• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //===--- PTHLexer.cpp - Lex from a token stream ---------------------------===//
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 the PTHLexer interface.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "clang/Lex/PTHLexer.h"
15 #include "clang/Basic/FileManager.h"
16 #include "clang/Basic/FileSystemStatCache.h"
17 #include "clang/Basic/IdentifierTable.h"
18 #include "clang/Basic/OnDiskHashTable.h"
19 #include "clang/Basic/TokenKinds.h"
20 #include "clang/Lex/LexDiagnostic.h"
21 #include "clang/Lex/PTHManager.h"
22 #include "clang/Lex/Preprocessor.h"
23 #include "clang/Lex/Token.h"
24 #include "llvm/ADT/OwningPtr.h"
25 #include "llvm/ADT/StringExtras.h"
26 #include "llvm/ADT/StringMap.h"
27 #include "llvm/Support/MemoryBuffer.h"
28 #include "llvm/Support/system_error.h"
29 using namespace clang;
30 using namespace clang::io;
31 
32 #define DISK_TOKEN_SIZE (1+1+2+4+4)
33 
34 //===----------------------------------------------------------------------===//
35 // PTHLexer methods.
36 //===----------------------------------------------------------------------===//
37 
PTHLexer(Preprocessor & PP,FileID FID,const unsigned char * D,const unsigned char * ppcond,PTHManager & PM)38 PTHLexer::PTHLexer(Preprocessor &PP, FileID FID, const unsigned char *D,
39                    const unsigned char *ppcond, PTHManager &PM)
40   : PreprocessorLexer(&PP, FID), TokBuf(D), CurPtr(D), LastHashTokPtr(0),
41     PPCond(ppcond), CurPPCondPtr(ppcond), PTHMgr(PM) {
42 
43   FileStartLoc = PP.getSourceManager().getLocForStartOfFile(FID);
44 }
45 
Lex(Token & Tok)46 void PTHLexer::Lex(Token& Tok) {
47 LexNextToken:
48 
49   //===--------------------------------------==//
50   // Read the raw token data.
51   //===--------------------------------------==//
52 
53   // Shadow CurPtr into an automatic variable.
54   const unsigned char *CurPtrShadow = CurPtr;
55 
56   // Read in the data for the token.
57   unsigned Word0 = ReadLE32(CurPtrShadow);
58   uint32_t IdentifierID = ReadLE32(CurPtrShadow);
59   uint32_t FileOffset = ReadLE32(CurPtrShadow);
60 
61   tok::TokenKind TKind = (tok::TokenKind) (Word0 & 0xFF);
62   Token::TokenFlags TFlags = (Token::TokenFlags) ((Word0 >> 8) & 0xFF);
63   uint32_t Len = Word0 >> 16;
64 
65   CurPtr = CurPtrShadow;
66 
67   //===--------------------------------------==//
68   // Construct the token itself.
69   //===--------------------------------------==//
70 
71   Tok.startToken();
72   Tok.setKind(TKind);
73   Tok.setFlag(TFlags);
74   assert(!LexingRawMode);
75   Tok.setLocation(FileStartLoc.getLocWithOffset(FileOffset));
76   Tok.setLength(Len);
77 
78   // Handle identifiers.
79   if (Tok.isLiteral()) {
80     Tok.setLiteralData((const char*) (PTHMgr.SpellingBase + IdentifierID));
81   }
82   else if (IdentifierID) {
83     MIOpt.ReadToken();
84     IdentifierInfo *II = PTHMgr.GetIdentifierInfo(IdentifierID-1);
85 
86     Tok.setIdentifierInfo(II);
87 
88     // Change the kind of this identifier to the appropriate token kind, e.g.
89     // turning "for" into a keyword.
90     Tok.setKind(II->getTokenID());
91 
92     if (II->isHandleIdentifierCase())
93       PP->HandleIdentifier(Tok);
94     return;
95   }
96 
97   //===--------------------------------------==//
98   // Process the token.
99   //===--------------------------------------==//
100   if (TKind == tok::eof) {
101     // Save the end-of-file token.
102     EofToken = Tok;
103 
104     // Save 'PP' to 'PPCache' as LexEndOfFile can delete 'this'.
105     Preprocessor *PPCache = PP;
106 
107     assert(!ParsingPreprocessorDirective);
108     assert(!LexingRawMode);
109 
110     if (LexEndOfFile(Tok))
111       return;
112 
113     return PPCache->Lex(Tok);
114   }
115 
116   if (TKind == tok::hash && Tok.isAtStartOfLine()) {
117     LastHashTokPtr = CurPtr - DISK_TOKEN_SIZE;
118     assert(!LexingRawMode);
119     PP->HandleDirective(Tok);
120 
121     if (PP->isCurrentLexer(this))
122       goto LexNextToken;
123 
124     return PP->Lex(Tok);
125   }
126 
127   if (TKind == tok::eod) {
128     assert(ParsingPreprocessorDirective);
129     ParsingPreprocessorDirective = false;
130     return;
131   }
132 
133   MIOpt.ReadToken();
134 }
135 
LexEndOfFile(Token & Result)136 bool PTHLexer::LexEndOfFile(Token &Result) {
137   // If we hit the end of the file while parsing a preprocessor directive,
138   // end the preprocessor directive first.  The next token returned will
139   // then be the end of file.
140   if (ParsingPreprocessorDirective) {
141     ParsingPreprocessorDirective = false; // Done parsing the "line".
142     return true;  // Have a token.
143   }
144 
145   assert(!LexingRawMode);
146 
147   // If we are in a #if directive, emit an error.
148   while (!ConditionalStack.empty()) {
149     if (PP->getCodeCompletionFileLoc() != FileStartLoc)
150       PP->Diag(ConditionalStack.back().IfLoc,
151                diag::err_pp_unterminated_conditional);
152     ConditionalStack.pop_back();
153   }
154 
155   // Finally, let the preprocessor handle this.
156   return PP->HandleEndOfFile(Result);
157 }
158 
159 // FIXME: We can just grab the last token instead of storing a copy
160 // into EofToken.
getEOF(Token & Tok)161 void PTHLexer::getEOF(Token& Tok) {
162   assert(EofToken.is(tok::eof));
163   Tok = EofToken;
164 }
165 
DiscardToEndOfLine()166 void PTHLexer::DiscardToEndOfLine() {
167   assert(ParsingPreprocessorDirective && ParsingFilename == false &&
168          "Must be in a preprocessing directive!");
169 
170   // We assume that if the preprocessor wishes to discard to the end of
171   // the line that it also means to end the current preprocessor directive.
172   ParsingPreprocessorDirective = false;
173 
174   // Skip tokens by only peeking at their token kind and the flags.
175   // We don't need to actually reconstruct full tokens from the token buffer.
176   // This saves some copies and it also reduces IdentifierInfo* lookup.
177   const unsigned char* p = CurPtr;
178   while (1) {
179     // Read the token kind.  Are we at the end of the file?
180     tok::TokenKind x = (tok::TokenKind) (uint8_t) *p;
181     if (x == tok::eof) break;
182 
183     // Read the token flags.  Are we at the start of the next line?
184     Token::TokenFlags y = (Token::TokenFlags) (uint8_t) p[1];
185     if (y & Token::StartOfLine) break;
186 
187     // Skip to the next token.
188     p += DISK_TOKEN_SIZE;
189   }
190 
191   CurPtr = p;
192 }
193 
194 /// SkipBlock - Used by Preprocessor to skip the current conditional block.
SkipBlock()195 bool PTHLexer::SkipBlock() {
196   assert(CurPPCondPtr && "No cached PP conditional information.");
197   assert(LastHashTokPtr && "No known '#' token.");
198 
199   const unsigned char* HashEntryI = 0;
200   uint32_t TableIdx;
201 
202   do {
203     // Read the token offset from the side-table.
204     uint32_t Offset = ReadLE32(CurPPCondPtr);
205 
206     // Read the target table index from the side-table.
207     TableIdx = ReadLE32(CurPPCondPtr);
208 
209     // Compute the actual memory address of the '#' token data for this entry.
210     HashEntryI = TokBuf + Offset;
211 
212     // Optmization: "Sibling jumping".  #if...#else...#endif blocks can
213     //  contain nested blocks.  In the side-table we can jump over these
214     //  nested blocks instead of doing a linear search if the next "sibling"
215     //  entry is not at a location greater than LastHashTokPtr.
216     if (HashEntryI < LastHashTokPtr && TableIdx) {
217       // In the side-table we are still at an entry for a '#' token that
218       // is earlier than the last one we saw.  Check if the location we would
219       // stride gets us closer.
220       const unsigned char* NextPPCondPtr =
221         PPCond + TableIdx*(sizeof(uint32_t)*2);
222       assert(NextPPCondPtr >= CurPPCondPtr);
223       // Read where we should jump to.
224       const unsigned char* HashEntryJ = TokBuf + ReadLE32(NextPPCondPtr);
225 
226       if (HashEntryJ <= LastHashTokPtr) {
227         // Jump directly to the next entry in the side table.
228         HashEntryI = HashEntryJ;
229         TableIdx = ReadLE32(NextPPCondPtr);
230         CurPPCondPtr = NextPPCondPtr;
231       }
232     }
233   }
234   while (HashEntryI < LastHashTokPtr);
235   assert(HashEntryI == LastHashTokPtr && "No PP-cond entry found for '#'");
236   assert(TableIdx && "No jumping from #endifs.");
237 
238   // Update our side-table iterator.
239   const unsigned char* NextPPCondPtr = PPCond + TableIdx*(sizeof(uint32_t)*2);
240   assert(NextPPCondPtr >= CurPPCondPtr);
241   CurPPCondPtr = NextPPCondPtr;
242 
243   // Read where we should jump to.
244   HashEntryI = TokBuf + ReadLE32(NextPPCondPtr);
245   uint32_t NextIdx = ReadLE32(NextPPCondPtr);
246 
247   // By construction NextIdx will be zero if this is a #endif.  This is useful
248   // to know to obviate lexing another token.
249   bool isEndif = NextIdx == 0;
250 
251   // This case can occur when we see something like this:
252   //
253   //  #if ...
254   //   /* a comment or nothing */
255   //  #elif
256   //
257   // If we are skipping the first #if block it will be the case that CurPtr
258   // already points 'elif'.  Just return.
259 
260   if (CurPtr > HashEntryI) {
261     assert(CurPtr == HashEntryI + DISK_TOKEN_SIZE);
262     // Did we reach a #endif?  If so, go ahead and consume that token as well.
263     if (isEndif)
264       CurPtr += DISK_TOKEN_SIZE*2;
265     else
266       LastHashTokPtr = HashEntryI;
267 
268     return isEndif;
269   }
270 
271   // Otherwise, we need to advance.  Update CurPtr to point to the '#' token.
272   CurPtr = HashEntryI;
273 
274   // Update the location of the last observed '#'.  This is useful if we
275   // are skipping multiple blocks.
276   LastHashTokPtr = CurPtr;
277 
278   // Skip the '#' token.
279   assert(((tok::TokenKind)*CurPtr) == tok::hash);
280   CurPtr += DISK_TOKEN_SIZE;
281 
282   // Did we reach a #endif?  If so, go ahead and consume that token as well.
283   if (isEndif) { CurPtr += DISK_TOKEN_SIZE*2; }
284 
285   return isEndif;
286 }
287 
getSourceLocation()288 SourceLocation PTHLexer::getSourceLocation() {
289   // getSourceLocation is not on the hot path.  It is used to get the location
290   // of the next token when transitioning back to this lexer when done
291   // handling a #included file.  Just read the necessary data from the token
292   // data buffer to construct the SourceLocation object.
293   // NOTE: This is a virtual function; hence it is defined out-of-line.
294   const unsigned char *OffsetPtr = CurPtr + (DISK_TOKEN_SIZE - 4);
295   uint32_t Offset = ReadLE32(OffsetPtr);
296   return FileStartLoc.getLocWithOffset(Offset);
297 }
298 
299 //===----------------------------------------------------------------------===//
300 // PTH file lookup: map from strings to file data.
301 //===----------------------------------------------------------------------===//
302 
303 /// PTHFileLookup - This internal data structure is used by the PTHManager
304 ///  to map from FileEntry objects managed by FileManager to offsets within
305 ///  the PTH file.
306 namespace {
307 class PTHFileData {
308   const uint32_t TokenOff;
309   const uint32_t PPCondOff;
310 public:
PTHFileData(uint32_t tokenOff,uint32_t ppCondOff)311   PTHFileData(uint32_t tokenOff, uint32_t ppCondOff)
312     : TokenOff(tokenOff), PPCondOff(ppCondOff) {}
313 
getTokenOffset() const314   uint32_t getTokenOffset() const { return TokenOff; }
getPPCondOffset() const315   uint32_t getPPCondOffset() const { return PPCondOff; }
316 };
317 
318 
319 class PTHFileLookupCommonTrait {
320 public:
321   typedef std::pair<unsigned char, const char*> internal_key_type;
322 
ComputeHash(internal_key_type x)323   static unsigned ComputeHash(internal_key_type x) {
324     return llvm::HashString(x.second);
325   }
326 
327   static std::pair<unsigned, unsigned>
ReadKeyDataLength(const unsigned char * & d)328   ReadKeyDataLength(const unsigned char*& d) {
329     unsigned keyLen = (unsigned) ReadUnalignedLE16(d);
330     unsigned dataLen = (unsigned) *(d++);
331     return std::make_pair(keyLen, dataLen);
332   }
333 
ReadKey(const unsigned char * d,unsigned)334   static internal_key_type ReadKey(const unsigned char* d, unsigned) {
335     unsigned char k = *(d++); // Read the entry kind.
336     return std::make_pair(k, (const char*) d);
337   }
338 };
339 
340 class PTHFileLookupTrait : public PTHFileLookupCommonTrait {
341 public:
342   typedef const FileEntry* external_key_type;
343   typedef PTHFileData      data_type;
344 
GetInternalKey(const FileEntry * FE)345   static internal_key_type GetInternalKey(const FileEntry* FE) {
346     return std::make_pair((unsigned char) 0x1, FE->getName());
347   }
348 
EqualKey(internal_key_type a,internal_key_type b)349   static bool EqualKey(internal_key_type a, internal_key_type b) {
350     return a.first == b.first && strcmp(a.second, b.second) == 0;
351   }
352 
ReadData(const internal_key_type & k,const unsigned char * d,unsigned)353   static PTHFileData ReadData(const internal_key_type& k,
354                               const unsigned char* d, unsigned) {
355     assert(k.first == 0x1 && "Only file lookups can match!");
356     uint32_t x = ::ReadUnalignedLE32(d);
357     uint32_t y = ::ReadUnalignedLE32(d);
358     return PTHFileData(x, y);
359   }
360 };
361 
362 class PTHStringLookupTrait {
363 public:
364   typedef uint32_t
365           data_type;
366 
367   typedef const std::pair<const char*, unsigned>
368           external_key_type;
369 
370   typedef external_key_type internal_key_type;
371 
EqualKey(const internal_key_type & a,const internal_key_type & b)372   static bool EqualKey(const internal_key_type& a,
373                        const internal_key_type& b) {
374     return (a.second == b.second) ? memcmp(a.first, b.first, a.second) == 0
375                                   : false;
376   }
377 
ComputeHash(const internal_key_type & a)378   static unsigned ComputeHash(const internal_key_type& a) {
379     return llvm::HashString(StringRef(a.first, a.second));
380   }
381 
382   // This hopefully will just get inlined and removed by the optimizer.
383   static const internal_key_type&
GetInternalKey(const external_key_type & x)384   GetInternalKey(const external_key_type& x) { return x; }
385 
386   static std::pair<unsigned, unsigned>
ReadKeyDataLength(const unsigned char * & d)387   ReadKeyDataLength(const unsigned char*& d) {
388     return std::make_pair((unsigned) ReadUnalignedLE16(d), sizeof(uint32_t));
389   }
390 
391   static std::pair<const char*, unsigned>
ReadKey(const unsigned char * d,unsigned n)392   ReadKey(const unsigned char* d, unsigned n) {
393       assert(n >= 2 && d[n-1] == '\0');
394       return std::make_pair((const char*) d, n-1);
395     }
396 
ReadData(const internal_key_type & k,const unsigned char * d,unsigned)397   static uint32_t ReadData(const internal_key_type& k, const unsigned char* d,
398                            unsigned) {
399     return ::ReadUnalignedLE32(d);
400   }
401 };
402 
403 } // end anonymous namespace
404 
405 typedef OnDiskChainedHashTable<PTHFileLookupTrait>   PTHFileLookup;
406 typedef OnDiskChainedHashTable<PTHStringLookupTrait> PTHStringIdLookup;
407 
408 //===----------------------------------------------------------------------===//
409 // PTHManager methods.
410 //===----------------------------------------------------------------------===//
411 
PTHManager(const llvm::MemoryBuffer * buf,void * fileLookup,const unsigned char * idDataTable,IdentifierInfo ** perIDCache,void * stringIdLookup,unsigned numIds,const unsigned char * spellingBase,const char * originalSourceFile)412 PTHManager::PTHManager(const llvm::MemoryBuffer* buf, void* fileLookup,
413                        const unsigned char* idDataTable,
414                        IdentifierInfo** perIDCache,
415                        void* stringIdLookup, unsigned numIds,
416                        const unsigned char* spellingBase,
417                        const char* originalSourceFile)
418 : Buf(buf), PerIDCache(perIDCache), FileLookup(fileLookup),
419   IdDataTable(idDataTable), StringIdLookup(stringIdLookup),
420   NumIds(numIds), PP(0), SpellingBase(spellingBase),
421   OriginalSourceFile(originalSourceFile) {}
422 
~PTHManager()423 PTHManager::~PTHManager() {
424   delete Buf;
425   delete (PTHFileLookup*) FileLookup;
426   delete (PTHStringIdLookup*) StringIdLookup;
427   free(PerIDCache);
428 }
429 
InvalidPTH(DiagnosticsEngine & Diags,const char * Msg)430 static void InvalidPTH(DiagnosticsEngine &Diags, const char *Msg) {
431   Diags.Report(Diags.getCustomDiagID(DiagnosticsEngine::Error, Msg));
432 }
433 
Create(const std::string & file,DiagnosticsEngine & Diags)434 PTHManager *PTHManager::Create(const std::string &file,
435                                DiagnosticsEngine &Diags) {
436   // Memory map the PTH file.
437   OwningPtr<llvm::MemoryBuffer> File;
438 
439   if (llvm::MemoryBuffer::getFile(file, File)) {
440     // FIXME: Add ec.message() to this diag.
441     Diags.Report(diag::err_invalid_pth_file) << file;
442     return 0;
443   }
444 
445   // Get the buffer ranges and check if there are at least three 32-bit
446   // words at the end of the file.
447   const unsigned char *BufBeg = (const unsigned char*)File->getBufferStart();
448   const unsigned char *BufEnd = (const unsigned char*)File->getBufferEnd();
449 
450   // Check the prologue of the file.
451   if ((BufEnd - BufBeg) < (signed)(sizeof("cfe-pth") + 4 + 4) ||
452       memcmp(BufBeg, "cfe-pth", sizeof("cfe-pth")) != 0) {
453     Diags.Report(diag::err_invalid_pth_file) << file;
454     return 0;
455   }
456 
457   // Read the PTH version.
458   const unsigned char *p = BufBeg + (sizeof("cfe-pth"));
459   unsigned Version = ReadLE32(p);
460 
461   if (Version < PTHManager::Version) {
462     InvalidPTH(Diags,
463         Version < PTHManager::Version
464         ? "PTH file uses an older PTH format that is no longer supported"
465         : "PTH file uses a newer PTH format that cannot be read");
466     return 0;
467   }
468 
469   // Compute the address of the index table at the end of the PTH file.
470   const unsigned char *PrologueOffset = p;
471 
472   if (PrologueOffset >= BufEnd) {
473     Diags.Report(diag::err_invalid_pth_file) << file;
474     return 0;
475   }
476 
477   // Construct the file lookup table.  This will be used for mapping from
478   // FileEntry*'s to cached tokens.
479   const unsigned char* FileTableOffset = PrologueOffset + sizeof(uint32_t)*2;
480   const unsigned char* FileTable = BufBeg + ReadLE32(FileTableOffset);
481 
482   if (!(FileTable > BufBeg && FileTable < BufEnd)) {
483     Diags.Report(diag::err_invalid_pth_file) << file;
484     return 0; // FIXME: Proper error diagnostic?
485   }
486 
487   OwningPtr<PTHFileLookup> FL(PTHFileLookup::Create(FileTable, BufBeg));
488 
489   // Warn if the PTH file is empty.  We still want to create a PTHManager
490   // as the PTH could be used with -include-pth.
491   if (FL->isEmpty())
492     InvalidPTH(Diags, "PTH file contains no cached source data");
493 
494   // Get the location of the table mapping from persistent ids to the
495   // data needed to reconstruct identifiers.
496   const unsigned char* IDTableOffset = PrologueOffset + sizeof(uint32_t)*0;
497   const unsigned char* IData = BufBeg + ReadLE32(IDTableOffset);
498 
499   if (!(IData >= BufBeg && IData < BufEnd)) {
500     Diags.Report(diag::err_invalid_pth_file) << file;
501     return 0;
502   }
503 
504   // Get the location of the hashtable mapping between strings and
505   // persistent IDs.
506   const unsigned char* StringIdTableOffset = PrologueOffset + sizeof(uint32_t)*1;
507   const unsigned char* StringIdTable = BufBeg + ReadLE32(StringIdTableOffset);
508   if (!(StringIdTable >= BufBeg && StringIdTable < BufEnd)) {
509     Diags.Report(diag::err_invalid_pth_file) << file;
510     return 0;
511   }
512 
513   OwningPtr<PTHStringIdLookup> SL(PTHStringIdLookup::Create(StringIdTable,
514                                                                   BufBeg));
515 
516   // Get the location of the spelling cache.
517   const unsigned char* spellingBaseOffset = PrologueOffset + sizeof(uint32_t)*3;
518   const unsigned char* spellingBase = BufBeg + ReadLE32(spellingBaseOffset);
519   if (!(spellingBase >= BufBeg && spellingBase < BufEnd)) {
520     Diags.Report(diag::err_invalid_pth_file) << file;
521     return 0;
522   }
523 
524   // Get the number of IdentifierInfos and pre-allocate the identifier cache.
525   uint32_t NumIds = ReadLE32(IData);
526 
527   // Pre-allocate the persistent ID -> IdentifierInfo* cache.  We use calloc()
528   // so that we in the best case only zero out memory once when the OS returns
529   // us new pages.
530   IdentifierInfo** PerIDCache = 0;
531 
532   if (NumIds) {
533     PerIDCache = (IdentifierInfo**)calloc(NumIds, sizeof(*PerIDCache));
534     if (!PerIDCache) {
535       InvalidPTH(Diags, "Could not allocate memory for processing PTH file");
536       return 0;
537     }
538   }
539 
540   // Compute the address of the original source file.
541   const unsigned char* originalSourceBase = PrologueOffset + sizeof(uint32_t)*4;
542   unsigned len = ReadUnalignedLE16(originalSourceBase);
543   if (!len) originalSourceBase = 0;
544 
545   // Create the new PTHManager.
546   return new PTHManager(File.take(), FL.take(), IData, PerIDCache,
547                         SL.take(), NumIds, spellingBase,
548                         (const char*) originalSourceBase);
549 }
550 
LazilyCreateIdentifierInfo(unsigned PersistentID)551 IdentifierInfo* PTHManager::LazilyCreateIdentifierInfo(unsigned PersistentID) {
552   // Look in the PTH file for the string data for the IdentifierInfo object.
553   const unsigned char* TableEntry = IdDataTable + sizeof(uint32_t)*PersistentID;
554   const unsigned char* IDData =
555     (const unsigned char*)Buf->getBufferStart() + ReadLE32(TableEntry);
556   assert(IDData < (const unsigned char*)Buf->getBufferEnd());
557 
558   // Allocate the object.
559   std::pair<IdentifierInfo,const unsigned char*> *Mem =
560     Alloc.Allocate<std::pair<IdentifierInfo,const unsigned char*> >();
561 
562   Mem->second = IDData;
563   assert(IDData[0] != '\0');
564   IdentifierInfo *II = new ((void*) Mem) IdentifierInfo();
565 
566   // Store the new IdentifierInfo in the cache.
567   PerIDCache[PersistentID] = II;
568   assert(II->getNameStart() && II->getNameStart()[0] != '\0');
569   return II;
570 }
571 
get(StringRef Name)572 IdentifierInfo* PTHManager::get(StringRef Name) {
573   PTHStringIdLookup& SL = *((PTHStringIdLookup*)StringIdLookup);
574   // Double check our assumption that the last character isn't '\0'.
575   assert(Name.empty() || Name.back() != '\0');
576   PTHStringIdLookup::iterator I = SL.find(std::make_pair(Name.data(),
577                                                          Name.size()));
578   if (I == SL.end()) // No identifier found?
579     return 0;
580 
581   // Match found.  Return the identifier!
582   assert(*I > 0);
583   return GetIdentifierInfo(*I-1);
584 }
585 
CreateLexer(FileID FID)586 PTHLexer *PTHManager::CreateLexer(FileID FID) {
587   const FileEntry *FE = PP->getSourceManager().getFileEntryForID(FID);
588   if (!FE)
589     return 0;
590 
591   // Lookup the FileEntry object in our file lookup data structure.  It will
592   // return a variant that indicates whether or not there is an offset within
593   // the PTH file that contains cached tokens.
594   PTHFileLookup& PFL = *((PTHFileLookup*)FileLookup);
595   PTHFileLookup::iterator I = PFL.find(FE);
596 
597   if (I == PFL.end()) // No tokens available?
598     return 0;
599 
600   const PTHFileData& FileData = *I;
601 
602   const unsigned char *BufStart = (const unsigned char *)Buf->getBufferStart();
603   // Compute the offset of the token data within the buffer.
604   const unsigned char* data = BufStart + FileData.getTokenOffset();
605 
606   // Get the location of pp-conditional table.
607   const unsigned char* ppcond = BufStart + FileData.getPPCondOffset();
608   uint32_t Len = ReadLE32(ppcond);
609   if (Len == 0) ppcond = 0;
610 
611   assert(PP && "No preprocessor set yet!");
612   return new PTHLexer(*PP, FID, data, ppcond, *this);
613 }
614 
615 //===----------------------------------------------------------------------===//
616 // 'stat' caching.
617 //===----------------------------------------------------------------------===//
618 
619 namespace {
620 class PTHStatData {
621 public:
622   const bool HasData;
623   uint64_t Size;
624   time_t ModTime;
625   llvm::sys::fs::UniqueID UniqueID;
626   bool IsDirectory;
627 
PTHStatData(uint64_t Size,time_t ModTime,llvm::sys::fs::UniqueID UniqueID,bool IsDirectory)628   PTHStatData(uint64_t Size, time_t ModTime, llvm::sys::fs::UniqueID UniqueID,
629               bool IsDirectory)
630       : HasData(true), Size(Size), ModTime(ModTime), UniqueID(UniqueID),
631         IsDirectory(IsDirectory) {}
632 
PTHStatData()633   PTHStatData() : HasData(false) {}
634 };
635 
636 class PTHStatLookupTrait : public PTHFileLookupCommonTrait {
637 public:
638   typedef const char* external_key_type;  // const char*
639   typedef PTHStatData data_type;
640 
GetInternalKey(const char * path)641   static internal_key_type GetInternalKey(const char *path) {
642     // The key 'kind' doesn't matter here because it is ignored in EqualKey.
643     return std::make_pair((unsigned char) 0x0, path);
644   }
645 
EqualKey(internal_key_type a,internal_key_type b)646   static bool EqualKey(internal_key_type a, internal_key_type b) {
647     // When doing 'stat' lookups we don't care about the kind of 'a' and 'b',
648     // just the paths.
649     return strcmp(a.second, b.second) == 0;
650   }
651 
ReadData(const internal_key_type & k,const unsigned char * d,unsigned)652   static data_type ReadData(const internal_key_type& k, const unsigned char* d,
653                             unsigned) {
654 
655     if (k.first /* File or Directory */) {
656       bool IsDirectory = true;
657       if (k.first == 0x1 /* File */) {
658         IsDirectory = false;
659         d += 4 * 2; // Skip the first 2 words.
660       }
661 
662       uint64_t File = ReadUnalignedLE64(d);
663       uint64_t Device = ReadUnalignedLE64(d);
664       llvm::sys::fs::UniqueID UniqueID(File, Device);
665       time_t ModTime = ReadUnalignedLE64(d);
666       uint64_t Size = ReadUnalignedLE64(d);
667       return data_type(Size, ModTime, UniqueID, IsDirectory);
668     }
669 
670     // Negative stat.  Don't read anything.
671     return data_type();
672   }
673 };
674 
675 class PTHStatCache : public FileSystemStatCache {
676   typedef OnDiskChainedHashTable<PTHStatLookupTrait> CacheTy;
677   CacheTy Cache;
678 
679 public:
PTHStatCache(PTHFileLookup & FL)680   PTHStatCache(PTHFileLookup &FL) :
681     Cache(FL.getNumBuckets(), FL.getNumEntries(), FL.getBuckets(),
682           FL.getBase()) {}
683 
~PTHStatCache()684   ~PTHStatCache() {}
685 
getStat(const char * Path,FileData & Data,bool isFile,int * FileDescriptor)686   LookupResult getStat(const char *Path, FileData &Data, bool isFile,
687                        int *FileDescriptor) {
688     // Do the lookup for the file's data in the PTH file.
689     CacheTy::iterator I = Cache.find(Path);
690 
691     // If we don't get a hit in the PTH file just forward to 'stat'.
692     if (I == Cache.end())
693       return statChained(Path, Data, isFile, FileDescriptor);
694 
695     const PTHStatData &D = *I;
696 
697     if (!D.HasData)
698       return CacheMissing;
699 
700     Data.Size = D.Size;
701     Data.ModTime = D.ModTime;
702     Data.UniqueID = D.UniqueID;
703     Data.IsDirectory = D.IsDirectory;
704     Data.IsNamedPipe = false;
705     Data.InPCH = true;
706 
707     return CacheExists;
708   }
709 };
710 } // end anonymous namespace
711 
createStatCache()712 FileSystemStatCache *PTHManager::createStatCache() {
713   return new PTHStatCache(*((PTHFileLookup*) FileLookup));
714 }
715