clang  7.0.0svn
PPLexerChange.cpp
Go to the documentation of this file.
1 //===--- PPLexerChange.cpp - Handle changing lexers in the preprocessor ---===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements pieces of the Preprocessor interface that manage the
11 // current lexer stack.
12 //
13 //===----------------------------------------------------------------------===//
14 
15 #include "clang/Lex/Preprocessor.h"
18 #include "clang/Lex/HeaderSearch.h"
20 #include "clang/Lex/MacroInfo.h"
21 #include "clang/Lex/PTHManager.h"
22 #include "llvm/ADT/StringSwitch.h"
23 #include "llvm/Support/FileSystem.h"
24 #include "llvm/Support/MemoryBuffer.h"
25 #include "llvm/Support/Path.h"
26 using namespace clang;
27 
29 
30 //===----------------------------------------------------------------------===//
31 // Miscellaneous Methods.
32 //===----------------------------------------------------------------------===//
33 
34 /// isInPrimaryFile - Return true if we're in the top-level file, not in a
35 /// \#include. This looks through macro expansions and active _Pragma lexers.
37  if (IsFileLexer())
38  return IncludeMacroStack.empty();
39 
40  // If there are any stacked lexers, we're in a #include.
41  assert(IsFileLexer(IncludeMacroStack[0]) &&
42  "Top level include stack isn't our primary lexer?");
43  return std::none_of(
44  IncludeMacroStack.begin() + 1, IncludeMacroStack.end(),
45  [&](const IncludeStackInfo &ISI) -> bool { return IsFileLexer(ISI); });
46 }
47 
48 /// getCurrentLexer - Return the current file lexer being lexed from. Note
49 /// that this ignores any potentially active macro expansions and _Pragma
50 /// expansions going on at the time.
52  if (IsFileLexer())
53  return CurPPLexer;
54 
55  // Look for a stacked lexer.
56  for (const IncludeStackInfo &ISI : llvm::reverse(IncludeMacroStack)) {
57  if (IsFileLexer(ISI))
58  return ISI.ThePPLexer;
59  }
60  return nullptr;
61 }
62 
63 
64 //===----------------------------------------------------------------------===//
65 // Methods for Entering and Callbacks for leaving various contexts
66 //===----------------------------------------------------------------------===//
67 
68 /// EnterSourceFile - Add a source file to the top of the include stack and
69 /// start lexing tokens from it instead of the current buffer.
71  SourceLocation Loc) {
72  assert(!CurTokenLexer && "Cannot #include a file inside a macro!");
73  ++NumEnteredSourceFiles;
74 
75  if (MaxIncludeStackDepth < IncludeMacroStack.size())
76  MaxIncludeStackDepth = IncludeMacroStack.size();
77 
78  if (PTH) {
79  if (PTHLexer *PL = PTH->CreateLexer(FID)) {
80  EnterSourceFileWithPTH(PL, CurDir);
81  return false;
82  }
83  }
84 
85  // Get the MemoryBuffer for this FID, if it fails, we fail.
86  bool Invalid = false;
87  const llvm::MemoryBuffer *InputFile =
88  getSourceManager().getBuffer(FID, Loc, &Invalid);
89  if (Invalid) {
90  SourceLocation FileStart = SourceMgr.getLocForStartOfFile(FID);
91  Diag(Loc, diag::err_pp_error_opening_file)
92  << std::string(SourceMgr.getBufferName(FileStart)) << "";
93  return true;
94  }
95 
96  if (isCodeCompletionEnabled() &&
97  SourceMgr.getFileEntryForID(FID) == CodeCompletionFile) {
98  CodeCompletionFileLoc = SourceMgr.getLocForStartOfFile(FID);
99  CodeCompletionLoc =
100  CodeCompletionFileLoc.getLocWithOffset(CodeCompletionOffset);
101  }
102 
103  EnterSourceFileWithLexer(new Lexer(FID, InputFile, *this), CurDir);
104  return false;
105 }
106 
107 /// EnterSourceFileWithLexer - Add a source file to the top of the include stack
108 /// and start lexing tokens from it instead of the current buffer.
109 void Preprocessor::EnterSourceFileWithLexer(Lexer *TheLexer,
110  const DirectoryLookup *CurDir) {
111 
112  // Add the current lexer to the include stack.
113  if (CurPPLexer || CurTokenLexer)
114  PushIncludeMacroStack();
115 
116  CurLexer.reset(TheLexer);
117  CurPPLexer = TheLexer;
118  CurDirLookup = CurDir;
119  CurLexerSubmodule = nullptr;
120  if (CurLexerKind != CLK_LexAfterModuleImport)
121  CurLexerKind = CLK_Lexer;
122 
123  // Notify the client, if desired, that we are in a new source file.
124  if (Callbacks && !CurLexer->Is_PragmaLexer) {
125  SrcMgr::CharacteristicKind FileType =
126  SourceMgr.getFileCharacteristic(CurLexer->getFileLoc());
127 
128  Callbacks->FileChanged(CurLexer->getFileLoc(),
129  PPCallbacks::EnterFile, FileType);
130  }
131 }
132 
133 /// EnterSourceFileWithPTH - Add a source file to the top of the include stack
134 /// and start getting tokens from it using the PTH cache.
135 void Preprocessor::EnterSourceFileWithPTH(PTHLexer *PL,
136  const DirectoryLookup *CurDir) {
137 
138  if (CurPPLexer || CurTokenLexer)
139  PushIncludeMacroStack();
140 
141  CurDirLookup = CurDir;
142  CurPTHLexer.reset(PL);
143  CurPPLexer = CurPTHLexer.get();
144  CurLexerSubmodule = nullptr;
145  if (CurLexerKind != CLK_LexAfterModuleImport)
146  CurLexerKind = CLK_PTHLexer;
147 
148  // Notify the client, if desired, that we are in a new source file.
149  if (Callbacks) {
150  FileID FID = CurPPLexer->getFileID();
151  SourceLocation EnterLoc = SourceMgr.getLocForStartOfFile(FID);
152  SrcMgr::CharacteristicKind FileType =
153  SourceMgr.getFileCharacteristic(EnterLoc);
154  Callbacks->FileChanged(EnterLoc, PPCallbacks::EnterFile, FileType);
155  }
156 }
157 
158 /// EnterMacro - Add a Macro to the top of the include stack and start lexing
159 /// tokens from it instead of the current buffer.
161  MacroInfo *Macro, MacroArgs *Args) {
162  std::unique_ptr<TokenLexer> TokLexer;
163  if (NumCachedTokenLexers == 0) {
164  TokLexer = llvm::make_unique<TokenLexer>(Tok, ILEnd, Macro, Args, *this);
165  } else {
166  TokLexer = std::move(TokenLexerCache[--NumCachedTokenLexers]);
167  TokLexer->Init(Tok, ILEnd, Macro, Args);
168  }
169 
170  PushIncludeMacroStack();
171  CurDirLookup = nullptr;
172  CurTokenLexer = std::move(TokLexer);
173  if (CurLexerKind != CLK_LexAfterModuleImport)
174  CurLexerKind = CLK_TokenLexer;
175 }
176 
177 /// EnterTokenStream - Add a "macro" context to the top of the include stack,
178 /// which will cause the lexer to start returning the specified tokens.
179 ///
180 /// If DisableMacroExpansion is true, tokens lexed from the token stream will
181 /// not be subject to further macro expansion. Otherwise, these tokens will
182 /// be re-macro-expanded when/if expansion is enabled.
183 ///
184 /// If OwnsTokens is false, this method assumes that the specified stream of
185 /// tokens has a permanent owner somewhere, so they do not need to be copied.
186 /// If it is true, it assumes the array of tokens is allocated with new[] and
187 /// must be freed.
188 ///
189 void Preprocessor::EnterTokenStream(const Token *Toks, unsigned NumToks,
190  bool DisableMacroExpansion,
191  bool OwnsTokens) {
192  if (CurLexerKind == CLK_CachingLexer) {
193  if (CachedLexPos < CachedTokens.size()) {
194  // We're entering tokens into the middle of our cached token stream. We
195  // can't represent that, so just insert the tokens into the buffer.
196  CachedTokens.insert(CachedTokens.begin() + CachedLexPos,
197  Toks, Toks + NumToks);
198  if (OwnsTokens)
199  delete [] Toks;
200  return;
201  }
202 
203  // New tokens are at the end of the cached token sequnece; insert the
204  // token stream underneath the caching lexer.
205  ExitCachingLexMode();
206  EnterTokenStream(Toks, NumToks, DisableMacroExpansion, OwnsTokens);
207  EnterCachingLexMode();
208  return;
209  }
210 
211  // Create a macro expander to expand from the specified token stream.
212  std::unique_ptr<TokenLexer> TokLexer;
213  if (NumCachedTokenLexers == 0) {
214  TokLexer = llvm::make_unique<TokenLexer>(
215  Toks, NumToks, DisableMacroExpansion, OwnsTokens, *this);
216  } else {
217  TokLexer = std::move(TokenLexerCache[--NumCachedTokenLexers]);
218  TokLexer->Init(Toks, NumToks, DisableMacroExpansion, OwnsTokens);
219  }
220 
221  // Save our current state.
222  PushIncludeMacroStack();
223  CurDirLookup = nullptr;
224  CurTokenLexer = std::move(TokLexer);
225  if (CurLexerKind != CLK_LexAfterModuleImport)
226  CurLexerKind = CLK_TokenLexer;
227 }
228 
229 /// Compute the relative path that names the given file relative to
230 /// the given directory.
231 static void computeRelativePath(FileManager &FM, const DirectoryEntry *Dir,
232  const FileEntry *File,
234  Result.clear();
235 
236  StringRef FilePath = File->getDir()->getName();
237  StringRef Path = FilePath;
238  while (!Path.empty()) {
239  if (const DirectoryEntry *CurDir = FM.getDirectory(Path)) {
240  if (CurDir == Dir) {
241  Result = FilePath.substr(Path.size());
242  llvm::sys::path::append(Result,
243  llvm::sys::path::filename(File->getName()));
244  return;
245  }
246  }
247 
248  Path = llvm::sys::path::parent_path(Path);
249  }
250 
251  Result = File->getName();
252 }
253 
254 void Preprocessor::PropagateLineStartLeadingSpaceInfo(Token &Result) {
255  if (CurTokenLexer) {
256  CurTokenLexer->PropagateLineStartLeadingSpaceInfo(Result);
257  return;
258  }
259  if (CurLexer) {
260  CurLexer->PropagateLineStartLeadingSpaceInfo(Result);
261  return;
262  }
263  // FIXME: Handle other kinds of lexers? It generally shouldn't matter,
264  // but it might if they're empty?
265 }
266 
267 /// Determine the location to use as the end of the buffer for a lexer.
268 ///
269 /// If the file ends with a newline, form the EOF token on the newline itself,
270 /// rather than "on the line following it", which doesn't exist. This makes
271 /// diagnostics relating to the end of file include the last file that the user
272 /// actually typed, which is goodness.
273 const char *Preprocessor::getCurLexerEndPos() {
274  const char *EndPos = CurLexer->BufferEnd;
275  if (EndPos != CurLexer->BufferStart &&
276  (EndPos[-1] == '\n' || EndPos[-1] == '\r')) {
277  --EndPos;
278 
279  // Handle \n\r and \r\n:
280  if (EndPos != CurLexer->BufferStart &&
281  (EndPos[-1] == '\n' || EndPos[-1] == '\r') &&
282  EndPos[-1] != EndPos[0])
283  --EndPos;
284  }
285 
286  return EndPos;
287 }
288 
290  const Module &Mod, SmallVectorImpl<const Module *> &SubMods) {
291  if (Mod.getUmbrellaHeader())
292  SubMods.push_back(&Mod);
293  for (auto *M : Mod.submodules())
295 }
296 
297 void Preprocessor::diagnoseMissingHeaderInUmbrellaDir(const Module &Mod) {
298  assert(Mod.getUmbrellaHeader() && "Module must use umbrella header");
299  SourceLocation StartLoc =
300  SourceMgr.getLocForStartOfFile(SourceMgr.getMainFileID());
301  if (getDiagnostics().isIgnored(diag::warn_uncovered_module_header, StartLoc))
302  return;
303 
304  ModuleMap &ModMap = getHeaderSearchInfo().getModuleMap();
305  const DirectoryEntry *Dir = Mod.getUmbrellaDir().Entry;
306  vfs::FileSystem &FS = *FileMgr.getVirtualFileSystem();
307  std::error_code EC;
308  for (vfs::recursive_directory_iterator Entry(FS, Dir->getName(), EC), End;
309  Entry != End && !EC; Entry.increment(EC)) {
310  using llvm::StringSwitch;
311 
312  // Check whether this entry has an extension typically associated with
313  // headers.
314  if (!StringSwitch<bool>(llvm::sys::path::extension(Entry->getName()))
315  .Cases(".h", ".H", ".hh", ".hpp", true)
316  .Default(false))
317  continue;
318 
319  if (const FileEntry *Header = getFileManager().getFile(Entry->getName()))
320  if (!getSourceManager().hasFileInfo(Header)) {
321  if (!ModMap.isHeaderInUnavailableModule(Header)) {
322  // Find the relative path that would access this header.
323  SmallString<128> RelativePath;
324  computeRelativePath(FileMgr, Dir, Header, RelativePath);
325  Diag(StartLoc, diag::warn_uncovered_module_header)
326  << Mod.getFullModuleName() << RelativePath;
327  }
328  }
329  }
330 }
331 
332 /// HandleEndOfFile - This callback is invoked when the lexer hits the end of
333 /// the current file. This either returns the EOF token or pops a level off
334 /// the include stack and keeps going.
335 bool Preprocessor::HandleEndOfFile(Token &Result, bool isEndOfMacro) {
336  assert(!CurTokenLexer &&
337  "Ending a file when currently in a macro!");
338 
339  // If we have an unclosed module region from a pragma at the end of a
340  // module, complain and close it now.
341  // FIXME: This is not correct if we are building a module from PTH.
342  const bool LeavingSubmodule = CurLexer && CurLexerSubmodule;
343  if ((LeavingSubmodule || IncludeMacroStack.empty()) &&
344  !BuildingSubmoduleStack.empty() &&
345  BuildingSubmoduleStack.back().IsPragma) {
346  Diag(BuildingSubmoduleStack.back().ImportLoc,
347  diag::err_pp_module_begin_without_module_end);
348  Module *M = LeaveSubmodule(/*ForPragma*/true);
349 
350  Result.startToken();
351  const char *EndPos = getCurLexerEndPos();
352  CurLexer->BufferPtr = EndPos;
353  CurLexer->FormTokenWithChars(Result, EndPos, tok::annot_module_end);
354  Result.setAnnotationEndLoc(Result.getLocation());
355  Result.setAnnotationValue(M);
356  return true;
357  }
358 
359  // See if this file had a controlling macro.
360  if (CurPPLexer) { // Not ending a macro, ignore it.
361  if (const IdentifierInfo *ControllingMacro =
362  CurPPLexer->MIOpt.GetControllingMacroAtEndOfFile()) {
363  // Okay, this has a controlling macro, remember in HeaderFileInfo.
364  if (const FileEntry *FE = CurPPLexer->getFileEntry()) {
365  HeaderInfo.SetFileControllingMacro(FE, ControllingMacro);
366  if (MacroInfo *MI =
367  getMacroInfo(const_cast<IdentifierInfo*>(ControllingMacro)))
368  MI->setUsedForHeaderGuard(true);
369  if (const IdentifierInfo *DefinedMacro =
370  CurPPLexer->MIOpt.GetDefinedMacro()) {
371  if (!isMacroDefined(ControllingMacro) &&
372  DefinedMacro != ControllingMacro &&
373  HeaderInfo.FirstTimeLexingFile(FE)) {
374 
375  // If the edit distance between the two macros is more than 50%,
376  // DefinedMacro may not be header guard, or can be header guard of
377  // another header file. Therefore, it maybe defining something
378  // completely different. This can be observed in the wild when
379  // handling feature macros or header guards in different files.
380 
381  const StringRef ControllingMacroName = ControllingMacro->getName();
382  const StringRef DefinedMacroName = DefinedMacro->getName();
383  const size_t MaxHalfLength = std::max(ControllingMacroName.size(),
384  DefinedMacroName.size()) / 2;
385  const unsigned ED = ControllingMacroName.edit_distance(
386  DefinedMacroName, true, MaxHalfLength);
387  if (ED <= MaxHalfLength) {
388  // Emit a warning for a bad header guard.
389  Diag(CurPPLexer->MIOpt.GetMacroLocation(),
390  diag::warn_header_guard)
391  << CurPPLexer->MIOpt.GetMacroLocation() << ControllingMacro;
392  Diag(CurPPLexer->MIOpt.GetDefinedLocation(),
393  diag::note_header_guard)
394  << CurPPLexer->MIOpt.GetDefinedLocation() << DefinedMacro
395  << ControllingMacro
397  CurPPLexer->MIOpt.GetDefinedLocation(),
398  ControllingMacro->getName());
399  }
400  }
401  }
402  }
403  }
404  }
405 
406  // Complain about reaching a true EOF within arc_cf_code_audited.
407  // We don't want to complain about reaching the end of a macro
408  // instantiation or a _Pragma.
409  if (PragmaARCCFCodeAuditedLoc.isValid() &&
410  !isEndOfMacro && !(CurLexer && CurLexer->Is_PragmaLexer)) {
411  Diag(PragmaARCCFCodeAuditedLoc, diag::err_pp_eof_in_arc_cf_code_audited);
412 
413  // Recover by leaving immediately.
414  PragmaARCCFCodeAuditedLoc = SourceLocation();
415  }
416 
417  // Complain about reaching a true EOF within assume_nonnull.
418  // We don't want to complain about reaching the end of a macro
419  // instantiation or a _Pragma.
420  if (PragmaAssumeNonNullLoc.isValid() &&
421  !isEndOfMacro && !(CurLexer && CurLexer->Is_PragmaLexer)) {
422  Diag(PragmaAssumeNonNullLoc, diag::err_pp_eof_in_assume_nonnull);
423 
424  // Recover by leaving immediately.
425  PragmaAssumeNonNullLoc = SourceLocation();
426  }
427 
428  // If this is a #include'd file, pop it off the include stack and continue
429  // lexing the #includer file.
430  if (!IncludeMacroStack.empty()) {
431 
432  // If we lexed the code-completion file, act as if we reached EOF.
433  if (isCodeCompletionEnabled() && CurPPLexer &&
434  SourceMgr.getLocForStartOfFile(CurPPLexer->getFileID()) ==
435  CodeCompletionFileLoc) {
436  if (CurLexer) {
437  Result.startToken();
438  CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof);
439  CurLexer.reset();
440  } else {
441  assert(CurPTHLexer && "Got EOF but no current lexer set!");
442  CurPTHLexer->getEOF(Result);
443  CurPTHLexer.reset();
444  }
445 
446  CurPPLexer = nullptr;
447  recomputeCurLexerKind();
448  return true;
449  }
450 
451  if (!isEndOfMacro && CurPPLexer &&
452  SourceMgr.getIncludeLoc(CurPPLexer->getFileID()).isValid()) {
453  // Notify SourceManager to record the number of FileIDs that were created
454  // during lexing of the #include'd file.
455  unsigned NumFIDs =
456  SourceMgr.local_sloc_entry_size() -
457  CurPPLexer->getInitialNumSLocEntries() + 1/*#include'd file*/;
458  SourceMgr.setNumCreatedFIDsForFileID(CurPPLexer->getFileID(), NumFIDs);
459  }
460 
461  bool ExitedFromPredefinesFile = false;
462  FileID ExitedFID;
463  if (!isEndOfMacro && CurPPLexer) {
464  ExitedFID = CurPPLexer->getFileID();
465 
466  assert(PredefinesFileID.isValid() &&
467  "HandleEndOfFile is called before PredefinesFileId is set");
468  ExitedFromPredefinesFile = (PredefinesFileID == ExitedFID);
469  }
470 
471  if (LeavingSubmodule) {
472  // We're done with this submodule.
473  Module *M = LeaveSubmodule(/*ForPragma*/false);
474 
475  // Notify the parser that we've left the module.
476  const char *EndPos = getCurLexerEndPos();
477  Result.startToken();
478  CurLexer->BufferPtr = EndPos;
479  CurLexer->FormTokenWithChars(Result, EndPos, tok::annot_module_end);
480  Result.setAnnotationEndLoc(Result.getLocation());
481  Result.setAnnotationValue(M);
482  }
483 
484  // We're done with the #included file.
485  RemoveTopOfLexerStack();
486 
487  // Propagate info about start-of-line/leading white-space/etc.
488  PropagateLineStartLeadingSpaceInfo(Result);
489 
490  // Notify the client, if desired, that we are in a new source file.
491  if (Callbacks && !isEndOfMacro && CurPPLexer) {
492  SrcMgr::CharacteristicKind FileType =
493  SourceMgr.getFileCharacteristic(CurPPLexer->getSourceLocation());
494  Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
495  PPCallbacks::ExitFile, FileType, ExitedFID);
496  }
497 
498  // Restore conditional stack from the preamble right after exiting from the
499  // predefines file.
500  if (ExitedFromPredefinesFile)
501  replayPreambleConditionalStack();
502 
503  // Client should lex another token unless we generated an EOM.
504  return LeavingSubmodule;
505  }
506 
507  // If this is the end of the main file, form an EOF token.
508  if (CurLexer) {
509  const char *EndPos = getCurLexerEndPos();
510  Result.startToken();
511  CurLexer->BufferPtr = EndPos;
512  CurLexer->FormTokenWithChars(Result, EndPos, tok::eof);
513 
514  if (isCodeCompletionEnabled()) {
515  // Inserting the code-completion point increases the source buffer by 1,
516  // but the main FileID was created before inserting the point.
517  // Compensate by reducing the EOF location by 1, otherwise the location
518  // will point to the next FileID.
519  // FIXME: This is hacky, the code-completion point should probably be
520  // inserted before the main FileID is created.
521  if (CurLexer->getFileLoc() == CodeCompletionFileLoc)
522  Result.setLocation(Result.getLocation().getLocWithOffset(-1));
523  }
524 
525  if (!isIncrementalProcessingEnabled())
526  // We're done with lexing.
527  CurLexer.reset();
528  } else {
529  assert(CurPTHLexer && "Got EOF but no current lexer set!");
530  CurPTHLexer->getEOF(Result);
531  CurPTHLexer.reset();
532  }
533 
534  if (!isIncrementalProcessingEnabled())
535  CurPPLexer = nullptr;
536 
537  if (TUKind == TU_Complete) {
538  // This is the end of the top-level file. 'WarnUnusedMacroLocs' has
539  // collected all macro locations that we need to warn because they are not
540  // used.
541  for (WarnUnusedMacroLocsTy::iterator
542  I=WarnUnusedMacroLocs.begin(), E=WarnUnusedMacroLocs.end();
543  I!=E; ++I)
544  Diag(*I, diag::pp_macro_not_used);
545  }
546 
547  // If we are building a module that has an umbrella header, make sure that
548  // each of the headers within the directory, including all submodules, is
549  // covered by the umbrella header was actually included by the umbrella
550  // header.
551  if (Module *Mod = getCurrentModule()) {
554  for (auto *M : AllMods)
555  diagnoseMissingHeaderInUmbrellaDir(*M);
556  }
557 
558  return true;
559 }
560 
561 /// HandleEndOfTokenLexer - This callback is invoked when the current TokenLexer
562 /// hits the end of its token stream.
564  assert(CurTokenLexer && !CurPPLexer &&
565  "Ending a macro when currently in a #include file!");
566 
567  if (!MacroExpandingLexersStack.empty() &&
568  MacroExpandingLexersStack.back().first == CurTokenLexer.get())
569  removeCachedMacroExpandedTokensOfLastLexer();
570 
571  // Delete or cache the now-dead macro expander.
572  if (NumCachedTokenLexers == TokenLexerCacheSize)
573  CurTokenLexer.reset();
574  else
575  TokenLexerCache[NumCachedTokenLexers++] = std::move(CurTokenLexer);
576 
577  // Handle this like a #include file being popped off the stack.
578  return HandleEndOfFile(Result, true);
579 }
580 
581 /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the
582 /// lexer stack. This should only be used in situations where the current
583 /// state of the top-of-stack lexer is unknown.
585  assert(!IncludeMacroStack.empty() && "Ran out of stack entries to load");
586 
587  if (CurTokenLexer) {
588  // Delete or cache the now-dead macro expander.
589  if (NumCachedTokenLexers == TokenLexerCacheSize)
590  CurTokenLexer.reset();
591  else
592  TokenLexerCache[NumCachedTokenLexers++] = std::move(CurTokenLexer);
593  }
594 
595  PopIncludeMacroStack();
596 }
597 
598 /// HandleMicrosoftCommentPaste - When the macro expander pastes together a
599 /// comment (/##/) in microsoft mode, this method handles updating the current
600 /// state, returning the token on the next source line.
602  assert(CurTokenLexer && !CurPPLexer &&
603  "Pasted comment can only be formed from macro");
604  // We handle this by scanning for the closest real lexer, switching it to
605  // raw mode and preprocessor mode. This will cause it to return \n as an
606  // explicit EOD token.
607  PreprocessorLexer *FoundLexer = nullptr;
608  bool LexerWasInPPMode = false;
609  for (const IncludeStackInfo &ISI : llvm::reverse(IncludeMacroStack)) {
610  if (ISI.ThePPLexer == nullptr) continue; // Scan for a real lexer.
611 
612  // Once we find a real lexer, mark it as raw mode (disabling macro
613  // expansions) and preprocessor mode (return EOD). We know that the lexer
614  // was *not* in raw mode before, because the macro that the comment came
615  // from was expanded. However, it could have already been in preprocessor
616  // mode (#if COMMENT) in which case we have to return it to that mode and
617  // return EOD.
618  FoundLexer = ISI.ThePPLexer;
619  FoundLexer->LexingRawMode = true;
620  LexerWasInPPMode = FoundLexer->ParsingPreprocessorDirective;
621  FoundLexer->ParsingPreprocessorDirective = true;
622  break;
623  }
624 
625  // Okay, we either found and switched over the lexer, or we didn't find a
626  // lexer. In either case, finish off the macro the comment came from, getting
627  // the next token.
628  if (!HandleEndOfTokenLexer(Tok)) Lex(Tok);
629 
630  // Discarding comments as long as we don't have EOF or EOD. This 'comments
631  // out' the rest of the line, including any tokens that came from other macros
632  // that were active, as in:
633  // #define submacro a COMMENT b
634  // submacro c
635  // which should lex to 'a' only: 'b' and 'c' should be removed.
636  while (Tok.isNot(tok::eod) && Tok.isNot(tok::eof))
637  Lex(Tok);
638 
639  // If we got an eod token, then we successfully found the end of the line.
640  if (Tok.is(tok::eod)) {
641  assert(FoundLexer && "Can't get end of line without an active lexer");
642  // Restore the lexer back to normal mode instead of raw mode.
643  FoundLexer->LexingRawMode = false;
644 
645  // If the lexer was already in preprocessor mode, just return the EOD token
646  // to finish the preprocessor line.
647  if (LexerWasInPPMode) return;
648 
649  // Otherwise, switch out of PP mode and return the next lexed token.
650  FoundLexer->ParsingPreprocessorDirective = false;
651  return Lex(Tok);
652  }
653 
654  // If we got an EOF token, then we reached the end of the token stream but
655  // didn't find an explicit \n. This can only happen if there was no lexer
656  // active (an active lexer would return EOD at EOF if there was no \n in
657  // preprocessor directive mode), so just return EOF as our token.
658  assert(!FoundLexer && "Lexer should return EOD before EOF in PP mode");
659 }
660 
662  bool ForPragma) {
663  if (!getLangOpts().ModulesLocalVisibility) {
664  // Just track that we entered this submodule.
665  BuildingSubmoduleStack.push_back(
666  BuildingSubmoduleInfo(M, ImportLoc, ForPragma, CurSubmoduleState,
667  PendingModuleMacroNames.size()));
668  return;
669  }
670 
671  // Resolve as much of the module definition as we can now, before we enter
672  // one of its headers.
673  // FIXME: Can we enable Complain here?
674  // FIXME: Can we do this when local visibility is disabled?
675  ModuleMap &ModMap = getHeaderSearchInfo().getModuleMap();
676  ModMap.resolveExports(M, /*Complain=*/false);
677  ModMap.resolveUses(M, /*Complain=*/false);
678  ModMap.resolveConflicts(M, /*Complain=*/false);
679 
680  // If this is the first time we've entered this module, set up its state.
681  auto R = Submodules.insert(std::make_pair(M, SubmoduleState()));
682  auto &State = R.first->second;
683  bool FirstTime = R.second;
684  if (FirstTime) {
685  // Determine the set of starting macros for this submodule; take these
686  // from the "null" module (the predefines buffer).
687  //
688  // FIXME: If we have local visibility but not modules enabled, the
689  // NullSubmoduleState is polluted by #defines in the top-level source
690  // file.
691  auto &StartingMacros = NullSubmoduleState.Macros;
692 
693  // Restore to the starting state.
694  // FIXME: Do this lazily, when each macro name is first referenced.
695  for (auto &Macro : StartingMacros) {
696  // Skip uninteresting macros.
697  if (!Macro.second.getLatest() &&
698  Macro.second.getOverriddenMacros().empty())
699  continue;
700 
701  MacroState MS(Macro.second.getLatest());
702  MS.setOverriddenMacros(*this, Macro.second.getOverriddenMacros());
703  State.Macros.insert(std::make_pair(Macro.first, std::move(MS)));
704  }
705  }
706 
707  // Track that we entered this module.
708  BuildingSubmoduleStack.push_back(
709  BuildingSubmoduleInfo(M, ImportLoc, ForPragma, CurSubmoduleState,
710  PendingModuleMacroNames.size()));
711 
712  // Switch to this submodule as the current submodule.
713  CurSubmoduleState = &State;
714 
715  // This module is visible to itself.
716  if (FirstTime)
717  makeModuleVisible(M, ImportLoc);
718 }
719 
720 bool Preprocessor::needModuleMacros() const {
721  // If we're not within a submodule, we never need to create ModuleMacros.
722  if (BuildingSubmoduleStack.empty())
723  return false;
724  // If we are tracking module macro visibility even for textually-included
725  // headers, we need ModuleMacros.
726  if (getLangOpts().ModulesLocalVisibility)
727  return true;
728  // Otherwise, we only need module macros if we're actually compiling a module
729  // interface.
730  return getLangOpts().isCompilingModule();
731 }
732 
734  if (BuildingSubmoduleStack.empty() ||
735  BuildingSubmoduleStack.back().IsPragma != ForPragma) {
736  assert(ForPragma && "non-pragma module enter/leave mismatch");
737  return nullptr;
738  }
739 
740  auto &Info = BuildingSubmoduleStack.back();
741 
742  Module *LeavingMod = Info.M;
743  SourceLocation ImportLoc = Info.ImportLoc;
744 
745  if (!needModuleMacros() ||
746  (!getLangOpts().ModulesLocalVisibility &&
747  LeavingMod->getTopLevelModuleName() != getLangOpts().CurrentModule)) {
748  // If we don't need module macros, or this is not a module for which we
749  // are tracking macro visibility, don't build any, and preserve the list
750  // of pending names for the surrounding submodule.
751  BuildingSubmoduleStack.pop_back();
752  makeModuleVisible(LeavingMod, ImportLoc);
753  return LeavingMod;
754  }
755 
756  // Create ModuleMacros for any macros defined in this submodule.
757  llvm::SmallPtrSet<const IdentifierInfo*, 8> VisitedMacros;
758  for (unsigned I = Info.OuterPendingModuleMacroNames;
759  I != PendingModuleMacroNames.size(); ++I) {
760  auto *II = const_cast<IdentifierInfo*>(PendingModuleMacroNames[I]);
761  if (!VisitedMacros.insert(II).second)
762  continue;
763 
764  auto MacroIt = CurSubmoduleState->Macros.find(II);
765  if (MacroIt == CurSubmoduleState->Macros.end())
766  continue;
767  auto &Macro = MacroIt->second;
768 
769  // Find the starting point for the MacroDirective chain in this submodule.
770  MacroDirective *OldMD = nullptr;
771  auto *OldState = Info.OuterSubmoduleState;
772  if (getLangOpts().ModulesLocalVisibility)
773  OldState = &NullSubmoduleState;
774  if (OldState && OldState != CurSubmoduleState) {
775  // FIXME: It'd be better to start at the state from when we most recently
776  // entered this submodule, but it doesn't really matter.
777  auto &OldMacros = OldState->Macros;
778  auto OldMacroIt = OldMacros.find(II);
779  if (OldMacroIt == OldMacros.end())
780  OldMD = nullptr;
781  else
782  OldMD = OldMacroIt->second.getLatest();
783  }
784 
785  // This module may have exported a new macro. If so, create a ModuleMacro
786  // representing that fact.
787  bool ExplicitlyPublic = false;
788  for (auto *MD = Macro.getLatest(); MD != OldMD; MD = MD->getPrevious()) {
789  assert(MD && "broken macro directive chain");
790 
791  if (auto *VisMD = dyn_cast<VisibilityMacroDirective>(MD)) {
792  // The latest visibility directive for a name in a submodule affects
793  // all the directives that come before it.
794  if (VisMD->isPublic())
795  ExplicitlyPublic = true;
796  else if (!ExplicitlyPublic)
797  // Private with no following public directive: not exported.
798  break;
799  } else {
800  MacroInfo *Def = nullptr;
801  if (DefMacroDirective *DefMD = dyn_cast<DefMacroDirective>(MD))
802  Def = DefMD->getInfo();
803 
804  // FIXME: Issue a warning if multiple headers for the same submodule
805  // define a macro, rather than silently ignoring all but the first.
806  bool IsNew;
807  // Don't bother creating a module macro if it would represent a #undef
808  // that doesn't override anything.
809  if (Def || !Macro.getOverriddenMacros().empty())
810  addModuleMacro(LeavingMod, II, Def,
811  Macro.getOverriddenMacros(), IsNew);
812 
813  if (!getLangOpts().ModulesLocalVisibility) {
814  // This macro is exposed to the rest of this compilation as a
815  // ModuleMacro; we don't need to track its MacroDirective any more.
816  Macro.setLatest(nullptr);
817  Macro.setOverriddenMacros(*this, {});
818  }
819  break;
820  }
821  }
822  }
823  PendingModuleMacroNames.resize(Info.OuterPendingModuleMacroNames);
824 
825  // FIXME: Before we leave this submodule, we should parse all the other
826  // headers within it. Otherwise, we're left with an inconsistent state
827  // where we've made the module visible but don't yet have its complete
828  // contents.
829 
830  // Put back the outer module's state, if we're tracking it.
831  if (getLangOpts().ModulesLocalVisibility)
832  CurSubmoduleState = Info.OuterSubmoduleState;
833 
834  BuildingSubmoduleStack.pop_back();
835 
836  // A nested #include makes the included submodule visible.
837  makeModuleVisible(LeavingMod, ImportLoc);
838  return LeavingMod;
839 }
static DiagnosticBuilder Diag(DiagnosticsEngine *Diags, const LangOptions &Features, FullSourceLoc TokLoc, const char *TokBegin, const char *TokRangeBegin, const char *TokRangeEnd, unsigned DiagID)
Produce a diagnostic highlighting some portion of a literal.
Lexer - This provides a simple interface that turns a text buffer into a stream of tokens...
Definition: Lexer.h:77
SourceLocation getLocWithOffset(int Offset) const
Return a source location with the specified offset from this SourceLocation.
Implements support for file system lookup, file system caching, and directory search management...
Definition: FileManager.h:116
Defines the clang::FileManager interface and associated types.
bool is(tok::TokenKind K) const
is/isNot - Predicates to check if this token is a specific kind, as in "if (Tok.is(tok::l_brace)) {...
Definition: Token.h:95
Defines the SourceManager interface.
bool isInPrimaryFile() const
Return true if we&#39;re in the top-level file, not in a #include.
Defines the clang::MacroInfo and clang::MacroDirective classes.
bool resolveUses(Module *Mod, bool Complain)
Resolve all of the unresolved uses in the given module.
Definition: ModuleMap.cpp:1236
CharacteristicKind
Indicates whether a file or directory holds normal user code, system code, or system code which is im...
Definition: SourceManager.h:78
const MacroDirective * getPrevious() const
Get previous definition of the macro with the same name.
Definition: MacroInfo.h:329
bool HandleEndOfTokenLexer(Token &Result)
Callback invoked when the current TokenLexer hits the end of its token stream.
The virtual file system interface.
One of these records is kept for each identifier that is lexed.
A directive for a defined macro or a macro imported from a module.
Definition: MacroInfo.h:406
bool ParsingPreprocessorDirective
True when parsing #XXX; turns &#39;\n&#39; into a tok::eod token.
bool resolveConflicts(Module *Mod, bool Complain)
Resolve all of the unresolved conflicts in the given module.
Definition: ModuleMap.cpp:1249
LineState State
void EnterSubmodule(Module *M, SourceLocation ImportLoc, bool ForPragma)
Module * LeaveSubmodule(bool ForPragma)
Header getUmbrellaHeader() const
Retrieve the header that serves as the umbrella header for this module.
Definition: Module.h:483
Token - This structure provides full information about a lexed token.
Definition: Token.h:35
An input iterator over the recursive contents of a virtual path, similar to llvm::sys::fs::recursive_...
Describes a module or submodule.
Definition: Module.h:65
StringRef getTopLevelModuleName() const
Retrieve the name of the top-level module.
Definition: Module.h:461
std::string getFullModuleName(bool AllowStringLiterals=false) const
Retrieve the full name of this module, including the path from its top-level module.
Definition: Module.cpp:182
const FormatToken & Tok
const DirectoryEntry * getDirectory(StringRef DirName, bool CacheFailure=true)
Lookup, cache, and verify the specified directory (real or virtual).
const DirectoryEntry * Entry
Definition: Module.h:162
void setAnnotationValue(void *val)
Definition: Token.h:228
bool LexingRawMode
True if in raw mode.
SourceLocation End
MacroArgs - An instance of this class captures information about the formal arguments specified to a ...
Definition: MacroArgs.h:30
SourceLocation getLocation() const
Return a source location identifier for the specified offset in the current file. ...
Definition: Token.h:124
Defines the clang::Preprocessor interface.
void HandleMicrosoftCommentPaste(Token &Tok)
When the macro expander pastes together a comment (/##/) in Microsoft mode, this method handles updat...
bool resolveExports(Module *Mod, bool Complain)
Resolve all of the unresolved exports in the given module.
Definition: ModuleMap.cpp:1223
static void computeRelativePath(FileManager &FM, const DirectoryEntry *Dir, const FileEntry *File, SmallString< 128 > &Result)
Compute the relative path that names the given file relative to the given directory.
bool HandleEndOfFile(Token &Result, bool isEndOfMacro=false)
Callback invoked when the lexer hits the end of the current file.
The result type of a method or function.
DirectoryLookup - This class represents one entry in the search list that specifies the search order ...
const DirectoryEntry * getDir() const
Return the directory the file lives in.
Definition: FileManager.h:94
Encapsulates changes to the "macros namespace" (the location where the macro name became active...
Definition: MacroInfo.h:291
Encodes a location in the source.
StringRef getName() const
Definition: FileManager.h:84
void setAnnotationEndLoc(SourceLocation L)
Definition: Token.h:142
Cached information about one file (either on disk or in the virtual file system). ...
Definition: FileManager.h:59
void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroInfo *Macro, MacroArgs *Args)
Add a Macro to the top of the include stack and start lexing tokens from it instead of the current bu...
An opaque identifier used by SourceManager which refers to a source file (MemoryBuffer) along with it...
bool isNot(tok::TokenKind K) const
Definition: Token.h:96
Dataflow Directional Tag Classes.
llvm::iterator_range< submodule_iterator > submodules()
Definition: Module.h:561
Encapsulates the data about a macro definition (e.g.
Definition: MacroInfo.h:40
DirectoryName getUmbrellaDir() const
Retrieve the directory for which this module serves as the umbrella.
Definition: Module.cpp:207
Cached information about one directory (either on disk or in the virtual file system).
Definition: FileManager.h:45
static void collectAllSubModulesWithUmbrellaHeader(const Module &Mod, SmallVectorImpl< const Module *> &SubMods)
static bool isMacroDefined(const Sema &S, SourceLocation Loc, StringRef Name)
void RemoveTopOfLexerStack()
Pop the current lexer/macro exp off the top of the lexer stack.
static FixItHint CreateReplacement(CharSourceRange RemoveRange, StringRef Code)
Create a code modification hint that replaces the given source range with the given code string...
Definition: Diagnostic.h:129
The translation unit is a complete translation unit.
Definition: LangOptions.h:303
bool isHeaderInUnavailableModule(const FileEntry *Header) const
Determine whether the given header is part of a module marked &#39;unavailable&#39;.
Definition: ModuleMap.cpp:666
__DEVICE__ int max(int __a, int __b)
void setLocation(SourceLocation L)
Definition: Token.h:132
PreprocessorLexer * getCurrentFileLexer() const
Return the current file lexer being lexed from.
void startToken()
Reset all flags to cleared.
Definition: Token.h:169
StringRef getName() const
Definition: FileManager.h:51
bool EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir, SourceLocation Loc)
Add a source file to the top of the include stack and start lexing tokens from it instead of the curr...