• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 /// \file
11 /// \brief Implements # directive processing for the Preprocessor.
12 ///
13 //===----------------------------------------------------------------------===//
14 
15 #include "clang/Lex/Preprocessor.h"
16 #include "clang/Lex/LiteralSupport.h"
17 #include "clang/Lex/HeaderSearch.h"
18 #include "clang/Lex/MacroInfo.h"
19 #include "clang/Lex/LexDiagnostic.h"
20 #include "clang/Lex/CodeCompletionHandler.h"
21 #include "clang/Lex/ModuleLoader.h"
22 #include "clang/Lex/Pragma.h"
23 #include "clang/Basic/FileManager.h"
24 #include "clang/Basic/SourceManager.h"
25 #include "llvm/ADT/APInt.h"
26 #include "llvm/Support/ErrorHandling.h"
27 using namespace clang;
28 
29 //===----------------------------------------------------------------------===//
30 // Utility Methods for Preprocessor Directive Handling.
31 //===----------------------------------------------------------------------===//
32 
AllocateMacroInfo()33 MacroInfo *Preprocessor::AllocateMacroInfo() {
34   MacroInfoChain *MIChain;
35 
36   if (MICache) {
37     MIChain = MICache;
38     MICache = MICache->Next;
39   }
40   else {
41     MIChain = BP.Allocate<MacroInfoChain>();
42   }
43 
44   MIChain->Next = MIChainHead;
45   MIChain->Prev = 0;
46   if (MIChainHead)
47     MIChainHead->Prev = MIChain;
48   MIChainHead = MIChain;
49 
50   return &(MIChain->MI);
51 }
52 
AllocateMacroInfo(SourceLocation L)53 MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) {
54   MacroInfo *MI = AllocateMacroInfo();
55   new (MI) MacroInfo(L);
56   return MI;
57 }
58 
CloneMacroInfo(const MacroInfo & MacroToClone)59 MacroInfo *Preprocessor::CloneMacroInfo(const MacroInfo &MacroToClone) {
60   MacroInfo *MI = AllocateMacroInfo();
61   new (MI) MacroInfo(MacroToClone, BP);
62   return MI;
63 }
64 
65 /// \brief Release the specified MacroInfo to be reused for allocating
66 /// new MacroInfo objects.
ReleaseMacroInfo(MacroInfo * MI)67 void Preprocessor::ReleaseMacroInfo(MacroInfo *MI) {
68   MacroInfoChain *MIChain = (MacroInfoChain*) MI;
69   if (MacroInfoChain *Prev = MIChain->Prev) {
70     MacroInfoChain *Next = MIChain->Next;
71     Prev->Next = Next;
72     if (Next)
73       Next->Prev = Prev;
74   }
75   else {
76     assert(MIChainHead == MIChain);
77     MIChainHead = MIChain->Next;
78     MIChainHead->Prev = 0;
79   }
80   MIChain->Next = MICache;
81   MICache = MIChain;
82 
83   MI->Destroy();
84 }
85 
86 /// \brief Read and discard all tokens remaining on the current line until
87 /// the tok::eod token is found.
DiscardUntilEndOfDirective()88 void Preprocessor::DiscardUntilEndOfDirective() {
89   Token Tmp;
90   do {
91     LexUnexpandedToken(Tmp);
92     assert(Tmp.isNot(tok::eof) && "EOF seen while discarding directive tokens");
93   } while (Tmp.isNot(tok::eod));
94 }
95 
96 /// \brief Lex and validate a macro name, which occurs after a
97 /// \#define or \#undef.
98 ///
99 /// This sets the token kind to eod and discards the rest
100 /// of the macro line if the macro name is invalid.  \p isDefineUndef is 1 if
101 /// this is due to a a \#define, 2 if \#undef directive, 0 if it is something
102 /// else (e.g. \#ifdef).
ReadMacroName(Token & MacroNameTok,char isDefineUndef)103 void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) {
104   // Read the token, don't allow macro expansion on it.
105   LexUnexpandedToken(MacroNameTok);
106 
107   if (MacroNameTok.is(tok::code_completion)) {
108     if (CodeComplete)
109       CodeComplete->CodeCompleteMacroName(isDefineUndef == 1);
110     setCodeCompletionReached();
111     LexUnexpandedToken(MacroNameTok);
112   }
113 
114   // Missing macro name?
115   if (MacroNameTok.is(tok::eod)) {
116     Diag(MacroNameTok, diag::err_pp_missing_macro_name);
117     return;
118   }
119 
120   IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
121   if (II == 0) {
122     bool Invalid = false;
123     std::string Spelling = getSpelling(MacroNameTok, &Invalid);
124     if (Invalid)
125       return;
126 
127     const IdentifierInfo &Info = Identifiers.get(Spelling);
128 
129     // Allow #defining |and| and friends in microsoft mode.
130     if (Info.isCPlusPlusOperatorKeyword() && getLangOpts().MicrosoftMode) {
131       MacroNameTok.setIdentifierInfo(getIdentifierInfo(Spelling));
132       return;
133     }
134 
135     if (Info.isCPlusPlusOperatorKeyword())
136       // C++ 2.5p2: Alternative tokens behave the same as its primary token
137       // except for their spellings.
138       Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling;
139     else
140       Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
141     // Fall through on error.
142   } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) {
143     // Error if defining "defined": C99 6.10.8.4.
144     Diag(MacroNameTok, diag::err_defined_macro_name);
145   } else if (isDefineUndef && II->hasMacroDefinition() &&
146              getMacroInfo(II)->isBuiltinMacro()) {
147     // Error if defining "__LINE__" and other builtins: C99 6.10.8.4.
148     if (isDefineUndef == 1)
149       Diag(MacroNameTok, diag::pp_redef_builtin_macro);
150     else
151       Diag(MacroNameTok, diag::pp_undef_builtin_macro);
152   } else {
153     // Okay, we got a good identifier node.  Return it.
154     return;
155   }
156 
157   // Invalid macro name, read and discard the rest of the line.  Then set the
158   // token kind to tok::eod.
159   MacroNameTok.setKind(tok::eod);
160   return DiscardUntilEndOfDirective();
161 }
162 
163 /// \brief Ensure that the next token is a tok::eod token.
164 ///
165 /// If not, emit a diagnostic and consume up until the eod.  If EnableMacros is
166 /// true, then we consider macros that expand to zero tokens as being ok.
CheckEndOfDirective(const char * DirType,bool EnableMacros)167 void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) {
168   Token Tmp;
169   // Lex unexpanded tokens for most directives: macros might expand to zero
170   // tokens, causing us to miss diagnosing invalid lines.  Some directives (like
171   // #line) allow empty macros.
172   if (EnableMacros)
173     Lex(Tmp);
174   else
175     LexUnexpandedToken(Tmp);
176 
177   // There should be no tokens after the directive, but we allow them as an
178   // extension.
179   while (Tmp.is(tok::comment))  // Skip comments in -C mode.
180     LexUnexpandedToken(Tmp);
181 
182   if (Tmp.isNot(tok::eod)) {
183     // Add a fixit in GNU/C99/C++ mode.  Don't offer a fixit for strict-C89,
184     // or if this is a macro-style preprocessing directive, because it is more
185     // trouble than it is worth to insert /**/ and check that there is no /**/
186     // in the range also.
187     FixItHint Hint;
188     if ((LangOpts.GNUMode || LangOpts.C99 || LangOpts.CPlusPlus) &&
189         !CurTokenLexer)
190       Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//");
191     Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint;
192     DiscardUntilEndOfDirective();
193   }
194 }
195 
196 
197 
198 /// SkipExcludedConditionalBlock - We just read a \#if or related directive and
199 /// decided that the subsequent tokens are in the \#if'd out portion of the
200 /// file.  Lex the rest of the file, until we see an \#endif.  If
201 /// FoundNonSkipPortion is true, then we have already emitted code for part of
202 /// this \#if directive, so \#else/\#elif blocks should never be entered.
203 /// If ElseOk is true, then \#else directives are ok, if not, then we have
204 /// already seen one so a \#else directive is a duplicate.  When this returns,
205 /// the caller can lex the first valid token.
SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,bool FoundNonSkipPortion,bool FoundElse,SourceLocation ElseLoc)206 void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
207                                                 bool FoundNonSkipPortion,
208                                                 bool FoundElse,
209                                                 SourceLocation ElseLoc) {
210   ++NumSkipped;
211   assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?");
212 
213   CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false,
214                                  FoundNonSkipPortion, FoundElse);
215 
216   if (CurPTHLexer) {
217     PTHSkipExcludedConditionalBlock();
218     return;
219   }
220 
221   // Enter raw mode to disable identifier lookup (and thus macro expansion),
222   // disabling warnings, etc.
223   CurPPLexer->LexingRawMode = true;
224   Token Tok;
225   while (1) {
226     CurLexer->Lex(Tok);
227 
228     if (Tok.is(tok::code_completion)) {
229       if (CodeComplete)
230         CodeComplete->CodeCompleteInConditionalExclusion();
231       setCodeCompletionReached();
232       continue;
233     }
234 
235     // If this is the end of the buffer, we have an error.
236     if (Tok.is(tok::eof)) {
237       // Emit errors for each unterminated conditional on the stack, including
238       // the current one.
239       while (!CurPPLexer->ConditionalStack.empty()) {
240         if (CurLexer->getFileLoc() != CodeCompletionFileLoc)
241           Diag(CurPPLexer->ConditionalStack.back().IfLoc,
242                diag::err_pp_unterminated_conditional);
243         CurPPLexer->ConditionalStack.pop_back();
244       }
245 
246       // Just return and let the caller lex after this #include.
247       break;
248     }
249 
250     // If this token is not a preprocessor directive, just skip it.
251     if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine())
252       continue;
253 
254     // We just parsed a # character at the start of a line, so we're in
255     // directive mode.  Tell the lexer this so any newlines we see will be
256     // converted into an EOD token (this terminates the macro).
257     CurPPLexer->ParsingPreprocessorDirective = true;
258     if (CurLexer) CurLexer->SetCommentRetentionState(false);
259 
260 
261     // Read the next token, the directive flavor.
262     LexUnexpandedToken(Tok);
263 
264     // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
265     // something bogus), skip it.
266     if (Tok.isNot(tok::raw_identifier)) {
267       CurPPLexer->ParsingPreprocessorDirective = false;
268       // Restore comment saving mode.
269       if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
270       continue;
271     }
272 
273     // If the first letter isn't i or e, it isn't intesting to us.  We know that
274     // this is safe in the face of spelling differences, because there is no way
275     // to spell an i/e in a strange way that is another letter.  Skipping this
276     // allows us to avoid looking up the identifier info for #define/#undef and
277     // other common directives.
278     const char *RawCharData = Tok.getRawIdentifierData();
279 
280     char FirstChar = RawCharData[0];
281     if (FirstChar >= 'a' && FirstChar <= 'z' &&
282         FirstChar != 'i' && FirstChar != 'e') {
283       CurPPLexer->ParsingPreprocessorDirective = false;
284       // Restore comment saving mode.
285       if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
286       continue;
287     }
288 
289     // Get the identifier name without trigraphs or embedded newlines.  Note
290     // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
291     // when skipping.
292     char DirectiveBuf[20];
293     StringRef Directive;
294     if (!Tok.needsCleaning() && Tok.getLength() < 20) {
295       Directive = StringRef(RawCharData, Tok.getLength());
296     } else {
297       std::string DirectiveStr = getSpelling(Tok);
298       unsigned IdLen = DirectiveStr.size();
299       if (IdLen >= 20) {
300         CurPPLexer->ParsingPreprocessorDirective = false;
301         // Restore comment saving mode.
302         if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
303         continue;
304       }
305       memcpy(DirectiveBuf, &DirectiveStr[0], IdLen);
306       Directive = StringRef(DirectiveBuf, IdLen);
307     }
308 
309     if (Directive.startswith("if")) {
310       StringRef Sub = Directive.substr(2);
311       if (Sub.empty() ||   // "if"
312           Sub == "def" ||   // "ifdef"
313           Sub == "ndef") {  // "ifndef"
314         // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
315         // bother parsing the condition.
316         DiscardUntilEndOfDirective();
317         CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true,
318                                        /*foundnonskip*/false,
319                                        /*foundelse*/false);
320       }
321     } else if (Directive[0] == 'e') {
322       StringRef Sub = Directive.substr(1);
323       if (Sub == "ndif") {  // "endif"
324         PPConditionalInfo CondInfo;
325         CondInfo.WasSkipping = true; // Silence bogus warning.
326         bool InCond = CurPPLexer->popConditionalLevel(CondInfo);
327         (void)InCond;  // Silence warning in no-asserts mode.
328         assert(!InCond && "Can't be skipping if not in a conditional!");
329 
330         // If we popped the outermost skipping block, we're done skipping!
331         if (!CondInfo.WasSkipping) {
332           // Restore the value of LexingRawMode so that trailing comments
333           // are handled correctly, if we've reached the outermost block.
334           CurPPLexer->LexingRawMode = false;
335           CheckEndOfDirective("endif");
336           CurPPLexer->LexingRawMode = true;
337           if (Callbacks)
338             Callbacks->Endif(Tok.getLocation(), CondInfo.IfLoc);
339           break;
340         } else {
341           DiscardUntilEndOfDirective();
342         }
343       } else if (Sub == "lse") { // "else".
344         // #else directive in a skipping conditional.  If not in some other
345         // skipping conditional, and if #else hasn't already been seen, enter it
346         // as a non-skipping conditional.
347         PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
348 
349         // If this is a #else with a #else before it, report the error.
350         if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else);
351 
352         // Note that we've seen a #else in this conditional.
353         CondInfo.FoundElse = true;
354 
355         // If the conditional is at the top level, and the #if block wasn't
356         // entered, enter the #else block now.
357         if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
358           CondInfo.FoundNonSkip = true;
359           // Restore the value of LexingRawMode so that trailing comments
360           // are handled correctly.
361           CurPPLexer->LexingRawMode = false;
362           CheckEndOfDirective("else");
363           CurPPLexer->LexingRawMode = true;
364           if (Callbacks)
365             Callbacks->Else(Tok.getLocation(), CondInfo.IfLoc);
366           break;
367         } else {
368           DiscardUntilEndOfDirective();  // C99 6.10p4.
369         }
370       } else if (Sub == "lif") {  // "elif".
371         PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
372 
373         bool ShouldEnter;
374         const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
375         // If this is in a skipping block or if we're already handled this #if
376         // block, don't bother parsing the condition.
377         if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
378           DiscardUntilEndOfDirective();
379           ShouldEnter = false;
380         } else {
381           // Restore the value of LexingRawMode so that identifiers are
382           // looked up, etc, inside the #elif expression.
383           assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
384           CurPPLexer->LexingRawMode = false;
385           IdentifierInfo *IfNDefMacro = 0;
386           ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
387           CurPPLexer->LexingRawMode = true;
388         }
389         const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
390 
391         // If this is a #elif with a #else before it, report the error.
392         if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else);
393 
394         // If this condition is true, enter it!
395         if (ShouldEnter) {
396           CondInfo.FoundNonSkip = true;
397           if (Callbacks)
398             Callbacks->Elif(Tok.getLocation(),
399                             SourceRange(ConditionalBegin, ConditionalEnd),
400                             CondInfo.IfLoc);
401           break;
402         }
403       }
404     }
405 
406     CurPPLexer->ParsingPreprocessorDirective = false;
407     // Restore comment saving mode.
408     if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
409   }
410 
411   // Finally, if we are out of the conditional (saw an #endif or ran off the end
412   // of the file, just stop skipping and return to lexing whatever came after
413   // the #if block.
414   CurPPLexer->LexingRawMode = false;
415 
416   if (Callbacks) {
417     SourceLocation BeginLoc = ElseLoc.isValid() ? ElseLoc : IfTokenLoc;
418     Callbacks->SourceRangeSkipped(SourceRange(BeginLoc, Tok.getLocation()));
419   }
420 }
421 
PTHSkipExcludedConditionalBlock()422 void Preprocessor::PTHSkipExcludedConditionalBlock() {
423 
424   while (1) {
425     assert(CurPTHLexer);
426     assert(CurPTHLexer->LexingRawMode == false);
427 
428     // Skip to the next '#else', '#elif', or #endif.
429     if (CurPTHLexer->SkipBlock()) {
430       // We have reached an #endif.  Both the '#' and 'endif' tokens
431       // have been consumed by the PTHLexer.  Just pop off the condition level.
432       PPConditionalInfo CondInfo;
433       bool InCond = CurPTHLexer->popConditionalLevel(CondInfo);
434       (void)InCond;  // Silence warning in no-asserts mode.
435       assert(!InCond && "Can't be skipping if not in a conditional!");
436       break;
437     }
438 
439     // We have reached a '#else' or '#elif'.  Lex the next token to get
440     // the directive flavor.
441     Token Tok;
442     LexUnexpandedToken(Tok);
443 
444     // We can actually look up the IdentifierInfo here since we aren't in
445     // raw mode.
446     tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID();
447 
448     if (K == tok::pp_else) {
449       // #else: Enter the else condition.  We aren't in a nested condition
450       //  since we skip those. We're always in the one matching the last
451       //  blocked we skipped.
452       PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
453       // Note that we've seen a #else in this conditional.
454       CondInfo.FoundElse = true;
455 
456       // If the #if block wasn't entered then enter the #else block now.
457       if (!CondInfo.FoundNonSkip) {
458         CondInfo.FoundNonSkip = true;
459 
460         // Scan until the eod token.
461         CurPTHLexer->ParsingPreprocessorDirective = true;
462         DiscardUntilEndOfDirective();
463         CurPTHLexer->ParsingPreprocessorDirective = false;
464 
465         break;
466       }
467 
468       // Otherwise skip this block.
469       continue;
470     }
471 
472     assert(K == tok::pp_elif);
473     PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
474 
475     // If this is a #elif with a #else before it, report the error.
476     if (CondInfo.FoundElse)
477       Diag(Tok, diag::pp_err_elif_after_else);
478 
479     // If this is in a skipping block or if we're already handled this #if
480     // block, don't bother parsing the condition.  We just skip this block.
481     if (CondInfo.FoundNonSkip)
482       continue;
483 
484     // Evaluate the condition of the #elif.
485     IdentifierInfo *IfNDefMacro = 0;
486     CurPTHLexer->ParsingPreprocessorDirective = true;
487     bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
488     CurPTHLexer->ParsingPreprocessorDirective = false;
489 
490     // If this condition is true, enter it!
491     if (ShouldEnter) {
492       CondInfo.FoundNonSkip = true;
493       break;
494     }
495 
496     // Otherwise, skip this block and go to the next one.
497     continue;
498   }
499 }
500 
LookupFile(StringRef Filename,bool isAngled,const DirectoryLookup * FromDir,const DirectoryLookup * & CurDir,SmallVectorImpl<char> * SearchPath,SmallVectorImpl<char> * RelativePath,Module ** SuggestedModule,bool SkipCache)501 const FileEntry *Preprocessor::LookupFile(
502     StringRef Filename,
503     bool isAngled,
504     const DirectoryLookup *FromDir,
505     const DirectoryLookup *&CurDir,
506     SmallVectorImpl<char> *SearchPath,
507     SmallVectorImpl<char> *RelativePath,
508     Module **SuggestedModule,
509     bool SkipCache) {
510   // If the header lookup mechanism may be relative to the current file, pass in
511   // info about where the current file is.
512   const FileEntry *CurFileEnt = 0;
513   if (!FromDir) {
514     FileID FID = getCurrentFileLexer()->getFileID();
515     CurFileEnt = SourceMgr.getFileEntryForID(FID);
516 
517     // If there is no file entry associated with this file, it must be the
518     // predefines buffer.  Any other file is not lexed with a normal lexer, so
519     // it won't be scanned for preprocessor directives.   If we have the
520     // predefines buffer, resolve #include references (which come from the
521     // -include command line argument) as if they came from the main file, this
522     // affects file lookup etc.
523     if (CurFileEnt == 0) {
524       FID = SourceMgr.getMainFileID();
525       CurFileEnt = SourceMgr.getFileEntryForID(FID);
526     }
527   }
528 
529   // Do a standard file entry lookup.
530   CurDir = CurDirLookup;
531   const FileEntry *FE = HeaderInfo.LookupFile(
532       Filename, isAngled, FromDir, CurDir, CurFileEnt,
533       SearchPath, RelativePath, SuggestedModule, SkipCache);
534   if (FE) return FE;
535 
536   // Otherwise, see if this is a subframework header.  If so, this is relative
537   // to one of the headers on the #include stack.  Walk the list of the current
538   // headers on the #include stack and pass them to HeaderInfo.
539   // FIXME: SuggestedModule!
540   if (IsFileLexer()) {
541     if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID())))
542       if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt,
543                                                     SearchPath, RelativePath)))
544         return FE;
545   }
546 
547   for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
548     IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1];
549     if (IsFileLexer(ISEntry)) {
550       if ((CurFileEnt =
551            SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID())))
552         if ((FE = HeaderInfo.LookupSubframeworkHeader(
553                 Filename, CurFileEnt, SearchPath, RelativePath)))
554           return FE;
555     }
556   }
557 
558   // Otherwise, we really couldn't find the file.
559   return 0;
560 }
561 
562 
563 //===----------------------------------------------------------------------===//
564 // Preprocessor Directive Handling.
565 //===----------------------------------------------------------------------===//
566 
567 class Preprocessor::ResetMacroExpansionHelper {
568 public:
ResetMacroExpansionHelper(Preprocessor * pp)569   ResetMacroExpansionHelper(Preprocessor *pp)
570     : PP(pp), save(pp->DisableMacroExpansion) {
571     if (pp->MacroExpansionInDirectivesOverride)
572       pp->DisableMacroExpansion = false;
573   }
~ResetMacroExpansionHelper()574   ~ResetMacroExpansionHelper() {
575     PP->DisableMacroExpansion = save;
576   }
577 private:
578   Preprocessor *PP;
579   bool save;
580 };
581 
582 /// HandleDirective - This callback is invoked when the lexer sees a # token
583 /// at the start of a line.  This consumes the directive, modifies the
584 /// lexer/preprocessor state, and advances the lexer(s) so that the next token
585 /// read is the correct one.
HandleDirective(Token & Result)586 void Preprocessor::HandleDirective(Token &Result) {
587   // FIXME: Traditional: # with whitespace before it not recognized by K&R?
588 
589   // We just parsed a # character at the start of a line, so we're in directive
590   // mode.  Tell the lexer this so any newlines we see will be converted into an
591   // EOD token (which terminates the directive).
592   CurPPLexer->ParsingPreprocessorDirective = true;
593 
594   ++NumDirectives;
595 
596   // We are about to read a token.  For the multiple-include optimization FA to
597   // work, we have to remember if we had read any tokens *before* this
598   // pp-directive.
599   bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal();
600 
601   // Save the '#' token in case we need to return it later.
602   Token SavedHash = Result;
603 
604   // Read the next token, the directive flavor.  This isn't expanded due to
605   // C99 6.10.3p8.
606   LexUnexpandedToken(Result);
607 
608   // C99 6.10.3p11: Is this preprocessor directive in macro invocation?  e.g.:
609   //   #define A(x) #x
610   //   A(abc
611   //     #warning blah
612   //   def)
613   // If so, the user is relying on undefined behavior, emit a diagnostic. Do
614   // not support this for #include-like directives, since that can result in
615   // terrible diagnostics, and does not work in GCC.
616   if (InMacroArgs) {
617     if (IdentifierInfo *II = Result.getIdentifierInfo()) {
618       switch (II->getPPKeywordID()) {
619       case tok::pp_include:
620       case tok::pp_import:
621       case tok::pp_include_next:
622       case tok::pp___include_macros:
623         Diag(Result, diag::err_embedded_include) << II->getName();
624         DiscardUntilEndOfDirective();
625         return;
626       default:
627         break;
628       }
629     }
630     Diag(Result, diag::ext_embedded_directive);
631   }
632 
633   // Temporarily enable macro expansion if set so
634   // and reset to previous state when returning from this function.
635   ResetMacroExpansionHelper helper(this);
636 
637 TryAgain:
638   switch (Result.getKind()) {
639   case tok::eod:
640     return;   // null directive.
641   case tok::comment:
642     // Handle stuff like "# /*foo*/ define X" in -E -C mode.
643     LexUnexpandedToken(Result);
644     goto TryAgain;
645   case tok::code_completion:
646     if (CodeComplete)
647       CodeComplete->CodeCompleteDirective(
648                                     CurPPLexer->getConditionalStackDepth() > 0);
649     setCodeCompletionReached();
650     return;
651   case tok::numeric_constant:  // # 7  GNU line marker directive.
652     if (getLangOpts().AsmPreprocessor)
653       break;  // # 4 is not a preprocessor directive in .S files.
654     return HandleDigitDirective(Result);
655   default:
656     IdentifierInfo *II = Result.getIdentifierInfo();
657     if (II == 0) break;  // Not an identifier.
658 
659     // Ask what the preprocessor keyword ID is.
660     switch (II->getPPKeywordID()) {
661     default: break;
662     // C99 6.10.1 - Conditional Inclusion.
663     case tok::pp_if:
664       return HandleIfDirective(Result, ReadAnyTokensBeforeDirective);
665     case tok::pp_ifdef:
666       return HandleIfdefDirective(Result, false, true/*not valid for miopt*/);
667     case tok::pp_ifndef:
668       return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective);
669     case tok::pp_elif:
670       return HandleElifDirective(Result);
671     case tok::pp_else:
672       return HandleElseDirective(Result);
673     case tok::pp_endif:
674       return HandleEndifDirective(Result);
675 
676     // C99 6.10.2 - Source File Inclusion.
677     case tok::pp_include:
678       // Handle #include.
679       return HandleIncludeDirective(SavedHash.getLocation(), Result);
680     case tok::pp___include_macros:
681       // Handle -imacros.
682       return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result);
683 
684     // C99 6.10.3 - Macro Replacement.
685     case tok::pp_define:
686       return HandleDefineDirective(Result);
687     case tok::pp_undef:
688       return HandleUndefDirective(Result);
689 
690     // C99 6.10.4 - Line Control.
691     case tok::pp_line:
692       return HandleLineDirective(Result);
693 
694     // C99 6.10.5 - Error Directive.
695     case tok::pp_error:
696       return HandleUserDiagnosticDirective(Result, false);
697 
698     // C99 6.10.6 - Pragma Directive.
699     case tok::pp_pragma:
700       return HandlePragmaDirective(PIK_HashPragma);
701 
702     // GNU Extensions.
703     case tok::pp_import:
704       return HandleImportDirective(SavedHash.getLocation(), Result);
705     case tok::pp_include_next:
706       return HandleIncludeNextDirective(SavedHash.getLocation(), Result);
707 
708     case tok::pp_warning:
709       Diag(Result, diag::ext_pp_warning_directive);
710       return HandleUserDiagnosticDirective(Result, true);
711     case tok::pp_ident:
712       return HandleIdentSCCSDirective(Result);
713     case tok::pp_sccs:
714       return HandleIdentSCCSDirective(Result);
715     case tok::pp_assert:
716       //isExtension = true;  // FIXME: implement #assert
717       break;
718     case tok::pp_unassert:
719       //isExtension = true;  // FIXME: implement #unassert
720       break;
721 
722     case tok::pp___public_macro:
723       if (getLangOpts().Modules)
724         return HandleMacroPublicDirective(Result);
725       break;
726 
727     case tok::pp___private_macro:
728       if (getLangOpts().Modules)
729         return HandleMacroPrivateDirective(Result);
730       break;
731     }
732     break;
733   }
734 
735   // If this is a .S file, treat unknown # directives as non-preprocessor
736   // directives.  This is important because # may be a comment or introduce
737   // various pseudo-ops.  Just return the # token and push back the following
738   // token to be lexed next time.
739   if (getLangOpts().AsmPreprocessor) {
740     Token *Toks = new Token[2];
741     // Return the # and the token after it.
742     Toks[0] = SavedHash;
743     Toks[1] = Result;
744 
745     // If the second token is a hashhash token, then we need to translate it to
746     // unknown so the token lexer doesn't try to perform token pasting.
747     if (Result.is(tok::hashhash))
748       Toks[1].setKind(tok::unknown);
749 
750     // Enter this token stream so that we re-lex the tokens.  Make sure to
751     // enable macro expansion, in case the token after the # is an identifier
752     // that is expanded.
753     EnterTokenStream(Toks, 2, false, true);
754     return;
755   }
756 
757   // If we reached here, the preprocessing token is not valid!
758   Diag(Result, diag::err_pp_invalid_directive);
759 
760   // Read the rest of the PP line.
761   DiscardUntilEndOfDirective();
762 
763   // Okay, we're done parsing the directive.
764 }
765 
766 /// GetLineValue - Convert a numeric token into an unsigned value, emitting
767 /// Diagnostic DiagID if it is invalid, and returning the value in Val.
GetLineValue(Token & DigitTok,unsigned & Val,unsigned DiagID,Preprocessor & PP)768 static bool GetLineValue(Token &DigitTok, unsigned &Val,
769                          unsigned DiagID, Preprocessor &PP) {
770   if (DigitTok.isNot(tok::numeric_constant)) {
771     PP.Diag(DigitTok, DiagID);
772 
773     if (DigitTok.isNot(tok::eod))
774       PP.DiscardUntilEndOfDirective();
775     return true;
776   }
777 
778   SmallString<64> IntegerBuffer;
779   IntegerBuffer.resize(DigitTok.getLength());
780   const char *DigitTokBegin = &IntegerBuffer[0];
781   bool Invalid = false;
782   unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid);
783   if (Invalid)
784     return true;
785 
786   // Verify that we have a simple digit-sequence, and compute the value.  This
787   // is always a simple digit string computed in decimal, so we do this manually
788   // here.
789   Val = 0;
790   for (unsigned i = 0; i != ActualLength; ++i) {
791     if (!isdigit(DigitTokBegin[i])) {
792       PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i),
793               diag::err_pp_line_digit_sequence);
794       PP.DiscardUntilEndOfDirective();
795       return true;
796     }
797 
798     unsigned NextVal = Val*10+(DigitTokBegin[i]-'0');
799     if (NextVal < Val) { // overflow.
800       PP.Diag(DigitTok, DiagID);
801       PP.DiscardUntilEndOfDirective();
802       return true;
803     }
804     Val = NextVal;
805   }
806 
807   if (DigitTokBegin[0] == '0' && Val)
808     PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal);
809 
810   return false;
811 }
812 
813 /// \brief Handle a \#line directive: C99 6.10.4.
814 ///
815 /// The two acceptable forms are:
816 /// \verbatim
817 ///   # line digit-sequence
818 ///   # line digit-sequence "s-char-sequence"
819 /// \endverbatim
HandleLineDirective(Token & Tok)820 void Preprocessor::HandleLineDirective(Token &Tok) {
821   // Read the line # and string argument.  Per C99 6.10.4p5, these tokens are
822   // expanded.
823   Token DigitTok;
824   Lex(DigitTok);
825 
826   // Validate the number and convert it to an unsigned.
827   unsigned LineNo;
828   if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this))
829     return;
830 
831   if (LineNo == 0)
832     Diag(DigitTok, diag::ext_pp_line_zero);
833 
834   // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
835   // number greater than 2147483647".  C90 requires that the line # be <= 32767.
836   unsigned LineLimit = 32768U;
837   if (LangOpts.C99 || LangOpts.CPlusPlus0x)
838     LineLimit = 2147483648U;
839   if (LineNo >= LineLimit)
840     Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit;
841   else if (LangOpts.CPlusPlus0x && LineNo >= 32768U)
842     Diag(DigitTok, diag::warn_cxx98_compat_pp_line_too_big);
843 
844   int FilenameID = -1;
845   Token StrTok;
846   Lex(StrTok);
847 
848   // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
849   // string followed by eod.
850   if (StrTok.is(tok::eod))
851     ; // ok
852   else if (StrTok.isNot(tok::string_literal)) {
853     Diag(StrTok, diag::err_pp_line_invalid_filename);
854     return DiscardUntilEndOfDirective();
855   } else if (StrTok.hasUDSuffix()) {
856     Diag(StrTok, diag::err_invalid_string_udl);
857     return DiscardUntilEndOfDirective();
858   } else {
859     // Parse and validate the string, converting it into a unique ID.
860     StringLiteralParser Literal(&StrTok, 1, *this);
861     assert(Literal.isAscii() && "Didn't allow wide strings in");
862     if (Literal.hadError)
863       return DiscardUntilEndOfDirective();
864     if (Literal.Pascal) {
865       Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
866       return DiscardUntilEndOfDirective();
867     }
868     FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
869 
870     // Verify that there is nothing after the string, other than EOD.  Because
871     // of C99 6.10.4p5, macros that expand to empty tokens are ok.
872     CheckEndOfDirective("line", true);
873   }
874 
875   SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID);
876 
877   if (Callbacks)
878     Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
879                            PPCallbacks::RenameFile,
880                            SrcMgr::C_User);
881 }
882 
883 /// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
884 /// marker directive.
ReadLineMarkerFlags(bool & IsFileEntry,bool & IsFileExit,bool & IsSystemHeader,bool & IsExternCHeader,Preprocessor & PP)885 static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
886                                 bool &IsSystemHeader, bool &IsExternCHeader,
887                                 Preprocessor &PP) {
888   unsigned FlagVal;
889   Token FlagTok;
890   PP.Lex(FlagTok);
891   if (FlagTok.is(tok::eod)) return false;
892   if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
893     return true;
894 
895   if (FlagVal == 1) {
896     IsFileEntry = true;
897 
898     PP.Lex(FlagTok);
899     if (FlagTok.is(tok::eod)) return false;
900     if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
901       return true;
902   } else if (FlagVal == 2) {
903     IsFileExit = true;
904 
905     SourceManager &SM = PP.getSourceManager();
906     // If we are leaving the current presumed file, check to make sure the
907     // presumed include stack isn't empty!
908     FileID CurFileID =
909       SM.getDecomposedExpansionLoc(FlagTok.getLocation()).first;
910     PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation());
911     if (PLoc.isInvalid())
912       return true;
913 
914     // If there is no include loc (main file) or if the include loc is in a
915     // different physical file, then we aren't in a "1" line marker flag region.
916     SourceLocation IncLoc = PLoc.getIncludeLoc();
917     if (IncLoc.isInvalid() ||
918         SM.getDecomposedExpansionLoc(IncLoc).first != CurFileID) {
919       PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop);
920       PP.DiscardUntilEndOfDirective();
921       return true;
922     }
923 
924     PP.Lex(FlagTok);
925     if (FlagTok.is(tok::eod)) return false;
926     if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
927       return true;
928   }
929 
930   // We must have 3 if there are still flags.
931   if (FlagVal != 3) {
932     PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
933     PP.DiscardUntilEndOfDirective();
934     return true;
935   }
936 
937   IsSystemHeader = true;
938 
939   PP.Lex(FlagTok);
940   if (FlagTok.is(tok::eod)) return false;
941   if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
942     return true;
943 
944   // We must have 4 if there is yet another flag.
945   if (FlagVal != 4) {
946     PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
947     PP.DiscardUntilEndOfDirective();
948     return true;
949   }
950 
951   IsExternCHeader = true;
952 
953   PP.Lex(FlagTok);
954   if (FlagTok.is(tok::eod)) return false;
955 
956   // There are no more valid flags here.
957   PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
958   PP.DiscardUntilEndOfDirective();
959   return true;
960 }
961 
962 /// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
963 /// one of the following forms:
964 ///
965 ///     # 42
966 ///     # 42 "file" ('1' | '2')?
967 ///     # 42 "file" ('1' | '2')? '3' '4'?
968 ///
HandleDigitDirective(Token & DigitTok)969 void Preprocessor::HandleDigitDirective(Token &DigitTok) {
970   // Validate the number and convert it to an unsigned.  GNU does not have a
971   // line # limit other than it fit in 32-bits.
972   unsigned LineNo;
973   if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer,
974                    *this))
975     return;
976 
977   Token StrTok;
978   Lex(StrTok);
979 
980   bool IsFileEntry = false, IsFileExit = false;
981   bool IsSystemHeader = false, IsExternCHeader = false;
982   int FilenameID = -1;
983 
984   // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
985   // string followed by eod.
986   if (StrTok.is(tok::eod))
987     ; // ok
988   else if (StrTok.isNot(tok::string_literal)) {
989     Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
990     return DiscardUntilEndOfDirective();
991   } else if (StrTok.hasUDSuffix()) {
992     Diag(StrTok, diag::err_invalid_string_udl);
993     return DiscardUntilEndOfDirective();
994   } else {
995     // Parse and validate the string, converting it into a unique ID.
996     StringLiteralParser Literal(&StrTok, 1, *this);
997     assert(Literal.isAscii() && "Didn't allow wide strings in");
998     if (Literal.hadError)
999       return DiscardUntilEndOfDirective();
1000     if (Literal.Pascal) {
1001       Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1002       return DiscardUntilEndOfDirective();
1003     }
1004     FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1005 
1006     // If a filename was present, read any flags that are present.
1007     if (ReadLineMarkerFlags(IsFileEntry, IsFileExit,
1008                             IsSystemHeader, IsExternCHeader, *this))
1009       return;
1010   }
1011 
1012   // Create a line note with this information.
1013   SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID,
1014                         IsFileEntry, IsFileExit,
1015                         IsSystemHeader, IsExternCHeader);
1016 
1017   // If the preprocessor has callbacks installed, notify them of the #line
1018   // change.  This is used so that the line marker comes out in -E mode for
1019   // example.
1020   if (Callbacks) {
1021     PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile;
1022     if (IsFileEntry)
1023       Reason = PPCallbacks::EnterFile;
1024     else if (IsFileExit)
1025       Reason = PPCallbacks::ExitFile;
1026     SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User;
1027     if (IsExternCHeader)
1028       FileKind = SrcMgr::C_ExternCSystem;
1029     else if (IsSystemHeader)
1030       FileKind = SrcMgr::C_System;
1031 
1032     Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind);
1033   }
1034 }
1035 
1036 
1037 /// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
1038 ///
HandleUserDiagnosticDirective(Token & Tok,bool isWarning)1039 void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
1040                                                  bool isWarning) {
1041   // PTH doesn't emit #warning or #error directives.
1042   if (CurPTHLexer)
1043     return CurPTHLexer->DiscardToEndOfLine();
1044 
1045   // Read the rest of the line raw.  We do this because we don't want macros
1046   // to be expanded and we don't require that the tokens be valid preprocessing
1047   // tokens.  For example, this is allowed: "#warning `   'foo".  GCC does
1048   // collapse multiple consequtive white space between tokens, but this isn't
1049   // specified by the standard.
1050   SmallString<128> Message;
1051   CurLexer->ReadToEndOfLine(&Message);
1052 
1053   // Find the first non-whitespace character, so that we can make the
1054   // diagnostic more succinct.
1055   StringRef Msg = Message.str().ltrim(" ");
1056 
1057   if (isWarning)
1058     Diag(Tok, diag::pp_hash_warning) << Msg;
1059   else
1060     Diag(Tok, diag::err_pp_hash_error) << Msg;
1061 }
1062 
1063 /// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
1064 ///
HandleIdentSCCSDirective(Token & Tok)1065 void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
1066   // Yes, this directive is an extension.
1067   Diag(Tok, diag::ext_pp_ident_directive);
1068 
1069   // Read the string argument.
1070   Token StrTok;
1071   Lex(StrTok);
1072 
1073   // If the token kind isn't a string, it's a malformed directive.
1074   if (StrTok.isNot(tok::string_literal) &&
1075       StrTok.isNot(tok::wide_string_literal)) {
1076     Diag(StrTok, diag::err_pp_malformed_ident);
1077     if (StrTok.isNot(tok::eod))
1078       DiscardUntilEndOfDirective();
1079     return;
1080   }
1081 
1082   if (StrTok.hasUDSuffix()) {
1083     Diag(StrTok, diag::err_invalid_string_udl);
1084     return DiscardUntilEndOfDirective();
1085   }
1086 
1087   // Verify that there is nothing after the string, other than EOD.
1088   CheckEndOfDirective("ident");
1089 
1090   if (Callbacks) {
1091     bool Invalid = false;
1092     std::string Str = getSpelling(StrTok, &Invalid);
1093     if (!Invalid)
1094       Callbacks->Ident(Tok.getLocation(), Str);
1095   }
1096 }
1097 
1098 /// \brief Handle a #public directive.
HandleMacroPublicDirective(Token & Tok)1099 void Preprocessor::HandleMacroPublicDirective(Token &Tok) {
1100   Token MacroNameTok;
1101   ReadMacroName(MacroNameTok, 2);
1102 
1103   // Error reading macro name?  If so, diagnostic already issued.
1104   if (MacroNameTok.is(tok::eod))
1105     return;
1106 
1107   // Check to see if this is the last token on the #__public_macro line.
1108   CheckEndOfDirective("__public_macro");
1109 
1110   // Okay, we finally have a valid identifier to undef.
1111   MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo());
1112 
1113   // If the macro is not defined, this is an error.
1114   if (MI == 0) {
1115     Diag(MacroNameTok, diag::err_pp_visibility_non_macro)
1116       << MacroNameTok.getIdentifierInfo();
1117     return;
1118   }
1119 
1120   // Note that this macro has now been exported.
1121   MI->setVisibility(/*IsPublic=*/true, MacroNameTok.getLocation());
1122 
1123   // If this macro definition came from a PCH file, mark it
1124   // as having changed since serialization.
1125   if (MI->isFromAST())
1126     MI->setChangedAfterLoad();
1127 }
1128 
1129 /// \brief Handle a #private directive.
HandleMacroPrivateDirective(Token & Tok)1130 void Preprocessor::HandleMacroPrivateDirective(Token &Tok) {
1131   Token MacroNameTok;
1132   ReadMacroName(MacroNameTok, 2);
1133 
1134   // Error reading macro name?  If so, diagnostic already issued.
1135   if (MacroNameTok.is(tok::eod))
1136     return;
1137 
1138   // Check to see if this is the last token on the #__private_macro line.
1139   CheckEndOfDirective("__private_macro");
1140 
1141   // Okay, we finally have a valid identifier to undef.
1142   MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo());
1143 
1144   // If the macro is not defined, this is an error.
1145   if (MI == 0) {
1146     Diag(MacroNameTok, diag::err_pp_visibility_non_macro)
1147       << MacroNameTok.getIdentifierInfo();
1148     return;
1149   }
1150 
1151   // Note that this macro has now been marked private.
1152   MI->setVisibility(/*IsPublic=*/false, MacroNameTok.getLocation());
1153 
1154   // If this macro definition came from a PCH file, mark it
1155   // as having changed since serialization.
1156   if (MI->isFromAST())
1157     MI->setChangedAfterLoad();
1158 }
1159 
1160 //===----------------------------------------------------------------------===//
1161 // Preprocessor Include Directive Handling.
1162 //===----------------------------------------------------------------------===//
1163 
1164 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
1165 /// checked and spelled filename, e.g. as an operand of \#include. This returns
1166 /// true if the input filename was in <>'s or false if it were in ""'s.  The
1167 /// caller is expected to provide a buffer that is large enough to hold the
1168 /// spelling of the filename, but is also expected to handle the case when
1169 /// this method decides to use a different buffer.
GetIncludeFilenameSpelling(SourceLocation Loc,StringRef & Buffer)1170 bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
1171                                               StringRef &Buffer) {
1172   // Get the text form of the filename.
1173   assert(!Buffer.empty() && "Can't have tokens with empty spellings!");
1174 
1175   // Make sure the filename is <x> or "x".
1176   bool isAngled;
1177   if (Buffer[0] == '<') {
1178     if (Buffer.back() != '>') {
1179       Diag(Loc, diag::err_pp_expects_filename);
1180       Buffer = StringRef();
1181       return true;
1182     }
1183     isAngled = true;
1184   } else if (Buffer[0] == '"') {
1185     if (Buffer.back() != '"') {
1186       Diag(Loc, diag::err_pp_expects_filename);
1187       Buffer = StringRef();
1188       return true;
1189     }
1190     isAngled = false;
1191   } else {
1192     Diag(Loc, diag::err_pp_expects_filename);
1193     Buffer = StringRef();
1194     return true;
1195   }
1196 
1197   // Diagnose #include "" as invalid.
1198   if (Buffer.size() <= 2) {
1199     Diag(Loc, diag::err_pp_empty_filename);
1200     Buffer = StringRef();
1201     return true;
1202   }
1203 
1204   // Skip the brackets.
1205   Buffer = Buffer.substr(1, Buffer.size()-2);
1206   return isAngled;
1207 }
1208 
1209 /// \brief Handle cases where the \#include name is expanded from a macro
1210 /// as multiple tokens, which need to be glued together.
1211 ///
1212 /// This occurs for code like:
1213 /// \code
1214 ///    \#define FOO <a/b.h>
1215 ///    \#include FOO
1216 /// \endcode
1217 /// because in this case, "<a/b.h>" is returned as 7 tokens, not one.
1218 ///
1219 /// This code concatenates and consumes tokens up to the '>' token.  It returns
1220 /// false if the > was found, otherwise it returns true if it finds and consumes
1221 /// the EOD marker.
ConcatenateIncludeName(SmallString<128> & FilenameBuffer,SourceLocation & End)1222 bool Preprocessor::ConcatenateIncludeName(
1223                                         SmallString<128> &FilenameBuffer,
1224                                           SourceLocation &End) {
1225   Token CurTok;
1226 
1227   Lex(CurTok);
1228   while (CurTok.isNot(tok::eod)) {
1229     End = CurTok.getLocation();
1230 
1231     // FIXME: Provide code completion for #includes.
1232     if (CurTok.is(tok::code_completion)) {
1233       setCodeCompletionReached();
1234       Lex(CurTok);
1235       continue;
1236     }
1237 
1238     // Append the spelling of this token to the buffer. If there was a space
1239     // before it, add it now.
1240     if (CurTok.hasLeadingSpace())
1241       FilenameBuffer.push_back(' ');
1242 
1243     // Get the spelling of the token, directly into FilenameBuffer if possible.
1244     unsigned PreAppendSize = FilenameBuffer.size();
1245     FilenameBuffer.resize(PreAppendSize+CurTok.getLength());
1246 
1247     const char *BufPtr = &FilenameBuffer[PreAppendSize];
1248     unsigned ActualLen = getSpelling(CurTok, BufPtr);
1249 
1250     // If the token was spelled somewhere else, copy it into FilenameBuffer.
1251     if (BufPtr != &FilenameBuffer[PreAppendSize])
1252       memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1253 
1254     // Resize FilenameBuffer to the correct size.
1255     if (CurTok.getLength() != ActualLen)
1256       FilenameBuffer.resize(PreAppendSize+ActualLen);
1257 
1258     // If we found the '>' marker, return success.
1259     if (CurTok.is(tok::greater))
1260       return false;
1261 
1262     Lex(CurTok);
1263   }
1264 
1265   // If we hit the eod marker, emit an error and return true so that the caller
1266   // knows the EOD has been read.
1267   Diag(CurTok.getLocation(), diag::err_pp_expects_filename);
1268   return true;
1269 }
1270 
1271 /// HandleIncludeDirective - The "\#include" tokens have just been read, read
1272 /// the file to be included from the lexer, then include it!  This is a common
1273 /// routine with functionality shared between \#include, \#include_next and
1274 /// \#import.  LookupFrom is set when this is a \#include_next directive, it
1275 /// specifies the file to start searching from.
HandleIncludeDirective(SourceLocation HashLoc,Token & IncludeTok,const DirectoryLookup * LookupFrom,bool isImport)1276 void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc,
1277                                           Token &IncludeTok,
1278                                           const DirectoryLookup *LookupFrom,
1279                                           bool isImport) {
1280 
1281   Token FilenameTok;
1282   CurPPLexer->LexIncludeFilename(FilenameTok);
1283 
1284   // Reserve a buffer to get the spelling.
1285   SmallString<128> FilenameBuffer;
1286   StringRef Filename;
1287   SourceLocation End;
1288   SourceLocation CharEnd; // the end of this directive, in characters
1289 
1290   switch (FilenameTok.getKind()) {
1291   case tok::eod:
1292     // If the token kind is EOD, the error has already been diagnosed.
1293     return;
1294 
1295   case tok::angle_string_literal:
1296   case tok::string_literal:
1297     Filename = getSpelling(FilenameTok, FilenameBuffer);
1298     End = FilenameTok.getLocation();
1299     CharEnd = End.getLocWithOffset(Filename.size());
1300     break;
1301 
1302   case tok::less:
1303     // This could be a <foo/bar.h> file coming from a macro expansion.  In this
1304     // case, glue the tokens together into FilenameBuffer and interpret those.
1305     FilenameBuffer.push_back('<');
1306     if (ConcatenateIncludeName(FilenameBuffer, End))
1307       return;   // Found <eod> but no ">"?  Diagnostic already emitted.
1308     Filename = FilenameBuffer.str();
1309     CharEnd = getLocForEndOfToken(End);
1310     break;
1311   default:
1312     Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
1313     DiscardUntilEndOfDirective();
1314     return;
1315   }
1316 
1317   StringRef OriginalFilename = Filename;
1318   bool isAngled =
1319     GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename);
1320   // If GetIncludeFilenameSpelling set the start ptr to null, there was an
1321   // error.
1322   if (Filename.empty()) {
1323     DiscardUntilEndOfDirective();
1324     return;
1325   }
1326 
1327   // Verify that there is nothing after the filename, other than EOD.  Note that
1328   // we allow macros that expand to nothing after the filename, because this
1329   // falls into the category of "#include pp-tokens new-line" specified in
1330   // C99 6.10.2p4.
1331   CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true);
1332 
1333   // Check that we don't have infinite #include recursion.
1334   if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
1335     Diag(FilenameTok, diag::err_pp_include_too_deep);
1336     return;
1337   }
1338 
1339   // Complain about attempts to #include files in an audit pragma.
1340   if (PragmaARCCFCodeAuditedLoc.isValid()) {
1341     Diag(HashLoc, diag::err_pp_include_in_arc_cf_code_audited);
1342     Diag(PragmaARCCFCodeAuditedLoc, diag::note_pragma_entered_here);
1343 
1344     // Immediately leave the pragma.
1345     PragmaARCCFCodeAuditedLoc = SourceLocation();
1346   }
1347 
1348   if (HeaderInfo.HasIncludeAliasMap()) {
1349     // Map the filename with the brackets still attached.  If the name doesn't
1350     // map to anything, fall back on the filename we've already gotten the
1351     // spelling for.
1352     StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(OriginalFilename);
1353     if (!NewName.empty())
1354       Filename = NewName;
1355   }
1356 
1357   // Search include directories.
1358   const DirectoryLookup *CurDir;
1359   SmallString<1024> SearchPath;
1360   SmallString<1024> RelativePath;
1361   // We get the raw path only if we have 'Callbacks' to which we later pass
1362   // the path.
1363   Module *SuggestedModule = 0;
1364   const FileEntry *File = LookupFile(
1365       Filename, isAngled, LookupFrom, CurDir,
1366       Callbacks ? &SearchPath : NULL, Callbacks ? &RelativePath : NULL,
1367       getLangOpts().Modules? &SuggestedModule : 0);
1368 
1369   if (Callbacks) {
1370     if (!File) {
1371       // Give the clients a chance to recover.
1372       SmallString<128> RecoveryPath;
1373       if (Callbacks->FileNotFound(Filename, RecoveryPath)) {
1374         if (const DirectoryEntry *DE = FileMgr.getDirectory(RecoveryPath)) {
1375           // Add the recovery path to the list of search paths.
1376           DirectoryLookup DL(DE, SrcMgr::C_User, true, false);
1377           HeaderInfo.AddSearchPath(DL, isAngled);
1378 
1379           // Try the lookup again, skipping the cache.
1380           File = LookupFile(Filename, isAngled, LookupFrom, CurDir, 0, 0,
1381                             getLangOpts().Modules? &SuggestedModule : 0,
1382                             /*SkipCache*/true);
1383         }
1384       }
1385     }
1386 
1387     // Notify the callback object that we've seen an inclusion directive.
1388     Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled, File,
1389                                   End, SearchPath, RelativePath);
1390   }
1391 
1392   if (File == 0) {
1393     if (!SuppressIncludeNotFoundError) {
1394       // If the file could not be located and it was included via angle
1395       // brackets, we can attempt a lookup as though it were a quoted path to
1396       // provide the user with a possible fixit.
1397       if (isAngled) {
1398         File = LookupFile(Filename, false, LookupFrom, CurDir,
1399                           Callbacks ? &SearchPath : 0,
1400                           Callbacks ? &RelativePath : 0,
1401                           getLangOpts().Modules ? &SuggestedModule : 0);
1402         if (File) {
1403           SourceRange Range(FilenameTok.getLocation(), CharEnd);
1404           Diag(FilenameTok, diag::err_pp_file_not_found_not_fatal) <<
1405             Filename <<
1406             FixItHint::CreateReplacement(Range, "\"" + Filename.str() + "\"");
1407         }
1408       }
1409       // If the file is still not found, just go with the vanilla diagnostic
1410       if (!File)
1411         Diag(FilenameTok, diag::err_pp_file_not_found) << Filename;
1412     }
1413     if (!File)
1414       return;
1415   }
1416 
1417   // If we are supposed to import a module rather than including the header,
1418   // do so now.
1419   if (SuggestedModule) {
1420     // Compute the module access path corresponding to this module.
1421     // FIXME: Should we have a second loadModule() overload to avoid this
1422     // extra lookup step?
1423     llvm::SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> Path;
1424     for (Module *Mod = SuggestedModule; Mod; Mod = Mod->Parent)
1425       Path.push_back(std::make_pair(getIdentifierInfo(Mod->Name),
1426                                     FilenameTok.getLocation()));
1427     std::reverse(Path.begin(), Path.end());
1428 
1429     // Warn that we're replacing the include/import with a module import.
1430     SmallString<128> PathString;
1431     for (unsigned I = 0, N = Path.size(); I != N; ++I) {
1432       if (I)
1433         PathString += '.';
1434       PathString += Path[I].first->getName();
1435     }
1436     int IncludeKind = 0;
1437 
1438     switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) {
1439     case tok::pp_include:
1440       IncludeKind = 0;
1441       break;
1442 
1443     case tok::pp_import:
1444       IncludeKind = 1;
1445       break;
1446 
1447     case tok::pp_include_next:
1448       IncludeKind = 2;
1449       break;
1450 
1451     case tok::pp___include_macros:
1452       IncludeKind = 3;
1453       break;
1454 
1455     default:
1456       llvm_unreachable("unknown include directive kind");
1457     }
1458 
1459     // Determine whether we are actually building the module that this
1460     // include directive maps to.
1461     bool BuildingImportedModule
1462       = Path[0].first->getName() == getLangOpts().CurrentModule;
1463 
1464     if (!BuildingImportedModule && getLangOpts().ObjC2) {
1465       // If we're not building the imported module, warn that we're going
1466       // to automatically turn this inclusion directive into a module import.
1467       // We only do this in Objective-C, where we have a module-import syntax.
1468       CharSourceRange ReplaceRange(SourceRange(HashLoc, CharEnd),
1469                                    /*IsTokenRange=*/false);
1470       Diag(HashLoc, diag::warn_auto_module_import)
1471         << IncludeKind << PathString
1472         << FixItHint::CreateReplacement(ReplaceRange,
1473              "@__experimental_modules_import " + PathString.str().str() + ";");
1474     }
1475 
1476     // Load the module.
1477     // If this was an #__include_macros directive, only make macros visible.
1478     Module::NameVisibilityKind Visibility
1479       = (IncludeKind == 3)? Module::MacrosVisible : Module::AllVisible;
1480     Module *Imported
1481       = TheModuleLoader.loadModule(IncludeTok.getLocation(), Path, Visibility,
1482                                    /*IsIncludeDirective=*/true);
1483 
1484     // If this header isn't part of the module we're building, we're done.
1485     if (!BuildingImportedModule && Imported)
1486       return;
1487   }
1488 
1489   // The #included file will be considered to be a system header if either it is
1490   // in a system include directory, or if the #includer is a system include
1491   // header.
1492   SrcMgr::CharacteristicKind FileCharacter =
1493     std::max(HeaderInfo.getFileDirFlavor(File),
1494              SourceMgr.getFileCharacteristic(FilenameTok.getLocation()));
1495 
1496   // Ask HeaderInfo if we should enter this #include file.  If not, #including
1497   // this file will have no effect.
1498   if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) {
1499     if (Callbacks)
1500       Callbacks->FileSkipped(*File, FilenameTok, FileCharacter);
1501     return;
1502   }
1503 
1504   // Look up the file, create a File ID for it.
1505   SourceLocation IncludePos = End;
1506   // If the filename string was the result of macro expansions, set the include
1507   // position on the file where it will be included and after the expansions.
1508   if (IncludePos.isMacroID())
1509     IncludePos = SourceMgr.getExpansionRange(IncludePos).second;
1510   FileID FID = SourceMgr.createFileID(File, IncludePos, FileCharacter);
1511   assert(!FID.isInvalid() && "Expected valid file ID");
1512 
1513   // Finally, if all is good, enter the new file!
1514   EnterSourceFile(FID, CurDir, FilenameTok.getLocation());
1515 }
1516 
1517 /// HandleIncludeNextDirective - Implements \#include_next.
1518 ///
HandleIncludeNextDirective(SourceLocation HashLoc,Token & IncludeNextTok)1519 void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc,
1520                                               Token &IncludeNextTok) {
1521   Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
1522 
1523   // #include_next is like #include, except that we start searching after
1524   // the current found directory.  If we can't do this, issue a
1525   // diagnostic.
1526   const DirectoryLookup *Lookup = CurDirLookup;
1527   if (isInPrimaryFile()) {
1528     Lookup = 0;
1529     Diag(IncludeNextTok, diag::pp_include_next_in_primary);
1530   } else if (Lookup == 0) {
1531     Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
1532   } else {
1533     // Start looking up in the next directory.
1534     ++Lookup;
1535   }
1536 
1537   return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup);
1538 }
1539 
1540 /// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode
HandleMicrosoftImportDirective(Token & Tok)1541 void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) {
1542   // The Microsoft #import directive takes a type library and generates header
1543   // files from it, and includes those.  This is beyond the scope of what clang
1544   // does, so we ignore it and error out.  However, #import can optionally have
1545   // trailing attributes that span multiple lines.  We're going to eat those
1546   // so we can continue processing from there.
1547   Diag(Tok, diag::err_pp_import_directive_ms );
1548 
1549   // Read tokens until we get to the end of the directive.  Note that the
1550   // directive can be split over multiple lines using the backslash character.
1551   DiscardUntilEndOfDirective();
1552 }
1553 
1554 /// HandleImportDirective - Implements \#import.
1555 ///
HandleImportDirective(SourceLocation HashLoc,Token & ImportTok)1556 void Preprocessor::HandleImportDirective(SourceLocation HashLoc,
1557                                          Token &ImportTok) {
1558   if (!LangOpts.ObjC1) {  // #import is standard for ObjC.
1559     if (LangOpts.MicrosoftMode)
1560       return HandleMicrosoftImportDirective(ImportTok);
1561     Diag(ImportTok, diag::ext_pp_import_directive);
1562   }
1563   return HandleIncludeDirective(HashLoc, ImportTok, 0, true);
1564 }
1565 
1566 /// HandleIncludeMacrosDirective - The -imacros command line option turns into a
1567 /// pseudo directive in the predefines buffer.  This handles it by sucking all
1568 /// tokens through the preprocessor and discarding them (only keeping the side
1569 /// effects on the preprocessor).
HandleIncludeMacrosDirective(SourceLocation HashLoc,Token & IncludeMacrosTok)1570 void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc,
1571                                                 Token &IncludeMacrosTok) {
1572   // This directive should only occur in the predefines buffer.  If not, emit an
1573   // error and reject it.
1574   SourceLocation Loc = IncludeMacrosTok.getLocation();
1575   if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) {
1576     Diag(IncludeMacrosTok.getLocation(),
1577          diag::pp_include_macros_out_of_predefines);
1578     DiscardUntilEndOfDirective();
1579     return;
1580   }
1581 
1582   // Treat this as a normal #include for checking purposes.  If this is
1583   // successful, it will push a new lexer onto the include stack.
1584   HandleIncludeDirective(HashLoc, IncludeMacrosTok, 0, false);
1585 
1586   Token TmpTok;
1587   do {
1588     Lex(TmpTok);
1589     assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!");
1590   } while (TmpTok.isNot(tok::hashhash));
1591 }
1592 
1593 //===----------------------------------------------------------------------===//
1594 // Preprocessor Macro Directive Handling.
1595 //===----------------------------------------------------------------------===//
1596 
1597 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1598 /// definition has just been read.  Lex the rest of the arguments and the
1599 /// closing ), updating MI with what we learn.  Return true if an error occurs
1600 /// parsing the arg list.
ReadMacroDefinitionArgList(MacroInfo * MI,Token & Tok)1601 bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI, Token &Tok) {
1602   SmallVector<IdentifierInfo*, 32> Arguments;
1603 
1604   while (1) {
1605     LexUnexpandedToken(Tok);
1606     switch (Tok.getKind()) {
1607     case tok::r_paren:
1608       // Found the end of the argument list.
1609       if (Arguments.empty())  // #define FOO()
1610         return false;
1611       // Otherwise we have #define FOO(A,)
1612       Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
1613       return true;
1614     case tok::ellipsis:  // #define X(... -> C99 varargs
1615       if (!LangOpts.C99)
1616         Diag(Tok, LangOpts.CPlusPlus0x ?
1617              diag::warn_cxx98_compat_variadic_macro :
1618              diag::ext_variadic_macro);
1619 
1620       // Lex the token after the identifier.
1621       LexUnexpandedToken(Tok);
1622       if (Tok.isNot(tok::r_paren)) {
1623         Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1624         return true;
1625       }
1626       // Add the __VA_ARGS__ identifier as an argument.
1627       Arguments.push_back(Ident__VA_ARGS__);
1628       MI->setIsC99Varargs();
1629       MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1630       return false;
1631     case tok::eod:  // #define X(
1632       Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1633       return true;
1634     default:
1635       // Handle keywords and identifiers here to accept things like
1636       // #define Foo(for) for.
1637       IdentifierInfo *II = Tok.getIdentifierInfo();
1638       if (II == 0) {
1639         // #define X(1
1640         Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
1641         return true;
1642       }
1643 
1644       // If this is already used as an argument, it is used multiple times (e.g.
1645       // #define X(A,A.
1646       if (std::find(Arguments.begin(), Arguments.end(), II) !=
1647           Arguments.end()) {  // C99 6.10.3p6
1648         Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II;
1649         return true;
1650       }
1651 
1652       // Add the argument to the macro info.
1653       Arguments.push_back(II);
1654 
1655       // Lex the token after the identifier.
1656       LexUnexpandedToken(Tok);
1657 
1658       switch (Tok.getKind()) {
1659       default:          // #define X(A B
1660         Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
1661         return true;
1662       case tok::r_paren: // #define X(A)
1663         MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1664         return false;
1665       case tok::comma:  // #define X(A,
1666         break;
1667       case tok::ellipsis:  // #define X(A... -> GCC extension
1668         // Diagnose extension.
1669         Diag(Tok, diag::ext_named_variadic_macro);
1670 
1671         // Lex the token after the identifier.
1672         LexUnexpandedToken(Tok);
1673         if (Tok.isNot(tok::r_paren)) {
1674           Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1675           return true;
1676         }
1677 
1678         MI->setIsGNUVarargs();
1679         MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1680         return false;
1681       }
1682     }
1683   }
1684 }
1685 
1686 /// HandleDefineDirective - Implements \#define.  This consumes the entire macro
1687 /// line then lets the caller lex the next real token.
HandleDefineDirective(Token & DefineTok)1688 void Preprocessor::HandleDefineDirective(Token &DefineTok) {
1689   ++NumDefined;
1690 
1691   Token MacroNameTok;
1692   ReadMacroName(MacroNameTok, 1);
1693 
1694   // Error reading macro name?  If so, diagnostic already issued.
1695   if (MacroNameTok.is(tok::eod))
1696     return;
1697 
1698   Token LastTok = MacroNameTok;
1699 
1700   // If we are supposed to keep comments in #defines, reenable comment saving
1701   // mode.
1702   if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
1703 
1704   // Create the new macro.
1705   MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation());
1706 
1707   Token Tok;
1708   LexUnexpandedToken(Tok);
1709 
1710   // If this is a function-like macro definition, parse the argument list,
1711   // marking each of the identifiers as being used as macro arguments.  Also,
1712   // check other constraints on the first token of the macro body.
1713   if (Tok.is(tok::eod)) {
1714     // If there is no body to this macro, we have no special handling here.
1715   } else if (Tok.hasLeadingSpace()) {
1716     // This is a normal token with leading space.  Clear the leading space
1717     // marker on the first token to get proper expansion.
1718     Tok.clearFlag(Token::LeadingSpace);
1719   } else if (Tok.is(tok::l_paren)) {
1720     // This is a function-like macro definition.  Read the argument list.
1721     MI->setIsFunctionLike();
1722     if (ReadMacroDefinitionArgList(MI, LastTok)) {
1723       // Forget about MI.
1724       ReleaseMacroInfo(MI);
1725       // Throw away the rest of the line.
1726       if (CurPPLexer->ParsingPreprocessorDirective)
1727         DiscardUntilEndOfDirective();
1728       return;
1729     }
1730 
1731     // If this is a definition of a variadic C99 function-like macro, not using
1732     // the GNU named varargs extension, enabled __VA_ARGS__.
1733 
1734     // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
1735     // This gets unpoisoned where it is allowed.
1736     assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!");
1737     if (MI->isC99Varargs())
1738       Ident__VA_ARGS__->setIsPoisoned(false);
1739 
1740     // Read the first token after the arg list for down below.
1741     LexUnexpandedToken(Tok);
1742   } else if (LangOpts.C99 || LangOpts.CPlusPlus0x) {
1743     // C99 requires whitespace between the macro definition and the body.  Emit
1744     // a diagnostic for something like "#define X+".
1745     Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
1746   } else {
1747     // C90 6.8 TC1 says: "In the definition of an object-like macro, if the
1748     // first character of a replacement list is not a character required by
1749     // subclause 5.2.1, then there shall be white-space separation between the
1750     // identifier and the replacement list.".  5.2.1 lists this set:
1751     //   "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which
1752     // is irrelevant here.
1753     bool isInvalid = false;
1754     if (Tok.is(tok::at)) // @ is not in the list above.
1755       isInvalid = true;
1756     else if (Tok.is(tok::unknown)) {
1757       // If we have an unknown token, it is something strange like "`".  Since
1758       // all of valid characters would have lexed into a single character
1759       // token of some sort, we know this is not a valid case.
1760       isInvalid = true;
1761     }
1762     if (isInvalid)
1763       Diag(Tok, diag::ext_missing_whitespace_after_macro_name);
1764     else
1765       Diag(Tok, diag::warn_missing_whitespace_after_macro_name);
1766   }
1767 
1768   if (!Tok.is(tok::eod))
1769     LastTok = Tok;
1770 
1771   // Read the rest of the macro body.
1772   if (MI->isObjectLike()) {
1773     // Object-like macros are very simple, just read their body.
1774     while (Tok.isNot(tok::eod)) {
1775       LastTok = Tok;
1776       MI->AddTokenToBody(Tok);
1777       // Get the next token of the macro.
1778       LexUnexpandedToken(Tok);
1779     }
1780 
1781   } else {
1782     // Otherwise, read the body of a function-like macro.  While we are at it,
1783     // check C99 6.10.3.2p1: ensure that # operators are followed by macro
1784     // parameters in function-like macro expansions.
1785     while (Tok.isNot(tok::eod)) {
1786       LastTok = Tok;
1787 
1788       if (Tok.isNot(tok::hash)) {
1789         MI->AddTokenToBody(Tok);
1790 
1791         // Get the next token of the macro.
1792         LexUnexpandedToken(Tok);
1793         continue;
1794       }
1795 
1796       // Get the next token of the macro.
1797       LexUnexpandedToken(Tok);
1798 
1799       // Check for a valid macro arg identifier.
1800       if (Tok.getIdentifierInfo() == 0 ||
1801           MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) {
1802 
1803         // If this is assembler-with-cpp mode, we accept random gibberish after
1804         // the '#' because '#' is often a comment character.  However, change
1805         // the kind of the token to tok::unknown so that the preprocessor isn't
1806         // confused.
1807         if (getLangOpts().AsmPreprocessor && Tok.isNot(tok::eod)) {
1808           LastTok.setKind(tok::unknown);
1809         } else {
1810           Diag(Tok, diag::err_pp_stringize_not_parameter);
1811           ReleaseMacroInfo(MI);
1812 
1813           // Disable __VA_ARGS__ again.
1814           Ident__VA_ARGS__->setIsPoisoned(true);
1815           return;
1816         }
1817       }
1818 
1819       // Things look ok, add the '#' and param name tokens to the macro.
1820       MI->AddTokenToBody(LastTok);
1821       MI->AddTokenToBody(Tok);
1822       LastTok = Tok;
1823 
1824       // Get the next token of the macro.
1825       LexUnexpandedToken(Tok);
1826     }
1827   }
1828 
1829 
1830   // Disable __VA_ARGS__ again.
1831   Ident__VA_ARGS__->setIsPoisoned(true);
1832 
1833   // Check that there is no paste (##) operator at the beginning or end of the
1834   // replacement list.
1835   unsigned NumTokens = MI->getNumTokens();
1836   if (NumTokens != 0) {
1837     if (MI->getReplacementToken(0).is(tok::hashhash)) {
1838       Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
1839       ReleaseMacroInfo(MI);
1840       return;
1841     }
1842     if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
1843       Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
1844       ReleaseMacroInfo(MI);
1845       return;
1846     }
1847   }
1848 
1849   MI->setDefinitionEndLoc(LastTok.getLocation());
1850 
1851   // Finally, if this identifier already had a macro defined for it, verify that
1852   // the macro bodies are identical, and issue diagnostics if they are not.
1853   if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) {
1854     // It is very common for system headers to have tons of macro redefinitions
1855     // and for warnings to be disabled in system headers.  If this is the case,
1856     // then don't bother calling MacroInfo::isIdenticalTo.
1857     if (!getDiagnostics().getSuppressSystemWarnings() ||
1858         !SourceMgr.isInSystemHeader(DefineTok.getLocation())) {
1859       if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused())
1860         Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
1861 
1862       // Macros must be identical.  This means all tokens and whitespace
1863       // separation must be the same.  C99 6.10.3.2.
1864       if (!OtherMI->isAllowRedefinitionsWithoutWarning() &&
1865           !MI->isIdenticalTo(*OtherMI, *this)) {
1866         Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
1867           << MacroNameTok.getIdentifierInfo();
1868         Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition);
1869       }
1870     }
1871     if (OtherMI->isWarnIfUnused())
1872       WarnUnusedMacroLocs.erase(OtherMI->getDefinitionLoc());
1873   }
1874 
1875   setMacroInfo(MacroNameTok.getIdentifierInfo(), MI);
1876 
1877   assert(!MI->isUsed());
1878   // If we need warning for not using the macro, add its location in the
1879   // warn-because-unused-macro set. If it gets used it will be removed from set.
1880   if (isInPrimaryFile() && // don't warn for include'd macros.
1881       Diags->getDiagnosticLevel(diag::pp_macro_not_used,
1882           MI->getDefinitionLoc()) != DiagnosticsEngine::Ignored) {
1883     MI->setIsWarnIfUnused(true);
1884     WarnUnusedMacroLocs.insert(MI->getDefinitionLoc());
1885   }
1886 
1887   // If the callbacks want to know, tell them about the macro definition.
1888   if (Callbacks)
1889     Callbacks->MacroDefined(MacroNameTok, MI);
1890 }
1891 
1892 /// HandleUndefDirective - Implements \#undef.
1893 ///
HandleUndefDirective(Token & UndefTok)1894 void Preprocessor::HandleUndefDirective(Token &UndefTok) {
1895   ++NumUndefined;
1896 
1897   Token MacroNameTok;
1898   ReadMacroName(MacroNameTok, 2);
1899 
1900   // Error reading macro name?  If so, diagnostic already issued.
1901   if (MacroNameTok.is(tok::eod))
1902     return;
1903 
1904   // Check to see if this is the last token on the #undef line.
1905   CheckEndOfDirective("undef");
1906 
1907   // Okay, we finally have a valid identifier to undef.
1908   MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo());
1909 
1910   // If the macro is not defined, this is a noop undef, just return.
1911   if (MI == 0) return;
1912 
1913   if (!MI->isUsed() && MI->isWarnIfUnused())
1914     Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
1915 
1916   // If the callbacks want to know, tell them about the macro #undef.
1917   if (Callbacks)
1918     Callbacks->MacroUndefined(MacroNameTok, MI);
1919 
1920   if (MI->isWarnIfUnused())
1921     WarnUnusedMacroLocs.erase(MI->getDefinitionLoc());
1922 
1923   MI->setUndefLoc(MacroNameTok.getLocation());
1924   clearMacroInfo(MacroNameTok.getIdentifierInfo());
1925 }
1926 
1927 
1928 //===----------------------------------------------------------------------===//
1929 // Preprocessor Conditional Directive Handling.
1930 //===----------------------------------------------------------------------===//
1931 
1932 /// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive.  isIfndef
1933 /// is true when this is a \#ifndef directive.  ReadAnyTokensBeforeDirective is
1934 /// true if any tokens have been returned or pp-directives activated before this
1935 /// \#ifndef has been lexed.
1936 ///
HandleIfdefDirective(Token & Result,bool isIfndef,bool ReadAnyTokensBeforeDirective)1937 void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef,
1938                                         bool ReadAnyTokensBeforeDirective) {
1939   ++NumIf;
1940   Token DirectiveTok = Result;
1941 
1942   Token MacroNameTok;
1943   ReadMacroName(MacroNameTok);
1944 
1945   // Error reading macro name?  If so, diagnostic already issued.
1946   if (MacroNameTok.is(tok::eod)) {
1947     // Skip code until we get to #endif.  This helps with recovery by not
1948     // emitting an error when the #endif is reached.
1949     SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
1950                                  /*Foundnonskip*/false, /*FoundElse*/false);
1951     return;
1952   }
1953 
1954   // Check to see if this is the last token on the #if[n]def line.
1955   CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef");
1956 
1957   IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
1958   MacroInfo *MI = getMacroInfo(MII);
1959 
1960   if (CurPPLexer->getConditionalStackDepth() == 0) {
1961     // If the start of a top-level #ifdef and if the macro is not defined,
1962     // inform MIOpt that this might be the start of a proper include guard.
1963     // Otherwise it is some other form of unknown conditional which we can't
1964     // handle.
1965     if (!ReadAnyTokensBeforeDirective && MI == 0) {
1966       assert(isIfndef && "#ifdef shouldn't reach here");
1967       CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MII);
1968     } else
1969       CurPPLexer->MIOpt.EnterTopLevelConditional();
1970   }
1971 
1972   // If there is a macro, process it.
1973   if (MI)  // Mark it used.
1974     markMacroAsUsed(MI);
1975 
1976   if (Callbacks) {
1977     if (isIfndef)
1978       Callbacks->Ifndef(DirectiveTok.getLocation(), MacroNameTok);
1979     else
1980       Callbacks->Ifdef(DirectiveTok.getLocation(), MacroNameTok);
1981   }
1982 
1983   // Should we include the stuff contained by this directive?
1984   if (!MI == isIfndef) {
1985     // Yes, remember that we are inside a conditional, then lex the next token.
1986     CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(),
1987                                      /*wasskip*/false, /*foundnonskip*/true,
1988                                      /*foundelse*/false);
1989   } else {
1990     // No, skip the contents of this block.
1991     SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
1992                                  /*Foundnonskip*/false,
1993                                  /*FoundElse*/false);
1994   }
1995 }
1996 
1997 /// HandleIfDirective - Implements the \#if directive.
1998 ///
HandleIfDirective(Token & IfToken,bool ReadAnyTokensBeforeDirective)1999 void Preprocessor::HandleIfDirective(Token &IfToken,
2000                                      bool ReadAnyTokensBeforeDirective) {
2001   ++NumIf;
2002 
2003   // Parse and evaluate the conditional expression.
2004   IdentifierInfo *IfNDefMacro = 0;
2005   const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2006   const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro);
2007   const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2008 
2009   // If this condition is equivalent to #ifndef X, and if this is the first
2010   // directive seen, handle it for the multiple-include optimization.
2011   if (CurPPLexer->getConditionalStackDepth() == 0) {
2012     if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue)
2013       CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro);
2014     else
2015       CurPPLexer->MIOpt.EnterTopLevelConditional();
2016   }
2017 
2018   if (Callbacks)
2019     Callbacks->If(IfToken.getLocation(),
2020                   SourceRange(ConditionalBegin, ConditionalEnd));
2021 
2022   // Should we include the stuff contained by this directive?
2023   if (ConditionalTrue) {
2024     // Yes, remember that we are inside a conditional, then lex the next token.
2025     CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
2026                                    /*foundnonskip*/true, /*foundelse*/false);
2027   } else {
2028     // No, skip the contents of this block.
2029     SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false,
2030                                  /*FoundElse*/false);
2031   }
2032 }
2033 
2034 /// HandleEndifDirective - Implements the \#endif directive.
2035 ///
HandleEndifDirective(Token & EndifToken)2036 void Preprocessor::HandleEndifDirective(Token &EndifToken) {
2037   ++NumEndif;
2038 
2039   // Check that this is the whole directive.
2040   CheckEndOfDirective("endif");
2041 
2042   PPConditionalInfo CondInfo;
2043   if (CurPPLexer->popConditionalLevel(CondInfo)) {
2044     // No conditionals on the stack: this is an #endif without an #if.
2045     Diag(EndifToken, diag::err_pp_endif_without_if);
2046     return;
2047   }
2048 
2049   // If this the end of a top-level #endif, inform MIOpt.
2050   if (CurPPLexer->getConditionalStackDepth() == 0)
2051     CurPPLexer->MIOpt.ExitTopLevelConditional();
2052 
2053   assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
2054          "This code should only be reachable in the non-skipping case!");
2055 
2056   if (Callbacks)
2057     Callbacks->Endif(EndifToken.getLocation(), CondInfo.IfLoc);
2058 }
2059 
2060 /// HandleElseDirective - Implements the \#else directive.
2061 ///
HandleElseDirective(Token & Result)2062 void Preprocessor::HandleElseDirective(Token &Result) {
2063   ++NumElse;
2064 
2065   // #else directive in a non-skipping conditional... start skipping.
2066   CheckEndOfDirective("else");
2067 
2068   PPConditionalInfo CI;
2069   if (CurPPLexer->popConditionalLevel(CI)) {
2070     Diag(Result, diag::pp_err_else_without_if);
2071     return;
2072   }
2073 
2074   // If this is a top-level #else, inform the MIOpt.
2075   if (CurPPLexer->getConditionalStackDepth() == 0)
2076     CurPPLexer->MIOpt.EnterTopLevelConditional();
2077 
2078   // If this is a #else with a #else before it, report the error.
2079   if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
2080 
2081   if (Callbacks)
2082     Callbacks->Else(Result.getLocation(), CI.IfLoc);
2083 
2084   // Finally, skip the rest of the contents of this block.
2085   SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2086                                /*FoundElse*/true, Result.getLocation());
2087 }
2088 
2089 /// HandleElifDirective - Implements the \#elif directive.
2090 ///
HandleElifDirective(Token & ElifToken)2091 void Preprocessor::HandleElifDirective(Token &ElifToken) {
2092   ++NumElse;
2093 
2094   // #elif directive in a non-skipping conditional... start skipping.
2095   // We don't care what the condition is, because we will always skip it (since
2096   // the block immediately before it was included).
2097   const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2098   DiscardUntilEndOfDirective();
2099   const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2100 
2101   PPConditionalInfo CI;
2102   if (CurPPLexer->popConditionalLevel(CI)) {
2103     Diag(ElifToken, diag::pp_err_elif_without_if);
2104     return;
2105   }
2106 
2107   // If this is a top-level #elif, inform the MIOpt.
2108   if (CurPPLexer->getConditionalStackDepth() == 0)
2109     CurPPLexer->MIOpt.EnterTopLevelConditional();
2110 
2111   // If this is a #elif with a #else before it, report the error.
2112   if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
2113 
2114   if (Callbacks)
2115     Callbacks->Elif(ElifToken.getLocation(),
2116                     SourceRange(ConditionalBegin, ConditionalEnd), CI.IfLoc);
2117 
2118   // Finally, skip the rest of the contents of this block.
2119   SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2120                                /*FoundElse*/CI.FoundElse,
2121                                ElifToken.getLocation());
2122 }
2123