PPLexerChange.cpp revision 39d9841ed4c0568d4b44dfbc12ac04491f60a374
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/SourceManager.h" 20#include "llvm/Support/MemoryBuffer.h" 21using namespace clang; 22 23PPCallbacks::~PPCallbacks() {} 24 25//===----------------------------------------------------------------------===// 26// Miscellaneous Methods. 27//===----------------------------------------------------------------------===// 28 29/// isInPrimaryFile - Return true if we're in the top-level file, not in a 30/// #include. This looks through macro expansions and active _Pragma lexers. 31bool Preprocessor::isInPrimaryFile() const { 32 if (IsFileLexer()) 33 return IncludeMacroStack.empty(); 34 35 // If there are any stacked lexers, we're in a #include. 36 assert(IsFileLexer(IncludeMacroStack[0]) && 37 "Top level include stack isn't our primary lexer?"); 38 for (unsigned i = 1, e = IncludeMacroStack.size(); i != e; ++i) 39 if (IsFileLexer(IncludeMacroStack[i])) 40 return false; 41 return true; 42} 43 44/// getCurrentLexer - Return the current file lexer being lexed from. Note 45/// that this ignores any potentially active macro expansions and _Pragma 46/// expansions going on at the time. 47PreprocessorLexer *Preprocessor::getCurrentFileLexer() const { 48 if (IsFileLexer()) 49 return CurPPLexer; 50 51 // Look for a stacked lexer. 52 for (unsigned i = IncludeMacroStack.size(); i != 0; --i) { 53 const IncludeStackInfo& ISI = IncludeMacroStack[i-1]; 54 if (IsFileLexer(ISI)) 55 return ISI.ThePPLexer; 56 } 57 return 0; 58} 59 60 61//===----------------------------------------------------------------------===// 62// Methods for Entering and Callbacks for leaving various contexts 63//===----------------------------------------------------------------------===// 64 65/// EnterSourceFile - Add a source file to the top of the include stack and 66/// start lexing tokens from it instead of the current buffer. 67bool Preprocessor::EnterSourceFile(FileID FID, const DirectoryLookup *CurDir, 68 std::string *ErrorStr) { 69 assert(CurTokenLexer == 0 && "Cannot #include a file inside a macro!"); 70 ++NumEnteredSourceFiles; 71 72 if (MaxIncludeStackDepth < IncludeMacroStack.size()) 73 MaxIncludeStackDepth = IncludeMacroStack.size(); 74 75 if (PTH) { 76 if (PTHLexer *PL = PTH->CreateLexer(FID)) { 77 EnterSourceFileWithPTH(PL, CurDir); 78 return false; 79 } 80 } 81 82 // Get the MemoryBuffer for this FID, if it fails, we fail. 83 const llvm::MemoryBuffer *InputFile = 84 getSourceManager().getBuffer(FID, ErrorStr); 85 if (InputFile == 0) 86 return true; 87 88 EnterSourceFileWithLexer(new Lexer(FID, InputFile, *this), CurDir); 89 return false; 90} 91 92/// EnterSourceFileWithLexer - Add a source file to the top of the include stack 93/// and start lexing tokens from it instead of the current buffer. 94void Preprocessor::EnterSourceFileWithLexer(Lexer *TheLexer, 95 const DirectoryLookup *CurDir) { 96 97 // Add the current lexer to the include stack. 98 if (CurPPLexer || CurTokenLexer) 99 PushIncludeMacroStack(); 100 101 CurLexer.reset(TheLexer); 102 CurPPLexer = TheLexer; 103 CurDirLookup = CurDir; 104 105 // Notify the client, if desired, that we are in a new source file. 106 if (Callbacks && !CurLexer->Is_PragmaLexer) { 107 SrcMgr::CharacteristicKind FileType = 108 SourceMgr.getFileCharacteristic(CurLexer->getFileLoc()); 109 110 Callbacks->FileChanged(CurLexer->getFileLoc(), 111 PPCallbacks::EnterFile, FileType); 112 } 113} 114 115/// EnterSourceFileWithPTH - Add a source file to the top of the include stack 116/// and start getting tokens from it using the PTH cache. 117void Preprocessor::EnterSourceFileWithPTH(PTHLexer *PL, 118 const DirectoryLookup *CurDir) { 119 120 if (CurPPLexer || CurTokenLexer) 121 PushIncludeMacroStack(); 122 123 CurDirLookup = CurDir; 124 CurPTHLexer.reset(PL); 125 CurPPLexer = CurPTHLexer.get(); 126 127 // Notify the client, if desired, that we are in a new source file. 128 if (Callbacks) { 129 FileID FID = CurPPLexer->getFileID(); 130 SourceLocation EnterLoc = SourceMgr.getLocForStartOfFile(FID); 131 SrcMgr::CharacteristicKind FileType = 132 SourceMgr.getFileCharacteristic(EnterLoc); 133 Callbacks->FileChanged(EnterLoc, PPCallbacks::EnterFile, FileType); 134 } 135} 136 137/// EnterMacro - Add a Macro to the top of the include stack and start lexing 138/// tokens from it instead of the current buffer. 139void Preprocessor::EnterMacro(Token &Tok, SourceLocation ILEnd, 140 MacroArgs *Args) { 141 PushIncludeMacroStack(); 142 CurDirLookup = 0; 143 144 if (NumCachedTokenLexers == 0) { 145 CurTokenLexer.reset(new TokenLexer(Tok, ILEnd, Args, *this)); 146 } else { 147 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]); 148 CurTokenLexer->Init(Tok, ILEnd, Args); 149 } 150} 151 152/// EnterTokenStream - Add a "macro" context to the top of the include stack, 153/// which will cause the lexer to start returning the specified tokens. 154/// 155/// If DisableMacroExpansion is true, tokens lexed from the token stream will 156/// not be subject to further macro expansion. Otherwise, these tokens will 157/// be re-macro-expanded when/if expansion is enabled. 158/// 159/// If OwnsTokens is false, this method assumes that the specified stream of 160/// tokens has a permanent owner somewhere, so they do not need to be copied. 161/// If it is true, it assumes the array of tokens is allocated with new[] and 162/// must be freed. 163/// 164void Preprocessor::EnterTokenStream(const Token *Toks, unsigned NumToks, 165 bool DisableMacroExpansion, 166 bool OwnsTokens) { 167 // Save our current state. 168 PushIncludeMacroStack(); 169 CurDirLookup = 0; 170 171 // Create a macro expander to expand from the specified token stream. 172 if (NumCachedTokenLexers == 0) { 173 CurTokenLexer.reset(new TokenLexer(Toks, NumToks, DisableMacroExpansion, 174 OwnsTokens, *this)); 175 } else { 176 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]); 177 CurTokenLexer->Init(Toks, NumToks, DisableMacroExpansion, OwnsTokens); 178 } 179} 180 181/// HandleEndOfFile - This callback is invoked when the lexer hits the end of 182/// the current file. This either returns the EOF token or pops a level off 183/// the include stack and keeps going. 184bool Preprocessor::HandleEndOfFile(Token &Result, bool isEndOfMacro) { 185 assert(!CurTokenLexer && 186 "Ending a file when currently in a macro!"); 187 188 // See if this file had a controlling macro. 189 if (CurPPLexer) { // Not ending a macro, ignore it. 190 if (const IdentifierInfo *ControllingMacro = 191 CurPPLexer->MIOpt.GetControllingMacroAtEndOfFile()) { 192 // Okay, this has a controlling macro, remember in HeaderFileInfo. 193 if (const FileEntry *FE = 194 SourceMgr.getFileEntryForID(CurPPLexer->getFileID())) 195 HeaderInfo.SetFileControllingMacro(FE, ControllingMacro); 196 } 197 } 198 199 // If this is a #include'd file, pop it off the include stack and continue 200 // lexing the #includer file. 201 if (!IncludeMacroStack.empty()) { 202 // We're done with the #included file. 203 RemoveTopOfLexerStack(); 204 205 // Notify the client, if desired, that we are in a new source file. 206 if (Callbacks && !isEndOfMacro && CurPPLexer) { 207 SrcMgr::CharacteristicKind FileType = 208 SourceMgr.getFileCharacteristic(CurPPLexer->getSourceLocation()); 209 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), 210 PPCallbacks::ExitFile, FileType); 211 } 212 213 // Client should lex another token. 214 return false; 215 } 216 217 // If the file ends with a newline, form the EOF token on the newline itself, 218 // rather than "on the line following it", which doesn't exist. This makes 219 // diagnostics relating to the end of file include the last file that the user 220 // actually typed, which is goodness. 221 if (CurLexer) { 222 const char *EndPos = CurLexer->BufferEnd; 223 if (EndPos != CurLexer->BufferStart && 224 (EndPos[-1] == '\n' || EndPos[-1] == '\r')) { 225 --EndPos; 226 227 // Handle \n\r and \r\n: 228 if (EndPos != CurLexer->BufferStart && 229 (EndPos[-1] == '\n' || EndPos[-1] == '\r') && 230 EndPos[-1] != EndPos[0]) 231 --EndPos; 232 } 233 234 Result.startToken(); 235 CurLexer->BufferPtr = EndPos; 236 CurLexer->FormTokenWithChars(Result, EndPos, tok::eof); 237 238 // We're done with the #included file. 239 CurLexer.reset(); 240 } else { 241 assert(CurPTHLexer && "Got EOF but no current lexer set!"); 242 CurPTHLexer->getEOF(Result); 243 CurPTHLexer.reset(); 244 } 245 246 CurPPLexer = 0; 247 248 // This is the end of the top-level file. If the diag::pp_macro_not_used 249 // diagnostic is enabled, look for macros that have not been used. 250 if (getDiagnostics().getDiagnosticLevel(diag::pp_macro_not_used) != 251 Diagnostic::Ignored) { 252 for (macro_iterator I = macro_begin(), E = macro_end(); I != E; ++I) 253 if (!I->second->isUsed()) 254 Diag(I->second->getDefinitionLoc(), diag::pp_macro_not_used); 255 } 256 return true; 257} 258 259/// HandleEndOfTokenLexer - This callback is invoked when the current TokenLexer 260/// hits the end of its token stream. 261bool Preprocessor::HandleEndOfTokenLexer(Token &Result) { 262 assert(CurTokenLexer && !CurPPLexer && 263 "Ending a macro when currently in a #include file!"); 264 265 // Delete or cache the now-dead macro expander. 266 if (NumCachedTokenLexers == TokenLexerCacheSize) 267 CurTokenLexer.reset(); 268 else 269 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take(); 270 271 // Handle this like a #include file being popped off the stack. 272 return HandleEndOfFile(Result, true); 273} 274 275/// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the 276/// lexer stack. This should only be used in situations where the current 277/// state of the top-of-stack lexer is unknown. 278void Preprocessor::RemoveTopOfLexerStack() { 279 assert(!IncludeMacroStack.empty() && "Ran out of stack entries to load"); 280 281 if (CurTokenLexer) { 282 // Delete or cache the now-dead macro expander. 283 if (NumCachedTokenLexers == TokenLexerCacheSize) 284 CurTokenLexer.reset(); 285 else 286 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take(); 287 } 288 289 PopIncludeMacroStack(); 290} 291 292/// HandleMicrosoftCommentPaste - When the macro expander pastes together a 293/// comment (/##/) in microsoft mode, this method handles updating the current 294/// state, returning the token on the next source line. 295void Preprocessor::HandleMicrosoftCommentPaste(Token &Tok) { 296 assert(CurTokenLexer && !CurPPLexer && 297 "Pasted comment can only be formed from macro"); 298 299 // We handle this by scanning for the closest real lexer, switching it to 300 // raw mode and preprocessor mode. This will cause it to return \n as an 301 // explicit EOM token. 302 PreprocessorLexer *FoundLexer = 0; 303 bool LexerWasInPPMode = false; 304 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 305 IncludeStackInfo &ISI = *(IncludeMacroStack.end()-i-1); 306 if (ISI.ThePPLexer == 0) continue; // Scan for a real lexer. 307 308 // Once we find a real lexer, mark it as raw mode (disabling macro 309 // expansions) and preprocessor mode (return EOM). We know that the lexer 310 // was *not* in raw mode before, because the macro that the comment came 311 // from was expanded. However, it could have already been in preprocessor 312 // mode (#if COMMENT) in which case we have to return it to that mode and 313 // return EOM. 314 FoundLexer = ISI.ThePPLexer; 315 FoundLexer->LexingRawMode = true; 316 LexerWasInPPMode = FoundLexer->ParsingPreprocessorDirective; 317 FoundLexer->ParsingPreprocessorDirective = true; 318 break; 319 } 320 321 // Okay, we either found and switched over the lexer, or we didn't find a 322 // lexer. In either case, finish off the macro the comment came from, getting 323 // the next token. 324 if (!HandleEndOfTokenLexer(Tok)) Lex(Tok); 325 326 // Discarding comments as long as we don't have EOF or EOM. This 'comments 327 // out' the rest of the line, including any tokens that came from other macros 328 // that were active, as in: 329 // #define submacro a COMMENT b 330 // submacro c 331 // which should lex to 'a' only: 'b' and 'c' should be removed. 332 while (Tok.isNot(tok::eom) && Tok.isNot(tok::eof)) 333 Lex(Tok); 334 335 // If we got an eom token, then we successfully found the end of the line. 336 if (Tok.is(tok::eom)) { 337 assert(FoundLexer && "Can't get end of line without an active lexer"); 338 // Restore the lexer back to normal mode instead of raw mode. 339 FoundLexer->LexingRawMode = false; 340 341 // If the lexer was already in preprocessor mode, just return the EOM token 342 // to finish the preprocessor line. 343 if (LexerWasInPPMode) return; 344 345 // Otherwise, switch out of PP mode and return the next lexed token. 346 FoundLexer->ParsingPreprocessorDirective = false; 347 return Lex(Tok); 348 } 349 350 // If we got an EOF token, then we reached the end of the token stream but 351 // didn't find an explicit \n. This can only happen if there was no lexer 352 // active (an active lexer would return EOM at EOF if there was no \n in 353 // preprocessor directive mode), so just return EOF as our token. 354 assert(!FoundLexer && "Lexer should return EOM before EOF in PP mode"); 355} 356