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/Basic/FileManager.h"
17 #include "clang/Basic/SourceManager.h"
18 #include "clang/Lex/CodeCompletionHandler.h"
19 #include "clang/Lex/HeaderSearch.h"
20 #include "clang/Lex/HeaderSearchOptions.h"
21 #include "clang/Lex/LexDiagnostic.h"
22 #include "clang/Lex/LiteralSupport.h"
23 #include "clang/Lex/MacroInfo.h"
24 #include "clang/Lex/ModuleLoader.h"
25 #include "clang/Lex/Pragma.h"
26 #include "llvm/ADT/APInt.h"
27 #include "llvm/Support/ErrorHandling.h"
28 #include "llvm/Support/Path.h"
29 #include "llvm/Support/SaveAndRestore.h"
30 using namespace clang;
31
32 //===----------------------------------------------------------------------===//
33 // Utility Methods for Preprocessor Directive Handling.
34 //===----------------------------------------------------------------------===//
35
AllocateMacroInfo()36 MacroInfo *Preprocessor::AllocateMacroInfo() {
37 MacroInfoChain *MIChain = BP.Allocate<MacroInfoChain>();
38 MIChain->Next = MIChainHead;
39 MIChainHead = MIChain;
40 return &MIChain->MI;
41 }
42
AllocateMacroInfo(SourceLocation L)43 MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) {
44 MacroInfo *MI = AllocateMacroInfo();
45 new (MI) MacroInfo(L);
46 return MI;
47 }
48
AllocateDeserializedMacroInfo(SourceLocation L,unsigned SubModuleID)49 MacroInfo *Preprocessor::AllocateDeserializedMacroInfo(SourceLocation L,
50 unsigned SubModuleID) {
51 static_assert(llvm::AlignOf<MacroInfo>::Alignment >= sizeof(SubModuleID),
52 "alignment for MacroInfo is less than the ID");
53 DeserializedMacroInfoChain *MIChain =
54 BP.Allocate<DeserializedMacroInfoChain>();
55 MIChain->Next = DeserialMIChainHead;
56 DeserialMIChainHead = MIChain;
57
58 MacroInfo *MI = &MIChain->MI;
59 new (MI) MacroInfo(L);
60 MI->FromASTFile = true;
61 MI->setOwningModuleID(SubModuleID);
62 return MI;
63 }
64
AllocateDefMacroDirective(MacroInfo * MI,SourceLocation Loc)65 DefMacroDirective *Preprocessor::AllocateDefMacroDirective(MacroInfo *MI,
66 SourceLocation Loc) {
67 return new (BP) DefMacroDirective(MI, Loc);
68 }
69
70 UndefMacroDirective *
AllocateUndefMacroDirective(SourceLocation UndefLoc)71 Preprocessor::AllocateUndefMacroDirective(SourceLocation UndefLoc) {
72 return new (BP) UndefMacroDirective(UndefLoc);
73 }
74
75 VisibilityMacroDirective *
AllocateVisibilityMacroDirective(SourceLocation Loc,bool isPublic)76 Preprocessor::AllocateVisibilityMacroDirective(SourceLocation Loc,
77 bool isPublic) {
78 return new (BP) VisibilityMacroDirective(Loc, isPublic);
79 }
80
81 /// \brief Read and discard all tokens remaining on the current line until
82 /// the tok::eod token is found.
DiscardUntilEndOfDirective()83 void Preprocessor::DiscardUntilEndOfDirective() {
84 Token Tmp;
85 do {
86 LexUnexpandedToken(Tmp);
87 assert(Tmp.isNot(tok::eof) && "EOF seen while discarding directive tokens");
88 } while (Tmp.isNot(tok::eod));
89 }
90
91 /// \brief Enumerates possible cases of #define/#undef a reserved identifier.
92 enum MacroDiag {
93 MD_NoWarn, //> Not a reserved identifier
94 MD_KeywordDef, //> Macro hides keyword, enabled by default
95 MD_ReservedMacro //> #define of #undef reserved id, disabled by default
96 };
97
98 /// \brief Checks if the specified identifier is reserved in the specified
99 /// language.
100 /// This function does not check if the identifier is a keyword.
isReservedId(StringRef Text,const LangOptions & Lang)101 static bool isReservedId(StringRef Text, const LangOptions &Lang) {
102 // C++ [macro.names], C11 7.1.3:
103 // All identifiers that begin with an underscore and either an uppercase
104 // letter or another underscore are always reserved for any use.
105 if (Text.size() >= 2 && Text[0] == '_' &&
106 (isUppercase(Text[1]) || Text[1] == '_'))
107 return true;
108 // C++ [global.names]
109 // Each name that contains a double underscore ... is reserved to the
110 // implementation for any use.
111 if (Lang.CPlusPlus) {
112 if (Text.find("__") != StringRef::npos)
113 return true;
114 }
115 return false;
116 }
117
shouldWarnOnMacroDef(Preprocessor & PP,IdentifierInfo * II)118 static MacroDiag shouldWarnOnMacroDef(Preprocessor &PP, IdentifierInfo *II) {
119 const LangOptions &Lang = PP.getLangOpts();
120 StringRef Text = II->getName();
121 if (isReservedId(Text, Lang))
122 return MD_ReservedMacro;
123 if (II->isKeyword(Lang))
124 return MD_KeywordDef;
125 if (Lang.CPlusPlus11 && (Text.equals("override") || Text.equals("final")))
126 return MD_KeywordDef;
127 return MD_NoWarn;
128 }
129
shouldWarnOnMacroUndef(Preprocessor & PP,IdentifierInfo * II)130 static MacroDiag shouldWarnOnMacroUndef(Preprocessor &PP, IdentifierInfo *II) {
131 const LangOptions &Lang = PP.getLangOpts();
132 StringRef Text = II->getName();
133 // Do not warn on keyword undef. It is generally harmless and widely used.
134 if (isReservedId(Text, Lang))
135 return MD_ReservedMacro;
136 return MD_NoWarn;
137 }
138
CheckMacroName(Token & MacroNameTok,MacroUse isDefineUndef,bool * ShadowFlag)139 bool Preprocessor::CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
140 bool *ShadowFlag) {
141 // Missing macro name?
142 if (MacroNameTok.is(tok::eod))
143 return Diag(MacroNameTok, diag::err_pp_missing_macro_name);
144
145 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
146 if (!II) {
147 bool Invalid = false;
148 std::string Spelling = getSpelling(MacroNameTok, &Invalid);
149 if (Invalid)
150 return Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
151 II = getIdentifierInfo(Spelling);
152
153 if (!II->isCPlusPlusOperatorKeyword())
154 return Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
155
156 // C++ 2.5p2: Alternative tokens behave the same as its primary token
157 // except for their spellings.
158 Diag(MacroNameTok, getLangOpts().MicrosoftExt
159 ? diag::ext_pp_operator_used_as_macro_name
160 : diag::err_pp_operator_used_as_macro_name)
161 << II << MacroNameTok.getKind();
162
163 // Allow #defining |and| and friends for Microsoft compatibility or
164 // recovery when legacy C headers are included in C++.
165 MacroNameTok.setIdentifierInfo(II);
166 }
167
168 if ((isDefineUndef != MU_Other) && II->getPPKeywordID() == tok::pp_defined) {
169 // Error if defining "defined": C99 6.10.8/4, C++ [cpp.predefined]p4.
170 return Diag(MacroNameTok, diag::err_defined_macro_name);
171 }
172
173 if (isDefineUndef == MU_Undef) {
174 auto *MI = getMacroInfo(II);
175 if (MI && MI->isBuiltinMacro()) {
176 // Warn if undefining "__LINE__" and other builtins, per C99 6.10.8/4
177 // and C++ [cpp.predefined]p4], but allow it as an extension.
178 Diag(MacroNameTok, diag::ext_pp_undef_builtin_macro);
179 }
180 }
181
182 // If defining/undefining reserved identifier or a keyword, we need to issue
183 // a warning.
184 SourceLocation MacroNameLoc = MacroNameTok.getLocation();
185 if (ShadowFlag)
186 *ShadowFlag = false;
187 if (!SourceMgr.isInSystemHeader(MacroNameLoc) &&
188 (strcmp(SourceMgr.getBufferName(MacroNameLoc), "<built-in>") != 0)) {
189 MacroDiag D = MD_NoWarn;
190 if (isDefineUndef == MU_Define) {
191 D = shouldWarnOnMacroDef(*this, II);
192 }
193 else if (isDefineUndef == MU_Undef)
194 D = shouldWarnOnMacroUndef(*this, II);
195 if (D == MD_KeywordDef) {
196 // We do not want to warn on some patterns widely used in configuration
197 // scripts. This requires analyzing next tokens, so do not issue warnings
198 // now, only inform caller.
199 if (ShadowFlag)
200 *ShadowFlag = true;
201 }
202 if (D == MD_ReservedMacro)
203 Diag(MacroNameTok, diag::warn_pp_macro_is_reserved_id);
204 }
205
206 // Okay, we got a good identifier.
207 return false;
208 }
209
210 /// \brief Lex and validate a macro name, which occurs after a
211 /// \#define or \#undef.
212 ///
213 /// This sets the token kind to eod and discards the rest of the macro line if
214 /// the macro name is invalid.
215 ///
216 /// \param MacroNameTok Token that is expected to be a macro name.
217 /// \param isDefineUndef Context in which macro is used.
218 /// \param ShadowFlag Points to a flag that is set if macro shadows a keyword.
ReadMacroName(Token & MacroNameTok,MacroUse isDefineUndef,bool * ShadowFlag)219 void Preprocessor::ReadMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
220 bool *ShadowFlag) {
221 // Read the token, don't allow macro expansion on it.
222 LexUnexpandedToken(MacroNameTok);
223
224 if (MacroNameTok.is(tok::code_completion)) {
225 if (CodeComplete)
226 CodeComplete->CodeCompleteMacroName(isDefineUndef == MU_Define);
227 setCodeCompletionReached();
228 LexUnexpandedToken(MacroNameTok);
229 }
230
231 if (!CheckMacroName(MacroNameTok, isDefineUndef, ShadowFlag))
232 return;
233
234 // Invalid macro name, read and discard the rest of the line and set the
235 // token kind to tok::eod if necessary.
236 if (MacroNameTok.isNot(tok::eod)) {
237 MacroNameTok.setKind(tok::eod);
238 DiscardUntilEndOfDirective();
239 }
240 }
241
242 /// \brief Ensure that the next token is a tok::eod token.
243 ///
244 /// If not, emit a diagnostic and consume up until the eod. If EnableMacros is
245 /// true, then we consider macros that expand to zero tokens as being ok.
CheckEndOfDirective(const char * DirType,bool EnableMacros)246 void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) {
247 Token Tmp;
248 // Lex unexpanded tokens for most directives: macros might expand to zero
249 // tokens, causing us to miss diagnosing invalid lines. Some directives (like
250 // #line) allow empty macros.
251 if (EnableMacros)
252 Lex(Tmp);
253 else
254 LexUnexpandedToken(Tmp);
255
256 // There should be no tokens after the directive, but we allow them as an
257 // extension.
258 while (Tmp.is(tok::comment)) // Skip comments in -C mode.
259 LexUnexpandedToken(Tmp);
260
261 if (Tmp.isNot(tok::eod)) {
262 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89,
263 // or if this is a macro-style preprocessing directive, because it is more
264 // trouble than it is worth to insert /**/ and check that there is no /**/
265 // in the range also.
266 FixItHint Hint;
267 if ((LangOpts.GNUMode || LangOpts.C99 || LangOpts.CPlusPlus) &&
268 !CurTokenLexer)
269 Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//");
270 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint;
271 DiscardUntilEndOfDirective();
272 }
273 }
274
275
276
277 /// SkipExcludedConditionalBlock - We just read a \#if or related directive and
278 /// decided that the subsequent tokens are in the \#if'd out portion of the
279 /// file. Lex the rest of the file, until we see an \#endif. If
280 /// FoundNonSkipPortion is true, then we have already emitted code for part of
281 /// this \#if directive, so \#else/\#elif blocks should never be entered.
282 /// If ElseOk is true, then \#else directives are ok, if not, then we have
283 /// already seen one so a \#else directive is a duplicate. When this returns,
284 /// the caller can lex the first valid token.
SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,bool FoundNonSkipPortion,bool FoundElse,SourceLocation ElseLoc)285 void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
286 bool FoundNonSkipPortion,
287 bool FoundElse,
288 SourceLocation ElseLoc) {
289 ++NumSkipped;
290 assert(!CurTokenLexer && CurPPLexer && "Lexing a macro, not a file?");
291
292 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false,
293 FoundNonSkipPortion, FoundElse);
294
295 if (CurPTHLexer) {
296 PTHSkipExcludedConditionalBlock();
297 return;
298 }
299
300 // Enter raw mode to disable identifier lookup (and thus macro expansion),
301 // disabling warnings, etc.
302 CurPPLexer->LexingRawMode = true;
303 Token Tok;
304 while (1) {
305 CurLexer->Lex(Tok);
306
307 if (Tok.is(tok::code_completion)) {
308 if (CodeComplete)
309 CodeComplete->CodeCompleteInConditionalExclusion();
310 setCodeCompletionReached();
311 continue;
312 }
313
314 // If this is the end of the buffer, we have an error.
315 if (Tok.is(tok::eof)) {
316 // Emit errors for each unterminated conditional on the stack, including
317 // the current one.
318 while (!CurPPLexer->ConditionalStack.empty()) {
319 if (CurLexer->getFileLoc() != CodeCompletionFileLoc)
320 Diag(CurPPLexer->ConditionalStack.back().IfLoc,
321 diag::err_pp_unterminated_conditional);
322 CurPPLexer->ConditionalStack.pop_back();
323 }
324
325 // Just return and let the caller lex after this #include.
326 break;
327 }
328
329 // If this token is not a preprocessor directive, just skip it.
330 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine())
331 continue;
332
333 // We just parsed a # character at the start of a line, so we're in
334 // directive mode. Tell the lexer this so any newlines we see will be
335 // converted into an EOD token (this terminates the macro).
336 CurPPLexer->ParsingPreprocessorDirective = true;
337 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
338
339
340 // Read the next token, the directive flavor.
341 LexUnexpandedToken(Tok);
342
343 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
344 // something bogus), skip it.
345 if (Tok.isNot(tok::raw_identifier)) {
346 CurPPLexer->ParsingPreprocessorDirective = false;
347 // Restore comment saving mode.
348 if (CurLexer) CurLexer->resetExtendedTokenMode();
349 continue;
350 }
351
352 // If the first letter isn't i or e, it isn't intesting to us. We know that
353 // this is safe in the face of spelling differences, because there is no way
354 // to spell an i/e in a strange way that is another letter. Skipping this
355 // allows us to avoid looking up the identifier info for #define/#undef and
356 // other common directives.
357 StringRef RI = Tok.getRawIdentifier();
358
359 char FirstChar = RI[0];
360 if (FirstChar >= 'a' && FirstChar <= 'z' &&
361 FirstChar != 'i' && FirstChar != 'e') {
362 CurPPLexer->ParsingPreprocessorDirective = false;
363 // Restore comment saving mode.
364 if (CurLexer) CurLexer->resetExtendedTokenMode();
365 continue;
366 }
367
368 // Get the identifier name without trigraphs or embedded newlines. Note
369 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
370 // when skipping.
371 char DirectiveBuf[20];
372 StringRef Directive;
373 if (!Tok.needsCleaning() && RI.size() < 20) {
374 Directive = RI;
375 } else {
376 std::string DirectiveStr = getSpelling(Tok);
377 unsigned IdLen = DirectiveStr.size();
378 if (IdLen >= 20) {
379 CurPPLexer->ParsingPreprocessorDirective = false;
380 // Restore comment saving mode.
381 if (CurLexer) CurLexer->resetExtendedTokenMode();
382 continue;
383 }
384 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen);
385 Directive = StringRef(DirectiveBuf, IdLen);
386 }
387
388 if (Directive.startswith("if")) {
389 StringRef Sub = Directive.substr(2);
390 if (Sub.empty() || // "if"
391 Sub == "def" || // "ifdef"
392 Sub == "ndef") { // "ifndef"
393 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
394 // bother parsing the condition.
395 DiscardUntilEndOfDirective();
396 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true,
397 /*foundnonskip*/false,
398 /*foundelse*/false);
399 }
400 } else if (Directive[0] == 'e') {
401 StringRef Sub = Directive.substr(1);
402 if (Sub == "ndif") { // "endif"
403 PPConditionalInfo CondInfo;
404 CondInfo.WasSkipping = true; // Silence bogus warning.
405 bool InCond = CurPPLexer->popConditionalLevel(CondInfo);
406 (void)InCond; // Silence warning in no-asserts mode.
407 assert(!InCond && "Can't be skipping if not in a conditional!");
408
409 // If we popped the outermost skipping block, we're done skipping!
410 if (!CondInfo.WasSkipping) {
411 // Restore the value of LexingRawMode so that trailing comments
412 // are handled correctly, if we've reached the outermost block.
413 CurPPLexer->LexingRawMode = false;
414 CheckEndOfDirective("endif");
415 CurPPLexer->LexingRawMode = true;
416 if (Callbacks)
417 Callbacks->Endif(Tok.getLocation(), CondInfo.IfLoc);
418 break;
419 } else {
420 DiscardUntilEndOfDirective();
421 }
422 } else if (Sub == "lse") { // "else".
423 // #else directive in a skipping conditional. If not in some other
424 // skipping conditional, and if #else hasn't already been seen, enter it
425 // as a non-skipping conditional.
426 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
427
428 // If this is a #else with a #else before it, report the error.
429 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else);
430
431 // Note that we've seen a #else in this conditional.
432 CondInfo.FoundElse = true;
433
434 // If the conditional is at the top level, and the #if block wasn't
435 // entered, enter the #else block now.
436 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
437 CondInfo.FoundNonSkip = true;
438 // Restore the value of LexingRawMode so that trailing comments
439 // are handled correctly.
440 CurPPLexer->LexingRawMode = false;
441 CheckEndOfDirective("else");
442 CurPPLexer->LexingRawMode = true;
443 if (Callbacks)
444 Callbacks->Else(Tok.getLocation(), CondInfo.IfLoc);
445 break;
446 } else {
447 DiscardUntilEndOfDirective(); // C99 6.10p4.
448 }
449 } else if (Sub == "lif") { // "elif".
450 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
451
452 // If this is a #elif with a #else before it, report the error.
453 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else);
454
455 // If this is in a skipping block or if we're already handled this #if
456 // block, don't bother parsing the condition.
457 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
458 DiscardUntilEndOfDirective();
459 } else {
460 const SourceLocation CondBegin = CurPPLexer->getSourceLocation();
461 // Restore the value of LexingRawMode so that identifiers are
462 // looked up, etc, inside the #elif expression.
463 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
464 CurPPLexer->LexingRawMode = false;
465 IdentifierInfo *IfNDefMacro = nullptr;
466 const bool CondValue = EvaluateDirectiveExpression(IfNDefMacro);
467 CurPPLexer->LexingRawMode = true;
468 if (Callbacks) {
469 const SourceLocation CondEnd = CurPPLexer->getSourceLocation();
470 Callbacks->Elif(Tok.getLocation(),
471 SourceRange(CondBegin, CondEnd),
472 (CondValue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False), CondInfo.IfLoc);
473 }
474 // If this condition is true, enter it!
475 if (CondValue) {
476 CondInfo.FoundNonSkip = true;
477 break;
478 }
479 }
480 }
481 }
482
483 CurPPLexer->ParsingPreprocessorDirective = false;
484 // Restore comment saving mode.
485 if (CurLexer) CurLexer->resetExtendedTokenMode();
486 }
487
488 // Finally, if we are out of the conditional (saw an #endif or ran off the end
489 // of the file, just stop skipping and return to lexing whatever came after
490 // the #if block.
491 CurPPLexer->LexingRawMode = false;
492
493 if (Callbacks) {
494 SourceLocation BeginLoc = ElseLoc.isValid() ? ElseLoc : IfTokenLoc;
495 Callbacks->SourceRangeSkipped(SourceRange(BeginLoc, Tok.getLocation()));
496 }
497 }
498
PTHSkipExcludedConditionalBlock()499 void Preprocessor::PTHSkipExcludedConditionalBlock() {
500
501 while (1) {
502 assert(CurPTHLexer);
503 assert(CurPTHLexer->LexingRawMode == false);
504
505 // Skip to the next '#else', '#elif', or #endif.
506 if (CurPTHLexer->SkipBlock()) {
507 // We have reached an #endif. Both the '#' and 'endif' tokens
508 // have been consumed by the PTHLexer. Just pop off the condition level.
509 PPConditionalInfo CondInfo;
510 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo);
511 (void)InCond; // Silence warning in no-asserts mode.
512 assert(!InCond && "Can't be skipping if not in a conditional!");
513 break;
514 }
515
516 // We have reached a '#else' or '#elif'. Lex the next token to get
517 // the directive flavor.
518 Token Tok;
519 LexUnexpandedToken(Tok);
520
521 // We can actually look up the IdentifierInfo here since we aren't in
522 // raw mode.
523 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID();
524
525 if (K == tok::pp_else) {
526 // #else: Enter the else condition. We aren't in a nested condition
527 // since we skip those. We're always in the one matching the last
528 // blocked we skipped.
529 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
530 // Note that we've seen a #else in this conditional.
531 CondInfo.FoundElse = true;
532
533 // If the #if block wasn't entered then enter the #else block now.
534 if (!CondInfo.FoundNonSkip) {
535 CondInfo.FoundNonSkip = true;
536
537 // Scan until the eod token.
538 CurPTHLexer->ParsingPreprocessorDirective = true;
539 DiscardUntilEndOfDirective();
540 CurPTHLexer->ParsingPreprocessorDirective = false;
541
542 break;
543 }
544
545 // Otherwise skip this block.
546 continue;
547 }
548
549 assert(K == tok::pp_elif);
550 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
551
552 // If this is a #elif with a #else before it, report the error.
553 if (CondInfo.FoundElse)
554 Diag(Tok, diag::pp_err_elif_after_else);
555
556 // If this is in a skipping block or if we're already handled this #if
557 // block, don't bother parsing the condition. We just skip this block.
558 if (CondInfo.FoundNonSkip)
559 continue;
560
561 // Evaluate the condition of the #elif.
562 IdentifierInfo *IfNDefMacro = nullptr;
563 CurPTHLexer->ParsingPreprocessorDirective = true;
564 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
565 CurPTHLexer->ParsingPreprocessorDirective = false;
566
567 // If this condition is true, enter it!
568 if (ShouldEnter) {
569 CondInfo.FoundNonSkip = true;
570 break;
571 }
572
573 // Otherwise, skip this block and go to the next one.
574 continue;
575 }
576 }
577
getModuleForLocation(SourceLocation Loc)578 Module *Preprocessor::getModuleForLocation(SourceLocation Loc) {
579 ModuleMap &ModMap = HeaderInfo.getModuleMap();
580 if (SourceMgr.isInMainFile(Loc)) {
581 if (Module *CurMod = getCurrentModule())
582 return CurMod; // Compiling a module.
583 return HeaderInfo.getModuleMap().SourceModule; // Compiling a source.
584 }
585 // Try to determine the module of the include directive.
586 // FIXME: Look into directly passing the FileEntry from LookupFile instead.
587 FileID IDOfIncl = SourceMgr.getFileID(SourceMgr.getExpansionLoc(Loc));
588 if (const FileEntry *EntryOfIncl = SourceMgr.getFileEntryForID(IDOfIncl)) {
589 // The include comes from a file.
590 return ModMap.findModuleForHeader(EntryOfIncl).getModule();
591 } else {
592 // The include does not come from a file,
593 // so it is probably a module compilation.
594 return getCurrentModule();
595 }
596 }
597
getModuleContainingLocation(SourceLocation Loc)598 Module *Preprocessor::getModuleContainingLocation(SourceLocation Loc) {
599 return HeaderInfo.getModuleMap().inferModuleFromLocation(
600 FullSourceLoc(Loc, SourceMgr));
601 }
602
LookupFile(SourceLocation FilenameLoc,StringRef Filename,bool isAngled,const DirectoryLookup * FromDir,const FileEntry * FromFile,const DirectoryLookup * & CurDir,SmallVectorImpl<char> * SearchPath,SmallVectorImpl<char> * RelativePath,ModuleMap::KnownHeader * SuggestedModule,bool SkipCache)603 const FileEntry *Preprocessor::LookupFile(
604 SourceLocation FilenameLoc,
605 StringRef Filename,
606 bool isAngled,
607 const DirectoryLookup *FromDir,
608 const FileEntry *FromFile,
609 const DirectoryLookup *&CurDir,
610 SmallVectorImpl<char> *SearchPath,
611 SmallVectorImpl<char> *RelativePath,
612 ModuleMap::KnownHeader *SuggestedModule,
613 bool SkipCache) {
614 // If the header lookup mechanism may be relative to the current inclusion
615 // stack, record the parent #includes.
616 SmallVector<std::pair<const FileEntry *, const DirectoryEntry *>, 16>
617 Includers;
618 if (!FromDir && !FromFile) {
619 FileID FID = getCurrentFileLexer()->getFileID();
620 const FileEntry *FileEnt = SourceMgr.getFileEntryForID(FID);
621
622 // If there is no file entry associated with this file, it must be the
623 // predefines buffer or the module includes buffer. Any other file is not
624 // lexed with a normal lexer, so it won't be scanned for preprocessor
625 // directives.
626 //
627 // If we have the predefines buffer, resolve #include references (which come
628 // from the -include command line argument) from the current working
629 // directory instead of relative to the main file.
630 //
631 // If we have the module includes buffer, resolve #include references (which
632 // come from header declarations in the module map) relative to the module
633 // map file.
634 if (!FileEnt) {
635 if (FID == SourceMgr.getMainFileID() && MainFileDir)
636 Includers.push_back(std::make_pair(nullptr, MainFileDir));
637 else if ((FileEnt =
638 SourceMgr.getFileEntryForID(SourceMgr.getMainFileID())))
639 Includers.push_back(std::make_pair(FileEnt, FileMgr.getDirectory(".")));
640 } else {
641 Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir()));
642 }
643
644 // MSVC searches the current include stack from top to bottom for
645 // headers included by quoted include directives.
646 // See: http://msdn.microsoft.com/en-us/library/36k2cdd4.aspx
647 if (LangOpts.MSVCCompat && !isAngled) {
648 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
649 IncludeStackInfo &ISEntry = IncludeMacroStack[e - i - 1];
650 if (IsFileLexer(ISEntry))
651 if ((FileEnt = SourceMgr.getFileEntryForID(
652 ISEntry.ThePPLexer->getFileID())))
653 Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir()));
654 }
655 }
656 }
657
658 CurDir = CurDirLookup;
659
660 if (FromFile) {
661 // We're supposed to start looking from after a particular file. Search
662 // the include path until we find that file or run out of files.
663 const DirectoryLookup *TmpCurDir = CurDir;
664 const DirectoryLookup *TmpFromDir = nullptr;
665 while (const FileEntry *FE = HeaderInfo.LookupFile(
666 Filename, FilenameLoc, isAngled, TmpFromDir, TmpCurDir,
667 Includers, SearchPath, RelativePath, SuggestedModule,
668 SkipCache)) {
669 // Keep looking as if this file did a #include_next.
670 TmpFromDir = TmpCurDir;
671 ++TmpFromDir;
672 if (FE == FromFile) {
673 // Found it.
674 FromDir = TmpFromDir;
675 CurDir = TmpCurDir;
676 break;
677 }
678 }
679 }
680
681 // Do a standard file entry lookup.
682 const FileEntry *FE = HeaderInfo.LookupFile(
683 Filename, FilenameLoc, isAngled, FromDir, CurDir, Includers, SearchPath,
684 RelativePath, SuggestedModule, SkipCache);
685 if (FE) {
686 if (SuggestedModule && !LangOpts.AsmPreprocessor)
687 HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
688 getModuleForLocation(FilenameLoc), FilenameLoc, Filename, FE);
689 return FE;
690 }
691
692 const FileEntry *CurFileEnt;
693 // Otherwise, see if this is a subframework header. If so, this is relative
694 // to one of the headers on the #include stack. Walk the list of the current
695 // headers on the #include stack and pass them to HeaderInfo.
696 if (IsFileLexer()) {
697 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) {
698 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt,
699 SearchPath, RelativePath,
700 SuggestedModule))) {
701 if (SuggestedModule && !LangOpts.AsmPreprocessor)
702 HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
703 getModuleForLocation(FilenameLoc), FilenameLoc, Filename, FE);
704 return FE;
705 }
706 }
707 }
708
709 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
710 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1];
711 if (IsFileLexer(ISEntry)) {
712 if ((CurFileEnt =
713 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) {
714 if ((FE = HeaderInfo.LookupSubframeworkHeader(
715 Filename, CurFileEnt, SearchPath, RelativePath,
716 SuggestedModule))) {
717 if (SuggestedModule && !LangOpts.AsmPreprocessor)
718 HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
719 getModuleForLocation(FilenameLoc), FilenameLoc, Filename, FE);
720 return FE;
721 }
722 }
723 }
724 }
725
726 // Otherwise, we really couldn't find the file.
727 return nullptr;
728 }
729
730
731 //===----------------------------------------------------------------------===//
732 // Preprocessor Directive Handling.
733 //===----------------------------------------------------------------------===//
734
735 class Preprocessor::ResetMacroExpansionHelper {
736 public:
ResetMacroExpansionHelper(Preprocessor * pp)737 ResetMacroExpansionHelper(Preprocessor *pp)
738 : PP(pp), save(pp->DisableMacroExpansion) {
739 if (pp->MacroExpansionInDirectivesOverride)
740 pp->DisableMacroExpansion = false;
741 }
~ResetMacroExpansionHelper()742 ~ResetMacroExpansionHelper() {
743 PP->DisableMacroExpansion = save;
744 }
745 private:
746 Preprocessor *PP;
747 bool save;
748 };
749
750 /// HandleDirective - This callback is invoked when the lexer sees a # token
751 /// at the start of a line. This consumes the directive, modifies the
752 /// lexer/preprocessor state, and advances the lexer(s) so that the next token
753 /// read is the correct one.
HandleDirective(Token & Result)754 void Preprocessor::HandleDirective(Token &Result) {
755 // FIXME: Traditional: # with whitespace before it not recognized by K&R?
756
757 // We just parsed a # character at the start of a line, so we're in directive
758 // mode. Tell the lexer this so any newlines we see will be converted into an
759 // EOD token (which terminates the directive).
760 CurPPLexer->ParsingPreprocessorDirective = true;
761 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
762
763 bool ImmediatelyAfterTopLevelIfndef =
764 CurPPLexer->MIOpt.getImmediatelyAfterTopLevelIfndef();
765 CurPPLexer->MIOpt.resetImmediatelyAfterTopLevelIfndef();
766
767 ++NumDirectives;
768
769 // We are about to read a token. For the multiple-include optimization FA to
770 // work, we have to remember if we had read any tokens *before* this
771 // pp-directive.
772 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal();
773
774 // Save the '#' token in case we need to return it later.
775 Token SavedHash = Result;
776
777 // Read the next token, the directive flavor. This isn't expanded due to
778 // C99 6.10.3p8.
779 LexUnexpandedToken(Result);
780
781 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.:
782 // #define A(x) #x
783 // A(abc
784 // #warning blah
785 // def)
786 // If so, the user is relying on undefined behavior, emit a diagnostic. Do
787 // not support this for #include-like directives, since that can result in
788 // terrible diagnostics, and does not work in GCC.
789 if (InMacroArgs) {
790 if (IdentifierInfo *II = Result.getIdentifierInfo()) {
791 switch (II->getPPKeywordID()) {
792 case tok::pp_include:
793 case tok::pp_import:
794 case tok::pp_include_next:
795 case tok::pp___include_macros:
796 case tok::pp_pragma:
797 Diag(Result, diag::err_embedded_directive) << II->getName();
798 DiscardUntilEndOfDirective();
799 return;
800 default:
801 break;
802 }
803 }
804 Diag(Result, diag::ext_embedded_directive);
805 }
806
807 // Temporarily enable macro expansion if set so
808 // and reset to previous state when returning from this function.
809 ResetMacroExpansionHelper helper(this);
810
811 switch (Result.getKind()) {
812 case tok::eod:
813 return; // null directive.
814 case tok::code_completion:
815 if (CodeComplete)
816 CodeComplete->CodeCompleteDirective(
817 CurPPLexer->getConditionalStackDepth() > 0);
818 setCodeCompletionReached();
819 return;
820 case tok::numeric_constant: // # 7 GNU line marker directive.
821 if (getLangOpts().AsmPreprocessor)
822 break; // # 4 is not a preprocessor directive in .S files.
823 return HandleDigitDirective(Result);
824 default:
825 IdentifierInfo *II = Result.getIdentifierInfo();
826 if (!II) break; // Not an identifier.
827
828 // Ask what the preprocessor keyword ID is.
829 switch (II->getPPKeywordID()) {
830 default: break;
831 // C99 6.10.1 - Conditional Inclusion.
832 case tok::pp_if:
833 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective);
834 case tok::pp_ifdef:
835 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/);
836 case tok::pp_ifndef:
837 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective);
838 case tok::pp_elif:
839 return HandleElifDirective(Result);
840 case tok::pp_else:
841 return HandleElseDirective(Result);
842 case tok::pp_endif:
843 return HandleEndifDirective(Result);
844
845 // C99 6.10.2 - Source File Inclusion.
846 case tok::pp_include:
847 // Handle #include.
848 return HandleIncludeDirective(SavedHash.getLocation(), Result);
849 case tok::pp___include_macros:
850 // Handle -imacros.
851 return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result);
852
853 // C99 6.10.3 - Macro Replacement.
854 case tok::pp_define:
855 return HandleDefineDirective(Result, ImmediatelyAfterTopLevelIfndef);
856 case tok::pp_undef:
857 return HandleUndefDirective(Result);
858
859 // C99 6.10.4 - Line Control.
860 case tok::pp_line:
861 return HandleLineDirective(Result);
862
863 // C99 6.10.5 - Error Directive.
864 case tok::pp_error:
865 return HandleUserDiagnosticDirective(Result, false);
866
867 // C99 6.10.6 - Pragma Directive.
868 case tok::pp_pragma:
869 return HandlePragmaDirective(SavedHash.getLocation(), PIK_HashPragma);
870
871 // GNU Extensions.
872 case tok::pp_import:
873 return HandleImportDirective(SavedHash.getLocation(), Result);
874 case tok::pp_include_next:
875 return HandleIncludeNextDirective(SavedHash.getLocation(), Result);
876
877 case tok::pp_warning:
878 Diag(Result, diag::ext_pp_warning_directive);
879 return HandleUserDiagnosticDirective(Result, true);
880 case tok::pp_ident:
881 return HandleIdentSCCSDirective(Result);
882 case tok::pp_sccs:
883 return HandleIdentSCCSDirective(Result);
884 case tok::pp_assert:
885 //isExtension = true; // FIXME: implement #assert
886 break;
887 case tok::pp_unassert:
888 //isExtension = true; // FIXME: implement #unassert
889 break;
890
891 case tok::pp___public_macro:
892 if (getLangOpts().Modules)
893 return HandleMacroPublicDirective(Result);
894 break;
895
896 case tok::pp___private_macro:
897 if (getLangOpts().Modules)
898 return HandleMacroPrivateDirective(Result);
899 break;
900 }
901 break;
902 }
903
904 // If this is a .S file, treat unknown # directives as non-preprocessor
905 // directives. This is important because # may be a comment or introduce
906 // various pseudo-ops. Just return the # token and push back the following
907 // token to be lexed next time.
908 if (getLangOpts().AsmPreprocessor) {
909 Token *Toks = new Token[2];
910 // Return the # and the token after it.
911 Toks[0] = SavedHash;
912 Toks[1] = Result;
913
914 // If the second token is a hashhash token, then we need to translate it to
915 // unknown so the token lexer doesn't try to perform token pasting.
916 if (Result.is(tok::hashhash))
917 Toks[1].setKind(tok::unknown);
918
919 // Enter this token stream so that we re-lex the tokens. Make sure to
920 // enable macro expansion, in case the token after the # is an identifier
921 // that is expanded.
922 EnterTokenStream(Toks, 2, false, true);
923 return;
924 }
925
926 // If we reached here, the preprocessing token is not valid!
927 Diag(Result, diag::err_pp_invalid_directive);
928
929 // Read the rest of the PP line.
930 DiscardUntilEndOfDirective();
931
932 // Okay, we're done parsing the directive.
933 }
934
935 /// GetLineValue - Convert a numeric token into an unsigned value, emitting
936 /// Diagnostic DiagID if it is invalid, and returning the value in Val.
GetLineValue(Token & DigitTok,unsigned & Val,unsigned DiagID,Preprocessor & PP,bool IsGNULineDirective=false)937 static bool GetLineValue(Token &DigitTok, unsigned &Val,
938 unsigned DiagID, Preprocessor &PP,
939 bool IsGNULineDirective=false) {
940 if (DigitTok.isNot(tok::numeric_constant)) {
941 PP.Diag(DigitTok, DiagID);
942
943 if (DigitTok.isNot(tok::eod))
944 PP.DiscardUntilEndOfDirective();
945 return true;
946 }
947
948 SmallString<64> IntegerBuffer;
949 IntegerBuffer.resize(DigitTok.getLength());
950 const char *DigitTokBegin = &IntegerBuffer[0];
951 bool Invalid = false;
952 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid);
953 if (Invalid)
954 return true;
955
956 // Verify that we have a simple digit-sequence, and compute the value. This
957 // is always a simple digit string computed in decimal, so we do this manually
958 // here.
959 Val = 0;
960 for (unsigned i = 0; i != ActualLength; ++i) {
961 // C++1y [lex.fcon]p1:
962 // Optional separating single quotes in a digit-sequence are ignored
963 if (DigitTokBegin[i] == '\'')
964 continue;
965
966 if (!isDigit(DigitTokBegin[i])) {
967 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i),
968 diag::err_pp_line_digit_sequence) << IsGNULineDirective;
969 PP.DiscardUntilEndOfDirective();
970 return true;
971 }
972
973 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0');
974 if (NextVal < Val) { // overflow.
975 PP.Diag(DigitTok, DiagID);
976 PP.DiscardUntilEndOfDirective();
977 return true;
978 }
979 Val = NextVal;
980 }
981
982 if (DigitTokBegin[0] == '0' && Val)
983 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal)
984 << IsGNULineDirective;
985
986 return false;
987 }
988
989 /// \brief Handle a \#line directive: C99 6.10.4.
990 ///
991 /// The two acceptable forms are:
992 /// \verbatim
993 /// # line digit-sequence
994 /// # line digit-sequence "s-char-sequence"
995 /// \endverbatim
HandleLineDirective(Token & Tok)996 void Preprocessor::HandleLineDirective(Token &Tok) {
997 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are
998 // expanded.
999 Token DigitTok;
1000 Lex(DigitTok);
1001
1002 // Validate the number and convert it to an unsigned.
1003 unsigned LineNo;
1004 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this))
1005 return;
1006
1007 if (LineNo == 0)
1008 Diag(DigitTok, diag::ext_pp_line_zero);
1009
1010 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
1011 // number greater than 2147483647". C90 requires that the line # be <= 32767.
1012 unsigned LineLimit = 32768U;
1013 if (LangOpts.C99 || LangOpts.CPlusPlus11)
1014 LineLimit = 2147483648U;
1015 if (LineNo >= LineLimit)
1016 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit;
1017 else if (LangOpts.CPlusPlus11 && LineNo >= 32768U)
1018 Diag(DigitTok, diag::warn_cxx98_compat_pp_line_too_big);
1019
1020 int FilenameID = -1;
1021 Token StrTok;
1022 Lex(StrTok);
1023
1024 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a
1025 // string followed by eod.
1026 if (StrTok.is(tok::eod))
1027 ; // ok
1028 else if (StrTok.isNot(tok::string_literal)) {
1029 Diag(StrTok, diag::err_pp_line_invalid_filename);
1030 return DiscardUntilEndOfDirective();
1031 } else if (StrTok.hasUDSuffix()) {
1032 Diag(StrTok, diag::err_invalid_string_udl);
1033 return DiscardUntilEndOfDirective();
1034 } else {
1035 // Parse and validate the string, converting it into a unique ID.
1036 StringLiteralParser Literal(StrTok, *this);
1037 assert(Literal.isAscii() && "Didn't allow wide strings in");
1038 if (Literal.hadError)
1039 return DiscardUntilEndOfDirective();
1040 if (Literal.Pascal) {
1041 Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1042 return DiscardUntilEndOfDirective();
1043 }
1044 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1045
1046 // Verify that there is nothing after the string, other than EOD. Because
1047 // of C99 6.10.4p5, macros that expand to empty tokens are ok.
1048 CheckEndOfDirective("line", true);
1049 }
1050
1051 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID);
1052
1053 if (Callbacks)
1054 Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
1055 PPCallbacks::RenameFile,
1056 SrcMgr::C_User);
1057 }
1058
1059 /// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
1060 /// marker directive.
ReadLineMarkerFlags(bool & IsFileEntry,bool & IsFileExit,bool & IsSystemHeader,bool & IsExternCHeader,Preprocessor & PP)1061 static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
1062 bool &IsSystemHeader, bool &IsExternCHeader,
1063 Preprocessor &PP) {
1064 unsigned FlagVal;
1065 Token FlagTok;
1066 PP.Lex(FlagTok);
1067 if (FlagTok.is(tok::eod)) return false;
1068 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
1069 return true;
1070
1071 if (FlagVal == 1) {
1072 IsFileEntry = true;
1073
1074 PP.Lex(FlagTok);
1075 if (FlagTok.is(tok::eod)) return false;
1076 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
1077 return true;
1078 } else if (FlagVal == 2) {
1079 IsFileExit = true;
1080
1081 SourceManager &SM = PP.getSourceManager();
1082 // If we are leaving the current presumed file, check to make sure the
1083 // presumed include stack isn't empty!
1084 FileID CurFileID =
1085 SM.getDecomposedExpansionLoc(FlagTok.getLocation()).first;
1086 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation());
1087 if (PLoc.isInvalid())
1088 return true;
1089
1090 // If there is no include loc (main file) or if the include loc is in a
1091 // different physical file, then we aren't in a "1" line marker flag region.
1092 SourceLocation IncLoc = PLoc.getIncludeLoc();
1093 if (IncLoc.isInvalid() ||
1094 SM.getDecomposedExpansionLoc(IncLoc).first != CurFileID) {
1095 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop);
1096 PP.DiscardUntilEndOfDirective();
1097 return true;
1098 }
1099
1100 PP.Lex(FlagTok);
1101 if (FlagTok.is(tok::eod)) return false;
1102 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
1103 return true;
1104 }
1105
1106 // We must have 3 if there are still flags.
1107 if (FlagVal != 3) {
1108 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1109 PP.DiscardUntilEndOfDirective();
1110 return true;
1111 }
1112
1113 IsSystemHeader = true;
1114
1115 PP.Lex(FlagTok);
1116 if (FlagTok.is(tok::eod)) return false;
1117 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
1118 return true;
1119
1120 // We must have 4 if there is yet another flag.
1121 if (FlagVal != 4) {
1122 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1123 PP.DiscardUntilEndOfDirective();
1124 return true;
1125 }
1126
1127 IsExternCHeader = true;
1128
1129 PP.Lex(FlagTok);
1130 if (FlagTok.is(tok::eod)) return false;
1131
1132 // There are no more valid flags here.
1133 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1134 PP.DiscardUntilEndOfDirective();
1135 return true;
1136 }
1137
1138 /// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
1139 /// one of the following forms:
1140 ///
1141 /// # 42
1142 /// # 42 "file" ('1' | '2')?
1143 /// # 42 "file" ('1' | '2')? '3' '4'?
1144 ///
HandleDigitDirective(Token & DigitTok)1145 void Preprocessor::HandleDigitDirective(Token &DigitTok) {
1146 // Validate the number and convert it to an unsigned. GNU does not have a
1147 // line # limit other than it fit in 32-bits.
1148 unsigned LineNo;
1149 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer,
1150 *this, true))
1151 return;
1152
1153 Token StrTok;
1154 Lex(StrTok);
1155
1156 bool IsFileEntry = false, IsFileExit = false;
1157 bool IsSystemHeader = false, IsExternCHeader = false;
1158 int FilenameID = -1;
1159
1160 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a
1161 // string followed by eod.
1162 if (StrTok.is(tok::eod))
1163 ; // ok
1164 else if (StrTok.isNot(tok::string_literal)) {
1165 Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1166 return DiscardUntilEndOfDirective();
1167 } else if (StrTok.hasUDSuffix()) {
1168 Diag(StrTok, diag::err_invalid_string_udl);
1169 return DiscardUntilEndOfDirective();
1170 } else {
1171 // Parse and validate the string, converting it into a unique ID.
1172 StringLiteralParser Literal(StrTok, *this);
1173 assert(Literal.isAscii() && "Didn't allow wide strings in");
1174 if (Literal.hadError)
1175 return DiscardUntilEndOfDirective();
1176 if (Literal.Pascal) {
1177 Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1178 return DiscardUntilEndOfDirective();
1179 }
1180 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1181
1182 // If a filename was present, read any flags that are present.
1183 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit,
1184 IsSystemHeader, IsExternCHeader, *this))
1185 return;
1186 }
1187
1188 // Create a line note with this information.
1189 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID,
1190 IsFileEntry, IsFileExit,
1191 IsSystemHeader, IsExternCHeader);
1192
1193 // If the preprocessor has callbacks installed, notify them of the #line
1194 // change. This is used so that the line marker comes out in -E mode for
1195 // example.
1196 if (Callbacks) {
1197 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile;
1198 if (IsFileEntry)
1199 Reason = PPCallbacks::EnterFile;
1200 else if (IsFileExit)
1201 Reason = PPCallbacks::ExitFile;
1202 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User;
1203 if (IsExternCHeader)
1204 FileKind = SrcMgr::C_ExternCSystem;
1205 else if (IsSystemHeader)
1206 FileKind = SrcMgr::C_System;
1207
1208 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind);
1209 }
1210 }
1211
1212
1213 /// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
1214 ///
HandleUserDiagnosticDirective(Token & Tok,bool isWarning)1215 void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
1216 bool isWarning) {
1217 // PTH doesn't emit #warning or #error directives.
1218 if (CurPTHLexer)
1219 return CurPTHLexer->DiscardToEndOfLine();
1220
1221 // Read the rest of the line raw. We do this because we don't want macros
1222 // to be expanded and we don't require that the tokens be valid preprocessing
1223 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does
1224 // collapse multiple consequtive white space between tokens, but this isn't
1225 // specified by the standard.
1226 SmallString<128> Message;
1227 CurLexer->ReadToEndOfLine(&Message);
1228
1229 // Find the first non-whitespace character, so that we can make the
1230 // diagnostic more succinct.
1231 StringRef Msg = StringRef(Message).ltrim(" ");
1232
1233 if (isWarning)
1234 Diag(Tok, diag::pp_hash_warning) << Msg;
1235 else
1236 Diag(Tok, diag::err_pp_hash_error) << Msg;
1237 }
1238
1239 /// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
1240 ///
HandleIdentSCCSDirective(Token & Tok)1241 void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
1242 // Yes, this directive is an extension.
1243 Diag(Tok, diag::ext_pp_ident_directive);
1244
1245 // Read the string argument.
1246 Token StrTok;
1247 Lex(StrTok);
1248
1249 // If the token kind isn't a string, it's a malformed directive.
1250 if (StrTok.isNot(tok::string_literal) &&
1251 StrTok.isNot(tok::wide_string_literal)) {
1252 Diag(StrTok, diag::err_pp_malformed_ident);
1253 if (StrTok.isNot(tok::eod))
1254 DiscardUntilEndOfDirective();
1255 return;
1256 }
1257
1258 if (StrTok.hasUDSuffix()) {
1259 Diag(StrTok, diag::err_invalid_string_udl);
1260 return DiscardUntilEndOfDirective();
1261 }
1262
1263 // Verify that there is nothing after the string, other than EOD.
1264 CheckEndOfDirective("ident");
1265
1266 if (Callbacks) {
1267 bool Invalid = false;
1268 std::string Str = getSpelling(StrTok, &Invalid);
1269 if (!Invalid)
1270 Callbacks->Ident(Tok.getLocation(), Str);
1271 }
1272 }
1273
1274 /// \brief Handle a #public directive.
HandleMacroPublicDirective(Token & Tok)1275 void Preprocessor::HandleMacroPublicDirective(Token &Tok) {
1276 Token MacroNameTok;
1277 ReadMacroName(MacroNameTok, MU_Undef);
1278
1279 // Error reading macro name? If so, diagnostic already issued.
1280 if (MacroNameTok.is(tok::eod))
1281 return;
1282
1283 // Check to see if this is the last token on the #__public_macro line.
1284 CheckEndOfDirective("__public_macro");
1285
1286 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1287 // Okay, we finally have a valid identifier to undef.
1288 MacroDirective *MD = getLocalMacroDirective(II);
1289
1290 // If the macro is not defined, this is an error.
1291 if (!MD) {
1292 Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1293 return;
1294 }
1295
1296 // Note that this macro has now been exported.
1297 appendMacroDirective(II, AllocateVisibilityMacroDirective(
1298 MacroNameTok.getLocation(), /*IsPublic=*/true));
1299 }
1300
1301 /// \brief Handle a #private directive.
HandleMacroPrivateDirective(Token & Tok)1302 void Preprocessor::HandleMacroPrivateDirective(Token &Tok) {
1303 Token MacroNameTok;
1304 ReadMacroName(MacroNameTok, MU_Undef);
1305
1306 // Error reading macro name? If so, diagnostic already issued.
1307 if (MacroNameTok.is(tok::eod))
1308 return;
1309
1310 // Check to see if this is the last token on the #__private_macro line.
1311 CheckEndOfDirective("__private_macro");
1312
1313 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1314 // Okay, we finally have a valid identifier to undef.
1315 MacroDirective *MD = getLocalMacroDirective(II);
1316
1317 // If the macro is not defined, this is an error.
1318 if (!MD) {
1319 Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1320 return;
1321 }
1322
1323 // Note that this macro has now been marked private.
1324 appendMacroDirective(II, AllocateVisibilityMacroDirective(
1325 MacroNameTok.getLocation(), /*IsPublic=*/false));
1326 }
1327
1328 //===----------------------------------------------------------------------===//
1329 // Preprocessor Include Directive Handling.
1330 //===----------------------------------------------------------------------===//
1331
1332 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
1333 /// checked and spelled filename, e.g. as an operand of \#include. This returns
1334 /// true if the input filename was in <>'s or false if it were in ""'s. The
1335 /// caller is expected to provide a buffer that is large enough to hold the
1336 /// spelling of the filename, but is also expected to handle the case when
1337 /// this method decides to use a different buffer.
GetIncludeFilenameSpelling(SourceLocation Loc,StringRef & Buffer)1338 bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
1339 StringRef &Buffer) {
1340 // Get the text form of the filename.
1341 assert(!Buffer.empty() && "Can't have tokens with empty spellings!");
1342
1343 // Make sure the filename is <x> or "x".
1344 bool isAngled;
1345 if (Buffer[0] == '<') {
1346 if (Buffer.back() != '>') {
1347 Diag(Loc, diag::err_pp_expects_filename);
1348 Buffer = StringRef();
1349 return true;
1350 }
1351 isAngled = true;
1352 } else if (Buffer[0] == '"') {
1353 if (Buffer.back() != '"') {
1354 Diag(Loc, diag::err_pp_expects_filename);
1355 Buffer = StringRef();
1356 return true;
1357 }
1358 isAngled = false;
1359 } else {
1360 Diag(Loc, diag::err_pp_expects_filename);
1361 Buffer = StringRef();
1362 return true;
1363 }
1364
1365 // Diagnose #include "" as invalid.
1366 if (Buffer.size() <= 2) {
1367 Diag(Loc, diag::err_pp_empty_filename);
1368 Buffer = StringRef();
1369 return true;
1370 }
1371
1372 // Skip the brackets.
1373 Buffer = Buffer.substr(1, Buffer.size()-2);
1374 return isAngled;
1375 }
1376
1377 // \brief Handle cases where the \#include name is expanded from a macro
1378 // as multiple tokens, which need to be glued together.
1379 //
1380 // This occurs for code like:
1381 // \code
1382 // \#define FOO <a/b.h>
1383 // \#include FOO
1384 // \endcode
1385 // because in this case, "<a/b.h>" is returned as 7 tokens, not one.
1386 //
1387 // This code concatenates and consumes tokens up to the '>' token. It returns
1388 // false if the > was found, otherwise it returns true if it finds and consumes
1389 // the EOD marker.
ConcatenateIncludeName(SmallString<128> & FilenameBuffer,SourceLocation & End)1390 bool Preprocessor::ConcatenateIncludeName(SmallString<128> &FilenameBuffer,
1391 SourceLocation &End) {
1392 Token CurTok;
1393
1394 Lex(CurTok);
1395 while (CurTok.isNot(tok::eod)) {
1396 End = CurTok.getLocation();
1397
1398 // FIXME: Provide code completion for #includes.
1399 if (CurTok.is(tok::code_completion)) {
1400 setCodeCompletionReached();
1401 Lex(CurTok);
1402 continue;
1403 }
1404
1405 // Append the spelling of this token to the buffer. If there was a space
1406 // before it, add it now.
1407 if (CurTok.hasLeadingSpace())
1408 FilenameBuffer.push_back(' ');
1409
1410 // Get the spelling of the token, directly into FilenameBuffer if possible.
1411 unsigned PreAppendSize = FilenameBuffer.size();
1412 FilenameBuffer.resize(PreAppendSize+CurTok.getLength());
1413
1414 const char *BufPtr = &FilenameBuffer[PreAppendSize];
1415 unsigned ActualLen = getSpelling(CurTok, BufPtr);
1416
1417 // If the token was spelled somewhere else, copy it into FilenameBuffer.
1418 if (BufPtr != &FilenameBuffer[PreAppendSize])
1419 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1420
1421 // Resize FilenameBuffer to the correct size.
1422 if (CurTok.getLength() != ActualLen)
1423 FilenameBuffer.resize(PreAppendSize+ActualLen);
1424
1425 // If we found the '>' marker, return success.
1426 if (CurTok.is(tok::greater))
1427 return false;
1428
1429 Lex(CurTok);
1430 }
1431
1432 // If we hit the eod marker, emit an error and return true so that the caller
1433 // knows the EOD has been read.
1434 Diag(CurTok.getLocation(), diag::err_pp_expects_filename);
1435 return true;
1436 }
1437
1438 /// \brief Push a token onto the token stream containing an annotation.
EnterAnnotationToken(Preprocessor & PP,SourceLocation Begin,SourceLocation End,tok::TokenKind Kind,void * AnnotationVal)1439 static void EnterAnnotationToken(Preprocessor &PP,
1440 SourceLocation Begin, SourceLocation End,
1441 tok::TokenKind Kind, void *AnnotationVal) {
1442 // FIXME: Produce this as the current token directly, rather than
1443 // allocating a new token for it.
1444 Token *Tok = new Token[1];
1445 Tok[0].startToken();
1446 Tok[0].setKind(Kind);
1447 Tok[0].setLocation(Begin);
1448 Tok[0].setAnnotationEndLoc(End);
1449 Tok[0].setAnnotationValue(AnnotationVal);
1450 PP.EnterTokenStream(Tok, 1, true, true);
1451 }
1452
1453 /// \brief Produce a diagnostic informing the user that a #include or similar
1454 /// was implicitly treated as a module import.
diagnoseAutoModuleImport(Preprocessor & PP,SourceLocation HashLoc,Token & IncludeTok,ArrayRef<std::pair<IdentifierInfo *,SourceLocation>> Path,SourceLocation PathEnd)1455 static void diagnoseAutoModuleImport(
1456 Preprocessor &PP, SourceLocation HashLoc, Token &IncludeTok,
1457 ArrayRef<std::pair<IdentifierInfo *, SourceLocation>> Path,
1458 SourceLocation PathEnd) {
1459 assert(PP.getLangOpts().ObjC2 && "no import syntax available");
1460
1461 SmallString<128> PathString;
1462 for (unsigned I = 0, N = Path.size(); I != N; ++I) {
1463 if (I)
1464 PathString += '.';
1465 PathString += Path[I].first->getName();
1466 }
1467 int IncludeKind = 0;
1468
1469 switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) {
1470 case tok::pp_include:
1471 IncludeKind = 0;
1472 break;
1473
1474 case tok::pp_import:
1475 IncludeKind = 1;
1476 break;
1477
1478 case tok::pp_include_next:
1479 IncludeKind = 2;
1480 break;
1481
1482 case tok::pp___include_macros:
1483 IncludeKind = 3;
1484 break;
1485
1486 default:
1487 llvm_unreachable("unknown include directive kind");
1488 }
1489
1490 CharSourceRange ReplaceRange(SourceRange(HashLoc, PathEnd),
1491 /*IsTokenRange=*/false);
1492 PP.Diag(HashLoc, diag::warn_auto_module_import)
1493 << IncludeKind << PathString
1494 << FixItHint::CreateReplacement(ReplaceRange,
1495 ("@import " + PathString + ";").str());
1496 }
1497
1498 /// HandleIncludeDirective - The "\#include" tokens have just been read, read
1499 /// the file to be included from the lexer, then include it! This is a common
1500 /// routine with functionality shared between \#include, \#include_next and
1501 /// \#import. LookupFrom is set when this is a \#include_next directive, it
1502 /// specifies the file to start searching from.
HandleIncludeDirective(SourceLocation HashLoc,Token & IncludeTok,const DirectoryLookup * LookupFrom,const FileEntry * LookupFromFile,bool isImport)1503 void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc,
1504 Token &IncludeTok,
1505 const DirectoryLookup *LookupFrom,
1506 const FileEntry *LookupFromFile,
1507 bool isImport) {
1508
1509 Token FilenameTok;
1510 CurPPLexer->LexIncludeFilename(FilenameTok);
1511
1512 // Reserve a buffer to get the spelling.
1513 SmallString<128> FilenameBuffer;
1514 StringRef Filename;
1515 SourceLocation End;
1516 SourceLocation CharEnd; // the end of this directive, in characters
1517
1518 switch (FilenameTok.getKind()) {
1519 case tok::eod:
1520 // If the token kind is EOD, the error has already been diagnosed.
1521 return;
1522
1523 case tok::angle_string_literal:
1524 case tok::string_literal:
1525 Filename = getSpelling(FilenameTok, FilenameBuffer);
1526 End = FilenameTok.getLocation();
1527 CharEnd = End.getLocWithOffset(FilenameTok.getLength());
1528 break;
1529
1530 case tok::less:
1531 // This could be a <foo/bar.h> file coming from a macro expansion. In this
1532 // case, glue the tokens together into FilenameBuffer and interpret those.
1533 FilenameBuffer.push_back('<');
1534 if (ConcatenateIncludeName(FilenameBuffer, End))
1535 return; // Found <eod> but no ">"? Diagnostic already emitted.
1536 Filename = FilenameBuffer;
1537 CharEnd = End.getLocWithOffset(1);
1538 break;
1539 default:
1540 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
1541 DiscardUntilEndOfDirective();
1542 return;
1543 }
1544
1545 CharSourceRange FilenameRange
1546 = CharSourceRange::getCharRange(FilenameTok.getLocation(), CharEnd);
1547 StringRef OriginalFilename = Filename;
1548 bool isAngled =
1549 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename);
1550 // If GetIncludeFilenameSpelling set the start ptr to null, there was an
1551 // error.
1552 if (Filename.empty()) {
1553 DiscardUntilEndOfDirective();
1554 return;
1555 }
1556
1557 // Verify that there is nothing after the filename, other than EOD. Note that
1558 // we allow macros that expand to nothing after the filename, because this
1559 // falls into the category of "#include pp-tokens new-line" specified in
1560 // C99 6.10.2p4.
1561 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true);
1562
1563 // Check that we don't have infinite #include recursion.
1564 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
1565 Diag(FilenameTok, diag::err_pp_include_too_deep);
1566 return;
1567 }
1568
1569 // Complain about attempts to #include files in an audit pragma.
1570 if (PragmaARCCFCodeAuditedLoc.isValid()) {
1571 Diag(HashLoc, diag::err_pp_include_in_arc_cf_code_audited);
1572 Diag(PragmaARCCFCodeAuditedLoc, diag::note_pragma_entered_here);
1573
1574 // Immediately leave the pragma.
1575 PragmaARCCFCodeAuditedLoc = SourceLocation();
1576 }
1577
1578 // Complain about attempts to #include files in an assume-nonnull pragma.
1579 if (PragmaAssumeNonNullLoc.isValid()) {
1580 Diag(HashLoc, diag::err_pp_include_in_assume_nonnull);
1581 Diag(PragmaAssumeNonNullLoc, diag::note_pragma_entered_here);
1582
1583 // Immediately leave the pragma.
1584 PragmaAssumeNonNullLoc = SourceLocation();
1585 }
1586
1587 if (HeaderInfo.HasIncludeAliasMap()) {
1588 // Map the filename with the brackets still attached. If the name doesn't
1589 // map to anything, fall back on the filename we've already gotten the
1590 // spelling for.
1591 StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(OriginalFilename);
1592 if (!NewName.empty())
1593 Filename = NewName;
1594 }
1595
1596 // Search include directories.
1597 const DirectoryLookup *CurDir;
1598 SmallString<1024> SearchPath;
1599 SmallString<1024> RelativePath;
1600 // We get the raw path only if we have 'Callbacks' to which we later pass
1601 // the path.
1602 ModuleMap::KnownHeader SuggestedModule;
1603 SourceLocation FilenameLoc = FilenameTok.getLocation();
1604 SmallString<128> NormalizedPath;
1605 if (LangOpts.MSVCCompat) {
1606 NormalizedPath = Filename.str();
1607 #ifndef LLVM_ON_WIN32
1608 llvm::sys::path::native(NormalizedPath);
1609 #endif
1610 }
1611 const FileEntry *File = LookupFile(
1612 FilenameLoc, LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename,
1613 isAngled, LookupFrom, LookupFromFile, CurDir,
1614 Callbacks ? &SearchPath : nullptr, Callbacks ? &RelativePath : nullptr,
1615 &SuggestedModule);
1616
1617 if (!File) {
1618 if (Callbacks) {
1619 // Give the clients a chance to recover.
1620 SmallString<128> RecoveryPath;
1621 if (Callbacks->FileNotFound(Filename, RecoveryPath)) {
1622 if (const DirectoryEntry *DE = FileMgr.getDirectory(RecoveryPath)) {
1623 // Add the recovery path to the list of search paths.
1624 DirectoryLookup DL(DE, SrcMgr::C_User, false);
1625 HeaderInfo.AddSearchPath(DL, isAngled);
1626
1627 // Try the lookup again, skipping the cache.
1628 File = LookupFile(
1629 FilenameLoc,
1630 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, isAngled,
1631 LookupFrom, LookupFromFile, CurDir, nullptr, nullptr,
1632 &SuggestedModule, /*SkipCache*/ true);
1633 }
1634 }
1635 }
1636
1637 if (!SuppressIncludeNotFoundError) {
1638 // If the file could not be located and it was included via angle
1639 // brackets, we can attempt a lookup as though it were a quoted path to
1640 // provide the user with a possible fixit.
1641 if (isAngled) {
1642 File = LookupFile(
1643 FilenameLoc,
1644 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, false,
1645 LookupFrom, LookupFromFile, CurDir,
1646 Callbacks ? &SearchPath : nullptr,
1647 Callbacks ? &RelativePath : nullptr,
1648 &SuggestedModule);
1649 if (File) {
1650 SourceRange Range(FilenameTok.getLocation(), CharEnd);
1651 Diag(FilenameTok, diag::err_pp_file_not_found_not_fatal) <<
1652 Filename <<
1653 FixItHint::CreateReplacement(Range, "\"" + Filename.str() + "\"");
1654 }
1655 }
1656
1657 // If the file is still not found, just go with the vanilla diagnostic
1658 if (!File)
1659 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename;
1660 }
1661 }
1662
1663 // Should we enter the source file? Set to false if either the source file is
1664 // known to have no effect beyond its effect on module visibility -- that is,
1665 // if it's got an include guard that is already defined or is a modular header
1666 // we've imported or already built.
1667 bool ShouldEnter = true;
1668
1669 // Determine whether we should try to import the module for this #include, if
1670 // there is one. Don't do so if precompiled module support is disabled or we
1671 // are processing this module textually (because we're building the module).
1672 if (File && SuggestedModule && getLangOpts().Modules &&
1673 SuggestedModule.getModule()->getTopLevelModuleName() !=
1674 getLangOpts().CurrentModule &&
1675 SuggestedModule.getModule()->getTopLevelModuleName() !=
1676 getLangOpts().ImplementationOfModule) {
1677 // Compute the module access path corresponding to this module.
1678 // FIXME: Should we have a second loadModule() overload to avoid this
1679 // extra lookup step?
1680 SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> Path;
1681 for (Module *Mod = SuggestedModule.getModule(); Mod; Mod = Mod->Parent)
1682 Path.push_back(std::make_pair(getIdentifierInfo(Mod->Name),
1683 FilenameTok.getLocation()));
1684 std::reverse(Path.begin(), Path.end());
1685
1686 // Warn that we're replacing the include/import with a module import.
1687 // We only do this in Objective-C, where we have a module-import syntax.
1688 if (getLangOpts().ObjC2)
1689 diagnoseAutoModuleImport(*this, HashLoc, IncludeTok, Path, CharEnd);
1690
1691 // Load the module to import its macros. We'll make the declarations
1692 // visible when the parser gets here.
1693 // FIXME: Pass SuggestedModule in here rather than converting it to a path
1694 // and making the module loader convert it back again.
1695 ModuleLoadResult Imported = TheModuleLoader.loadModule(
1696 IncludeTok.getLocation(), Path, Module::Hidden,
1697 /*IsIncludeDirective=*/true);
1698 assert((Imported == nullptr || Imported == SuggestedModule.getModule()) &&
1699 "the imported module is different than the suggested one");
1700
1701 if (Imported)
1702 ShouldEnter = false;
1703 else if (Imported.isMissingExpected()) {
1704 // We failed to find a submodule that we assumed would exist (because it
1705 // was in the directory of an umbrella header, for instance), but no
1706 // actual module exists for it (because the umbrella header is
1707 // incomplete). Treat this as a textual inclusion.
1708 SuggestedModule = ModuleMap::KnownHeader();
1709 } else {
1710 // We hit an error processing the import. Bail out.
1711 if (hadModuleLoaderFatalFailure()) {
1712 // With a fatal failure in the module loader, we abort parsing.
1713 Token &Result = IncludeTok;
1714 if (CurLexer) {
1715 Result.startToken();
1716 CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof);
1717 CurLexer->cutOffLexing();
1718 } else {
1719 assert(CurPTHLexer && "#include but no current lexer set!");
1720 CurPTHLexer->getEOF(Result);
1721 }
1722 }
1723 return;
1724 }
1725 }
1726
1727 if (Callbacks) {
1728 // Notify the callback object that we've seen an inclusion directive.
1729 Callbacks->InclusionDirective(
1730 HashLoc, IncludeTok,
1731 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, isAngled,
1732 FilenameRange, File, SearchPath, RelativePath,
1733 ShouldEnter ? nullptr : SuggestedModule.getModule());
1734 }
1735
1736 if (!File)
1737 return;
1738
1739 // The #included file will be considered to be a system header if either it is
1740 // in a system include directory, or if the #includer is a system include
1741 // header.
1742 SrcMgr::CharacteristicKind FileCharacter =
1743 std::max(HeaderInfo.getFileDirFlavor(File),
1744 SourceMgr.getFileCharacteristic(FilenameTok.getLocation()));
1745
1746 // FIXME: If we have a suggested module, and we've already visited this file,
1747 // don't bother entering it again. We know it has no further effect.
1748
1749 // Ask HeaderInfo if we should enter this #include file. If not, #including
1750 // this file will have no effect.
1751 if (ShouldEnter &&
1752 !HeaderInfo.ShouldEnterIncludeFile(*this, File, isImport,
1753 SuggestedModule.getModule())) {
1754 ShouldEnter = false;
1755 if (Callbacks)
1756 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter);
1757 }
1758
1759 // If we don't need to enter the file, stop now.
1760 if (!ShouldEnter) {
1761 // If this is a module import, make it visible if needed.
1762 if (auto *M = SuggestedModule.getModule()) {
1763 makeModuleVisible(M, HashLoc);
1764
1765 if (IncludeTok.getIdentifierInfo()->getPPKeywordID() !=
1766 tok::pp___include_macros)
1767 EnterAnnotationToken(*this, HashLoc, End, tok::annot_module_include, M);
1768 }
1769 return;
1770 }
1771
1772 // Look up the file, create a File ID for it.
1773 SourceLocation IncludePos = End;
1774 // If the filename string was the result of macro expansions, set the include
1775 // position on the file where it will be included and after the expansions.
1776 if (IncludePos.isMacroID())
1777 IncludePos = SourceMgr.getExpansionRange(IncludePos).second;
1778 FileID FID = SourceMgr.createFileID(File, IncludePos, FileCharacter);
1779 assert(!FID.isInvalid() && "Expected valid file ID");
1780
1781 // If all is good, enter the new file!
1782 if (EnterSourceFile(FID, CurDir, FilenameTok.getLocation()))
1783 return;
1784
1785 // Determine if we're switching to building a new submodule, and which one.
1786 if (auto *M = SuggestedModule.getModule()) {
1787 assert(!CurSubmodule && "should not have marked this as a module yet");
1788 CurSubmodule = M;
1789
1790 // Let the macro handling code know that any future macros are within
1791 // the new submodule.
1792 EnterSubmodule(M, HashLoc);
1793
1794 // Let the parser know that any future declarations are within the new
1795 // submodule.
1796 // FIXME: There's no point doing this if we're handling a #__include_macros
1797 // directive.
1798 EnterAnnotationToken(*this, HashLoc, End, tok::annot_module_begin, M);
1799 }
1800 }
1801
1802 /// HandleIncludeNextDirective - Implements \#include_next.
1803 ///
HandleIncludeNextDirective(SourceLocation HashLoc,Token & IncludeNextTok)1804 void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc,
1805 Token &IncludeNextTok) {
1806 Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
1807
1808 // #include_next is like #include, except that we start searching after
1809 // the current found directory. If we can't do this, issue a
1810 // diagnostic.
1811 const DirectoryLookup *Lookup = CurDirLookup;
1812 const FileEntry *LookupFromFile = nullptr;
1813 if (isInPrimaryFile()) {
1814 Lookup = nullptr;
1815 Diag(IncludeNextTok, diag::pp_include_next_in_primary);
1816 } else if (CurSubmodule) {
1817 // Start looking up in the directory *after* the one in which the current
1818 // file would be found, if any.
1819 assert(CurPPLexer && "#include_next directive in macro?");
1820 LookupFromFile = CurPPLexer->getFileEntry();
1821 Lookup = nullptr;
1822 } else if (!Lookup) {
1823 Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
1824 } else {
1825 // Start looking up in the next directory.
1826 ++Lookup;
1827 }
1828
1829 return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup,
1830 LookupFromFile);
1831 }
1832
1833 /// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode
HandleMicrosoftImportDirective(Token & Tok)1834 void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) {
1835 // The Microsoft #import directive takes a type library and generates header
1836 // files from it, and includes those. This is beyond the scope of what clang
1837 // does, so we ignore it and error out. However, #import can optionally have
1838 // trailing attributes that span multiple lines. We're going to eat those
1839 // so we can continue processing from there.
1840 Diag(Tok, diag::err_pp_import_directive_ms );
1841
1842 // Read tokens until we get to the end of the directive. Note that the
1843 // directive can be split over multiple lines using the backslash character.
1844 DiscardUntilEndOfDirective();
1845 }
1846
1847 /// HandleImportDirective - Implements \#import.
1848 ///
HandleImportDirective(SourceLocation HashLoc,Token & ImportTok)1849 void Preprocessor::HandleImportDirective(SourceLocation HashLoc,
1850 Token &ImportTok) {
1851 if (!LangOpts.ObjC1) { // #import is standard for ObjC.
1852 if (LangOpts.MSVCCompat)
1853 return HandleMicrosoftImportDirective(ImportTok);
1854 Diag(ImportTok, diag::ext_pp_import_directive);
1855 }
1856 return HandleIncludeDirective(HashLoc, ImportTok, nullptr, nullptr, true);
1857 }
1858
1859 /// HandleIncludeMacrosDirective - The -imacros command line option turns into a
1860 /// pseudo directive in the predefines buffer. This handles it by sucking all
1861 /// tokens through the preprocessor and discarding them (only keeping the side
1862 /// effects on the preprocessor).
HandleIncludeMacrosDirective(SourceLocation HashLoc,Token & IncludeMacrosTok)1863 void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc,
1864 Token &IncludeMacrosTok) {
1865 // This directive should only occur in the predefines buffer. If not, emit an
1866 // error and reject it.
1867 SourceLocation Loc = IncludeMacrosTok.getLocation();
1868 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) {
1869 Diag(IncludeMacrosTok.getLocation(),
1870 diag::pp_include_macros_out_of_predefines);
1871 DiscardUntilEndOfDirective();
1872 return;
1873 }
1874
1875 // Treat this as a normal #include for checking purposes. If this is
1876 // successful, it will push a new lexer onto the include stack.
1877 HandleIncludeDirective(HashLoc, IncludeMacrosTok);
1878
1879 Token TmpTok;
1880 do {
1881 Lex(TmpTok);
1882 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!");
1883 } while (TmpTok.isNot(tok::hashhash));
1884 }
1885
1886 //===----------------------------------------------------------------------===//
1887 // Preprocessor Macro Directive Handling.
1888 //===----------------------------------------------------------------------===//
1889
1890 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1891 /// definition has just been read. Lex the rest of the arguments and the
1892 /// closing ), updating MI with what we learn. Return true if an error occurs
1893 /// parsing the arg list.
ReadMacroDefinitionArgList(MacroInfo * MI,Token & Tok)1894 bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI, Token &Tok) {
1895 SmallVector<IdentifierInfo*, 32> Arguments;
1896
1897 while (1) {
1898 LexUnexpandedToken(Tok);
1899 switch (Tok.getKind()) {
1900 case tok::r_paren:
1901 // Found the end of the argument list.
1902 if (Arguments.empty()) // #define FOO()
1903 return false;
1904 // Otherwise we have #define FOO(A,)
1905 Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
1906 return true;
1907 case tok::ellipsis: // #define X(... -> C99 varargs
1908 if (!LangOpts.C99)
1909 Diag(Tok, LangOpts.CPlusPlus11 ?
1910 diag::warn_cxx98_compat_variadic_macro :
1911 diag::ext_variadic_macro);
1912
1913 // OpenCL v1.2 s6.9.e: variadic macros are not supported.
1914 if (LangOpts.OpenCL) {
1915 Diag(Tok, diag::err_pp_opencl_variadic_macros);
1916 return true;
1917 }
1918
1919 // Lex the token after the identifier.
1920 LexUnexpandedToken(Tok);
1921 if (Tok.isNot(tok::r_paren)) {
1922 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1923 return true;
1924 }
1925 // Add the __VA_ARGS__ identifier as an argument.
1926 Arguments.push_back(Ident__VA_ARGS__);
1927 MI->setIsC99Varargs();
1928 MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1929 return false;
1930 case tok::eod: // #define X(
1931 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1932 return true;
1933 default:
1934 // Handle keywords and identifiers here to accept things like
1935 // #define Foo(for) for.
1936 IdentifierInfo *II = Tok.getIdentifierInfo();
1937 if (!II) {
1938 // #define X(1
1939 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
1940 return true;
1941 }
1942
1943 // If this is already used as an argument, it is used multiple times (e.g.
1944 // #define X(A,A.
1945 if (std::find(Arguments.begin(), Arguments.end(), II) !=
1946 Arguments.end()) { // C99 6.10.3p6
1947 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II;
1948 return true;
1949 }
1950
1951 // Add the argument to the macro info.
1952 Arguments.push_back(II);
1953
1954 // Lex the token after the identifier.
1955 LexUnexpandedToken(Tok);
1956
1957 switch (Tok.getKind()) {
1958 default: // #define X(A B
1959 Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
1960 return true;
1961 case tok::r_paren: // #define X(A)
1962 MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1963 return false;
1964 case tok::comma: // #define X(A,
1965 break;
1966 case tok::ellipsis: // #define X(A... -> GCC extension
1967 // Diagnose extension.
1968 Diag(Tok, diag::ext_named_variadic_macro);
1969
1970 // Lex the token after the identifier.
1971 LexUnexpandedToken(Tok);
1972 if (Tok.isNot(tok::r_paren)) {
1973 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1974 return true;
1975 }
1976
1977 MI->setIsGNUVarargs();
1978 MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1979 return false;
1980 }
1981 }
1982 }
1983 }
1984
isConfigurationPattern(Token & MacroName,MacroInfo * MI,const LangOptions & LOptions)1985 static bool isConfigurationPattern(Token &MacroName, MacroInfo *MI,
1986 const LangOptions &LOptions) {
1987 if (MI->getNumTokens() == 1) {
1988 const Token &Value = MI->getReplacementToken(0);
1989
1990 // Macro that is identity, like '#define inline inline' is a valid pattern.
1991 if (MacroName.getKind() == Value.getKind())
1992 return true;
1993
1994 // Macro that maps a keyword to the same keyword decorated with leading/
1995 // trailing underscores is a valid pattern:
1996 // #define inline __inline
1997 // #define inline __inline__
1998 // #define inline _inline (in MS compatibility mode)
1999 StringRef MacroText = MacroName.getIdentifierInfo()->getName();
2000 if (IdentifierInfo *II = Value.getIdentifierInfo()) {
2001 if (!II->isKeyword(LOptions))
2002 return false;
2003 StringRef ValueText = II->getName();
2004 StringRef TrimmedValue = ValueText;
2005 if (!ValueText.startswith("__")) {
2006 if (ValueText.startswith("_"))
2007 TrimmedValue = TrimmedValue.drop_front(1);
2008 else
2009 return false;
2010 } else {
2011 TrimmedValue = TrimmedValue.drop_front(2);
2012 if (TrimmedValue.endswith("__"))
2013 TrimmedValue = TrimmedValue.drop_back(2);
2014 }
2015 return TrimmedValue.equals(MacroText);
2016 } else {
2017 return false;
2018 }
2019 }
2020
2021 // #define inline
2022 if (MacroName.isOneOf(tok::kw_extern, tok::kw_inline, tok::kw_static,
2023 tok::kw_const) &&
2024 MI->getNumTokens() == 0) {
2025 return true;
2026 }
2027
2028 return false;
2029 }
2030
2031 /// HandleDefineDirective - Implements \#define. This consumes the entire macro
2032 /// line then lets the caller lex the next real token.
HandleDefineDirective(Token & DefineTok,bool ImmediatelyAfterHeaderGuard)2033 void Preprocessor::HandleDefineDirective(Token &DefineTok,
2034 bool ImmediatelyAfterHeaderGuard) {
2035 ++NumDefined;
2036
2037 Token MacroNameTok;
2038 bool MacroShadowsKeyword;
2039 ReadMacroName(MacroNameTok, MU_Define, &MacroShadowsKeyword);
2040
2041 // Error reading macro name? If so, diagnostic already issued.
2042 if (MacroNameTok.is(tok::eod))
2043 return;
2044
2045 Token LastTok = MacroNameTok;
2046
2047 // If we are supposed to keep comments in #defines, reenable comment saving
2048 // mode.
2049 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
2050
2051 // Create the new macro.
2052 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation());
2053
2054 Token Tok;
2055 LexUnexpandedToken(Tok);
2056
2057 // If this is a function-like macro definition, parse the argument list,
2058 // marking each of the identifiers as being used as macro arguments. Also,
2059 // check other constraints on the first token of the macro body.
2060 if (Tok.is(tok::eod)) {
2061 if (ImmediatelyAfterHeaderGuard) {
2062 // Save this macro information since it may part of a header guard.
2063 CurPPLexer->MIOpt.SetDefinedMacro(MacroNameTok.getIdentifierInfo(),
2064 MacroNameTok.getLocation());
2065 }
2066 // If there is no body to this macro, we have no special handling here.
2067 } else if (Tok.hasLeadingSpace()) {
2068 // This is a normal token with leading space. Clear the leading space
2069 // marker on the first token to get proper expansion.
2070 Tok.clearFlag(Token::LeadingSpace);
2071 } else if (Tok.is(tok::l_paren)) {
2072 // This is a function-like macro definition. Read the argument list.
2073 MI->setIsFunctionLike();
2074 if (ReadMacroDefinitionArgList(MI, LastTok)) {
2075 // Throw away the rest of the line.
2076 if (CurPPLexer->ParsingPreprocessorDirective)
2077 DiscardUntilEndOfDirective();
2078 return;
2079 }
2080
2081 // If this is a definition of a variadic C99 function-like macro, not using
2082 // the GNU named varargs extension, enabled __VA_ARGS__.
2083
2084 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
2085 // This gets unpoisoned where it is allowed.
2086 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!");
2087 if (MI->isC99Varargs())
2088 Ident__VA_ARGS__->setIsPoisoned(false);
2089
2090 // Read the first token after the arg list for down below.
2091 LexUnexpandedToken(Tok);
2092 } else if (LangOpts.C99 || LangOpts.CPlusPlus11) {
2093 // C99 requires whitespace between the macro definition and the body. Emit
2094 // a diagnostic for something like "#define X+".
2095 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
2096 } else {
2097 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the
2098 // first character of a replacement list is not a character required by
2099 // subclause 5.2.1, then there shall be white-space separation between the
2100 // identifier and the replacement list.". 5.2.1 lists this set:
2101 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which
2102 // is irrelevant here.
2103 bool isInvalid = false;
2104 if (Tok.is(tok::at)) // @ is not in the list above.
2105 isInvalid = true;
2106 else if (Tok.is(tok::unknown)) {
2107 // If we have an unknown token, it is something strange like "`". Since
2108 // all of valid characters would have lexed into a single character
2109 // token of some sort, we know this is not a valid case.
2110 isInvalid = true;
2111 }
2112 if (isInvalid)
2113 Diag(Tok, diag::ext_missing_whitespace_after_macro_name);
2114 else
2115 Diag(Tok, diag::warn_missing_whitespace_after_macro_name);
2116 }
2117
2118 if (!Tok.is(tok::eod))
2119 LastTok = Tok;
2120
2121 // Read the rest of the macro body.
2122 if (MI->isObjectLike()) {
2123 // Object-like macros are very simple, just read their body.
2124 while (Tok.isNot(tok::eod)) {
2125 LastTok = Tok;
2126 MI->AddTokenToBody(Tok);
2127 // Get the next token of the macro.
2128 LexUnexpandedToken(Tok);
2129 }
2130
2131 } else {
2132 // Otherwise, read the body of a function-like macro. While we are at it,
2133 // check C99 6.10.3.2p1: ensure that # operators are followed by macro
2134 // parameters in function-like macro expansions.
2135 while (Tok.isNot(tok::eod)) {
2136 LastTok = Tok;
2137
2138 if (Tok.isNot(tok::hash) && Tok.isNot(tok::hashhash)) {
2139 MI->AddTokenToBody(Tok);
2140
2141 // Get the next token of the macro.
2142 LexUnexpandedToken(Tok);
2143 continue;
2144 }
2145
2146 // If we're in -traditional mode, then we should ignore stringification
2147 // and token pasting. Mark the tokens as unknown so as not to confuse
2148 // things.
2149 if (getLangOpts().TraditionalCPP) {
2150 Tok.setKind(tok::unknown);
2151 MI->AddTokenToBody(Tok);
2152
2153 // Get the next token of the macro.
2154 LexUnexpandedToken(Tok);
2155 continue;
2156 }
2157
2158 if (Tok.is(tok::hashhash)) {
2159
2160 // If we see token pasting, check if it looks like the gcc comma
2161 // pasting extension. We'll use this information to suppress
2162 // diagnostics later on.
2163
2164 // Get the next token of the macro.
2165 LexUnexpandedToken(Tok);
2166
2167 if (Tok.is(tok::eod)) {
2168 MI->AddTokenToBody(LastTok);
2169 break;
2170 }
2171
2172 unsigned NumTokens = MI->getNumTokens();
2173 if (NumTokens && Tok.getIdentifierInfo() == Ident__VA_ARGS__ &&
2174 MI->getReplacementToken(NumTokens-1).is(tok::comma))
2175 MI->setHasCommaPasting();
2176
2177 // Things look ok, add the '##' token to the macro.
2178 MI->AddTokenToBody(LastTok);
2179 continue;
2180 }
2181
2182 // Get the next token of the macro.
2183 LexUnexpandedToken(Tok);
2184
2185 // Check for a valid macro arg identifier.
2186 if (Tok.getIdentifierInfo() == nullptr ||
2187 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) {
2188
2189 // If this is assembler-with-cpp mode, we accept random gibberish after
2190 // the '#' because '#' is often a comment character. However, change
2191 // the kind of the token to tok::unknown so that the preprocessor isn't
2192 // confused.
2193 if (getLangOpts().AsmPreprocessor && Tok.isNot(tok::eod)) {
2194 LastTok.setKind(tok::unknown);
2195 MI->AddTokenToBody(LastTok);
2196 continue;
2197 } else {
2198 Diag(Tok, diag::err_pp_stringize_not_parameter);
2199
2200 // Disable __VA_ARGS__ again.
2201 Ident__VA_ARGS__->setIsPoisoned(true);
2202 return;
2203 }
2204 }
2205
2206 // Things look ok, add the '#' and param name tokens to the macro.
2207 MI->AddTokenToBody(LastTok);
2208 MI->AddTokenToBody(Tok);
2209 LastTok = Tok;
2210
2211 // Get the next token of the macro.
2212 LexUnexpandedToken(Tok);
2213 }
2214 }
2215
2216 if (MacroShadowsKeyword &&
2217 !isConfigurationPattern(MacroNameTok, MI, getLangOpts())) {
2218 Diag(MacroNameTok, diag::warn_pp_macro_hides_keyword);
2219 }
2220
2221 // Disable __VA_ARGS__ again.
2222 Ident__VA_ARGS__->setIsPoisoned(true);
2223
2224 // Check that there is no paste (##) operator at the beginning or end of the
2225 // replacement list.
2226 unsigned NumTokens = MI->getNumTokens();
2227 if (NumTokens != 0) {
2228 if (MI->getReplacementToken(0).is(tok::hashhash)) {
2229 Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
2230 return;
2231 }
2232 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
2233 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
2234 return;
2235 }
2236 }
2237
2238 MI->setDefinitionEndLoc(LastTok.getLocation());
2239
2240 // Finally, if this identifier already had a macro defined for it, verify that
2241 // the macro bodies are identical, and issue diagnostics if they are not.
2242 if (const MacroInfo *OtherMI=getMacroInfo(MacroNameTok.getIdentifierInfo())) {
2243 // It is very common for system headers to have tons of macro redefinitions
2244 // and for warnings to be disabled in system headers. If this is the case,
2245 // then don't bother calling MacroInfo::isIdenticalTo.
2246 if (!getDiagnostics().getSuppressSystemWarnings() ||
2247 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) {
2248 if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused())
2249 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
2250
2251 // Warn if defining "__LINE__" and other builtins, per C99 6.10.8/4 and
2252 // C++ [cpp.predefined]p4, but allow it as an extension.
2253 if (OtherMI->isBuiltinMacro())
2254 Diag(MacroNameTok, diag::ext_pp_redef_builtin_macro);
2255 // Macros must be identical. This means all tokens and whitespace
2256 // separation must be the same. C99 6.10.3p2.
2257 else if (!OtherMI->isAllowRedefinitionsWithoutWarning() &&
2258 !MI->isIdenticalTo(*OtherMI, *this, /*Syntactic=*/LangOpts.MicrosoftExt)) {
2259 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
2260 << MacroNameTok.getIdentifierInfo();
2261 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition);
2262 }
2263 }
2264 if (OtherMI->isWarnIfUnused())
2265 WarnUnusedMacroLocs.erase(OtherMI->getDefinitionLoc());
2266 }
2267
2268 DefMacroDirective *MD =
2269 appendDefMacroDirective(MacroNameTok.getIdentifierInfo(), MI);
2270
2271 assert(!MI->isUsed());
2272 // If we need warning for not using the macro, add its location in the
2273 // warn-because-unused-macro set. If it gets used it will be removed from set.
2274 if (getSourceManager().isInMainFile(MI->getDefinitionLoc()) &&
2275 !Diags->isIgnored(diag::pp_macro_not_used, MI->getDefinitionLoc())) {
2276 MI->setIsWarnIfUnused(true);
2277 WarnUnusedMacroLocs.insert(MI->getDefinitionLoc());
2278 }
2279
2280 // If the callbacks want to know, tell them about the macro definition.
2281 if (Callbacks)
2282 Callbacks->MacroDefined(MacroNameTok, MD);
2283 }
2284
2285 /// HandleUndefDirective - Implements \#undef.
2286 ///
HandleUndefDirective(Token & UndefTok)2287 void Preprocessor::HandleUndefDirective(Token &UndefTok) {
2288 ++NumUndefined;
2289
2290 Token MacroNameTok;
2291 ReadMacroName(MacroNameTok, MU_Undef);
2292
2293 // Error reading macro name? If so, diagnostic already issued.
2294 if (MacroNameTok.is(tok::eod))
2295 return;
2296
2297 // Check to see if this is the last token on the #undef line.
2298 CheckEndOfDirective("undef");
2299
2300 // Okay, we have a valid identifier to undef.
2301 auto *II = MacroNameTok.getIdentifierInfo();
2302 auto MD = getMacroDefinition(II);
2303
2304 // If the callbacks want to know, tell them about the macro #undef.
2305 // Note: no matter if the macro was defined or not.
2306 if (Callbacks)
2307 Callbacks->MacroUndefined(MacroNameTok, MD);
2308
2309 // If the macro is not defined, this is a noop undef, just return.
2310 const MacroInfo *MI = MD.getMacroInfo();
2311 if (!MI)
2312 return;
2313
2314 if (!MI->isUsed() && MI->isWarnIfUnused())
2315 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
2316
2317 if (MI->isWarnIfUnused())
2318 WarnUnusedMacroLocs.erase(MI->getDefinitionLoc());
2319
2320 appendMacroDirective(MacroNameTok.getIdentifierInfo(),
2321 AllocateUndefMacroDirective(MacroNameTok.getLocation()));
2322 }
2323
2324
2325 //===----------------------------------------------------------------------===//
2326 // Preprocessor Conditional Directive Handling.
2327 //===----------------------------------------------------------------------===//
2328
2329 /// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive. isIfndef
2330 /// is true when this is a \#ifndef directive. ReadAnyTokensBeforeDirective is
2331 /// true if any tokens have been returned or pp-directives activated before this
2332 /// \#ifndef has been lexed.
2333 ///
HandleIfdefDirective(Token & Result,bool isIfndef,bool ReadAnyTokensBeforeDirective)2334 void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef,
2335 bool ReadAnyTokensBeforeDirective) {
2336 ++NumIf;
2337 Token DirectiveTok = Result;
2338
2339 Token MacroNameTok;
2340 ReadMacroName(MacroNameTok);
2341
2342 // Error reading macro name? If so, diagnostic already issued.
2343 if (MacroNameTok.is(tok::eod)) {
2344 // Skip code until we get to #endif. This helps with recovery by not
2345 // emitting an error when the #endif is reached.
2346 SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
2347 /*Foundnonskip*/false, /*FoundElse*/false);
2348 return;
2349 }
2350
2351 // Check to see if this is the last token on the #if[n]def line.
2352 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef");
2353
2354 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
2355 auto MD = getMacroDefinition(MII);
2356 MacroInfo *MI = MD.getMacroInfo();
2357
2358 if (CurPPLexer->getConditionalStackDepth() == 0) {
2359 // If the start of a top-level #ifdef and if the macro is not defined,
2360 // inform MIOpt that this might be the start of a proper include guard.
2361 // Otherwise it is some other form of unknown conditional which we can't
2362 // handle.
2363 if (!ReadAnyTokensBeforeDirective && !MI) {
2364 assert(isIfndef && "#ifdef shouldn't reach here");
2365 CurPPLexer->MIOpt.EnterTopLevelIfndef(MII, MacroNameTok.getLocation());
2366 } else
2367 CurPPLexer->MIOpt.EnterTopLevelConditional();
2368 }
2369
2370 // If there is a macro, process it.
2371 if (MI) // Mark it used.
2372 markMacroAsUsed(MI);
2373
2374 if (Callbacks) {
2375 if (isIfndef)
2376 Callbacks->Ifndef(DirectiveTok.getLocation(), MacroNameTok, MD);
2377 else
2378 Callbacks->Ifdef(DirectiveTok.getLocation(), MacroNameTok, MD);
2379 }
2380
2381 // Should we include the stuff contained by this directive?
2382 if (!MI == isIfndef) {
2383 // Yes, remember that we are inside a conditional, then lex the next token.
2384 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(),
2385 /*wasskip*/false, /*foundnonskip*/true,
2386 /*foundelse*/false);
2387 } else {
2388 // No, skip the contents of this block.
2389 SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
2390 /*Foundnonskip*/false,
2391 /*FoundElse*/false);
2392 }
2393 }
2394
2395 /// HandleIfDirective - Implements the \#if directive.
2396 ///
HandleIfDirective(Token & IfToken,bool ReadAnyTokensBeforeDirective)2397 void Preprocessor::HandleIfDirective(Token &IfToken,
2398 bool ReadAnyTokensBeforeDirective) {
2399 ++NumIf;
2400
2401 // Parse and evaluate the conditional expression.
2402 IdentifierInfo *IfNDefMacro = nullptr;
2403 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2404 const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro);
2405 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2406
2407 // If this condition is equivalent to #ifndef X, and if this is the first
2408 // directive seen, handle it for the multiple-include optimization.
2409 if (CurPPLexer->getConditionalStackDepth() == 0) {
2410 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue)
2411 // FIXME: Pass in the location of the macro name, not the 'if' token.
2412 CurPPLexer->MIOpt.EnterTopLevelIfndef(IfNDefMacro, IfToken.getLocation());
2413 else
2414 CurPPLexer->MIOpt.EnterTopLevelConditional();
2415 }
2416
2417 if (Callbacks)
2418 Callbacks->If(IfToken.getLocation(),
2419 SourceRange(ConditionalBegin, ConditionalEnd),
2420 (ConditionalTrue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False));
2421
2422 // Should we include the stuff contained by this directive?
2423 if (ConditionalTrue) {
2424 // Yes, remember that we are inside a conditional, then lex the next token.
2425 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
2426 /*foundnonskip*/true, /*foundelse*/false);
2427 } else {
2428 // No, skip the contents of this block.
2429 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false,
2430 /*FoundElse*/false);
2431 }
2432 }
2433
2434 /// HandleEndifDirective - Implements the \#endif directive.
2435 ///
HandleEndifDirective(Token & EndifToken)2436 void Preprocessor::HandleEndifDirective(Token &EndifToken) {
2437 ++NumEndif;
2438
2439 // Check that this is the whole directive.
2440 CheckEndOfDirective("endif");
2441
2442 PPConditionalInfo CondInfo;
2443 if (CurPPLexer->popConditionalLevel(CondInfo)) {
2444 // No conditionals on the stack: this is an #endif without an #if.
2445 Diag(EndifToken, diag::err_pp_endif_without_if);
2446 return;
2447 }
2448
2449 // If this the end of a top-level #endif, inform MIOpt.
2450 if (CurPPLexer->getConditionalStackDepth() == 0)
2451 CurPPLexer->MIOpt.ExitTopLevelConditional();
2452
2453 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
2454 "This code should only be reachable in the non-skipping case!");
2455
2456 if (Callbacks)
2457 Callbacks->Endif(EndifToken.getLocation(), CondInfo.IfLoc);
2458 }
2459
2460 /// HandleElseDirective - Implements the \#else directive.
2461 ///
HandleElseDirective(Token & Result)2462 void Preprocessor::HandleElseDirective(Token &Result) {
2463 ++NumElse;
2464
2465 // #else directive in a non-skipping conditional... start skipping.
2466 CheckEndOfDirective("else");
2467
2468 PPConditionalInfo CI;
2469 if (CurPPLexer->popConditionalLevel(CI)) {
2470 Diag(Result, diag::pp_err_else_without_if);
2471 return;
2472 }
2473
2474 // If this is a top-level #else, inform the MIOpt.
2475 if (CurPPLexer->getConditionalStackDepth() == 0)
2476 CurPPLexer->MIOpt.EnterTopLevelConditional();
2477
2478 // If this is a #else with a #else before it, report the error.
2479 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
2480
2481 if (Callbacks)
2482 Callbacks->Else(Result.getLocation(), CI.IfLoc);
2483
2484 // Finally, skip the rest of the contents of this block.
2485 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2486 /*FoundElse*/true, Result.getLocation());
2487 }
2488
2489 /// HandleElifDirective - Implements the \#elif directive.
2490 ///
HandleElifDirective(Token & ElifToken)2491 void Preprocessor::HandleElifDirective(Token &ElifToken) {
2492 ++NumElse;
2493
2494 // #elif directive in a non-skipping conditional... start skipping.
2495 // We don't care what the condition is, because we will always skip it (since
2496 // the block immediately before it was included).
2497 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2498 DiscardUntilEndOfDirective();
2499 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2500
2501 PPConditionalInfo CI;
2502 if (CurPPLexer->popConditionalLevel(CI)) {
2503 Diag(ElifToken, diag::pp_err_elif_without_if);
2504 return;
2505 }
2506
2507 // If this is a top-level #elif, inform the MIOpt.
2508 if (CurPPLexer->getConditionalStackDepth() == 0)
2509 CurPPLexer->MIOpt.EnterTopLevelConditional();
2510
2511 // If this is a #elif with a #else before it, report the error.
2512 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
2513
2514 if (Callbacks)
2515 Callbacks->Elif(ElifToken.getLocation(),
2516 SourceRange(ConditionalBegin, ConditionalEnd),
2517 PPCallbacks::CVK_NotEvaluated, CI.IfLoc);
2518
2519 // Finally, skip the rest of the contents of this block.
2520 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2521 /*FoundElse*/CI.FoundElse,
2522 ElifToken.getLocation());
2523 }
2524