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