // Skip the brackets.
552  ++BufStart;
553  --BufEnd;
554  return isAngled;
555}
556
557/// ConcatenateIncludeName - Handle cases where the #include name is expanded
558/// from a macro as multiple tokens, which need to be glued together.  This
559/// occurs for code like:
560///    #define FOO <a/b.h>
561///    #include FOO
562/// because in this case, "<a/b.h>" is returned as 7 tokens, not one.
563///
564/// This code concatenates and consumes tokens up to the '>' token.  It returns
565/// false if the > was found, otherwise it returns true if it finds and consumes
566/// the EOM marker.
567static bool ConcatenateIncludeName(llvm::SmallVector<char, 128> &FilenameBuffer,
568                                   Preprocessor &PP) {
569  Token CurTok;
570
571  PP.Lex(CurTok);
572  while (CurTok.isNot(tok::eom)) {
573    // Append the spelling of this token to the buffer. If there was a space
574    // before it, add it now.
575    if (CurTok.hasLeadingSpace())
576      FilenameBuffer.push_back(' ');
577
578    // Get the spelling of the token, directly into FilenameBuffer if possible.
579    unsigned PreAppendSize = FilenameBuffer.size();
580    FilenameBuffer.resize(PreAppendSize+CurTok.getLength());
581
582    const char *BufPtr = &FilenameBuffer[PreAppendSize];
583    unsigned ActualLen = PP.getSpelling(CurTok, BufPtr);
584
585    // If the token was spelled somewhere else, copy it into FilenameBuffer.
586    if (BufPtr != &FilenameBuffer[PreAppendSize])
587      memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
588
589    // Resize FilenameBuffer to the correct size.
590    if (CurTok.getLength() != ActualLen)
591      FilenameBuffer.resize(PreAppendSize+ActualLen);
592
593    // If we found the '>' marker, return success.
594    if (CurTok.is(tok::greater))
595      return false;
596
597    PP.Lex(CurTok);
598  }
599
600  // If we hit the eom marker, emit an error and return true so that the caller
601  // knows the EOM has been read.
602  PP.Diag(CurTok.getLocation(), diag::err_pp_expects_filename);
603  return true;
604}
605
606/// HandleIncludeDirective - The "#include" tokens have just been read, read the
607/// file to be included from the lexer, then include it!  This is a common
608/// routine with functionality shared between #include, #include_next and
609/// #import.  LookupFrom is set when this is a #include_next directive, it
610/// specifies the file to start searching from.
611void Preprocessor::HandleIncludeDirective(Token &IncludeTok,
612                                          const DirectoryLookup *LookupFrom,
613                                          bool isImport) {
614
615  Token FilenameTok;
616  CurPPLexer->LexIncludeFilename(FilenameTok);
617
618  // Reserve a buffer to get the spelling.
619  llvm::SmallVector<char, 128> FilenameBuffer;
620  const char *FilenameStart, *FilenameEnd;
621
622  switch (FilenameTok.getKind()) {
623  case tok::eom:
624    // If the token kind is EOM, the error has already been diagnosed.
625    return;
626
627  case tok::angle_string_literal:
628  case tok::string_literal: {
629    FilenameBuffer.resize(FilenameTok.getLength());
630    FilenameStart = &FilenameBuffer[0];
631    unsigned Len = getSpelling(FilenameTok, FilenameStart);
632    FilenameEnd = FilenameStart+Len;
633    break;
634  }
635
636  case tok::less:
637    // This could be a <foo/bar.h> file coming from a macro expansion.  In this
638    // case, glue the tokens together into FilenameBuffer and interpret those.
639    FilenameBuffer.push_back('<');
640    if (ConcatenateIncludeName(FilenameBuffer, *this))
641      return;   // Found <eom> but no ">"?  Diagnostic already emitted.
642    FilenameStart = &FilenameBuffer[0];
643    FilenameEnd = &FilenameBuffer[FilenameBuffer.size()];
644    break;
645  default:
646    Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
647    DiscardUntilEndOfDirective();
648    return;
649  }
650
651  bool isAngled = GetIncludeFilenameSpelling(FilenameTok.getLocation(),
652                                             FilenameStart, FilenameEnd);
653  // If GetIncludeFilenameSpelling set the start ptr to null, there was an
654  // error.
655  if (FilenameStart == 0) {
656    DiscardUntilEndOfDirective();
657    return;
658  }
659
660  // Verify that there is nothing after the filename, other than EOM.  Use the
661  // preprocessor to lex this in case lexing the filename entered a macro.
662  CheckEndOfDirective("#include");
663
664  // Check that we don't have infinite #include recursion.
665  if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
666    Diag(FilenameTok, diag::err_pp_include_too_deep);
667    return;
668  }
669
670  // Search include directories.
671  const DirectoryLookup *CurDir;
672  const FileEntry *File = LookupFile(FilenameStart, FilenameEnd,
673                                     isAngled, LookupFrom, CurDir);
674  if (File == 0) {
675    Diag(FilenameTok, diag::err_pp_file_not_found)
676       << std::string(FilenameStart, FilenameEnd);
677    return;
678  }
679
680  // Ask HeaderInfo if we should enter this #include file.  If not, #including
681  // this file will have no effect.
682  if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport))
683    return;
684
685  // The #included file will be considered to be a system header if either it is
686  // in a system include directory, or if the #includer is a system include
687  // header.
688  SrcMgr::CharacteristicKind FileCharacter =
689    std::max(HeaderInfo.getFileDirFlavor(File),
690          SourceMgr.getFileCharacteristic(getCurrentFileLexer()->getFileLoc()));
691
692  // Look up the file, create a File ID for it.
693  unsigned FileID = SourceMgr.createFileID(File, FilenameTok.getLocation(),
694                                           FileCharacter);
695  if (FileID == 0) {
696    Diag(FilenameTok, diag::err_pp_file_not_found)
697      << std::string(FilenameStart, FilenameEnd);
698    return;
699  }
700
701  // Finally, if all is good, enter the new file!
702  EnterSourceFile(FileID, CurDir);
703}
704
705/// HandleIncludeNextDirective - Implements #include_next.
706///
707void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) {
708  Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
709
710  // #include_next is like #include, except that we start searching after
711  // the current found directory.  If we can't do this, issue a
712  // diagnostic.
713  const DirectoryLookup *Lookup = CurDirLookup;
714  if (isInPrimaryFile()) {
715    Lookup = 0;
716    Diag(IncludeNextTok, diag::pp_include_next_in_primary);
717  } else if (Lookup == 0) {
718    Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
719  } else {
720    // Start looking up in the next directory.
721    ++Lookup;
722  }
723
724  return HandleIncludeDirective(IncludeNextTok, Lookup);
725}
726
727/// HandleImportDirective - Implements #import.
728///
729void Preprocessor::HandleImportDirective(Token &ImportTok) {
730  Diag(ImportTok, diag::ext_pp_import_directive);
731
732  return HandleIncludeDirective(ImportTok, 0, true);
733}
734
735//===----------------------------------------------------------------------===//
736// Preprocessor Macro Directive Handling.
737//===----------------------------------------------------------------------===//
738
739/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
740/// definition has just been read.  Lex the rest of the arguments and the
741/// closing ), updating MI with what we learn.  Return true if an error occurs
742/// parsing the arg list.
743bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) {
744  llvm::SmallVector<IdentifierInfo*, 32> Arguments;
745
746  Token Tok;
747  while (1) {
748    LexUnexpandedToken(Tok);
749    switch (Tok.getKind()) {
750    case tok::r_paren:
751      // Found the end of the argument list.
752      if (Arguments.empty()) {  // #define FOO()
753        MI->setArgumentList(Arguments.begin(), Arguments.end());
754        return false;
755      }
756      // Otherwise we have #define FOO(A,)
757      Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
758      return true;
759    case tok::ellipsis:  // #define X(... -> C99 varargs
760      // Warn if use of C99 feature in non-C99 mode.
761      if (!Features.C99) Diag(Tok, diag::ext_variadic_macro);
762
763      // Lex the token after the identifier.
764      LexUnexpandedToken(Tok);
765      if (Tok.isNot(tok::r_paren)) {
766        Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
767        return true;
768      }
769      // Add the __VA_ARGS__ identifier as an argument.
770      Arguments.push_back(Ident__VA_ARGS__);
771      MI->setIsC99Varargs();
772      MI->setArgumentList(Arguments.begin(), Arguments.end());
773      return false;
774    case tok::eom:  // #define X(
775      Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
776      return true;
777    default:
778      // Handle keywords and identifiers here to accept things like
779      // #define Foo(for) for.
780      IdentifierInfo *II = Tok.getIdentifierInfo();
781      if (II == 0) {
782        // #define X(1
783        Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
784        return true;
785      }
786
787      // If this is already used as an argument, it is used multiple times (e.g.
788      // #define X(A,A.
789      if (std::find(Arguments.begin(), Arguments.end(), II) !=
790          Arguments.end()) {  // C99 6.10.3p6
791        Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II->getName();
792        return true;
793      }
794
795      // Add the argument to the macro info.
796      Arguments.push_back(II);
797
798      // Lex the token after the identifier.
799      LexUnexpandedToken(Tok);
800
801      switch (Tok.getKind()) {
802      default:          // #define X(A B
803        Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
804        return true;
805      case tok::r_paren: // #define X(A)
806        MI->setArgumentList(Arguments.begin(), Arguments.end());
807        return false;
808      case tok::comma:  // #define X(A,
809        break;
810      case tok::ellipsis:  // #define X(A... -> GCC extension
811        // Diagnose extension.
812        Diag(Tok, diag::ext_named_variadic_macro);
813
814        // Lex the token after the identifier.
815        LexUnexpandedToken(Tok);
816        if (Tok.isNot(tok::r_paren)) {
817          Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
818          return true;
819        }
820
821        MI->setIsGNUVarargs();
822        MI->setArgumentList(Arguments.begin(), Arguments.end());
823        return false;
824      }
825    }
826  }
827}
828
829/// HandleDefineDirective - Implements #define.  This consumes the entire macro
830/// line then lets the caller lex the next real token.
831void Preprocessor::HandleDefineDirective(Token &DefineTok) {
832  ++NumDefined;
833
834  Token MacroNameTok;
835  ReadMacroName(MacroNameTok, 1);
836
837  // Error reading macro name?  If so, diagnostic already issued.
838  if (MacroNameTok.is(tok::eom))
839    return;
840
841  // If we are supposed to keep comments in #defines, reenable comment saving
842  // mode.
843  if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
844
845  // Create the new macro.
846  MacroInfo *MI = new MacroInfo(MacroNameTok.getLocation());
847
848  Token Tok;
849  LexUnexpandedToken(Tok);
850
851  // If this is a function-like macro definition, parse the argument list,
852  // marking each of the identifiers as being used as macro arguments.  Also,
853  // check other constraints on the first token of the macro body.
854  if (Tok.is(tok::eom)) {
855    // If there is no body to this macro, we have no special handling here.
856  } else if (Tok.is(tok::l_paren) && !Tok.hasLeadingSpace()) {
857    // This is a function-like macro definition.  Read the argument list.
858    MI->setIsFunctionLike();
859    if (ReadMacroDefinitionArgList(MI)) {
860      // Forget about MI.
861      delete MI;
862      // Throw away the rest of the line.
863      if (CurPPLexer->ParsingPreprocessorDirective)
864        DiscardUntilEndOfDirective();
865      return;
866    }
867
868    // Read the first token after the arg list for down below.
869    LexUnexpandedToken(Tok);
870  } else if (!Tok.hasLeadingSpace()) {
871    // C99 requires whitespace between the macro definition and the body.  Emit
872    // a diagnostic for something like "#define X+".
873    if (Features.C99) {
874      Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
875    } else {
876      // FIXME: C90/C++ do not get this diagnostic, but it does get a similar
877      // one in some cases!
878    }
879  } else {
880    // This is a normal token with leading space.  Clear the leading space
881    // marker on the first token to get proper expansion.
882    Tok.clearFlag(Token::LeadingSpace);
883  }
884
885  // If this is a definition of a variadic C99 function-like macro, not using
886  // the GNU named varargs extension, enabled __VA_ARGS__.
887
888  // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
889  // This gets unpoisoned where it is allowed.
890  assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!");
891  if (MI->isC99Varargs())
892    Ident__VA_ARGS__->setIsPoisoned(false);
893
894  // Read the rest of the macro body.
895  if (MI->isObjectLike()) {
896    // Object-like macros are very simple, just read their body.
897    while (Tok.isNot(tok::eom)) {
898      MI->AddTokenToBody(Tok);
899      // Get the next token of the macro.
900      LexUnexpandedToken(Tok);
901    }
902
903  } else {
904    // Otherwise, read the body of a function-like macro.  This has to validate
905    // the # (stringize) operator.
906    while (Tok.isNot(tok::eom)) {
907      MI->AddTokenToBody(Tok);
908
909      // Check C99 6.10.3.2p1: ensure that # operators are followed by macro
910      // parameters in function-like macro expansions.
911      if (Tok.isNot(tok::hash)) {
912        // Get the next token of the macro.
913        LexUnexpandedToken(Tok);
914        continue;
915      }
916
917      // Get the next token of the macro.
918      LexUnexpandedToken(Tok);
919
920      // Not a macro arg identifier?
921      if (!Tok.getIdentifierInfo() ||
922          MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) {
923        Diag(Tok, diag::err_pp_stringize_not_parameter);
924        delete MI;
925
926        // Disable __VA_ARGS__ again.
927        Ident__VA_ARGS__->setIsPoisoned(true);
928        return;
929      }
930
931      // Things look ok, add the param name token to the macro.
932      MI->AddTokenToBody(Tok);
933
934      // Get the next token of the macro.
935      LexUnexpandedToken(Tok);
936    }
937  }
938
939
940  // Disable __VA_ARGS__ again.
941  Ident__VA_ARGS__->setIsPoisoned(true);
942
943  // Check that there is no paste (##) operator at the begining or end of the
944  // replacement list.
945  unsigned NumTokens = MI->getNumTokens();
946  if (NumTokens != 0) {
947    if (MI->getReplacementToken(0).is(tok::hashhash)) {
948      Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
949      delete MI;
950      return;
951    }
952    if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
953      Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
954      delete MI;
955      return;
956    }
957  }
958
959  // If this is the primary source file, remember that this macro hasn't been
960  // used yet.
961  if (isInPrimaryFile())
962    MI->setIsUsed(false);
963
964  // Finally, if this identifier already had a macro defined for it, verify that
965  // the macro bodies are identical and free the old definition.
966  if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) {
967    if (!OtherMI->isUsed())
968      Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
969
970    // Macros must be identical.  This means all tokes and whitespace separation
971    // must be the same.  C99 6.10.3.2.
972    if (!MI->isIdenticalTo(*OtherMI, *this)) {
973      Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
974        << MacroNameTok.getIdentifierInfo()->getName();
975      Diag(OtherMI->getDefinitionLoc(), diag::ext_pp_macro_redef2);
976    }
977    delete OtherMI;
978  }
979
980  setMacroInfo(MacroNameTok.getIdentifierInfo(), MI);
981}
982
983/// HandleUndefDirective - Implements #undef.
984///
985void Preprocessor::HandleUndefDirective(Token &UndefTok) {
986  ++NumUndefined;
987
988  Token MacroNameTok;
989  ReadMacroName(MacroNameTok, 2);
990
991  // Error reading macro name?  If so, diagnostic already issued.
992  if (MacroNameTok.is(tok::eom))
993    return;
994
995  // Check to see if this is the last token on the #undef line.
996  CheckEndOfDirective("#undef");
997
998  // Okay, we finally have a valid identifier to undef.
999  MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo());
1000
1001  // If the macro is not defined, this is a noop undef, just return.
1002  if (MI == 0) return;
1003
1004  if (!MI->isUsed())
1005    Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
1006
1007  // Free macro definition.
1008  delete MI;
1009  setMacroInfo(MacroNameTok.getIdentifierInfo(), 0);
1010}
1011
1012
1013//===----------------------------------------------------------------------===//
1014// Preprocessor Conditional Directive Handling.
1015//===----------------------------------------------------------------------===//
1016
1017/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive.  isIfndef is
1018/// true when this is a #ifndef directive.  ReadAnyTokensBeforeDirective is true
1019/// if any tokens have been returned or pp-directives activated before this
1020/// #ifndef has been lexed.
1021///
1022void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef,
1023                                        bool ReadAnyTokensBeforeDirective) {
1024  ++NumIf;
1025  Token DirectiveTok = Result;
1026
1027  Token MacroNameTok;
1028  ReadMacroName(MacroNameTok);
1029
1030  // Error reading macro name?  If so, diagnostic already issued.
1031  if (MacroNameTok.is(tok::eom)) {
1032    // Skip code until we get to #endif.  This helps with recovery by not
1033    // emitting an error when the #endif is reached.
1034    SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
1035                                 /*Foundnonskip*/false, /*FoundElse*/false);
1036    return;
1037  }
1038
1039  // Check to see if this is the last token on the #if[n]def line.
1040  CheckEndOfDirective(isIfndef ? "#ifndef" : "#ifdef");
1041
1042  if (CurPPLexer->getConditionalStackDepth() == 0) {
1043    // If the start of a top-level #ifdef, inform MIOpt.
1044    if (!ReadAnyTokensBeforeDirective) {
1045      assert(isIfndef && "#ifdef shouldn't reach here");
1046      CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MacroNameTok.getIdentifierInfo());
1047    } else
1048      CurPPLexer->MIOpt.EnterTopLevelConditional();
1049  }
1050
1051  IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
1052  MacroInfo *MI = getMacroInfo(MII);
1053
1054  // If there is a macro, process it.
1055  if (MI)  // Mark it used.
1056    MI->setIsUsed(true);
1057
1058  // Should we include the stuff contained by this directive?
1059  if (!MI == isIfndef) {
1060    // Yes, remember that we are inside a conditional, then lex the next token.
1061    CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), /*wasskip*/false,
1062                                   /*foundnonskip*/true, /*foundelse*/false);
1063  } else {
1064    // No, skip the contents of this block and return the first token after it.
1065    SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
1066                                 /*Foundnonskip*/false,
1067                                 /*FoundElse*/false);
1068  }
1069}
1070
1071/// HandleIfDirective - Implements the #if directive.
1072///
1073void Preprocessor::HandleIfDirective(Token &IfToken,
1074                                     bool ReadAnyTokensBeforeDirective) {
1075  ++NumIf;
1076
1077  // Parse and evaluation the conditional expression.
1078  IdentifierInfo *IfNDefMacro = 0;
1079  bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro);
1080
1081
1082  // If this condition is equivalent to #ifndef X, and if this is the first
1083  // directive seen, handle it for the multiple-include optimization.
1084  if (CurPPLexer->getConditionalStackDepth() == 0) {
1085    if (!ReadAnyTokensBeforeDirective && IfNDefMacro)
1086      CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro);
1087    else
1088      CurPPLexer->MIOpt.EnterTopLevelConditional();
1089  }
1090
1091  // Should we include the stuff contained by this directive?
1092  if (ConditionalTrue) {
1093    // Yes, remember that we are inside a conditional, then lex the next token.
1094    CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
1095                                   /*foundnonskip*/true, /*foundelse*/false);
1096  } else {
1097    // No, skip the contents of this block and return the first token after it.
1098    SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false,
1099                                 /*FoundElse*/false);
1100  }
1101}
1102
1103/// HandleEndifDirective - Implements the #endif directive.
1104///
1105void Preprocessor::HandleEndifDirective(Token &EndifToken) {
1106  ++NumEndif;
1107
1108  // Check that this is the whole directive.
1109  CheckEndOfDirective("#endif");
1110
1111  PPConditionalInfo CondInfo;
1112  if (CurPPLexer->popConditionalLevel(CondInfo)) {
1113    // No conditionals on the stack: this is an #endif without an #if.
1114    Diag(EndifToken, diag::err_pp_endif_without_if);
1115    return;
1116  }
1117
1118  // If this the end of a top-level #endif, inform MIOpt.
1119  if (CurPPLexer->getConditionalStackDepth() == 0)
1120    CurPPLexer->MIOpt.ExitTopLevelConditional();
1121
1122  assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
1123         "This code should only be reachable in the non-skipping case!");
1124}
1125
1126
1127void Preprocessor::HandleElseDirective(Token &Result) {
1128  ++NumElse;
1129
1130  // #else directive in a non-skipping conditional... start skipping.
1131  CheckEndOfDirective("#else");
1132
1133  PPConditionalInfo CI;
1134  if (CurPPLexer->popConditionalLevel(CI)) {
1135    Diag(Result, diag::pp_err_else_without_if);
1136    return;
1137  }
1138
1139  // If this is a top-level #else, inform the MIOpt.
1140  if (CurPPLexer->getConditionalStackDepth() == 0)
1141    CurPPLexer->MIOpt.EnterTopLevelConditional();
1142
1143  // If this is a #else with a #else before it, report the error.
1144  if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
1145
1146  // Finally, skip the rest of the contents of this block and return the first
1147  // token after it.
1148  return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
1149                                      /*FoundElse*/true);
1150}
1151
1152void Preprocessor::HandleElifDirective(Token &ElifToken) {
1153  ++NumElse;
1154
1155  // #elif directive in a non-skipping conditional... start skipping.
1156  // We don't care what the condition is, because we will always skip it (since
1157  // the block immediately before it was included).
1158  DiscardUntilEndOfDirective();
1159
1160  PPConditionalInfo CI;
1161  if (CurPPLexer->popConditionalLevel(CI)) {
1162    Diag(ElifToken, diag::pp_err_elif_without_if);
1163    return;
1164  }
1165
1166  // If this is a top-level #elif, inform the MIOpt.
1167  if (CurPPLexer->getConditionalStackDepth() == 0)
1168    CurPPLexer->MIOpt.EnterTopLevelConditional();
1169
1170  // If this is a #elif with a #else before it, report the error.
1171  if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
1172
1173  // Finally, skip the rest of the contents of this block and return the first
1174  // token after it.
1175  return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
1176                                      /*FoundElse*/CI.FoundElse);
1177}
1178
1179