blob: 6399e867b3825f30cfc86d1d731c46164366730c [file] [log] [blame]
Chris Lattner1543e9c2008-03-09 02:18:51 +00001//===--- TokenLexer.cpp - Lex from a token stream -------------------------===//
Reid Spencer5f016e22007-07-11 17:01:13 +00002//
3// The LLVM Compiler Infrastructure
4//
Chris Lattner0bc735f2007-12-29 19:59:25 +00005// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
Reid Spencer5f016e22007-07-11 17:01:13 +00007//
8//===----------------------------------------------------------------------===//
9//
Chris Lattner1543e9c2008-03-09 02:18:51 +000010// This file implements the TokenLexer interface.
Reid Spencer5f016e22007-07-11 17:01:13 +000011//
12//===----------------------------------------------------------------------===//
13
Chris Lattner5d75de02008-03-09 02:22:57 +000014#include "clang/Lex/TokenLexer.h"
Chris Lattnere5c8ffe2008-03-09 02:55:12 +000015#include "MacroArgs.h"
Reid Spencer5f016e22007-07-11 17:01:13 +000016#include "clang/Lex/MacroInfo.h"
17#include "clang/Lex/Preprocessor.h"
18#include "clang/Basic/SourceManager.h"
19#include "clang/Basic/Diagnostic.h"
20#include "llvm/ADT/SmallVector.h"
21using namespace clang;
22
Reid Spencer5f016e22007-07-11 17:01:13 +000023
Chris Lattner1543e9c2008-03-09 02:18:51 +000024/// Create a TokenLexer for the specified macro with the specified actual
Reid Spencer5f016e22007-07-11 17:01:13 +000025/// arguments. Note that this ctor takes ownership of the ActualArgs pointer.
Chris Lattner1543e9c2008-03-09 02:18:51 +000026void TokenLexer::Init(Token &Tok, MacroArgs *Actuals) {
27 // If the client is reusing a TokenLexer, make sure to free any memory
Chris Lattner9594acf2007-07-15 00:25:26 +000028 // associated with it.
29 destroy();
30
Chris Lattnercc1a8752007-10-07 08:44:20 +000031 Macro = PP.getMacroInfo(Tok.getIdentifierInfo());
Chris Lattner9594acf2007-07-15 00:25:26 +000032 ActualArgs = Actuals;
33 CurToken = 0;
34 InstantiateLoc = Tok.getLocation();
35 AtStartOfLine = Tok.isAtStartOfLine();
36 HasLeadingSpace = Tok.hasLeadingSpace();
Chris Lattner8d896432008-03-09 02:07:49 +000037 Tokens = &*Macro->tokens_begin();
38 OwnsTokens = false;
Chris Lattner6b884502008-03-10 06:06:04 +000039 DisableMacroExpansion = false;
Chris Lattner8d896432008-03-09 02:07:49 +000040 NumTokens = Macro->tokens_end()-Macro->tokens_begin();
Reid Spencer5f016e22007-07-11 17:01:13 +000041
42 // If this is a function-like macro, expand the arguments and change
Chris Lattner8d896432008-03-09 02:07:49 +000043 // Tokens to point to the expanded tokens.
Reid Spencer5f016e22007-07-11 17:01:13 +000044 if (Macro->isFunctionLike() && Macro->getNumArgs())
45 ExpandFunctionArguments();
46
47 // Mark the macro as currently disabled, so that it is not recursively
48 // expanded. The macro must be disabled only after argument pre-expansion of
49 // function-like macro arguments occurs.
50 Macro->DisableMacro();
51}
52
Chris Lattner9594acf2007-07-15 00:25:26 +000053
54
Chris Lattner1543e9c2008-03-09 02:18:51 +000055/// Create a TokenLexer for the specified token stream. This does not
Reid Spencer5f016e22007-07-11 17:01:13 +000056/// take ownership of the specified token vector.
Chris Lattner6b884502008-03-10 06:06:04 +000057void TokenLexer::Init(const Token *TokArray, unsigned NumToks,
58 bool disableMacroExpansion, bool ownsTokens) {
Chris Lattner1543e9c2008-03-09 02:18:51 +000059 // If the client is reusing a TokenLexer, make sure to free any memory
Chris Lattner9594acf2007-07-15 00:25:26 +000060 // associated with it.
61 destroy();
62
63 Macro = 0;
64 ActualArgs = 0;
Chris Lattner8d896432008-03-09 02:07:49 +000065 Tokens = TokArray;
Chris Lattner6b884502008-03-10 06:06:04 +000066 OwnsTokens = ownsTokens;
67 DisableMacroExpansion = disableMacroExpansion;
Chris Lattner8d896432008-03-09 02:07:49 +000068 NumTokens = NumToks;
Chris Lattner9594acf2007-07-15 00:25:26 +000069 CurToken = 0;
70 InstantiateLoc = SourceLocation();
71 AtStartOfLine = false;
72 HasLeadingSpace = false;
Reid Spencer5f016e22007-07-11 17:01:13 +000073
74 // Set HasLeadingSpace/AtStartOfLine so that the first token will be
75 // returned unmodified.
76 if (NumToks != 0) {
77 AtStartOfLine = TokArray[0].isAtStartOfLine();
78 HasLeadingSpace = TokArray[0].hasLeadingSpace();
79 }
80}
81
82
Chris Lattner1543e9c2008-03-09 02:18:51 +000083void TokenLexer::destroy() {
Reid Spencer5f016e22007-07-11 17:01:13 +000084 // If this was a function-like macro that actually uses its arguments, delete
85 // the expanded tokens.
Chris Lattner8d896432008-03-09 02:07:49 +000086 if (OwnsTokens) {
87 delete [] Tokens;
88 Tokens = 0;
Chris Lattner9c683062007-07-22 01:16:55 +000089 }
Reid Spencer5f016e22007-07-11 17:01:13 +000090
Chris Lattner1543e9c2008-03-09 02:18:51 +000091 // TokenLexer owns its formal arguments.
Reid Spencer5f016e22007-07-11 17:01:13 +000092 if (ActualArgs) ActualArgs->destroy();
93}
94
95/// Expand the arguments of a function-like macro so that we can quickly
Chris Lattner8d896432008-03-09 02:07:49 +000096/// return preexpanded tokens from Tokens.
Chris Lattner1543e9c2008-03-09 02:18:51 +000097void TokenLexer::ExpandFunctionArguments() {
Chris Lattnerd2177732007-07-20 16:59:19 +000098 llvm::SmallVector<Token, 128> ResultToks;
Reid Spencer5f016e22007-07-11 17:01:13 +000099
Chris Lattner8d896432008-03-09 02:07:49 +0000100 // Loop through 'Tokens', expanding them into ResultToks. Keep
Reid Spencer5f016e22007-07-11 17:01:13 +0000101 // track of whether we change anything. If not, no need to keep them. If so,
Chris Lattner8d896432008-03-09 02:07:49 +0000102 // we install the newly expanded sequence as the new 'Tokens' list.
Reid Spencer5f016e22007-07-11 17:01:13 +0000103 bool MadeChange = false;
104
105 // NextTokGetsSpace - When this is true, the next token appended to the
106 // output list will get a leading space, regardless of whether it had one to
107 // begin with or not. This is used for placemarker support.
108 bool NextTokGetsSpace = false;
109
Chris Lattner8d896432008-03-09 02:07:49 +0000110 for (unsigned i = 0, e = NumTokens; i != e; ++i) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000111 // If we found the stringify operator, get the argument stringified. The
112 // preprocessor already verified that the following token is a macro name
113 // when the #define was parsed.
Chris Lattner8d896432008-03-09 02:07:49 +0000114 const Token &CurTok = Tokens[i];
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000115 if (CurTok.is(tok::hash) || CurTok.is(tok::hashat)) {
Chris Lattner8d896432008-03-09 02:07:49 +0000116 int ArgNo = Macro->getArgumentNum(Tokens[i+1].getIdentifierInfo());
Reid Spencer5f016e22007-07-11 17:01:13 +0000117 assert(ArgNo != -1 && "Token following # is not an argument?");
118
Chris Lattnerd2177732007-07-20 16:59:19 +0000119 Token Res;
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000120 if (CurTok.is(tok::hash)) // Stringify
Reid Spencer5f016e22007-07-11 17:01:13 +0000121 Res = ActualArgs->getStringifiedArgument(ArgNo, PP);
122 else {
123 // 'charify': don't bother caching these.
Chris Lattnere5c8ffe2008-03-09 02:55:12 +0000124 Res = MacroArgs::StringifyArgument(ActualArgs->getUnexpArgument(ArgNo),
125 PP, true);
Reid Spencer5f016e22007-07-11 17:01:13 +0000126 }
127
128 // The stringified/charified string leading space flag gets set to match
129 // the #/#@ operator.
130 if (CurTok.hasLeadingSpace() || NextTokGetsSpace)
Chris Lattnerd2177732007-07-20 16:59:19 +0000131 Res.setFlag(Token::LeadingSpace);
Reid Spencer5f016e22007-07-11 17:01:13 +0000132
133 ResultToks.push_back(Res);
134 MadeChange = true;
135 ++i; // Skip arg name.
136 NextTokGetsSpace = false;
137 continue;
138 }
139
140 // Otherwise, if this is not an argument token, just add the token to the
141 // output buffer.
142 IdentifierInfo *II = CurTok.getIdentifierInfo();
143 int ArgNo = II ? Macro->getArgumentNum(II) : -1;
144 if (ArgNo == -1) {
145 // This isn't an argument, just add it.
146 ResultToks.push_back(CurTok);
147
148 if (NextTokGetsSpace) {
Chris Lattnerd2177732007-07-20 16:59:19 +0000149 ResultToks.back().setFlag(Token::LeadingSpace);
Reid Spencer5f016e22007-07-11 17:01:13 +0000150 NextTokGetsSpace = false;
151 }
152 continue;
153 }
154
155 // An argument is expanded somehow, the result is different than the
156 // input.
157 MadeChange = true;
158
159 // Otherwise, this is a use of the argument. Find out if there is a paste
160 // (##) operator before or after the argument.
161 bool PasteBefore =
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000162 !ResultToks.empty() && ResultToks.back().is(tok::hashhash);
Chris Lattner8d896432008-03-09 02:07:49 +0000163 bool PasteAfter = i+1 != e && Tokens[i+1].is(tok::hashhash);
Reid Spencer5f016e22007-07-11 17:01:13 +0000164
165 // If it is not the LHS/RHS of a ## operator, we must pre-expand the
166 // argument and substitute the expanded tokens into the result. This is
167 // C99 6.10.3.1p1.
168 if (!PasteBefore && !PasteAfter) {
Chris Lattnerd2177732007-07-20 16:59:19 +0000169 const Token *ResultArgToks;
Reid Spencer5f016e22007-07-11 17:01:13 +0000170
171 // Only preexpand the argument if it could possibly need it. This
172 // avoids some work in common cases.
Chris Lattnerd2177732007-07-20 16:59:19 +0000173 const Token *ArgTok = ActualArgs->getUnexpArgument(ArgNo);
Chris Lattnercc1a8752007-10-07 08:44:20 +0000174 if (ActualArgs->ArgNeedsPreexpansion(ArgTok, PP))
Reid Spencer5f016e22007-07-11 17:01:13 +0000175 ResultArgToks = &ActualArgs->getPreExpArgument(ArgNo, PP)[0];
176 else
177 ResultArgToks = ArgTok; // Use non-preexpanded tokens.
178
179 // If the arg token expanded into anything, append it.
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000180 if (ResultArgToks->isNot(tok::eof)) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000181 unsigned FirstResult = ResultToks.size();
182 unsigned NumToks = MacroArgs::getArgLength(ResultArgToks);
183 ResultToks.append(ResultArgToks, ResultArgToks+NumToks);
184
185 // If any tokens were substituted from the argument, the whitespace
186 // before the first token should match the whitespace of the arg
187 // identifier.
Chris Lattnerd2177732007-07-20 16:59:19 +0000188 ResultToks[FirstResult].setFlagValue(Token::LeadingSpace,
Reid Spencer5f016e22007-07-11 17:01:13 +0000189 CurTok.hasLeadingSpace() ||
190 NextTokGetsSpace);
191 NextTokGetsSpace = false;
192 } else {
193 // If this is an empty argument, and if there was whitespace before the
194 // formal token, make sure the next token gets whitespace before it.
195 NextTokGetsSpace = CurTok.hasLeadingSpace();
196 }
197 continue;
198 }
199
200 // Okay, we have a token that is either the LHS or RHS of a paste (##)
201 // argument. It gets substituted as its non-pre-expanded tokens.
Chris Lattnerd2177732007-07-20 16:59:19 +0000202 const Token *ArgToks = ActualArgs->getUnexpArgument(ArgNo);
Reid Spencer5f016e22007-07-11 17:01:13 +0000203 unsigned NumToks = MacroArgs::getArgLength(ArgToks);
204 if (NumToks) { // Not an empty argument?
Chris Lattner71a3a8d2008-01-29 07:54:23 +0000205 // If this is the GNU ", ## __VA_ARG__" extension, and we just learned
206 // that __VA_ARG__ expands to multiple tokens, avoid a pasting error when
207 // the expander trys to paste ',' with the first token of the __VA_ARG__
208 // expansion.
209 if (PasteBefore && ResultToks.size() >= 2 &&
210 ResultToks[ResultToks.size()-2].is(tok::comma) &&
211 (unsigned)ArgNo == Macro->getNumArgs()-1 &&
212 Macro->isVariadic()) {
213 // Remove the paste operator, report use of the extension.
214 PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma);
215 ResultToks.pop_back();
216 }
217
Reid Spencer5f016e22007-07-11 17:01:13 +0000218 ResultToks.append(ArgToks, ArgToks+NumToks);
219
220 // If the next token was supposed to get leading whitespace, ensure it has
221 // it now.
222 if (NextTokGetsSpace) {
Chris Lattnerd2177732007-07-20 16:59:19 +0000223 ResultToks[ResultToks.size()-NumToks].setFlag(Token::LeadingSpace);
Reid Spencer5f016e22007-07-11 17:01:13 +0000224 NextTokGetsSpace = false;
225 }
226 continue;
227 }
228
229 // If an empty argument is on the LHS or RHS of a paste, the standard (C99
230 // 6.10.3.3p2,3) calls for a bunch of placemarker stuff to occur. We
231 // implement this by eating ## operators when a LHS or RHS expands to
232 // empty.
233 NextTokGetsSpace |= CurTok.hasLeadingSpace();
234 if (PasteAfter) {
235 // Discard the argument token and skip (don't copy to the expansion
236 // buffer) the paste operator after it.
Chris Lattner8d896432008-03-09 02:07:49 +0000237 NextTokGetsSpace |= Tokens[i+1].hasLeadingSpace();
Reid Spencer5f016e22007-07-11 17:01:13 +0000238 ++i;
239 continue;
240 }
241
242 // If this is on the RHS of a paste operator, we've already copied the
243 // paste operator to the ResultToks list. Remove it.
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000244 assert(PasteBefore && ResultToks.back().is(tok::hashhash));
Reid Spencer5f016e22007-07-11 17:01:13 +0000245 NextTokGetsSpace |= ResultToks.back().hasLeadingSpace();
246 ResultToks.pop_back();
247
248 // If this is the __VA_ARGS__ token, and if the argument wasn't provided,
249 // and if the macro had at least one real argument, and if the token before
250 // the ## was a comma, remove the comma.
251 if ((unsigned)ArgNo == Macro->getNumArgs()-1 && // is __VA_ARGS__
252 ActualArgs->isVarargsElidedUse() && // Argument elided.
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000253 !ResultToks.empty() && ResultToks.back().is(tok::comma)) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000254 // Never add a space, even if the comma, ##, or arg had a space.
255 NextTokGetsSpace = false;
Chris Lattner71a3a8d2008-01-29 07:54:23 +0000256 // Remove the paste operator, report use of the extension.
257 PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma);
Reid Spencer5f016e22007-07-11 17:01:13 +0000258 ResultToks.pop_back();
259 }
260 continue;
261 }
262
Chris Lattner8d896432008-03-09 02:07:49 +0000263 // If anything changed, install this as the new Tokens list.
Reid Spencer5f016e22007-07-11 17:01:13 +0000264 if (MadeChange) {
265 // This is deleted in the dtor.
Chris Lattner8d896432008-03-09 02:07:49 +0000266 NumTokens = ResultToks.size();
Chris Lattnerd2177732007-07-20 16:59:19 +0000267 Token *Res = new Token[ResultToks.size()];
Chris Lattner8d896432008-03-09 02:07:49 +0000268 if (NumTokens)
269 memcpy(Res, &ResultToks[0], NumTokens*sizeof(Token));
270 Tokens = Res;
271 OwnsTokens = true;
Reid Spencer5f016e22007-07-11 17:01:13 +0000272 }
273}
274
275/// Lex - Lex and return a token from this macro stream.
276///
Chris Lattner1543e9c2008-03-09 02:18:51 +0000277void TokenLexer::Lex(Token &Tok) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000278 // Lexing off the end of the macro, pop this macro off the expansion stack.
279 if (isAtEnd()) {
280 // If this is a macro (not a token stream), mark the macro enabled now
281 // that it is no longer being expanded.
282 if (Macro) Macro->EnableMacro();
283
284 // Pop this context off the preprocessors lexer stack and get the next
285 // token. This will delete "this" so remember the PP instance var.
286 Preprocessor &PPCache = PP;
Chris Lattnerfde2bf92008-03-09 03:04:16 +0000287 if (PP.HandleEndOfTokenLexer(Tok))
Reid Spencer5f016e22007-07-11 17:01:13 +0000288 return;
289
Chris Lattnerfde2bf92008-03-09 03:04:16 +0000290 // HandleEndOfTokenLexer may not return a token. If it doesn't, lex
291 // whatever is next.
Reid Spencer5f016e22007-07-11 17:01:13 +0000292 return PPCache.Lex(Tok);
293 }
294
295 // If this is the first token of the expanded result, we inherit spacing
296 // properties later.
297 bool isFirstToken = CurToken == 0;
298
299 // Get the next token to return.
Chris Lattner8d896432008-03-09 02:07:49 +0000300 Tok = Tokens[CurToken++];
Reid Spencer5f016e22007-07-11 17:01:13 +0000301
302 // If this token is followed by a token paste (##) operator, paste the tokens!
Chris Lattner8d896432008-03-09 02:07:49 +0000303 if (!isAtEnd() && Tokens[CurToken].is(tok::hashhash))
Chris Lattner3f1cc832008-02-07 06:03:59 +0000304 if (PasteTokens(Tok)) {
305 // When handling the microsoft /##/ extension, the final token is
306 // returned by PasteTokens, not the pasted token.
307 return;
308 }
Reid Spencer5f016e22007-07-11 17:01:13 +0000309
310 // The token's current location indicate where the token was lexed from. We
311 // need this information to compute the spelling of the token, but any
312 // diagnostics for the expanded token should appear as if they came from
313 // InstantiationLoc. Pull this information together into a new SourceLocation
314 // that captures all of this.
315 if (InstantiateLoc.isValid()) { // Don't do this for token streams.
316 SourceManager &SrcMgr = PP.getSourceManager();
Chris Lattnerabca2bb2007-07-15 06:35:27 +0000317 Tok.setLocation(SrcMgr.getInstantiationLoc(Tok.getLocation(),
318 InstantiateLoc));
Reid Spencer5f016e22007-07-11 17:01:13 +0000319 }
320
321 // If this is the first token, set the lexical properties of the token to
322 // match the lexical properties of the macro identifier.
323 if (isFirstToken) {
Chris Lattnerd2177732007-07-20 16:59:19 +0000324 Tok.setFlagValue(Token::StartOfLine , AtStartOfLine);
325 Tok.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
Reid Spencer5f016e22007-07-11 17:01:13 +0000326 }
327
328 // Handle recursive expansion!
Chris Lattner6b884502008-03-10 06:06:04 +0000329 if (Tok.getIdentifierInfo() && !DisableMacroExpansion)
Reid Spencer5f016e22007-07-11 17:01:13 +0000330 return PP.HandleIdentifier(Tok);
331
332 // Otherwise, return a normal token.
333}
334
335/// PasteTokens - Tok is the LHS of a ## operator, and CurToken is the ##
336/// operator. Read the ## and RHS, and paste the LHS/RHS together. If there
Nico Weber48002c82008-09-29 00:25:48 +0000337/// are more ## after it, chomp them iteratively. Return the result as Tok.
Chris Lattner3f1cc832008-02-07 06:03:59 +0000338/// If this returns true, the caller should immediately return the token.
Chris Lattner1543e9c2008-03-09 02:18:51 +0000339bool TokenLexer::PasteTokens(Token &Tok) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000340 llvm::SmallVector<char, 128> Buffer;
341 do {
342 // Consume the ## operator.
Chris Lattner8d896432008-03-09 02:07:49 +0000343 SourceLocation PasteOpLoc = Tokens[CurToken].getLocation();
Reid Spencer5f016e22007-07-11 17:01:13 +0000344 ++CurToken;
345 assert(!isAtEnd() && "No token on the RHS of a paste operator!");
346
347 // Get the RHS token.
Chris Lattner8d896432008-03-09 02:07:49 +0000348 const Token &RHS = Tokens[CurToken];
Reid Spencer5f016e22007-07-11 17:01:13 +0000349
350 bool isInvalid = false;
351
352 // Allocate space for the result token. This is guaranteed to be enough for
353 // the two tokens and a null terminator.
354 Buffer.resize(Tok.getLength() + RHS.getLength() + 1);
355
356 // Get the spelling of the LHS token in Buffer.
357 const char *BufPtr = &Buffer[0];
358 unsigned LHSLen = PP.getSpelling(Tok, BufPtr);
359 if (BufPtr != &Buffer[0]) // Really, we want the chars in Buffer!
360 memcpy(&Buffer[0], BufPtr, LHSLen);
361
362 BufPtr = &Buffer[LHSLen];
363 unsigned RHSLen = PP.getSpelling(RHS, BufPtr);
364 if (BufPtr != &Buffer[LHSLen]) // Really, we want the chars in Buffer!
365 memcpy(&Buffer[LHSLen], BufPtr, RHSLen);
366
367 // Add null terminator.
368 Buffer[LHSLen+RHSLen] = '\0';
369
370 // Trim excess space.
371 Buffer.resize(LHSLen+RHSLen+1);
372
373 // Plop the pasted result (including the trailing newline and null) into a
374 // scratch buffer where we can lex it.
375 SourceLocation ResultTokLoc = PP.CreateString(&Buffer[0], Buffer.size());
376
377 // Lex the resultant pasted token into Result.
Chris Lattnerd2177732007-07-20 16:59:19 +0000378 Token Result;
Reid Spencer5f016e22007-07-11 17:01:13 +0000379
Chris Lattner0af57422008-10-12 01:31:51 +0000380 if (Tok.is(tok::identifier) && RHS.is(tok::identifier)) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000381 // Common paste case: identifier+identifier = identifier. Avoid creating
382 // a lexer and other overhead.
383 PP.IncrementPasteCounter(true);
384 Result.startToken();
385 Result.setKind(tok::identifier);
386 Result.setLocation(ResultTokLoc);
387 Result.setLength(LHSLen+RHSLen);
388 } else {
389 PP.IncrementPasteCounter(false);
390
391 // Make a lexer to lex this string from.
392 SourceManager &SourceMgr = PP.getSourceManager();
393 const char *ResultStrData = SourceMgr.getCharacterData(ResultTokLoc);
394
Chris Lattner590f0cc2008-10-12 01:15:46 +0000395 const llvm::MemoryBuffer *Buffer =
396 SourceMgr.getBuffer(ResultTokLoc.getFileID());
397
Reid Spencer5f016e22007-07-11 17:01:13 +0000398 // Make a lexer object so that we lex and expand the paste result.
Chris Lattner590f0cc2008-10-12 01:15:46 +0000399 Lexer TL(ResultTokLoc, PP.getLangOptions(), ResultStrData,
400 ResultStrData+LHSLen+RHSLen /*don't include null*/, Buffer);
Reid Spencer5f016e22007-07-11 17:01:13 +0000401
402 // Lex a token in raw mode. This way it won't look up identifiers
403 // automatically, lexing off the end will return an eof token, and
404 // warnings are disabled. This returns true if the result token is the
405 // entire buffer.
Chris Lattner590f0cc2008-10-12 01:15:46 +0000406 bool IsComplete = TL.LexFromRawLexer(Result);
Reid Spencer5f016e22007-07-11 17:01:13 +0000407
408 // If we got an EOF token, we didn't form even ONE token. For example, we
409 // did "/ ## /" to get "//".
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000410 IsComplete &= Result.isNot(tok::eof);
Reid Spencer5f016e22007-07-11 17:01:13 +0000411 isInvalid = !IsComplete;
Reid Spencer5f016e22007-07-11 17:01:13 +0000412 }
413
414 // If pasting the two tokens didn't form a full new token, this is an error.
415 // This occurs with "x ## +" and other stuff. Return with Tok unmodified
416 // and with RHS as the next token to lex.
417 if (isInvalid) {
Chris Lattner3f1cc832008-02-07 06:03:59 +0000418 // Test for the Microsoft extension of /##/ turning into // here on the
419 // error path.
420 if (PP.getLangOptions().Microsoft && Tok.is(tok::slash) &&
421 RHS.is(tok::slash)) {
422 HandleMicrosoftCommentPaste(Tok);
423 return true;
424 } else {
425 // TODO: If not in assembler language mode.
Chris Lattner204b2fe2008-11-18 21:48:13 +0000426 PP.Diag(PasteOpLoc, diag::err_pp_bad_paste)
427 << std::string(Buffer.begin(), Buffer.end()-1);
Chris Lattner3f1cc832008-02-07 06:03:59 +0000428 return false;
429 }
Reid Spencer5f016e22007-07-11 17:01:13 +0000430 }
431
Chris Lattner1543e9c2008-03-09 02:18:51 +0000432 // Turn ## into 'unknown' to avoid # ## # from looking like a paste
433 // operator.
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000434 if (Result.is(tok::hashhash))
Reid Spencer5f016e22007-07-11 17:01:13 +0000435 Result.setKind(tok::unknown);
Chris Lattner71a3a8d2008-01-29 07:54:23 +0000436 // FIXME: Turn __VA_ARGS__ into "not a token"?
Reid Spencer5f016e22007-07-11 17:01:13 +0000437
438 // Transfer properties of the LHS over the the Result.
Chris Lattnerd2177732007-07-20 16:59:19 +0000439 Result.setFlagValue(Token::StartOfLine , Tok.isAtStartOfLine());
440 Result.setFlagValue(Token::LeadingSpace, Tok.hasLeadingSpace());
Reid Spencer5f016e22007-07-11 17:01:13 +0000441
442 // Finally, replace LHS with the result, consume the RHS, and iterate.
443 ++CurToken;
444 Tok = Result;
Chris Lattner8d896432008-03-09 02:07:49 +0000445 } while (!isAtEnd() && Tokens[CurToken].is(tok::hashhash));
Reid Spencer5f016e22007-07-11 17:01:13 +0000446
447 // Now that we got the result token, it will be subject to expansion. Since
448 // token pasting re-lexes the result token in raw mode, identifier information
449 // isn't looked up. As such, if the result is an identifier, look up id info.
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000450 if (Tok.is(tok::identifier)) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000451 // Look up the identifier info for the token. We disabled identifier lookup
452 // by saying we're skipping contents, so we need to do this manually.
453 Tok.setIdentifierInfo(PP.LookUpIdentifierInfo(Tok));
454 }
Chris Lattner3f1cc832008-02-07 06:03:59 +0000455 return false;
Reid Spencer5f016e22007-07-11 17:01:13 +0000456}
457
458/// isNextTokenLParen - If the next token lexed will pop this macro off the
459/// expansion stack, return 2. If the next unexpanded token is a '(', return
460/// 1, otherwise return 0.
Chris Lattner1543e9c2008-03-09 02:18:51 +0000461unsigned TokenLexer::isNextTokenLParen() const {
Reid Spencer5f016e22007-07-11 17:01:13 +0000462 // Out of tokens?
463 if (isAtEnd())
464 return 2;
Chris Lattner8d896432008-03-09 02:07:49 +0000465 return Tokens[CurToken].is(tok::l_paren);
Reid Spencer5f016e22007-07-11 17:01:13 +0000466}
Chris Lattner3f1cc832008-02-07 06:03:59 +0000467
468
469/// HandleMicrosoftCommentPaste - In microsoft compatibility mode, /##/ pastes
470/// together to form a comment that comments out everything in the current
471/// macro, other active macros, and anything left on the current physical
472/// source line of the instantiated buffer. Handle this by returning the
473/// first token on the next line.
Chris Lattner1543e9c2008-03-09 02:18:51 +0000474void TokenLexer::HandleMicrosoftCommentPaste(Token &Tok) {
Chris Lattner3f1cc832008-02-07 06:03:59 +0000475 // We 'comment out' the rest of this macro by just ignoring the rest of the
476 // tokens that have not been lexed yet, if any.
477
478 // Since this must be a macro, mark the macro enabled now that it is no longer
479 // being expanded.
480 assert(Macro && "Token streams can't paste comments");
481 Macro->EnableMacro();
482
483 PP.HandleMicrosoftCommentPaste(Tok);
484}