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/Basic/CharInfo.h" 16 #include "clang/Basic/FileManager.h" 17 #include "clang/Basic/IdentifierTable.h" 18 #include "clang/Basic/LangOptions.h" 19 #include "clang/Basic/Module.h" 20 #include "clang/Basic/SourceLocation.h" 21 #include "clang/Basic/SourceManager.h" 22 #include "clang/Basic/TokenKinds.h" 23 #include "clang/Lex/CodeCompletionHandler.h" 24 #include "clang/Lex/HeaderSearch.h" 25 #include "clang/Lex/LexDiagnostic.h" 26 #include "clang/Lex/LiteralSupport.h" 27 #include "clang/Lex/MacroInfo.h" 28 #include "clang/Lex/ModuleLoader.h" 29 #include "clang/Lex/ModuleMap.h" 30 #include "clang/Lex/PPCallbacks.h" 31 #include "clang/Lex/Pragma.h" 32 #include "clang/Lex/Preprocessor.h" 33 #include "clang/Lex/PTHLexer.h" 34 #include "clang/Lex/Token.h" 35 #include "llvm/ADT/ArrayRef.h" 36 #include "llvm/ADT/SmallString.h" 37 #include "llvm/ADT/SmallVector.h" 38 #include "llvm/ADT/STLExtras.h" 39 #include "llvm/ADT/StringSwitch.h" 40 #include "llvm/ADT/StringRef.h" 41 #include "llvm/Support/AlignOf.h" 42 #include "llvm/Support/ErrorHandling.h" 43 #include "llvm/Support/Path.h" 44 #include <algorithm> 45 #include <cassert> 46 #include <cstring> 47 #include <new> 48 #include <string> 49 #include <utility> 50 51 using namespace clang; 52 53 //===----------------------------------------------------------------------===// 54 // Utility Methods for Preprocessor Directive Handling. 55 //===----------------------------------------------------------------------===// 56 57 MacroInfo *Preprocessor::AllocateMacroInfo() { 58 MacroInfoChain *MIChain = BP.Allocate<MacroInfoChain>(); 59 MIChain->Next = MIChainHead; 60 MIChainHead = MIChain; 61 return &MIChain->MI; 62 } 63 64 MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) { 65 MacroInfo *MI = AllocateMacroInfo(); 66 new (MI) MacroInfo(L); 67 return MI; 68 } 69 70 MacroInfo *Preprocessor::AllocateDeserializedMacroInfo(SourceLocation L, 71 unsigned SubModuleID) { 72 static_assert(alignof(MacroInfo) >= sizeof(SubModuleID), 73 "alignment for MacroInfo is less than the ID"); 74 DeserializedMacroInfoChain *MIChain = 75 BP.Allocate<DeserializedMacroInfoChain>(); 76 MIChain->Next = DeserialMIChainHead; 77 DeserialMIChainHead = MIChain; 78 79 MacroInfo *MI = &MIChain->MI; 80 new (MI) MacroInfo(L); 81 MI->FromASTFile = true; 82 MI->setOwningModuleID(SubModuleID); 83 return MI; 84 } 85 86 DefMacroDirective *Preprocessor::AllocateDefMacroDirective(MacroInfo *MI, 87 SourceLocation Loc) { 88 return new (BP) DefMacroDirective(MI, Loc); 89 } 90 91 UndefMacroDirective * 92 Preprocessor::AllocateUndefMacroDirective(SourceLocation UndefLoc) { 93 return new (BP) UndefMacroDirective(UndefLoc); 94 } 95 96 VisibilityMacroDirective * 97 Preprocessor::AllocateVisibilityMacroDirective(SourceLocation Loc, 98 bool isPublic) { 99 return new (BP) VisibilityMacroDirective(Loc, isPublic); 100 } 101 102 /// \brief Read and discard all tokens remaining on the current line until 103 /// the tok::eod token is found. 104 void Preprocessor::DiscardUntilEndOfDirective() { 105 Token Tmp; 106 do { 107 LexUnexpandedToken(Tmp); 108 assert(Tmp.isNot(tok::eof) && "EOF seen while discarding directive tokens"); 109 } while (Tmp.isNot(tok::eod)); 110 } 111 112 /// \brief Enumerates possible cases of #define/#undef a reserved identifier. 113 enum MacroDiag { 114 MD_NoWarn, //> Not a reserved identifier 115 MD_KeywordDef, //> Macro hides keyword, enabled by default 116 MD_ReservedMacro //> #define of #undef reserved id, disabled by default 117 }; 118 119 /// \brief Checks if the specified identifier is reserved in the specified 120 /// language. 121 /// This function does not check if the identifier is a keyword. 122 static bool isReservedId(StringRef Text, const LangOptions &Lang) { 123 // C++ [macro.names], C11 7.1.3: 124 // All identifiers that begin with an underscore and either an uppercase 125 // letter or another underscore are always reserved for any use. 126 if (Text.size() >= 2 && Text[0] == '_' && 127 (isUppercase(Text[1]) || Text[1] == '_')) 128 return true; 129 // C++ [global.names] 130 // Each name that contains a double underscore ... is reserved to the 131 // implementation for any use. 132 if (Lang.CPlusPlus) { 133 if (Text.find("__") != StringRef::npos) 134 return true; 135 } 136 return false; 137 } 138 139 static MacroDiag shouldWarnOnMacroDef(Preprocessor &PP, IdentifierInfo *II) { 140 const LangOptions &Lang = PP.getLangOpts(); 141 StringRef Text = II->getName(); 142 if (isReservedId(Text, Lang)) 143 return MD_ReservedMacro; 144 if (II->isKeyword(Lang)) 145 return MD_KeywordDef; 146 if (Lang.CPlusPlus11 && (Text.equals("override") || Text.equals("final"))) 147 return MD_KeywordDef; 148 return MD_NoWarn; 149 } 150 151 static MacroDiag shouldWarnOnMacroUndef(Preprocessor &PP, IdentifierInfo *II) { 152 const LangOptions &Lang = PP.getLangOpts(); 153 StringRef Text = II->getName(); 154 // Do not warn on keyword undef. It is generally harmless and widely used. 155 if (isReservedId(Text, Lang)) 156 return MD_ReservedMacro; 157 return MD_NoWarn; 158 } 159 160 // Return true if we want to issue a diagnostic by default if we 161 // encounter this name in a #include with the wrong case. For now, 162 // this includes the standard C and C++ headers, Posix headers, 163 // and Boost headers. Improper case for these #includes is a 164 // potential portability issue. 165 static bool warnByDefaultOnWrongCase(StringRef Include) { 166 // If the first component of the path is "boost", treat this like a standard header 167 // for the purposes of diagnostics. 168 if (::llvm::sys::path::begin(Include)->equals_lower("boost")) 169 return true; 170 171 // "condition_variable" is the longest standard header name at 18 characters. 172 // If the include file name is longer than that, it can't be a standard header. 173 static const size_t MaxStdHeaderNameLen = 18u; 174 if (Include.size() > MaxStdHeaderNameLen) 175 return false; 176 177 // Lowercase and normalize the search string. 178 SmallString<32> LowerInclude{Include}; 179 for (char &Ch : LowerInclude) { 180 // In the ASCII range? 181 if (static_cast<unsigned char>(Ch) > 0x7f) 182 return false; // Can't be a standard header 183 // ASCII lowercase: 184 if (Ch >= 'A' && Ch <= 'Z') 185 Ch += 'a' - 'A'; 186 // Normalize path separators for comparison purposes. 187 else if (::llvm::sys::path::is_separator(Ch)) 188 Ch = '/'; 189 } 190 191 // The standard C/C++ and Posix headers 192 return llvm::StringSwitch<bool>(LowerInclude) 193 // C library headers 194 .Cases("assert.h", "complex.h", "ctype.h", "errno.h", "fenv.h", true) 195 .Cases("float.h", "inttypes.h", "iso646.h", "limits.h", "locale.h", true) 196 .Cases("math.h", "setjmp.h", "signal.h", "stdalign.h", "stdarg.h", true) 197 .Cases("stdatomic.h", "stdbool.h", "stddef.h", "stdint.h", "stdio.h", true) 198 .Cases("stdlib.h", "stdnoreturn.h", "string.h", "tgmath.h", "threads.h", true) 199 .Cases("time.h", "uchar.h", "wchar.h", "wctype.h", true) 200 201 // C++ headers for C library facilities 202 .Cases("cassert", "ccomplex", "cctype", "cerrno", "cfenv", true) 203 .Cases("cfloat", "cinttypes", "ciso646", "climits", "clocale", true) 204 .Cases("cmath", "csetjmp", "csignal", "cstdalign", "cstdarg", true) 205 .Cases("cstdbool", "cstddef", "cstdint", "cstdio", "cstdlib", true) 206 .Cases("cstring", "ctgmath", "ctime", "cuchar", "cwchar", true) 207 .Case("cwctype", true) 208 209 // C++ library headers 210 .Cases("algorithm", "fstream", "list", "regex", "thread", true) 211 .Cases("array", "functional", "locale", "scoped_allocator", "tuple", true) 212 .Cases("atomic", "future", "map", "set", "type_traits", true) 213 .Cases("bitset", "initializer_list", "memory", "shared_mutex", "typeindex", true) 214 .Cases("chrono", "iomanip", "mutex", "sstream", "typeinfo", true) 215 .Cases("codecvt", "ios", "new", "stack", "unordered_map", true) 216 .Cases("complex", "iosfwd", "numeric", "stdexcept", "unordered_set", true) 217 .Cases("condition_variable", "iostream", "ostream", "streambuf", "utility", true) 218 .Cases("deque", "istream", "queue", "string", "valarray", true) 219 .Cases("exception", "iterator", "random", "strstream", "vector", true) 220 .Cases("forward_list", "limits", "ratio", "system_error", true) 221 222 // POSIX headers (which aren't also C headers) 223 .Cases("aio.h", "arpa/inet.h", "cpio.h", "dirent.h", "dlfcn.h", true) 224 .Cases("fcntl.h", "fmtmsg.h", "fnmatch.h", "ftw.h", "glob.h", true) 225 .Cases("grp.h", "iconv.h", "langinfo.h", "libgen.h", "monetary.h", true) 226 .Cases("mqueue.h", "ndbm.h", "net/if.h", "netdb.h", "netinet/in.h", true) 227 .Cases("netinet/tcp.h", "nl_types.h", "poll.h", "pthread.h", "pwd.h", true) 228 .Cases("regex.h", "sched.h", "search.h", "semaphore.h", "spawn.h", true) 229 .Cases("strings.h", "stropts.h", "sys/ipc.h", "sys/mman.h", "sys/msg.h", true) 230 .Cases("sys/resource.h", "sys/select.h", "sys/sem.h", "sys/shm.h", "sys/socket.h", true) 231 .Cases("sys/stat.h", "sys/statvfs.h", "sys/time.h", "sys/times.h", "sys/types.h", true) 232 .Cases("sys/uio.h", "sys/un.h", "sys/utsname.h", "sys/wait.h", "syslog.h", true) 233 .Cases("tar.h", "termios.h", "trace.h", "ulimit.h", true) 234 .Cases("unistd.h", "utime.h", "utmpx.h", "wordexp.h", true) 235 .Default(false); 236 } 237 238 bool Preprocessor::CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef, 239 bool *ShadowFlag) { 240 // Missing macro name? 241 if (MacroNameTok.is(tok::eod)) 242 return Diag(MacroNameTok, diag::err_pp_missing_macro_name); 243 244 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 245 if (!II) { 246 bool Invalid = false; 247 std::string Spelling = getSpelling(MacroNameTok, &Invalid); 248 if (Invalid) 249 return Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 250 II = getIdentifierInfo(Spelling); 251 252 if (!II->isCPlusPlusOperatorKeyword()) 253 return Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 254 255 // C++ 2.5p2: Alternative tokens behave the same as its primary token 256 // except for their spellings. 257 Diag(MacroNameTok, getLangOpts().MicrosoftExt 258 ? diag::ext_pp_operator_used_as_macro_name 259 : diag::err_pp_operator_used_as_macro_name) 260 << II << MacroNameTok.getKind(); 261 262 // Allow #defining |and| and friends for Microsoft compatibility or 263 // recovery when legacy C headers are included in C++. 264 MacroNameTok.setIdentifierInfo(II); 265 } 266 267 if ((isDefineUndef != MU_Other) && II->getPPKeywordID() == tok::pp_defined) { 268 // Error if defining "defined": C99 6.10.8/4, C++ [cpp.predefined]p4. 269 return Diag(MacroNameTok, diag::err_defined_macro_name); 270 } 271 272 if (isDefineUndef == MU_Undef) { 273 auto *MI = getMacroInfo(II); 274 if (MI && MI->isBuiltinMacro()) { 275 // Warn if undefining "__LINE__" and other builtins, per C99 6.10.8/4 276 // and C++ [cpp.predefined]p4], but allow it as an extension. 277 Diag(MacroNameTok, diag::ext_pp_undef_builtin_macro); 278 } 279 } 280 281 // If defining/undefining reserved identifier or a keyword, we need to issue 282 // a warning. 283 SourceLocation MacroNameLoc = MacroNameTok.getLocation(); 284 if (ShadowFlag) 285 *ShadowFlag = false; 286 if (!SourceMgr.isInSystemHeader(MacroNameLoc) && 287 (SourceMgr.getBufferName(MacroNameLoc) != "<built-in>")) { 288 MacroDiag D = MD_NoWarn; 289 if (isDefineUndef == MU_Define) { 290 D = shouldWarnOnMacroDef(*this, II); 291 } 292 else if (isDefineUndef == MU_Undef) 293 D = shouldWarnOnMacroUndef(*this, II); 294 if (D == MD_KeywordDef) { 295 // We do not want to warn on some patterns widely used in configuration 296 // scripts. This requires analyzing next tokens, so do not issue warnings 297 // now, only inform caller. 298 if (ShadowFlag) 299 *ShadowFlag = true; 300 } 301 if (D == MD_ReservedMacro) 302 Diag(MacroNameTok, diag::warn_pp_macro_is_reserved_id); 303 } 304 305 // Okay, we got a good identifier. 306 return false; 307 } 308 309 /// \brief Lex and validate a macro name, which occurs after a 310 /// \#define or \#undef. 311 /// 312 /// This sets the token kind to eod and discards the rest of the macro line if 313 /// the macro name is invalid. 314 /// 315 /// \param MacroNameTok Token that is expected to be a macro name. 316 /// \param isDefineUndef Context in which macro is used. 317 /// \param ShadowFlag Points to a flag that is set if macro shadows a keyword. 318 void Preprocessor::ReadMacroName(Token &MacroNameTok, MacroUse isDefineUndef, 319 bool *ShadowFlag) { 320 // Read the token, don't allow macro expansion on it. 321 LexUnexpandedToken(MacroNameTok); 322 323 if (MacroNameTok.is(tok::code_completion)) { 324 if (CodeComplete) 325 CodeComplete->CodeCompleteMacroName(isDefineUndef == MU_Define); 326 setCodeCompletionReached(); 327 LexUnexpandedToken(MacroNameTok); 328 } 329 330 if (!CheckMacroName(MacroNameTok, isDefineUndef, ShadowFlag)) 331 return; 332 333 // Invalid macro name, read and discard the rest of the line and set the 334 // token kind to tok::eod if necessary. 335 if (MacroNameTok.isNot(tok::eod)) { 336 MacroNameTok.setKind(tok::eod); 337 DiscardUntilEndOfDirective(); 338 } 339 } 340 341 /// \brief Ensure that the next token is a tok::eod token. 342 /// 343 /// If not, emit a diagnostic and consume up until the eod. If EnableMacros is 344 /// true, then we consider macros that expand to zero tokens as being ok. 345 void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) { 346 Token Tmp; 347 // Lex unexpanded tokens for most directives: macros might expand to zero 348 // tokens, causing us to miss diagnosing invalid lines. Some directives (like 349 // #line) allow empty macros. 350 if (EnableMacros) 351 Lex(Tmp); 352 else 353 LexUnexpandedToken(Tmp); 354 355 // There should be no tokens after the directive, but we allow them as an 356 // extension. 357 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 358 LexUnexpandedToken(Tmp); 359 360 if (Tmp.isNot(tok::eod)) { 361 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89, 362 // or if this is a macro-style preprocessing directive, because it is more 363 // trouble than it is worth to insert /**/ and check that there is no /**/ 364 // in the range also. 365 FixItHint Hint; 366 if ((LangOpts.GNUMode || LangOpts.C99 || LangOpts.CPlusPlus) && 367 !CurTokenLexer) 368 Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//"); 369 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint; 370 DiscardUntilEndOfDirective(); 371 } 372 } 373 374 /// SkipExcludedConditionalBlock - We just read a \#if or related directive and 375 /// decided that the subsequent tokens are in the \#if'd out portion of the 376 /// file. Lex the rest of the file, until we see an \#endif. If 377 /// FoundNonSkipPortion is true, then we have already emitted code for part of 378 /// this \#if directive, so \#else/\#elif blocks should never be entered. 379 /// If ElseOk is true, then \#else directives are ok, if not, then we have 380 /// already seen one so a \#else directive is a duplicate. When this returns, 381 /// the caller can lex the first valid token. 382 void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 383 bool FoundNonSkipPortion, 384 bool FoundElse, 385 SourceLocation ElseLoc) { 386 ++NumSkipped; 387 assert(!CurTokenLexer && CurPPLexer && "Lexing a macro, not a file?"); 388 389 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false, 390 FoundNonSkipPortion, FoundElse); 391 392 if (CurPTHLexer) { 393 PTHSkipExcludedConditionalBlock(); 394 return; 395 } 396 397 // Enter raw mode to disable identifier lookup (and thus macro expansion), 398 // disabling warnings, etc. 399 CurPPLexer->LexingRawMode = true; 400 Token Tok; 401 while (true) { 402 CurLexer->Lex(Tok); 403 404 if (Tok.is(tok::code_completion)) { 405 if (CodeComplete) 406 CodeComplete->CodeCompleteInConditionalExclusion(); 407 setCodeCompletionReached(); 408 continue; 409 } 410 411 // If this is the end of the buffer, we have an error. 412 if (Tok.is(tok::eof)) { 413 // Emit errors for each unterminated conditional on the stack, including 414 // the current one. 415 while (!CurPPLexer->ConditionalStack.empty()) { 416 if (CurLexer->getFileLoc() != CodeCompletionFileLoc) 417 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 418 diag::err_pp_unterminated_conditional); 419 CurPPLexer->ConditionalStack.pop_back(); 420 } 421 422 // Just return and let the caller lex after this #include. 423 break; 424 } 425 426 // If this token is not a preprocessor directive, just skip it. 427 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 428 continue; 429 430 // We just parsed a # character at the start of a line, so we're in 431 // directive mode. Tell the lexer this so any newlines we see will be 432 // converted into an EOD token (this terminates the macro). 433 CurPPLexer->ParsingPreprocessorDirective = true; 434 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false); 435 436 437 // Read the next token, the directive flavor. 438 LexUnexpandedToken(Tok); 439 440 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 441 // something bogus), skip it. 442 if (Tok.isNot(tok::raw_identifier)) { 443 CurPPLexer->ParsingPreprocessorDirective = false; 444 // Restore comment saving mode. 445 if (CurLexer) CurLexer->resetExtendedTokenMode(); 446 continue; 447 } 448 449 // If the first letter isn't i or e, it isn't intesting to us. We know that 450 // this is safe in the face of spelling differences, because there is no way 451 // to spell an i/e in a strange way that is another letter. Skipping this 452 // allows us to avoid looking up the identifier info for #define/#undef and 453 // other common directives. 454 StringRef RI = Tok.getRawIdentifier(); 455 456 char FirstChar = RI[0]; 457 if (FirstChar >= 'a' && FirstChar <= 'z' && 458 FirstChar != 'i' && FirstChar != 'e') { 459 CurPPLexer->ParsingPreprocessorDirective = false; 460 // Restore comment saving mode. 461 if (CurLexer) CurLexer->resetExtendedTokenMode(); 462 continue; 463 } 464 465 // Get the identifier name without trigraphs or embedded newlines. Note 466 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 467 // when skipping. 468 char DirectiveBuf[20]; 469 StringRef Directive; 470 if (!Tok.needsCleaning() && RI.size() < 20) { 471 Directive = RI; 472 } else { 473 std::string DirectiveStr = getSpelling(Tok); 474 size_t IdLen = DirectiveStr.size(); 475 if (IdLen >= 20) { 476 CurPPLexer->ParsingPreprocessorDirective = false; 477 // Restore comment saving mode. 478 if (CurLexer) CurLexer->resetExtendedTokenMode(); 479 continue; 480 } 481 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen); 482 Directive = StringRef(DirectiveBuf, IdLen); 483 } 484 485 if (Directive.startswith("if")) { 486 StringRef Sub = Directive.substr(2); 487 if (Sub.empty() || // "if" 488 Sub == "def" || // "ifdef" 489 Sub == "ndef") { // "ifndef" 490 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 491 // bother parsing the condition. 492 DiscardUntilEndOfDirective(); 493 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 494 /*foundnonskip*/false, 495 /*foundelse*/false); 496 } 497 } else if (Directive[0] == 'e') { 498 StringRef Sub = Directive.substr(1); 499 if (Sub == "ndif") { // "endif" 500 PPConditionalInfo CondInfo; 501 CondInfo.WasSkipping = true; // Silence bogus warning. 502 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 503 (void)InCond; // Silence warning in no-asserts mode. 504 assert(!InCond && "Can't be skipping if not in a conditional!"); 505 506 // If we popped the outermost skipping block, we're done skipping! 507 if (!CondInfo.WasSkipping) { 508 // Restore the value of LexingRawMode so that trailing comments 509 // are handled correctly, if we've reached the outermost block. 510 CurPPLexer->LexingRawMode = false; 511 CheckEndOfDirective("endif"); 512 CurPPLexer->LexingRawMode = true; 513 if (Callbacks) 514 Callbacks->Endif(Tok.getLocation(), CondInfo.IfLoc); 515 break; 516 } else { 517 DiscardUntilEndOfDirective(); 518 } 519 } else if (Sub == "lse") { // "else". 520 // #else directive in a skipping conditional. If not in some other 521 // skipping conditional, and if #else hasn't already been seen, enter it 522 // as a non-skipping conditional. 523 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 524 525 // If this is a #else with a #else before it, report the error. 526 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 527 528 // Note that we've seen a #else in this conditional. 529 CondInfo.FoundElse = true; 530 531 // If the conditional is at the top level, and the #if block wasn't 532 // entered, enter the #else block now. 533 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 534 CondInfo.FoundNonSkip = true; 535 // Restore the value of LexingRawMode so that trailing comments 536 // are handled correctly. 537 CurPPLexer->LexingRawMode = false; 538 CheckEndOfDirective("else"); 539 CurPPLexer->LexingRawMode = true; 540 if (Callbacks) 541 Callbacks->Else(Tok.getLocation(), CondInfo.IfLoc); 542 break; 543 } else { 544 DiscardUntilEndOfDirective(); // C99 6.10p4. 545 } 546 } else if (Sub == "lif") { // "elif". 547 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 548 549 // If this is a #elif with a #else before it, report the error. 550 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 551 552 // If this is in a skipping block or if we're already handled this #if 553 // block, don't bother parsing the condition. 554 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 555 DiscardUntilEndOfDirective(); 556 } else { 557 const SourceLocation CondBegin = CurPPLexer->getSourceLocation(); 558 // Restore the value of LexingRawMode so that identifiers are 559 // looked up, etc, inside the #elif expression. 560 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 561 CurPPLexer->LexingRawMode = false; 562 IdentifierInfo *IfNDefMacro = nullptr; 563 const bool CondValue = EvaluateDirectiveExpression(IfNDefMacro); 564 CurPPLexer->LexingRawMode = true; 565 if (Callbacks) { 566 const SourceLocation CondEnd = CurPPLexer->getSourceLocation(); 567 Callbacks->Elif(Tok.getLocation(), 568 SourceRange(CondBegin, CondEnd), 569 (CondValue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False), CondInfo.IfLoc); 570 } 571 // If this condition is true, enter it! 572 if (CondValue) { 573 CondInfo.FoundNonSkip = true; 574 break; 575 } 576 } 577 } 578 } 579 580 CurPPLexer->ParsingPreprocessorDirective = false; 581 // Restore comment saving mode. 582 if (CurLexer) CurLexer->resetExtendedTokenMode(); 583 } 584 585 // Finally, if we are out of the conditional (saw an #endif or ran off the end 586 // of the file, just stop skipping and return to lexing whatever came after 587 // the #if block. 588 CurPPLexer->LexingRawMode = false; 589 590 if (Callbacks) { 591 SourceLocation BeginLoc = ElseLoc.isValid() ? ElseLoc : IfTokenLoc; 592 Callbacks->SourceRangeSkipped(SourceRange(BeginLoc, Tok.getLocation())); 593 } 594 } 595 596 void Preprocessor::PTHSkipExcludedConditionalBlock() { 597 while (true) { 598 assert(CurPTHLexer); 599 assert(CurPTHLexer->LexingRawMode == false); 600 601 // Skip to the next '#else', '#elif', or #endif. 602 if (CurPTHLexer->SkipBlock()) { 603 // We have reached an #endif. Both the '#' and 'endif' tokens 604 // have been consumed by the PTHLexer. Just pop off the condition level. 605 PPConditionalInfo CondInfo; 606 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 607 (void)InCond; // Silence warning in no-asserts mode. 608 assert(!InCond && "Can't be skipping if not in a conditional!"); 609 break; 610 } 611 612 // We have reached a '#else' or '#elif'. Lex the next token to get 613 // the directive flavor. 614 Token Tok; 615 LexUnexpandedToken(Tok); 616 617 // We can actually look up the IdentifierInfo here since we aren't in 618 // raw mode. 619 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 620 621 if (K == tok::pp_else) { 622 // #else: Enter the else condition. We aren't in a nested condition 623 // since we skip those. We're always in the one matching the last 624 // blocked we skipped. 625 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 626 // Note that we've seen a #else in this conditional. 627 CondInfo.FoundElse = true; 628 629 // If the #if block wasn't entered then enter the #else block now. 630 if (!CondInfo.FoundNonSkip) { 631 CondInfo.FoundNonSkip = true; 632 633 // Scan until the eod token. 634 CurPTHLexer->ParsingPreprocessorDirective = true; 635 DiscardUntilEndOfDirective(); 636 CurPTHLexer->ParsingPreprocessorDirective = false; 637 638 break; 639 } 640 641 // Otherwise skip this block. 642 continue; 643 } 644 645 assert(K == tok::pp_elif); 646 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 647 648 // If this is a #elif with a #else before it, report the error. 649 if (CondInfo.FoundElse) 650 Diag(Tok, diag::pp_err_elif_after_else); 651 652 // If this is in a skipping block or if we're already handled this #if 653 // block, don't bother parsing the condition. We just skip this block. 654 if (CondInfo.FoundNonSkip) 655 continue; 656 657 // Evaluate the condition of the #elif. 658 IdentifierInfo *IfNDefMacro = nullptr; 659 CurPTHLexer->ParsingPreprocessorDirective = true; 660 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 661 CurPTHLexer->ParsingPreprocessorDirective = false; 662 663 // If this condition is true, enter it! 664 if (ShouldEnter) { 665 CondInfo.FoundNonSkip = true; 666 break; 667 } 668 669 // Otherwise, skip this block and go to the next one. 670 } 671 } 672 673 Module *Preprocessor::getModuleForLocation(SourceLocation Loc) { 674 if (!SourceMgr.isInMainFile(Loc)) { 675 // Try to determine the module of the include directive. 676 // FIXME: Look into directly passing the FileEntry from LookupFile instead. 677 FileID IDOfIncl = SourceMgr.getFileID(SourceMgr.getExpansionLoc(Loc)); 678 if (const FileEntry *EntryOfIncl = SourceMgr.getFileEntryForID(IDOfIncl)) { 679 // The include comes from an included file. 680 return HeaderInfo.getModuleMap() 681 .findModuleForHeader(EntryOfIncl) 682 .getModule(); 683 } 684 } 685 686 // This is either in the main file or not in a file at all. It belongs 687 // to the current module, if there is one. 688 return getLangOpts().CurrentModule.empty() 689 ? nullptr 690 : HeaderInfo.lookupModule(getLangOpts().CurrentModule); 691 } 692 693 Module *Preprocessor::getModuleContainingLocation(SourceLocation Loc) { 694 return HeaderInfo.getModuleMap().inferModuleFromLocation( 695 FullSourceLoc(Loc, SourceMgr)); 696 } 697 698 const FileEntry * 699 Preprocessor::getModuleHeaderToIncludeForDiagnostics(SourceLocation IncLoc, 700 SourceLocation Loc) { 701 // If we have a module import syntax, we shouldn't include a header to 702 // make a particular module visible. 703 if (getLangOpts().ObjC2) 704 return nullptr; 705 706 // Figure out which module we'd want to import. 707 Module *M = getModuleContainingLocation(Loc); 708 if (!M) 709 return nullptr; 710 711 Module *TopM = M->getTopLevelModule(); 712 Module *IncM = getModuleForLocation(IncLoc); 713 714 // Walk up through the include stack, looking through textual headers of M 715 // until we hit a non-textual header that we can #include. (We assume textual 716 // headers of a module with non-textual headers aren't meant to be used to 717 // import entities from the module.) 718 auto &SM = getSourceManager(); 719 while (!Loc.isInvalid() && !SM.isInMainFile(Loc)) { 720 auto ID = SM.getFileID(SM.getExpansionLoc(Loc)); 721 auto *FE = SM.getFileEntryForID(ID); 722 723 bool InTextualHeader = false; 724 for (auto Header : HeaderInfo.getModuleMap().findAllModulesForHeader(FE)) { 725 if (!Header.getModule()->isSubModuleOf(TopM)) 726 continue; 727 728 if (!(Header.getRole() & ModuleMap::TextualHeader)) { 729 // If this is an accessible, non-textual header of M's top-level module 730 // that transitively includes the given location and makes the 731 // corresponding module visible, this is the thing to #include. 732 if (Header.isAccessibleFrom(IncM)) 733 return FE; 734 735 // It's in a private header; we can't #include it. 736 // FIXME: If there's a public header in some module that re-exports it, 737 // then we could suggest including that, but it's not clear that's the 738 // expected way to make this entity visible. 739 continue; 740 } 741 742 InTextualHeader = true; 743 } 744 745 if (!InTextualHeader) 746 break; 747 748 Loc = SM.getIncludeLoc(ID); 749 } 750 751 return nullptr; 752 } 753 754 const FileEntry *Preprocessor::LookupFile( 755 SourceLocation FilenameLoc, 756 StringRef Filename, 757 bool isAngled, 758 const DirectoryLookup *FromDir, 759 const FileEntry *FromFile, 760 const DirectoryLookup *&CurDir, 761 SmallVectorImpl<char> *SearchPath, 762 SmallVectorImpl<char> *RelativePath, 763 ModuleMap::KnownHeader *SuggestedModule, 764 bool SkipCache) { 765 Module *RequestingModule = getModuleForLocation(FilenameLoc); 766 bool RequestingModuleIsModuleInterface = !SourceMgr.isInMainFile(FilenameLoc); 767 768 // If the header lookup mechanism may be relative to the current inclusion 769 // stack, record the parent #includes. 770 SmallVector<std::pair<const FileEntry *, const DirectoryEntry *>, 16> 771 Includers; 772 bool BuildSystemModule = false; 773 if (!FromDir && !FromFile) { 774 FileID FID = getCurrentFileLexer()->getFileID(); 775 const FileEntry *FileEnt = SourceMgr.getFileEntryForID(FID); 776 777 // If there is no file entry associated with this file, it must be the 778 // predefines buffer or the module includes buffer. Any other file is not 779 // lexed with a normal lexer, so it won't be scanned for preprocessor 780 // directives. 781 // 782 // If we have the predefines buffer, resolve #include references (which come 783 // from the -include command line argument) from the current working 784 // directory instead of relative to the main file. 785 // 786 // If we have the module includes buffer, resolve #include references (which 787 // come from header declarations in the module map) relative to the module 788 // map file. 789 if (!FileEnt) { 790 if (FID == SourceMgr.getMainFileID() && MainFileDir) { 791 Includers.push_back(std::make_pair(nullptr, MainFileDir)); 792 BuildSystemModule = getCurrentModule()->IsSystem; 793 } else if ((FileEnt = 794 SourceMgr.getFileEntryForID(SourceMgr.getMainFileID()))) 795 Includers.push_back(std::make_pair(FileEnt, FileMgr.getDirectory("."))); 796 } else { 797 Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir())); 798 } 799 800 // MSVC searches the current include stack from top to bottom for 801 // headers included by quoted include directives. 802 // See: http://msdn.microsoft.com/en-us/library/36k2cdd4.aspx 803 if (LangOpts.MSVCCompat && !isAngled) { 804 for (IncludeStackInfo &ISEntry : llvm::reverse(IncludeMacroStack)) { 805 if (IsFileLexer(ISEntry)) 806 if ((FileEnt = ISEntry.ThePPLexer->getFileEntry())) 807 Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir())); 808 } 809 } 810 } 811 812 CurDir = CurDirLookup; 813 814 if (FromFile) { 815 // We're supposed to start looking from after a particular file. Search 816 // the include path until we find that file or run out of files. 817 const DirectoryLookup *TmpCurDir = CurDir; 818 const DirectoryLookup *TmpFromDir = nullptr; 819 while (const FileEntry *FE = HeaderInfo.LookupFile( 820 Filename, FilenameLoc, isAngled, TmpFromDir, TmpCurDir, 821 Includers, SearchPath, RelativePath, RequestingModule, 822 SuggestedModule, SkipCache)) { 823 // Keep looking as if this file did a #include_next. 824 TmpFromDir = TmpCurDir; 825 ++TmpFromDir; 826 if (FE == FromFile) { 827 // Found it. 828 FromDir = TmpFromDir; 829 CurDir = TmpCurDir; 830 break; 831 } 832 } 833 } 834 835 // Do a standard file entry lookup. 836 const FileEntry *FE = HeaderInfo.LookupFile( 837 Filename, FilenameLoc, isAngled, FromDir, CurDir, Includers, SearchPath, 838 RelativePath, RequestingModule, SuggestedModule, SkipCache, 839 BuildSystemModule); 840 if (FE) { 841 if (SuggestedModule && !LangOpts.AsmPreprocessor) 842 HeaderInfo.getModuleMap().diagnoseHeaderInclusion( 843 RequestingModule, RequestingModuleIsModuleInterface, FilenameLoc, 844 Filename, FE); 845 return FE; 846 } 847 848 const FileEntry *CurFileEnt; 849 // Otherwise, see if this is a subframework header. If so, this is relative 850 // to one of the headers on the #include stack. Walk the list of the current 851 // headers on the #include stack and pass them to HeaderInfo. 852 if (IsFileLexer()) { 853 if ((CurFileEnt = CurPPLexer->getFileEntry())) { 854 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt, 855 SearchPath, RelativePath, 856 RequestingModule, 857 SuggestedModule))) { 858 if (SuggestedModule && !LangOpts.AsmPreprocessor) 859 HeaderInfo.getModuleMap().diagnoseHeaderInclusion( 860 RequestingModule, RequestingModuleIsModuleInterface, FilenameLoc, 861 Filename, FE); 862 return FE; 863 } 864 } 865 } 866 867 for (IncludeStackInfo &ISEntry : llvm::reverse(IncludeMacroStack)) { 868 if (IsFileLexer(ISEntry)) { 869 if ((CurFileEnt = ISEntry.ThePPLexer->getFileEntry())) { 870 if ((FE = HeaderInfo.LookupSubframeworkHeader( 871 Filename, CurFileEnt, SearchPath, RelativePath, 872 RequestingModule, SuggestedModule))) { 873 if (SuggestedModule && !LangOpts.AsmPreprocessor) 874 HeaderInfo.getModuleMap().diagnoseHeaderInclusion( 875 RequestingModule, RequestingModuleIsModuleInterface, 876 FilenameLoc, Filename, FE); 877 return FE; 878 } 879 } 880 } 881 } 882 883 // Otherwise, we really couldn't find the file. 884 return nullptr; 885 } 886 887 //===----------------------------------------------------------------------===// 888 // Preprocessor Directive Handling. 889 //===----------------------------------------------------------------------===// 890 891 class Preprocessor::ResetMacroExpansionHelper { 892 public: 893 ResetMacroExpansionHelper(Preprocessor *pp) 894 : PP(pp), save(pp->DisableMacroExpansion) { 895 if (pp->MacroExpansionInDirectivesOverride) 896 pp->DisableMacroExpansion = false; 897 } 898 899 ~ResetMacroExpansionHelper() { 900 PP->DisableMacroExpansion = save; 901 } 902 903 private: 904 Preprocessor *PP; 905 bool save; 906 }; 907 908 /// HandleDirective - This callback is invoked when the lexer sees a # token 909 /// at the start of a line. This consumes the directive, modifies the 910 /// lexer/preprocessor state, and advances the lexer(s) so that the next token 911 /// read is the correct one. 912 void Preprocessor::HandleDirective(Token &Result) { 913 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 914 915 // We just parsed a # character at the start of a line, so we're in directive 916 // mode. Tell the lexer this so any newlines we see will be converted into an 917 // EOD token (which terminates the directive). 918 CurPPLexer->ParsingPreprocessorDirective = true; 919 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false); 920 921 bool ImmediatelyAfterTopLevelIfndef = 922 CurPPLexer->MIOpt.getImmediatelyAfterTopLevelIfndef(); 923 CurPPLexer->MIOpt.resetImmediatelyAfterTopLevelIfndef(); 924 925 ++NumDirectives; 926 927 // We are about to read a token. For the multiple-include optimization FA to 928 // work, we have to remember if we had read any tokens *before* this 929 // pp-directive. 930 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 931 932 // Save the '#' token in case we need to return it later. 933 Token SavedHash = Result; 934 935 // Read the next token, the directive flavor. This isn't expanded due to 936 // C99 6.10.3p8. 937 LexUnexpandedToken(Result); 938 939 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 940 // #define A(x) #x 941 // A(abc 942 // #warning blah 943 // def) 944 // If so, the user is relying on undefined behavior, emit a diagnostic. Do 945 // not support this for #include-like directives, since that can result in 946 // terrible diagnostics, and does not work in GCC. 947 if (InMacroArgs) { 948 if (IdentifierInfo *II = Result.getIdentifierInfo()) { 949 switch (II->getPPKeywordID()) { 950 case tok::pp_include: 951 case tok::pp_import: 952 case tok::pp_include_next: 953 case tok::pp___include_macros: 954 case tok::pp_pragma: 955 Diag(Result, diag::err_embedded_directive) << II->getName(); 956 DiscardUntilEndOfDirective(); 957 return; 958 default: 959 break; 960 } 961 } 962 Diag(Result, diag::ext_embedded_directive); 963 } 964 965 // Temporarily enable macro expansion if set so 966 // and reset to previous state when returning from this function. 967 ResetMacroExpansionHelper helper(this); 968 969 switch (Result.getKind()) { 970 case tok::eod: 971 return; // null directive. 972 case tok::code_completion: 973 if (CodeComplete) 974 CodeComplete->CodeCompleteDirective( 975 CurPPLexer->getConditionalStackDepth() > 0); 976 setCodeCompletionReached(); 977 return; 978 case tok::numeric_constant: // # 7 GNU line marker directive. 979 if (getLangOpts().AsmPreprocessor) 980 break; // # 4 is not a preprocessor directive in .S files. 981 return HandleDigitDirective(Result); 982 default: 983 IdentifierInfo *II = Result.getIdentifierInfo(); 984 if (!II) break; // Not an identifier. 985 986 // Ask what the preprocessor keyword ID is. 987 switch (II->getPPKeywordID()) { 988 default: break; 989 // C99 6.10.1 - Conditional Inclusion. 990 case tok::pp_if: 991 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 992 case tok::pp_ifdef: 993 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 994 case tok::pp_ifndef: 995 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 996 case tok::pp_elif: 997 return HandleElifDirective(Result); 998 case tok::pp_else: 999 return HandleElseDirective(Result); 1000 case tok::pp_endif: 1001 return HandleEndifDirective(Result); 1002 1003 // C99 6.10.2 - Source File Inclusion. 1004 case tok::pp_include: 1005 // Handle #include. 1006 return HandleIncludeDirective(SavedHash.getLocation(), Result); 1007 case tok::pp___include_macros: 1008 // Handle -imacros. 1009 return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result); 1010 1011 // C99 6.10.3 - Macro Replacement. 1012 case tok::pp_define: 1013 return HandleDefineDirective(Result, ImmediatelyAfterTopLevelIfndef); 1014 case tok::pp_undef: 1015 return HandleUndefDirective(); 1016 1017 // C99 6.10.4 - Line Control. 1018 case tok::pp_line: 1019 return HandleLineDirective(); 1020 1021 // C99 6.10.5 - Error Directive. 1022 case tok::pp_error: 1023 return HandleUserDiagnosticDirective(Result, false); 1024 1025 // C99 6.10.6 - Pragma Directive. 1026 case tok::pp_pragma: 1027 return HandlePragmaDirective(SavedHash.getLocation(), PIK_HashPragma); 1028 1029 // GNU Extensions. 1030 case tok::pp_import: 1031 return HandleImportDirective(SavedHash.getLocation(), Result); 1032 case tok::pp_include_next: 1033 return HandleIncludeNextDirective(SavedHash.getLocation(), Result); 1034 1035 case tok::pp_warning: 1036 Diag(Result, diag::ext_pp_warning_directive); 1037 return HandleUserDiagnosticDirective(Result, true); 1038 case tok::pp_ident: 1039 return HandleIdentSCCSDirective(Result); 1040 case tok::pp_sccs: 1041 return HandleIdentSCCSDirective(Result); 1042 case tok::pp_assert: 1043 //isExtension = true; // FIXME: implement #assert 1044 break; 1045 case tok::pp_unassert: 1046 //isExtension = true; // FIXME: implement #unassert 1047 break; 1048 1049 case tok::pp___public_macro: 1050 if (getLangOpts().Modules) 1051 return HandleMacroPublicDirective(Result); 1052 break; 1053 1054 case tok::pp___private_macro: 1055 if (getLangOpts().Modules) 1056 return HandleMacroPrivateDirective(); 1057 break; 1058 } 1059 break; 1060 } 1061 1062 // If this is a .S file, treat unknown # directives as non-preprocessor 1063 // directives. This is important because # may be a comment or introduce 1064 // various pseudo-ops. Just return the # token and push back the following 1065 // token to be lexed next time. 1066 if (getLangOpts().AsmPreprocessor) { 1067 auto Toks = llvm::make_unique<Token[]>(2); 1068 // Return the # and the token after it. 1069 Toks[0] = SavedHash; 1070 Toks[1] = Result; 1071 1072 // If the second token is a hashhash token, then we need to translate it to 1073 // unknown so the token lexer doesn't try to perform token pasting. 1074 if (Result.is(tok::hashhash)) 1075 Toks[1].setKind(tok::unknown); 1076 1077 // Enter this token stream so that we re-lex the tokens. Make sure to 1078 // enable macro expansion, in case the token after the # is an identifier 1079 // that is expanded. 1080 EnterTokenStream(std::move(Toks), 2, false); 1081 return; 1082 } 1083 1084 // If we reached here, the preprocessing token is not valid! 1085 Diag(Result, diag::err_pp_invalid_directive); 1086 1087 // Read the rest of the PP line. 1088 DiscardUntilEndOfDirective(); 1089 1090 // Okay, we're done parsing the directive. 1091 } 1092 1093 /// GetLineValue - Convert a numeric token into an unsigned value, emitting 1094 /// Diagnostic DiagID if it is invalid, and returning the value in Val. 1095 static bool GetLineValue(Token &DigitTok, unsigned &Val, 1096 unsigned DiagID, Preprocessor &PP, 1097 bool IsGNULineDirective=false) { 1098 if (DigitTok.isNot(tok::numeric_constant)) { 1099 PP.Diag(DigitTok, DiagID); 1100 1101 if (DigitTok.isNot(tok::eod)) 1102 PP.DiscardUntilEndOfDirective(); 1103 return true; 1104 } 1105 1106 SmallString<64> IntegerBuffer; 1107 IntegerBuffer.resize(DigitTok.getLength()); 1108 const char *DigitTokBegin = &IntegerBuffer[0]; 1109 bool Invalid = false; 1110 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid); 1111 if (Invalid) 1112 return true; 1113 1114 // Verify that we have a simple digit-sequence, and compute the value. This 1115 // is always a simple digit string computed in decimal, so we do this manually 1116 // here. 1117 Val = 0; 1118 for (unsigned i = 0; i != ActualLength; ++i) { 1119 // C++1y [lex.fcon]p1: 1120 // Optional separating single quotes in a digit-sequence are ignored 1121 if (DigitTokBegin[i] == '\'') 1122 continue; 1123 1124 if (!isDigit(DigitTokBegin[i])) { 1125 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i), 1126 diag::err_pp_line_digit_sequence) << IsGNULineDirective; 1127 PP.DiscardUntilEndOfDirective(); 1128 return true; 1129 } 1130 1131 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0'); 1132 if (NextVal < Val) { // overflow. 1133 PP.Diag(DigitTok, DiagID); 1134 PP.DiscardUntilEndOfDirective(); 1135 return true; 1136 } 1137 Val = NextVal; 1138 } 1139 1140 if (DigitTokBegin[0] == '0' && Val) 1141 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal) 1142 << IsGNULineDirective; 1143 1144 return false; 1145 } 1146 1147 /// \brief Handle a \#line directive: C99 6.10.4. 1148 /// 1149 /// The two acceptable forms are: 1150 /// \verbatim 1151 /// # line digit-sequence 1152 /// # line digit-sequence "s-char-sequence" 1153 /// \endverbatim 1154 void Preprocessor::HandleLineDirective() { 1155 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 1156 // expanded. 1157 Token DigitTok; 1158 Lex(DigitTok); 1159 1160 // Validate the number and convert it to an unsigned. 1161 unsigned LineNo; 1162 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this)) 1163 return; 1164 1165 if (LineNo == 0) 1166 Diag(DigitTok, diag::ext_pp_line_zero); 1167 1168 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 1169 // number greater than 2147483647". C90 requires that the line # be <= 32767. 1170 unsigned LineLimit = 32768U; 1171 if (LangOpts.C99 || LangOpts.CPlusPlus11) 1172 LineLimit = 2147483648U; 1173 if (LineNo >= LineLimit) 1174 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 1175 else if (LangOpts.CPlusPlus11 && LineNo >= 32768U) 1176 Diag(DigitTok, diag::warn_cxx98_compat_pp_line_too_big); 1177 1178 int FilenameID = -1; 1179 Token StrTok; 1180 Lex(StrTok); 1181 1182 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a 1183 // string followed by eod. 1184 if (StrTok.is(tok::eod)) 1185 ; // ok 1186 else if (StrTok.isNot(tok::string_literal)) { 1187 Diag(StrTok, diag::err_pp_line_invalid_filename); 1188 return DiscardUntilEndOfDirective(); 1189 } else if (StrTok.hasUDSuffix()) { 1190 Diag(StrTok, diag::err_invalid_string_udl); 1191 return DiscardUntilEndOfDirective(); 1192 } else { 1193 // Parse and validate the string, converting it into a unique ID. 1194 StringLiteralParser Literal(StrTok, *this); 1195 assert(Literal.isAscii() && "Didn't allow wide strings in"); 1196 if (Literal.hadError) 1197 return DiscardUntilEndOfDirective(); 1198 if (Literal.Pascal) { 1199 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 1200 return DiscardUntilEndOfDirective(); 1201 } 1202 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString()); 1203 1204 // Verify that there is nothing after the string, other than EOD. Because 1205 // of C99 6.10.4p5, macros that expand to empty tokens are ok. 1206 CheckEndOfDirective("line", true); 1207 } 1208 1209 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 1210 1211 if (Callbacks) 1212 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), 1213 PPCallbacks::RenameFile, 1214 SrcMgr::C_User); 1215 } 1216 1217 /// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 1218 /// marker directive. 1219 static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 1220 bool &IsSystemHeader, bool &IsExternCHeader, 1221 Preprocessor &PP) { 1222 unsigned FlagVal; 1223 Token FlagTok; 1224 PP.Lex(FlagTok); 1225 if (FlagTok.is(tok::eod)) return false; 1226 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 1227 return true; 1228 1229 if (FlagVal == 1) { 1230 IsFileEntry = true; 1231 1232 PP.Lex(FlagTok); 1233 if (FlagTok.is(tok::eod)) return false; 1234 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 1235 return true; 1236 } else if (FlagVal == 2) { 1237 IsFileExit = true; 1238 1239 SourceManager &SM = PP.getSourceManager(); 1240 // If we are leaving the current presumed file, check to make sure the 1241 // presumed include stack isn't empty! 1242 FileID CurFileID = 1243 SM.getDecomposedExpansionLoc(FlagTok.getLocation()).first; 1244 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 1245 if (PLoc.isInvalid()) 1246 return true; 1247 1248 // If there is no include loc (main file) or if the include loc is in a 1249 // different physical file, then we aren't in a "1" line marker flag region. 1250 SourceLocation IncLoc = PLoc.getIncludeLoc(); 1251 if (IncLoc.isInvalid() || 1252 SM.getDecomposedExpansionLoc(IncLoc).first != CurFileID) { 1253 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 1254 PP.DiscardUntilEndOfDirective(); 1255 return true; 1256 } 1257 1258 PP.Lex(FlagTok); 1259 if (FlagTok.is(tok::eod)) return false; 1260 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 1261 return true; 1262 } 1263 1264 // We must have 3 if there are still flags. 1265 if (FlagVal != 3) { 1266 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 1267 PP.DiscardUntilEndOfDirective(); 1268 return true; 1269 } 1270 1271 IsSystemHeader = true; 1272 1273 PP.Lex(FlagTok); 1274 if (FlagTok.is(tok::eod)) return false; 1275 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 1276 return true; 1277 1278 // We must have 4 if there is yet another flag. 1279 if (FlagVal != 4) { 1280 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 1281 PP.DiscardUntilEndOfDirective(); 1282 return true; 1283 } 1284 1285 IsExternCHeader = true; 1286 1287 PP.Lex(FlagTok); 1288 if (FlagTok.is(tok::eod)) return false; 1289 1290 // There are no more valid flags here. 1291 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 1292 PP.DiscardUntilEndOfDirective(); 1293 return true; 1294 } 1295 1296 /// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 1297 /// one of the following forms: 1298 /// 1299 /// # 42 1300 /// # 42 "file" ('1' | '2')? 1301 /// # 42 "file" ('1' | '2')? '3' '4'? 1302 /// 1303 void Preprocessor::HandleDigitDirective(Token &DigitTok) { 1304 // Validate the number and convert it to an unsigned. GNU does not have a 1305 // line # limit other than it fit in 32-bits. 1306 unsigned LineNo; 1307 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 1308 *this, true)) 1309 return; 1310 1311 Token StrTok; 1312 Lex(StrTok); 1313 1314 bool IsFileEntry = false, IsFileExit = false; 1315 bool IsSystemHeader = false, IsExternCHeader = false; 1316 int FilenameID = -1; 1317 1318 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a 1319 // string followed by eod. 1320 if (StrTok.is(tok::eod)) 1321 ; // ok 1322 else if (StrTok.isNot(tok::string_literal)) { 1323 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 1324 return DiscardUntilEndOfDirective(); 1325 } else if (StrTok.hasUDSuffix()) { 1326 Diag(StrTok, diag::err_invalid_string_udl); 1327 return DiscardUntilEndOfDirective(); 1328 } else { 1329 // Parse and validate the string, converting it into a unique ID. 1330 StringLiteralParser Literal(StrTok, *this); 1331 assert(Literal.isAscii() && "Didn't allow wide strings in"); 1332 if (Literal.hadError) 1333 return DiscardUntilEndOfDirective(); 1334 if (Literal.Pascal) { 1335 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 1336 return DiscardUntilEndOfDirective(); 1337 } 1338 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString()); 1339 1340 // If a filename was present, read any flags that are present. 1341 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 1342 IsSystemHeader, IsExternCHeader, *this)) 1343 return; 1344 } 1345 1346 // Create a line note with this information. 1347 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 1348 IsFileEntry, IsFileExit, 1349 IsSystemHeader, IsExternCHeader); 1350 1351 // If the preprocessor has callbacks installed, notify them of the #line 1352 // change. This is used so that the line marker comes out in -E mode for 1353 // example. 1354 if (Callbacks) { 1355 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 1356 if (IsFileEntry) 1357 Reason = PPCallbacks::EnterFile; 1358 else if (IsFileExit) 1359 Reason = PPCallbacks::ExitFile; 1360 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 1361 if (IsExternCHeader) 1362 FileKind = SrcMgr::C_ExternCSystem; 1363 else if (IsSystemHeader) 1364 FileKind = SrcMgr::C_System; 1365 1366 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind); 1367 } 1368 } 1369 1370 /// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 1371 /// 1372 void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 1373 bool isWarning) { 1374 // PTH doesn't emit #warning or #error directives. 1375 if (CurPTHLexer) 1376 return CurPTHLexer->DiscardToEndOfLine(); 1377 1378 // Read the rest of the line raw. We do this because we don't want macros 1379 // to be expanded and we don't require that the tokens be valid preprocessing 1380 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 1381 // collapse multiple consequtive white space between tokens, but this isn't 1382 // specified by the standard. 1383 SmallString<128> Message; 1384 CurLexer->ReadToEndOfLine(&Message); 1385 1386 // Find the first non-whitespace character, so that we can make the 1387 // diagnostic more succinct. 1388 StringRef Msg = StringRef(Message).ltrim(' '); 1389 1390 if (isWarning) 1391 Diag(Tok, diag::pp_hash_warning) << Msg; 1392 else 1393 Diag(Tok, diag::err_pp_hash_error) << Msg; 1394 } 1395 1396 /// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 1397 /// 1398 void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 1399 // Yes, this directive is an extension. 1400 Diag(Tok, diag::ext_pp_ident_directive); 1401 1402 // Read the string argument. 1403 Token StrTok; 1404 Lex(StrTok); 1405 1406 // If the token kind isn't a string, it's a malformed directive. 1407 if (StrTok.isNot(tok::string_literal) && 1408 StrTok.isNot(tok::wide_string_literal)) { 1409 Diag(StrTok, diag::err_pp_malformed_ident); 1410 if (StrTok.isNot(tok::eod)) 1411 DiscardUntilEndOfDirective(); 1412 return; 1413 } 1414 1415 if (StrTok.hasUDSuffix()) { 1416 Diag(StrTok, diag::err_invalid_string_udl); 1417 return DiscardUntilEndOfDirective(); 1418 } 1419 1420 // Verify that there is nothing after the string, other than EOD. 1421 CheckEndOfDirective("ident"); 1422 1423 if (Callbacks) { 1424 bool Invalid = false; 1425 std::string Str = getSpelling(StrTok, &Invalid); 1426 if (!Invalid) 1427 Callbacks->Ident(Tok.getLocation(), Str); 1428 } 1429 } 1430 1431 /// \brief Handle a #public directive. 1432 void Preprocessor::HandleMacroPublicDirective(Token &Tok) { 1433 Token MacroNameTok; 1434 ReadMacroName(MacroNameTok, MU_Undef); 1435 1436 // Error reading macro name? If so, diagnostic already issued. 1437 if (MacroNameTok.is(tok::eod)) 1438 return; 1439 1440 // Check to see if this is the last token on the #__public_macro line. 1441 CheckEndOfDirective("__public_macro"); 1442 1443 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 1444 // Okay, we finally have a valid identifier to undef. 1445 MacroDirective *MD = getLocalMacroDirective(II); 1446 1447 // If the macro is not defined, this is an error. 1448 if (!MD) { 1449 Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II; 1450 return; 1451 } 1452 1453 // Note that this macro has now been exported. 1454 appendMacroDirective(II, AllocateVisibilityMacroDirective( 1455 MacroNameTok.getLocation(), /*IsPublic=*/true)); 1456 } 1457 1458 /// \brief Handle a #private directive. 1459 void Preprocessor::HandleMacroPrivateDirective() { 1460 Token MacroNameTok; 1461 ReadMacroName(MacroNameTok, MU_Undef); 1462 1463 // Error reading macro name? If so, diagnostic already issued. 1464 if (MacroNameTok.is(tok::eod)) 1465 return; 1466 1467 // Check to see if this is the last token on the #__private_macro line. 1468 CheckEndOfDirective("__private_macro"); 1469 1470 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 1471 // Okay, we finally have a valid identifier to undef. 1472 MacroDirective *MD = getLocalMacroDirective(II); 1473 1474 // If the macro is not defined, this is an error. 1475 if (!MD) { 1476 Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II; 1477 return; 1478 } 1479 1480 // Note that this macro has now been marked private. 1481 appendMacroDirective(II, AllocateVisibilityMacroDirective( 1482 MacroNameTok.getLocation(), /*IsPublic=*/false)); 1483 } 1484 1485 //===----------------------------------------------------------------------===// 1486 // Preprocessor Include Directive Handling. 1487 //===----------------------------------------------------------------------===// 1488 1489 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 1490 /// checked and spelled filename, e.g. as an operand of \#include. This returns 1491 /// true if the input filename was in <>'s or false if it were in ""'s. The 1492 /// caller is expected to provide a buffer that is large enough to hold the 1493 /// spelling of the filename, but is also expected to handle the case when 1494 /// this method decides to use a different buffer. 1495 bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 1496 StringRef &Buffer) { 1497 // Get the text form of the filename. 1498 assert(!Buffer.empty() && "Can't have tokens with empty spellings!"); 1499 1500 // Make sure the filename is <x> or "x". 1501 bool isAngled; 1502 if (Buffer[0] == '<') { 1503 if (Buffer.back() != '>') { 1504 Diag(Loc, diag::err_pp_expects_filename); 1505 Buffer = StringRef(); 1506 return true; 1507 } 1508 isAngled = true; 1509 } else if (Buffer[0] == '"') { 1510 if (Buffer.back() != '"') { 1511 Diag(Loc, diag::err_pp_expects_filename); 1512 Buffer = StringRef(); 1513 return true; 1514 } 1515 isAngled = false; 1516 } else { 1517 Diag(Loc, diag::err_pp_expects_filename); 1518 Buffer = StringRef(); 1519 return true; 1520 } 1521 1522 // Diagnose #include "" as invalid. 1523 if (Buffer.size() <= 2) { 1524 Diag(Loc, diag::err_pp_empty_filename); 1525 Buffer = StringRef(); 1526 return true; 1527 } 1528 1529 // Skip the brackets. 1530 Buffer = Buffer.substr(1, Buffer.size()-2); 1531 return isAngled; 1532 } 1533 1534 // \brief Handle cases where the \#include name is expanded from a macro 1535 // as multiple tokens, which need to be glued together. 1536 // 1537 // This occurs for code like: 1538 // \code 1539 // \#define FOO <a/b.h> 1540 // \#include FOO 1541 // \endcode 1542 // because in this case, "<a/b.h>" is returned as 7 tokens, not one. 1543 // 1544 // This code concatenates and consumes tokens up to the '>' token. It returns 1545 // false if the > was found, otherwise it returns true if it finds and consumes 1546 // the EOD marker. 1547 bool Preprocessor::ConcatenateIncludeName(SmallString<128> &FilenameBuffer, 1548 SourceLocation &End) { 1549 Token CurTok; 1550 1551 Lex(CurTok); 1552 while (CurTok.isNot(tok::eod)) { 1553 End = CurTok.getLocation(); 1554 1555 // FIXME: Provide code completion for #includes. 1556 if (CurTok.is(tok::code_completion)) { 1557 setCodeCompletionReached(); 1558 Lex(CurTok); 1559 continue; 1560 } 1561 1562 // Append the spelling of this token to the buffer. If there was a space 1563 // before it, add it now. 1564 if (CurTok.hasLeadingSpace()) 1565 FilenameBuffer.push_back(' '); 1566 1567 // Get the spelling of the token, directly into FilenameBuffer if possible. 1568 size_t PreAppendSize = FilenameBuffer.size(); 1569 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 1570 1571 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 1572 unsigned ActualLen = getSpelling(CurTok, BufPtr); 1573 1574 // If the token was spelled somewhere else, copy it into FilenameBuffer. 1575 if (BufPtr != &FilenameBuffer[PreAppendSize]) 1576 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 1577 1578 // Resize FilenameBuffer to the correct size. 1579 if (CurTok.getLength() != ActualLen) 1580 FilenameBuffer.resize(PreAppendSize+ActualLen); 1581 1582 // If we found the '>' marker, return success. 1583 if (CurTok.is(tok::greater)) 1584 return false; 1585 1586 Lex(CurTok); 1587 } 1588 1589 // If we hit the eod marker, emit an error and return true so that the caller 1590 // knows the EOD has been read. 1591 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1592 return true; 1593 } 1594 1595 /// \brief Push a token onto the token stream containing an annotation. 1596 static void EnterAnnotationToken(Preprocessor &PP, 1597 SourceLocation Begin, SourceLocation End, 1598 tok::TokenKind Kind, void *AnnotationVal) { 1599 // FIXME: Produce this as the current token directly, rather than 1600 // allocating a new token for it. 1601 auto Tok = llvm::make_unique<Token[]>(1); 1602 Tok[0].startToken(); 1603 Tok[0].setKind(Kind); 1604 Tok[0].setLocation(Begin); 1605 Tok[0].setAnnotationEndLoc(End); 1606 Tok[0].setAnnotationValue(AnnotationVal); 1607 PP.EnterTokenStream(std::move(Tok), 1, true); 1608 } 1609 1610 /// \brief Produce a diagnostic informing the user that a #include or similar 1611 /// was implicitly treated as a module import. 1612 static void diagnoseAutoModuleImport( 1613 Preprocessor &PP, SourceLocation HashLoc, Token &IncludeTok, 1614 ArrayRef<std::pair<IdentifierInfo *, SourceLocation>> Path, 1615 SourceLocation PathEnd) { 1616 assert(PP.getLangOpts().ObjC2 && "no import syntax available"); 1617 1618 SmallString<128> PathString; 1619 for (size_t I = 0, N = Path.size(); I != N; ++I) { 1620 if (I) 1621 PathString += '.'; 1622 PathString += Path[I].first->getName(); 1623 } 1624 int IncludeKind = 0; 1625 1626 switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) { 1627 case tok::pp_include: 1628 IncludeKind = 0; 1629 break; 1630 1631 case tok::pp_import: 1632 IncludeKind = 1; 1633 break; 1634 1635 case tok::pp_include_next: 1636 IncludeKind = 2; 1637 break; 1638 1639 case tok::pp___include_macros: 1640 IncludeKind = 3; 1641 break; 1642 1643 default: 1644 llvm_unreachable("unknown include directive kind"); 1645 } 1646 1647 CharSourceRange ReplaceRange(SourceRange(HashLoc, PathEnd), 1648 /*IsTokenRange=*/false); 1649 PP.Diag(HashLoc, diag::warn_auto_module_import) 1650 << IncludeKind << PathString 1651 << FixItHint::CreateReplacement(ReplaceRange, 1652 ("@import " + PathString + ";").str()); 1653 } 1654 1655 // Given a vector of path components and a string containing the real 1656 // path to the file, build a properly-cased replacement in the vector, 1657 // and return true if the replacement should be suggested. 1658 static bool trySimplifyPath(SmallVectorImpl<StringRef> &Components, 1659 StringRef RealPathName) { 1660 auto RealPathComponentIter = llvm::sys::path::rbegin(RealPathName); 1661 auto RealPathComponentEnd = llvm::sys::path::rend(RealPathName); 1662 int Cnt = 0; 1663 bool SuggestReplacement = false; 1664 // Below is a best-effort to handle ".." in paths. It is admittedly 1665 // not 100% correct in the presence of symlinks. 1666 for (auto &Component : llvm::reverse(Components)) { 1667 if ("." == Component) { 1668 } else if (".." == Component) { 1669 ++Cnt; 1670 } else if (Cnt) { 1671 --Cnt; 1672 } else if (RealPathComponentIter != RealPathComponentEnd) { 1673 if (Component != *RealPathComponentIter) { 1674 // If these path components differ by more than just case, then we 1675 // may be looking at symlinked paths. Bail on this diagnostic to avoid 1676 // noisy false positives. 1677 SuggestReplacement = RealPathComponentIter->equals_lower(Component); 1678 if (!SuggestReplacement) 1679 break; 1680 Component = *RealPathComponentIter; 1681 } 1682 ++RealPathComponentIter; 1683 } 1684 } 1685 return SuggestReplacement; 1686 } 1687 1688 /// HandleIncludeDirective - The "\#include" tokens have just been read, read 1689 /// the file to be included from the lexer, then include it! This is a common 1690 /// routine with functionality shared between \#include, \#include_next and 1691 /// \#import. LookupFrom is set when this is a \#include_next directive, it 1692 /// specifies the file to start searching from. 1693 void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc, 1694 Token &IncludeTok, 1695 const DirectoryLookup *LookupFrom, 1696 const FileEntry *LookupFromFile, 1697 bool isImport) { 1698 Token FilenameTok; 1699 CurPPLexer->LexIncludeFilename(FilenameTok); 1700 1701 // Reserve a buffer to get the spelling. 1702 SmallString<128> FilenameBuffer; 1703 StringRef Filename; 1704 SourceLocation End; 1705 SourceLocation CharEnd; // the end of this directive, in characters 1706 1707 switch (FilenameTok.getKind()) { 1708 case tok::eod: 1709 // If the token kind is EOD, the error has already been diagnosed. 1710 return; 1711 1712 case tok::angle_string_literal: 1713 case tok::string_literal: 1714 Filename = getSpelling(FilenameTok, FilenameBuffer); 1715 End = FilenameTok.getLocation(); 1716 CharEnd = End.getLocWithOffset(FilenameTok.getLength()); 1717 break; 1718 1719 case tok::less: 1720 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1721 // case, glue the tokens together into FilenameBuffer and interpret those. 1722 FilenameBuffer.push_back('<'); 1723 if (ConcatenateIncludeName(FilenameBuffer, End)) 1724 return; // Found <eod> but no ">"? Diagnostic already emitted. 1725 Filename = FilenameBuffer; 1726 CharEnd = End.getLocWithOffset(1); 1727 break; 1728 default: 1729 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1730 DiscardUntilEndOfDirective(); 1731 return; 1732 } 1733 1734 CharSourceRange FilenameRange 1735 = CharSourceRange::getCharRange(FilenameTok.getLocation(), CharEnd); 1736 StringRef OriginalFilename = Filename; 1737 bool isAngled = 1738 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename); 1739 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1740 // error. 1741 if (Filename.empty()) { 1742 DiscardUntilEndOfDirective(); 1743 return; 1744 } 1745 1746 // Verify that there is nothing after the filename, other than EOD. Note that 1747 // we allow macros that expand to nothing after the filename, because this 1748 // falls into the category of "#include pp-tokens new-line" specified in 1749 // C99 6.10.2p4. 1750 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1751 1752 // Check that we don't have infinite #include recursion. 1753 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1754 Diag(FilenameTok, diag::err_pp_include_too_deep); 1755 return; 1756 } 1757 1758 // Complain about attempts to #include files in an audit pragma. 1759 if (PragmaARCCFCodeAuditedLoc.isValid()) { 1760 Diag(HashLoc, diag::err_pp_include_in_arc_cf_code_audited); 1761 Diag(PragmaARCCFCodeAuditedLoc, diag::note_pragma_entered_here); 1762 1763 // Immediately leave the pragma. 1764 PragmaARCCFCodeAuditedLoc = SourceLocation(); 1765 } 1766 1767 // Complain about attempts to #include files in an assume-nonnull pragma. 1768 if (PragmaAssumeNonNullLoc.isValid()) { 1769 Diag(HashLoc, diag::err_pp_include_in_assume_nonnull); 1770 Diag(PragmaAssumeNonNullLoc, diag::note_pragma_entered_here); 1771 1772 // Immediately leave the pragma. 1773 PragmaAssumeNonNullLoc = SourceLocation(); 1774 } 1775 1776 if (HeaderInfo.HasIncludeAliasMap()) { 1777 // Map the filename with the brackets still attached. If the name doesn't 1778 // map to anything, fall back on the filename we've already gotten the 1779 // spelling for. 1780 StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(OriginalFilename); 1781 if (!NewName.empty()) 1782 Filename = NewName; 1783 } 1784 1785 // Search include directories. 1786 const DirectoryLookup *CurDir; 1787 SmallString<1024> SearchPath; 1788 SmallString<1024> RelativePath; 1789 // We get the raw path only if we have 'Callbacks' to which we later pass 1790 // the path. 1791 ModuleMap::KnownHeader SuggestedModule; 1792 SourceLocation FilenameLoc = FilenameTok.getLocation(); 1793 SmallString<128> NormalizedPath; 1794 if (LangOpts.MSVCCompat) { 1795 NormalizedPath = Filename.str(); 1796 #ifndef LLVM_ON_WIN32 1797 llvm::sys::path::native(NormalizedPath); 1798 #endif 1799 } 1800 const FileEntry *File = LookupFile( 1801 FilenameLoc, LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, 1802 isAngled, LookupFrom, LookupFromFile, CurDir, 1803 Callbacks ? &SearchPath : nullptr, Callbacks ? &RelativePath : nullptr, 1804 &SuggestedModule); 1805 1806 if (!File) { 1807 if (Callbacks) { 1808 // Give the clients a chance to recover. 1809 SmallString<128> RecoveryPath; 1810 if (Callbacks->FileNotFound(Filename, RecoveryPath)) { 1811 if (const DirectoryEntry *DE = FileMgr.getDirectory(RecoveryPath)) { 1812 // Add the recovery path to the list of search paths. 1813 DirectoryLookup DL(DE, SrcMgr::C_User, false); 1814 HeaderInfo.AddSearchPath(DL, isAngled); 1815 1816 // Try the lookup again, skipping the cache. 1817 File = LookupFile( 1818 FilenameLoc, 1819 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, isAngled, 1820 LookupFrom, LookupFromFile, CurDir, nullptr, nullptr, 1821 &SuggestedModule, /*SkipCache*/ true); 1822 } 1823 } 1824 } 1825 1826 if (!SuppressIncludeNotFoundError) { 1827 // If the file could not be located and it was included via angle 1828 // brackets, we can attempt a lookup as though it were a quoted path to 1829 // provide the user with a possible fixit. 1830 if (isAngled) { 1831 File = LookupFile( 1832 FilenameLoc, 1833 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, false, 1834 LookupFrom, LookupFromFile, CurDir, 1835 Callbacks ? &SearchPath : nullptr, 1836 Callbacks ? &RelativePath : nullptr, 1837 &SuggestedModule); 1838 if (File) { 1839 SourceRange Range(FilenameTok.getLocation(), CharEnd); 1840 Diag(FilenameTok, diag::err_pp_file_not_found_not_fatal) << 1841 Filename << 1842 FixItHint::CreateReplacement(Range, "\"" + Filename.str() + "\""); 1843 } 1844 } 1845 1846 // If the file is still not found, just go with the vanilla diagnostic 1847 if (!File) 1848 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename 1849 << FilenameRange; 1850 } 1851 } 1852 1853 // Should we enter the source file? Set to false if either the source file is 1854 // known to have no effect beyond its effect on module visibility -- that is, 1855 // if it's got an include guard that is already defined or is a modular header 1856 // we've imported or already built. 1857 bool ShouldEnter = true; 1858 1859 // Determine whether we should try to import the module for this #include, if 1860 // there is one. Don't do so if precompiled module support is disabled or we 1861 // are processing this module textually (because we're building the module). 1862 if (File && SuggestedModule && getLangOpts().Modules && 1863 SuggestedModule.getModule()->getTopLevelModuleName() != 1864 getLangOpts().CurrentModule) { 1865 // If this include corresponds to a module but that module is 1866 // unavailable, diagnose the situation and bail out. 1867 // FIXME: Remove this; loadModule does the same check (but produces 1868 // slightly worse diagnostics). 1869 if (!SuggestedModule.getModule()->isAvailable()) { 1870 Module::Requirement Requirement; 1871 Module::UnresolvedHeaderDirective MissingHeader; 1872 Module *M = SuggestedModule.getModule(); 1873 // Identify the cause. 1874 (void)M->isAvailable(getLangOpts(), getTargetInfo(), Requirement, 1875 MissingHeader); 1876 if (MissingHeader.FileNameLoc.isValid()) { 1877 Diag(MissingHeader.FileNameLoc, diag::err_module_header_missing) 1878 << MissingHeader.IsUmbrella << MissingHeader.FileName; 1879 } else { 1880 Diag(M->DefinitionLoc, diag::err_module_unavailable) 1881 << M->getFullModuleName() << Requirement.second << Requirement.first; 1882 } 1883 Diag(FilenameTok.getLocation(), 1884 diag::note_implicit_top_level_module_import_here) 1885 << M->getTopLevelModuleName(); 1886 return; 1887 } 1888 1889 // Compute the module access path corresponding to this module. 1890 // FIXME: Should we have a second loadModule() overload to avoid this 1891 // extra lookup step? 1892 SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> Path; 1893 for (Module *Mod = SuggestedModule.getModule(); Mod; Mod = Mod->Parent) 1894 Path.push_back(std::make_pair(getIdentifierInfo(Mod->Name), 1895 FilenameTok.getLocation())); 1896 std::reverse(Path.begin(), Path.end()); 1897 1898 // Warn that we're replacing the include/import with a module import. 1899 // We only do this in Objective-C, where we have a module-import syntax. 1900 if (getLangOpts().ObjC2) 1901 diagnoseAutoModuleImport(*this, HashLoc, IncludeTok, Path, CharEnd); 1902 1903 // Load the module to import its macros. We'll make the declarations 1904 // visible when the parser gets here. 1905 // FIXME: Pass SuggestedModule in here rather than converting it to a path 1906 // and making the module loader convert it back again. 1907 ModuleLoadResult Imported = TheModuleLoader.loadModule( 1908 IncludeTok.getLocation(), Path, Module::Hidden, 1909 /*IsIncludeDirective=*/true); 1910 assert((Imported == nullptr || Imported == SuggestedModule.getModule()) && 1911 "the imported module is different than the suggested one"); 1912 1913 if (Imported) 1914 ShouldEnter = false; 1915 else if (Imported.isMissingExpected()) { 1916 // We failed to find a submodule that we assumed would exist (because it 1917 // was in the directory of an umbrella header, for instance), but no 1918 // actual module containing it exists (because the umbrella header is 1919 // incomplete). Treat this as a textual inclusion. 1920 SuggestedModule = ModuleMap::KnownHeader(); 1921 } else if (Imported.isConfigMismatch()) { 1922 // On a configuration mismatch, enter the header textually. We still know 1923 // that it's part of the corresponding module. 1924 } else { 1925 // We hit an error processing the import. Bail out. 1926 if (hadModuleLoaderFatalFailure()) { 1927 // With a fatal failure in the module loader, we abort parsing. 1928 Token &Result = IncludeTok; 1929 if (CurLexer) { 1930 Result.startToken(); 1931 CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof); 1932 CurLexer->cutOffLexing(); 1933 } else { 1934 assert(CurPTHLexer && "#include but no current lexer set!"); 1935 CurPTHLexer->getEOF(Result); 1936 } 1937 } 1938 return; 1939 } 1940 } 1941 1942 if (Callbacks) { 1943 // Notify the callback object that we've seen an inclusion directive. 1944 Callbacks->InclusionDirective( 1945 HashLoc, IncludeTok, 1946 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, isAngled, 1947 FilenameRange, File, SearchPath, RelativePath, 1948 ShouldEnter ? nullptr : SuggestedModule.getModule()); 1949 } 1950 1951 if (!File) 1952 return; 1953 1954 // The #included file will be considered to be a system header if either it is 1955 // in a system include directory, or if the #includer is a system include 1956 // header. 1957 SrcMgr::CharacteristicKind FileCharacter = 1958 std::max(HeaderInfo.getFileDirFlavor(File), 1959 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1960 1961 // FIXME: If we have a suggested module, and we've already visited this file, 1962 // don't bother entering it again. We know it has no further effect. 1963 1964 // Issue a diagnostic if the name of the file on disk has a different case 1965 // than the one we're about to open. 1966 const bool CheckIncludePathPortability = 1967 File && !File->tryGetRealPathName().empty(); 1968 1969 if (CheckIncludePathPortability) { 1970 StringRef Name = LangOpts.MSVCCompat ? NormalizedPath.str() : Filename; 1971 StringRef RealPathName = File->tryGetRealPathName(); 1972 SmallVector<StringRef, 16> Components(llvm::sys::path::begin(Name), 1973 llvm::sys::path::end(Name)); 1974 1975 if (trySimplifyPath(Components, RealPathName)) { 1976 SmallString<128> Path; 1977 Path.reserve(Name.size()+2); 1978 Path.push_back(isAngled ? '<' : '"'); 1979 for (auto Component : Components) { 1980 Path.append(Component); 1981 // Append the separator the user used, or the close quote 1982 Path.push_back( 1983 Path.size() <= Filename.size() ? Filename[Path.size()-1] : 1984 (isAngled ? '>' : '"')); 1985 } 1986 auto Replacement = Path.str().str(); 1987 // For user files and known standard headers, by default we issue a diagnostic. 1988 // For other system headers, we don't. They can be controlled separately. 1989 auto DiagId = (FileCharacter == SrcMgr::C_User || warnByDefaultOnWrongCase(Name)) ? 1990 diag::pp_nonportable_path : diag::pp_nonportable_system_path; 1991 SourceRange Range(FilenameTok.getLocation(), CharEnd); 1992 Diag(FilenameTok, DiagId) << Replacement << 1993 FixItHint::CreateReplacement(Range, Replacement); 1994 } 1995 } 1996 1997 // Ask HeaderInfo if we should enter this #include file. If not, #including 1998 // this file will have no effect. 1999 if (ShouldEnter && 2000 !HeaderInfo.ShouldEnterIncludeFile(*this, File, isImport, 2001 SuggestedModule.getModule())) { 2002 ShouldEnter = false; 2003 if (Callbacks) 2004 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter); 2005 } 2006 2007 // If we don't need to enter the file, stop now. 2008 if (!ShouldEnter) { 2009 // If this is a module import, make it visible if needed. 2010 if (auto *M = SuggestedModule.getModule()) { 2011 makeModuleVisible(M, HashLoc); 2012 2013 if (IncludeTok.getIdentifierInfo()->getPPKeywordID() != 2014 tok::pp___include_macros) 2015 EnterAnnotationToken(*this, HashLoc, End, tok::annot_module_include, M); 2016 } 2017 return; 2018 } 2019 2020 // Look up the file, create a File ID for it. 2021 SourceLocation IncludePos = End; 2022 // If the filename string was the result of macro expansions, set the include 2023 // position on the file where it will be included and after the expansions. 2024 if (IncludePos.isMacroID()) 2025 IncludePos = SourceMgr.getExpansionRange(IncludePos).second; 2026 FileID FID = SourceMgr.createFileID(File, IncludePos, FileCharacter); 2027 assert(FID.isValid() && "Expected valid file ID"); 2028 2029 // If all is good, enter the new file! 2030 if (EnterSourceFile(FID, CurDir, FilenameTok.getLocation())) 2031 return; 2032 2033 // Determine if we're switching to building a new submodule, and which one. 2034 if (auto *M = SuggestedModule.getModule()) { 2035 assert(!CurSubmodule && "should not have marked this as a module yet"); 2036 CurSubmodule = M; 2037 2038 // Let the macro handling code know that any future macros are within 2039 // the new submodule. 2040 EnterSubmodule(M, HashLoc); 2041 2042 // Let the parser know that any future declarations are within the new 2043 // submodule. 2044 // FIXME: There's no point doing this if we're handling a #__include_macros 2045 // directive. 2046 EnterAnnotationToken(*this, HashLoc, End, tok::annot_module_begin, M); 2047 } 2048 } 2049 2050 /// HandleIncludeNextDirective - Implements \#include_next. 2051 /// 2052 void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc, 2053 Token &IncludeNextTok) { 2054 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 2055 2056 // #include_next is like #include, except that we start searching after 2057 // the current found directory. If we can't do this, issue a 2058 // diagnostic. 2059 const DirectoryLookup *Lookup = CurDirLookup; 2060 const FileEntry *LookupFromFile = nullptr; 2061 if (isInPrimaryFile() && LangOpts.IsHeaderFile) { 2062 // If the main file is a header, then it's either for PCH/AST generation, 2063 // or libclang opened it. Either way, handle it as a normal include below 2064 // and do not complain about include_next. 2065 } else if (isInPrimaryFile()) { 2066 Lookup = nullptr; 2067 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 2068 } else if (CurSubmodule) { 2069 // Start looking up in the directory *after* the one in which the current 2070 // file would be found, if any. 2071 assert(CurPPLexer && "#include_next directive in macro?"); 2072 LookupFromFile = CurPPLexer->getFileEntry(); 2073 Lookup = nullptr; 2074 } else if (!Lookup) { 2075 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 2076 } else { 2077 // Start looking up in the next directory. 2078 ++Lookup; 2079 } 2080 2081 return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup, 2082 LookupFromFile); 2083 } 2084 2085 /// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode 2086 void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) { 2087 // The Microsoft #import directive takes a type library and generates header 2088 // files from it, and includes those. This is beyond the scope of what clang 2089 // does, so we ignore it and error out. However, #import can optionally have 2090 // trailing attributes that span multiple lines. We're going to eat those 2091 // so we can continue processing from there. 2092 Diag(Tok, diag::err_pp_import_directive_ms ); 2093 2094 // Read tokens until we get to the end of the directive. Note that the 2095 // directive can be split over multiple lines using the backslash character. 2096 DiscardUntilEndOfDirective(); 2097 } 2098 2099 /// HandleImportDirective - Implements \#import. 2100 /// 2101 void Preprocessor::HandleImportDirective(SourceLocation HashLoc, 2102 Token &ImportTok) { 2103 if (!LangOpts.ObjC1) { // #import is standard for ObjC. 2104 if (LangOpts.MSVCCompat) 2105 return HandleMicrosoftImportDirective(ImportTok); 2106 Diag(ImportTok, diag::ext_pp_import_directive); 2107 } 2108 return HandleIncludeDirective(HashLoc, ImportTok, nullptr, nullptr, true); 2109 } 2110 2111 /// HandleIncludeMacrosDirective - The -imacros command line option turns into a 2112 /// pseudo directive in the predefines buffer. This handles it by sucking all 2113 /// tokens through the preprocessor and discarding them (only keeping the side 2114 /// effects on the preprocessor). 2115 void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc, 2116 Token &IncludeMacrosTok) { 2117 // This directive should only occur in the predefines buffer. If not, emit an 2118 // error and reject it. 2119 SourceLocation Loc = IncludeMacrosTok.getLocation(); 2120 if (SourceMgr.getBufferName(Loc) != "<built-in>") { 2121 Diag(IncludeMacrosTok.getLocation(), 2122 diag::pp_include_macros_out_of_predefines); 2123 DiscardUntilEndOfDirective(); 2124 return; 2125 } 2126 2127 // Treat this as a normal #include for checking purposes. If this is 2128 // successful, it will push a new lexer onto the include stack. 2129 HandleIncludeDirective(HashLoc, IncludeMacrosTok); 2130 2131 Token TmpTok; 2132 do { 2133 Lex(TmpTok); 2134 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 2135 } while (TmpTok.isNot(tok::hashhash)); 2136 } 2137 2138 //===----------------------------------------------------------------------===// 2139 // Preprocessor Macro Directive Handling. 2140 //===----------------------------------------------------------------------===// 2141 2142 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 2143 /// definition has just been read. Lex the rest of the arguments and the 2144 /// closing ), updating MI with what we learn. Return true if an error occurs 2145 /// parsing the arg list. 2146 bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI, Token &Tok) { 2147 SmallVector<IdentifierInfo*, 32> Arguments; 2148 2149 while (true) { 2150 LexUnexpandedToken(Tok); 2151 switch (Tok.getKind()) { 2152 case tok::r_paren: 2153 // Found the end of the argument list. 2154 if (Arguments.empty()) // #define FOO() 2155 return false; 2156 // Otherwise we have #define FOO(A,) 2157 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 2158 return true; 2159 case tok::ellipsis: // #define X(... -> C99 varargs 2160 if (!LangOpts.C99) 2161 Diag(Tok, LangOpts.CPlusPlus11 ? 2162 diag::warn_cxx98_compat_variadic_macro : 2163 diag::ext_variadic_macro); 2164 2165 // OpenCL v1.2 s6.9.e: variadic macros are not supported. 2166 if (LangOpts.OpenCL) { 2167 Diag(Tok, diag::err_pp_opencl_variadic_macros); 2168 return true; 2169 } 2170 2171 // Lex the token after the identifier. 2172 LexUnexpandedToken(Tok); 2173 if (Tok.isNot(tok::r_paren)) { 2174 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 2175 return true; 2176 } 2177 // Add the __VA_ARGS__ identifier as an argument. 2178 Arguments.push_back(Ident__VA_ARGS__); 2179 MI->setIsC99Varargs(); 2180 MI->setArgumentList(Arguments, BP); 2181 return false; 2182 case tok::eod: // #define X( 2183 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 2184 return true; 2185 default: 2186 // Handle keywords and identifiers here to accept things like 2187 // #define Foo(for) for. 2188 IdentifierInfo *II = Tok.getIdentifierInfo(); 2189 if (!II) { 2190 // #define X(1 2191 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 2192 return true; 2193 } 2194 2195 // If this is already used as an argument, it is used multiple times (e.g. 2196 // #define X(A,A. 2197 if (std::find(Arguments.begin(), Arguments.end(), II) != 2198 Arguments.end()) { // C99 6.10.3p6 2199 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 2200 return true; 2201 } 2202 2203 // Add the argument to the macro info. 2204 Arguments.push_back(II); 2205 2206 // Lex the token after the identifier. 2207 LexUnexpandedToken(Tok); 2208 2209 switch (Tok.getKind()) { 2210 default: // #define X(A B 2211 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 2212 return true; 2213 case tok::r_paren: // #define X(A) 2214 MI->setArgumentList(Arguments, BP); 2215 return false; 2216 case tok::comma: // #define X(A, 2217 break; 2218 case tok::ellipsis: // #define X(A... -> GCC extension 2219 // Diagnose extension. 2220 Diag(Tok, diag::ext_named_variadic_macro); 2221 2222 // Lex the token after the identifier. 2223 LexUnexpandedToken(Tok); 2224 if (Tok.isNot(tok::r_paren)) { 2225 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 2226 return true; 2227 } 2228 2229 MI->setIsGNUVarargs(); 2230 MI->setArgumentList(Arguments, BP); 2231 return false; 2232 } 2233 } 2234 } 2235 } 2236 2237 static bool isConfigurationPattern(Token &MacroName, MacroInfo *MI, 2238 const LangOptions &LOptions) { 2239 if (MI->getNumTokens() == 1) { 2240 const Token &Value = MI->getReplacementToken(0); 2241 2242 // Macro that is identity, like '#define inline inline' is a valid pattern. 2243 if (MacroName.getKind() == Value.getKind()) 2244 return true; 2245 2246 // Macro that maps a keyword to the same keyword decorated with leading/ 2247 // trailing underscores is a valid pattern: 2248 // #define inline __inline 2249 // #define inline __inline__ 2250 // #define inline _inline (in MS compatibility mode) 2251 StringRef MacroText = MacroName.getIdentifierInfo()->getName(); 2252 if (IdentifierInfo *II = Value.getIdentifierInfo()) { 2253 if (!II->isKeyword(LOptions)) 2254 return false; 2255 StringRef ValueText = II->getName(); 2256 StringRef TrimmedValue = ValueText; 2257 if (!ValueText.startswith("__")) { 2258 if (ValueText.startswith("_")) 2259 TrimmedValue = TrimmedValue.drop_front(1); 2260 else 2261 return false; 2262 } else { 2263 TrimmedValue = TrimmedValue.drop_front(2); 2264 if (TrimmedValue.endswith("__")) 2265 TrimmedValue = TrimmedValue.drop_back(2); 2266 } 2267 return TrimmedValue.equals(MacroText); 2268 } else { 2269 return false; 2270 } 2271 } 2272 2273 // #define inline 2274 return MacroName.isOneOf(tok::kw_extern, tok::kw_inline, tok::kw_static, 2275 tok::kw_const) && 2276 MI->getNumTokens() == 0; 2277 } 2278 2279 /// HandleDefineDirective - Implements \#define. This consumes the entire macro 2280 /// line then lets the caller lex the next real token. 2281 void Preprocessor::HandleDefineDirective(Token &DefineTok, 2282 bool ImmediatelyAfterHeaderGuard) { 2283 ++NumDefined; 2284 2285 Token MacroNameTok; 2286 bool MacroShadowsKeyword; 2287 ReadMacroName(MacroNameTok, MU_Define, &MacroShadowsKeyword); 2288 2289 // Error reading macro name? If so, diagnostic already issued. 2290 if (MacroNameTok.is(tok::eod)) 2291 return; 2292 2293 Token LastTok = MacroNameTok; 2294 2295 // If we are supposed to keep comments in #defines, reenable comment saving 2296 // mode. 2297 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 2298 2299 // Create the new macro. 2300 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 2301 2302 Token Tok; 2303 LexUnexpandedToken(Tok); 2304 2305 // If this is a function-like macro definition, parse the argument list, 2306 // marking each of the identifiers as being used as macro arguments. Also, 2307 // check other constraints on the first token of the macro body. 2308 if (Tok.is(tok::eod)) { 2309 if (ImmediatelyAfterHeaderGuard) { 2310 // Save this macro information since it may part of a header guard. 2311 CurPPLexer->MIOpt.SetDefinedMacro(MacroNameTok.getIdentifierInfo(), 2312 MacroNameTok.getLocation()); 2313 } 2314 // If there is no body to this macro, we have no special handling here. 2315 } else if (Tok.hasLeadingSpace()) { 2316 // This is a normal token with leading space. Clear the leading space 2317 // marker on the first token to get proper expansion. 2318 Tok.clearFlag(Token::LeadingSpace); 2319 } else if (Tok.is(tok::l_paren)) { 2320 // This is a function-like macro definition. Read the argument list. 2321 MI->setIsFunctionLike(); 2322 if (ReadMacroDefinitionArgList(MI, LastTok)) { 2323 // Throw away the rest of the line. 2324 if (CurPPLexer->ParsingPreprocessorDirective) 2325 DiscardUntilEndOfDirective(); 2326 return; 2327 } 2328 2329 // If this is a definition of a variadic C99 function-like macro, not using 2330 // the GNU named varargs extension, enabled __VA_ARGS__. 2331 2332 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 2333 // This gets unpoisoned where it is allowed. 2334 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 2335 if (MI->isC99Varargs()) 2336 Ident__VA_ARGS__->setIsPoisoned(false); 2337 2338 // Read the first token after the arg list for down below. 2339 LexUnexpandedToken(Tok); 2340 } else if (LangOpts.C99 || LangOpts.CPlusPlus11) { 2341 // C99 requires whitespace between the macro definition and the body. Emit 2342 // a diagnostic for something like "#define X+". 2343 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 2344 } else { 2345 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 2346 // first character of a replacement list is not a character required by 2347 // subclause 5.2.1, then there shall be white-space separation between the 2348 // identifier and the replacement list.". 5.2.1 lists this set: 2349 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 2350 // is irrelevant here. 2351 bool isInvalid = false; 2352 if (Tok.is(tok::at)) // @ is not in the list above. 2353 isInvalid = true; 2354 else if (Tok.is(tok::unknown)) { 2355 // If we have an unknown token, it is something strange like "`". Since 2356 // all of valid characters would have lexed into a single character 2357 // token of some sort, we know this is not a valid case. 2358 isInvalid = true; 2359 } 2360 if (isInvalid) 2361 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 2362 else 2363 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 2364 } 2365 2366 if (!Tok.is(tok::eod)) 2367 LastTok = Tok; 2368 2369 // Read the rest of the macro body. 2370 if (MI->isObjectLike()) { 2371 // Object-like macros are very simple, just read their body. 2372 while (Tok.isNot(tok::eod)) { 2373 LastTok = Tok; 2374 MI->AddTokenToBody(Tok); 2375 // Get the next token of the macro. 2376 LexUnexpandedToken(Tok); 2377 } 2378 } else { 2379 // Otherwise, read the body of a function-like macro. While we are at it, 2380 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 2381 // parameters in function-like macro expansions. 2382 while (Tok.isNot(tok::eod)) { 2383 LastTok = Tok; 2384 2385 if (!Tok.isOneOf(tok::hash, tok::hashat, tok::hashhash)) { 2386 MI->AddTokenToBody(Tok); 2387 2388 // Get the next token of the macro. 2389 LexUnexpandedToken(Tok); 2390 continue; 2391 } 2392 2393 // If we're in -traditional mode, then we should ignore stringification 2394 // and token pasting. Mark the tokens as unknown so as not to confuse 2395 // things. 2396 if (getLangOpts().TraditionalCPP) { 2397 Tok.setKind(tok::unknown); 2398 MI->AddTokenToBody(Tok); 2399 2400 // Get the next token of the macro. 2401 LexUnexpandedToken(Tok); 2402 continue; 2403 } 2404 2405 if (Tok.is(tok::hashhash)) { 2406 // If we see token pasting, check if it looks like the gcc comma 2407 // pasting extension. We'll use this information to suppress 2408 // diagnostics later on. 2409 2410 // Get the next token of the macro. 2411 LexUnexpandedToken(Tok); 2412 2413 if (Tok.is(tok::eod)) { 2414 MI->AddTokenToBody(LastTok); 2415 break; 2416 } 2417 2418 unsigned NumTokens = MI->getNumTokens(); 2419 if (NumTokens && Tok.getIdentifierInfo() == Ident__VA_ARGS__ && 2420 MI->getReplacementToken(NumTokens-1).is(tok::comma)) 2421 MI->setHasCommaPasting(); 2422 2423 // Things look ok, add the '##' token to the macro. 2424 MI->AddTokenToBody(LastTok); 2425 continue; 2426 } 2427 2428 // Get the next token of the macro. 2429 LexUnexpandedToken(Tok); 2430 2431 // Check for a valid macro arg identifier. 2432 if (Tok.getIdentifierInfo() == nullptr || 2433 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 2434 2435 // If this is assembler-with-cpp mode, we accept random gibberish after 2436 // the '#' because '#' is often a comment character. However, change 2437 // the kind of the token to tok::unknown so that the preprocessor isn't 2438 // confused. 2439 if (getLangOpts().AsmPreprocessor && Tok.isNot(tok::eod)) { 2440 LastTok.setKind(tok::unknown); 2441 MI->AddTokenToBody(LastTok); 2442 continue; 2443 } else { 2444 Diag(Tok, diag::err_pp_stringize_not_parameter) 2445 << LastTok.is(tok::hashat); 2446 2447 // Disable __VA_ARGS__ again. 2448 Ident__VA_ARGS__->setIsPoisoned(true); 2449 return; 2450 } 2451 } 2452 2453 // Things look ok, add the '#' and param name tokens to the macro. 2454 MI->AddTokenToBody(LastTok); 2455 MI->AddTokenToBody(Tok); 2456 LastTok = Tok; 2457 2458 // Get the next token of the macro. 2459 LexUnexpandedToken(Tok); 2460 } 2461 } 2462 2463 if (MacroShadowsKeyword && 2464 !isConfigurationPattern(MacroNameTok, MI, getLangOpts())) { 2465 Diag(MacroNameTok, diag::warn_pp_macro_hides_keyword); 2466 } 2467 2468 // Disable __VA_ARGS__ again. 2469 Ident__VA_ARGS__->setIsPoisoned(true); 2470 2471 // Check that there is no paste (##) operator at the beginning or end of the 2472 // replacement list. 2473 unsigned NumTokens = MI->getNumTokens(); 2474 if (NumTokens != 0) { 2475 if (MI->getReplacementToken(0).is(tok::hashhash)) { 2476 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 2477 return; 2478 } 2479 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 2480 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 2481 return; 2482 } 2483 } 2484 2485 MI->setDefinitionEndLoc(LastTok.getLocation()); 2486 2487 // Finally, if this identifier already had a macro defined for it, verify that 2488 // the macro bodies are identical, and issue diagnostics if they are not. 2489 if (const MacroInfo *OtherMI=getMacroInfo(MacroNameTok.getIdentifierInfo())) { 2490 // In Objective-C, ignore attempts to directly redefine the builtin 2491 // definitions of the ownership qualifiers. It's still possible to 2492 // #undef them. 2493 auto isObjCProtectedMacro = [](const IdentifierInfo *II) -> bool { 2494 return II->isStr("__strong") || 2495 II->isStr("__weak") || 2496 II->isStr("__unsafe_unretained") || 2497 II->isStr("__autoreleasing"); 2498 }; 2499 if (getLangOpts().ObjC1 && 2500 SourceMgr.getFileID(OtherMI->getDefinitionLoc()) 2501 == getPredefinesFileID() && 2502 isObjCProtectedMacro(MacroNameTok.getIdentifierInfo())) { 2503 // Warn if it changes the tokens. 2504 if ((!getDiagnostics().getSuppressSystemWarnings() || 2505 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) && 2506 !MI->isIdenticalTo(*OtherMI, *this, 2507 /*Syntactic=*/LangOpts.MicrosoftExt)) { 2508 Diag(MI->getDefinitionLoc(), diag::warn_pp_objc_macro_redef_ignored); 2509 } 2510 assert(!OtherMI->isWarnIfUnused()); 2511 return; 2512 } 2513 2514 // It is very common for system headers to have tons of macro redefinitions 2515 // and for warnings to be disabled in system headers. If this is the case, 2516 // then don't bother calling MacroInfo::isIdenticalTo. 2517 if (!getDiagnostics().getSuppressSystemWarnings() || 2518 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 2519 if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused()) 2520 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 2521 2522 // Warn if defining "__LINE__" and other builtins, per C99 6.10.8/4 and 2523 // C++ [cpp.predefined]p4, but allow it as an extension. 2524 if (OtherMI->isBuiltinMacro()) 2525 Diag(MacroNameTok, diag::ext_pp_redef_builtin_macro); 2526 // Macros must be identical. This means all tokens and whitespace 2527 // separation must be the same. C99 6.10.3p2. 2528 else if (!OtherMI->isAllowRedefinitionsWithoutWarning() && 2529 !MI->isIdenticalTo(*OtherMI, *this, /*Syntactic=*/LangOpts.MicrosoftExt)) { 2530 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 2531 << MacroNameTok.getIdentifierInfo(); 2532 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 2533 } 2534 } 2535 if (OtherMI->isWarnIfUnused()) 2536 WarnUnusedMacroLocs.erase(OtherMI->getDefinitionLoc()); 2537 } 2538 2539 DefMacroDirective *MD = 2540 appendDefMacroDirective(MacroNameTok.getIdentifierInfo(), MI); 2541 2542 assert(!MI->isUsed()); 2543 // If we need warning for not using the macro, add its location in the 2544 // warn-because-unused-macro set. If it gets used it will be removed from set. 2545 if (getSourceManager().isInMainFile(MI->getDefinitionLoc()) && 2546 !Diags->isIgnored(diag::pp_macro_not_used, MI->getDefinitionLoc())) { 2547 MI->setIsWarnIfUnused(true); 2548 WarnUnusedMacroLocs.insert(MI->getDefinitionLoc()); 2549 } 2550 2551 // If the callbacks want to know, tell them about the macro definition. 2552 if (Callbacks) 2553 Callbacks->MacroDefined(MacroNameTok, MD); 2554 } 2555 2556 /// HandleUndefDirective - Implements \#undef. 2557 /// 2558 void Preprocessor::HandleUndefDirective() { 2559 ++NumUndefined; 2560 2561 Token MacroNameTok; 2562 ReadMacroName(MacroNameTok, MU_Undef); 2563 2564 // Error reading macro name? If so, diagnostic already issued. 2565 if (MacroNameTok.is(tok::eod)) 2566 return; 2567 2568 // Check to see if this is the last token on the #undef line. 2569 CheckEndOfDirective("undef"); 2570 2571 // Okay, we have a valid identifier to undef. 2572 auto *II = MacroNameTok.getIdentifierInfo(); 2573 auto MD = getMacroDefinition(II); 2574 2575 // If the callbacks want to know, tell them about the macro #undef. 2576 // Note: no matter if the macro was defined or not. 2577 if (Callbacks) 2578 Callbacks->MacroUndefined(MacroNameTok, MD); 2579 2580 // If the macro is not defined, this is a noop undef, just return. 2581 const MacroInfo *MI = MD.getMacroInfo(); 2582 if (!MI) 2583 return; 2584 2585 if (!MI->isUsed() && MI->isWarnIfUnused()) 2586 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 2587 2588 if (MI->isWarnIfUnused()) 2589 WarnUnusedMacroLocs.erase(MI->getDefinitionLoc()); 2590 2591 appendMacroDirective(MacroNameTok.getIdentifierInfo(), 2592 AllocateUndefMacroDirective(MacroNameTok.getLocation())); 2593 } 2594 2595 //===----------------------------------------------------------------------===// 2596 // Preprocessor Conditional Directive Handling. 2597 //===----------------------------------------------------------------------===// 2598 2599 /// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive. isIfndef 2600 /// is true when this is a \#ifndef directive. ReadAnyTokensBeforeDirective is 2601 /// true if any tokens have been returned or pp-directives activated before this 2602 /// \#ifndef has been lexed. 2603 /// 2604 void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 2605 bool ReadAnyTokensBeforeDirective) { 2606 ++NumIf; 2607 Token DirectiveTok = Result; 2608 2609 Token MacroNameTok; 2610 ReadMacroName(MacroNameTok); 2611 2612 // Error reading macro name? If so, diagnostic already issued. 2613 if (MacroNameTok.is(tok::eod)) { 2614 // Skip code until we get to #endif. This helps with recovery by not 2615 // emitting an error when the #endif is reached. 2616 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 2617 /*Foundnonskip*/false, /*FoundElse*/false); 2618 return; 2619 } 2620 2621 // Check to see if this is the last token on the #if[n]def line. 2622 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 2623 2624 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 2625 auto MD = getMacroDefinition(MII); 2626 MacroInfo *MI = MD.getMacroInfo(); 2627 2628 if (CurPPLexer->getConditionalStackDepth() == 0) { 2629 // If the start of a top-level #ifdef and if the macro is not defined, 2630 // inform MIOpt that this might be the start of a proper include guard. 2631 // Otherwise it is some other form of unknown conditional which we can't 2632 // handle. 2633 if (!ReadAnyTokensBeforeDirective && !MI) { 2634 assert(isIfndef && "#ifdef shouldn't reach here"); 2635 CurPPLexer->MIOpt.EnterTopLevelIfndef(MII, MacroNameTok.getLocation()); 2636 } else 2637 CurPPLexer->MIOpt.EnterTopLevelConditional(); 2638 } 2639 2640 // If there is a macro, process it. 2641 if (MI) // Mark it used. 2642 markMacroAsUsed(MI); 2643 2644 if (Callbacks) { 2645 if (isIfndef) 2646 Callbacks->Ifndef(DirectiveTok.getLocation(), MacroNameTok, MD); 2647 else 2648 Callbacks->Ifdef(DirectiveTok.getLocation(), MacroNameTok, MD); 2649 } 2650 2651 // Should we include the stuff contained by this directive? 2652 if (!MI == isIfndef) { 2653 // Yes, remember that we are inside a conditional, then lex the next token. 2654 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 2655 /*wasskip*/false, /*foundnonskip*/true, 2656 /*foundelse*/false); 2657 } else { 2658 // No, skip the contents of this block. 2659 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 2660 /*Foundnonskip*/false, 2661 /*FoundElse*/false); 2662 } 2663 } 2664 2665 /// HandleIfDirective - Implements the \#if directive. 2666 /// 2667 void Preprocessor::HandleIfDirective(Token &IfToken, 2668 bool ReadAnyTokensBeforeDirective) { 2669 ++NumIf; 2670 2671 // Parse and evaluate the conditional expression. 2672 IdentifierInfo *IfNDefMacro = nullptr; 2673 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 2674 const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 2675 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 2676 2677 // If this condition is equivalent to #ifndef X, and if this is the first 2678 // directive seen, handle it for the multiple-include optimization. 2679 if (CurPPLexer->getConditionalStackDepth() == 0) { 2680 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue) 2681 // FIXME: Pass in the location of the macro name, not the 'if' token. 2682 CurPPLexer->MIOpt.EnterTopLevelIfndef(IfNDefMacro, IfToken.getLocation()); 2683 else 2684 CurPPLexer->MIOpt.EnterTopLevelConditional(); 2685 } 2686 2687 if (Callbacks) 2688 Callbacks->If(IfToken.getLocation(), 2689 SourceRange(ConditionalBegin, ConditionalEnd), 2690 (ConditionalTrue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False)); 2691 2692 // Should we include the stuff contained by this directive? 2693 if (ConditionalTrue) { 2694 // Yes, remember that we are inside a conditional, then lex the next token. 2695 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 2696 /*foundnonskip*/true, /*foundelse*/false); 2697 } else { 2698 // No, skip the contents of this block. 2699 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 2700 /*FoundElse*/false); 2701 } 2702 } 2703 2704 /// HandleEndifDirective - Implements the \#endif directive. 2705 /// 2706 void Preprocessor::HandleEndifDirective(Token &EndifToken) { 2707 ++NumEndif; 2708 2709 // Check that this is the whole directive. 2710 CheckEndOfDirective("endif"); 2711 2712 PPConditionalInfo CondInfo; 2713 if (CurPPLexer->popConditionalLevel(CondInfo)) { 2714 // No conditionals on the stack: this is an #endif without an #if. 2715 Diag(EndifToken, diag::err_pp_endif_without_if); 2716 return; 2717 } 2718 2719 // If this the end of a top-level #endif, inform MIOpt. 2720 if (CurPPLexer->getConditionalStackDepth() == 0) 2721 CurPPLexer->MIOpt.ExitTopLevelConditional(); 2722 2723 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 2724 "This code should only be reachable in the non-skipping case!"); 2725 2726 if (Callbacks) 2727 Callbacks->Endif(EndifToken.getLocation(), CondInfo.IfLoc); 2728 } 2729 2730 /// HandleElseDirective - Implements the \#else directive. 2731 /// 2732 void Preprocessor::HandleElseDirective(Token &Result) { 2733 ++NumElse; 2734 2735 // #else directive in a non-skipping conditional... start skipping. 2736 CheckEndOfDirective("else"); 2737 2738 PPConditionalInfo CI; 2739 if (CurPPLexer->popConditionalLevel(CI)) { 2740 Diag(Result, diag::pp_err_else_without_if); 2741 return; 2742 } 2743 2744 // If this is a top-level #else, inform the MIOpt. 2745 if (CurPPLexer->getConditionalStackDepth() == 0) 2746 CurPPLexer->MIOpt.EnterTopLevelConditional(); 2747 2748 // If this is a #else with a #else before it, report the error. 2749 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 2750 2751 if (Callbacks) 2752 Callbacks->Else(Result.getLocation(), CI.IfLoc); 2753 2754 // Finally, skip the rest of the contents of this block. 2755 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 2756 /*FoundElse*/true, Result.getLocation()); 2757 } 2758 2759 /// HandleElifDirective - Implements the \#elif directive. 2760 /// 2761 void Preprocessor::HandleElifDirective(Token &ElifToken) { 2762 ++NumElse; 2763 2764 // #elif directive in a non-skipping conditional... start skipping. 2765 // We don't care what the condition is, because we will always skip it (since 2766 // the block immediately before it was included). 2767 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 2768 DiscardUntilEndOfDirective(); 2769 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 2770 2771 PPConditionalInfo CI; 2772 if (CurPPLexer->popConditionalLevel(CI)) { 2773 Diag(ElifToken, diag::pp_err_elif_without_if); 2774 return; 2775 } 2776 2777 // If this is a top-level #elif, inform the MIOpt. 2778 if (CurPPLexer->getConditionalStackDepth() == 0) 2779 CurPPLexer->MIOpt.EnterTopLevelConditional(); 2780 2781 // If this is a #elif with a #else before it, report the error. 2782 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 2783 2784 if (Callbacks) 2785 Callbacks->Elif(ElifToken.getLocation(), 2786 SourceRange(ConditionalBegin, ConditionalEnd), 2787 PPCallbacks::CVK_NotEvaluated, CI.IfLoc); 2788 2789 // Finally, skip the rest of the contents of this block. 2790 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 2791 /*FoundElse*/CI.FoundElse, 2792 ElifToken.getLocation()); 2793 } 2794