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"
16 #include "clang/Lex/HeaderSearch.h"
17 #include "clang/Lex/MacroInfo.h"
18 #include "clang/Lex/LexDiagnostic.h"
19 #include "clang/Basic/FileManager.h"
20 #include "clang/Basic/SourceManager.h"
21 #include "llvm/Support/FileSystem.h"
22 #include "llvm/Support/MemoryBuffer.h"
23 #include "llvm/Support/PathV2.h"
24 #include "llvm/ADT/StringSwitch.h"
25 using namespace clang;
26
~PPCallbacks()27 PPCallbacks::~PPCallbacks() {}
28
29 //===----------------------------------------------------------------------===//
30 // Miscellaneous Methods.
31 //===----------------------------------------------------------------------===//
32
33 /// isInPrimaryFile - Return true if we're in the top-level file, not in a
34 /// #include. This looks through macro expansions and active _Pragma lexers.
isInPrimaryFile() const35 bool Preprocessor::isInPrimaryFile() const {
36 if (IsFileLexer())
37 return IncludeMacroStack.empty();
38
39 // If there are any stacked lexers, we're in a #include.
40 assert(IsFileLexer(IncludeMacroStack[0]) &&
41 "Top level include stack isn't our primary lexer?");
42 for (unsigned i = 1, e = IncludeMacroStack.size(); i != e; ++i)
43 if (IsFileLexer(IncludeMacroStack[i]))
44 return false;
45 return true;
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.
getCurrentFileLexer() const51 PreprocessorLexer *Preprocessor::getCurrentFileLexer() const {
52 if (IsFileLexer())
53 return CurPPLexer;
54
55 // Look for a stacked lexer.
56 for (unsigned i = IncludeMacroStack.size(); i != 0; --i) {
57 const IncludeStackInfo& ISI = IncludeMacroStack[i-1];
58 if (IsFileLexer(ISI))
59 return ISI.ThePPLexer;
60 }
61 return 0;
62 }
63
64
65 //===----------------------------------------------------------------------===//
66 // Methods for Entering and Callbacks for leaving various contexts
67 //===----------------------------------------------------------------------===//
68
69 /// EnterSourceFile - Add a source file to the top of the include stack and
70 /// start lexing tokens from it instead of the current buffer.
EnterSourceFile(FileID FID,const DirectoryLookup * CurDir,SourceLocation Loc)71 void Preprocessor::EnterSourceFile(FileID FID, const DirectoryLookup *CurDir,
72 SourceLocation Loc) {
73 assert(CurTokenLexer == 0 && "Cannot #include a file inside a macro!");
74 ++NumEnteredSourceFiles;
75
76 if (MaxIncludeStackDepth < IncludeMacroStack.size())
77 MaxIncludeStackDepth = IncludeMacroStack.size();
78
79 if (PTH) {
80 if (PTHLexer *PL = PTH->CreateLexer(FID)) {
81 EnterSourceFileWithPTH(PL, CurDir);
82 return;
83 }
84 }
85
86 // Get the MemoryBuffer for this FID, if it fails, we fail.
87 bool Invalid = false;
88 const llvm::MemoryBuffer *InputFile =
89 getSourceManager().getBuffer(FID, Loc, &Invalid);
90 if (Invalid) {
91 SourceLocation FileStart = SourceMgr.getLocForStartOfFile(FID);
92 Diag(Loc, diag::err_pp_error_opening_file)
93 << std::string(SourceMgr.getBufferName(FileStart)) << "";
94 return;
95 }
96
97 if (isCodeCompletionEnabled() &&
98 SourceMgr.getFileEntryForID(FID) == CodeCompletionFile) {
99 CodeCompletionFileLoc = SourceMgr.getLocForStartOfFile(FID);
100 CodeCompletionLoc =
101 CodeCompletionFileLoc.getLocWithOffset(CodeCompletionOffset);
102 }
103
104 EnterSourceFileWithLexer(new Lexer(FID, InputFile, *this), CurDir);
105 return;
106 }
107
108 /// EnterSourceFileWithLexer - Add a source file to the top of the include stack
109 /// and start lexing tokens from it instead of the current buffer.
EnterSourceFileWithLexer(Lexer * TheLexer,const DirectoryLookup * CurDir)110 void Preprocessor::EnterSourceFileWithLexer(Lexer *TheLexer,
111 const DirectoryLookup *CurDir) {
112
113 // Add the current lexer to the include stack.
114 if (CurPPLexer || CurTokenLexer)
115 PushIncludeMacroStack();
116
117 CurLexer.reset(TheLexer);
118 CurPPLexer = TheLexer;
119 CurDirLookup = CurDir;
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.
EnterSourceFileWithPTH(PTHLexer * PL,const DirectoryLookup * CurDir)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 if (CurLexerKind != CLK_LexAfterModuleImport)
145 CurLexerKind = CLK_PTHLexer;
146
147 // Notify the client, if desired, that we are in a new source file.
148 if (Callbacks) {
149 FileID FID = CurPPLexer->getFileID();
150 SourceLocation EnterLoc = SourceMgr.getLocForStartOfFile(FID);
151 SrcMgr::CharacteristicKind FileType =
152 SourceMgr.getFileCharacteristic(EnterLoc);
153 Callbacks->FileChanged(EnterLoc, PPCallbacks::EnterFile, FileType);
154 }
155 }
156
157 /// EnterMacro - Add a Macro to the top of the include stack and start lexing
158 /// tokens from it instead of the current buffer.
EnterMacro(Token & Tok,SourceLocation ILEnd,MacroArgs * Args)159 void Preprocessor::EnterMacro(Token &Tok, SourceLocation ILEnd,
160 MacroArgs *Args) {
161 PushIncludeMacroStack();
162 CurDirLookup = 0;
163
164 if (NumCachedTokenLexers == 0) {
165 CurTokenLexer.reset(new TokenLexer(Tok, ILEnd, Args, *this));
166 } else {
167 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]);
168 CurTokenLexer->Init(Tok, ILEnd, Args);
169 }
170 if (CurLexerKind != CLK_LexAfterModuleImport)
171 CurLexerKind = CLK_TokenLexer;
172 }
173
174 /// EnterTokenStream - Add a "macro" context to the top of the include stack,
175 /// which will cause the lexer to start returning the specified tokens.
176 ///
177 /// If DisableMacroExpansion is true, tokens lexed from the token stream will
178 /// not be subject to further macro expansion. Otherwise, these tokens will
179 /// be re-macro-expanded when/if expansion is enabled.
180 ///
181 /// If OwnsTokens is false, this method assumes that the specified stream of
182 /// tokens has a permanent owner somewhere, so they do not need to be copied.
183 /// If it is true, it assumes the array of tokens is allocated with new[] and
184 /// must be freed.
185 ///
EnterTokenStream(const Token * Toks,unsigned NumToks,bool DisableMacroExpansion,bool OwnsTokens)186 void Preprocessor::EnterTokenStream(const Token *Toks, unsigned NumToks,
187 bool DisableMacroExpansion,
188 bool OwnsTokens) {
189 // Save our current state.
190 PushIncludeMacroStack();
191 CurDirLookup = 0;
192
193 // Create a macro expander to expand from the specified token stream.
194 if (NumCachedTokenLexers == 0) {
195 CurTokenLexer.reset(new TokenLexer(Toks, NumToks, DisableMacroExpansion,
196 OwnsTokens, *this));
197 } else {
198 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]);
199 CurTokenLexer->Init(Toks, NumToks, DisableMacroExpansion, OwnsTokens);
200 }
201 if (CurLexerKind != CLK_LexAfterModuleImport)
202 CurLexerKind = CLK_TokenLexer;
203 }
204
205 /// \brief Compute the relative path that names the given file relative to
206 /// the given directory.
computeRelativePath(FileManager & FM,const DirectoryEntry * Dir,const FileEntry * File,SmallString<128> & Result)207 static void computeRelativePath(FileManager &FM, const DirectoryEntry *Dir,
208 const FileEntry *File,
209 SmallString<128> &Result) {
210 Result.clear();
211
212 StringRef FilePath = File->getDir()->getName();
213 StringRef Path = FilePath;
214 while (!Path.empty()) {
215 if (const DirectoryEntry *CurDir = FM.getDirectory(Path)) {
216 if (CurDir == Dir) {
217 Result = FilePath.substr(Path.size());
218 llvm::sys::path::append(Result,
219 llvm::sys::path::filename(File->getName()));
220 return;
221 }
222 }
223
224 Path = llvm::sys::path::parent_path(Path);
225 }
226
227 Result = File->getName();
228 }
229
230 /// HandleEndOfFile - This callback is invoked when the lexer hits the end of
231 /// the current file. This either returns the EOF token or pops a level off
232 /// the include stack and keeps going.
HandleEndOfFile(Token & Result,bool isEndOfMacro)233 bool Preprocessor::HandleEndOfFile(Token &Result, bool isEndOfMacro) {
234 assert(!CurTokenLexer &&
235 "Ending a file when currently in a macro!");
236
237 // See if this file had a controlling macro.
238 if (CurPPLexer) { // Not ending a macro, ignore it.
239 if (const IdentifierInfo *ControllingMacro =
240 CurPPLexer->MIOpt.GetControllingMacroAtEndOfFile()) {
241 // Okay, this has a controlling macro, remember in HeaderFileInfo.
242 if (const FileEntry *FE =
243 SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))
244 HeaderInfo.SetFileControllingMacro(FE, ControllingMacro);
245 }
246 }
247
248 // Complain about reaching a true EOF within arc_cf_code_audited.
249 // We don't want to complain about reaching the end of a macro
250 // instantiation or a _Pragma.
251 if (PragmaARCCFCodeAuditedLoc.isValid() &&
252 !isEndOfMacro && !(CurLexer && CurLexer->Is_PragmaLexer)) {
253 Diag(PragmaARCCFCodeAuditedLoc, diag::err_pp_eof_in_arc_cf_code_audited);
254
255 // Recover by leaving immediately.
256 PragmaARCCFCodeAuditedLoc = SourceLocation();
257 }
258
259 // If this is a #include'd file, pop it off the include stack and continue
260 // lexing the #includer file.
261 if (!IncludeMacroStack.empty()) {
262
263 // If we lexed the code-completion file, act as if we reached EOF.
264 if (isCodeCompletionEnabled() && CurPPLexer &&
265 SourceMgr.getLocForStartOfFile(CurPPLexer->getFileID()) ==
266 CodeCompletionFileLoc) {
267 if (CurLexer) {
268 Result.startToken();
269 CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof);
270 CurLexer.reset();
271 } else {
272 assert(CurPTHLexer && "Got EOF but no current lexer set!");
273 CurPTHLexer->getEOF(Result);
274 CurPTHLexer.reset();
275 }
276
277 CurPPLexer = 0;
278 return true;
279 }
280
281 if (!isEndOfMacro && CurPPLexer &&
282 SourceMgr.getIncludeLoc(CurPPLexer->getFileID()).isValid()) {
283 // Notify SourceManager to record the number of FileIDs that were created
284 // during lexing of the #include'd file.
285 unsigned NumFIDs =
286 SourceMgr.local_sloc_entry_size() -
287 CurPPLexer->getInitialNumSLocEntries() + 1/*#include'd file*/;
288 SourceMgr.setNumCreatedFIDsForFileID(CurPPLexer->getFileID(), NumFIDs);
289 }
290
291 FileID ExitedFID;
292 if (Callbacks && !isEndOfMacro && CurPPLexer)
293 ExitedFID = CurPPLexer->getFileID();
294
295 // We're done with the #included file.
296 RemoveTopOfLexerStack();
297
298 // Notify the client, if desired, that we are in a new source file.
299 if (Callbacks && !isEndOfMacro && CurPPLexer) {
300 SrcMgr::CharacteristicKind FileType =
301 SourceMgr.getFileCharacteristic(CurPPLexer->getSourceLocation());
302 Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
303 PPCallbacks::ExitFile, FileType, ExitedFID);
304 }
305
306 // Client should lex another token.
307 return false;
308 }
309
310 // If the file ends with a newline, form the EOF token on the newline itself,
311 // rather than "on the line following it", which doesn't exist. This makes
312 // diagnostics relating to the end of file include the last file that the user
313 // actually typed, which is goodness.
314 if (CurLexer) {
315 const char *EndPos = CurLexer->BufferEnd;
316 if (EndPos != CurLexer->BufferStart &&
317 (EndPos[-1] == '\n' || EndPos[-1] == '\r')) {
318 --EndPos;
319
320 // Handle \n\r and \r\n:
321 if (EndPos != CurLexer->BufferStart &&
322 (EndPos[-1] == '\n' || EndPos[-1] == '\r') &&
323 EndPos[-1] != EndPos[0])
324 --EndPos;
325 }
326
327 Result.startToken();
328 CurLexer->BufferPtr = EndPos;
329 CurLexer->FormTokenWithChars(Result, EndPos, tok::eof);
330
331 if (!isIncrementalProcessingEnabled())
332 // We're done with lexing.
333 CurLexer.reset();
334 } else {
335 assert(CurPTHLexer && "Got EOF but no current lexer set!");
336 CurPTHLexer->getEOF(Result);
337 CurPTHLexer.reset();
338 }
339
340 if (!isIncrementalProcessingEnabled())
341 CurPPLexer = 0;
342
343 // This is the end of the top-level file. 'WarnUnusedMacroLocs' has collected
344 // all macro locations that we need to warn because they are not used.
345 for (WarnUnusedMacroLocsTy::iterator
346 I=WarnUnusedMacroLocs.begin(), E=WarnUnusedMacroLocs.end(); I!=E; ++I)
347 Diag(*I, diag::pp_macro_not_used);
348
349 // If we are building a module that has an umbrella header, make sure that
350 // each of the headers within the directory covered by the umbrella header
351 // was actually included by the umbrella header.
352 if (Module *Mod = getCurrentModule()) {
353 if (Mod->getUmbrellaHeader()) {
354 SourceLocation StartLoc
355 = SourceMgr.getLocForStartOfFile(SourceMgr.getMainFileID());
356
357 if (getDiagnostics().getDiagnosticLevel(
358 diag::warn_uncovered_module_header,
359 StartLoc) != DiagnosticsEngine::Ignored) {
360 ModuleMap &ModMap = getHeaderSearchInfo().getModuleMap();
361 typedef llvm::sys::fs::recursive_directory_iterator
362 recursive_directory_iterator;
363 const DirectoryEntry *Dir = Mod->getUmbrellaDir();
364 llvm::error_code EC;
365 for (recursive_directory_iterator Entry(Dir->getName(), EC), End;
366 Entry != End && !EC; Entry.increment(EC)) {
367 using llvm::StringSwitch;
368
369 // Check whether this entry has an extension typically associated with
370 // headers.
371 if (!StringSwitch<bool>(llvm::sys::path::extension(Entry->path()))
372 .Cases(".h", ".H", ".hh", ".hpp", true)
373 .Default(false))
374 continue;
375
376 if (const FileEntry *Header = getFileManager().getFile(Entry->path()))
377 if (!getSourceManager().hasFileInfo(Header)) {
378 if (!ModMap.isHeaderInUnavailableModule(Header)) {
379 // Find the relative path that would access this header.
380 SmallString<128> RelativePath;
381 computeRelativePath(FileMgr, Dir, Header, RelativePath);
382 Diag(StartLoc, diag::warn_uncovered_module_header)
383 << RelativePath;
384 }
385 }
386 }
387 }
388 }
389 }
390
391 return true;
392 }
393
394 /// HandleEndOfTokenLexer - This callback is invoked when the current TokenLexer
395 /// hits the end of its token stream.
HandleEndOfTokenLexer(Token & Result)396 bool Preprocessor::HandleEndOfTokenLexer(Token &Result) {
397 assert(CurTokenLexer && !CurPPLexer &&
398 "Ending a macro when currently in a #include file!");
399
400 if (!MacroExpandingLexersStack.empty() &&
401 MacroExpandingLexersStack.back().first == CurTokenLexer.get())
402 removeCachedMacroExpandedTokensOfLastLexer();
403
404 // Delete or cache the now-dead macro expander.
405 if (NumCachedTokenLexers == TokenLexerCacheSize)
406 CurTokenLexer.reset();
407 else
408 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take();
409
410 // Handle this like a #include file being popped off the stack.
411 return HandleEndOfFile(Result, true);
412 }
413
414 /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the
415 /// lexer stack. This should only be used in situations where the current
416 /// state of the top-of-stack lexer is unknown.
RemoveTopOfLexerStack()417 void Preprocessor::RemoveTopOfLexerStack() {
418 assert(!IncludeMacroStack.empty() && "Ran out of stack entries to load");
419
420 if (CurTokenLexer) {
421 // Delete or cache the now-dead macro expander.
422 if (NumCachedTokenLexers == TokenLexerCacheSize)
423 CurTokenLexer.reset();
424 else
425 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take();
426 }
427
428 PopIncludeMacroStack();
429 }
430
431 /// HandleMicrosoftCommentPaste - When the macro expander pastes together a
432 /// comment (/##/) in microsoft mode, this method handles updating the current
433 /// state, returning the token on the next source line.
HandleMicrosoftCommentPaste(Token & Tok)434 void Preprocessor::HandleMicrosoftCommentPaste(Token &Tok) {
435 assert(CurTokenLexer && !CurPPLexer &&
436 "Pasted comment can only be formed from macro");
437
438 // We handle this by scanning for the closest real lexer, switching it to
439 // raw mode and preprocessor mode. This will cause it to return \n as an
440 // explicit EOD token.
441 PreprocessorLexer *FoundLexer = 0;
442 bool LexerWasInPPMode = false;
443 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
444 IncludeStackInfo &ISI = *(IncludeMacroStack.end()-i-1);
445 if (ISI.ThePPLexer == 0) continue; // Scan for a real lexer.
446
447 // Once we find a real lexer, mark it as raw mode (disabling macro
448 // expansions) and preprocessor mode (return EOD). We know that the lexer
449 // was *not* in raw mode before, because the macro that the comment came
450 // from was expanded. However, it could have already been in preprocessor
451 // mode (#if COMMENT) in which case we have to return it to that mode and
452 // return EOD.
453 FoundLexer = ISI.ThePPLexer;
454 FoundLexer->LexingRawMode = true;
455 LexerWasInPPMode = FoundLexer->ParsingPreprocessorDirective;
456 FoundLexer->ParsingPreprocessorDirective = true;
457 break;
458 }
459
460 // Okay, we either found and switched over the lexer, or we didn't find a
461 // lexer. In either case, finish off the macro the comment came from, getting
462 // the next token.
463 if (!HandleEndOfTokenLexer(Tok)) Lex(Tok);
464
465 // Discarding comments as long as we don't have EOF or EOD. This 'comments
466 // out' the rest of the line, including any tokens that came from other macros
467 // that were active, as in:
468 // #define submacro a COMMENT b
469 // submacro c
470 // which should lex to 'a' only: 'b' and 'c' should be removed.
471 while (Tok.isNot(tok::eod) && Tok.isNot(tok::eof))
472 Lex(Tok);
473
474 // If we got an eod token, then we successfully found the end of the line.
475 if (Tok.is(tok::eod)) {
476 assert(FoundLexer && "Can't get end of line without an active lexer");
477 // Restore the lexer back to normal mode instead of raw mode.
478 FoundLexer->LexingRawMode = false;
479
480 // If the lexer was already in preprocessor mode, just return the EOD token
481 // to finish the preprocessor line.
482 if (LexerWasInPPMode) return;
483
484 // Otherwise, switch out of PP mode and return the next lexed token.
485 FoundLexer->ParsingPreprocessorDirective = false;
486 return Lex(Tok);
487 }
488
489 // If we got an EOF token, then we reached the end of the token stream but
490 // didn't find an explicit \n. This can only happen if there was no lexer
491 // active (an active lexer would return EOD at EOF if there was no \n in
492 // preprocessor directive mode), so just return EOF as our token.
493 assert(!FoundLexer && "Lexer should return EOD before EOF in PP mode");
494 }
495