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 /// 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/PreprocessorOptions.h"
34 #include "clang/Lex/Token.h"
35 #include "clang/Lex/VariadicMacroSupport.h"
36 #include "llvm/ADT/ArrayRef.h"
37 #include "llvm/ADT/SmallString.h"
38 #include "llvm/ADT/SmallVector.h"
39 #include "llvm/ADT/STLExtras.h"
40 #include "llvm/ADT/StringSwitch.h"
41 #include "llvm/ADT/StringRef.h"
42 #include "llvm/Support/AlignOf.h"
43 #include "llvm/Support/ErrorHandling.h"
44 #include "llvm/Support/Path.h"
45 #include <algorithm>
46 #include <cassert>
47 #include <cstring>
48 #include <new>
49 #include <string>
50 #include <utility>
51
52 using namespace clang;
53
54 //===----------------------------------------------------------------------===//
55 // Utility Methods for Preprocessor Directive Handling.
56 //===----------------------------------------------------------------------===//
57
AllocateMacroInfo(SourceLocation L)58 MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) {
59 auto *MIChain = new (BP) MacroInfoChain{L, MIChainHead};
60 MIChainHead = MIChain;
61 return &MIChain->MI;
62 }
63
AllocateDefMacroDirective(MacroInfo * MI,SourceLocation Loc)64 DefMacroDirective *Preprocessor::AllocateDefMacroDirective(MacroInfo *MI,
65 SourceLocation Loc) {
66 return new (BP) DefMacroDirective(MI, Loc);
67 }
68
69 UndefMacroDirective *
AllocateUndefMacroDirective(SourceLocation UndefLoc)70 Preprocessor::AllocateUndefMacroDirective(SourceLocation UndefLoc) {
71 return new (BP) UndefMacroDirective(UndefLoc);
72 }
73
74 VisibilityMacroDirective *
AllocateVisibilityMacroDirective(SourceLocation Loc,bool isPublic)75 Preprocessor::AllocateVisibilityMacroDirective(SourceLocation Loc,
76 bool isPublic) {
77 return new (BP) VisibilityMacroDirective(Loc, isPublic);
78 }
79
80 /// Read and discard all tokens remaining on the current line until
81 /// the tok::eod token is found.
DiscardUntilEndOfDirective()82 void Preprocessor::DiscardUntilEndOfDirective() {
83 Token Tmp;
84 do {
85 LexUnexpandedToken(Tmp);
86 assert(Tmp.isNot(tok::eof) && "EOF seen while discarding directive tokens");
87 } while (Tmp.isNot(tok::eod));
88 }
89
90 /// Enumerates possible cases of #define/#undef a reserved identifier.
91 enum MacroDiag {
92 MD_NoWarn, //> Not a reserved identifier
93 MD_KeywordDef, //> Macro hides keyword, enabled by default
94 MD_ReservedMacro //> #define of #undef reserved id, disabled by default
95 };
96
97 /// Checks if the specified identifier is reserved in the specified
98 /// language.
99 /// This function does not check if the identifier is a keyword.
isReservedId(StringRef Text,const LangOptions & Lang)100 static bool isReservedId(StringRef Text, const LangOptions &Lang) {
101 // C++ [macro.names], C11 7.1.3:
102 // All identifiers that begin with an underscore and either an uppercase
103 // letter or another underscore are always reserved for any use.
104 if (Text.size() >= 2 && Text[0] == '_' &&
105 (isUppercase(Text[1]) || Text[1] == '_'))
106 return true;
107 // C++ [global.names]
108 // Each name that contains a double underscore ... is reserved to the
109 // implementation for any use.
110 if (Lang.CPlusPlus) {
111 if (Text.find("__") != StringRef::npos)
112 return true;
113 }
114 return false;
115 }
116
117 // The -fmodule-name option tells the compiler to textually include headers in
118 // the specified module, meaning clang won't build the specified module. This is
119 // useful in a number of situations, for instance, when building a library that
120 // vends a module map, one might want to avoid hitting intermediate build
121 // products containimg the the module map or avoid finding the system installed
122 // modulemap for that library.
isForModuleBuilding(Module * M,StringRef CurrentModule,StringRef ModuleName)123 static bool isForModuleBuilding(Module *M, StringRef CurrentModule,
124 StringRef ModuleName) {
125 StringRef TopLevelName = M->getTopLevelModuleName();
126
127 // When building framework Foo, we wanna make sure that Foo *and* Foo_Private
128 // are textually included and no modules are built for both.
129 if (M->getTopLevelModule()->IsFramework && CurrentModule == ModuleName &&
130 !CurrentModule.endswith("_Private") && TopLevelName.endswith("_Private"))
131 TopLevelName = TopLevelName.drop_back(8);
132
133 return TopLevelName == CurrentModule;
134 }
135
shouldWarnOnMacroDef(Preprocessor & PP,IdentifierInfo * II)136 static MacroDiag shouldWarnOnMacroDef(Preprocessor &PP, IdentifierInfo *II) {
137 const LangOptions &Lang = PP.getLangOpts();
138 StringRef Text = II->getName();
139 if (isReservedId(Text, Lang))
140 return MD_ReservedMacro;
141 if (II->isKeyword(Lang))
142 return MD_KeywordDef;
143 if (Lang.CPlusPlus11 && (Text.equals("override") || Text.equals("final")))
144 return MD_KeywordDef;
145 return MD_NoWarn;
146 }
147
shouldWarnOnMacroUndef(Preprocessor & PP,IdentifierInfo * II)148 static MacroDiag shouldWarnOnMacroUndef(Preprocessor &PP, IdentifierInfo *II) {
149 const LangOptions &Lang = PP.getLangOpts();
150 StringRef Text = II->getName();
151 // Do not warn on keyword undef. It is generally harmless and widely used.
152 if (isReservedId(Text, Lang))
153 return MD_ReservedMacro;
154 return MD_NoWarn;
155 }
156
157 // Return true if we want to issue a diagnostic by default if we
158 // encounter this name in a #include with the wrong case. For now,
159 // this includes the standard C and C++ headers, Posix headers,
160 // and Boost headers. Improper case for these #includes is a
161 // potential portability issue.
warnByDefaultOnWrongCase(StringRef Include)162 static bool warnByDefaultOnWrongCase(StringRef Include) {
163 // If the first component of the path is "boost", treat this like a standard header
164 // for the purposes of diagnostics.
165 if (::llvm::sys::path::begin(Include)->equals_lower("boost"))
166 return true;
167
168 // "condition_variable" is the longest standard header name at 18 characters.
169 // If the include file name is longer than that, it can't be a standard header.
170 static const size_t MaxStdHeaderNameLen = 18u;
171 if (Include.size() > MaxStdHeaderNameLen)
172 return false;
173
174 // Lowercase and normalize the search string.
175 SmallString<32> LowerInclude{Include};
176 for (char &Ch : LowerInclude) {
177 // In the ASCII range?
178 if (static_cast<unsigned char>(Ch) > 0x7f)
179 return false; // Can't be a standard header
180 // ASCII lowercase:
181 if (Ch >= 'A' && Ch <= 'Z')
182 Ch += 'a' - 'A';
183 // Normalize path separators for comparison purposes.
184 else if (::llvm::sys::path::is_separator(Ch))
185 Ch = '/';
186 }
187
188 // The standard C/C++ and Posix headers
189 return llvm::StringSwitch<bool>(LowerInclude)
190 // C library headers
191 .Cases("assert.h", "complex.h", "ctype.h", "errno.h", "fenv.h", true)
192 .Cases("float.h", "inttypes.h", "iso646.h", "limits.h", "locale.h", true)
193 .Cases("math.h", "setjmp.h", "signal.h", "stdalign.h", "stdarg.h", true)
194 .Cases("stdatomic.h", "stdbool.h", "stddef.h", "stdint.h", "stdio.h", true)
195 .Cases("stdlib.h", "stdnoreturn.h", "string.h", "tgmath.h", "threads.h", true)
196 .Cases("time.h", "uchar.h", "wchar.h", "wctype.h", true)
197
198 // C++ headers for C library facilities
199 .Cases("cassert", "ccomplex", "cctype", "cerrno", "cfenv", true)
200 .Cases("cfloat", "cinttypes", "ciso646", "climits", "clocale", true)
201 .Cases("cmath", "csetjmp", "csignal", "cstdalign", "cstdarg", true)
202 .Cases("cstdbool", "cstddef", "cstdint", "cstdio", "cstdlib", true)
203 .Cases("cstring", "ctgmath", "ctime", "cuchar", "cwchar", true)
204 .Case("cwctype", true)
205
206 // C++ library headers
207 .Cases("algorithm", "fstream", "list", "regex", "thread", true)
208 .Cases("array", "functional", "locale", "scoped_allocator", "tuple", true)
209 .Cases("atomic", "future", "map", "set", "type_traits", true)
210 .Cases("bitset", "initializer_list", "memory", "shared_mutex", "typeindex", true)
211 .Cases("chrono", "iomanip", "mutex", "sstream", "typeinfo", true)
212 .Cases("codecvt", "ios", "new", "stack", "unordered_map", true)
213 .Cases("complex", "iosfwd", "numeric", "stdexcept", "unordered_set", true)
214 .Cases("condition_variable", "iostream", "ostream", "streambuf", "utility", true)
215 .Cases("deque", "istream", "queue", "string", "valarray", true)
216 .Cases("exception", "iterator", "random", "strstream", "vector", true)
217 .Cases("forward_list", "limits", "ratio", "system_error", true)
218
219 // POSIX headers (which aren't also C headers)
220 .Cases("aio.h", "arpa/inet.h", "cpio.h", "dirent.h", "dlfcn.h", true)
221 .Cases("fcntl.h", "fmtmsg.h", "fnmatch.h", "ftw.h", "glob.h", true)
222 .Cases("grp.h", "iconv.h", "langinfo.h", "libgen.h", "monetary.h", true)
223 .Cases("mqueue.h", "ndbm.h", "net/if.h", "netdb.h", "netinet/in.h", true)
224 .Cases("netinet/tcp.h", "nl_types.h", "poll.h", "pthread.h", "pwd.h", true)
225 .Cases("regex.h", "sched.h", "search.h", "semaphore.h", "spawn.h", true)
226 .Cases("strings.h", "stropts.h", "sys/ipc.h", "sys/mman.h", "sys/msg.h", true)
227 .Cases("sys/resource.h", "sys/select.h", "sys/sem.h", "sys/shm.h", "sys/socket.h", true)
228 .Cases("sys/stat.h", "sys/statvfs.h", "sys/time.h", "sys/times.h", "sys/types.h", true)
229 .Cases("sys/uio.h", "sys/un.h", "sys/utsname.h", "sys/wait.h", "syslog.h", true)
230 .Cases("tar.h", "termios.h", "trace.h", "ulimit.h", true)
231 .Cases("unistd.h", "utime.h", "utmpx.h", "wordexp.h", true)
232 .Default(false);
233 }
234
CheckMacroName(Token & MacroNameTok,MacroUse isDefineUndef,bool * ShadowFlag)235 bool Preprocessor::CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
236 bool *ShadowFlag) {
237 // Missing macro name?
238 if (MacroNameTok.is(tok::eod))
239 return Diag(MacroNameTok, diag::err_pp_missing_macro_name);
240
241 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
242 if (!II)
243 return Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
244
245 if (II->isCPlusPlusOperatorKeyword()) {
246 // C++ 2.5p2: Alternative tokens behave the same as its primary token
247 // except for their spellings.
248 Diag(MacroNameTok, getLangOpts().MicrosoftExt
249 ? diag::ext_pp_operator_used_as_macro_name
250 : diag::err_pp_operator_used_as_macro_name)
251 << II << MacroNameTok.getKind();
252 // Allow #defining |and| and friends for Microsoft compatibility or
253 // recovery when legacy C headers are included in C++.
254 }
255
256 if ((isDefineUndef != MU_Other) && II->getPPKeywordID() == tok::pp_defined) {
257 // Error if defining "defined": C99 6.10.8/4, C++ [cpp.predefined]p4.
258 return Diag(MacroNameTok, diag::err_defined_macro_name);
259 }
260
261 if (isDefineUndef == MU_Undef) {
262 auto *MI = getMacroInfo(II);
263 if (MI && MI->isBuiltinMacro()) {
264 // Warn if undefining "__LINE__" and other builtins, per C99 6.10.8/4
265 // and C++ [cpp.predefined]p4], but allow it as an extension.
266 Diag(MacroNameTok, diag::ext_pp_undef_builtin_macro);
267 }
268 }
269
270 // If defining/undefining reserved identifier or a keyword, we need to issue
271 // a warning.
272 SourceLocation MacroNameLoc = MacroNameTok.getLocation();
273 if (ShadowFlag)
274 *ShadowFlag = false;
275 if (!SourceMgr.isInSystemHeader(MacroNameLoc) &&
276 (SourceMgr.getBufferName(MacroNameLoc) != "<built-in>")) {
277 MacroDiag D = MD_NoWarn;
278 if (isDefineUndef == MU_Define) {
279 D = shouldWarnOnMacroDef(*this, II);
280 }
281 else if (isDefineUndef == MU_Undef)
282 D = shouldWarnOnMacroUndef(*this, II);
283 if (D == MD_KeywordDef) {
284 // We do not want to warn on some patterns widely used in configuration
285 // scripts. This requires analyzing next tokens, so do not issue warnings
286 // now, only inform caller.
287 if (ShadowFlag)
288 *ShadowFlag = true;
289 }
290 if (D == MD_ReservedMacro)
291 Diag(MacroNameTok, diag::warn_pp_macro_is_reserved_id);
292 }
293
294 // Okay, we got a good identifier.
295 return false;
296 }
297
298 /// Lex and validate a macro name, which occurs after a
299 /// \#define or \#undef.
300 ///
301 /// This sets the token kind to eod and discards the rest of the macro line if
302 /// the macro name is invalid.
303 ///
304 /// \param MacroNameTok Token that is expected to be a macro name.
305 /// \param isDefineUndef Context in which macro is used.
306 /// \param ShadowFlag Points to a flag that is set if macro shadows a keyword.
ReadMacroName(Token & MacroNameTok,MacroUse isDefineUndef,bool * ShadowFlag)307 void Preprocessor::ReadMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
308 bool *ShadowFlag) {
309 // Read the token, don't allow macro expansion on it.
310 LexUnexpandedToken(MacroNameTok);
311
312 if (MacroNameTok.is(tok::code_completion)) {
313 if (CodeComplete)
314 CodeComplete->CodeCompleteMacroName(isDefineUndef == MU_Define);
315 setCodeCompletionReached();
316 LexUnexpandedToken(MacroNameTok);
317 }
318
319 if (!CheckMacroName(MacroNameTok, isDefineUndef, ShadowFlag))
320 return;
321
322 // Invalid macro name, read and discard the rest of the line and set the
323 // token kind to tok::eod if necessary.
324 if (MacroNameTok.isNot(tok::eod)) {
325 MacroNameTok.setKind(tok::eod);
326 DiscardUntilEndOfDirective();
327 }
328 }
329
330 /// Ensure that the next token is a tok::eod token.
331 ///
332 /// If not, emit a diagnostic and consume up until the eod. If EnableMacros is
333 /// true, then we consider macros that expand to zero tokens as being ok.
CheckEndOfDirective(const char * DirType,bool EnableMacros)334 void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) {
335 Token Tmp;
336 // Lex unexpanded tokens for most directives: macros might expand to zero
337 // tokens, causing us to miss diagnosing invalid lines. Some directives (like
338 // #line) allow empty macros.
339 if (EnableMacros)
340 Lex(Tmp);
341 else
342 LexUnexpandedToken(Tmp);
343
344 // There should be no tokens after the directive, but we allow them as an
345 // extension.
346 while (Tmp.is(tok::comment)) // Skip comments in -C mode.
347 LexUnexpandedToken(Tmp);
348
349 if (Tmp.isNot(tok::eod)) {
350 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89,
351 // or if this is a macro-style preprocessing directive, because it is more
352 // trouble than it is worth to insert /**/ and check that there is no /**/
353 // in the range also.
354 FixItHint Hint;
355 if ((LangOpts.GNUMode || LangOpts.C99 || LangOpts.CPlusPlus) &&
356 !CurTokenLexer)
357 Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//");
358 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint;
359 DiscardUntilEndOfDirective();
360 }
361 }
362
363 /// SkipExcludedConditionalBlock - We just read a \#if or related directive and
364 /// decided that the subsequent tokens are in the \#if'd out portion of the
365 /// file. Lex the rest of the file, until we see an \#endif. If
366 /// FoundNonSkipPortion is true, then we have already emitted code for part of
367 /// this \#if directive, so \#else/\#elif blocks should never be entered.
368 /// If ElseOk is true, then \#else directives are ok, if not, then we have
369 /// already seen one so a \#else directive is a duplicate. When this returns,
370 /// the caller can lex the first valid token.
SkipExcludedConditionalBlock(SourceLocation HashTokenLoc,SourceLocation IfTokenLoc,bool FoundNonSkipPortion,bool FoundElse,SourceLocation ElseLoc)371 void Preprocessor::SkipExcludedConditionalBlock(SourceLocation HashTokenLoc,
372 SourceLocation IfTokenLoc,
373 bool FoundNonSkipPortion,
374 bool FoundElse,
375 SourceLocation ElseLoc) {
376 ++NumSkipped;
377 assert(!CurTokenLexer && CurPPLexer && "Lexing a macro, not a file?");
378
379 if (PreambleConditionalStack.reachedEOFWhileSkipping())
380 PreambleConditionalStack.clearSkipInfo();
381 else
382 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/ false,
383 FoundNonSkipPortion, FoundElse);
384
385 // Enter raw mode to disable identifier lookup (and thus macro expansion),
386 // disabling warnings, etc.
387 CurPPLexer->LexingRawMode = true;
388 Token Tok;
389 while (true) {
390 CurLexer->Lex(Tok);
391
392 if (Tok.is(tok::code_completion)) {
393 if (CodeComplete)
394 CodeComplete->CodeCompleteInConditionalExclusion();
395 setCodeCompletionReached();
396 continue;
397 }
398
399 // If this is the end of the buffer, we have an error.
400 if (Tok.is(tok::eof)) {
401 // We don't emit errors for unterminated conditionals here,
402 // Lexer::LexEndOfFile can do that properly.
403 // Just return and let the caller lex after this #include.
404 if (PreambleConditionalStack.isRecording())
405 PreambleConditionalStack.SkipInfo.emplace(
406 HashTokenLoc, IfTokenLoc, FoundNonSkipPortion, FoundElse, ElseLoc);
407 break;
408 }
409
410 // If this token is not a preprocessor directive, just skip it.
411 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine())
412 continue;
413
414 // We just parsed a # character at the start of a line, so we're in
415 // directive mode. Tell the lexer this so any newlines we see will be
416 // converted into an EOD token (this terminates the macro).
417 CurPPLexer->ParsingPreprocessorDirective = true;
418 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
419
420
421 // Read the next token, the directive flavor.
422 LexUnexpandedToken(Tok);
423
424 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
425 // something bogus), skip it.
426 if (Tok.isNot(tok::raw_identifier)) {
427 CurPPLexer->ParsingPreprocessorDirective = false;
428 // Restore comment saving mode.
429 if (CurLexer) CurLexer->resetExtendedTokenMode();
430 continue;
431 }
432
433 // If the first letter isn't i or e, it isn't intesting to us. We know that
434 // this is safe in the face of spelling differences, because there is no way
435 // to spell an i/e in a strange way that is another letter. Skipping this
436 // allows us to avoid looking up the identifier info for #define/#undef and
437 // other common directives.
438 StringRef RI = Tok.getRawIdentifier();
439
440 char FirstChar = RI[0];
441 if (FirstChar >= 'a' && FirstChar <= 'z' &&
442 FirstChar != 'i' && FirstChar != 'e') {
443 CurPPLexer->ParsingPreprocessorDirective = false;
444 // Restore comment saving mode.
445 if (CurLexer) CurLexer->resetExtendedTokenMode();
446 continue;
447 }
448
449 // Get the identifier name without trigraphs or embedded newlines. Note
450 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
451 // when skipping.
452 char DirectiveBuf[20];
453 StringRef Directive;
454 if (!Tok.needsCleaning() && RI.size() < 20) {
455 Directive = RI;
456 } else {
457 std::string DirectiveStr = getSpelling(Tok);
458 size_t IdLen = DirectiveStr.size();
459 if (IdLen >= 20) {
460 CurPPLexer->ParsingPreprocessorDirective = false;
461 // Restore comment saving mode.
462 if (CurLexer) CurLexer->resetExtendedTokenMode();
463 continue;
464 }
465 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen);
466 Directive = StringRef(DirectiveBuf, IdLen);
467 }
468
469 if (Directive.startswith("if")) {
470 StringRef Sub = Directive.substr(2);
471 if (Sub.empty() || // "if"
472 Sub == "def" || // "ifdef"
473 Sub == "ndef") { // "ifndef"
474 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
475 // bother parsing the condition.
476 DiscardUntilEndOfDirective();
477 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true,
478 /*foundnonskip*/false,
479 /*foundelse*/false);
480 }
481 } else if (Directive[0] == 'e') {
482 StringRef Sub = Directive.substr(1);
483 if (Sub == "ndif") { // "endif"
484 PPConditionalInfo CondInfo;
485 CondInfo.WasSkipping = true; // Silence bogus warning.
486 bool InCond = CurPPLexer->popConditionalLevel(CondInfo);
487 (void)InCond; // Silence warning in no-asserts mode.
488 assert(!InCond && "Can't be skipping if not in a conditional!");
489
490 // If we popped the outermost skipping block, we're done skipping!
491 if (!CondInfo.WasSkipping) {
492 // Restore the value of LexingRawMode so that trailing comments
493 // are handled correctly, if we've reached the outermost block.
494 CurPPLexer->LexingRawMode = false;
495 CheckEndOfDirective("endif");
496 CurPPLexer->LexingRawMode = true;
497 if (Callbacks)
498 Callbacks->Endif(Tok.getLocation(), CondInfo.IfLoc);
499 break;
500 } else {
501 DiscardUntilEndOfDirective();
502 }
503 } else if (Sub == "lse") { // "else".
504 // #else directive in a skipping conditional. If not in some other
505 // skipping conditional, and if #else hasn't already been seen, enter it
506 // as a non-skipping conditional.
507 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
508
509 // If this is a #else with a #else before it, report the error.
510 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else);
511
512 // Note that we've seen a #else in this conditional.
513 CondInfo.FoundElse = true;
514
515 // If the conditional is at the top level, and the #if block wasn't
516 // entered, enter the #else block now.
517 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
518 CondInfo.FoundNonSkip = true;
519 // Restore the value of LexingRawMode so that trailing comments
520 // are handled correctly.
521 CurPPLexer->LexingRawMode = false;
522 CheckEndOfDirective("else");
523 CurPPLexer->LexingRawMode = true;
524 if (Callbacks)
525 Callbacks->Else(Tok.getLocation(), CondInfo.IfLoc);
526 break;
527 } else {
528 DiscardUntilEndOfDirective(); // C99 6.10p4.
529 }
530 } else if (Sub == "lif") { // "elif".
531 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
532
533 // If this is a #elif with a #else before it, report the error.
534 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else);
535
536 // If this is in a skipping block or if we're already handled this #if
537 // block, don't bother parsing the condition.
538 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
539 DiscardUntilEndOfDirective();
540 } else {
541 const SourceLocation CondBegin = CurPPLexer->getSourceLocation();
542 // Restore the value of LexingRawMode so that identifiers are
543 // looked up, etc, inside the #elif expression.
544 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
545 CurPPLexer->LexingRawMode = false;
546 IdentifierInfo *IfNDefMacro = nullptr;
547 const bool CondValue = EvaluateDirectiveExpression(IfNDefMacro).Conditional;
548 CurPPLexer->LexingRawMode = true;
549 if (Callbacks) {
550 const SourceLocation CondEnd = CurPPLexer->getSourceLocation();
551 Callbacks->Elif(Tok.getLocation(),
552 SourceRange(CondBegin, CondEnd),
553 (CondValue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False), CondInfo.IfLoc);
554 }
555 // If this condition is true, enter it!
556 if (CondValue) {
557 CondInfo.FoundNonSkip = true;
558 break;
559 }
560 }
561 }
562 }
563
564 CurPPLexer->ParsingPreprocessorDirective = false;
565 // Restore comment saving mode.
566 if (CurLexer) CurLexer->resetExtendedTokenMode();
567 }
568
569 // Finally, if we are out of the conditional (saw an #endif or ran off the end
570 // of the file, just stop skipping and return to lexing whatever came after
571 // the #if block.
572 CurPPLexer->LexingRawMode = false;
573
574 // The last skipped range isn't actually skipped yet if it's truncated
575 // by the end of the preamble; we'll resume parsing after the preamble.
576 if (Callbacks && (Tok.isNot(tok::eof) || !isRecordingPreamble()))
577 Callbacks->SourceRangeSkipped(
578 SourceRange(HashTokenLoc, CurPPLexer->getSourceLocation()),
579 Tok.getLocation());
580 }
581
getModuleForLocation(SourceLocation Loc)582 Module *Preprocessor::getModuleForLocation(SourceLocation Loc) {
583 if (!SourceMgr.isInMainFile(Loc)) {
584 // Try to determine the module of the include directive.
585 // FIXME: Look into directly passing the FileEntry from LookupFile instead.
586 FileID IDOfIncl = SourceMgr.getFileID(SourceMgr.getExpansionLoc(Loc));
587 if (const FileEntry *EntryOfIncl = SourceMgr.getFileEntryForID(IDOfIncl)) {
588 // The include comes from an included file.
589 return HeaderInfo.getModuleMap()
590 .findModuleForHeader(EntryOfIncl)
591 .getModule();
592 }
593 }
594
595 // This is either in the main file or not in a file at all. It belongs
596 // to the current module, if there is one.
597 return getLangOpts().CurrentModule.empty()
598 ? nullptr
599 : HeaderInfo.lookupModule(getLangOpts().CurrentModule);
600 }
601
602 const FileEntry *
getModuleHeaderToIncludeForDiagnostics(SourceLocation IncLoc,Module * M,SourceLocation Loc)603 Preprocessor::getModuleHeaderToIncludeForDiagnostics(SourceLocation IncLoc,
604 Module *M,
605 SourceLocation Loc) {
606 assert(M && "no module to include");
607
608 // If we have a module import syntax, we shouldn't include a header to
609 // make a particular module visible.
610 if (getLangOpts().ObjC)
611 return nullptr;
612
613 Module *TopM = M->getTopLevelModule();
614 Module *IncM = getModuleForLocation(IncLoc);
615
616 // Walk up through the include stack, looking through textual headers of M
617 // until we hit a non-textual header that we can #include. (We assume textual
618 // headers of a module with non-textual headers aren't meant to be used to
619 // import entities from the module.)
620 auto &SM = getSourceManager();
621 while (!Loc.isInvalid() && !SM.isInMainFile(Loc)) {
622 auto ID = SM.getFileID(SM.getExpansionLoc(Loc));
623 auto *FE = SM.getFileEntryForID(ID);
624 if (!FE)
625 break;
626
627 bool InTextualHeader = false;
628 for (auto Header : HeaderInfo.getModuleMap().findAllModulesForHeader(FE)) {
629 if (!Header.getModule()->isSubModuleOf(TopM))
630 continue;
631
632 if (!(Header.getRole() & ModuleMap::TextualHeader)) {
633 // If this is an accessible, non-textual header of M's top-level module
634 // that transitively includes the given location and makes the
635 // corresponding module visible, this is the thing to #include.
636 if (Header.isAccessibleFrom(IncM))
637 return FE;
638
639 // It's in a private header; we can't #include it.
640 // FIXME: If there's a public header in some module that re-exports it,
641 // then we could suggest including that, but it's not clear that's the
642 // expected way to make this entity visible.
643 continue;
644 }
645
646 InTextualHeader = true;
647 }
648
649 if (!InTextualHeader)
650 break;
651
652 Loc = SM.getIncludeLoc(ID);
653 }
654
655 return nullptr;
656 }
657
LookupFile(SourceLocation FilenameLoc,StringRef Filename,bool isAngled,const DirectoryLookup * FromDir,const FileEntry * FromFile,const DirectoryLookup * & CurDir,SmallVectorImpl<char> * SearchPath,SmallVectorImpl<char> * RelativePath,ModuleMap::KnownHeader * SuggestedModule,bool * IsMapped,bool SkipCache)658 const FileEntry *Preprocessor::LookupFile(
659 SourceLocation FilenameLoc, StringRef Filename, bool isAngled,
660 const DirectoryLookup *FromDir, const FileEntry *FromFile,
661 const DirectoryLookup *&CurDir, SmallVectorImpl<char> *SearchPath,
662 SmallVectorImpl<char> *RelativePath,
663 ModuleMap::KnownHeader *SuggestedModule, bool *IsMapped, bool SkipCache) {
664 Module *RequestingModule = getModuleForLocation(FilenameLoc);
665 bool RequestingModuleIsModuleInterface = !SourceMgr.isInMainFile(FilenameLoc);
666
667 // If the header lookup mechanism may be relative to the current inclusion
668 // stack, record the parent #includes.
669 SmallVector<std::pair<const FileEntry *, const DirectoryEntry *>, 16>
670 Includers;
671 bool BuildSystemModule = false;
672 if (!FromDir && !FromFile) {
673 FileID FID = getCurrentFileLexer()->getFileID();
674 const FileEntry *FileEnt = SourceMgr.getFileEntryForID(FID);
675
676 // If there is no file entry associated with this file, it must be the
677 // predefines buffer or the module includes buffer. Any other file is not
678 // lexed with a normal lexer, so it won't be scanned for preprocessor
679 // directives.
680 //
681 // If we have the predefines buffer, resolve #include references (which come
682 // from the -include command line argument) from the current working
683 // directory instead of relative to the main file.
684 //
685 // If we have the module includes buffer, resolve #include references (which
686 // come from header declarations in the module map) relative to the module
687 // map file.
688 if (!FileEnt) {
689 if (FID == SourceMgr.getMainFileID() && MainFileDir) {
690 Includers.push_back(std::make_pair(nullptr, MainFileDir));
691 BuildSystemModule = getCurrentModule()->IsSystem;
692 } else if ((FileEnt =
693 SourceMgr.getFileEntryForID(SourceMgr.getMainFileID())))
694 Includers.push_back(std::make_pair(FileEnt, FileMgr.getDirectory(".")));
695 } else {
696 Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir()));
697 }
698
699 // MSVC searches the current include stack from top to bottom for
700 // headers included by quoted include directives.
701 // See: http://msdn.microsoft.com/en-us/library/36k2cdd4.aspx
702 if (LangOpts.MSVCCompat && !isAngled) {
703 for (IncludeStackInfo &ISEntry : llvm::reverse(IncludeMacroStack)) {
704 if (IsFileLexer(ISEntry))
705 if ((FileEnt = ISEntry.ThePPLexer->getFileEntry()))
706 Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir()));
707 }
708 }
709 }
710
711 CurDir = CurDirLookup;
712
713 if (FromFile) {
714 // We're supposed to start looking from after a particular file. Search
715 // the include path until we find that file or run out of files.
716 const DirectoryLookup *TmpCurDir = CurDir;
717 const DirectoryLookup *TmpFromDir = nullptr;
718 while (const FileEntry *FE = HeaderInfo.LookupFile(
719 Filename, FilenameLoc, isAngled, TmpFromDir, TmpCurDir,
720 Includers, SearchPath, RelativePath, RequestingModule,
721 SuggestedModule, /*IsMapped=*/nullptr, SkipCache)) {
722 // Keep looking as if this file did a #include_next.
723 TmpFromDir = TmpCurDir;
724 ++TmpFromDir;
725 if (FE == FromFile) {
726 // Found it.
727 FromDir = TmpFromDir;
728 CurDir = TmpCurDir;
729 break;
730 }
731 }
732 }
733
734 // Do a standard file entry lookup.
735 const FileEntry *FE = HeaderInfo.LookupFile(
736 Filename, FilenameLoc, isAngled, FromDir, CurDir, Includers, SearchPath,
737 RelativePath, RequestingModule, SuggestedModule, IsMapped, SkipCache,
738 BuildSystemModule);
739 if (FE) {
740 if (SuggestedModule && !LangOpts.AsmPreprocessor)
741 HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
742 RequestingModule, RequestingModuleIsModuleInterface, FilenameLoc,
743 Filename, FE);
744 return FE;
745 }
746
747 const FileEntry *CurFileEnt;
748 // Otherwise, see if this is a subframework header. If so, this is relative
749 // to one of the headers on the #include stack. Walk the list of the current
750 // headers on the #include stack and pass them to HeaderInfo.
751 if (IsFileLexer()) {
752 if ((CurFileEnt = CurPPLexer->getFileEntry())) {
753 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt,
754 SearchPath, RelativePath,
755 RequestingModule,
756 SuggestedModule))) {
757 if (SuggestedModule && !LangOpts.AsmPreprocessor)
758 HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
759 RequestingModule, RequestingModuleIsModuleInterface, FilenameLoc,
760 Filename, FE);
761 return FE;
762 }
763 }
764 }
765
766 for (IncludeStackInfo &ISEntry : llvm::reverse(IncludeMacroStack)) {
767 if (IsFileLexer(ISEntry)) {
768 if ((CurFileEnt = ISEntry.ThePPLexer->getFileEntry())) {
769 if ((FE = HeaderInfo.LookupSubframeworkHeader(
770 Filename, CurFileEnt, SearchPath, RelativePath,
771 RequestingModule, SuggestedModule))) {
772 if (SuggestedModule && !LangOpts.AsmPreprocessor)
773 HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
774 RequestingModule, RequestingModuleIsModuleInterface,
775 FilenameLoc, Filename, FE);
776 return FE;
777 }
778 }
779 }
780 }
781
782 // Otherwise, we really couldn't find the file.
783 return nullptr;
784 }
785
786 //===----------------------------------------------------------------------===//
787 // Preprocessor Directive Handling.
788 //===----------------------------------------------------------------------===//
789
790 class Preprocessor::ResetMacroExpansionHelper {
791 public:
ResetMacroExpansionHelper(Preprocessor * pp)792 ResetMacroExpansionHelper(Preprocessor *pp)
793 : PP(pp), save(pp->DisableMacroExpansion) {
794 if (pp->MacroExpansionInDirectivesOverride)
795 pp->DisableMacroExpansion = false;
796 }
797
~ResetMacroExpansionHelper()798 ~ResetMacroExpansionHelper() {
799 PP->DisableMacroExpansion = save;
800 }
801
802 private:
803 Preprocessor *PP;
804 bool save;
805 };
806
807 /// Process a directive while looking for the through header or a #pragma
808 /// hdrstop. The following directives are handled:
809 /// #include (to check if it is the through header)
810 /// #define (to warn about macros that don't match the PCH)
811 /// #pragma (to check for pragma hdrstop).
812 /// All other directives are completely discarded.
HandleSkippedDirectiveWhileUsingPCH(Token & Result,SourceLocation HashLoc)813 void Preprocessor::HandleSkippedDirectiveWhileUsingPCH(Token &Result,
814 SourceLocation HashLoc) {
815 if (const IdentifierInfo *II = Result.getIdentifierInfo()) {
816 if (II->getPPKeywordID() == tok::pp_define) {
817 return HandleDefineDirective(Result,
818 /*ImmediatelyAfterHeaderGuard=*/false);
819 }
820 if (SkippingUntilPCHThroughHeader &&
821 II->getPPKeywordID() == tok::pp_include) {
822 return HandleIncludeDirective(HashLoc, Result);
823 }
824 if (SkippingUntilPragmaHdrStop && II->getPPKeywordID() == tok::pp_pragma) {
825 Token P = LookAhead(0);
826 auto *II = P.getIdentifierInfo();
827 if (II && II->getName() == "hdrstop")
828 return HandlePragmaDirective(HashLoc, PIK_HashPragma);
829 }
830 }
831 DiscardUntilEndOfDirective();
832 }
833
834 /// HandleDirective - This callback is invoked when the lexer sees a # token
835 /// at the start of a line. This consumes the directive, modifies the
836 /// lexer/preprocessor state, and advances the lexer(s) so that the next token
837 /// read is the correct one.
HandleDirective(Token & Result)838 void Preprocessor::HandleDirective(Token &Result) {
839 // FIXME: Traditional: # with whitespace before it not recognized by K&R?
840
841 // We just parsed a # character at the start of a line, so we're in directive
842 // mode. Tell the lexer this so any newlines we see will be converted into an
843 // EOD token (which terminates the directive).
844 CurPPLexer->ParsingPreprocessorDirective = true;
845 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
846
847 bool ImmediatelyAfterTopLevelIfndef =
848 CurPPLexer->MIOpt.getImmediatelyAfterTopLevelIfndef();
849 CurPPLexer->MIOpt.resetImmediatelyAfterTopLevelIfndef();
850
851 ++NumDirectives;
852
853 // We are about to read a token. For the multiple-include optimization FA to
854 // work, we have to remember if we had read any tokens *before* this
855 // pp-directive.
856 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal();
857
858 // Save the '#' token in case we need to return it later.
859 Token SavedHash = Result;
860
861 // Read the next token, the directive flavor. This isn't expanded due to
862 // C99 6.10.3p8.
863 LexUnexpandedToken(Result);
864
865 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.:
866 // #define A(x) #x
867 // A(abc
868 // #warning blah
869 // def)
870 // If so, the user is relying on undefined behavior, emit a diagnostic. Do
871 // not support this for #include-like directives, since that can result in
872 // terrible diagnostics, and does not work in GCC.
873 if (InMacroArgs) {
874 if (IdentifierInfo *II = Result.getIdentifierInfo()) {
875 switch (II->getPPKeywordID()) {
876 case tok::pp_include:
877 case tok::pp_import:
878 case tok::pp_include_next:
879 case tok::pp___include_macros:
880 case tok::pp_pragma:
881 Diag(Result, diag::err_embedded_directive) << II->getName();
882 DiscardUntilEndOfDirective();
883 return;
884 default:
885 break;
886 }
887 }
888 Diag(Result, diag::ext_embedded_directive);
889 }
890
891 // Temporarily enable macro expansion if set so
892 // and reset to previous state when returning from this function.
893 ResetMacroExpansionHelper helper(this);
894
895 if (SkippingUntilPCHThroughHeader || SkippingUntilPragmaHdrStop)
896 return HandleSkippedDirectiveWhileUsingPCH(Result, SavedHash.getLocation());
897
898 switch (Result.getKind()) {
899 case tok::eod:
900 return; // null directive.
901 case tok::code_completion:
902 if (CodeComplete)
903 CodeComplete->CodeCompleteDirective(
904 CurPPLexer->getConditionalStackDepth() > 0);
905 setCodeCompletionReached();
906 return;
907 case tok::numeric_constant: // # 7 GNU line marker directive.
908 if (getLangOpts().AsmPreprocessor)
909 break; // # 4 is not a preprocessor directive in .S files.
910 return HandleDigitDirective(Result);
911 default:
912 IdentifierInfo *II = Result.getIdentifierInfo();
913 if (!II) break; // Not an identifier.
914
915 // Ask what the preprocessor keyword ID is.
916 switch (II->getPPKeywordID()) {
917 default: break;
918 // C99 6.10.1 - Conditional Inclusion.
919 case tok::pp_if:
920 return HandleIfDirective(Result, SavedHash, ReadAnyTokensBeforeDirective);
921 case tok::pp_ifdef:
922 return HandleIfdefDirective(Result, SavedHash, false,
923 true /*not valid for miopt*/);
924 case tok::pp_ifndef:
925 return HandleIfdefDirective(Result, SavedHash, true,
926 ReadAnyTokensBeforeDirective);
927 case tok::pp_elif:
928 return HandleElifDirective(Result, SavedHash);
929 case tok::pp_else:
930 return HandleElseDirective(Result, SavedHash);
931 case tok::pp_endif:
932 return HandleEndifDirective(Result);
933
934 // C99 6.10.2 - Source File Inclusion.
935 case tok::pp_include:
936 // Handle #include.
937 return HandleIncludeDirective(SavedHash.getLocation(), Result);
938 case tok::pp___include_macros:
939 // Handle -imacros.
940 return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result);
941
942 // C99 6.10.3 - Macro Replacement.
943 case tok::pp_define:
944 return HandleDefineDirective(Result, ImmediatelyAfterTopLevelIfndef);
945 case tok::pp_undef:
946 return HandleUndefDirective();
947
948 // C99 6.10.4 - Line Control.
949 case tok::pp_line:
950 return HandleLineDirective();
951
952 // C99 6.10.5 - Error Directive.
953 case tok::pp_error:
954 return HandleUserDiagnosticDirective(Result, false);
955
956 // C99 6.10.6 - Pragma Directive.
957 case tok::pp_pragma:
958 return HandlePragmaDirective(SavedHash.getLocation(), PIK_HashPragma);
959
960 // GNU Extensions.
961 case tok::pp_import:
962 return HandleImportDirective(SavedHash.getLocation(), Result);
963 case tok::pp_include_next:
964 return HandleIncludeNextDirective(SavedHash.getLocation(), Result);
965
966 case tok::pp_warning:
967 Diag(Result, diag::ext_pp_warning_directive);
968 return HandleUserDiagnosticDirective(Result, true);
969 case tok::pp_ident:
970 return HandleIdentSCCSDirective(Result);
971 case tok::pp_sccs:
972 return HandleIdentSCCSDirective(Result);
973 case tok::pp_assert:
974 //isExtension = true; // FIXME: implement #assert
975 break;
976 case tok::pp_unassert:
977 //isExtension = true; // FIXME: implement #unassert
978 break;
979
980 case tok::pp___public_macro:
981 if (getLangOpts().Modules)
982 return HandleMacroPublicDirective(Result);
983 break;
984
985 case tok::pp___private_macro:
986 if (getLangOpts().Modules)
987 return HandleMacroPrivateDirective();
988 break;
989 }
990 break;
991 }
992
993 // If this is a .S file, treat unknown # directives as non-preprocessor
994 // directives. This is important because # may be a comment or introduce
995 // various pseudo-ops. Just return the # token and push back the following
996 // token to be lexed next time.
997 if (getLangOpts().AsmPreprocessor) {
998 auto Toks = llvm::make_unique<Token[]>(2);
999 // Return the # and the token after it.
1000 Toks[0] = SavedHash;
1001 Toks[1] = Result;
1002
1003 // If the second token is a hashhash token, then we need to translate it to
1004 // unknown so the token lexer doesn't try to perform token pasting.
1005 if (Result.is(tok::hashhash))
1006 Toks[1].setKind(tok::unknown);
1007
1008 // Enter this token stream so that we re-lex the tokens. Make sure to
1009 // enable macro expansion, in case the token after the # is an identifier
1010 // that is expanded.
1011 EnterTokenStream(std::move(Toks), 2, false);
1012 return;
1013 }
1014
1015 // If we reached here, the preprocessing token is not valid!
1016 Diag(Result, diag::err_pp_invalid_directive);
1017
1018 // Read the rest of the PP line.
1019 DiscardUntilEndOfDirective();
1020
1021 // Okay, we're done parsing the directive.
1022 }
1023
1024 /// GetLineValue - Convert a numeric token into an unsigned value, emitting
1025 /// Diagnostic DiagID if it is invalid, and returning the value in Val.
GetLineValue(Token & DigitTok,unsigned & Val,unsigned DiagID,Preprocessor & PP,bool IsGNULineDirective=false)1026 static bool GetLineValue(Token &DigitTok, unsigned &Val,
1027 unsigned DiagID, Preprocessor &PP,
1028 bool IsGNULineDirective=false) {
1029 if (DigitTok.isNot(tok::numeric_constant)) {
1030 PP.Diag(DigitTok, DiagID);
1031
1032 if (DigitTok.isNot(tok::eod))
1033 PP.DiscardUntilEndOfDirective();
1034 return true;
1035 }
1036
1037 SmallString<64> IntegerBuffer;
1038 IntegerBuffer.resize(DigitTok.getLength());
1039 const char *DigitTokBegin = &IntegerBuffer[0];
1040 bool Invalid = false;
1041 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid);
1042 if (Invalid)
1043 return true;
1044
1045 // Verify that we have a simple digit-sequence, and compute the value. This
1046 // is always a simple digit string computed in decimal, so we do this manually
1047 // here.
1048 Val = 0;
1049 for (unsigned i = 0; i != ActualLength; ++i) {
1050 // C++1y [lex.fcon]p1:
1051 // Optional separating single quotes in a digit-sequence are ignored
1052 if (DigitTokBegin[i] == '\'')
1053 continue;
1054
1055 if (!isDigit(DigitTokBegin[i])) {
1056 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i),
1057 diag::err_pp_line_digit_sequence) << IsGNULineDirective;
1058 PP.DiscardUntilEndOfDirective();
1059 return true;
1060 }
1061
1062 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0');
1063 if (NextVal < Val) { // overflow.
1064 PP.Diag(DigitTok, DiagID);
1065 PP.DiscardUntilEndOfDirective();
1066 return true;
1067 }
1068 Val = NextVal;
1069 }
1070
1071 if (DigitTokBegin[0] == '0' && Val)
1072 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal)
1073 << IsGNULineDirective;
1074
1075 return false;
1076 }
1077
1078 /// Handle a \#line directive: C99 6.10.4.
1079 ///
1080 /// The two acceptable forms are:
1081 /// \verbatim
1082 /// # line digit-sequence
1083 /// # line digit-sequence "s-char-sequence"
1084 /// \endverbatim
HandleLineDirective()1085 void Preprocessor::HandleLineDirective() {
1086 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are
1087 // expanded.
1088 Token DigitTok;
1089 Lex(DigitTok);
1090
1091 // Validate the number and convert it to an unsigned.
1092 unsigned LineNo;
1093 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this))
1094 return;
1095
1096 if (LineNo == 0)
1097 Diag(DigitTok, diag::ext_pp_line_zero);
1098
1099 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
1100 // number greater than 2147483647". C90 requires that the line # be <= 32767.
1101 unsigned LineLimit = 32768U;
1102 if (LangOpts.C99 || LangOpts.CPlusPlus11)
1103 LineLimit = 2147483648U;
1104 if (LineNo >= LineLimit)
1105 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit;
1106 else if (LangOpts.CPlusPlus11 && LineNo >= 32768U)
1107 Diag(DigitTok, diag::warn_cxx98_compat_pp_line_too_big);
1108
1109 int FilenameID = -1;
1110 Token StrTok;
1111 Lex(StrTok);
1112
1113 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a
1114 // string followed by eod.
1115 if (StrTok.is(tok::eod))
1116 ; // ok
1117 else if (StrTok.isNot(tok::string_literal)) {
1118 Diag(StrTok, diag::err_pp_line_invalid_filename);
1119 return DiscardUntilEndOfDirective();
1120 } else if (StrTok.hasUDSuffix()) {
1121 Diag(StrTok, diag::err_invalid_string_udl);
1122 return DiscardUntilEndOfDirective();
1123 } else {
1124 // Parse and validate the string, converting it into a unique ID.
1125 StringLiteralParser Literal(StrTok, *this);
1126 assert(Literal.isAscii() && "Didn't allow wide strings in");
1127 if (Literal.hadError)
1128 return DiscardUntilEndOfDirective();
1129 if (Literal.Pascal) {
1130 Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1131 return DiscardUntilEndOfDirective();
1132 }
1133 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1134
1135 // Verify that there is nothing after the string, other than EOD. Because
1136 // of C99 6.10.4p5, macros that expand to empty tokens are ok.
1137 CheckEndOfDirective("line", true);
1138 }
1139
1140 // Take the file kind of the file containing the #line directive. #line
1141 // directives are often used for generated sources from the same codebase, so
1142 // the new file should generally be classified the same way as the current
1143 // file. This is visible in GCC's pre-processed output, which rewrites #line
1144 // to GNU line markers.
1145 SrcMgr::CharacteristicKind FileKind =
1146 SourceMgr.getFileCharacteristic(DigitTok.getLocation());
1147
1148 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, false,
1149 false, FileKind);
1150
1151 if (Callbacks)
1152 Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
1153 PPCallbacks::RenameFile, FileKind);
1154 }
1155
1156 /// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
1157 /// marker directive.
ReadLineMarkerFlags(bool & IsFileEntry,bool & IsFileExit,SrcMgr::CharacteristicKind & FileKind,Preprocessor & PP)1158 static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
1159 SrcMgr::CharacteristicKind &FileKind,
1160 Preprocessor &PP) {
1161 unsigned FlagVal;
1162 Token FlagTok;
1163 PP.Lex(FlagTok);
1164 if (FlagTok.is(tok::eod)) return false;
1165 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
1166 return true;
1167
1168 if (FlagVal == 1) {
1169 IsFileEntry = true;
1170
1171 PP.Lex(FlagTok);
1172 if (FlagTok.is(tok::eod)) return false;
1173 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
1174 return true;
1175 } else if (FlagVal == 2) {
1176 IsFileExit = true;
1177
1178 SourceManager &SM = PP.getSourceManager();
1179 // If we are leaving the current presumed file, check to make sure the
1180 // presumed include stack isn't empty!
1181 FileID CurFileID =
1182 SM.getDecomposedExpansionLoc(FlagTok.getLocation()).first;
1183 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation());
1184 if (PLoc.isInvalid())
1185 return true;
1186
1187 // If there is no include loc (main file) or if the include loc is in a
1188 // different physical file, then we aren't in a "1" line marker flag region.
1189 SourceLocation IncLoc = PLoc.getIncludeLoc();
1190 if (IncLoc.isInvalid() ||
1191 SM.getDecomposedExpansionLoc(IncLoc).first != CurFileID) {
1192 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop);
1193 PP.DiscardUntilEndOfDirective();
1194 return true;
1195 }
1196
1197 PP.Lex(FlagTok);
1198 if (FlagTok.is(tok::eod)) return false;
1199 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
1200 return true;
1201 }
1202
1203 // We must have 3 if there are still flags.
1204 if (FlagVal != 3) {
1205 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1206 PP.DiscardUntilEndOfDirective();
1207 return true;
1208 }
1209
1210 FileKind = SrcMgr::C_System;
1211
1212 PP.Lex(FlagTok);
1213 if (FlagTok.is(tok::eod)) return false;
1214 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
1215 return true;
1216
1217 // We must have 4 if there is yet another flag.
1218 if (FlagVal != 4) {
1219 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1220 PP.DiscardUntilEndOfDirective();
1221 return true;
1222 }
1223
1224 FileKind = SrcMgr::C_ExternCSystem;
1225
1226 PP.Lex(FlagTok);
1227 if (FlagTok.is(tok::eod)) return false;
1228
1229 // There are no more valid flags here.
1230 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1231 PP.DiscardUntilEndOfDirective();
1232 return true;
1233 }
1234
1235 /// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
1236 /// one of the following forms:
1237 ///
1238 /// # 42
1239 /// # 42 "file" ('1' | '2')?
1240 /// # 42 "file" ('1' | '2')? '3' '4'?
1241 ///
HandleDigitDirective(Token & DigitTok)1242 void Preprocessor::HandleDigitDirective(Token &DigitTok) {
1243 // Validate the number and convert it to an unsigned. GNU does not have a
1244 // line # limit other than it fit in 32-bits.
1245 unsigned LineNo;
1246 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer,
1247 *this, true))
1248 return;
1249
1250 Token StrTok;
1251 Lex(StrTok);
1252
1253 bool IsFileEntry = false, IsFileExit = false;
1254 int FilenameID = -1;
1255 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User;
1256
1257 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a
1258 // string followed by eod.
1259 if (StrTok.is(tok::eod)) {
1260 // Treat this like "#line NN", which doesn't change file characteristics.
1261 FileKind = SourceMgr.getFileCharacteristic(DigitTok.getLocation());
1262 } else if (StrTok.isNot(tok::string_literal)) {
1263 Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1264 return DiscardUntilEndOfDirective();
1265 } else if (StrTok.hasUDSuffix()) {
1266 Diag(StrTok, diag::err_invalid_string_udl);
1267 return DiscardUntilEndOfDirective();
1268 } else {
1269 // Parse and validate the string, converting it into a unique ID.
1270 StringLiteralParser Literal(StrTok, *this);
1271 assert(Literal.isAscii() && "Didn't allow wide strings in");
1272 if (Literal.hadError)
1273 return DiscardUntilEndOfDirective();
1274 if (Literal.Pascal) {
1275 Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1276 return DiscardUntilEndOfDirective();
1277 }
1278 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1279
1280 // If a filename was present, read any flags that are present.
1281 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, FileKind, *this))
1282 return;
1283 }
1284
1285 // Create a line note with this information.
1286 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, IsFileEntry,
1287 IsFileExit, FileKind);
1288
1289 // If the preprocessor has callbacks installed, notify them of the #line
1290 // change. This is used so that the line marker comes out in -E mode for
1291 // example.
1292 if (Callbacks) {
1293 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile;
1294 if (IsFileEntry)
1295 Reason = PPCallbacks::EnterFile;
1296 else if (IsFileExit)
1297 Reason = PPCallbacks::ExitFile;
1298
1299 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind);
1300 }
1301 }
1302
1303 /// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
1304 ///
HandleUserDiagnosticDirective(Token & Tok,bool isWarning)1305 void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
1306 bool isWarning) {
1307 // Read the rest of the line raw. We do this because we don't want macros
1308 // to be expanded and we don't require that the tokens be valid preprocessing
1309 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does
1310 // collapse multiple consecutive white space between tokens, but this isn't
1311 // specified by the standard.
1312 SmallString<128> Message;
1313 CurLexer->ReadToEndOfLine(&Message);
1314
1315 // Find the first non-whitespace character, so that we can make the
1316 // diagnostic more succinct.
1317 StringRef Msg = StringRef(Message).ltrim(' ');
1318
1319 if (isWarning)
1320 Diag(Tok, diag::pp_hash_warning) << Msg;
1321 else
1322 Diag(Tok, diag::err_pp_hash_error) << Msg;
1323 }
1324
1325 /// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
1326 ///
HandleIdentSCCSDirective(Token & Tok)1327 void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
1328 // Yes, this directive is an extension.
1329 Diag(Tok, diag::ext_pp_ident_directive);
1330
1331 // Read the string argument.
1332 Token StrTok;
1333 Lex(StrTok);
1334
1335 // If the token kind isn't a string, it's a malformed directive.
1336 if (StrTok.isNot(tok::string_literal) &&
1337 StrTok.isNot(tok::wide_string_literal)) {
1338 Diag(StrTok, diag::err_pp_malformed_ident);
1339 if (StrTok.isNot(tok::eod))
1340 DiscardUntilEndOfDirective();
1341 return;
1342 }
1343
1344 if (StrTok.hasUDSuffix()) {
1345 Diag(StrTok, diag::err_invalid_string_udl);
1346 return DiscardUntilEndOfDirective();
1347 }
1348
1349 // Verify that there is nothing after the string, other than EOD.
1350 CheckEndOfDirective("ident");
1351
1352 if (Callbacks) {
1353 bool Invalid = false;
1354 std::string Str = getSpelling(StrTok, &Invalid);
1355 if (!Invalid)
1356 Callbacks->Ident(Tok.getLocation(), Str);
1357 }
1358 }
1359
1360 /// Handle a #public directive.
HandleMacroPublicDirective(Token & Tok)1361 void Preprocessor::HandleMacroPublicDirective(Token &Tok) {
1362 Token MacroNameTok;
1363 ReadMacroName(MacroNameTok, MU_Undef);
1364
1365 // Error reading macro name? If so, diagnostic already issued.
1366 if (MacroNameTok.is(tok::eod))
1367 return;
1368
1369 // Check to see if this is the last token on the #__public_macro line.
1370 CheckEndOfDirective("__public_macro");
1371
1372 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1373 // Okay, we finally have a valid identifier to undef.
1374 MacroDirective *MD = getLocalMacroDirective(II);
1375
1376 // If the macro is not defined, this is an error.
1377 if (!MD) {
1378 Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1379 return;
1380 }
1381
1382 // Note that this macro has now been exported.
1383 appendMacroDirective(II, AllocateVisibilityMacroDirective(
1384 MacroNameTok.getLocation(), /*IsPublic=*/true));
1385 }
1386
1387 /// Handle a #private directive.
HandleMacroPrivateDirective()1388 void Preprocessor::HandleMacroPrivateDirective() {
1389 Token MacroNameTok;
1390 ReadMacroName(MacroNameTok, MU_Undef);
1391
1392 // Error reading macro name? If so, diagnostic already issued.
1393 if (MacroNameTok.is(tok::eod))
1394 return;
1395
1396 // Check to see if this is the last token on the #__private_macro line.
1397 CheckEndOfDirective("__private_macro");
1398
1399 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1400 // Okay, we finally have a valid identifier to undef.
1401 MacroDirective *MD = getLocalMacroDirective(II);
1402
1403 // If the macro is not defined, this is an error.
1404 if (!MD) {
1405 Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1406 return;
1407 }
1408
1409 // Note that this macro has now been marked private.
1410 appendMacroDirective(II, AllocateVisibilityMacroDirective(
1411 MacroNameTok.getLocation(), /*IsPublic=*/false));
1412 }
1413
1414 //===----------------------------------------------------------------------===//
1415 // Preprocessor Include Directive Handling.
1416 //===----------------------------------------------------------------------===//
1417
1418 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
1419 /// checked and spelled filename, e.g. as an operand of \#include. This returns
1420 /// true if the input filename was in <>'s or false if it were in ""'s. The
1421 /// caller is expected to provide a buffer that is large enough to hold the
1422 /// spelling of the filename, but is also expected to handle the case when
1423 /// this method decides to use a different buffer.
GetIncludeFilenameSpelling(SourceLocation Loc,StringRef & Buffer)1424 bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
1425 StringRef &Buffer) {
1426 // Get the text form of the filename.
1427 assert(!Buffer.empty() && "Can't have tokens with empty spellings!");
1428
1429 // Make sure the filename is <x> or "x".
1430 bool isAngled;
1431 if (Buffer[0] == '<') {
1432 if (Buffer.back() != '>') {
1433 Diag(Loc, diag::err_pp_expects_filename);
1434 Buffer = StringRef();
1435 return true;
1436 }
1437 isAngled = true;
1438 } else if (Buffer[0] == '"') {
1439 if (Buffer.back() != '"') {
1440 Diag(Loc, diag::err_pp_expects_filename);
1441 Buffer = StringRef();
1442 return true;
1443 }
1444 isAngled = false;
1445 } else {
1446 Diag(Loc, diag::err_pp_expects_filename);
1447 Buffer = StringRef();
1448 return true;
1449 }
1450
1451 // Diagnose #include "" as invalid.
1452 if (Buffer.size() <= 2) {
1453 Diag(Loc, diag::err_pp_empty_filename);
1454 Buffer = StringRef();
1455 return true;
1456 }
1457
1458 // Skip the brackets.
1459 Buffer = Buffer.substr(1, Buffer.size()-2);
1460 return isAngled;
1461 }
1462
1463 // Handle cases where the \#include name is expanded from a macro
1464 // as multiple tokens, which need to be glued together.
1465 //
1466 // This occurs for code like:
1467 // \code
1468 // \#define FOO <a/b.h>
1469 // \#include FOO
1470 // \endcode
1471 // because in this case, "<a/b.h>" is returned as 7 tokens, not one.
1472 //
1473 // This code concatenates and consumes tokens up to the '>' token. It returns
1474 // false if the > was found, otherwise it returns true if it finds and consumes
1475 // the EOD marker.
ConcatenateIncludeName(SmallString<128> & FilenameBuffer,SourceLocation & End)1476 bool Preprocessor::ConcatenateIncludeName(SmallString<128> &FilenameBuffer,
1477 SourceLocation &End) {
1478 Token CurTok;
1479
1480 Lex(CurTok);
1481 while (CurTok.isNot(tok::eod)) {
1482 End = CurTok.getLocation();
1483
1484 // FIXME: Provide code completion for #includes.
1485 if (CurTok.is(tok::code_completion)) {
1486 setCodeCompletionReached();
1487 Lex(CurTok);
1488 continue;
1489 }
1490
1491 // Append the spelling of this token to the buffer. If there was a space
1492 // before it, add it now.
1493 if (CurTok.hasLeadingSpace())
1494 FilenameBuffer.push_back(' ');
1495
1496 // Get the spelling of the token, directly into FilenameBuffer if possible.
1497 size_t PreAppendSize = FilenameBuffer.size();
1498 FilenameBuffer.resize(PreAppendSize+CurTok.getLength());
1499
1500 const char *BufPtr = &FilenameBuffer[PreAppendSize];
1501 unsigned ActualLen = getSpelling(CurTok, BufPtr);
1502
1503 // If the token was spelled somewhere else, copy it into FilenameBuffer.
1504 if (BufPtr != &FilenameBuffer[PreAppendSize])
1505 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1506
1507 // Resize FilenameBuffer to the correct size.
1508 if (CurTok.getLength() != ActualLen)
1509 FilenameBuffer.resize(PreAppendSize+ActualLen);
1510
1511 // If we found the '>' marker, return success.
1512 if (CurTok.is(tok::greater))
1513 return false;
1514
1515 Lex(CurTok);
1516 }
1517
1518 // If we hit the eod marker, emit an error and return true so that the caller
1519 // knows the EOD has been read.
1520 Diag(CurTok.getLocation(), diag::err_pp_expects_filename);
1521 return true;
1522 }
1523
1524 /// Push a token onto the token stream containing an annotation.
EnterAnnotationToken(SourceRange Range,tok::TokenKind Kind,void * AnnotationVal)1525 void Preprocessor::EnterAnnotationToken(SourceRange Range,
1526 tok::TokenKind Kind,
1527 void *AnnotationVal) {
1528 // FIXME: Produce this as the current token directly, rather than
1529 // allocating a new token for it.
1530 auto Tok = llvm::make_unique<Token[]>(1);
1531 Tok[0].startToken();
1532 Tok[0].setKind(Kind);
1533 Tok[0].setLocation(Range.getBegin());
1534 Tok[0].setAnnotationEndLoc(Range.getEnd());
1535 Tok[0].setAnnotationValue(AnnotationVal);
1536 EnterTokenStream(std::move(Tok), 1, true);
1537 }
1538
1539 /// Produce a diagnostic informing the user that a #include or similar
1540 /// was implicitly treated as a module import.
diagnoseAutoModuleImport(Preprocessor & PP,SourceLocation HashLoc,Token & IncludeTok,ArrayRef<std::pair<IdentifierInfo *,SourceLocation>> Path,SourceLocation PathEnd)1541 static void diagnoseAutoModuleImport(
1542 Preprocessor &PP, SourceLocation HashLoc, Token &IncludeTok,
1543 ArrayRef<std::pair<IdentifierInfo *, SourceLocation>> Path,
1544 SourceLocation PathEnd) {
1545 assert(PP.getLangOpts().ObjC && "no import syntax available");
1546
1547 SmallString<128> PathString;
1548 for (size_t I = 0, N = Path.size(); I != N; ++I) {
1549 if (I)
1550 PathString += '.';
1551 PathString += Path[I].first->getName();
1552 }
1553 int IncludeKind = 0;
1554
1555 switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) {
1556 case tok::pp_include:
1557 IncludeKind = 0;
1558 break;
1559
1560 case tok::pp_import:
1561 IncludeKind = 1;
1562 break;
1563
1564 case tok::pp_include_next:
1565 IncludeKind = 2;
1566 break;
1567
1568 case tok::pp___include_macros:
1569 IncludeKind = 3;
1570 break;
1571
1572 default:
1573 llvm_unreachable("unknown include directive kind");
1574 }
1575
1576 CharSourceRange ReplaceRange(SourceRange(HashLoc, PathEnd),
1577 /*IsTokenRange=*/false);
1578 PP.Diag(HashLoc, diag::warn_auto_module_import)
1579 << IncludeKind << PathString
1580 << FixItHint::CreateReplacement(ReplaceRange,
1581 ("@import " + PathString + ";").str());
1582 }
1583
1584 // Given a vector of path components and a string containing the real
1585 // path to the file, build a properly-cased replacement in the vector,
1586 // and return true if the replacement should be suggested.
trySimplifyPath(SmallVectorImpl<StringRef> & Components,StringRef RealPathName)1587 static bool trySimplifyPath(SmallVectorImpl<StringRef> &Components,
1588 StringRef RealPathName) {
1589 auto RealPathComponentIter = llvm::sys::path::rbegin(RealPathName);
1590 auto RealPathComponentEnd = llvm::sys::path::rend(RealPathName);
1591 int Cnt = 0;
1592 bool SuggestReplacement = false;
1593 // Below is a best-effort to handle ".." in paths. It is admittedly
1594 // not 100% correct in the presence of symlinks.
1595 for (auto &Component : llvm::reverse(Components)) {
1596 if ("." == Component) {
1597 } else if (".." == Component) {
1598 ++Cnt;
1599 } else if (Cnt) {
1600 --Cnt;
1601 } else if (RealPathComponentIter != RealPathComponentEnd) {
1602 if (Component != *RealPathComponentIter) {
1603 // If these path components differ by more than just case, then we
1604 // may be looking at symlinked paths. Bail on this diagnostic to avoid
1605 // noisy false positives.
1606 SuggestReplacement = RealPathComponentIter->equals_lower(Component);
1607 if (!SuggestReplacement)
1608 break;
1609 Component = *RealPathComponentIter;
1610 }
1611 ++RealPathComponentIter;
1612 }
1613 }
1614 return SuggestReplacement;
1615 }
1616
checkModuleIsAvailable(const LangOptions & LangOpts,const TargetInfo & TargetInfo,DiagnosticsEngine & Diags,Module * M)1617 bool Preprocessor::checkModuleIsAvailable(const LangOptions &LangOpts,
1618 const TargetInfo &TargetInfo,
1619 DiagnosticsEngine &Diags, Module *M) {
1620 Module::Requirement Requirement;
1621 Module::UnresolvedHeaderDirective MissingHeader;
1622 Module *ShadowingModule = nullptr;
1623 if (M->isAvailable(LangOpts, TargetInfo, Requirement, MissingHeader,
1624 ShadowingModule))
1625 return false;
1626
1627 if (MissingHeader.FileNameLoc.isValid()) {
1628 Diags.Report(MissingHeader.FileNameLoc, diag::err_module_header_missing)
1629 << MissingHeader.IsUmbrella << MissingHeader.FileName;
1630 } else if (ShadowingModule) {
1631 Diags.Report(M->DefinitionLoc, diag::err_module_shadowed) << M->Name;
1632 Diags.Report(ShadowingModule->DefinitionLoc,
1633 diag::note_previous_definition);
1634 } else {
1635 // FIXME: Track the location at which the requirement was specified, and
1636 // use it here.
1637 Diags.Report(M->DefinitionLoc, diag::err_module_unavailable)
1638 << M->getFullModuleName() << Requirement.second << Requirement.first;
1639 }
1640 return true;
1641 }
1642
1643 /// HandleIncludeDirective - The "\#include" tokens have just been read, read
1644 /// the file to be included from the lexer, then include it! This is a common
1645 /// routine with functionality shared between \#include, \#include_next and
1646 /// \#import. LookupFrom is set when this is a \#include_next directive, it
1647 /// specifies the file to start searching from.
HandleIncludeDirective(SourceLocation HashLoc,Token & IncludeTok,const DirectoryLookup * LookupFrom,const FileEntry * LookupFromFile,bool isImport)1648 void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc,
1649 Token &IncludeTok,
1650 const DirectoryLookup *LookupFrom,
1651 const FileEntry *LookupFromFile,
1652 bool isImport) {
1653 Token FilenameTok;
1654 CurPPLexer->LexIncludeFilename(FilenameTok);
1655
1656 // Reserve a buffer to get the spelling.
1657 SmallString<128> FilenameBuffer;
1658 StringRef Filename;
1659 SourceLocation End;
1660 SourceLocation CharEnd; // the end of this directive, in characters
1661
1662 switch (FilenameTok.getKind()) {
1663 case tok::eod:
1664 // If the token kind is EOD, the error has already been diagnosed.
1665 return;
1666
1667 case tok::angle_string_literal:
1668 case tok::string_literal:
1669 Filename = getSpelling(FilenameTok, FilenameBuffer);
1670 End = FilenameTok.getLocation();
1671 CharEnd = End.getLocWithOffset(FilenameTok.getLength());
1672 break;
1673
1674 case tok::less:
1675 // This could be a <foo/bar.h> file coming from a macro expansion. In this
1676 // case, glue the tokens together into FilenameBuffer and interpret those.
1677 FilenameBuffer.push_back('<');
1678 if (ConcatenateIncludeName(FilenameBuffer, End))
1679 return; // Found <eod> but no ">"? Diagnostic already emitted.
1680 Filename = FilenameBuffer;
1681 CharEnd = End.getLocWithOffset(1);
1682 break;
1683 default:
1684 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
1685 DiscardUntilEndOfDirective();
1686 return;
1687 }
1688
1689 CharSourceRange FilenameRange
1690 = CharSourceRange::getCharRange(FilenameTok.getLocation(), CharEnd);
1691 StringRef OriginalFilename = Filename;
1692 bool isAngled =
1693 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename);
1694 // If GetIncludeFilenameSpelling set the start ptr to null, there was an
1695 // error.
1696 if (Filename.empty()) {
1697 DiscardUntilEndOfDirective();
1698 return;
1699 }
1700
1701 // Verify that there is nothing after the filename, other than EOD. Note that
1702 // we allow macros that expand to nothing after the filename, because this
1703 // falls into the category of "#include pp-tokens new-line" specified in
1704 // C99 6.10.2p4.
1705 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true);
1706
1707 // Check that we don't have infinite #include recursion.
1708 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
1709 Diag(FilenameTok, diag::err_pp_include_too_deep);
1710 HasReachedMaxIncludeDepth = true;
1711 return;
1712 }
1713
1714 // Complain about attempts to #include files in an audit pragma.
1715 if (PragmaARCCFCodeAuditedLoc.isValid()) {
1716 Diag(HashLoc, diag::err_pp_include_in_arc_cf_code_audited);
1717 Diag(PragmaARCCFCodeAuditedLoc, diag::note_pragma_entered_here);
1718
1719 // Immediately leave the pragma.
1720 PragmaARCCFCodeAuditedLoc = SourceLocation();
1721 }
1722
1723 // Complain about attempts to #include files in an assume-nonnull pragma.
1724 if (PragmaAssumeNonNullLoc.isValid()) {
1725 Diag(HashLoc, diag::err_pp_include_in_assume_nonnull);
1726 Diag(PragmaAssumeNonNullLoc, diag::note_pragma_entered_here);
1727
1728 // Immediately leave the pragma.
1729 PragmaAssumeNonNullLoc = SourceLocation();
1730 }
1731
1732 if (HeaderInfo.HasIncludeAliasMap()) {
1733 // Map the filename with the brackets still attached. If the name doesn't
1734 // map to anything, fall back on the filename we've already gotten the
1735 // spelling for.
1736 StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(OriginalFilename);
1737 if (!NewName.empty())
1738 Filename = NewName;
1739 }
1740
1741 // Search include directories.
1742 bool IsMapped = false;
1743 const DirectoryLookup *CurDir;
1744 SmallString<1024> SearchPath;
1745 SmallString<1024> RelativePath;
1746 // We get the raw path only if we have 'Callbacks' to which we later pass
1747 // the path.
1748 ModuleMap::KnownHeader SuggestedModule;
1749 SourceLocation FilenameLoc = FilenameTok.getLocation();
1750 SmallString<128> NormalizedPath;
1751 if (LangOpts.MSVCCompat) {
1752 NormalizedPath = Filename.str();
1753 #ifndef _WIN32
1754 llvm::sys::path::native(NormalizedPath);
1755 #endif
1756 }
1757 const FileEntry *File = LookupFile(
1758 FilenameLoc, LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename,
1759 isAngled, LookupFrom, LookupFromFile, CurDir,
1760 Callbacks ? &SearchPath : nullptr, Callbacks ? &RelativePath : nullptr,
1761 &SuggestedModule, &IsMapped);
1762
1763 if (!File) {
1764 if (Callbacks) {
1765 // Give the clients a chance to recover.
1766 SmallString<128> RecoveryPath;
1767 if (Callbacks->FileNotFound(Filename, RecoveryPath)) {
1768 if (const DirectoryEntry *DE = FileMgr.getDirectory(RecoveryPath)) {
1769 // Add the recovery path to the list of search paths.
1770 DirectoryLookup DL(DE, SrcMgr::C_User, false);
1771 HeaderInfo.AddSearchPath(DL, isAngled);
1772
1773 // Try the lookup again, skipping the cache.
1774 File = LookupFile(
1775 FilenameLoc,
1776 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, isAngled,
1777 LookupFrom, LookupFromFile, CurDir, nullptr, nullptr,
1778 &SuggestedModule, &IsMapped, /*SkipCache*/ true);
1779 }
1780 }
1781 }
1782
1783 if (!SuppressIncludeNotFoundError) {
1784 // If the file could not be located and it was included via angle
1785 // brackets, we can attempt a lookup as though it were a quoted path to
1786 // provide the user with a possible fixit.
1787 if (isAngled) {
1788 File = LookupFile(
1789 FilenameLoc,
1790 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, false,
1791 LookupFrom, LookupFromFile, CurDir,
1792 Callbacks ? &SearchPath : nullptr,
1793 Callbacks ? &RelativePath : nullptr, &SuggestedModule, &IsMapped);
1794 if (File) {
1795 SourceRange Range(FilenameTok.getLocation(), CharEnd);
1796 Diag(FilenameTok, diag::err_pp_file_not_found_angled_include_not_fatal) <<
1797 Filename <<
1798 FixItHint::CreateReplacement(Range, "\"" + Filename.str() + "\"");
1799 }
1800 }
1801
1802 // Check for likely typos due to leading or trailing non-isAlphanumeric
1803 // characters
1804 StringRef OriginalFilename = Filename;
1805 if (LangOpts.SpellChecking && !File) {
1806 // A heuristic to correct a typo file name by removing leading and
1807 // trailing non-isAlphanumeric characters.
1808 auto CorrectTypoFilename = [](llvm::StringRef Filename) {
1809 Filename = Filename.drop_until(isAlphanumeric);
1810 while (!Filename.empty() && !isAlphanumeric(Filename.back())) {
1811 Filename = Filename.drop_back();
1812 }
1813 return Filename;
1814 };
1815 StringRef TypoCorrectionName = CorrectTypoFilename(Filename);
1816 SmallString<128> NormalizedTypoCorrectionPath;
1817 if (LangOpts.MSVCCompat) {
1818 NormalizedTypoCorrectionPath = TypoCorrectionName.str();
1819 #ifndef _WIN32
1820 llvm::sys::path::native(NormalizedTypoCorrectionPath);
1821 #endif
1822 }
1823 File = LookupFile(
1824 FilenameLoc,
1825 LangOpts.MSVCCompat ? NormalizedTypoCorrectionPath.c_str()
1826 : TypoCorrectionName,
1827 isAngled, LookupFrom, LookupFromFile, CurDir,
1828 Callbacks ? &SearchPath : nullptr,
1829 Callbacks ? &RelativePath : nullptr, &SuggestedModule, &IsMapped);
1830 if (File) {
1831 SourceRange Range(FilenameTok.getLocation(), CharEnd);
1832 auto Hint = isAngled
1833 ? FixItHint::CreateReplacement(
1834 Range, "<" + TypoCorrectionName.str() + ">")
1835 : FixItHint::CreateReplacement(
1836 Range, "\"" + TypoCorrectionName.str() + "\"");
1837 Diag(FilenameTok, diag::err_pp_file_not_found_typo_not_fatal)
1838 << OriginalFilename << TypoCorrectionName << Hint;
1839 // We found the file, so set the Filename to the name after typo
1840 // correction.
1841 Filename = TypoCorrectionName;
1842 }
1843 }
1844
1845 // If the file is still not found, just go with the vanilla diagnostic
1846 if (!File)
1847 Diag(FilenameTok, diag::err_pp_file_not_found) << OriginalFilename
1848 << FilenameRange;
1849 }
1850 }
1851
1852 if (usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) {
1853 if (isPCHThroughHeader(File))
1854 SkippingUntilPCHThroughHeader = false;
1855 return;
1856 }
1857
1858 // Should we enter the source file? Set to false if either the source file is
1859 // known to have no effect beyond its effect on module visibility -- that is,
1860 // if it's got an include guard that is already defined or is a modular header
1861 // we've imported or already built.
1862 bool ShouldEnter = true;
1863
1864 if (PPOpts->SingleFileParseMode)
1865 ShouldEnter = false;
1866
1867 // If we've reached the max allowed include depth, it is usually due to an
1868 // include cycle. Don't enter already processed files again as it can lead to
1869 // reaching the max allowed include depth again.
1870 if (ShouldEnter && HasReachedMaxIncludeDepth && File &&
1871 HeaderInfo.getFileInfo(File).NumIncludes)
1872 ShouldEnter = false;
1873
1874 // Determine whether we should try to import the module for this #include, if
1875 // there is one. Don't do so if precompiled module support is disabled or we
1876 // are processing this module textually (because we're building the module).
1877 if (ShouldEnter && File && SuggestedModule && getLangOpts().Modules &&
1878 !isForModuleBuilding(SuggestedModule.getModule(),
1879 getLangOpts().CurrentModule,
1880 getLangOpts().ModuleName)) {
1881 // If this include corresponds to a module but that module is
1882 // unavailable, diagnose the situation and bail out.
1883 // FIXME: Remove this; loadModule does the same check (but produces
1884 // slightly worse diagnostics).
1885 if (checkModuleIsAvailable(getLangOpts(), getTargetInfo(), getDiagnostics(),
1886 SuggestedModule.getModule())) {
1887 Diag(FilenameTok.getLocation(),
1888 diag::note_implicit_top_level_module_import_here)
1889 << SuggestedModule.getModule()->getTopLevelModuleName();
1890 return;
1891 }
1892
1893 // Compute the module access path corresponding to this module.
1894 // FIXME: Should we have a second loadModule() overload to avoid this
1895 // extra lookup step?
1896 SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> Path;
1897 for (Module *Mod = SuggestedModule.getModule(); Mod; Mod = Mod->Parent)
1898 Path.push_back(std::make_pair(getIdentifierInfo(Mod->Name),
1899 FilenameTok.getLocation()));
1900 std::reverse(Path.begin(), Path.end());
1901
1902 // Warn that we're replacing the include/import with a module import.
1903 // We only do this in Objective-C, where we have a module-import syntax.
1904 if (getLangOpts().ObjC)
1905 diagnoseAutoModuleImport(*this, HashLoc, IncludeTok, Path, CharEnd);
1906
1907 // Load the module to import its macros. We'll make the declarations
1908 // visible when the parser gets here.
1909 // FIXME: Pass SuggestedModule in here rather than converting it to a path
1910 // and making the module loader convert it back again.
1911 ModuleLoadResult Imported = TheModuleLoader.loadModule(
1912 IncludeTok.getLocation(), Path, Module::Hidden,
1913 /*IsIncludeDirective=*/true);
1914 assert((Imported == nullptr || Imported == SuggestedModule.getModule()) &&
1915 "the imported module is different than the suggested one");
1916
1917 if (Imported)
1918 ShouldEnter = false;
1919 else if (Imported.isMissingExpected()) {
1920 // We failed to find a submodule that we assumed would exist (because it
1921 // was in the directory of an umbrella header, for instance), but no
1922 // actual module containing it exists (because the umbrella header is
1923 // incomplete). Treat this as a textual inclusion.
1924 SuggestedModule = ModuleMap::KnownHeader();
1925 } else if (Imported.isConfigMismatch()) {
1926 // On a configuration mismatch, enter the header textually. We still know
1927 // that it's part of the corresponding module.
1928 } else {
1929 // We hit an error processing the import. Bail out.
1930 if (hadModuleLoaderFatalFailure()) {
1931 // With a fatal failure in the module loader, we abort parsing.
1932 Token &Result = IncludeTok;
1933 assert(CurLexer && "#include but no current lexer set!");
1934 Result.startToken();
1935 CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof);
1936 CurLexer->cutOffLexing();
1937 }
1938 return;
1939 }
1940 }
1941
1942 // The #included file will be considered to be a system header if either it is
1943 // in a system include directory, or if the #includer is a system include
1944 // header.
1945 SrcMgr::CharacteristicKind FileCharacter =
1946 SourceMgr.getFileCharacteristic(FilenameTok.getLocation());
1947 if (File)
1948 FileCharacter = std::max(HeaderInfo.getFileDirFlavor(File), FileCharacter);
1949
1950 // Ask HeaderInfo if we should enter this #include file. If not, #including
1951 // this file will have no effect.
1952 bool SkipHeader = false;
1953 if (ShouldEnter && File &&
1954 !HeaderInfo.ShouldEnterIncludeFile(*this, File, isImport,
1955 getLangOpts().Modules,
1956 SuggestedModule.getModule())) {
1957 ShouldEnter = false;
1958 SkipHeader = true;
1959 }
1960
1961 if (Callbacks) {
1962 // Notify the callback object that we've seen an inclusion directive.
1963 Callbacks->InclusionDirective(
1964 HashLoc, IncludeTok,
1965 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, isAngled,
1966 FilenameRange, File, SearchPath, RelativePath,
1967 ShouldEnter ? nullptr : SuggestedModule.getModule(), FileCharacter);
1968 if (SkipHeader && !SuggestedModule.getModule())
1969 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter);
1970 }
1971
1972 if (!File)
1973 return;
1974
1975 // FIXME: If we have a suggested module, and we've already visited this file,
1976 // don't bother entering it again. We know it has no further effect.
1977
1978 // Issue a diagnostic if the name of the file on disk has a different case
1979 // than the one we're about to open.
1980 const bool CheckIncludePathPortability =
1981 !IsMapped && File && !File->tryGetRealPathName().empty();
1982
1983 if (CheckIncludePathPortability) {
1984 StringRef Name = LangOpts.MSVCCompat ? NormalizedPath.str() : Filename;
1985 StringRef RealPathName = File->tryGetRealPathName();
1986 SmallVector<StringRef, 16> Components(llvm::sys::path::begin(Name),
1987 llvm::sys::path::end(Name));
1988
1989 if (trySimplifyPath(Components, RealPathName)) {
1990 SmallString<128> Path;
1991 Path.reserve(Name.size()+2);
1992 Path.push_back(isAngled ? '<' : '"');
1993 bool isLeadingSeparator = llvm::sys::path::is_absolute(Name);
1994 for (auto Component : Components) {
1995 if (isLeadingSeparator)
1996 isLeadingSeparator = false;
1997 else
1998 Path.append(Component);
1999 // Append the separator the user used, or the close quote
2000 Path.push_back(
2001 Path.size() <= Filename.size() ? Filename[Path.size()-1] :
2002 (isAngled ? '>' : '"'));
2003 }
2004 // For user files and known standard headers, by default we issue a diagnostic.
2005 // For other system headers, we don't. They can be controlled separately.
2006 auto DiagId = (FileCharacter == SrcMgr::C_User || warnByDefaultOnWrongCase(Name)) ?
2007 diag::pp_nonportable_path : diag::pp_nonportable_system_path;
2008 SourceRange Range(FilenameTok.getLocation(), CharEnd);
2009 Diag(FilenameTok, DiagId) << Path <<
2010 FixItHint::CreateReplacement(Range, Path);
2011 }
2012 }
2013
2014 // If we don't need to enter the file, stop now.
2015 if (!ShouldEnter) {
2016 // If this is a module import, make it visible if needed.
2017 if (auto *M = SuggestedModule.getModule()) {
2018 // When building a pch, -fmodule-name tells the compiler to textually
2019 // include headers in the specified module. But it is possible that
2020 // ShouldEnter is false because we are skipping the header. In that
2021 // case, We are not importing the specified module.
2022 if (SkipHeader && getLangOpts().CompilingPCH &&
2023 isForModuleBuilding(M, getLangOpts().CurrentModule,
2024 getLangOpts().ModuleName))
2025 return;
2026
2027 makeModuleVisible(M, HashLoc);
2028
2029 if (IncludeTok.getIdentifierInfo()->getPPKeywordID() !=
2030 tok::pp___include_macros)
2031 EnterAnnotationToken(SourceRange(HashLoc, End),
2032 tok::annot_module_include, M);
2033 }
2034 return;
2035 }
2036
2037 // Look up the file, create a File ID for it.
2038 SourceLocation IncludePos = End;
2039 // If the filename string was the result of macro expansions, set the include
2040 // position on the file where it will be included and after the expansions.
2041 if (IncludePos.isMacroID())
2042 IncludePos = SourceMgr.getExpansionRange(IncludePos).getEnd();
2043 FileID FID = SourceMgr.createFileID(File, IncludePos, FileCharacter);
2044 assert(FID.isValid() && "Expected valid file ID");
2045
2046 // If all is good, enter the new file!
2047 if (EnterSourceFile(FID, CurDir, FilenameTok.getLocation()))
2048 return;
2049
2050 // Determine if we're switching to building a new submodule, and which one.
2051 if (auto *M = SuggestedModule.getModule()) {
2052 if (M->getTopLevelModule()->ShadowingModule) {
2053 // We are building a submodule that belongs to a shadowed module. This
2054 // means we find header files in the shadowed module.
2055 Diag(M->DefinitionLoc, diag::err_module_build_shadowed_submodule)
2056 << M->getFullModuleName();
2057 Diag(M->getTopLevelModule()->ShadowingModule->DefinitionLoc,
2058 diag::note_previous_definition);
2059 return;
2060 }
2061 // When building a pch, -fmodule-name tells the compiler to textually
2062 // include headers in the specified module. We are not building the
2063 // specified module.
2064 if (getLangOpts().CompilingPCH &&
2065 isForModuleBuilding(M, getLangOpts().CurrentModule,
2066 getLangOpts().ModuleName))
2067 return;
2068
2069 assert(!CurLexerSubmodule && "should not have marked this as a module yet");
2070 CurLexerSubmodule = M;
2071
2072 // Let the macro handling code know that any future macros are within
2073 // the new submodule.
2074 EnterSubmodule(M, HashLoc, /*ForPragma*/false);
2075
2076 // Let the parser know that any future declarations are within the new
2077 // submodule.
2078 // FIXME: There's no point doing this if we're handling a #__include_macros
2079 // directive.
2080 EnterAnnotationToken(SourceRange(HashLoc, End), tok::annot_module_begin, M);
2081 }
2082 }
2083
2084 /// HandleIncludeNextDirective - Implements \#include_next.
2085 ///
HandleIncludeNextDirective(SourceLocation HashLoc,Token & IncludeNextTok)2086 void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc,
2087 Token &IncludeNextTok) {
2088 Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
2089
2090 // #include_next is like #include, except that we start searching after
2091 // the current found directory. If we can't do this, issue a
2092 // diagnostic.
2093 const DirectoryLookup *Lookup = CurDirLookup;
2094 const FileEntry *LookupFromFile = nullptr;
2095 if (isInPrimaryFile() && LangOpts.IsHeaderFile) {
2096 // If the main file is a header, then it's either for PCH/AST generation,
2097 // or libclang opened it. Either way, handle it as a normal include below
2098 // and do not complain about include_next.
2099 } else if (isInPrimaryFile()) {
2100 Lookup = nullptr;
2101 Diag(IncludeNextTok, diag::pp_include_next_in_primary);
2102 } else if (CurLexerSubmodule) {
2103 // Start looking up in the directory *after* the one in which the current
2104 // file would be found, if any.
2105 assert(CurPPLexer && "#include_next directive in macro?");
2106 LookupFromFile = CurPPLexer->getFileEntry();
2107 Lookup = nullptr;
2108 } else if (!Lookup) {
2109 Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
2110 } else {
2111 // Start looking up in the next directory.
2112 ++Lookup;
2113 }
2114
2115 return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup,
2116 LookupFromFile);
2117 }
2118
2119 /// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode
HandleMicrosoftImportDirective(Token & Tok)2120 void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) {
2121 // The Microsoft #import directive takes a type library and generates header
2122 // files from it, and includes those. This is beyond the scope of what clang
2123 // does, so we ignore it and error out. However, #import can optionally have
2124 // trailing attributes that span multiple lines. We're going to eat those
2125 // so we can continue processing from there.
2126 Diag(Tok, diag::err_pp_import_directive_ms );
2127
2128 // Read tokens until we get to the end of the directive. Note that the
2129 // directive can be split over multiple lines using the backslash character.
2130 DiscardUntilEndOfDirective();
2131 }
2132
2133 /// HandleImportDirective - Implements \#import.
2134 ///
HandleImportDirective(SourceLocation HashLoc,Token & ImportTok)2135 void Preprocessor::HandleImportDirective(SourceLocation HashLoc,
2136 Token &ImportTok) {
2137 if (!LangOpts.ObjC) { // #import is standard for ObjC.
2138 if (LangOpts.MSVCCompat)
2139 return HandleMicrosoftImportDirective(ImportTok);
2140 Diag(ImportTok, diag::ext_pp_import_directive);
2141 }
2142 return HandleIncludeDirective(HashLoc, ImportTok, nullptr, nullptr, true);
2143 }
2144
2145 /// HandleIncludeMacrosDirective - The -imacros command line option turns into a
2146 /// pseudo directive in the predefines buffer. This handles it by sucking all
2147 /// tokens through the preprocessor and discarding them (only keeping the side
2148 /// effects on the preprocessor).
HandleIncludeMacrosDirective(SourceLocation HashLoc,Token & IncludeMacrosTok)2149 void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc,
2150 Token &IncludeMacrosTok) {
2151 // This directive should only occur in the predefines buffer. If not, emit an
2152 // error and reject it.
2153 SourceLocation Loc = IncludeMacrosTok.getLocation();
2154 if (SourceMgr.getBufferName(Loc) != "<built-in>") {
2155 Diag(IncludeMacrosTok.getLocation(),
2156 diag::pp_include_macros_out_of_predefines);
2157 DiscardUntilEndOfDirective();
2158 return;
2159 }
2160
2161 // Treat this as a normal #include for checking purposes. If this is
2162 // successful, it will push a new lexer onto the include stack.
2163 HandleIncludeDirective(HashLoc, IncludeMacrosTok);
2164
2165 Token TmpTok;
2166 do {
2167 Lex(TmpTok);
2168 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!");
2169 } while (TmpTok.isNot(tok::hashhash));
2170 }
2171
2172 //===----------------------------------------------------------------------===//
2173 // Preprocessor Macro Directive Handling.
2174 //===----------------------------------------------------------------------===//
2175
2176 /// ReadMacroParameterList - The ( starting a parameter list of a macro
2177 /// definition has just been read. Lex the rest of the parameters and the
2178 /// closing ), updating MI with what we learn. Return true if an error occurs
2179 /// parsing the param list.
ReadMacroParameterList(MacroInfo * MI,Token & Tok)2180 bool Preprocessor::ReadMacroParameterList(MacroInfo *MI, Token &Tok) {
2181 SmallVector<IdentifierInfo*, 32> Parameters;
2182
2183 while (true) {
2184 LexUnexpandedToken(Tok);
2185 switch (Tok.getKind()) {
2186 case tok::r_paren:
2187 // Found the end of the parameter list.
2188 if (Parameters.empty()) // #define FOO()
2189 return false;
2190 // Otherwise we have #define FOO(A,)
2191 Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
2192 return true;
2193 case tok::ellipsis: // #define X(... -> C99 varargs
2194 if (!LangOpts.C99)
2195 Diag(Tok, LangOpts.CPlusPlus11 ?
2196 diag::warn_cxx98_compat_variadic_macro :
2197 diag::ext_variadic_macro);
2198
2199 // OpenCL v1.2 s6.9.e: variadic macros are not supported.
2200 if (LangOpts.OpenCL) {
2201 Diag(Tok, diag::err_pp_opencl_variadic_macros);
2202 return true;
2203 }
2204
2205 // Lex the token after the identifier.
2206 LexUnexpandedToken(Tok);
2207 if (Tok.isNot(tok::r_paren)) {
2208 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
2209 return true;
2210 }
2211 // Add the __VA_ARGS__ identifier as a parameter.
2212 Parameters.push_back(Ident__VA_ARGS__);
2213 MI->setIsC99Varargs();
2214 MI->setParameterList(Parameters, BP);
2215 return false;
2216 case tok::eod: // #define X(
2217 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
2218 return true;
2219 default:
2220 // Handle keywords and identifiers here to accept things like
2221 // #define Foo(for) for.
2222 IdentifierInfo *II = Tok.getIdentifierInfo();
2223 if (!II) {
2224 // #define X(1
2225 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
2226 return true;
2227 }
2228
2229 // If this is already used as a parameter, it is used multiple times (e.g.
2230 // #define X(A,A.
2231 if (std::find(Parameters.begin(), Parameters.end(), II) !=
2232 Parameters.end()) { // C99 6.10.3p6
2233 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II;
2234 return true;
2235 }
2236
2237 // Add the parameter to the macro info.
2238 Parameters.push_back(II);
2239
2240 // Lex the token after the identifier.
2241 LexUnexpandedToken(Tok);
2242
2243 switch (Tok.getKind()) {
2244 default: // #define X(A B
2245 Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
2246 return true;
2247 case tok::r_paren: // #define X(A)
2248 MI->setParameterList(Parameters, BP);
2249 return false;
2250 case tok::comma: // #define X(A,
2251 break;
2252 case tok::ellipsis: // #define X(A... -> GCC extension
2253 // Diagnose extension.
2254 Diag(Tok, diag::ext_named_variadic_macro);
2255
2256 // Lex the token after the identifier.
2257 LexUnexpandedToken(Tok);
2258 if (Tok.isNot(tok::r_paren)) {
2259 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
2260 return true;
2261 }
2262
2263 MI->setIsGNUVarargs();
2264 MI->setParameterList(Parameters, BP);
2265 return false;
2266 }
2267 }
2268 }
2269 }
2270
isConfigurationPattern(Token & MacroName,MacroInfo * MI,const LangOptions & LOptions)2271 static bool isConfigurationPattern(Token &MacroName, MacroInfo *MI,
2272 const LangOptions &LOptions) {
2273 if (MI->getNumTokens() == 1) {
2274 const Token &Value = MI->getReplacementToken(0);
2275
2276 // Macro that is identity, like '#define inline inline' is a valid pattern.
2277 if (MacroName.getKind() == Value.getKind())
2278 return true;
2279
2280 // Macro that maps a keyword to the same keyword decorated with leading/
2281 // trailing underscores is a valid pattern:
2282 // #define inline __inline
2283 // #define inline __inline__
2284 // #define inline _inline (in MS compatibility mode)
2285 StringRef MacroText = MacroName.getIdentifierInfo()->getName();
2286 if (IdentifierInfo *II = Value.getIdentifierInfo()) {
2287 if (!II->isKeyword(LOptions))
2288 return false;
2289 StringRef ValueText = II->getName();
2290 StringRef TrimmedValue = ValueText;
2291 if (!ValueText.startswith("__")) {
2292 if (ValueText.startswith("_"))
2293 TrimmedValue = TrimmedValue.drop_front(1);
2294 else
2295 return false;
2296 } else {
2297 TrimmedValue = TrimmedValue.drop_front(2);
2298 if (TrimmedValue.endswith("__"))
2299 TrimmedValue = TrimmedValue.drop_back(2);
2300 }
2301 return TrimmedValue.equals(MacroText);
2302 } else {
2303 return false;
2304 }
2305 }
2306
2307 // #define inline
2308 return MacroName.isOneOf(tok::kw_extern, tok::kw_inline, tok::kw_static,
2309 tok::kw_const) &&
2310 MI->getNumTokens() == 0;
2311 }
2312
2313 // ReadOptionalMacroParameterListAndBody - This consumes all (i.e. the
2314 // entire line) of the macro's tokens and adds them to MacroInfo, and while
2315 // doing so performs certain validity checks including (but not limited to):
2316 // - # (stringization) is followed by a macro parameter
2317 //
2318 // Returns a nullptr if an invalid sequence of tokens is encountered or returns
2319 // a pointer to a MacroInfo object.
2320
ReadOptionalMacroParameterListAndBody(const Token & MacroNameTok,const bool ImmediatelyAfterHeaderGuard)2321 MacroInfo *Preprocessor::ReadOptionalMacroParameterListAndBody(
2322 const Token &MacroNameTok, const bool ImmediatelyAfterHeaderGuard) {
2323
2324 Token LastTok = MacroNameTok;
2325 // Create the new macro.
2326 MacroInfo *const MI = AllocateMacroInfo(MacroNameTok.getLocation());
2327
2328 Token Tok;
2329 LexUnexpandedToken(Tok);
2330
2331 // Used to un-poison and then re-poison identifiers of the __VA_ARGS__ ilk
2332 // within their appropriate context.
2333 VariadicMacroScopeGuard VariadicMacroScopeGuard(*this);
2334
2335 // If this is a function-like macro definition, parse the argument list,
2336 // marking each of the identifiers as being used as macro arguments. Also,
2337 // check other constraints on the first token of the macro body.
2338 if (Tok.is(tok::eod)) {
2339 if (ImmediatelyAfterHeaderGuard) {
2340 // Save this macro information since it may part of a header guard.
2341 CurPPLexer->MIOpt.SetDefinedMacro(MacroNameTok.getIdentifierInfo(),
2342 MacroNameTok.getLocation());
2343 }
2344 // If there is no body to this macro, we have no special handling here.
2345 } else if (Tok.hasLeadingSpace()) {
2346 // This is a normal token with leading space. Clear the leading space
2347 // marker on the first token to get proper expansion.
2348 Tok.clearFlag(Token::LeadingSpace);
2349 } else if (Tok.is(tok::l_paren)) {
2350 // This is a function-like macro definition. Read the argument list.
2351 MI->setIsFunctionLike();
2352 if (ReadMacroParameterList(MI, LastTok)) {
2353 // Throw away the rest of the line.
2354 if (CurPPLexer->ParsingPreprocessorDirective)
2355 DiscardUntilEndOfDirective();
2356 return nullptr;
2357 }
2358
2359 // If this is a definition of an ISO C/C++ variadic function-like macro (not
2360 // using the GNU named varargs extension) inform our variadic scope guard
2361 // which un-poisons and re-poisons certain identifiers (e.g. __VA_ARGS__)
2362 // allowed only within the definition of a variadic macro.
2363
2364 if (MI->isC99Varargs()) {
2365 VariadicMacroScopeGuard.enterScope();
2366 }
2367
2368 // Read the first token after the arg list for down below.
2369 LexUnexpandedToken(Tok);
2370 } else if (LangOpts.C99 || LangOpts.CPlusPlus11) {
2371 // C99 requires whitespace between the macro definition and the body. Emit
2372 // a diagnostic for something like "#define X+".
2373 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
2374 } else {
2375 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the
2376 // first character of a replacement list is not a character required by
2377 // subclause 5.2.1, then there shall be white-space separation between the
2378 // identifier and the replacement list.". 5.2.1 lists this set:
2379 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which
2380 // is irrelevant here.
2381 bool isInvalid = false;
2382 if (Tok.is(tok::at)) // @ is not in the list above.
2383 isInvalid = true;
2384 else if (Tok.is(tok::unknown)) {
2385 // If we have an unknown token, it is something strange like "`". Since
2386 // all of valid characters would have lexed into a single character
2387 // token of some sort, we know this is not a valid case.
2388 isInvalid = true;
2389 }
2390 if (isInvalid)
2391 Diag(Tok, diag::ext_missing_whitespace_after_macro_name);
2392 else
2393 Diag(Tok, diag::warn_missing_whitespace_after_macro_name);
2394 }
2395
2396 if (!Tok.is(tok::eod))
2397 LastTok = Tok;
2398
2399 // Read the rest of the macro body.
2400 if (MI->isObjectLike()) {
2401 // Object-like macros are very simple, just read their body.
2402 while (Tok.isNot(tok::eod)) {
2403 LastTok = Tok;
2404 MI->AddTokenToBody(Tok);
2405 // Get the next token of the macro.
2406 LexUnexpandedToken(Tok);
2407 }
2408 } else {
2409 // Otherwise, read the body of a function-like macro. While we are at it,
2410 // check C99 6.10.3.2p1: ensure that # operators are followed by macro
2411 // parameters in function-like macro expansions.
2412
2413 VAOptDefinitionContext VAOCtx(*this);
2414
2415 while (Tok.isNot(tok::eod)) {
2416 LastTok = Tok;
2417
2418 if (!Tok.isOneOf(tok::hash, tok::hashat, tok::hashhash)) {
2419 MI->AddTokenToBody(Tok);
2420
2421 if (VAOCtx.isVAOptToken(Tok)) {
2422 // If we're already within a VAOPT, emit an error.
2423 if (VAOCtx.isInVAOpt()) {
2424 Diag(Tok, diag::err_pp_vaopt_nested_use);
2425 return nullptr;
2426 }
2427 // Ensure VAOPT is followed by a '(' .
2428 LexUnexpandedToken(Tok);
2429 if (Tok.isNot(tok::l_paren)) {
2430 Diag(Tok, diag::err_pp_missing_lparen_in_vaopt_use);
2431 return nullptr;
2432 }
2433 MI->AddTokenToBody(Tok);
2434 VAOCtx.sawVAOptFollowedByOpeningParens(Tok.getLocation());
2435 LexUnexpandedToken(Tok);
2436 if (Tok.is(tok::hashhash)) {
2437 Diag(Tok, diag::err_vaopt_paste_at_start);
2438 return nullptr;
2439 }
2440 continue;
2441 } else if (VAOCtx.isInVAOpt()) {
2442 if (Tok.is(tok::r_paren)) {
2443 if (VAOCtx.sawClosingParen()) {
2444 const unsigned NumTokens = MI->getNumTokens();
2445 assert(NumTokens >= 3 && "Must have seen at least __VA_OPT__( "
2446 "and a subsequent tok::r_paren");
2447 if (MI->getReplacementToken(NumTokens - 2).is(tok::hashhash)) {
2448 Diag(Tok, diag::err_vaopt_paste_at_end);
2449 return nullptr;
2450 }
2451 }
2452 } else if (Tok.is(tok::l_paren)) {
2453 VAOCtx.sawOpeningParen(Tok.getLocation());
2454 }
2455 }
2456 // Get the next token of the macro.
2457 LexUnexpandedToken(Tok);
2458 continue;
2459 }
2460
2461 // If we're in -traditional mode, then we should ignore stringification
2462 // and token pasting. Mark the tokens as unknown so as not to confuse
2463 // things.
2464 if (getLangOpts().TraditionalCPP) {
2465 Tok.setKind(tok::unknown);
2466 MI->AddTokenToBody(Tok);
2467
2468 // Get the next token of the macro.
2469 LexUnexpandedToken(Tok);
2470 continue;
2471 }
2472
2473 if (Tok.is(tok::hashhash)) {
2474 // If we see token pasting, check if it looks like the gcc comma
2475 // pasting extension. We'll use this information to suppress
2476 // diagnostics later on.
2477
2478 // Get the next token of the macro.
2479 LexUnexpandedToken(Tok);
2480
2481 if (Tok.is(tok::eod)) {
2482 MI->AddTokenToBody(LastTok);
2483 break;
2484 }
2485
2486 unsigned NumTokens = MI->getNumTokens();
2487 if (NumTokens && Tok.getIdentifierInfo() == Ident__VA_ARGS__ &&
2488 MI->getReplacementToken(NumTokens-1).is(tok::comma))
2489 MI->setHasCommaPasting();
2490
2491 // Things look ok, add the '##' token to the macro.
2492 MI->AddTokenToBody(LastTok);
2493 continue;
2494 }
2495
2496 // Our Token is a stringization operator.
2497 // Get the next token of the macro.
2498 LexUnexpandedToken(Tok);
2499
2500 // Check for a valid macro arg identifier or __VA_OPT__.
2501 if (!VAOCtx.isVAOptToken(Tok) &&
2502 (Tok.getIdentifierInfo() == nullptr ||
2503 MI->getParameterNum(Tok.getIdentifierInfo()) == -1)) {
2504
2505 // If this is assembler-with-cpp mode, we accept random gibberish after
2506 // the '#' because '#' is often a comment character. However, change
2507 // the kind of the token to tok::unknown so that the preprocessor isn't
2508 // confused.
2509 if (getLangOpts().AsmPreprocessor && Tok.isNot(tok::eod)) {
2510 LastTok.setKind(tok::unknown);
2511 MI->AddTokenToBody(LastTok);
2512 continue;
2513 } else {
2514 Diag(Tok, diag::err_pp_stringize_not_parameter)
2515 << LastTok.is(tok::hashat);
2516 return nullptr;
2517 }
2518 }
2519
2520 // Things look ok, add the '#' and param name tokens to the macro.
2521 MI->AddTokenToBody(LastTok);
2522
2523 // If the token following '#' is VAOPT, let the next iteration handle it
2524 // and check it for correctness, otherwise add the token and prime the
2525 // loop with the next one.
2526 if (!VAOCtx.isVAOptToken(Tok)) {
2527 MI->AddTokenToBody(Tok);
2528 LastTok = Tok;
2529
2530 // Get the next token of the macro.
2531 LexUnexpandedToken(Tok);
2532 }
2533 }
2534 if (VAOCtx.isInVAOpt()) {
2535 assert(Tok.is(tok::eod) && "Must be at End Of preprocessing Directive");
2536 Diag(Tok, diag::err_pp_expected_after)
2537 << LastTok.getKind() << tok::r_paren;
2538 Diag(VAOCtx.getUnmatchedOpeningParenLoc(), diag::note_matching) << tok::l_paren;
2539 return nullptr;
2540 }
2541 }
2542 MI->setDefinitionEndLoc(LastTok.getLocation());
2543 return MI;
2544 }
2545 /// HandleDefineDirective - Implements \#define. This consumes the entire macro
2546 /// line then lets the caller lex the next real token.
HandleDefineDirective(Token & DefineTok,const bool ImmediatelyAfterHeaderGuard)2547 void Preprocessor::HandleDefineDirective(
2548 Token &DefineTok, const bool ImmediatelyAfterHeaderGuard) {
2549 ++NumDefined;
2550
2551 Token MacroNameTok;
2552 bool MacroShadowsKeyword;
2553 ReadMacroName(MacroNameTok, MU_Define, &MacroShadowsKeyword);
2554
2555 // Error reading macro name? If so, diagnostic already issued.
2556 if (MacroNameTok.is(tok::eod))
2557 return;
2558
2559 // If we are supposed to keep comments in #defines, reenable comment saving
2560 // mode.
2561 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
2562
2563 MacroInfo *const MI = ReadOptionalMacroParameterListAndBody(
2564 MacroNameTok, ImmediatelyAfterHeaderGuard);
2565
2566 if (!MI) return;
2567
2568 if (MacroShadowsKeyword &&
2569 !isConfigurationPattern(MacroNameTok, MI, getLangOpts())) {
2570 Diag(MacroNameTok, diag::warn_pp_macro_hides_keyword);
2571 }
2572 // Check that there is no paste (##) operator at the beginning or end of the
2573 // replacement list.
2574 unsigned NumTokens = MI->getNumTokens();
2575 if (NumTokens != 0) {
2576 if (MI->getReplacementToken(0).is(tok::hashhash)) {
2577 Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
2578 return;
2579 }
2580 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
2581 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
2582 return;
2583 }
2584 }
2585
2586 // When skipping just warn about macros that do not match.
2587 if (SkippingUntilPCHThroughHeader) {
2588 const MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo());
2589 if (!OtherMI || !MI->isIdenticalTo(*OtherMI, *this,
2590 /*Syntactic=*/LangOpts.MicrosoftExt))
2591 Diag(MI->getDefinitionLoc(), diag::warn_pp_macro_def_mismatch_with_pch)
2592 << MacroNameTok.getIdentifierInfo();
2593 return;
2594 }
2595
2596 // Finally, if this identifier already had a macro defined for it, verify that
2597 // the macro bodies are identical, and issue diagnostics if they are not.
2598 if (const MacroInfo *OtherMI=getMacroInfo(MacroNameTok.getIdentifierInfo())) {
2599 // In Objective-C, ignore attempts to directly redefine the builtin
2600 // definitions of the ownership qualifiers. It's still possible to
2601 // #undef them.
2602 auto isObjCProtectedMacro = [](const IdentifierInfo *II) -> bool {
2603 return II->isStr("__strong") ||
2604 II->isStr("__weak") ||
2605 II->isStr("__unsafe_unretained") ||
2606 II->isStr("__autoreleasing");
2607 };
2608 if (getLangOpts().ObjC &&
2609 SourceMgr.getFileID(OtherMI->getDefinitionLoc())
2610 == getPredefinesFileID() &&
2611 isObjCProtectedMacro(MacroNameTok.getIdentifierInfo())) {
2612 // Warn if it changes the tokens.
2613 if ((!getDiagnostics().getSuppressSystemWarnings() ||
2614 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) &&
2615 !MI->isIdenticalTo(*OtherMI, *this,
2616 /*Syntactic=*/LangOpts.MicrosoftExt)) {
2617 Diag(MI->getDefinitionLoc(), diag::warn_pp_objc_macro_redef_ignored);
2618 }
2619 assert(!OtherMI->isWarnIfUnused());
2620 return;
2621 }
2622
2623 // It is very common for system headers to have tons of macro redefinitions
2624 // and for warnings to be disabled in system headers. If this is the case,
2625 // then don't bother calling MacroInfo::isIdenticalTo.
2626 if (!getDiagnostics().getSuppressSystemWarnings() ||
2627 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) {
2628 if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused())
2629 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
2630
2631 // Warn if defining "__LINE__" and other builtins, per C99 6.10.8/4 and
2632 // C++ [cpp.predefined]p4, but allow it as an extension.
2633 if (OtherMI->isBuiltinMacro())
2634 Diag(MacroNameTok, diag::ext_pp_redef_builtin_macro);
2635 // Macros must be identical. This means all tokens and whitespace
2636 // separation must be the same. C99 6.10.3p2.
2637 else if (!OtherMI->isAllowRedefinitionsWithoutWarning() &&
2638 !MI->isIdenticalTo(*OtherMI, *this, /*Syntactic=*/LangOpts.MicrosoftExt)) {
2639 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
2640 << MacroNameTok.getIdentifierInfo();
2641 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition);
2642 }
2643 }
2644 if (OtherMI->isWarnIfUnused())
2645 WarnUnusedMacroLocs.erase(OtherMI->getDefinitionLoc());
2646 }
2647
2648 DefMacroDirective *MD =
2649 appendDefMacroDirective(MacroNameTok.getIdentifierInfo(), MI);
2650
2651 assert(!MI->isUsed());
2652 // If we need warning for not using the macro, add its location in the
2653 // warn-because-unused-macro set. If it gets used it will be removed from set.
2654 if (getSourceManager().isInMainFile(MI->getDefinitionLoc()) &&
2655 !Diags->isIgnored(diag::pp_macro_not_used, MI->getDefinitionLoc())) {
2656 MI->setIsWarnIfUnused(true);
2657 WarnUnusedMacroLocs.insert(MI->getDefinitionLoc());
2658 }
2659
2660 // If the callbacks want to know, tell them about the macro definition.
2661 if (Callbacks)
2662 Callbacks->MacroDefined(MacroNameTok, MD);
2663 }
2664
2665 /// HandleUndefDirective - Implements \#undef.
2666 ///
HandleUndefDirective()2667 void Preprocessor::HandleUndefDirective() {
2668 ++NumUndefined;
2669
2670 Token MacroNameTok;
2671 ReadMacroName(MacroNameTok, MU_Undef);
2672
2673 // Error reading macro name? If so, diagnostic already issued.
2674 if (MacroNameTok.is(tok::eod))
2675 return;
2676
2677 // Check to see if this is the last token on the #undef line.
2678 CheckEndOfDirective("undef");
2679
2680 // Okay, we have a valid identifier to undef.
2681 auto *II = MacroNameTok.getIdentifierInfo();
2682 auto MD = getMacroDefinition(II);
2683 UndefMacroDirective *Undef = nullptr;
2684
2685 // If the macro is not defined, this is a noop undef.
2686 if (const MacroInfo *MI = MD.getMacroInfo()) {
2687 if (!MI->isUsed() && MI->isWarnIfUnused())
2688 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
2689
2690 if (MI->isWarnIfUnused())
2691 WarnUnusedMacroLocs.erase(MI->getDefinitionLoc());
2692
2693 Undef = AllocateUndefMacroDirective(MacroNameTok.getLocation());
2694 }
2695
2696 // If the callbacks want to know, tell them about the macro #undef.
2697 // Note: no matter if the macro was defined or not.
2698 if (Callbacks)
2699 Callbacks->MacroUndefined(MacroNameTok, MD, Undef);
2700
2701 if (Undef)
2702 appendMacroDirective(II, Undef);
2703 }
2704
2705 //===----------------------------------------------------------------------===//
2706 // Preprocessor Conditional Directive Handling.
2707 //===----------------------------------------------------------------------===//
2708
2709 /// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive. isIfndef
2710 /// is true when this is a \#ifndef directive. ReadAnyTokensBeforeDirective is
2711 /// true if any tokens have been returned or pp-directives activated before this
2712 /// \#ifndef has been lexed.
2713 ///
HandleIfdefDirective(Token & Result,const Token & HashToken,bool isIfndef,bool ReadAnyTokensBeforeDirective)2714 void Preprocessor::HandleIfdefDirective(Token &Result,
2715 const Token &HashToken,
2716 bool isIfndef,
2717 bool ReadAnyTokensBeforeDirective) {
2718 ++NumIf;
2719 Token DirectiveTok = Result;
2720
2721 Token MacroNameTok;
2722 ReadMacroName(MacroNameTok);
2723
2724 // Error reading macro name? If so, diagnostic already issued.
2725 if (MacroNameTok.is(tok::eod)) {
2726 // Skip code until we get to #endif. This helps with recovery by not
2727 // emitting an error when the #endif is reached.
2728 SkipExcludedConditionalBlock(HashToken.getLocation(),
2729 DirectiveTok.getLocation(),
2730 /*Foundnonskip*/ false, /*FoundElse*/ false);
2731 return;
2732 }
2733
2734 // Check to see if this is the last token on the #if[n]def line.
2735 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef");
2736
2737 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
2738 auto MD = getMacroDefinition(MII);
2739 MacroInfo *MI = MD.getMacroInfo();
2740
2741 if (CurPPLexer->getConditionalStackDepth() == 0) {
2742 // If the start of a top-level #ifdef and if the macro is not defined,
2743 // inform MIOpt that this might be the start of a proper include guard.
2744 // Otherwise it is some other form of unknown conditional which we can't
2745 // handle.
2746 if (!ReadAnyTokensBeforeDirective && !MI) {
2747 assert(isIfndef && "#ifdef shouldn't reach here");
2748 CurPPLexer->MIOpt.EnterTopLevelIfndef(MII, MacroNameTok.getLocation());
2749 } else
2750 CurPPLexer->MIOpt.EnterTopLevelConditional();
2751 }
2752
2753 // If there is a macro, process it.
2754 if (MI) // Mark it used.
2755 markMacroAsUsed(MI);
2756
2757 if (Callbacks) {
2758 if (isIfndef)
2759 Callbacks->Ifndef(DirectiveTok.getLocation(), MacroNameTok, MD);
2760 else
2761 Callbacks->Ifdef(DirectiveTok.getLocation(), MacroNameTok, MD);
2762 }
2763
2764 // Should we include the stuff contained by this directive?
2765 if (PPOpts->SingleFileParseMode && !MI) {
2766 // In 'single-file-parse mode' undefined identifiers trigger parsing of all
2767 // the directive blocks.
2768 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(),
2769 /*wasskip*/false, /*foundnonskip*/false,
2770 /*foundelse*/false);
2771 } else if (!MI == isIfndef) {
2772 // Yes, remember that we are inside a conditional, then lex the next token.
2773 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(),
2774 /*wasskip*/false, /*foundnonskip*/true,
2775 /*foundelse*/false);
2776 } else {
2777 // No, skip the contents of this block.
2778 SkipExcludedConditionalBlock(HashToken.getLocation(),
2779 DirectiveTok.getLocation(),
2780 /*Foundnonskip*/ false,
2781 /*FoundElse*/ false);
2782 }
2783 }
2784
2785 /// HandleIfDirective - Implements the \#if directive.
2786 ///
HandleIfDirective(Token & IfToken,const Token & HashToken,bool ReadAnyTokensBeforeDirective)2787 void Preprocessor::HandleIfDirective(Token &IfToken,
2788 const Token &HashToken,
2789 bool ReadAnyTokensBeforeDirective) {
2790 ++NumIf;
2791
2792 // Parse and evaluate the conditional expression.
2793 IdentifierInfo *IfNDefMacro = nullptr;
2794 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2795 const DirectiveEvalResult DER = EvaluateDirectiveExpression(IfNDefMacro);
2796 const bool ConditionalTrue = DER.Conditional;
2797 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2798
2799 // If this condition is equivalent to #ifndef X, and if this is the first
2800 // directive seen, handle it for the multiple-include optimization.
2801 if (CurPPLexer->getConditionalStackDepth() == 0) {
2802 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue)
2803 // FIXME: Pass in the location of the macro name, not the 'if' token.
2804 CurPPLexer->MIOpt.EnterTopLevelIfndef(IfNDefMacro, IfToken.getLocation());
2805 else
2806 CurPPLexer->MIOpt.EnterTopLevelConditional();
2807 }
2808
2809 if (Callbacks)
2810 Callbacks->If(IfToken.getLocation(),
2811 SourceRange(ConditionalBegin, ConditionalEnd),
2812 (ConditionalTrue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False));
2813
2814 // Should we include the stuff contained by this directive?
2815 if (PPOpts->SingleFileParseMode && DER.IncludedUndefinedIds) {
2816 // In 'single-file-parse mode' undefined identifiers trigger parsing of all
2817 // the directive blocks.
2818 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
2819 /*foundnonskip*/false, /*foundelse*/false);
2820 } else if (ConditionalTrue) {
2821 // Yes, remember that we are inside a conditional, then lex the next token.
2822 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
2823 /*foundnonskip*/true, /*foundelse*/false);
2824 } else {
2825 // No, skip the contents of this block.
2826 SkipExcludedConditionalBlock(HashToken.getLocation(), IfToken.getLocation(),
2827 /*Foundnonskip*/ false,
2828 /*FoundElse*/ false);
2829 }
2830 }
2831
2832 /// HandleEndifDirective - Implements the \#endif directive.
2833 ///
HandleEndifDirective(Token & EndifToken)2834 void Preprocessor::HandleEndifDirective(Token &EndifToken) {
2835 ++NumEndif;
2836
2837 // Check that this is the whole directive.
2838 CheckEndOfDirective("endif");
2839
2840 PPConditionalInfo CondInfo;
2841 if (CurPPLexer->popConditionalLevel(CondInfo)) {
2842 // No conditionals on the stack: this is an #endif without an #if.
2843 Diag(EndifToken, diag::err_pp_endif_without_if);
2844 return;
2845 }
2846
2847 // If this the end of a top-level #endif, inform MIOpt.
2848 if (CurPPLexer->getConditionalStackDepth() == 0)
2849 CurPPLexer->MIOpt.ExitTopLevelConditional();
2850
2851 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
2852 "This code should only be reachable in the non-skipping case!");
2853
2854 if (Callbacks)
2855 Callbacks->Endif(EndifToken.getLocation(), CondInfo.IfLoc);
2856 }
2857
2858 /// HandleElseDirective - Implements the \#else directive.
2859 ///
HandleElseDirective(Token & Result,const Token & HashToken)2860 void Preprocessor::HandleElseDirective(Token &Result, const Token &HashToken) {
2861 ++NumElse;
2862
2863 // #else directive in a non-skipping conditional... start skipping.
2864 CheckEndOfDirective("else");
2865
2866 PPConditionalInfo CI;
2867 if (CurPPLexer->popConditionalLevel(CI)) {
2868 Diag(Result, diag::pp_err_else_without_if);
2869 return;
2870 }
2871
2872 // If this is a top-level #else, inform the MIOpt.
2873 if (CurPPLexer->getConditionalStackDepth() == 0)
2874 CurPPLexer->MIOpt.EnterTopLevelConditional();
2875
2876 // If this is a #else with a #else before it, report the error.
2877 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
2878
2879 if (Callbacks)
2880 Callbacks->Else(Result.getLocation(), CI.IfLoc);
2881
2882 if (PPOpts->SingleFileParseMode && !CI.FoundNonSkip) {
2883 // In 'single-file-parse mode' undefined identifiers trigger parsing of all
2884 // the directive blocks.
2885 CurPPLexer->pushConditionalLevel(CI.IfLoc, /*wasskip*/false,
2886 /*foundnonskip*/false, /*foundelse*/true);
2887 return;
2888 }
2889
2890 // Finally, skip the rest of the contents of this block.
2891 SkipExcludedConditionalBlock(HashToken.getLocation(), CI.IfLoc,
2892 /*Foundnonskip*/ true,
2893 /*FoundElse*/ true, Result.getLocation());
2894 }
2895
2896 /// HandleElifDirective - Implements the \#elif directive.
2897 ///
HandleElifDirective(Token & ElifToken,const Token & HashToken)2898 void Preprocessor::HandleElifDirective(Token &ElifToken,
2899 const Token &HashToken) {
2900 ++NumElse;
2901
2902 // #elif directive in a non-skipping conditional... start skipping.
2903 // We don't care what the condition is, because we will always skip it (since
2904 // the block immediately before it was included).
2905 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2906 DiscardUntilEndOfDirective();
2907 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2908
2909 PPConditionalInfo CI;
2910 if (CurPPLexer->popConditionalLevel(CI)) {
2911 Diag(ElifToken, diag::pp_err_elif_without_if);
2912 return;
2913 }
2914
2915 // If this is a top-level #elif, inform the MIOpt.
2916 if (CurPPLexer->getConditionalStackDepth() == 0)
2917 CurPPLexer->MIOpt.EnterTopLevelConditional();
2918
2919 // If this is a #elif with a #else before it, report the error.
2920 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
2921
2922 if (Callbacks)
2923 Callbacks->Elif(ElifToken.getLocation(),
2924 SourceRange(ConditionalBegin, ConditionalEnd),
2925 PPCallbacks::CVK_NotEvaluated, CI.IfLoc);
2926
2927 if (PPOpts->SingleFileParseMode && !CI.FoundNonSkip) {
2928 // In 'single-file-parse mode' undefined identifiers trigger parsing of all
2929 // the directive blocks.
2930 CurPPLexer->pushConditionalLevel(ElifToken.getLocation(), /*wasskip*/false,
2931 /*foundnonskip*/false, /*foundelse*/false);
2932 return;
2933 }
2934
2935 // Finally, skip the rest of the contents of this block.
2936 SkipExcludedConditionalBlock(
2937 HashToken.getLocation(), CI.IfLoc, /*Foundnonskip*/ true,
2938 /*FoundElse*/ CI.FoundElse, ElifToken.getLocation());
2939 }
2940