blob: 5b45ddd35e9afd2a2d79920534d248d49e936830 [file] [log] [blame]
Chris Lattnera3b605e2008-03-09 03:13:06 +00001//===--- PPDirectives.cpp - Directive Handling for Preprocessor -----------===//
Chris Lattner141e71f2008-03-09 01:54:53 +00002//
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 # directive processing for the Preprocessor.
11//
12//===----------------------------------------------------------------------===//
13
14#include "clang/Lex/Preprocessor.h"
Chris Lattner359cc442009-01-26 05:29:08 +000015#include "clang/Lex/LiteralSupport.h"
Chris Lattner141e71f2008-03-09 01:54:53 +000016#include "clang/Lex/HeaderSearch.h"
17#include "clang/Lex/MacroInfo.h"
Chris Lattner141e71f2008-03-09 01:54:53 +000018#include "clang/Basic/Diagnostic.h"
19#include "clang/Basic/SourceManager.h"
Chris Lattner359cc442009-01-26 05:29:08 +000020#include "llvm/ADT/APInt.h"
Chris Lattner141e71f2008-03-09 01:54:53 +000021using namespace clang;
22
23//===----------------------------------------------------------------------===//
24// Utility Methods for Preprocessor Directive Handling.
25//===----------------------------------------------------------------------===//
26
Ted Kremenek0ea76722008-12-15 19:56:42 +000027MacroInfo* Preprocessor::AllocateMacroInfo(SourceLocation L) {
28 MacroInfo *MI;
29
30 if (!MICache.empty()) {
31 MI = MICache.back();
32 MICache.pop_back();
33 }
34 else MI = (MacroInfo*) BP.Allocate<MacroInfo>();
35 new (MI) MacroInfo(L);
36 return MI;
37}
38
Chris Lattner141e71f2008-03-09 01:54:53 +000039/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the
40/// current line until the tok::eom token is found.
41void Preprocessor::DiscardUntilEndOfDirective() {
42 Token Tmp;
43 do {
44 LexUnexpandedToken(Tmp);
45 } while (Tmp.isNot(tok::eom));
46}
47
Chris Lattner141e71f2008-03-09 01:54:53 +000048/// ReadMacroName - Lex and validate a macro name, which occurs after a
49/// #define or #undef. This sets the token kind to eom and discards the rest
50/// of the macro line if the macro name is invalid. isDefineUndef is 1 if
51/// this is due to a a #define, 2 if #undef directive, 0 if it is something
52/// else (e.g. #ifdef).
53void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) {
54 // Read the token, don't allow macro expansion on it.
55 LexUnexpandedToken(MacroNameTok);
56
57 // Missing macro name?
Chris Lattner3692b092008-11-18 07:59:24 +000058 if (MacroNameTok.is(tok::eom)) {
59 Diag(MacroNameTok, diag::err_pp_missing_macro_name);
60 return;
61 }
Chris Lattner141e71f2008-03-09 01:54:53 +000062
63 IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
64 if (II == 0) {
65 std::string Spelling = getSpelling(MacroNameTok);
Chris Lattner9485d232008-12-13 20:12:40 +000066 const IdentifierInfo &Info = Identifiers.get(Spelling);
67 if (Info.isCPlusPlusOperatorKeyword())
Chris Lattner141e71f2008-03-09 01:54:53 +000068 // C++ 2.5p2: Alternative tokens behave the same as its primary token
69 // except for their spellings.
Chris Lattner56b05c82008-11-18 08:02:48 +000070 Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling;
Chris Lattner141e71f2008-03-09 01:54:53 +000071 else
72 Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
73 // Fall through on error.
74 } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) {
75 // Error if defining "defined": C99 6.10.8.4.
76 Diag(MacroNameTok, diag::err_defined_macro_name);
77 } else if (isDefineUndef && II->hasMacroDefinition() &&
78 getMacroInfo(II)->isBuiltinMacro()) {
79 // Error if defining "__LINE__" and other builtins: C99 6.10.8.4.
80 if (isDefineUndef == 1)
81 Diag(MacroNameTok, diag::pp_redef_builtin_macro);
82 else
83 Diag(MacroNameTok, diag::pp_undef_builtin_macro);
84 } else {
85 // Okay, we got a good identifier node. Return it.
86 return;
87 }
88
89 // Invalid macro name, read and discard the rest of the line. Then set the
90 // token kind to tok::eom.
91 MacroNameTok.setKind(tok::eom);
92 return DiscardUntilEndOfDirective();
93}
94
95/// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If
96/// not, emit a diagnostic and consume up until the eom.
97void Preprocessor::CheckEndOfDirective(const char *DirType) {
98 Token Tmp;
99 // Lex unexpanded tokens: macros might expand to zero tokens, causing us to
100 // miss diagnosing invalid lines.
101 LexUnexpandedToken(Tmp);
102
103 // There should be no tokens after the directive, but we allow them as an
104 // extension.
105 while (Tmp.is(tok::comment)) // Skip comments in -C mode.
106 LexUnexpandedToken(Tmp);
107
108 if (Tmp.isNot(tok::eom)) {
Chris Lattner56b05c82008-11-18 08:02:48 +0000109 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType;
Chris Lattner141e71f2008-03-09 01:54:53 +0000110 DiscardUntilEndOfDirective();
111 }
112}
113
114
115
116/// SkipExcludedConditionalBlock - We just read a #if or related directive and
117/// decided that the subsequent tokens are in the #if'd out portion of the
118/// file. Lex the rest of the file, until we see an #endif. If
119/// FoundNonSkipPortion is true, then we have already emitted code for part of
120/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk
121/// is true, then #else directives are ok, if not, then we have already seen one
122/// so a #else directive is a duplicate. When this returns, the caller can lex
123/// the first valid token.
124void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
125 bool FoundNonSkipPortion,
126 bool FoundElse) {
127 ++NumSkipped;
Ted Kremenekf6452c52008-11-18 01:04:47 +0000128 assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?");
Chris Lattner141e71f2008-03-09 01:54:53 +0000129
Ted Kremenek60e45d42008-11-18 00:34:22 +0000130 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false,
Chris Lattner141e71f2008-03-09 01:54:53 +0000131 FoundNonSkipPortion, FoundElse);
132
Ted Kremenek268ee702008-12-12 18:34:08 +0000133 if (CurPTHLexer) {
134 PTHSkipExcludedConditionalBlock();
135 return;
136 }
137
Chris Lattner141e71f2008-03-09 01:54:53 +0000138 // Enter raw mode to disable identifier lookup (and thus macro expansion),
139 // disabling warnings, etc.
Ted Kremenek60e45d42008-11-18 00:34:22 +0000140 CurPPLexer->LexingRawMode = true;
Chris Lattner141e71f2008-03-09 01:54:53 +0000141 Token Tok;
142 while (1) {
Ted Kremenekf6452c52008-11-18 01:04:47 +0000143 if (CurLexer)
144 CurLexer->Lex(Tok);
145 else
146 CurPTHLexer->Lex(Tok);
Chris Lattner141e71f2008-03-09 01:54:53 +0000147
148 // If this is the end of the buffer, we have an error.
149 if (Tok.is(tok::eof)) {
150 // Emit errors for each unterminated conditional on the stack, including
151 // the current one.
Ted Kremenek60e45d42008-11-18 00:34:22 +0000152 while (!CurPPLexer->ConditionalStack.empty()) {
153 Diag(CurPPLexer->ConditionalStack.back().IfLoc,
Chris Lattner141e71f2008-03-09 01:54:53 +0000154 diag::err_pp_unterminated_conditional);
Ted Kremenek60e45d42008-11-18 00:34:22 +0000155 CurPPLexer->ConditionalStack.pop_back();
Chris Lattner141e71f2008-03-09 01:54:53 +0000156 }
157
158 // Just return and let the caller lex after this #include.
159 break;
160 }
161
162 // If this token is not a preprocessor directive, just skip it.
163 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine())
164 continue;
165
166 // We just parsed a # character at the start of a line, so we're in
167 // directive mode. Tell the lexer this so any newlines we see will be
168 // converted into an EOM token (this terminates the macro).
Ted Kremenek60e45d42008-11-18 00:34:22 +0000169 CurPPLexer->ParsingPreprocessorDirective = true;
Ted Kremenekac6b06d2008-11-18 00:43:07 +0000170 if (CurLexer) CurLexer->SetCommentRetentionState(false);
Chris Lattner141e71f2008-03-09 01:54:53 +0000171
172
173 // Read the next token, the directive flavor.
174 LexUnexpandedToken(Tok);
175
176 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
177 // something bogus), skip it.
178 if (Tok.isNot(tok::identifier)) {
Ted Kremenek60e45d42008-11-18 00:34:22 +0000179 CurPPLexer->ParsingPreprocessorDirective = false;
Chris Lattner141e71f2008-03-09 01:54:53 +0000180 // Restore comment saving mode.
Ted Kremenekac6b06d2008-11-18 00:43:07 +0000181 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
Chris Lattner141e71f2008-03-09 01:54:53 +0000182 continue;
183 }
184
185 // If the first letter isn't i or e, it isn't intesting to us. We know that
186 // this is safe in the face of spelling differences, because there is no way
187 // to spell an i/e in a strange way that is another letter. Skipping this
188 // allows us to avoid looking up the identifier info for #define/#undef and
189 // other common directives.
190 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation());
191 char FirstChar = RawCharData[0];
192 if (FirstChar >= 'a' && FirstChar <= 'z' &&
193 FirstChar != 'i' && FirstChar != 'e') {
Ted Kremenek60e45d42008-11-18 00:34:22 +0000194 CurPPLexer->ParsingPreprocessorDirective = false;
Chris Lattner141e71f2008-03-09 01:54:53 +0000195 // Restore comment saving mode.
Ted Kremenekac6b06d2008-11-18 00:43:07 +0000196 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
Chris Lattner141e71f2008-03-09 01:54:53 +0000197 continue;
198 }
199
200 // Get the identifier name without trigraphs or embedded newlines. Note
201 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
202 // when skipping.
203 // TODO: could do this with zero copies in the no-clean case by using
204 // strncmp below.
205 char Directive[20];
206 unsigned IdLen;
207 if (!Tok.needsCleaning() && Tok.getLength() < 20) {
208 IdLen = Tok.getLength();
209 memcpy(Directive, RawCharData, IdLen);
210 Directive[IdLen] = 0;
211 } else {
212 std::string DirectiveStr = getSpelling(Tok);
213 IdLen = DirectiveStr.size();
214 if (IdLen >= 20) {
Ted Kremenek60e45d42008-11-18 00:34:22 +0000215 CurPPLexer->ParsingPreprocessorDirective = false;
Chris Lattner141e71f2008-03-09 01:54:53 +0000216 // Restore comment saving mode.
Ted Kremenekac6b06d2008-11-18 00:43:07 +0000217 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
Chris Lattner141e71f2008-03-09 01:54:53 +0000218 continue;
219 }
220 memcpy(Directive, &DirectiveStr[0], IdLen);
221 Directive[IdLen] = 0;
222 }
223
224 if (FirstChar == 'i' && Directive[1] == 'f') {
225 if ((IdLen == 2) || // "if"
226 (IdLen == 5 && !strcmp(Directive+2, "def")) || // "ifdef"
227 (IdLen == 6 && !strcmp(Directive+2, "ndef"))) { // "ifndef"
228 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
229 // bother parsing the condition.
230 DiscardUntilEndOfDirective();
Ted Kremenek60e45d42008-11-18 00:34:22 +0000231 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true,
Chris Lattner141e71f2008-03-09 01:54:53 +0000232 /*foundnonskip*/false,
233 /*fnddelse*/false);
234 }
235 } else if (FirstChar == 'e') {
236 if (IdLen == 5 && !strcmp(Directive+1, "ndif")) { // "endif"
237 CheckEndOfDirective("#endif");
238 PPConditionalInfo CondInfo;
239 CondInfo.WasSkipping = true; // Silence bogus warning.
Ted Kremenek60e45d42008-11-18 00:34:22 +0000240 bool InCond = CurPPLexer->popConditionalLevel(CondInfo);
Chris Lattner141e71f2008-03-09 01:54:53 +0000241 InCond = InCond; // Silence warning in no-asserts mode.
242 assert(!InCond && "Can't be skipping if not in a conditional!");
243
244 // If we popped the outermost skipping block, we're done skipping!
245 if (!CondInfo.WasSkipping)
246 break;
247 } else if (IdLen == 4 && !strcmp(Directive+1, "lse")) { // "else".
248 // #else directive in a skipping conditional. If not in some other
249 // skipping conditional, and if #else hasn't already been seen, enter it
250 // as a non-skipping conditional.
251 CheckEndOfDirective("#else");
Ted Kremenek60e45d42008-11-18 00:34:22 +0000252 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
Chris Lattner141e71f2008-03-09 01:54:53 +0000253
254 // If this is a #else with a #else before it, report the error.
255 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else);
256
257 // Note that we've seen a #else in this conditional.
258 CondInfo.FoundElse = true;
259
260 // If the conditional is at the top level, and the #if block wasn't
261 // entered, enter the #else block now.
262 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
263 CondInfo.FoundNonSkip = true;
264 break;
265 }
266 } else if (IdLen == 4 && !strcmp(Directive+1, "lif")) { // "elif".
Ted Kremenek60e45d42008-11-18 00:34:22 +0000267 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
Chris Lattner141e71f2008-03-09 01:54:53 +0000268
269 bool ShouldEnter;
270 // If this is in a skipping block or if we're already handled this #if
271 // block, don't bother parsing the condition.
272 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
273 DiscardUntilEndOfDirective();
274 ShouldEnter = false;
275 } else {
276 // Restore the value of LexingRawMode so that identifiers are
277 // looked up, etc, inside the #elif expression.
Ted Kremenek60e45d42008-11-18 00:34:22 +0000278 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
279 CurPPLexer->LexingRawMode = false;
Chris Lattner141e71f2008-03-09 01:54:53 +0000280 IdentifierInfo *IfNDefMacro = 0;
281 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
Ted Kremenek60e45d42008-11-18 00:34:22 +0000282 CurPPLexer->LexingRawMode = true;
Chris Lattner141e71f2008-03-09 01:54:53 +0000283 }
284
285 // If this is a #elif with a #else before it, report the error.
286 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else);
287
288 // If this condition is true, enter it!
289 if (ShouldEnter) {
290 CondInfo.FoundNonSkip = true;
291 break;
292 }
293 }
294 }
295
Ted Kremenek60e45d42008-11-18 00:34:22 +0000296 CurPPLexer->ParsingPreprocessorDirective = false;
Chris Lattner141e71f2008-03-09 01:54:53 +0000297 // Restore comment saving mode.
Ted Kremenekac6b06d2008-11-18 00:43:07 +0000298 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments);
Chris Lattner141e71f2008-03-09 01:54:53 +0000299 }
300
301 // Finally, if we are out of the conditional (saw an #endif or ran off the end
302 // of the file, just stop skipping and return to lexing whatever came after
303 // the #if block.
Ted Kremenek60e45d42008-11-18 00:34:22 +0000304 CurPPLexer->LexingRawMode = false;
Chris Lattner141e71f2008-03-09 01:54:53 +0000305}
306
Ted Kremenek268ee702008-12-12 18:34:08 +0000307void Preprocessor::PTHSkipExcludedConditionalBlock() {
308
309 while(1) {
310 assert(CurPTHLexer);
311 assert(CurPTHLexer->LexingRawMode == false);
312
313 // Skip to the next '#else', '#elif', or #endif.
314 if (CurPTHLexer->SkipBlock()) {
315 // We have reached an #endif. Both the '#' and 'endif' tokens
316 // have been consumed by the PTHLexer. Just pop off the condition level.
317 PPConditionalInfo CondInfo;
318 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo);
319 InCond = InCond; // Silence warning in no-asserts mode.
320 assert(!InCond && "Can't be skipping if not in a conditional!");
321 break;
322 }
323
324 // We have reached a '#else' or '#elif'. Lex the next token to get
325 // the directive flavor.
326 Token Tok;
327 LexUnexpandedToken(Tok);
328
329 // We can actually look up the IdentifierInfo here since we aren't in
330 // raw mode.
331 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID();
332
333 if (K == tok::pp_else) {
334 // #else: Enter the else condition. We aren't in a nested condition
335 // since we skip those. We're always in the one matching the last
336 // blocked we skipped.
337 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
338 // Note that we've seen a #else in this conditional.
339 CondInfo.FoundElse = true;
340
341 // If the #if block wasn't entered then enter the #else block now.
342 if (!CondInfo.FoundNonSkip) {
343 CondInfo.FoundNonSkip = true;
Ted Kremeneke5680f32008-12-23 01:30:52 +0000344
345 // Consume the eom token.
346 CurPTHLexer->ParsingPreprocessorDirective = true;
347 LexUnexpandedToken(Tok);
348 assert(Tok.is(tok::eom));
349 CurPTHLexer->ParsingPreprocessorDirective = false;
350
Ted Kremenek268ee702008-12-12 18:34:08 +0000351 break;
352 }
353
354 // Otherwise skip this block.
355 continue;
356 }
357
358 assert(K == tok::pp_elif);
359 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
360
361 // If this is a #elif with a #else before it, report the error.
362 if (CondInfo.FoundElse)
363 Diag(Tok, diag::pp_err_elif_after_else);
364
365 // If this is in a skipping block or if we're already handled this #if
366 // block, don't bother parsing the condition. We just skip this block.
367 if (CondInfo.FoundNonSkip)
368 continue;
369
370 // Evaluate the condition of the #elif.
371 IdentifierInfo *IfNDefMacro = 0;
372 CurPTHLexer->ParsingPreprocessorDirective = true;
373 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
374 CurPTHLexer->ParsingPreprocessorDirective = false;
375
376 // If this condition is true, enter it!
377 if (ShouldEnter) {
378 CondInfo.FoundNonSkip = true;
379 break;
380 }
381
382 // Otherwise, skip this block and go to the next one.
383 continue;
384 }
385}
386
Chris Lattner10725092008-03-09 04:17:44 +0000387/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file,
388/// return null on failure. isAngled indicates whether the file reference is
389/// for system #include's or not (i.e. using <> instead of "").
390const FileEntry *Preprocessor::LookupFile(const char *FilenameStart,
391 const char *FilenameEnd,
392 bool isAngled,
393 const DirectoryLookup *FromDir,
394 const DirectoryLookup *&CurDir) {
395 // If the header lookup mechanism may be relative to the current file, pass in
396 // info about where the current file is.
397 const FileEntry *CurFileEnt = 0;
398 if (!FromDir) {
Chris Lattner2b2453a2009-01-17 06:22:33 +0000399 FileID FID = getCurrentFileLexer()->getFileID();
400 CurFileEnt = SourceMgr.getFileEntryForID(FID);
Chris Lattner10725092008-03-09 04:17:44 +0000401 }
402
403 // Do a standard file entry lookup.
404 CurDir = CurDirLookup;
405 const FileEntry *FE =
406 HeaderInfo.LookupFile(FilenameStart, FilenameEnd,
407 isAngled, FromDir, CurDir, CurFileEnt);
408 if (FE) return FE;
409
410 // Otherwise, see if this is a subframework header. If so, this is relative
411 // to one of the headers on the #include stack. Walk the list of the current
412 // headers on the #include stack and pass them to HeaderInfo.
Ted Kremenek81d24e12008-11-20 16:19:53 +0000413 if (IsFileLexer()) {
Ted Kremenek41938c82008-11-19 21:57:25 +0000414 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID())))
Chris Lattner10725092008-03-09 04:17:44 +0000415 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, FilenameEnd,
416 CurFileEnt)))
417 return FE;
418 }
419
420 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
421 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1];
Ted Kremenek81d24e12008-11-20 16:19:53 +0000422 if (IsFileLexer(ISEntry)) {
Chris Lattner10725092008-03-09 04:17:44 +0000423 if ((CurFileEnt =
Ted Kremenek41938c82008-11-19 21:57:25 +0000424 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID())))
Chris Lattner10725092008-03-09 04:17:44 +0000425 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart,
426 FilenameEnd, CurFileEnt)))
427 return FE;
428 }
429 }
430
431 // Otherwise, we really couldn't find the file.
432 return 0;
433}
434
Chris Lattner141e71f2008-03-09 01:54:53 +0000435
436//===----------------------------------------------------------------------===//
437// Preprocessor Directive Handling.
438//===----------------------------------------------------------------------===//
439
440/// HandleDirective - This callback is invoked when the lexer sees a # token
441/// at the start of a line. This consumes the directive, modifies the
442/// lexer/preprocessor state, and advances the lexer(s) so that the next token
443/// read is the correct one.
444void Preprocessor::HandleDirective(Token &Result) {
445 // FIXME: Traditional: # with whitespace before it not recognized by K&R?
446
447 // We just parsed a # character at the start of a line, so we're in directive
448 // mode. Tell the lexer this so any newlines we see will be converted into an
449 // EOM token (which terminates the directive).
Ted Kremenek60e45d42008-11-18 00:34:22 +0000450 CurPPLexer->ParsingPreprocessorDirective = true;
Chris Lattner141e71f2008-03-09 01:54:53 +0000451
452 ++NumDirectives;
453
454 // We are about to read a token. For the multiple-include optimization FA to
455 // work, we have to remember if we had read any tokens *before* this
456 // pp-directive.
Ted Kremenek60e45d42008-11-18 00:34:22 +0000457 bool ReadAnyTokensBeforeDirective = CurPPLexer->MIOpt.getHasReadAnyTokensVal();
Chris Lattner141e71f2008-03-09 01:54:53 +0000458
459 // Read the next token, the directive flavor. This isn't expanded due to
460 // C99 6.10.3p8.
461 LexUnexpandedToken(Result);
462
463 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.:
464 // #define A(x) #x
465 // A(abc
466 // #warning blah
467 // def)
468 // If so, the user is relying on non-portable behavior, emit a diagnostic.
469 if (InMacroArgs)
470 Diag(Result, diag::ext_embedded_directive);
471
472TryAgain:
473 switch (Result.getKind()) {
474 case tok::eom:
475 return; // null directive.
476 case tok::comment:
477 // Handle stuff like "# /*foo*/ define X" in -E -C mode.
478 LexUnexpandedToken(Result);
479 goto TryAgain;
480
Chris Lattner478a18e2009-01-26 06:19:46 +0000481 case tok::numeric_constant: // # 7 GNU line marker directive.
482 return HandleDigitDirective(Result);
Chris Lattner141e71f2008-03-09 01:54:53 +0000483 default:
484 IdentifierInfo *II = Result.getIdentifierInfo();
485 if (II == 0) break; // Not an identifier.
486
487 // Ask what the preprocessor keyword ID is.
488 switch (II->getPPKeywordID()) {
489 default: break;
490 // C99 6.10.1 - Conditional Inclusion.
491 case tok::pp_if:
492 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective);
493 case tok::pp_ifdef:
494 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/);
495 case tok::pp_ifndef:
496 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective);
497 case tok::pp_elif:
498 return HandleElifDirective(Result);
499 case tok::pp_else:
500 return HandleElseDirective(Result);
501 case tok::pp_endif:
502 return HandleEndifDirective(Result);
503
504 // C99 6.10.2 - Source File Inclusion.
505 case tok::pp_include:
506 return HandleIncludeDirective(Result); // Handle #include.
507
508 // C99 6.10.3 - Macro Replacement.
509 case tok::pp_define:
510 return HandleDefineDirective(Result);
511 case tok::pp_undef:
512 return HandleUndefDirective(Result);
513
514 // C99 6.10.4 - Line Control.
515 case tok::pp_line:
Chris Lattner359cc442009-01-26 05:29:08 +0000516 return HandleLineDirective(Result);
Chris Lattner141e71f2008-03-09 01:54:53 +0000517
518 // C99 6.10.5 - Error Directive.
519 case tok::pp_error:
520 return HandleUserDiagnosticDirective(Result, false);
521
522 // C99 6.10.6 - Pragma Directive.
523 case tok::pp_pragma:
524 return HandlePragmaDirective();
525
526 // GNU Extensions.
527 case tok::pp_import:
528 return HandleImportDirective(Result);
529 case tok::pp_include_next:
530 return HandleIncludeNextDirective(Result);
531
532 case tok::pp_warning:
533 Diag(Result, diag::ext_pp_warning_directive);
534 return HandleUserDiagnosticDirective(Result, true);
535 case tok::pp_ident:
536 return HandleIdentSCCSDirective(Result);
537 case tok::pp_sccs:
538 return HandleIdentSCCSDirective(Result);
539 case tok::pp_assert:
540 //isExtension = true; // FIXME: implement #assert
541 break;
542 case tok::pp_unassert:
543 //isExtension = true; // FIXME: implement #unassert
544 break;
545 }
546 break;
547 }
548
549 // If we reached here, the preprocessing token is not valid!
550 Diag(Result, diag::err_pp_invalid_directive);
551
552 // Read the rest of the PP line.
553 DiscardUntilEndOfDirective();
554
555 // Okay, we're done parsing the directive.
556}
557
Chris Lattner478a18e2009-01-26 06:19:46 +0000558/// GetLineValue - Convert a numeric token into an unsigned value, emitting
559/// Diagnostic DiagID if it is invalid, and returning the value in Val.
560static bool GetLineValue(Token &DigitTok, unsigned &Val,
561 unsigned DiagID, Preprocessor &PP) {
562 if (DigitTok.isNot(tok::numeric_constant)) {
563 PP.Diag(DigitTok, DiagID);
564
565 if (DigitTok.isNot(tok::eom))
566 PP.DiscardUntilEndOfDirective();
567 return true;
568 }
569
570 llvm::SmallString<64> IntegerBuffer;
571 IntegerBuffer.resize(DigitTok.getLength());
572 const char *DigitTokBegin = &IntegerBuffer[0];
573 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin);
574 NumericLiteralParser Literal(DigitTokBegin, DigitTokBegin+ActualLength,
575 DigitTok.getLocation(), PP);
576 if (Literal.hadError)
577 return true; // Error already emitted.
578
579 if (Literal.isFloatingLiteral() || Literal.isImaginary) {
580 PP.Diag(DigitTok, DiagID);
581 return true;
582 }
583
584 // Parse the integer literal into Result.
585 llvm::APInt APVal(32, 0);
586 if (Literal.GetIntegerValue(APVal)) {
587 // Overflow parsing integer literal.
588 PP.Diag(DigitTok, DiagID);
589 return true;
590 }
591 Val = APVal.getZExtValue();
592
593 // Reject 0, this is needed both by #line numbers and flags.
594 if (Val == 0) {
595 PP.Diag(DigitTok, DiagID);
596 PP.DiscardUntilEndOfDirective();
597 return true;
598 }
599
600 return false;
601}
602
Chris Lattner359cc442009-01-26 05:29:08 +0000603/// HandleLineDirective - Handle #line directive: C99 6.10.4. The two
604/// acceptable forms are:
605/// # line digit-sequence
606/// # line digit-sequence "s-char-sequence"
607void Preprocessor::HandleLineDirective(Token &Tok) {
608 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are
609 // expanded.
610 Token DigitTok;
611 Lex(DigitTok);
612
Chris Lattner359cc442009-01-26 05:29:08 +0000613 // Validate the number and convert it to an unsigned.
Chris Lattner478a18e2009-01-26 06:19:46 +0000614 unsigned LineNo;
615 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer, *this))
Chris Lattner359cc442009-01-26 05:29:08 +0000616 return;
Chris Lattner359cc442009-01-26 05:29:08 +0000617
Chris Lattner478a18e2009-01-26 06:19:46 +0000618 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
619 // number greater than 2147483647". C90 requires that the line # be <= 32767.
Chris Lattner359cc442009-01-26 05:29:08 +0000620 unsigned LineLimit = Features.C99 ? 2147483648U : 32768U;
621 if (LineNo >= LineLimit)
622 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit;
623
624 Token StrTok;
625 Lex(StrTok);
626
627 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a
628 // string followed by eom.
629 if (StrTok.is(tok::eom))
630 ; // ok
631 else if (StrTok.isNot(tok::string_literal)) {
632 Diag(StrTok, diag::err_pp_line_invalid_filename);
633 DiscardUntilEndOfDirective();
634 return;
635 } else {
636 // Verify that there is nothing after the string, other than EOM.
637 CheckEndOfDirective("#line");
638 }
639
640 // FIXME: do something with the #line info.
641}
642
Chris Lattner478a18e2009-01-26 06:19:46 +0000643/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
644/// marker directive.
645static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
646 bool &IsSystemHeader, bool &IsExternCHeader,
647 Preprocessor &PP) {
648 unsigned FlagVal;
649 Token FlagTok;
650 PP.Lex(FlagTok);
651 if (FlagTok.is(tok::eom)) return false;
652 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
653 return true;
654
655 if (FlagVal == 1) {
656 IsFileEntry = true;
657
658 PP.Lex(FlagTok);
659 if (FlagTok.is(tok::eom)) return false;
660 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
661 return true;
662 } else if (FlagVal == 2) {
663 IsFileExit = true;
664
665 PP.Lex(FlagTok);
666 if (FlagTok.is(tok::eom)) return false;
667 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
668 return true;
669 }
670
671 // We must have 3 if there are still flags.
672 if (FlagVal != 3) {
673 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
674 return true;
675 }
676
677 IsSystemHeader = true;
678
679 PP.Lex(FlagTok);
680 if (FlagTok.is(tok::eom)) return false;
681 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
682 return true;
683
684 // We must have 4 if there is yet another flag.
685 if (FlagVal != 4) {
686 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
687 return true;
688 }
689
690 IsExternCHeader = true;
691
692 PP.Lex(FlagTok);
693 if (FlagTok.is(tok::eom)) return false;
694
695 // There are no more valid flags here.
696 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
697 return true;
698}
699
700/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
701/// one of the following forms:
702///
703/// # 42
704/// # 42 "file" ('1' | '2')?
705/// # 42 "file" ('1' | '2')? '3' '4'?
706///
707void Preprocessor::HandleDigitDirective(Token &DigitTok) {
708 // Validate the number and convert it to an unsigned. GNU does not have a
709 // line # limit other than it fit in 32-bits.
710 unsigned LineNo;
711 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer,
712 *this))
713 return;
714
715 Token StrTok;
716 Lex(StrTok);
717
718 bool IsFileEntry = false, IsFileExit = false;
719 bool IsSystemHeader = false, IsExternCHeader = false;
720
721 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a
722 // string followed by eom.
723 if (StrTok.is(tok::eom))
724 ; // ok
725 else if (StrTok.isNot(tok::string_literal)) {
726 Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
727 DiscardUntilEndOfDirective();
728 return;
729 } else {
730 // If a filename was present, read any flags that are present.
731 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit,
732 IsSystemHeader, IsExternCHeader, *this)) {
733 DiscardUntilEndOfDirective();
734 return;
735 }
736 }
737
738 // FIXME: do something with the #line info.
739
740
741
742}
743
744
Chris Lattner099dd052009-01-26 05:30:54 +0000745/// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
746///
Chris Lattner141e71f2008-03-09 01:54:53 +0000747void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
748 bool isWarning) {
Chris Lattner099dd052009-01-26 05:30:54 +0000749 // PTH doesn't emit #warning or #error directives.
750 if (CurPTHLexer)
Chris Lattner359cc442009-01-26 05:29:08 +0000751 return CurPTHLexer->DiscardToEndOfLine();
752
Chris Lattner141e71f2008-03-09 01:54:53 +0000753 // Read the rest of the line raw. We do this because we don't want macros
754 // to be expanded and we don't require that the tokens be valid preprocessing
755 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does
756 // collapse multiple consequtive white space between tokens, but this isn't
757 // specified by the standard.
Chris Lattner359cc442009-01-26 05:29:08 +0000758 std::string Message = CurLexer->ReadToEndOfLine();
759 if (isWarning)
760 Diag(Tok, diag::pp_hash_warning) << Message;
761 else
762 Diag(Tok, diag::err_pp_hash_error) << Message;
Chris Lattner141e71f2008-03-09 01:54:53 +0000763}
764
765/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
766///
767void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
768 // Yes, this directive is an extension.
769 Diag(Tok, diag::ext_pp_ident_directive);
770
771 // Read the string argument.
772 Token StrTok;
773 Lex(StrTok);
774
775 // If the token kind isn't a string, it's a malformed directive.
776 if (StrTok.isNot(tok::string_literal) &&
Chris Lattner3692b092008-11-18 07:59:24 +0000777 StrTok.isNot(tok::wide_string_literal)) {
778 Diag(StrTok, diag::err_pp_malformed_ident);
Chris Lattner099dd052009-01-26 05:30:54 +0000779 if (StrTok.isNot(tok::eom))
780 DiscardUntilEndOfDirective();
Chris Lattner3692b092008-11-18 07:59:24 +0000781 return;
782 }
Chris Lattner141e71f2008-03-09 01:54:53 +0000783
784 // Verify that there is nothing after the string, other than EOM.
785 CheckEndOfDirective("#ident");
786
787 if (Callbacks)
788 Callbacks->Ident(Tok.getLocation(), getSpelling(StrTok));
789}
790
791//===----------------------------------------------------------------------===//
792// Preprocessor Include Directive Handling.
793//===----------------------------------------------------------------------===//
794
795/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
796/// checked and spelled filename, e.g. as an operand of #include. This returns
797/// true if the input filename was in <>'s or false if it were in ""'s. The
798/// caller is expected to provide a buffer that is large enough to hold the
799/// spelling of the filename, but is also expected to handle the case when
800/// this method decides to use a different buffer.
801bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
802 const char *&BufStart,
803 const char *&BufEnd) {
804 // Get the text form of the filename.
805 assert(BufStart != BufEnd && "Can't have tokens with empty spellings!");
806
807 // Make sure the filename is <x> or "x".
808 bool isAngled;
809 if (BufStart[0] == '<') {
810 if (BufEnd[-1] != '>') {
811 Diag(Loc, diag::err_pp_expects_filename);
812 BufStart = 0;
813 return true;
814 }
815 isAngled = true;
816 } else if (BufStart[0] == '"') {
817 if (BufEnd[-1] != '"') {
818 Diag(Loc, diag::err_pp_expects_filename);
819 BufStart = 0;
820 return true;
821 }
822 isAngled = false;
823 } else {
824 Diag(Loc, diag::err_pp_expects_filename);
825 BufStart = 0;
826 return true;
827 }
828
829 // Diagnose #include "" as invalid.
830 if (BufEnd-BufStart <= 2) {
831 Diag(Loc, diag::err_pp_empty_filename);
832 BufStart = 0;
833 return "";
834 }
835
836 // Skip the brackets.
837 ++BufStart;
838 --BufEnd;
839 return isAngled;
840}
841
842/// ConcatenateIncludeName - Handle cases where the #include name is expanded
843/// from a macro as multiple tokens, which need to be glued together. This
844/// occurs for code like:
845/// #define FOO <a/b.h>
846/// #include FOO
847/// because in this case, "<a/b.h>" is returned as 7 tokens, not one.
848///
849/// This code concatenates and consumes tokens up to the '>' token. It returns
850/// false if the > was found, otherwise it returns true if it finds and consumes
851/// the EOM marker.
852static bool ConcatenateIncludeName(llvm::SmallVector<char, 128> &FilenameBuffer,
853 Preprocessor &PP) {
854 Token CurTok;
855
856 PP.Lex(CurTok);
857 while (CurTok.isNot(tok::eom)) {
858 // Append the spelling of this token to the buffer. If there was a space
859 // before it, add it now.
860 if (CurTok.hasLeadingSpace())
861 FilenameBuffer.push_back(' ');
862
863 // Get the spelling of the token, directly into FilenameBuffer if possible.
864 unsigned PreAppendSize = FilenameBuffer.size();
865 FilenameBuffer.resize(PreAppendSize+CurTok.getLength());
866
867 const char *BufPtr = &FilenameBuffer[PreAppendSize];
868 unsigned ActualLen = PP.getSpelling(CurTok, BufPtr);
869
870 // If the token was spelled somewhere else, copy it into FilenameBuffer.
871 if (BufPtr != &FilenameBuffer[PreAppendSize])
872 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
873
874 // Resize FilenameBuffer to the correct size.
875 if (CurTok.getLength() != ActualLen)
876 FilenameBuffer.resize(PreAppendSize+ActualLen);
877
878 // If we found the '>' marker, return success.
879 if (CurTok.is(tok::greater))
880 return false;
881
882 PP.Lex(CurTok);
883 }
884
885 // If we hit the eom marker, emit an error and return true so that the caller
886 // knows the EOM has been read.
887 PP.Diag(CurTok.getLocation(), diag::err_pp_expects_filename);
888 return true;
889}
890
891/// HandleIncludeDirective - The "#include" tokens have just been read, read the
892/// file to be included from the lexer, then include it! This is a common
893/// routine with functionality shared between #include, #include_next and
Chris Lattner72181832008-09-26 20:12:23 +0000894/// #import. LookupFrom is set when this is a #include_next directive, it
895/// specifies the file to start searching from.
Chris Lattner141e71f2008-03-09 01:54:53 +0000896void Preprocessor::HandleIncludeDirective(Token &IncludeTok,
897 const DirectoryLookup *LookupFrom,
898 bool isImport) {
899
900 Token FilenameTok;
Ted Kremenek60e45d42008-11-18 00:34:22 +0000901 CurPPLexer->LexIncludeFilename(FilenameTok);
Chris Lattner141e71f2008-03-09 01:54:53 +0000902
903 // Reserve a buffer to get the spelling.
904 llvm::SmallVector<char, 128> FilenameBuffer;
905 const char *FilenameStart, *FilenameEnd;
906
907 switch (FilenameTok.getKind()) {
908 case tok::eom:
909 // If the token kind is EOM, the error has already been diagnosed.
910 return;
911
912 case tok::angle_string_literal:
913 case tok::string_literal: {
914 FilenameBuffer.resize(FilenameTok.getLength());
915 FilenameStart = &FilenameBuffer[0];
916 unsigned Len = getSpelling(FilenameTok, FilenameStart);
917 FilenameEnd = FilenameStart+Len;
918 break;
919 }
920
921 case tok::less:
922 // This could be a <foo/bar.h> file coming from a macro expansion. In this
923 // case, glue the tokens together into FilenameBuffer and interpret those.
924 FilenameBuffer.push_back('<');
925 if (ConcatenateIncludeName(FilenameBuffer, *this))
926 return; // Found <eom> but no ">"? Diagnostic already emitted.
927 FilenameStart = &FilenameBuffer[0];
928 FilenameEnd = &FilenameBuffer[FilenameBuffer.size()];
929 break;
930 default:
931 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
932 DiscardUntilEndOfDirective();
933 return;
934 }
935
936 bool isAngled = GetIncludeFilenameSpelling(FilenameTok.getLocation(),
937 FilenameStart, FilenameEnd);
938 // If GetIncludeFilenameSpelling set the start ptr to null, there was an
939 // error.
940 if (FilenameStart == 0) {
941 DiscardUntilEndOfDirective();
942 return;
943 }
944
945 // Verify that there is nothing after the filename, other than EOM. Use the
946 // preprocessor to lex this in case lexing the filename entered a macro.
947 CheckEndOfDirective("#include");
948
949 // Check that we don't have infinite #include recursion.
Chris Lattner3692b092008-11-18 07:59:24 +0000950 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
951 Diag(FilenameTok, diag::err_pp_include_too_deep);
952 return;
953 }
Chris Lattner141e71f2008-03-09 01:54:53 +0000954
955 // Search include directories.
956 const DirectoryLookup *CurDir;
957 const FileEntry *File = LookupFile(FilenameStart, FilenameEnd,
958 isAngled, LookupFrom, CurDir);
Chris Lattner3692b092008-11-18 07:59:24 +0000959 if (File == 0) {
960 Diag(FilenameTok, diag::err_pp_file_not_found)
961 << std::string(FilenameStart, FilenameEnd);
962 return;
963 }
Chris Lattner141e71f2008-03-09 01:54:53 +0000964
Chris Lattner72181832008-09-26 20:12:23 +0000965 // Ask HeaderInfo if we should enter this #include file. If not, #including
966 // this file will have no effect.
967 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport))
Chris Lattner141e71f2008-03-09 01:54:53 +0000968 return;
Chris Lattner72181832008-09-26 20:12:23 +0000969
970 // The #included file will be considered to be a system header if either it is
971 // in a system include directory, or if the #includer is a system include
972 // header.
Chris Lattner9d728512008-10-27 01:19:25 +0000973 SrcMgr::CharacteristicKind FileCharacter =
Chris Lattner0b9e7362008-09-26 21:18:42 +0000974 std::max(HeaderInfo.getFileDirFlavor(File),
Chris Lattner693faa62009-01-19 07:59:15 +0000975 SourceMgr.getFileCharacteristic(FilenameTok.getLocation()));
Chris Lattner72181832008-09-26 20:12:23 +0000976
Chris Lattner141e71f2008-03-09 01:54:53 +0000977 // Look up the file, create a File ID for it.
Chris Lattner2b2453a2009-01-17 06:22:33 +0000978 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(),
979 FileCharacter);
980 if (FID.isInvalid()) {
Chris Lattner56b05c82008-11-18 08:02:48 +0000981 Diag(FilenameTok, diag::err_pp_file_not_found)
982 << std::string(FilenameStart, FilenameEnd);
983 return;
984 }
Chris Lattner141e71f2008-03-09 01:54:53 +0000985
986 // Finally, if all is good, enter the new file!
Chris Lattner2b2453a2009-01-17 06:22:33 +0000987 EnterSourceFile(FID, CurDir);
Chris Lattner141e71f2008-03-09 01:54:53 +0000988}
989
990/// HandleIncludeNextDirective - Implements #include_next.
991///
992void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) {
993 Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
994
995 // #include_next is like #include, except that we start searching after
996 // the current found directory. If we can't do this, issue a
997 // diagnostic.
998 const DirectoryLookup *Lookup = CurDirLookup;
999 if (isInPrimaryFile()) {
1000 Lookup = 0;
1001 Diag(IncludeNextTok, diag::pp_include_next_in_primary);
1002 } else if (Lookup == 0) {
1003 Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
1004 } else {
1005 // Start looking up in the next directory.
1006 ++Lookup;
1007 }
1008
1009 return HandleIncludeDirective(IncludeNextTok, Lookup);
1010}
1011
1012/// HandleImportDirective - Implements #import.
1013///
1014void Preprocessor::HandleImportDirective(Token &ImportTok) {
1015 Diag(ImportTok, diag::ext_pp_import_directive);
1016
1017 return HandleIncludeDirective(ImportTok, 0, true);
1018}
1019
1020//===----------------------------------------------------------------------===//
1021// Preprocessor Macro Directive Handling.
1022//===----------------------------------------------------------------------===//
1023
1024/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1025/// definition has just been read. Lex the rest of the arguments and the
1026/// closing ), updating MI with what we learn. Return true if an error occurs
1027/// parsing the arg list.
1028bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) {
1029 llvm::SmallVector<IdentifierInfo*, 32> Arguments;
1030
1031 Token Tok;
1032 while (1) {
1033 LexUnexpandedToken(Tok);
1034 switch (Tok.getKind()) {
1035 case tok::r_paren:
1036 // Found the end of the argument list.
1037 if (Arguments.empty()) { // #define FOO()
1038 MI->setArgumentList(Arguments.begin(), Arguments.end());
1039 return false;
1040 }
1041 // Otherwise we have #define FOO(A,)
1042 Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
1043 return true;
1044 case tok::ellipsis: // #define X(... -> C99 varargs
1045 // Warn if use of C99 feature in non-C99 mode.
1046 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro);
1047
1048 // Lex the token after the identifier.
1049 LexUnexpandedToken(Tok);
1050 if (Tok.isNot(tok::r_paren)) {
1051 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1052 return true;
1053 }
1054 // Add the __VA_ARGS__ identifier as an argument.
1055 Arguments.push_back(Ident__VA_ARGS__);
1056 MI->setIsC99Varargs();
1057 MI->setArgumentList(Arguments.begin(), Arguments.end());
1058 return false;
1059 case tok::eom: // #define X(
1060 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1061 return true;
1062 default:
1063 // Handle keywords and identifiers here to accept things like
1064 // #define Foo(for) for.
1065 IdentifierInfo *II = Tok.getIdentifierInfo();
1066 if (II == 0) {
1067 // #define X(1
1068 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
1069 return true;
1070 }
1071
1072 // If this is already used as an argument, it is used multiple times (e.g.
1073 // #define X(A,A.
1074 if (std::find(Arguments.begin(), Arguments.end(), II) !=
1075 Arguments.end()) { // C99 6.10.3p6
Chris Lattner6cf3ed72008-11-19 07:33:58 +00001076 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II;
Chris Lattner141e71f2008-03-09 01:54:53 +00001077 return true;
1078 }
1079
1080 // Add the argument to the macro info.
1081 Arguments.push_back(II);
1082
1083 // Lex the token after the identifier.
1084 LexUnexpandedToken(Tok);
1085
1086 switch (Tok.getKind()) {
1087 default: // #define X(A B
1088 Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
1089 return true;
1090 case tok::r_paren: // #define X(A)
1091 MI->setArgumentList(Arguments.begin(), Arguments.end());
1092 return false;
1093 case tok::comma: // #define X(A,
1094 break;
1095 case tok::ellipsis: // #define X(A... -> GCC extension
1096 // Diagnose extension.
1097 Diag(Tok, diag::ext_named_variadic_macro);
1098
1099 // Lex the token after the identifier.
1100 LexUnexpandedToken(Tok);
1101 if (Tok.isNot(tok::r_paren)) {
1102 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1103 return true;
1104 }
1105
1106 MI->setIsGNUVarargs();
1107 MI->setArgumentList(Arguments.begin(), Arguments.end());
1108 return false;
1109 }
1110 }
1111 }
1112}
1113
1114/// HandleDefineDirective - Implements #define. This consumes the entire macro
1115/// line then lets the caller lex the next real token.
1116void Preprocessor::HandleDefineDirective(Token &DefineTok) {
1117 ++NumDefined;
1118
1119 Token MacroNameTok;
1120 ReadMacroName(MacroNameTok, 1);
1121
1122 // Error reading macro name? If so, diagnostic already issued.
1123 if (MacroNameTok.is(tok::eom))
1124 return;
1125
1126 // If we are supposed to keep comments in #defines, reenable comment saving
1127 // mode.
Ted Kremenekac6b06d2008-11-18 00:43:07 +00001128 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
Chris Lattner141e71f2008-03-09 01:54:53 +00001129
1130 // Create the new macro.
Ted Kremenek0ea76722008-12-15 19:56:42 +00001131 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation());
Chris Lattner141e71f2008-03-09 01:54:53 +00001132
1133 Token Tok;
1134 LexUnexpandedToken(Tok);
1135
1136 // If this is a function-like macro definition, parse the argument list,
1137 // marking each of the identifiers as being used as macro arguments. Also,
1138 // check other constraints on the first token of the macro body.
1139 if (Tok.is(tok::eom)) {
1140 // If there is no body to this macro, we have no special handling here.
1141 } else if (Tok.is(tok::l_paren) && !Tok.hasLeadingSpace()) {
1142 // This is a function-like macro definition. Read the argument list.
1143 MI->setIsFunctionLike();
1144 if (ReadMacroDefinitionArgList(MI)) {
1145 // Forget about MI.
Ted Kremenek0ea76722008-12-15 19:56:42 +00001146 ReleaseMacroInfo(MI);
Chris Lattner141e71f2008-03-09 01:54:53 +00001147 // Throw away the rest of the line.
Ted Kremenek60e45d42008-11-18 00:34:22 +00001148 if (CurPPLexer->ParsingPreprocessorDirective)
Chris Lattner141e71f2008-03-09 01:54:53 +00001149 DiscardUntilEndOfDirective();
1150 return;
1151 }
1152
1153 // Read the first token after the arg list for down below.
1154 LexUnexpandedToken(Tok);
1155 } else if (!Tok.hasLeadingSpace()) {
1156 // C99 requires whitespace between the macro definition and the body. Emit
1157 // a diagnostic for something like "#define X+".
1158 if (Features.C99) {
1159 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
1160 } else {
1161 // FIXME: C90/C++ do not get this diagnostic, but it does get a similar
1162 // one in some cases!
1163 }
1164 } else {
1165 // This is a normal token with leading space. Clear the leading space
1166 // marker on the first token to get proper expansion.
1167 Tok.clearFlag(Token::LeadingSpace);
1168 }
1169
1170 // If this is a definition of a variadic C99 function-like macro, not using
1171 // the GNU named varargs extension, enabled __VA_ARGS__.
1172
1173 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
1174 // This gets unpoisoned where it is allowed.
1175 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!");
1176 if (MI->isC99Varargs())
1177 Ident__VA_ARGS__->setIsPoisoned(false);
1178
1179 // Read the rest of the macro body.
1180 if (MI->isObjectLike()) {
1181 // Object-like macros are very simple, just read their body.
1182 while (Tok.isNot(tok::eom)) {
Chris Lattner9bd0d592009-01-26 04:06:48 +00001183 // If this token has a virtual location, resolve it down to its spelling
1184 // location. This is not strictly needed, but avoids extra resolutions
1185 // for macros that are expanded frequently.
1186 if (!Tok.getLocation().isFileID())
1187 Tok.setLocation(SourceMgr.getSpellingLoc(Tok.getLocation()));
1188
Chris Lattner141e71f2008-03-09 01:54:53 +00001189 MI->AddTokenToBody(Tok);
1190 // Get the next token of the macro.
1191 LexUnexpandedToken(Tok);
1192 }
1193
1194 } else {
1195 // Otherwise, read the body of a function-like macro. This has to validate
1196 // the # (stringize) operator.
1197 while (Tok.isNot(tok::eom)) {
Chris Lattner9bd0d592009-01-26 04:06:48 +00001198 // If this token has a virtual location, resolve it down to its spelling
1199 // location. This is not strictly needed, but avoids extra resolutions
1200 // for macros that are expanded frequently.
1201 if (!Tok.getLocation().isFileID())
1202 Tok.setLocation(SourceMgr.getSpellingLoc(Tok.getLocation()));
1203
Chris Lattner141e71f2008-03-09 01:54:53 +00001204 MI->AddTokenToBody(Tok);
1205
1206 // Check C99 6.10.3.2p1: ensure that # operators are followed by macro
1207 // parameters in function-like macro expansions.
1208 if (Tok.isNot(tok::hash)) {
1209 // Get the next token of the macro.
1210 LexUnexpandedToken(Tok);
1211 continue;
1212 }
1213
1214 // Get the next token of the macro.
1215 LexUnexpandedToken(Tok);
1216
1217 // Not a macro arg identifier?
1218 if (!Tok.getIdentifierInfo() ||
1219 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) {
1220 Diag(Tok, diag::err_pp_stringize_not_parameter);
Ted Kremenek0ea76722008-12-15 19:56:42 +00001221 ReleaseMacroInfo(MI);
Chris Lattner141e71f2008-03-09 01:54:53 +00001222
1223 // Disable __VA_ARGS__ again.
1224 Ident__VA_ARGS__->setIsPoisoned(true);
1225 return;
1226 }
1227
1228 // Things look ok, add the param name token to the macro.
1229 MI->AddTokenToBody(Tok);
1230
1231 // Get the next token of the macro.
1232 LexUnexpandedToken(Tok);
1233 }
1234 }
1235
1236
1237 // Disable __VA_ARGS__ again.
1238 Ident__VA_ARGS__->setIsPoisoned(true);
1239
1240 // Check that there is no paste (##) operator at the begining or end of the
1241 // replacement list.
1242 unsigned NumTokens = MI->getNumTokens();
1243 if (NumTokens != 0) {
1244 if (MI->getReplacementToken(0).is(tok::hashhash)) {
1245 Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
Ted Kremenek0ea76722008-12-15 19:56:42 +00001246 ReleaseMacroInfo(MI);
Chris Lattner141e71f2008-03-09 01:54:53 +00001247 return;
1248 }
1249 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
1250 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
Ted Kremenek0ea76722008-12-15 19:56:42 +00001251 ReleaseMacroInfo(MI);
Chris Lattner141e71f2008-03-09 01:54:53 +00001252 return;
1253 }
1254 }
1255
1256 // If this is the primary source file, remember that this macro hasn't been
1257 // used yet.
1258 if (isInPrimaryFile())
1259 MI->setIsUsed(false);
1260
1261 // Finally, if this identifier already had a macro defined for it, verify that
1262 // the macro bodies are identical and free the old definition.
1263 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) {
Chris Lattner41c3ae12009-01-16 19:50:11 +00001264 // It is very common for system headers to have tons of macro redefinitions
1265 // and for warnings to be disabled in system headers. If this is the case,
1266 // then don't bother calling MacroInfo::isIdenticalTo.
1267 if (!Diags.getSuppressSystemWarnings() ||
1268 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) {
1269 if (!OtherMI->isUsed())
1270 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
Chris Lattner141e71f2008-03-09 01:54:53 +00001271
Chris Lattner41c3ae12009-01-16 19:50:11 +00001272 // Macros must be identical. This means all tokes and whitespace
1273 // separation must be the same. C99 6.10.3.2.
1274 if (!MI->isIdenticalTo(*OtherMI, *this)) {
1275 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
1276 << MacroNameTok.getIdentifierInfo();
1277 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition);
1278 }
Chris Lattner141e71f2008-03-09 01:54:53 +00001279 }
Chris Lattner41c3ae12009-01-16 19:50:11 +00001280
Ted Kremenek0ea76722008-12-15 19:56:42 +00001281 ReleaseMacroInfo(OtherMI);
Chris Lattner141e71f2008-03-09 01:54:53 +00001282 }
1283
1284 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI);
1285}
1286
1287/// HandleUndefDirective - Implements #undef.
1288///
1289void Preprocessor::HandleUndefDirective(Token &UndefTok) {
1290 ++NumUndefined;
1291
1292 Token MacroNameTok;
1293 ReadMacroName(MacroNameTok, 2);
1294
1295 // Error reading macro name? If so, diagnostic already issued.
1296 if (MacroNameTok.is(tok::eom))
1297 return;
1298
1299 // Check to see if this is the last token on the #undef line.
1300 CheckEndOfDirective("#undef");
1301
1302 // Okay, we finally have a valid identifier to undef.
1303 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo());
1304
1305 // If the macro is not defined, this is a noop undef, just return.
1306 if (MI == 0) return;
1307
1308 if (!MI->isUsed())
1309 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
1310
1311 // Free macro definition.
Ted Kremenek0ea76722008-12-15 19:56:42 +00001312 ReleaseMacroInfo(MI);
Chris Lattner141e71f2008-03-09 01:54:53 +00001313 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0);
1314}
1315
1316
1317//===----------------------------------------------------------------------===//
1318// Preprocessor Conditional Directive Handling.
1319//===----------------------------------------------------------------------===//
1320
1321/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is
1322/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true
1323/// if any tokens have been returned or pp-directives activated before this
1324/// #ifndef has been lexed.
1325///
1326void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef,
1327 bool ReadAnyTokensBeforeDirective) {
1328 ++NumIf;
1329 Token DirectiveTok = Result;
1330
1331 Token MacroNameTok;
1332 ReadMacroName(MacroNameTok);
1333
1334 // Error reading macro name? If so, diagnostic already issued.
1335 if (MacroNameTok.is(tok::eom)) {
1336 // Skip code until we get to #endif. This helps with recovery by not
1337 // emitting an error when the #endif is reached.
1338 SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
1339 /*Foundnonskip*/false, /*FoundElse*/false);
1340 return;
1341 }
1342
1343 // Check to see if this is the last token on the #if[n]def line.
1344 CheckEndOfDirective(isIfndef ? "#ifndef" : "#ifdef");
1345
Ted Kremenek60e45d42008-11-18 00:34:22 +00001346 if (CurPPLexer->getConditionalStackDepth() == 0) {
Chris Lattner141e71f2008-03-09 01:54:53 +00001347 // If the start of a top-level #ifdef, inform MIOpt.
1348 if (!ReadAnyTokensBeforeDirective) {
1349 assert(isIfndef && "#ifdef shouldn't reach here");
Ted Kremenek60e45d42008-11-18 00:34:22 +00001350 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MacroNameTok.getIdentifierInfo());
Chris Lattner141e71f2008-03-09 01:54:53 +00001351 } else
Ted Kremenek60e45d42008-11-18 00:34:22 +00001352 CurPPLexer->MIOpt.EnterTopLevelConditional();
Chris Lattner141e71f2008-03-09 01:54:53 +00001353 }
1354
1355 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
1356 MacroInfo *MI = getMacroInfo(MII);
1357
1358 // If there is a macro, process it.
1359 if (MI) // Mark it used.
1360 MI->setIsUsed(true);
1361
1362 // Should we include the stuff contained by this directive?
1363 if (!MI == isIfndef) {
1364 // Yes, remember that we are inside a conditional, then lex the next token.
Ted Kremenek60e45d42008-11-18 00:34:22 +00001365 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), /*wasskip*/false,
Chris Lattner141e71f2008-03-09 01:54:53 +00001366 /*foundnonskip*/true, /*foundelse*/false);
1367 } else {
1368 // No, skip the contents of this block and return the first token after it.
1369 SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
1370 /*Foundnonskip*/false,
1371 /*FoundElse*/false);
1372 }
1373}
1374
1375/// HandleIfDirective - Implements the #if directive.
1376///
1377void Preprocessor::HandleIfDirective(Token &IfToken,
1378 bool ReadAnyTokensBeforeDirective) {
1379 ++NumIf;
1380
1381 // Parse and evaluation the conditional expression.
1382 IdentifierInfo *IfNDefMacro = 0;
1383 bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro);
1384
Nuno Lopes0049db62008-06-01 18:31:24 +00001385
1386 // If this condition is equivalent to #ifndef X, and if this is the first
1387 // directive seen, handle it for the multiple-include optimization.
Ted Kremenek60e45d42008-11-18 00:34:22 +00001388 if (CurPPLexer->getConditionalStackDepth() == 0) {
Nuno Lopes0049db62008-06-01 18:31:24 +00001389 if (!ReadAnyTokensBeforeDirective && IfNDefMacro)
Ted Kremenek60e45d42008-11-18 00:34:22 +00001390 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro);
Nuno Lopes0049db62008-06-01 18:31:24 +00001391 else
Ted Kremenek60e45d42008-11-18 00:34:22 +00001392 CurPPLexer->MIOpt.EnterTopLevelConditional();
Nuno Lopes0049db62008-06-01 18:31:24 +00001393 }
1394
Chris Lattner141e71f2008-03-09 01:54:53 +00001395 // Should we include the stuff contained by this directive?
1396 if (ConditionalTrue) {
Chris Lattner141e71f2008-03-09 01:54:53 +00001397 // Yes, remember that we are inside a conditional, then lex the next token.
Ted Kremenek60e45d42008-11-18 00:34:22 +00001398 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
Chris Lattner141e71f2008-03-09 01:54:53 +00001399 /*foundnonskip*/true, /*foundelse*/false);
1400 } else {
1401 // No, skip the contents of this block and return the first token after it.
1402 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false,
1403 /*FoundElse*/false);
1404 }
1405}
1406
1407/// HandleEndifDirective - Implements the #endif directive.
1408///
1409void Preprocessor::HandleEndifDirective(Token &EndifToken) {
1410 ++NumEndif;
1411
1412 // Check that this is the whole directive.
1413 CheckEndOfDirective("#endif");
1414
1415 PPConditionalInfo CondInfo;
Ted Kremenek60e45d42008-11-18 00:34:22 +00001416 if (CurPPLexer->popConditionalLevel(CondInfo)) {
Chris Lattner141e71f2008-03-09 01:54:53 +00001417 // No conditionals on the stack: this is an #endif without an #if.
Chris Lattner3692b092008-11-18 07:59:24 +00001418 Diag(EndifToken, diag::err_pp_endif_without_if);
1419 return;
Chris Lattner141e71f2008-03-09 01:54:53 +00001420 }
1421
1422 // If this the end of a top-level #endif, inform MIOpt.
Ted Kremenek60e45d42008-11-18 00:34:22 +00001423 if (CurPPLexer->getConditionalStackDepth() == 0)
1424 CurPPLexer->MIOpt.ExitTopLevelConditional();
Chris Lattner141e71f2008-03-09 01:54:53 +00001425
Ted Kremenek60e45d42008-11-18 00:34:22 +00001426 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
Chris Lattner141e71f2008-03-09 01:54:53 +00001427 "This code should only be reachable in the non-skipping case!");
1428}
1429
1430
1431void Preprocessor::HandleElseDirective(Token &Result) {
1432 ++NumElse;
1433
1434 // #else directive in a non-skipping conditional... start skipping.
1435 CheckEndOfDirective("#else");
1436
1437 PPConditionalInfo CI;
Chris Lattner3692b092008-11-18 07:59:24 +00001438 if (CurPPLexer->popConditionalLevel(CI)) {
1439 Diag(Result, diag::pp_err_else_without_if);
1440 return;
1441 }
Chris Lattner141e71f2008-03-09 01:54:53 +00001442
1443 // If this is a top-level #else, inform the MIOpt.
Ted Kremenek60e45d42008-11-18 00:34:22 +00001444 if (CurPPLexer->getConditionalStackDepth() == 0)
1445 CurPPLexer->MIOpt.EnterTopLevelConditional();
Chris Lattner141e71f2008-03-09 01:54:53 +00001446
1447 // If this is a #else with a #else before it, report the error.
1448 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
1449
1450 // Finally, skip the rest of the contents of this block and return the first
1451 // token after it.
1452 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
1453 /*FoundElse*/true);
1454}
1455
1456void Preprocessor::HandleElifDirective(Token &ElifToken) {
1457 ++NumElse;
1458
1459 // #elif directive in a non-skipping conditional... start skipping.
1460 // We don't care what the condition is, because we will always skip it (since
1461 // the block immediately before it was included).
1462 DiscardUntilEndOfDirective();
1463
1464 PPConditionalInfo CI;
Chris Lattner3692b092008-11-18 07:59:24 +00001465 if (CurPPLexer->popConditionalLevel(CI)) {
1466 Diag(ElifToken, diag::pp_err_elif_without_if);
1467 return;
1468 }
Chris Lattner141e71f2008-03-09 01:54:53 +00001469
1470 // If this is a top-level #elif, inform the MIOpt.
Ted Kremenek60e45d42008-11-18 00:34:22 +00001471 if (CurPPLexer->getConditionalStackDepth() == 0)
1472 CurPPLexer->MIOpt.EnterTopLevelConditional();
Chris Lattner141e71f2008-03-09 01:54:53 +00001473
1474 // If this is a #elif with a #else before it, report the error.
1475 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
1476
1477 // Finally, skip the rest of the contents of this block and return the first
1478 // token after it.
1479 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
1480 /*FoundElse*/CI.FoundElse);
1481}
1482