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 void Preprocessor::EnterSourceFile(FileID FID, const DirectoryLookup *CurDir,
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.
84 const llvm::MemoryBuffer *InputFile =
85 getSourceManager().getBuffer(FID, Loc, &Invalid);
87 SourceLocation FileStart = SourceMgr.getLocForStartOfFile(FID);
88 Diag(Loc, diag::err_pp_error_opening_file)
89 << std::string(SourceMgr.getBufferName(FileStart)) << "";
93 if (isCodeCompletionEnabled() &&
94 SourceMgr.getFileEntryForID(FID) == CodeCompletionFile) {
95 CodeCompletionFileLoc = SourceMgr.getLocForStartOfFile(FID);
97 CodeCompletionFileLoc.getLocWithOffset(CodeCompletionOffset);
100 EnterSourceFileWithLexer(new Lexer(FID, InputFile, *this), CurDir);
104 /// EnterSourceFileWithLexer - Add a source file to the top of the include stack
105 /// and start lexing tokens from it instead of the current buffer.
106 void Preprocessor::EnterSourceFileWithLexer(Lexer *TheLexer,
107 const DirectoryLookup *CurDir) {
109 // Add the current lexer to the include stack.
110 if (CurPPLexer || CurTokenLexer)
111 PushIncludeMacroStack();
113 CurLexer.reset(TheLexer);
114 CurPPLexer = TheLexer;
115 CurDirLookup = CurDir;
116 if (CurLexerKind != CLK_LexAfterModuleImport)
117 CurLexerKind = CLK_Lexer;
119 // Notify the client, if desired, that we are in a new source file.
120 if (Callbacks && !CurLexer->Is_PragmaLexer) {
121 SrcMgr::CharacteristicKind FileType =
122 SourceMgr.getFileCharacteristic(CurLexer->getFileLoc());
124 Callbacks->FileChanged(CurLexer->getFileLoc(),
125 PPCallbacks::EnterFile, FileType);
129 /// EnterSourceFileWithPTH - Add a source file to the top of the include stack
130 /// and start getting tokens from it using the PTH cache.
131 void Preprocessor::EnterSourceFileWithPTH(PTHLexer *PL,
132 const DirectoryLookup *CurDir) {
134 if (CurPPLexer || CurTokenLexer)
135 PushIncludeMacroStack();
137 CurDirLookup = CurDir;
138 CurPTHLexer.reset(PL);
139 CurPPLexer = CurPTHLexer.get();
140 if (CurLexerKind != CLK_LexAfterModuleImport)
141 CurLexerKind = CLK_PTHLexer;
143 // Notify the client, if desired, that we are in a new source file.
145 FileID FID = CurPPLexer->getFileID();
146 SourceLocation EnterLoc = SourceMgr.getLocForStartOfFile(FID);
147 SrcMgr::CharacteristicKind FileType =
148 SourceMgr.getFileCharacteristic(EnterLoc);
149 Callbacks->FileChanged(EnterLoc, PPCallbacks::EnterFile, FileType);
153 /// EnterMacro - Add a Macro to the top of the include stack and start lexing
154 /// tokens from it instead of the current buffer.
155 void Preprocessor::EnterMacro(Token &Tok, SourceLocation ILEnd,
157 PushIncludeMacroStack();
160 if (NumCachedTokenLexers == 0) {
161 CurTokenLexer.reset(new TokenLexer(Tok, ILEnd, Args, *this));
163 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]);
164 CurTokenLexer->Init(Tok, ILEnd, Args);
166 if (CurLexerKind != CLK_LexAfterModuleImport)
167 CurLexerKind = CLK_TokenLexer;
170 /// EnterTokenStream - Add a "macro" context to the top of the include stack,
171 /// which will cause the lexer to start returning the specified tokens.
173 /// If DisableMacroExpansion is true, tokens lexed from the token stream will
174 /// not be subject to further macro expansion. Otherwise, these tokens will
175 /// be re-macro-expanded when/if expansion is enabled.
177 /// If OwnsTokens is false, this method assumes that the specified stream of
178 /// tokens has a permanent owner somewhere, so they do not need to be copied.
179 /// If it is true, it assumes the array of tokens is allocated with new[] and
182 void Preprocessor::EnterTokenStream(const Token *Toks, unsigned NumToks,
183 bool DisableMacroExpansion,
185 // Save our current state.
186 PushIncludeMacroStack();
189 // Create a macro expander to expand from the specified token stream.
190 if (NumCachedTokenLexers == 0) {
191 CurTokenLexer.reset(new TokenLexer(Toks, NumToks, DisableMacroExpansion,
194 CurTokenLexer.reset(TokenLexerCache[--NumCachedTokenLexers]);
195 CurTokenLexer->Init(Toks, NumToks, DisableMacroExpansion, OwnsTokens);
197 if (CurLexerKind != CLK_LexAfterModuleImport)
198 CurLexerKind = CLK_TokenLexer;
201 /// HandleEndOfFile - This callback is invoked when the lexer hits the end of
202 /// the current file. This either returns the EOF token or pops a level off
203 /// the include stack and keeps going.
204 bool Preprocessor::HandleEndOfFile(Token &Result, bool isEndOfMacro) {
205 assert(!CurTokenLexer &&
206 "Ending a file when currently in a macro!");
208 // See if this file had a controlling macro.
209 if (CurPPLexer) { // Not ending a macro, ignore it.
210 if (const IdentifierInfo *ControllingMacro =
211 CurPPLexer->MIOpt.GetControllingMacroAtEndOfFile()) {
212 // Okay, this has a controlling macro, remember in HeaderFileInfo.
213 if (const FileEntry *FE =
214 SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))
215 HeaderInfo.SetFileControllingMacro(FE, ControllingMacro);
219 // Complain about reaching an EOF within arc_cf_code_audited.
220 if (PragmaARCCFCodeAuditedLoc.isValid()) {
221 Diag(PragmaARCCFCodeAuditedLoc, diag::err_pp_eof_in_arc_cf_code_audited);
223 // Recover by leaving immediately.
224 PragmaARCCFCodeAuditedLoc = SourceLocation();
227 // If this is a #include'd file, pop it off the include stack and continue
228 // lexing the #includer file.
229 if (!IncludeMacroStack.empty()) {
231 // If we lexed the code-completion file, act as if we reached EOF.
232 if (isCodeCompletionEnabled() && CurPPLexer &&
233 SourceMgr.getLocForStartOfFile(CurPPLexer->getFileID()) ==
234 CodeCompletionFileLoc) {
237 CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof);
240 assert(CurPTHLexer && "Got EOF but no current lexer set!");
241 CurPTHLexer->getEOF(Result);
249 if (!isEndOfMacro && CurPPLexer &&
250 SourceMgr.getIncludeLoc(CurPPLexer->getFileID()).isValid()) {
251 // Notify SourceManager to record the number of FileIDs that were created
252 // during lexing of the #include'd file.
254 SourceMgr.local_sloc_entry_size() -
255 CurPPLexer->getInitialNumSLocEntries() + 1/*#include'd file*/;
256 SourceMgr.setNumCreatedFIDsForFileID(CurPPLexer->getFileID(), NumFIDs);
260 if (Callbacks && !isEndOfMacro && CurPPLexer)
261 ExitedFID = CurPPLexer->getFileID();
263 // We're done with the #included file.
264 RemoveTopOfLexerStack();
266 // Notify the client, if desired, that we are in a new source file.
267 if (Callbacks && !isEndOfMacro && CurPPLexer) {
268 SrcMgr::CharacteristicKind FileType =
269 SourceMgr.getFileCharacteristic(CurPPLexer->getSourceLocation());
270 Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
271 PPCallbacks::ExitFile, FileType, ExitedFID);
274 // Client should lex another token.
278 // If the file ends with a newline, form the EOF token on the newline itself,
279 // rather than "on the line following it", which doesn't exist. This makes
280 // diagnostics relating to the end of file include the last file that the user
281 // actually typed, which is goodness.
283 const char *EndPos = CurLexer->BufferEnd;
284 if (EndPos != CurLexer->BufferStart &&
285 (EndPos[-1] == '\n' || EndPos[-1] == '\r')) {
288 // Handle \n\r and \r\n:
289 if (EndPos != CurLexer->BufferStart &&
290 (EndPos[-1] == '\n' || EndPos[-1] == '\r') &&
291 EndPos[-1] != EndPos[0])
296 CurLexer->BufferPtr = EndPos;
297 CurLexer->FormTokenWithChars(Result, EndPos, tok::eof);
299 // We're done with the #included file.
302 assert(CurPTHLexer && "Got EOF but no current lexer set!");
303 CurPTHLexer->getEOF(Result);
309 // This is the end of the top-level file. 'WarnUnusedMacroLocs' has collected
310 // all macro locations that we need to warn because they are not used.
311 for (WarnUnusedMacroLocsTy::iterator
312 I=WarnUnusedMacroLocs.begin(), E=WarnUnusedMacroLocs.end(); I!=E; ++I)
313 Diag(*I, diag::pp_macro_not_used);
318 /// HandleEndOfTokenLexer - This callback is invoked when the current TokenLexer
319 /// hits the end of its token stream.
320 bool Preprocessor::HandleEndOfTokenLexer(Token &Result) {
321 assert(CurTokenLexer && !CurPPLexer &&
322 "Ending a macro when currently in a #include file!");
324 if (!MacroExpandingLexersStack.empty() &&
325 MacroExpandingLexersStack.back().first == CurTokenLexer.get())
326 removeCachedMacroExpandedTokensOfLastLexer();
328 // Delete or cache the now-dead macro expander.
329 if (NumCachedTokenLexers == TokenLexerCacheSize)
330 CurTokenLexer.reset();
332 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take();
334 // Handle this like a #include file being popped off the stack.
335 return HandleEndOfFile(Result, true);
338 /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the
339 /// lexer stack. This should only be used in situations where the current
340 /// state of the top-of-stack lexer is unknown.
341 void Preprocessor::RemoveTopOfLexerStack() {
342 assert(!IncludeMacroStack.empty() && "Ran out of stack entries to load");
345 // Delete or cache the now-dead macro expander.
346 if (NumCachedTokenLexers == TokenLexerCacheSize)
347 CurTokenLexer.reset();
349 TokenLexerCache[NumCachedTokenLexers++] = CurTokenLexer.take();
352 PopIncludeMacroStack();
355 /// HandleMicrosoftCommentPaste - When the macro expander pastes together a
356 /// comment (/##/) in microsoft mode, this method handles updating the current
357 /// state, returning the token on the next source line.
358 void Preprocessor::HandleMicrosoftCommentPaste(Token &Tok) {
359 assert(CurTokenLexer && !CurPPLexer &&
360 "Pasted comment can only be formed from macro");
362 // We handle this by scanning for the closest real lexer, switching it to
363 // raw mode and preprocessor mode. This will cause it to return \n as an
364 // explicit EOD token.
365 PreprocessorLexer *FoundLexer = 0;
366 bool LexerWasInPPMode = false;
367 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
368 IncludeStackInfo &ISI = *(IncludeMacroStack.end()-i-1);
369 if (ISI.ThePPLexer == 0) continue; // Scan for a real lexer.
371 // Once we find a real lexer, mark it as raw mode (disabling macro
372 // expansions) and preprocessor mode (return EOD). We know that the lexer
373 // was *not* in raw mode before, because the macro that the comment came
374 // from was expanded. However, it could have already been in preprocessor
375 // mode (#if COMMENT) in which case we have to return it to that mode and
377 FoundLexer = ISI.ThePPLexer;
378 FoundLexer->LexingRawMode = true;
379 LexerWasInPPMode = FoundLexer->ParsingPreprocessorDirective;
380 FoundLexer->ParsingPreprocessorDirective = true;
384 // Okay, we either found and switched over the lexer, or we didn't find a
385 // lexer. In either case, finish off the macro the comment came from, getting
387 if (!HandleEndOfTokenLexer(Tok)) Lex(Tok);
389 // Discarding comments as long as we don't have EOF or EOD. This 'comments
390 // out' the rest of the line, including any tokens that came from other macros
391 // that were active, as in:
392 // #define submacro a COMMENT b
394 // which should lex to 'a' only: 'b' and 'c' should be removed.
395 while (Tok.isNot(tok::eod) && Tok.isNot(tok::eof))
398 // If we got an eod token, then we successfully found the end of the line.
399 if (Tok.is(tok::eod)) {
400 assert(FoundLexer && "Can't get end of line without an active lexer");
401 // Restore the lexer back to normal mode instead of raw mode.
402 FoundLexer->LexingRawMode = false;
404 // If the lexer was already in preprocessor mode, just return the EOD token
405 // to finish the preprocessor line.
406 if (LexerWasInPPMode) return;
408 // Otherwise, switch out of PP mode and return the next lexed token.
409 FoundLexer->ParsingPreprocessorDirective = false;
413 // If we got an EOF token, then we reached the end of the token stream but
414 // didn't find an explicit \n. This can only happen if there was no lexer
415 // active (an active lexer would return EOD at EOF if there was no \n in
416 // preprocessor directive mode), so just return EOF as our token.
417 assert(!FoundLexer && "Lexer should return EOD before EOF in PP mode");