PPLexerChange.cpp revision 6e2901407bff59aeb4cc301cc58b034723d0eb49
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 assert(CurTokenLexer == 0 && "Cannot #include a file inside a macro!"); 69 ++NumEnteredSourceFiles; 70 71 if (MaxIncludeStackDepth < IncludeMacroStack.size()) 72 MaxIncludeStackDepth = IncludeMacroStack.size(); 73 74 if (PTH) { 75 if (PTHLexer *PL = PTH->CreateLexer(FID)) { 76 EnterSourceFileWithPTH(PL, CurDir); 77 return false; 78 } 79 } 80 81 // Get the MemoryBuffer for this FID, if it fails, we fail. 82 const llvm::MemoryBuffer *InputFile = getSourceManager().getBuffer(FID); 83 if (InputFile == 0) 84 return true; 85 86 EnterSourceFileWithLexer(new Lexer(FID, InputFile, *this), CurDir); 87 return false; 88} 89 90/// EnterSourceFileWithLexer - Add a source file to the top of the include stack 91/// and start lexing tokens from it instead of the current buffer. 92void Preprocessor::EnterSourceFileWithLexer(Lexer *TheLexer, 93 const DirectoryLookup *CurDir) { 94 95 // Add the current lexer to the include stack. 96 if (CurPPLexer || CurTokenLexer) 97 PushIncludeMacroStack(); 98 99 CurLexer.reset(TheLexer); 100 CurPPLexer = TheLexer; 101 CurDirLookup = CurDir; 102 103 // Notify the client, if desired, that we are in a new source file. 104 if (Callbacks && !CurLexer->Is_PragmaLexer) { 105 SrcMgr::CharacteristicKind FileType = 106 SourceMgr.getFileCharacteristic(CurLexer->getFileLoc()); 107 108 Callbacks->FileChanged(CurLexer->getFileLoc(), 109 PPCallbacks::EnterFile, FileType); 110 } 111} 112 113/// EnterSourceFileWithPTH - Add a source file to the top of the include stack 114/// and start getting tokens from it using the PTH cache. 115void Preprocessor::EnterSourceFileWithPTH(PTHLexer *PL, 116 const DirectoryLookup *CurDir) { 117 118 if (CurPPLexer || CurTokenLexer) 119 PushIncludeMacroStack(); 120 121 CurDirLookup = CurDir; 122 CurPTHLexer.reset(PL); 123 CurPPLexer = CurPTHLexer.get(); 124 125 // Notify the client, if desired, that we are in a new source file. 126 if (Callbacks) { 127 FileID FID = CurPPLexer->getFileID(); 128 SourceLocation EnterLoc = SourceMgr.getLocForStartOfFile(FID); 129 SrcMgr::CharacteristicKind FileType = 130 SourceMgr.getFileCharacteristic(EnterLoc); 131 Callbacks->FileChanged(EnterLoc, PPCallbacks::EnterFile, FileType); 132 } 133} 134 135/// EnterMacro - Add a Macro to the top of the include stack and start lexing 136/// tokens from it instead of the current buffer. 137void Preprocessor::EnterMacro(Token &Tok, SourceLocation ILEnd, 138 MacroArgs *Args) { 139 PushIncludeMacroStack(); 140 CurDirLookup = 0; 141 142 if (NumCachedTokenLexers == 0) { 143 CurTokenLexer.reset(new TokenLexer(Tok, ILEnd, Args, *this)); 144 } else { 145 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]); 146 CurTokenLexer->Init(Tok, ILEnd, Args); 147 } 148} 149 150/// EnterTokenStream - Add a "macro" context to the top of the include stack, 151/// which will cause the lexer to start returning the specified tokens. 152/// 153/// If DisableMacroExpansion is true, tokens lexed from the token stream will 154/// not be subject to further macro expansion. Otherwise, these tokens will 155/// be re-macro-expanded when/if expansion is enabled. 156/// 157/// If OwnsTokens is false, this method assumes that the specified stream of 158/// tokens has a permanent owner somewhere, so they do not need to be copied. 159/// If it is true, it assumes the array of tokens is allocated with new[] and 160/// must be freed. 161/// 162void Preprocessor::EnterTokenStream(const Token *Toks, unsigned NumToks, 163 bool DisableMacroExpansion, 164 bool OwnsTokens) { 165 // Save our current state. 166 PushIncludeMacroStack(); 167 CurDirLookup = 0; 168 169 // Create a macro expander to expand from the specified token stream. 170 if (NumCachedTokenLexers == 0) { 171 CurTokenLexer.reset(new TokenLexer(Toks, NumToks, DisableMacroExpansion, 172 OwnsTokens, *this)); 173 } else { 174 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]); 175 CurTokenLexer->Init(Toks, NumToks, DisableMacroExpansion, OwnsTokens); 176 } 177} 178 179/// HandleEndOfFile - This callback is invoked when the lexer hits the end of 180/// the current file. This either returns the EOF token or pops a level off 181/// the include stack and keeps going. 182bool Preprocessor::HandleEndOfFile(Token &Result, bool isEndOfMacro) { 183 assert(!CurTokenLexer && 184 "Ending a file when currently in a macro!"); 185 186 // See if this file had a controlling macro. 187 if (CurPPLexer) { // Not ending a macro, ignore it. 188 if (const IdentifierInfo *ControllingMacro = 189 CurPPLexer->MIOpt.GetControllingMacroAtEndOfFile()) { 190 // Okay, this has a controlling macro, remember in HeaderFileInfo. 191 if (const FileEntry *FE = 192 SourceMgr.getFileEntryForID(CurPPLexer->getFileID())) 193 HeaderInfo.SetFileControllingMacro(FE, ControllingMacro); 194 } 195 } 196 197 // If this is a #include'd file, pop it off the include stack and continue 198 // lexing the #includer file. 199 if (!IncludeMacroStack.empty()) { 200 // We're done with the #included file. 201 RemoveTopOfLexerStack(); 202 203 // Notify the client, if desired, that we are in a new source file. 204 if (Callbacks && !isEndOfMacro && CurPPLexer) { 205 SrcMgr::CharacteristicKind FileType = 206 SourceMgr.getFileCharacteristic(CurPPLexer->getSourceLocation()); 207 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), 208 PPCallbacks::ExitFile, FileType); 209 } 210 211 // Client should lex another token. 212 return false; 213 } 214 215 // If the file ends with a newline, form the EOF token on the newline itself, 216 // rather than "on the line following it", which doesn't exist. This makes 217 // diagnostics relating to the end of file include the last file that the user 218 // actually typed, which is goodness. 219 if (CurLexer) { 220 const char *EndPos = CurLexer->BufferEnd; 221 if (EndPos != CurLexer->BufferStart && 222 (EndPos[-1] == '\n' || EndPos[-1] == '\r')) { 223 --EndPos; 224 225 // Handle \n\r and \r\n: 226 if (EndPos != CurLexer->BufferStart && 227 (EndPos[-1] == '\n' || EndPos[-1] == '\r') && 228 EndPos[-1] != EndPos[0]) 229 --EndPos; 230 } 231 232 Result.startToken(); 233 CurLexer->BufferPtr = EndPos; 234 CurLexer->FormTokenWithChars(Result, EndPos, tok::eof); 235 236 // We're done with the #included file. 237 CurLexer.reset(); 238 } else { 239 assert(CurPTHLexer && "Got EOF but no current lexer set!"); 240 CurPTHLexer->getEOF(Result); 241 CurPTHLexer.reset(); 242 } 243 244 CurPPLexer = 0; 245 246 // This is the end of the top-level file. If the diag::pp_macro_not_used 247 // diagnostic is enabled, look for macros that have not been used. 248 if (getDiagnostics().getDiagnosticLevel(diag::pp_macro_not_used) != 249 Diagnostic::Ignored) { 250 for (macro_iterator I = macro_begin(), E = macro_end(); I != E; ++I) 251 if (!I->second->isUsed()) 252 Diag(I->second->getDefinitionLoc(), diag::pp_macro_not_used); 253 } 254 return true; 255} 256 257/// HandleEndOfTokenLexer - This callback is invoked when the current TokenLexer 258/// hits the end of its token stream. 259bool Preprocessor::HandleEndOfTokenLexer(Token &Result) { 260 assert(CurTokenLexer && !CurPPLexer && 261 "Ending a macro when currently in a #include file!"); 262 263 // Delete or cache the now-dead macro expander. 264 if (NumCachedTokenLexers == TokenLexerCacheSize) 265 CurTokenLexer.reset(); 266 else 267 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take(); 268 269 // Handle this like a #include file being popped off the stack. 270 return HandleEndOfFile(Result, true); 271} 272 273/// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the 274/// lexer stack. This should only be used in situations where the current 275/// state of the top-of-stack lexer is unknown. 276void Preprocessor::RemoveTopOfLexerStack() { 277 assert(!IncludeMacroStack.empty() && "Ran out of stack entries to load"); 278 279 if (CurTokenLexer) { 280 // Delete or cache the now-dead macro expander. 281 if (NumCachedTokenLexers == TokenLexerCacheSize) 282 CurTokenLexer.reset(); 283 else 284 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take(); 285 } 286 287 PopIncludeMacroStack(); 288} 289 290/// HandleMicrosoftCommentPaste - When the macro expander pastes together a 291/// comment (/##/) in microsoft mode, this method handles updating the current 292/// state, returning the token on the next source line. 293void Preprocessor::HandleMicrosoftCommentPaste(Token &Tok) { 294 assert(CurTokenLexer && !CurPPLexer && 295 "Pasted comment can only be formed from macro"); 296 297 // We handle this by scanning for the closest real lexer, switching it to 298 // raw mode and preprocessor mode. This will cause it to return \n as an 299 // explicit EOM token. 300 PreprocessorLexer *FoundLexer = 0; 301 bool LexerWasInPPMode = false; 302 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 303 IncludeStackInfo &ISI = *(IncludeMacroStack.end()-i-1); 304 if (ISI.ThePPLexer == 0) continue; // Scan for a real lexer. 305 306 // Once we find a real lexer, mark it as raw mode (disabling macro 307 // expansions) and preprocessor mode (return EOM). We know that the lexer 308 // was *not* in raw mode before, because the macro that the comment came 309 // from was expanded. However, it could have already been in preprocessor 310 // mode (#if COMMENT) in which case we have to return it to that mode and 311 // return EOM. 312 FoundLexer = ISI.ThePPLexer; 313 FoundLexer->LexingRawMode = true; 314 LexerWasInPPMode = FoundLexer->ParsingPreprocessorDirective; 315 FoundLexer->ParsingPreprocessorDirective = true; 316 break; 317 } 318 319 // Okay, we either found and switched over the lexer, or we didn't find a 320 // lexer. In either case, finish off the macro the comment came from, getting 321 // the next token. 322 if (!HandleEndOfTokenLexer(Tok)) Lex(Tok); 323 324 // Discarding comments as long as we don't have EOF or EOM. This 'comments 325 // out' the rest of the line, including any tokens that came from other macros 326 // that were active, as in: 327 // #define submacro a COMMENT b 328 // submacro c 329 // which should lex to 'a' only: 'b' and 'c' should be removed. 330 while (Tok.isNot(tok::eom) && Tok.isNot(tok::eof)) 331 Lex(Tok); 332 333 // If we got an eom token, then we successfully found the end of the line. 334 if (Tok.is(tok::eom)) { 335 assert(FoundLexer && "Can't get end of line without an active lexer"); 336 // Restore the lexer back to normal mode instead of raw mode. 337 FoundLexer->LexingRawMode = false; 338 339 // If the lexer was already in preprocessor mode, just return the EOM token 340 // to finish the preprocessor line. 341 if (LexerWasInPPMode) return; 342 343 // Otherwise, switch out of PP mode and return the next lexed token. 344 FoundLexer->ParsingPreprocessorDirective = false; 345 return Lex(Tok); 346 } 347 348 // If we got an EOF token, then we reached the end of the token stream but 349 // didn't find an explicit \n. This can only happen if there was no lexer 350 // active (an active lexer would return EOM at EOF if there was no \n in 351 // preprocessor directive mode), so just return EOF as our token. 352 assert(!FoundLexer && "Lexer should return EOM before EOF in PP mode"); 353} 354