1 //===--- PPLexerChange.cpp - Handle changing lexers in the preprocessor ---===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements pieces of the Preprocessor interface that manage the
11 // current lexer stack.
13 //===----------------------------------------------------------------------===//
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"
21 using namespace clang;
23 PPCallbacks::~PPCallbacks() {}
25 //===----------------------------------------------------------------------===//
26 // Miscellaneous Methods.
27 //===----------------------------------------------------------------------===//
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.
31 bool Preprocessor::isInPrimaryFile() const {
33 return IncludeMacroStack.empty();
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]))
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.
47 PreprocessorLexer *Preprocessor::getCurrentFileLexer() const {
51 // Look for a stacked lexer.
52 for (unsigned i = IncludeMacroStack.size(); i != 0; --i) {
53 const IncludeStackInfo& ISI = IncludeMacroStack[i-1];
55 return ISI.ThePPLexer;
61 //===----------------------------------------------------------------------===//
62 // Methods for Entering and Callbacks for leaving various contexts
63 //===----------------------------------------------------------------------===//
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.
67 bool Preprocessor::EnterSourceFile(FileID FID, const DirectoryLookup *CurDir,
68 std::string &ErrorStr) {
69 assert(CurTokenLexer == 0 && "Cannot #include a file inside a macro!");
70 ++NumEnteredSourceFiles;
72 if (MaxIncludeStackDepth < IncludeMacroStack.size())
73 MaxIncludeStackDepth = IncludeMacroStack.size();
76 if (PTHLexer *PL = PTH->CreateLexer(FID)) {
77 EnterSourceFileWithPTH(PL, CurDir);
82 // Get the MemoryBuffer for this FID, if it fails, we fail.
83 const llvm::MemoryBuffer *InputFile = getSourceManager().getBuffer(FID);
87 EnterSourceFileWithLexer(new Lexer(FID, InputFile, *this), CurDir);
91 /// EnterSourceFileWithLexer - Add a source file to the top of the include stack
92 /// and start lexing tokens from it instead of the current buffer.
93 void Preprocessor::EnterSourceFileWithLexer(Lexer *TheLexer,
94 const DirectoryLookup *CurDir) {
96 // Add the current lexer to the include stack.
97 if (CurPPLexer || CurTokenLexer)
98 PushIncludeMacroStack();
100 CurLexer.reset(TheLexer);
101 CurPPLexer = TheLexer;
102 CurDirLookup = CurDir;
104 // Notify the client, if desired, that we are in a new source file.
105 if (Callbacks && !CurLexer->Is_PragmaLexer) {
106 SrcMgr::CharacteristicKind FileType =
107 SourceMgr.getFileCharacteristic(CurLexer->getFileLoc());
109 Callbacks->FileChanged(CurLexer->getFileLoc(),
110 PPCallbacks::EnterFile, FileType);
114 /// EnterSourceFileWithPTH - Add a source file to the top of the include stack
115 /// and start getting tokens from it using the PTH cache.
116 void Preprocessor::EnterSourceFileWithPTH(PTHLexer *PL,
117 const DirectoryLookup *CurDir) {
119 if (CurPPLexer || CurTokenLexer)
120 PushIncludeMacroStack();
122 CurDirLookup = CurDir;
123 CurPTHLexer.reset(PL);
124 CurPPLexer = CurPTHLexer.get();
126 // Notify the client, if desired, that we are in a new source file.
128 FileID FID = CurPPLexer->getFileID();
129 SourceLocation EnterLoc = SourceMgr.getLocForStartOfFile(FID);
130 SrcMgr::CharacteristicKind FileType =
131 SourceMgr.getFileCharacteristic(EnterLoc);
132 Callbacks->FileChanged(EnterLoc, PPCallbacks::EnterFile, FileType);
136 /// EnterMacro - Add a Macro to the top of the include stack and start lexing
137 /// tokens from it instead of the current buffer.
138 void Preprocessor::EnterMacro(Token &Tok, SourceLocation ILEnd,
140 PushIncludeMacroStack();
143 if (NumCachedTokenLexers == 0) {
144 CurTokenLexer.reset(new TokenLexer(Tok, ILEnd, Args, *this));
146 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]);
147 CurTokenLexer->Init(Tok, ILEnd, Args);
151 /// EnterTokenStream - Add a "macro" context to the top of the include stack,
152 /// which will cause the lexer to start returning the specified tokens.
154 /// If DisableMacroExpansion is true, tokens lexed from the token stream will
155 /// not be subject to further macro expansion. Otherwise, these tokens will
156 /// be re-macro-expanded when/if expansion is enabled.
158 /// If OwnsTokens is false, this method assumes that the specified stream of
159 /// tokens has a permanent owner somewhere, so they do not need to be copied.
160 /// If it is true, it assumes the array of tokens is allocated with new[] and
163 void Preprocessor::EnterTokenStream(const Token *Toks, unsigned NumToks,
164 bool DisableMacroExpansion,
166 // Save our current state.
167 PushIncludeMacroStack();
170 // Create a macro expander to expand from the specified token stream.
171 if (NumCachedTokenLexers == 0) {
172 CurTokenLexer.reset(new TokenLexer(Toks, NumToks, DisableMacroExpansion,
175 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]);
176 CurTokenLexer->Init(Toks, NumToks, DisableMacroExpansion, OwnsTokens);
180 /// HandleEndOfFile - This callback is invoked when the lexer hits the end of
181 /// the current file. This either returns the EOF token or pops a level off
182 /// the include stack and keeps going.
183 bool Preprocessor::HandleEndOfFile(Token &Result, bool isEndOfMacro) {
184 assert(!CurTokenLexer &&
185 "Ending a file when currently in a macro!");
187 // See if this file had a controlling macro.
188 if (CurPPLexer) { // Not ending a macro, ignore it.
189 if (const IdentifierInfo *ControllingMacro =
190 CurPPLexer->MIOpt.GetControllingMacroAtEndOfFile()) {
191 // Okay, this has a controlling macro, remember in HeaderFileInfo.
192 if (const FileEntry *FE =
193 SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))
194 HeaderInfo.SetFileControllingMacro(FE, ControllingMacro);
198 // If this is a #include'd file, pop it off the include stack and continue
199 // lexing the #includer file.
200 if (!IncludeMacroStack.empty()) {
201 // We're done with the #included file.
202 RemoveTopOfLexerStack();
204 // Notify the client, if desired, that we are in a new source file.
205 if (Callbacks && !isEndOfMacro && CurPPLexer) {
206 SrcMgr::CharacteristicKind FileType =
207 SourceMgr.getFileCharacteristic(CurPPLexer->getSourceLocation());
208 Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
209 PPCallbacks::ExitFile, FileType);
212 // Client should lex another token.
216 // If the file ends with a newline, form the EOF token on the newline itself,
217 // rather than "on the line following it", which doesn't exist. This makes
218 // diagnostics relating to the end of file include the last file that the user
219 // actually typed, which is goodness.
221 const char *EndPos = CurLexer->BufferEnd;
222 if (EndPos != CurLexer->BufferStart &&
223 (EndPos[-1] == '\n' || EndPos[-1] == '\r')) {
226 // Handle \n\r and \r\n:
227 if (EndPos != CurLexer->BufferStart &&
228 (EndPos[-1] == '\n' || EndPos[-1] == '\r') &&
229 EndPos[-1] != EndPos[0])
234 CurLexer->BufferPtr = EndPos;
235 CurLexer->FormTokenWithChars(Result, EndPos, tok::eof);
237 // We're done with the #included file.
240 assert(CurPTHLexer && "Got EOF but no current lexer set!");
241 CurPTHLexer->getEOF(Result);
247 // This is the end of the top-level file. If the diag::pp_macro_not_used
248 // diagnostic is enabled, look for macros that have not been used.
249 if (getDiagnostics().getDiagnosticLevel(diag::pp_macro_not_used) !=
250 Diagnostic::Ignored) {
251 for (macro_iterator I = macro_begin(false), E = macro_end(false);
253 if (!I->second->isUsed())
254 Diag(I->second->getDefinitionLoc(), diag::pp_macro_not_used);
259 /// HandleEndOfTokenLexer - This callback is invoked when the current TokenLexer
260 /// hits the end of its token stream.
261 bool Preprocessor::HandleEndOfTokenLexer(Token &Result) {
262 assert(CurTokenLexer && !CurPPLexer &&
263 "Ending a macro when currently in a #include file!");
265 // Delete or cache the now-dead macro expander.
266 if (NumCachedTokenLexers == TokenLexerCacheSize)
267 CurTokenLexer.reset();
269 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take();
271 // Handle this like a #include file being popped off the stack.
272 return HandleEndOfFile(Result, true);
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.
278 void Preprocessor::RemoveTopOfLexerStack() {
279 assert(!IncludeMacroStack.empty() && "Ran out of stack entries to load");
282 // Delete or cache the now-dead macro expander.
283 if (NumCachedTokenLexers == TokenLexerCacheSize)
284 CurTokenLexer.reset();
286 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take();
289 PopIncludeMacroStack();
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.
295 void Preprocessor::HandleMicrosoftCommentPaste(Token &Tok) {
296 assert(CurTokenLexer && !CurPPLexer &&
297 "Pasted comment can only be formed from macro");
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.
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
314 FoundLexer = ISI.ThePPLexer;
315 FoundLexer->LexingRawMode = true;
316 LexerWasInPPMode = FoundLexer->ParsingPreprocessorDirective;
317 FoundLexer->ParsingPreprocessorDirective = true;
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
324 if (!HandleEndOfTokenLexer(Tok)) Lex(Tok);
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
331 // which should lex to 'a' only: 'b' and 'c' should be removed.
332 while (Tok.isNot(tok::eom) && Tok.isNot(tok::eof))
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;
341 // If the lexer was already in preprocessor mode, just return the EOM token
342 // to finish the preprocessor line.
343 if (LexerWasInPPMode) return;
345 // Otherwise, switch out of PP mode and return the next lexed token.
346 FoundLexer->ParsingPreprocessorDirective = false;
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");