blob: 0a74b26482e5ccb16ab6ababe74c3c329a3f8ec6 [file] [log] [blame]
Chris Lattner22eb9722006-06-18 05:43:12 +00001//===--- Lexer.cpp - C Language Family Lexer ------------------------------===//
2//
3// The LLVM Compiler Infrastructure
4//
Chris Lattner5b12ab82007-12-29 19:59:25 +00005// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
Chris Lattner22eb9722006-06-18 05:43:12 +00007//
8//===----------------------------------------------------------------------===//
9//
Chris Lattner146762e2007-07-20 16:59:19 +000010// This file implements the Lexer and Token interfaces.
Chris Lattner22eb9722006-06-18 05:43:12 +000011//
12//===----------------------------------------------------------------------===//
13//
14// TODO: GCC Diagnostics emitted by the lexer:
15// PEDWARN: (form feed|vertical tab) in preprocessing directive
16//
17// Universal characters, unicode, char mapping:
18// WARNING: `%.*s' is not in NFKC
19// WARNING: `%.*s' is not in NFC
20//
21// Other:
Chris Lattner22eb9722006-06-18 05:43:12 +000022// TODO: Options to support:
23// -fexec-charset,-fwide-exec-charset
24//
25//===----------------------------------------------------------------------===//
26
27#include "clang/Lex/Lexer.h"
28#include "clang/Lex/Preprocessor.h"
Chris Lattner60f36222009-01-29 05:15:15 +000029#include "clang/Lex/LexDiagnostic.h"
Chris Lattnerdc5c0552007-07-20 16:37:10 +000030#include "clang/Basic/SourceManager.h"
Chris Lattner619c1742007-07-22 18:38:25 +000031#include "llvm/Support/Compiler.h"
Chris Lattner739e7392007-04-29 07:12:06 +000032#include "llvm/Support/MemoryBuffer.h"
Chris Lattner22eb9722006-06-18 05:43:12 +000033#include <cctype>
Chris Lattner22eb9722006-06-18 05:43:12 +000034using namespace clang;
35
Chris Lattner3dfff972009-12-17 05:29:40 +000036static void InitCharacterInfo();
Chris Lattner22eb9722006-06-18 05:43:12 +000037
Chris Lattner4894f482007-10-07 08:47:24 +000038//===----------------------------------------------------------------------===//
39// Token Class Implementation
40//===----------------------------------------------------------------------===//
41
Mike Stump11289f42009-09-09 15:08:12 +000042/// isObjCAtKeyword - Return true if we have an ObjC keyword identifier.
Chris Lattner4894f482007-10-07 08:47:24 +000043bool Token::isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const {
Douglas Gregor90abb6d2008-12-01 21:46:47 +000044 if (IdentifierInfo *II = getIdentifierInfo())
45 return II->getObjCKeywordID() == objcKey;
46 return false;
Chris Lattner4894f482007-10-07 08:47:24 +000047}
48
49/// getObjCKeywordID - Return the ObjC keyword kind.
50tok::ObjCKeywordKind Token::getObjCKeywordID() const {
51 IdentifierInfo *specId = getIdentifierInfo();
52 return specId ? specId->getObjCKeywordID() : tok::objc_not_keyword;
53}
54
Chris Lattner67671ed2007-12-13 01:59:49 +000055
Chris Lattner4894f482007-10-07 08:47:24 +000056//===----------------------------------------------------------------------===//
57// Lexer Class Implementation
58//===----------------------------------------------------------------------===//
59
Mike Stump11289f42009-09-09 15:08:12 +000060void Lexer::InitLexer(const char *BufStart, const char *BufPtr,
Chris Lattnerf76b9202009-01-17 06:55:17 +000061 const char *BufEnd) {
Chris Lattner3dfff972009-12-17 05:29:40 +000062 InitCharacterInfo();
Mike Stump11289f42009-09-09 15:08:12 +000063
Chris Lattnerf76b9202009-01-17 06:55:17 +000064 BufferStart = BufStart;
65 BufferPtr = BufPtr;
66 BufferEnd = BufEnd;
Mike Stump11289f42009-09-09 15:08:12 +000067
Chris Lattnerf76b9202009-01-17 06:55:17 +000068 assert(BufEnd[0] == 0 &&
69 "We assume that the input buffer has a null character at the end"
70 " to simplify lexing!");
Mike Stump11289f42009-09-09 15:08:12 +000071
Chris Lattnerf76b9202009-01-17 06:55:17 +000072 Is_PragmaLexer = false;
Chris Lattner7c027ee2009-12-14 06:16:57 +000073 IsInConflictMarker = false;
Douglas Gregor2436e712009-09-17 21:32:03 +000074
Chris Lattnerf76b9202009-01-17 06:55:17 +000075 // Start of the file is a start of line.
76 IsAtStartOfLine = true;
Mike Stump11289f42009-09-09 15:08:12 +000077
Chris Lattnerf76b9202009-01-17 06:55:17 +000078 // We are not after parsing a #.
79 ParsingPreprocessorDirective = false;
Mike Stump11289f42009-09-09 15:08:12 +000080
Chris Lattnerf76b9202009-01-17 06:55:17 +000081 // We are not after parsing #include.
82 ParsingFilename = false;
Mike Stump11289f42009-09-09 15:08:12 +000083
Chris Lattnerf76b9202009-01-17 06:55:17 +000084 // We are not in raw mode. Raw mode disables diagnostics and interpretation
85 // of tokens (e.g. identifiers, thus disabling macro expansion). It is used
86 // to quickly lex the tokens of the buffer, e.g. when handling a "#if 0" block
87 // or otherwise skipping over tokens.
88 LexingRawMode = false;
Mike Stump11289f42009-09-09 15:08:12 +000089
Chris Lattnerf76b9202009-01-17 06:55:17 +000090 // Default to not keeping comments.
91 ExtendedTokenMode = 0;
92}
93
Chris Lattner5965a282009-01-17 07:56:59 +000094/// Lexer constructor - Create a new lexer object for the specified buffer
95/// with the specified preprocessor managing the lexing process. This lexer
96/// assumes that the associated file buffer and Preprocessor objects will
97/// outlive it, so it doesn't take ownership of either of them.
Chris Lattner710bb872009-11-30 04:18:44 +000098Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *InputFile, Preprocessor &PP)
Chris Lattnerc8090892009-01-17 08:03:42 +000099 : PreprocessorLexer(&PP, FID),
100 FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
101 Features(PP.getLangOptions()) {
Mike Stump11289f42009-09-09 15:08:12 +0000102
Chris Lattner5965a282009-01-17 07:56:59 +0000103 InitLexer(InputFile->getBufferStart(), InputFile->getBufferStart(),
104 InputFile->getBufferEnd());
Mike Stump11289f42009-09-09 15:08:12 +0000105
Chris Lattner5965a282009-01-17 07:56:59 +0000106 // Default to keeping comments if the preprocessor wants them.
107 SetCommentRetentionState(PP.getCommentRetentionState());
108}
Chris Lattner4894f482007-10-07 08:47:24 +0000109
Chris Lattner02b436a2007-10-17 20:41:00 +0000110/// Lexer constructor - Create a new raw lexer object. This object is only
Chris Lattner50c90502008-10-12 01:15:46 +0000111/// suitable for calls to 'LexRawToken'. This lexer assumes that the text
112/// range will outlive it, so it doesn't take ownership of it.
Chris Lattner02b436a2007-10-17 20:41:00 +0000113Lexer::Lexer(SourceLocation fileloc, const LangOptions &features,
Chris Lattnerfcf64522009-01-17 07:42:27 +0000114 const char *BufStart, const char *BufPtr, const char *BufEnd)
Chris Lattner1abd2092009-01-17 03:48:08 +0000115 : FileLoc(fileloc), Features(features) {
Chris Lattnerf76b9202009-01-17 06:55:17 +0000116
Chris Lattnerf76b9202009-01-17 06:55:17 +0000117 InitLexer(BufStart, BufPtr, BufEnd);
Mike Stump11289f42009-09-09 15:08:12 +0000118
Chris Lattner02b436a2007-10-17 20:41:00 +0000119 // We *are* in raw mode.
120 LexingRawMode = true;
Chris Lattner02b436a2007-10-17 20:41:00 +0000121}
122
Chris Lattner08354fe2009-01-17 07:35:14 +0000123/// Lexer constructor - Create a new raw lexer object. This object is only
124/// suitable for calls to 'LexRawToken'. This lexer assumes that the text
125/// range will outlive it, so it doesn't take ownership of it.
Chris Lattner710bb872009-11-30 04:18:44 +0000126Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *FromFile,
127 const SourceManager &SM, const LangOptions &features)
Chris Lattner08354fe2009-01-17 07:35:14 +0000128 : FileLoc(SM.getLocForStartOfFile(FID)), Features(features) {
Chris Lattner08354fe2009-01-17 07:35:14 +0000129
Mike Stump11289f42009-09-09 15:08:12 +0000130 InitLexer(FromFile->getBufferStart(), FromFile->getBufferStart(),
Chris Lattner08354fe2009-01-17 07:35:14 +0000131 FromFile->getBufferEnd());
Mike Stump11289f42009-09-09 15:08:12 +0000132
Chris Lattner08354fe2009-01-17 07:35:14 +0000133 // We *are* in raw mode.
134 LexingRawMode = true;
135}
136
Chris Lattner757169b2009-01-17 08:27:52 +0000137/// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
138/// _Pragma expansion. This has a variety of magic semantics that this method
139/// sets up. It returns a new'd Lexer that must be delete'd when done.
140///
141/// On entrance to this routine, TokStartLoc is a macro location which has a
142/// spelling loc that indicates the bytes to be lexed for the token and an
143/// instantiation location that indicates where all lexed tokens should be
144/// "expanded from".
145///
146/// FIXME: It would really be nice to make _Pragma just be a wrapper around a
147/// normal lexer that remaps tokens as they fly by. This would require making
148/// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer
149/// interface that could handle this stuff. This would pull GetMappedTokenLoc
150/// out of the critical path of the lexer!
151///
Mike Stump11289f42009-09-09 15:08:12 +0000152Lexer *Lexer::Create_PragmaLexer(SourceLocation SpellingLoc,
Chris Lattner9dc9c202009-02-15 20:52:18 +0000153 SourceLocation InstantiationLocStart,
154 SourceLocation InstantiationLocEnd,
Chris Lattner29a2a192009-01-19 06:46:35 +0000155 unsigned TokLen, Preprocessor &PP) {
Chris Lattner757169b2009-01-17 08:27:52 +0000156 SourceManager &SM = PP.getSourceManager();
Chris Lattner757169b2009-01-17 08:27:52 +0000157
158 // Create the lexer as if we were going to lex the file normally.
Chris Lattnercbc35ecb2009-01-19 07:46:45 +0000159 FileID SpellingFID = SM.getFileID(SpellingLoc);
Chris Lattner710bb872009-11-30 04:18:44 +0000160 const llvm::MemoryBuffer *InputFile = SM.getBuffer(SpellingFID);
161 Lexer *L = new Lexer(SpellingFID, InputFile, PP);
Mike Stump11289f42009-09-09 15:08:12 +0000162
Chris Lattner757169b2009-01-17 08:27:52 +0000163 // Now that the lexer is created, change the start/end locations so that we
164 // just lex the subsection of the file that we want. This is lexing from a
165 // scratch buffer.
166 const char *StrData = SM.getCharacterData(SpellingLoc);
Mike Stump11289f42009-09-09 15:08:12 +0000167
Chris Lattner757169b2009-01-17 08:27:52 +0000168 L->BufferPtr = StrData;
169 L->BufferEnd = StrData+TokLen;
Chris Lattnerfa217bd2009-03-08 08:08:45 +0000170 assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!");
Chris Lattner757169b2009-01-17 08:27:52 +0000171
172 // Set the SourceLocation with the remapping information. This ensures that
173 // GetMappedTokenLoc will remap the tokens as they are lexed.
Chris Lattner4fa23622009-01-26 00:43:02 +0000174 L->FileLoc = SM.createInstantiationLoc(SM.getLocForStartOfFile(SpellingFID),
Chris Lattner9dc9c202009-02-15 20:52:18 +0000175 InstantiationLocStart,
176 InstantiationLocEnd, TokLen);
Mike Stump11289f42009-09-09 15:08:12 +0000177
Chris Lattner757169b2009-01-17 08:27:52 +0000178 // Ensure that the lexer thinks it is inside a directive, so that end \n will
179 // return an EOM token.
180 L->ParsingPreprocessorDirective = true;
Mike Stump11289f42009-09-09 15:08:12 +0000181
Chris Lattner757169b2009-01-17 08:27:52 +0000182 // This lexer really is for _Pragma.
183 L->Is_PragmaLexer = true;
184 return L;
185}
186
Chris Lattner02b436a2007-10-17 20:41:00 +0000187
Chris Lattnere3e81ea2006-07-03 01:13:26 +0000188/// Stringify - Convert the specified string into a C string, with surrounding
189/// ""'s, and with escaped \ and " characters.
Chris Lattnerecc39e92006-07-15 05:23:31 +0000190std::string Lexer::Stringify(const std::string &Str, bool Charify) {
Chris Lattnere3e81ea2006-07-03 01:13:26 +0000191 std::string Result = Str;
Chris Lattnerecc39e92006-07-15 05:23:31 +0000192 char Quote = Charify ? '\'' : '"';
Chris Lattnere3e81ea2006-07-03 01:13:26 +0000193 for (unsigned i = 0, e = Result.size(); i != e; ++i) {
Chris Lattnerecc39e92006-07-15 05:23:31 +0000194 if (Result[i] == '\\' || Result[i] == Quote) {
Chris Lattnere3e81ea2006-07-03 01:13:26 +0000195 Result.insert(Result.begin()+i, '\\');
196 ++i; ++e;
197 }
198 }
Chris Lattnerecc39e92006-07-15 05:23:31 +0000199 return Result;
Chris Lattnere3e81ea2006-07-03 01:13:26 +0000200}
201
Chris Lattner4c4a2452007-07-24 06:57:14 +0000202/// Stringify - Convert the specified string into a C string by escaping '\'
203/// and " characters. This does not add surrounding ""'s to the string.
204void Lexer::Stringify(llvm::SmallVectorImpl<char> &Str) {
205 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
206 if (Str[i] == '\\' || Str[i] == '"') {
207 Str.insert(Str.begin()+i, '\\');
208 ++i; ++e;
209 }
210 }
211}
212
Chris Lattner22eb9722006-06-18 05:43:12 +0000213
Chris Lattner8e129c22007-10-17 21:18:47 +0000214/// MeasureTokenLength - Relex the token at the specified location and return
215/// its length in bytes in the input file. If the token needs cleaning (e.g.
216/// includes a trigraph or an escaped newline) then this count includes bytes
217/// that are part of that.
218unsigned Lexer::MeasureTokenLength(SourceLocation Loc,
Chris Lattner184e65d2009-04-14 23:22:57 +0000219 const SourceManager &SM,
220 const LangOptions &LangOpts) {
Chris Lattner8e129c22007-10-17 21:18:47 +0000221 // TODO: this could be special cased for common tokens like identifiers, ')',
222 // etc to make this faster, if it mattered. Just look at StrData[0] to handle
Mike Stump11289f42009-09-09 15:08:12 +0000223 // all obviously single-char tokens. This could use
Chris Lattner8e129c22007-10-17 21:18:47 +0000224 // Lexer::isObviouslySimpleCharacter for example to handle identifiers or
225 // something.
Chris Lattner4fa23622009-01-26 00:43:02 +0000226
227 // If this comes from a macro expansion, we really do want the macro name, not
228 // the token this macro expanded to.
Chris Lattnerd3817212009-01-26 22:24:27 +0000229 Loc = SM.getInstantiationLoc(Loc);
230 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
Chris Lattner5509d532009-01-17 08:30:10 +0000231 std::pair<const char *,const char *> Buffer = SM.getBufferData(LocInfo.first);
232 const char *StrData = Buffer.first+LocInfo.second;
233
Chris Lattner8e129c22007-10-17 21:18:47 +0000234 // Create a lexer starting at the beginning of this token.
Chris Lattnerfcf64522009-01-17 07:42:27 +0000235 Lexer TheLexer(Loc, LangOpts, Buffer.first, StrData, Buffer.second);
Chris Lattnera3d4f162009-10-14 15:04:18 +0000236 TheLexer.SetCommentRetentionState(true);
Chris Lattner8e129c22007-10-17 21:18:47 +0000237 Token TheTok;
Chris Lattner50c90502008-10-12 01:15:46 +0000238 TheLexer.LexFromRawLexer(TheTok);
Chris Lattner8e129c22007-10-17 21:18:47 +0000239 return TheTok.getLength();
240}
241
Chris Lattner22eb9722006-06-18 05:43:12 +0000242//===----------------------------------------------------------------------===//
243// Character information.
244//===----------------------------------------------------------------------===//
245
Chris Lattner22eb9722006-06-18 05:43:12 +0000246enum {
247 CHAR_HORZ_WS = 0x01, // ' ', '\t', '\f', '\v'. Note, no '\0'
248 CHAR_VERT_WS = 0x02, // '\r', '\n'
249 CHAR_LETTER = 0x04, // a-z,A-Z
250 CHAR_NUMBER = 0x08, // 0-9
251 CHAR_UNDER = 0x10, // _
252 CHAR_PERIOD = 0x20 // .
253};
254
Chris Lattnerde50a0c2009-07-07 17:09:54 +0000255// Statically initialize CharInfo table based on ASCII character set
256// Reference: FreeBSD 7.2 /usr/share/misc/ascii
Chris Lattner3dfff972009-12-17 05:29:40 +0000257static const unsigned char CharInfo[256] =
Chris Lattnerde50a0c2009-07-07 17:09:54 +0000258{
259// 0 NUL 1 SOH 2 STX 3 ETX
260// 4 EOT 5 ENQ 6 ACK 7 BEL
261 0 , 0 , 0 , 0 ,
262 0 , 0 , 0 , 0 ,
263// 8 BS 9 HT 10 NL 11 VT
264//12 NP 13 CR 14 SO 15 SI
265 0 , CHAR_HORZ_WS, CHAR_VERT_WS, CHAR_HORZ_WS,
266 CHAR_HORZ_WS, CHAR_VERT_WS, 0 , 0 ,
267//16 DLE 17 DC1 18 DC2 19 DC3
268//20 DC4 21 NAK 22 SYN 23 ETB
269 0 , 0 , 0 , 0 ,
270 0 , 0 , 0 , 0 ,
271//24 CAN 25 EM 26 SUB 27 ESC
272//28 FS 29 GS 30 RS 31 US
273 0 , 0 , 0 , 0 ,
274 0 , 0 , 0 , 0 ,
275//32 SP 33 ! 34 " 35 #
276//36 $ 37 % 38 & 39 '
277 CHAR_HORZ_WS, 0 , 0 , 0 ,
278 0 , 0 , 0 , 0 ,
279//40 ( 41 ) 42 * 43 +
280//44 , 45 - 46 . 47 /
281 0 , 0 , 0 , 0 ,
282 0 , 0 , CHAR_PERIOD , 0 ,
283//48 0 49 1 50 2 51 3
284//52 4 53 5 54 6 55 7
285 CHAR_NUMBER , CHAR_NUMBER , CHAR_NUMBER , CHAR_NUMBER ,
286 CHAR_NUMBER , CHAR_NUMBER , CHAR_NUMBER , CHAR_NUMBER ,
287//56 8 57 9 58 : 59 ;
288//60 < 61 = 62 > 63 ?
289 CHAR_NUMBER , CHAR_NUMBER , 0 , 0 ,
290 0 , 0 , 0 , 0 ,
291//64 @ 65 A 66 B 67 C
292//68 D 69 E 70 F 71 G
293 0 , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
294 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
295//72 H 73 I 74 J 75 K
296//76 L 77 M 78 N 79 O
297 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
298 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
299//80 P 81 Q 82 R 83 S
300//84 T 85 U 86 V 87 W
301 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
302 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
303//88 X 89 Y 90 Z 91 [
304//92 \ 93 ] 94 ^ 95 _
305 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , 0 ,
306 0 , 0 , 0 , CHAR_UNDER ,
307//96 ` 97 a 98 b 99 c
308//100 d 101 e 102 f 103 g
309 0 , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
310 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
311//104 h 105 i 106 j 107 k
312//108 l 109 m 110 n 111 o
313 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
314 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
315//112 p 113 q 114 r 115 s
316//116 t 117 u 118 v 119 w
317 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
318 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
319//120 x 121 y 122 z 123 {
320//124 | 125 } 126 ~ 127 DEL
321 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , 0 ,
322 0 , 0 , 0 , 0
323};
324
Chris Lattner3dfff972009-12-17 05:29:40 +0000325static void InitCharacterInfo() {
Chris Lattner22eb9722006-06-18 05:43:12 +0000326 static bool isInited = false;
327 if (isInited) return;
Chris Lattnerde50a0c2009-07-07 17:09:54 +0000328 // check the statically-initialized CharInfo table
329 assert(CHAR_HORZ_WS == CharInfo[(int)' ']);
330 assert(CHAR_HORZ_WS == CharInfo[(int)'\t']);
331 assert(CHAR_HORZ_WS == CharInfo[(int)'\f']);
332 assert(CHAR_HORZ_WS == CharInfo[(int)'\v']);
333 assert(CHAR_VERT_WS == CharInfo[(int)'\n']);
334 assert(CHAR_VERT_WS == CharInfo[(int)'\r']);
335 assert(CHAR_UNDER == CharInfo[(int)'_']);
336 assert(CHAR_PERIOD == CharInfo[(int)'.']);
337 for (unsigned i = 'a'; i <= 'z'; ++i) {
338 assert(CHAR_LETTER == CharInfo[i]);
339 assert(CHAR_LETTER == CharInfo[i+'A'-'a']);
340 }
Chris Lattner22eb9722006-06-18 05:43:12 +0000341 for (unsigned i = '0'; i <= '9'; ++i)
Chris Lattnerde50a0c2009-07-07 17:09:54 +0000342 assert(CHAR_NUMBER == CharInfo[i]);
Steve Naroff04bc0182009-12-08 16:38:12 +0000343
Chris Lattnerde50a0c2009-07-07 17:09:54 +0000344 isInited = true;
Chris Lattner22eb9722006-06-18 05:43:12 +0000345}
346
Chris Lattnerde50a0c2009-07-07 17:09:54 +0000347
Chris Lattner22eb9722006-06-18 05:43:12 +0000348/// isIdentifierBody - Return true if this is the body character of an
349/// identifier, which is [a-zA-Z0-9_].
350static inline bool isIdentifierBody(unsigned char c) {
Hartmut Kaiserc8107e52007-10-18 12:47:01 +0000351 return (CharInfo[c] & (CHAR_LETTER|CHAR_NUMBER|CHAR_UNDER)) ? true : false;
Chris Lattner22eb9722006-06-18 05:43:12 +0000352}
353
354/// isHorizontalWhitespace - Return true if this character is horizontal
355/// whitespace: ' ', '\t', '\f', '\v'. Note that this returns false for '\0'.
356static inline bool isHorizontalWhitespace(unsigned char c) {
Hartmut Kaiserc8107e52007-10-18 12:47:01 +0000357 return (CharInfo[c] & CHAR_HORZ_WS) ? true : false;
Chris Lattner22eb9722006-06-18 05:43:12 +0000358}
359
360/// isWhitespace - Return true if this character is horizontal or vertical
361/// whitespace: ' ', '\t', '\f', '\v', '\n', '\r'. Note that this returns false
362/// for '\0'.
363static inline bool isWhitespace(unsigned char c) {
Hartmut Kaiserc8107e52007-10-18 12:47:01 +0000364 return (CharInfo[c] & (CHAR_HORZ_WS|CHAR_VERT_WS)) ? true : false;
Chris Lattner22eb9722006-06-18 05:43:12 +0000365}
366
367/// isNumberBody - Return true if this is the body character of an
368/// preprocessing number, which is [a-zA-Z0-9_.].
369static inline bool isNumberBody(unsigned char c) {
Mike Stump11289f42009-09-09 15:08:12 +0000370 return (CharInfo[c] & (CHAR_LETTER|CHAR_NUMBER|CHAR_UNDER|CHAR_PERIOD)) ?
Hartmut Kaiserc8107e52007-10-18 12:47:01 +0000371 true : false;
Chris Lattner22eb9722006-06-18 05:43:12 +0000372}
373
Chris Lattnerd01e2912006-06-18 16:22:51 +0000374
Chris Lattner22eb9722006-06-18 05:43:12 +0000375//===----------------------------------------------------------------------===//
376// Diagnostics forwarding code.
377//===----------------------------------------------------------------------===//
378
Chris Lattner619c1742007-07-22 18:38:25 +0000379/// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
380/// lexer buffer was all instantiated at a single point, perform the mapping.
381/// This is currently only used for _Pragma implementation, so it is the slow
382/// path of the hot getSourceLocation method. Do not allow it to be inlined.
Benjamin Kramer5e738282009-11-14 16:36:57 +0000383static DISABLE_INLINE SourceLocation GetMappedTokenLoc(Preprocessor &PP,
384 SourceLocation FileLoc,
385 unsigned CharNo,
386 unsigned TokLen);
Chris Lattner619c1742007-07-22 18:38:25 +0000387static SourceLocation GetMappedTokenLoc(Preprocessor &PP,
388 SourceLocation FileLoc,
Chris Lattner4fa23622009-01-26 00:43:02 +0000389 unsigned CharNo, unsigned TokLen) {
Chris Lattner9dc9c202009-02-15 20:52:18 +0000390 assert(FileLoc.isMacroID() && "Must be an instantiation");
Mike Stump11289f42009-09-09 15:08:12 +0000391
Chris Lattner619c1742007-07-22 18:38:25 +0000392 // Otherwise, we're lexing "mapped tokens". This is used for things like
393 // _Pragma handling. Combine the instantiation location of FileLoc with the
Chris Lattner53e384f2009-01-16 07:00:02 +0000394 // spelling location.
Chris Lattner9dc9c202009-02-15 20:52:18 +0000395 SourceManager &SM = PP.getSourceManager();
Mike Stump11289f42009-09-09 15:08:12 +0000396
Chris Lattner8a425862009-01-16 07:36:28 +0000397 // Create a new SLoc which is expanded from Instantiation(FileLoc) but whose
Chris Lattner53e384f2009-01-16 07:00:02 +0000398 // characters come from spelling(FileLoc)+Offset.
Chris Lattner9dc9c202009-02-15 20:52:18 +0000399 SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
Chris Lattner29a2a192009-01-19 06:46:35 +0000400 SpellingLoc = SpellingLoc.getFileLocWithOffset(CharNo);
Mike Stump11289f42009-09-09 15:08:12 +0000401
Chris Lattner9dc9c202009-02-15 20:52:18 +0000402 // Figure out the expansion loc range, which is the range covered by the
403 // original _Pragma(...) sequence.
404 std::pair<SourceLocation,SourceLocation> II =
405 SM.getImmediateInstantiationRange(FileLoc);
Mike Stump11289f42009-09-09 15:08:12 +0000406
Chris Lattner9dc9c202009-02-15 20:52:18 +0000407 return SM.createInstantiationLoc(SpellingLoc, II.first, II.second, TokLen);
Chris Lattner619c1742007-07-22 18:38:25 +0000408}
409
Chris Lattner22eb9722006-06-18 05:43:12 +0000410/// getSourceLocation - Return a source location identifier for the specified
411/// offset in the current file.
Chris Lattner4fa23622009-01-26 00:43:02 +0000412SourceLocation Lexer::getSourceLocation(const char *Loc,
413 unsigned TokLen) const {
Chris Lattner5d1c0272007-07-22 18:44:36 +0000414 assert(Loc >= BufferStart && Loc <= BufferEnd &&
Chris Lattner4cca5ba2006-07-02 20:05:54 +0000415 "Location out of range for this buffer!");
Chris Lattnerdc5c0552007-07-20 16:37:10 +0000416
417 // In the normal case, we're just lexing from a simple file buffer, return
418 // the file id from FileLoc with the offset specified.
Chris Lattner5d1c0272007-07-22 18:44:36 +0000419 unsigned CharNo = Loc-BufferStart;
Chris Lattnerdc5c0552007-07-20 16:37:10 +0000420 if (FileLoc.isFileID())
Chris Lattner29a2a192009-01-19 06:46:35 +0000421 return FileLoc.getFileLocWithOffset(CharNo);
Mike Stump11289f42009-09-09 15:08:12 +0000422
Chris Lattnerd32480d2009-01-17 06:22:33 +0000423 // Otherwise, this is the _Pragma lexer case, which pretends that all of the
424 // tokens are lexed from where the _Pragma was defined.
Chris Lattner02b436a2007-10-17 20:41:00 +0000425 assert(PP && "This doesn't work on raw lexers");
Chris Lattner4fa23622009-01-26 00:43:02 +0000426 return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
Chris Lattner22eb9722006-06-18 05:43:12 +0000427}
428
Chris Lattner22eb9722006-06-18 05:43:12 +0000429/// Diag - Forwarding function for diagnostics. This translate a source
430/// position in the current buffer into a SourceLocation object for rendering.
Chris Lattner427c9c12008-11-22 00:59:29 +0000431DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
Chris Lattner907dfe92008-11-18 07:59:24 +0000432 return PP->Diag(getSourceLocation(Loc), DiagID);
Chris Lattner22eb9722006-06-18 05:43:12 +0000433}
434
435//===----------------------------------------------------------------------===//
436// Trigraph and Escaped Newline Handling Code.
437//===----------------------------------------------------------------------===//
438
439/// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
440/// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
441static char GetTrigraphCharForLetter(char Letter) {
442 switch (Letter) {
443 default: return 0;
444 case '=': return '#';
445 case ')': return ']';
446 case '(': return '[';
447 case '!': return '|';
448 case '\'': return '^';
449 case '>': return '}';
450 case '/': return '\\';
451 case '<': return '{';
452 case '-': return '~';
453 }
454}
455
456/// DecodeTrigraphChar - If the specified character is a legal trigraph when
457/// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
458/// return the result character. Finally, emit a warning about trigraph use
459/// whether trigraphs are enabled or not.
460static char DecodeTrigraphChar(const char *CP, Lexer *L) {
461 char Res = GetTrigraphCharForLetter(*CP);
Chris Lattner907dfe92008-11-18 07:59:24 +0000462 if (!Res || !L) return Res;
Mike Stump11289f42009-09-09 15:08:12 +0000463
Chris Lattner907dfe92008-11-18 07:59:24 +0000464 if (!L->getFeatures().Trigraphs) {
Chris Lattner6d27a162008-11-22 02:02:22 +0000465 if (!L->isLexingRawMode())
466 L->Diag(CP-2, diag::trigraph_ignored);
Chris Lattner907dfe92008-11-18 07:59:24 +0000467 return 0;
Chris Lattner22eb9722006-06-18 05:43:12 +0000468 }
Mike Stump11289f42009-09-09 15:08:12 +0000469
Chris Lattner6d27a162008-11-22 02:02:22 +0000470 if (!L->isLexingRawMode())
471 L->Diag(CP-2, diag::trigraph_converted) << std::string()+Res;
Chris Lattner22eb9722006-06-18 05:43:12 +0000472 return Res;
473}
474
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000475/// getEscapedNewLineSize - Return the size of the specified escaped newline,
476/// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
Mike Stump11289f42009-09-09 15:08:12 +0000477/// trigraph equivalent on entry to this function.
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000478unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
479 unsigned Size = 0;
480 while (isWhitespace(Ptr[Size])) {
481 ++Size;
Mike Stump11289f42009-09-09 15:08:12 +0000482
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000483 if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
484 continue;
485
486 // If this is a \r\n or \n\r, skip the other half.
487 if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
488 Ptr[Size-1] != Ptr[Size])
489 ++Size;
Mike Stump11289f42009-09-09 15:08:12 +0000490
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000491 return Size;
Mike Stump11289f42009-09-09 15:08:12 +0000492 }
493
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000494 // Not an escaped newline, must be a \t or something else.
495 return 0;
496}
497
Chris Lattner38b2cde2009-04-18 22:27:02 +0000498/// SkipEscapedNewLines - If P points to an escaped newline (or a series of
499/// them), skip over them and return the first non-escaped-newline found,
500/// otherwise return P.
501const char *Lexer::SkipEscapedNewLines(const char *P) {
502 while (1) {
503 const char *AfterEscape;
504 if (*P == '\\') {
505 AfterEscape = P+1;
506 } else if (*P == '?') {
507 // If not a trigraph for escape, bail out.
508 if (P[1] != '?' || P[2] != '/')
509 return P;
510 AfterEscape = P+3;
511 } else {
512 return P;
513 }
Mike Stump11289f42009-09-09 15:08:12 +0000514
Chris Lattner38b2cde2009-04-18 22:27:02 +0000515 unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
516 if (NewLineSize == 0) return P;
517 P = AfterEscape+NewLineSize;
518 }
519}
520
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000521
Chris Lattner22eb9722006-06-18 05:43:12 +0000522/// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
523/// get its size, and return it. This is tricky in several cases:
524/// 1. If currently at the start of a trigraph, we warn about the trigraph,
525/// then either return the trigraph (skipping 3 chars) or the '?',
526/// depending on whether trigraphs are enabled or not.
527/// 2. If this is an escaped newline (potentially with whitespace between
528/// the backslash and newline), implicitly skip the newline and return
529/// the char after it.
Chris Lattner505c5472006-07-03 00:55:48 +0000530/// 3. If this is a UCN, return it. FIXME: C++ UCN's?
Chris Lattner22eb9722006-06-18 05:43:12 +0000531///
532/// This handles the slow/uncommon case of the getCharAndSize method. Here we
533/// know that we can accumulate into Size, and that we have already incremented
534/// Ptr by Size bytes.
535///
Chris Lattnerd01e2912006-06-18 16:22:51 +0000536/// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
537/// be updated to match.
Chris Lattner22eb9722006-06-18 05:43:12 +0000538///
539char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
Chris Lattner146762e2007-07-20 16:59:19 +0000540 Token *Tok) {
Chris Lattner22eb9722006-06-18 05:43:12 +0000541 // If we have a slash, look for an escaped newline.
542 if (Ptr[0] == '\\') {
543 ++Size;
544 ++Ptr;
545Slash:
546 // Common case, backslash-char where the char is not whitespace.
547 if (!isWhitespace(Ptr[0])) return '\\';
Mike Stump11289f42009-09-09 15:08:12 +0000548
Chris Lattnerc1835952009-06-23 05:15:06 +0000549 // See if we have optional whitespace characters between the slash and
550 // newline.
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000551 if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
552 // Remember that this token needs to be cleaned.
553 if (Tok) Tok->setFlag(Token::NeedsCleaning);
Chris Lattner22eb9722006-06-18 05:43:12 +0000554
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000555 // Warn if there was whitespace between the backslash and newline.
Chris Lattnerc1835952009-06-23 05:15:06 +0000556 if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000557 Diag(Ptr, diag::backslash_newline_space);
Mike Stump11289f42009-09-09 15:08:12 +0000558
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000559 // Found backslash<whitespace><newline>. Parse the char after it.
560 Size += EscapedNewLineSize;
561 Ptr += EscapedNewLineSize;
562 // Use slow version to accumulate a correct size field.
563 return getCharAndSizeSlow(Ptr, Size, Tok);
564 }
Mike Stump11289f42009-09-09 15:08:12 +0000565
Chris Lattner22eb9722006-06-18 05:43:12 +0000566 // Otherwise, this is not an escaped newline, just return the slash.
567 return '\\';
568 }
Mike Stump11289f42009-09-09 15:08:12 +0000569
Chris Lattner22eb9722006-06-18 05:43:12 +0000570 // If this is a trigraph, process it.
571 if (Ptr[0] == '?' && Ptr[1] == '?') {
572 // If this is actually a legal trigraph (not something like "??x"), emit
573 // a trigraph warning. If so, and if trigraphs are enabled, return it.
574 if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : 0)) {
575 // Remember that this token needs to be cleaned.
Chris Lattner146762e2007-07-20 16:59:19 +0000576 if (Tok) Tok->setFlag(Token::NeedsCleaning);
Chris Lattner22eb9722006-06-18 05:43:12 +0000577
578 Ptr += 3;
579 Size += 3;
580 if (C == '\\') goto Slash;
581 return C;
582 }
583 }
Mike Stump11289f42009-09-09 15:08:12 +0000584
Chris Lattner22eb9722006-06-18 05:43:12 +0000585 // If this is neither, return a single character.
586 ++Size;
587 return *Ptr;
588}
589
Chris Lattnerd01e2912006-06-18 16:22:51 +0000590
Chris Lattner22eb9722006-06-18 05:43:12 +0000591/// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
592/// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
593/// and that we have already incremented Ptr by Size bytes.
594///
Chris Lattnerd01e2912006-06-18 16:22:51 +0000595/// NOTE: When this method is updated, getCharAndSizeSlow (above) should
596/// be updated to match.
597char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
Chris Lattner22eb9722006-06-18 05:43:12 +0000598 const LangOptions &Features) {
599 // If we have a slash, look for an escaped newline.
600 if (Ptr[0] == '\\') {
601 ++Size;
602 ++Ptr;
603Slash:
604 // Common case, backslash-char where the char is not whitespace.
605 if (!isWhitespace(Ptr[0])) return '\\';
Mike Stump11289f42009-09-09 15:08:12 +0000606
Chris Lattner22eb9722006-06-18 05:43:12 +0000607 // See if we have optional whitespace characters followed by a newline.
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000608 if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
609 // Found backslash<whitespace><newline>. Parse the char after it.
610 Size += EscapedNewLineSize;
611 Ptr += EscapedNewLineSize;
Mike Stump11289f42009-09-09 15:08:12 +0000612
Chris Lattnerfbce7aa2009-04-18 22:05:41 +0000613 // Use slow version to accumulate a correct size field.
614 return getCharAndSizeSlowNoWarn(Ptr, Size, Features);
615 }
Mike Stump11289f42009-09-09 15:08:12 +0000616
Chris Lattner22eb9722006-06-18 05:43:12 +0000617 // Otherwise, this is not an escaped newline, just return the slash.
618 return '\\';
619 }
Mike Stump11289f42009-09-09 15:08:12 +0000620
Chris Lattner22eb9722006-06-18 05:43:12 +0000621 // If this is a trigraph, process it.
622 if (Features.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
623 // If this is actually a legal trigraph (not something like "??x"), return
624 // it.
625 if (char C = GetTrigraphCharForLetter(Ptr[2])) {
626 Ptr += 3;
627 Size += 3;
628 if (C == '\\') goto Slash;
629 return C;
630 }
631 }
Mike Stump11289f42009-09-09 15:08:12 +0000632
Chris Lattner22eb9722006-06-18 05:43:12 +0000633 // If this is neither, return a single character.
634 ++Size;
635 return *Ptr;
636}
637
Chris Lattner22eb9722006-06-18 05:43:12 +0000638//===----------------------------------------------------------------------===//
639// Helper methods for lexing.
640//===----------------------------------------------------------------------===//
641
Chris Lattner146762e2007-07-20 16:59:19 +0000642void Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
Chris Lattner22eb9722006-06-18 05:43:12 +0000643 // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
644 unsigned Size;
645 unsigned char C = *CurPtr++;
Chris Lattner21d9b9a2010-01-11 02:38:50 +0000646 while (isIdentifierBody(C))
Chris Lattner22eb9722006-06-18 05:43:12 +0000647 C = *CurPtr++;
Chris Lattner21d9b9a2010-01-11 02:38:50 +0000648
Chris Lattner22eb9722006-06-18 05:43:12 +0000649 --CurPtr; // Back up over the skipped character.
650
651 // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
652 // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
Chris Lattner505c5472006-07-03 00:55:48 +0000653 // FIXME: UCNs.
Chris Lattner21d9b9a2010-01-11 02:38:50 +0000654 //
655 // TODO: Could merge these checks into a CharInfo flag to make the comparison
656 // cheaper
Chris Lattner22eb9722006-06-18 05:43:12 +0000657 if (C != '\\' && C != '?' && (C != '$' || !Features.DollarIdents)) {
658FinishIdentifier:
Chris Lattnercefc7682006-07-08 08:28:12 +0000659 const char *IdStart = BufferPtr;
Chris Lattnerb11c3232008-10-12 04:51:35 +0000660 FormTokenWithChars(Result, CurPtr, tok::identifier);
Mike Stump11289f42009-09-09 15:08:12 +0000661
Chris Lattner0f1f5052006-07-20 04:16:23 +0000662 // If we are in raw mode, return this identifier raw. There is no need to
663 // look up identifier information or attempt to macro expand it.
664 if (LexingRawMode) return;
Mike Stump11289f42009-09-09 15:08:12 +0000665
Chris Lattnercefc7682006-07-08 08:28:12 +0000666 // Fill in Result.IdentifierInfo, looking up the identifier in the
667 // identifier table.
Chris Lattner8256b972009-01-21 07:45:14 +0000668 IdentifierInfo *II = PP->LookUpIdentifierInfo(Result, IdStart);
Mike Stump11289f42009-09-09 15:08:12 +0000669
Chris Lattner1f6c7fe2009-01-23 18:35:48 +0000670 // Change the kind of this identifier to the appropriate token kind, e.g.
671 // turning "for" into a keyword.
672 Result.setKind(II->getTokenID());
Mike Stump11289f42009-09-09 15:08:12 +0000673
Chris Lattnerc5a00062006-06-18 16:41:01 +0000674 // Finally, now that we know we have an identifier, pass this off to the
675 // preprocessor, which may macro expand it or something.
Chris Lattner8256b972009-01-21 07:45:14 +0000676 if (II->isHandleIdentifierCase())
Chris Lattnerad89ec02009-01-21 07:43:11 +0000677 PP->HandleIdentifier(Result);
678 return;
Chris Lattner22eb9722006-06-18 05:43:12 +0000679 }
Mike Stump11289f42009-09-09 15:08:12 +0000680
Chris Lattner22eb9722006-06-18 05:43:12 +0000681 // Otherwise, $,\,? in identifier found. Enter slower path.
Mike Stump11289f42009-09-09 15:08:12 +0000682
Chris Lattner22eb9722006-06-18 05:43:12 +0000683 C = getCharAndSize(CurPtr, Size);
684 while (1) {
685 if (C == '$') {
686 // If we hit a $ and they are not supported in identifiers, we are done.
687 if (!Features.DollarIdents) goto FinishIdentifier;
Mike Stump11289f42009-09-09 15:08:12 +0000688
Chris Lattner22eb9722006-06-18 05:43:12 +0000689 // Otherwise, emit a diagnostic and continue.
Chris Lattner6d27a162008-11-22 02:02:22 +0000690 if (!isLexingRawMode())
691 Diag(CurPtr, diag::ext_dollar_in_identifier);
Chris Lattner22eb9722006-06-18 05:43:12 +0000692 CurPtr = ConsumeChar(CurPtr, Size, Result);
693 C = getCharAndSize(CurPtr, Size);
694 continue;
Chris Lattner505c5472006-07-03 00:55:48 +0000695 } else if (!isIdentifierBody(C)) { // FIXME: UCNs.
Chris Lattner22eb9722006-06-18 05:43:12 +0000696 // Found end of identifier.
697 goto FinishIdentifier;
698 }
699
700 // Otherwise, this character is good, consume it.
701 CurPtr = ConsumeChar(CurPtr, Size, Result);
702
703 C = getCharAndSize(CurPtr, Size);
Chris Lattner505c5472006-07-03 00:55:48 +0000704 while (isIdentifierBody(C)) { // FIXME: UCNs.
Chris Lattner22eb9722006-06-18 05:43:12 +0000705 CurPtr = ConsumeChar(CurPtr, Size, Result);
706 C = getCharAndSize(CurPtr, Size);
707 }
708 }
709}
710
711
Nate Begeman5eee9332008-04-14 02:26:39 +0000712/// LexNumericConstant - Lex the remainder of a integer or floating point
Chris Lattner22eb9722006-06-18 05:43:12 +0000713/// constant. From[-1] is the first character lexed. Return the end of the
714/// constant.
Chris Lattner146762e2007-07-20 16:59:19 +0000715void Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
Chris Lattner22eb9722006-06-18 05:43:12 +0000716 unsigned Size;
717 char C = getCharAndSize(CurPtr, Size);
718 char PrevCh = 0;
Chris Lattner505c5472006-07-03 00:55:48 +0000719 while (isNumberBody(C)) { // FIXME: UCNs?
Chris Lattner22eb9722006-06-18 05:43:12 +0000720 CurPtr = ConsumeChar(CurPtr, Size, Result);
721 PrevCh = C;
722 C = getCharAndSize(CurPtr, Size);
723 }
Mike Stump11289f42009-09-09 15:08:12 +0000724
Chris Lattner22eb9722006-06-18 05:43:12 +0000725 // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
726 if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e'))
727 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
728
729 // If we have a hex FP constant, continue.
Alexis Hunt91b78382010-01-10 23:37:56 +0000730 if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p') &&
731 (!PP || !PP->getLangOptions().CPlusPlus0x))
Chris Lattner22eb9722006-06-18 05:43:12 +0000732 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
Mike Stump11289f42009-09-09 15:08:12 +0000733
Chris Lattnerd01e2912006-06-18 16:22:51 +0000734 // Update the location of token as well as BufferPtr.
Chris Lattner5a7971e2009-01-26 19:29:26 +0000735 const char *TokStart = BufferPtr;
Chris Lattnerb11c3232008-10-12 04:51:35 +0000736 FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
Chris Lattner5a7971e2009-01-26 19:29:26 +0000737 Result.setLiteralData(TokStart);
Chris Lattner22eb9722006-06-18 05:43:12 +0000738}
739
740/// LexStringLiteral - Lex the remainder of a string literal, after having lexed
741/// either " or L".
Chris Lattner4d963442008-10-12 04:05:48 +0000742void Lexer::LexStringLiteral(Token &Result, const char *CurPtr, bool Wide) {
Chris Lattner22eb9722006-06-18 05:43:12 +0000743 const char *NulCharacter = 0; // Does this string contain the \0 character?
Mike Stump11289f42009-09-09 15:08:12 +0000744
Chris Lattner22eb9722006-06-18 05:43:12 +0000745 char C = getAndAdvanceChar(CurPtr, Result);
746 while (C != '"') {
747 // Skip escaped characters.
748 if (C == '\\') {
749 // Skip the escaped character.
750 C = getAndAdvanceChar(CurPtr, Result);
751 } else if (C == '\n' || C == '\r' || // Newline.
752 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
Chris Lattnerd14705b2009-03-18 21:10:12 +0000753 if (!isLexingRawMode() && !Features.AsmPreprocessor)
Chris Lattner6d27a162008-11-22 02:02:22 +0000754 Diag(BufferPtr, diag::err_unterminated_string);
Chris Lattnerb11c3232008-10-12 04:51:35 +0000755 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
Chris Lattner5a78a022006-07-20 06:02:19 +0000756 return;
Chris Lattner22eb9722006-06-18 05:43:12 +0000757 } else if (C == 0) {
758 NulCharacter = CurPtr-1;
759 }
760 C = getAndAdvanceChar(CurPtr, Result);
761 }
Mike Stump11289f42009-09-09 15:08:12 +0000762
Chris Lattner5a78a022006-07-20 06:02:19 +0000763 // If a nul character existed in the string, warn about it.
Chris Lattner6d27a162008-11-22 02:02:22 +0000764 if (NulCharacter && !isLexingRawMode())
765 Diag(NulCharacter, diag::null_in_string);
Chris Lattner22eb9722006-06-18 05:43:12 +0000766
Chris Lattnerd01e2912006-06-18 16:22:51 +0000767 // Update the location of the token as well as the BufferPtr instance var.
Chris Lattner5a7971e2009-01-26 19:29:26 +0000768 const char *TokStart = BufferPtr;
Chris Lattnerb11c3232008-10-12 04:51:35 +0000769 FormTokenWithChars(Result, CurPtr,
770 Wide ? tok::wide_string_literal : tok::string_literal);
Chris Lattner5a7971e2009-01-26 19:29:26 +0000771 Result.setLiteralData(TokStart);
Chris Lattner22eb9722006-06-18 05:43:12 +0000772}
773
774/// LexAngledStringLiteral - Lex the remainder of an angled string literal,
775/// after having lexed the '<' character. This is used for #include filenames.
Chris Lattner146762e2007-07-20 16:59:19 +0000776void Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
Chris Lattner22eb9722006-06-18 05:43:12 +0000777 const char *NulCharacter = 0; // Does this string contain the \0 character?
Chris Lattnerb40289b2009-04-17 23:56:52 +0000778 const char *AfterLessPos = CurPtr;
Chris Lattner22eb9722006-06-18 05:43:12 +0000779 char C = getAndAdvanceChar(CurPtr, Result);
780 while (C != '>') {
781 // Skip escaped characters.
782 if (C == '\\') {
783 // Skip the escaped character.
784 C = getAndAdvanceChar(CurPtr, Result);
785 } else if (C == '\n' || C == '\r' || // Newline.
786 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
Chris Lattnerb40289b2009-04-17 23:56:52 +0000787 // If the filename is unterminated, then it must just be a lone <
788 // character. Return this as such.
789 FormTokenWithChars(Result, AfterLessPos, tok::less);
Chris Lattner5a78a022006-07-20 06:02:19 +0000790 return;
Chris Lattner22eb9722006-06-18 05:43:12 +0000791 } else if (C == 0) {
792 NulCharacter = CurPtr-1;
793 }
794 C = getAndAdvanceChar(CurPtr, Result);
795 }
Mike Stump11289f42009-09-09 15:08:12 +0000796
Chris Lattner5a78a022006-07-20 06:02:19 +0000797 // If a nul character existed in the string, warn about it.
Chris Lattner6d27a162008-11-22 02:02:22 +0000798 if (NulCharacter && !isLexingRawMode())
799 Diag(NulCharacter, diag::null_in_string);
Mike Stump11289f42009-09-09 15:08:12 +0000800
Chris Lattnerd01e2912006-06-18 16:22:51 +0000801 // Update the location of token as well as BufferPtr.
Chris Lattner5a7971e2009-01-26 19:29:26 +0000802 const char *TokStart = BufferPtr;
Chris Lattnerb11c3232008-10-12 04:51:35 +0000803 FormTokenWithChars(Result, CurPtr, tok::angle_string_literal);
Chris Lattner5a7971e2009-01-26 19:29:26 +0000804 Result.setLiteralData(TokStart);
Chris Lattner22eb9722006-06-18 05:43:12 +0000805}
806
807
808/// LexCharConstant - Lex the remainder of a character constant, after having
809/// lexed either ' or L'.
Chris Lattner146762e2007-07-20 16:59:19 +0000810void Lexer::LexCharConstant(Token &Result, const char *CurPtr) {
Chris Lattner22eb9722006-06-18 05:43:12 +0000811 const char *NulCharacter = 0; // Does this character contain the \0 character?
812
813 // Handle the common case of 'x' and '\y' efficiently.
814 char C = getAndAdvanceChar(CurPtr, Result);
815 if (C == '\'') {
Chris Lattnerd14705b2009-03-18 21:10:12 +0000816 if (!isLexingRawMode() && !Features.AsmPreprocessor)
Chris Lattner6d27a162008-11-22 02:02:22 +0000817 Diag(BufferPtr, diag::err_empty_character);
Chris Lattnerb11c3232008-10-12 04:51:35 +0000818 FormTokenWithChars(Result, CurPtr, tok::unknown);
Chris Lattner5a78a022006-07-20 06:02:19 +0000819 return;
Chris Lattner22eb9722006-06-18 05:43:12 +0000820 } else if (C == '\\') {
821 // Skip the escaped character.
822 // FIXME: UCN's.
823 C = getAndAdvanceChar(CurPtr, Result);
824 }
Mike Stump11289f42009-09-09 15:08:12 +0000825
Chris Lattner22eb9722006-06-18 05:43:12 +0000826 if (C && C != '\n' && C != '\r' && CurPtr[0] == '\'') {
827 ++CurPtr;
828 } else {
829 // Fall back on generic code for embedded nulls, newlines, wide chars.
830 do {
831 // Skip escaped characters.
832 if (C == '\\') {
833 // Skip the escaped character.
834 C = getAndAdvanceChar(CurPtr, Result);
835 } else if (C == '\n' || C == '\r' || // Newline.
836 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
Chris Lattnerd14705b2009-03-18 21:10:12 +0000837 if (!isLexingRawMode() && !Features.AsmPreprocessor)
Chris Lattner6d27a162008-11-22 02:02:22 +0000838 Diag(BufferPtr, diag::err_unterminated_char);
Chris Lattnerb11c3232008-10-12 04:51:35 +0000839 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
Chris Lattner5a78a022006-07-20 06:02:19 +0000840 return;
Chris Lattner22eb9722006-06-18 05:43:12 +0000841 } else if (C == 0) {
842 NulCharacter = CurPtr-1;
843 }
844 C = getAndAdvanceChar(CurPtr, Result);
845 } while (C != '\'');
846 }
Mike Stump11289f42009-09-09 15:08:12 +0000847
Chris Lattner6d27a162008-11-22 02:02:22 +0000848 if (NulCharacter && !isLexingRawMode())
849 Diag(NulCharacter, diag::null_in_char);
Chris Lattner22eb9722006-06-18 05:43:12 +0000850
Chris Lattnerd01e2912006-06-18 16:22:51 +0000851 // Update the location of token as well as BufferPtr.
Chris Lattner5a7971e2009-01-26 19:29:26 +0000852 const char *TokStart = BufferPtr;
Chris Lattnerb11c3232008-10-12 04:51:35 +0000853 FormTokenWithChars(Result, CurPtr, tok::char_constant);
Chris Lattner5a7971e2009-01-26 19:29:26 +0000854 Result.setLiteralData(TokStart);
Chris Lattner22eb9722006-06-18 05:43:12 +0000855}
856
857/// SkipWhitespace - Efficiently skip over a series of whitespace characters.
858/// Update BufferPtr to point to the next non-whitespace character and return.
Chris Lattner4d963442008-10-12 04:05:48 +0000859///
860/// This method forms a token and returns true if KeepWhitespaceMode is enabled.
861///
862bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr) {
Chris Lattner22eb9722006-06-18 05:43:12 +0000863 // Whitespace - Skip it, then return the token after the whitespace.
864 unsigned char Char = *CurPtr; // Skip consequtive spaces efficiently.
865 while (1) {
866 // Skip horizontal whitespace very aggressively.
867 while (isHorizontalWhitespace(Char))
868 Char = *++CurPtr;
Mike Stump11289f42009-09-09 15:08:12 +0000869
Daniel Dunbar5c4cc092008-11-25 00:20:22 +0000870 // Otherwise if we have something other than whitespace, we're done.
Chris Lattner22eb9722006-06-18 05:43:12 +0000871 if (Char != '\n' && Char != '\r')
872 break;
Mike Stump11289f42009-09-09 15:08:12 +0000873
Chris Lattner22eb9722006-06-18 05:43:12 +0000874 if (ParsingPreprocessorDirective) {
875 // End of preprocessor directive line, let LexTokenInternal handle this.
876 BufferPtr = CurPtr;
Chris Lattner4d963442008-10-12 04:05:48 +0000877 return false;
Chris Lattner22eb9722006-06-18 05:43:12 +0000878 }
Mike Stump11289f42009-09-09 15:08:12 +0000879
Chris Lattner22eb9722006-06-18 05:43:12 +0000880 // ok, but handle newline.
881 // The returned token is at the start of the line.
Chris Lattner146762e2007-07-20 16:59:19 +0000882 Result.setFlag(Token::StartOfLine);
Chris Lattner22eb9722006-06-18 05:43:12 +0000883 // No leading whitespace seen so far.
Chris Lattner146762e2007-07-20 16:59:19 +0000884 Result.clearFlag(Token::LeadingSpace);
Chris Lattner22eb9722006-06-18 05:43:12 +0000885 Char = *++CurPtr;
886 }
887
888 // If this isn't immediately after a newline, there is leading space.
889 char PrevChar = CurPtr[-1];
890 if (PrevChar != '\n' && PrevChar != '\r')
Chris Lattner146762e2007-07-20 16:59:19 +0000891 Result.setFlag(Token::LeadingSpace);
Chris Lattner22eb9722006-06-18 05:43:12 +0000892
Chris Lattner4d963442008-10-12 04:05:48 +0000893 // If the client wants us to return whitespace, return it now.
894 if (isKeepWhitespaceMode()) {
Chris Lattnerb11c3232008-10-12 04:51:35 +0000895 FormTokenWithChars(Result, CurPtr, tok::unknown);
Chris Lattner4d963442008-10-12 04:05:48 +0000896 return true;
897 }
Mike Stump11289f42009-09-09 15:08:12 +0000898
Chris Lattner22eb9722006-06-18 05:43:12 +0000899 BufferPtr = CurPtr;
Chris Lattner4d963442008-10-12 04:05:48 +0000900 return false;
Chris Lattner22eb9722006-06-18 05:43:12 +0000901}
902
903// SkipBCPLComment - We have just read the // characters from input. Skip until
904// we find the newline character thats terminate the comment. Then update
Chris Lattnere01e7582008-10-12 04:15:42 +0000905/// BufferPtr and return. If we're in KeepCommentMode, this will form the token
906/// and return true.
Chris Lattner146762e2007-07-20 16:59:19 +0000907bool Lexer::SkipBCPLComment(Token &Result, const char *CurPtr) {
Chris Lattner22eb9722006-06-18 05:43:12 +0000908 // If BCPL comments aren't explicitly enabled for this language, emit an
909 // extension warning.
Chris Lattner6d27a162008-11-22 02:02:22 +0000910 if (!Features.BCPLComment && !isLexingRawMode()) {
Chris Lattnerd01e2912006-06-18 16:22:51 +0000911 Diag(BufferPtr, diag::ext_bcpl_comment);
Mike Stump11289f42009-09-09 15:08:12 +0000912
Chris Lattner22eb9722006-06-18 05:43:12 +0000913 // Mark them enabled so we only emit one warning for this translation
914 // unit.
915 Features.BCPLComment = true;
916 }
Mike Stump11289f42009-09-09 15:08:12 +0000917
Chris Lattner22eb9722006-06-18 05:43:12 +0000918 // Scan over the body of the comment. The common case, when scanning, is that
919 // the comment contains normal ascii characters with nothing interesting in
920 // them. As such, optimize for this case with the inner loop.
921 char C;
922 do {
923 C = *CurPtr;
Chris Lattner505c5472006-07-03 00:55:48 +0000924 // FIXME: Speedup BCPL comment lexing. Just scan for a \n or \r character.
925 // If we find a \n character, scan backwards, checking to see if it's an
926 // escaped newline, like we do for block comments.
Mike Stump11289f42009-09-09 15:08:12 +0000927
Chris Lattner22eb9722006-06-18 05:43:12 +0000928 // Skip over characters in the fast loop.
929 while (C != 0 && // Potentially EOF.
930 C != '\\' && // Potentially escaped newline.
931 C != '?' && // Potentially trigraph.
932 C != '\n' && C != '\r') // Newline or DOS-style newline.
933 C = *++CurPtr;
934
935 // If this is a newline, we're done.
936 if (C == '\n' || C == '\r')
937 break; // Found the newline? Break out!
Mike Stump11289f42009-09-09 15:08:12 +0000938
Chris Lattner22eb9722006-06-18 05:43:12 +0000939 // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
Chris Lattnere141a9e2008-12-12 07:34:39 +0000940 // properly decode the character. Read it in raw mode to avoid emitting
941 // diagnostics about things like trigraphs. If we see an escaped newline,
942 // we'll handle it below.
Chris Lattner22eb9722006-06-18 05:43:12 +0000943 const char *OldPtr = CurPtr;
Chris Lattnere141a9e2008-12-12 07:34:39 +0000944 bool OldRawMode = isLexingRawMode();
945 LexingRawMode = true;
Chris Lattner22eb9722006-06-18 05:43:12 +0000946 C = getAndAdvanceChar(CurPtr, Result);
Chris Lattnere141a9e2008-12-12 07:34:39 +0000947 LexingRawMode = OldRawMode;
Chris Lattnerecdaf402009-04-05 00:26:41 +0000948
949 // If the char that we finally got was a \n, then we must have had something
950 // like \<newline><newline>. We don't want to have consumed the second
951 // newline, we want CurPtr, to end up pointing to it down below.
952 if (C == '\n' || C == '\r') {
953 --CurPtr;
954 C = 'x'; // doesn't matter what this is.
955 }
Mike Stump11289f42009-09-09 15:08:12 +0000956
Chris Lattner22eb9722006-06-18 05:43:12 +0000957 // If we read multiple characters, and one of those characters was a \r or
Chris Lattnerff591e22007-06-09 06:07:22 +0000958 // \n, then we had an escaped newline within the comment. Emit diagnostic
959 // unless the next line is also a // comment.
960 if (CurPtr != OldPtr+1 && C != '/' && CurPtr[0] != '/') {
Chris Lattner22eb9722006-06-18 05:43:12 +0000961 for (; OldPtr != CurPtr; ++OldPtr)
962 if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
Chris Lattnerff591e22007-06-09 06:07:22 +0000963 // Okay, we found a // comment that ends in a newline, if the next
964 // line is also a // comment, but has spaces, don't emit a diagnostic.
965 if (isspace(C)) {
966 const char *ForwardPtr = CurPtr;
967 while (isspace(*ForwardPtr)) // Skip whitespace.
968 ++ForwardPtr;
969 if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
970 break;
971 }
Mike Stump11289f42009-09-09 15:08:12 +0000972
Chris Lattner6d27a162008-11-22 02:02:22 +0000973 if (!isLexingRawMode())
974 Diag(OldPtr-1, diag::ext_multi_line_bcpl_comment);
Chris Lattnercb283342006-06-18 06:48:37 +0000975 break;
Chris Lattner22eb9722006-06-18 05:43:12 +0000976 }
977 }
Mike Stump11289f42009-09-09 15:08:12 +0000978
Chris Lattner457fc152006-07-29 06:30:25 +0000979 if (CurPtr == BufferEnd+1) { --CurPtr; break; }
Chris Lattner22eb9722006-06-18 05:43:12 +0000980 } while (C != '\n' && C != '\r');
981
Chris Lattner457fc152006-07-29 06:30:25 +0000982 // Found but did not consume the newline.
Douglas Gregorc6d5edd2009-07-02 17:08:52 +0000983 if (PP)
Mike Stump11289f42009-09-09 15:08:12 +0000984 PP->HandleComment(SourceRange(getSourceLocation(BufferPtr),
Douglas Gregorc6d5edd2009-07-02 17:08:52 +0000985 getSourceLocation(CurPtr)));
Mike Stump11289f42009-09-09 15:08:12 +0000986
Chris Lattner457fc152006-07-29 06:30:25 +0000987 // If we are returning comments as tokens, return this comment as a token.
Chris Lattner8637abd2008-10-12 03:22:02 +0000988 if (inKeepCommentMode())
Chris Lattner457fc152006-07-29 06:30:25 +0000989 return SaveBCPLComment(Result, CurPtr);
Chris Lattner22eb9722006-06-18 05:43:12 +0000990
991 // If we are inside a preprocessor directive and we see the end of line,
992 // return immediately, so that the lexer can return this as an EOM token.
Chris Lattner457fc152006-07-29 06:30:25 +0000993 if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
Chris Lattner22eb9722006-06-18 05:43:12 +0000994 BufferPtr = CurPtr;
Chris Lattnere01e7582008-10-12 04:15:42 +0000995 return false;
Chris Lattner22eb9722006-06-18 05:43:12 +0000996 }
Mike Stump11289f42009-09-09 15:08:12 +0000997
Chris Lattner22eb9722006-06-18 05:43:12 +0000998 // Otherwise, eat the \n character. We don't care if this is a \n\r or
Chris Lattner87e97ea2008-10-12 00:23:07 +0000999 // \r\n sequence. This is an efficiency hack (because we know the \n can't
Chris Lattner4d963442008-10-12 04:05:48 +00001000 // contribute to another token), it isn't needed for correctness. Note that
1001 // this is ok even in KeepWhitespaceMode, because we would have returned the
1002 /// comment above in that mode.
Chris Lattner22eb9722006-06-18 05:43:12 +00001003 ++CurPtr;
Mike Stump11289f42009-09-09 15:08:12 +00001004
Chris Lattner22eb9722006-06-18 05:43:12 +00001005 // The next returned token is at the start of the line.
Chris Lattner146762e2007-07-20 16:59:19 +00001006 Result.setFlag(Token::StartOfLine);
Chris Lattner22eb9722006-06-18 05:43:12 +00001007 // No leading whitespace seen so far.
Chris Lattner146762e2007-07-20 16:59:19 +00001008 Result.clearFlag(Token::LeadingSpace);
Chris Lattner22eb9722006-06-18 05:43:12 +00001009 BufferPtr = CurPtr;
Chris Lattnere01e7582008-10-12 04:15:42 +00001010 return false;
Chris Lattner457fc152006-07-29 06:30:25 +00001011}
Chris Lattner22eb9722006-06-18 05:43:12 +00001012
Chris Lattner457fc152006-07-29 06:30:25 +00001013/// SaveBCPLComment - If in save-comment mode, package up this BCPL comment in
1014/// an appropriate way and return it.
Chris Lattner146762e2007-07-20 16:59:19 +00001015bool Lexer::SaveBCPLComment(Token &Result, const char *CurPtr) {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001016 // If we're not in a preprocessor directive, just return the // comment
1017 // directly.
1018 FormTokenWithChars(Result, CurPtr, tok::comment);
Mike Stump11289f42009-09-09 15:08:12 +00001019
Chris Lattnerb11c3232008-10-12 04:51:35 +00001020 if (!ParsingPreprocessorDirective)
1021 return true;
Mike Stump11289f42009-09-09 15:08:12 +00001022
Chris Lattnerb11c3232008-10-12 04:51:35 +00001023 // If this BCPL-style comment is in a macro definition, transmogrify it into
1024 // a C-style block comment.
1025 std::string Spelling = PP->getSpelling(Result);
1026 assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not bcpl comment?");
1027 Spelling[1] = '*'; // Change prefix to "/*".
1028 Spelling += "*/"; // add suffix.
Mike Stump11289f42009-09-09 15:08:12 +00001029
Chris Lattnerb11c3232008-10-12 04:51:35 +00001030 Result.setKind(tok::comment);
Chris Lattner5a7971e2009-01-26 19:29:26 +00001031 PP->CreateString(&Spelling[0], Spelling.size(), Result,
1032 Result.getLocation());
Chris Lattnere01e7582008-10-12 04:15:42 +00001033 return true;
Chris Lattner22eb9722006-06-18 05:43:12 +00001034}
1035
Chris Lattnercb283342006-06-18 06:48:37 +00001036/// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
1037/// character (either \n or \r) is part of an escaped newline sequence. Issue a
Chris Lattner89770572008-12-12 07:14:34 +00001038/// diagnostic if so. We know that the newline is inside of a block comment.
Mike Stump11289f42009-09-09 15:08:12 +00001039static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
Chris Lattner1f583052006-06-18 06:53:56 +00001040 Lexer *L) {
Chris Lattner22eb9722006-06-18 05:43:12 +00001041 assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
Mike Stump11289f42009-09-09 15:08:12 +00001042
Chris Lattner22eb9722006-06-18 05:43:12 +00001043 // Back up off the newline.
1044 --CurPtr;
Mike Stump11289f42009-09-09 15:08:12 +00001045
Chris Lattner22eb9722006-06-18 05:43:12 +00001046 // If this is a two-character newline sequence, skip the other character.
1047 if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
1048 // \n\n or \r\r -> not escaped newline.
1049 if (CurPtr[0] == CurPtr[1])
1050 return false;
1051 // \n\r or \r\n -> skip the newline.
1052 --CurPtr;
1053 }
Mike Stump11289f42009-09-09 15:08:12 +00001054
Chris Lattner22eb9722006-06-18 05:43:12 +00001055 // If we have horizontal whitespace, skip over it. We allow whitespace
1056 // between the slash and newline.
1057 bool HasSpace = false;
1058 while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
1059 --CurPtr;
1060 HasSpace = true;
1061 }
Mike Stump11289f42009-09-09 15:08:12 +00001062
Chris Lattner22eb9722006-06-18 05:43:12 +00001063 // If we have a slash, we know this is an escaped newline.
1064 if (*CurPtr == '\\') {
Chris Lattnercb283342006-06-18 06:48:37 +00001065 if (CurPtr[-1] != '*') return false;
Chris Lattner22eb9722006-06-18 05:43:12 +00001066 } else {
1067 // It isn't a slash, is it the ?? / trigraph?
Chris Lattnercb283342006-06-18 06:48:37 +00001068 if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
1069 CurPtr[-3] != '*')
Chris Lattner22eb9722006-06-18 05:43:12 +00001070 return false;
Mike Stump11289f42009-09-09 15:08:12 +00001071
Chris Lattnercb283342006-06-18 06:48:37 +00001072 // This is the trigraph ending the comment. Emit a stern warning!
Chris Lattner22eb9722006-06-18 05:43:12 +00001073 CurPtr -= 2;
1074
1075 // If no trigraphs are enabled, warn that we ignored this trigraph and
1076 // ignore this * character.
Chris Lattner1f583052006-06-18 06:53:56 +00001077 if (!L->getFeatures().Trigraphs) {
Chris Lattner6d27a162008-11-22 02:02:22 +00001078 if (!L->isLexingRawMode())
1079 L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
Chris Lattnercb283342006-06-18 06:48:37 +00001080 return false;
Chris Lattner22eb9722006-06-18 05:43:12 +00001081 }
Chris Lattner6d27a162008-11-22 02:02:22 +00001082 if (!L->isLexingRawMode())
1083 L->Diag(CurPtr, diag::trigraph_ends_block_comment);
Chris Lattner22eb9722006-06-18 05:43:12 +00001084 }
Mike Stump11289f42009-09-09 15:08:12 +00001085
Chris Lattner22eb9722006-06-18 05:43:12 +00001086 // Warn about having an escaped newline between the */ characters.
Chris Lattner6d27a162008-11-22 02:02:22 +00001087 if (!L->isLexingRawMode())
1088 L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
Mike Stump11289f42009-09-09 15:08:12 +00001089
Chris Lattner22eb9722006-06-18 05:43:12 +00001090 // If there was space between the backslash and newline, warn about it.
Chris Lattner6d27a162008-11-22 02:02:22 +00001091 if (HasSpace && !L->isLexingRawMode())
1092 L->Diag(CurPtr, diag::backslash_newline_space);
Mike Stump11289f42009-09-09 15:08:12 +00001093
Chris Lattnercb283342006-06-18 06:48:37 +00001094 return true;
Chris Lattner22eb9722006-06-18 05:43:12 +00001095}
1096
Chris Lattneraded4a92006-10-27 04:42:31 +00001097#ifdef __SSE2__
1098#include <emmintrin.h>
Chris Lattner9f6604f2006-10-30 20:01:22 +00001099#elif __ALTIVEC__
1100#include <altivec.h>
1101#undef bool
Chris Lattneraded4a92006-10-27 04:42:31 +00001102#endif
1103
Chris Lattner22eb9722006-06-18 05:43:12 +00001104/// SkipBlockComment - We have just read the /* characters from input. Read
1105/// until we find the */ characters that terminate the comment. Note that we
1106/// don't bother decoding trigraphs or escaped newlines in block comments,
1107/// because they cannot cause the comment to end. The only thing that can
1108/// happen is the comment could end with an escaped newline between the */ end
1109/// of comment.
Chris Lattnere01e7582008-10-12 04:15:42 +00001110///
1111/// If KeepCommentMode is enabled, this forms a token from the comment and
1112/// returns true.
Chris Lattner146762e2007-07-20 16:59:19 +00001113bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr) {
Chris Lattner22eb9722006-06-18 05:43:12 +00001114 // Scan one character past where we should, looking for a '/' character. Once
1115 // we find it, check to see if it was preceeded by a *. This common
1116 // optimization helps people who like to put a lot of * characters in their
1117 // comments.
Chris Lattnerc850ad62007-07-21 23:43:37 +00001118
1119 // The first character we get with newlines and trigraphs skipped to handle
1120 // the degenerate /*/ case below correctly if the * has an escaped newline
1121 // after it.
1122 unsigned CharSize;
1123 unsigned char C = getCharAndSize(CurPtr, CharSize);
1124 CurPtr += CharSize;
Chris Lattner22eb9722006-06-18 05:43:12 +00001125 if (C == 0 && CurPtr == BufferEnd+1) {
Chris Lattner6d27a162008-11-22 02:02:22 +00001126 if (!isLexingRawMode())
Chris Lattner7c2e9802008-10-12 01:31:51 +00001127 Diag(BufferPtr, diag::err_unterminated_block_comment);
Chris Lattner99e7d232008-10-12 04:19:49 +00001128 --CurPtr;
Mike Stump11289f42009-09-09 15:08:12 +00001129
Chris Lattner99e7d232008-10-12 04:19:49 +00001130 // KeepWhitespaceMode should return this broken comment as a token. Since
1131 // it isn't a well formed comment, just return it as an 'unknown' token.
1132 if (isKeepWhitespaceMode()) {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001133 FormTokenWithChars(Result, CurPtr, tok::unknown);
Chris Lattner99e7d232008-10-12 04:19:49 +00001134 return true;
1135 }
Mike Stump11289f42009-09-09 15:08:12 +00001136
Chris Lattner99e7d232008-10-12 04:19:49 +00001137 BufferPtr = CurPtr;
Chris Lattnere01e7582008-10-12 04:15:42 +00001138 return false;
Chris Lattner22eb9722006-06-18 05:43:12 +00001139 }
Mike Stump11289f42009-09-09 15:08:12 +00001140
Chris Lattnerc850ad62007-07-21 23:43:37 +00001141 // Check to see if the first character after the '/*' is another /. If so,
1142 // then this slash does not end the block comment, it is part of it.
1143 if (C == '/')
1144 C = *CurPtr++;
Mike Stump11289f42009-09-09 15:08:12 +00001145
Chris Lattner22eb9722006-06-18 05:43:12 +00001146 while (1) {
Chris Lattner6cc3e362006-10-27 04:12:35 +00001147 // Skip over all non-interesting characters until we find end of buffer or a
1148 // (probably ending) '/' character.
Chris Lattner6cc3e362006-10-27 04:12:35 +00001149 if (CurPtr + 24 < BufferEnd) {
1150 // While not aligned to a 16-byte boundary.
1151 while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
1152 C = *CurPtr++;
Mike Stump11289f42009-09-09 15:08:12 +00001153
Chris Lattner6cc3e362006-10-27 04:12:35 +00001154 if (C == '/') goto FoundSlash;
Chris Lattneraded4a92006-10-27 04:42:31 +00001155
1156#ifdef __SSE2__
1157 __m128i Slashes = _mm_set_epi8('/', '/', '/', '/', '/', '/', '/', '/',
1158 '/', '/', '/', '/', '/', '/', '/', '/');
1159 while (CurPtr+16 <= BufferEnd &&
1160 _mm_movemask_epi8(_mm_cmpeq_epi8(*(__m128i*)CurPtr, Slashes)) == 0)
1161 CurPtr += 16;
Chris Lattner9f6604f2006-10-30 20:01:22 +00001162#elif __ALTIVEC__
1163 __vector unsigned char Slashes = {
Mike Stump11289f42009-09-09 15:08:12 +00001164 '/', '/', '/', '/', '/', '/', '/', '/',
Chris Lattner9f6604f2006-10-30 20:01:22 +00001165 '/', '/', '/', '/', '/', '/', '/', '/'
1166 };
1167 while (CurPtr+16 <= BufferEnd &&
1168 !vec_any_eq(*(vector unsigned char*)CurPtr, Slashes))
1169 CurPtr += 16;
Mike Stump11289f42009-09-09 15:08:12 +00001170#else
Chris Lattneraded4a92006-10-27 04:42:31 +00001171 // Scan for '/' quickly. Many block comments are very large.
Chris Lattner6cc3e362006-10-27 04:12:35 +00001172 while (CurPtr[0] != '/' &&
1173 CurPtr[1] != '/' &&
1174 CurPtr[2] != '/' &&
1175 CurPtr[3] != '/' &&
1176 CurPtr+4 < BufferEnd) {
1177 CurPtr += 4;
1178 }
Chris Lattneraded4a92006-10-27 04:42:31 +00001179#endif
Mike Stump11289f42009-09-09 15:08:12 +00001180
Chris Lattneraded4a92006-10-27 04:42:31 +00001181 // It has to be one of the bytes scanned, increment to it and read one.
Chris Lattner6cc3e362006-10-27 04:12:35 +00001182 C = *CurPtr++;
1183 }
Mike Stump11289f42009-09-09 15:08:12 +00001184
Chris Lattneraded4a92006-10-27 04:42:31 +00001185 // Loop to scan the remainder.
Chris Lattner22eb9722006-06-18 05:43:12 +00001186 while (C != '/' && C != '\0')
1187 C = *CurPtr++;
Mike Stump11289f42009-09-09 15:08:12 +00001188
Chris Lattner6cc3e362006-10-27 04:12:35 +00001189 FoundSlash:
Chris Lattner22eb9722006-06-18 05:43:12 +00001190 if (C == '/') {
Chris Lattner22eb9722006-06-18 05:43:12 +00001191 if (CurPtr[-2] == '*') // We found the final */. We're done!
1192 break;
Mike Stump11289f42009-09-09 15:08:12 +00001193
Chris Lattner22eb9722006-06-18 05:43:12 +00001194 if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
Chris Lattner1f583052006-06-18 06:53:56 +00001195 if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
Chris Lattner22eb9722006-06-18 05:43:12 +00001196 // We found the final */, though it had an escaped newline between the
1197 // * and /. We're done!
1198 break;
1199 }
1200 }
1201 if (CurPtr[0] == '*' && CurPtr[1] != '/') {
1202 // If this is a /* inside of the comment, emit a warning. Don't do this
1203 // if this is a /*/, which will end the comment. This misses cases with
1204 // embedded escaped newlines, but oh well.
Chris Lattner6d27a162008-11-22 02:02:22 +00001205 if (!isLexingRawMode())
1206 Diag(CurPtr-1, diag::warn_nested_block_comment);
Chris Lattner22eb9722006-06-18 05:43:12 +00001207 }
1208 } else if (C == 0 && CurPtr == BufferEnd+1) {
Chris Lattner6d27a162008-11-22 02:02:22 +00001209 if (!isLexingRawMode())
1210 Diag(BufferPtr, diag::err_unterminated_block_comment);
Chris Lattner22eb9722006-06-18 05:43:12 +00001211 // Note: the user probably forgot a */. We could continue immediately
1212 // after the /*, but this would involve lexing a lot of what really is the
1213 // comment, which surely would confuse the parser.
Chris Lattner99e7d232008-10-12 04:19:49 +00001214 --CurPtr;
Mike Stump11289f42009-09-09 15:08:12 +00001215
Chris Lattner99e7d232008-10-12 04:19:49 +00001216 // KeepWhitespaceMode should return this broken comment as a token. Since
1217 // it isn't a well formed comment, just return it as an 'unknown' token.
1218 if (isKeepWhitespaceMode()) {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001219 FormTokenWithChars(Result, CurPtr, tok::unknown);
Chris Lattner99e7d232008-10-12 04:19:49 +00001220 return true;
1221 }
Mike Stump11289f42009-09-09 15:08:12 +00001222
Chris Lattner99e7d232008-10-12 04:19:49 +00001223 BufferPtr = CurPtr;
Chris Lattnere01e7582008-10-12 04:15:42 +00001224 return false;
Chris Lattner22eb9722006-06-18 05:43:12 +00001225 }
1226 C = *CurPtr++;
1227 }
Mike Stump11289f42009-09-09 15:08:12 +00001228
1229 if (PP)
1230 PP->HandleComment(SourceRange(getSourceLocation(BufferPtr),
Douglas Gregorc6d5edd2009-07-02 17:08:52 +00001231 getSourceLocation(CurPtr)));
1232
Chris Lattner457fc152006-07-29 06:30:25 +00001233 // If we are returning comments as tokens, return this comment as a token.
Chris Lattner8637abd2008-10-12 03:22:02 +00001234 if (inKeepCommentMode()) {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001235 FormTokenWithChars(Result, CurPtr, tok::comment);
Chris Lattnere01e7582008-10-12 04:15:42 +00001236 return true;
Chris Lattner457fc152006-07-29 06:30:25 +00001237 }
Chris Lattner22eb9722006-06-18 05:43:12 +00001238
1239 // It is common for the tokens immediately after a /**/ comment to be
1240 // whitespace. Instead of going through the big switch, handle it
Chris Lattner4d963442008-10-12 04:05:48 +00001241 // efficiently now. This is safe even in KeepWhitespaceMode because we would
1242 // have already returned above with the comment as a token.
Chris Lattner22eb9722006-06-18 05:43:12 +00001243 if (isHorizontalWhitespace(*CurPtr)) {
Chris Lattner146762e2007-07-20 16:59:19 +00001244 Result.setFlag(Token::LeadingSpace);
Chris Lattner457fc152006-07-29 06:30:25 +00001245 SkipWhitespace(Result, CurPtr+1);
Chris Lattnere01e7582008-10-12 04:15:42 +00001246 return false;
Chris Lattner22eb9722006-06-18 05:43:12 +00001247 }
1248
1249 // Otherwise, just return so that the next character will be lexed as a token.
1250 BufferPtr = CurPtr;
Chris Lattner146762e2007-07-20 16:59:19 +00001251 Result.setFlag(Token::LeadingSpace);
Chris Lattnere01e7582008-10-12 04:15:42 +00001252 return false;
Chris Lattner22eb9722006-06-18 05:43:12 +00001253}
1254
1255//===----------------------------------------------------------------------===//
1256// Primary Lexing Entry Points
1257//===----------------------------------------------------------------------===//
1258
Chris Lattner22eb9722006-06-18 05:43:12 +00001259/// ReadToEndOfLine - Read the rest of the current preprocessor line as an
1260/// uninterpreted string. This switches the lexer out of directive mode.
1261std::string Lexer::ReadToEndOfLine() {
1262 assert(ParsingPreprocessorDirective && ParsingFilename == false &&
1263 "Must be in a preprocessing directive!");
1264 std::string Result;
Chris Lattner146762e2007-07-20 16:59:19 +00001265 Token Tmp;
Chris Lattner22eb9722006-06-18 05:43:12 +00001266
1267 // CurPtr - Cache BufferPtr in an automatic variable.
1268 const char *CurPtr = BufferPtr;
Chris Lattner22eb9722006-06-18 05:43:12 +00001269 while (1) {
1270 char Char = getAndAdvanceChar(CurPtr, Tmp);
1271 switch (Char) {
1272 default:
1273 Result += Char;
1274 break;
1275 case 0: // Null.
1276 // Found end of file?
1277 if (CurPtr-1 != BufferEnd) {
1278 // Nope, normal character, continue.
1279 Result += Char;
1280 break;
1281 }
1282 // FALL THROUGH.
1283 case '\r':
1284 case '\n':
1285 // Okay, we found the end of the line. First, back up past the \0, \r, \n.
1286 assert(CurPtr[-1] == Char && "Trigraphs for newline?");
1287 BufferPtr = CurPtr-1;
Mike Stump11289f42009-09-09 15:08:12 +00001288
Chris Lattner22eb9722006-06-18 05:43:12 +00001289 // Next, lex the character, which should handle the EOM transition.
Chris Lattnercb283342006-06-18 06:48:37 +00001290 Lex(Tmp);
Chris Lattner98c1f7c2007-10-09 18:02:16 +00001291 assert(Tmp.is(tok::eom) && "Unexpected token!");
Mike Stump11289f42009-09-09 15:08:12 +00001292
Chris Lattner22eb9722006-06-18 05:43:12 +00001293 // Finally, we're done, return the string we found.
1294 return Result;
1295 }
1296 }
1297}
1298
1299/// LexEndOfFile - CurPtr points to the end of this file. Handle this
1300/// condition, reporting diagnostics and handling other edge cases as required.
Chris Lattner2183a6e2006-07-18 06:36:12 +00001301/// This returns true if Result contains a token, false if PP.Lex should be
1302/// called again.
Chris Lattner146762e2007-07-20 16:59:19 +00001303bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
Chris Lattner22eb9722006-06-18 05:43:12 +00001304 // If we hit the end of the file while parsing a preprocessor directive,
1305 // end the preprocessor directive first. The next token returned will
1306 // then be the end of file.
1307 if (ParsingPreprocessorDirective) {
1308 // Done parsing the "line".
1309 ParsingPreprocessorDirective = false;
Chris Lattnerd01e2912006-06-18 16:22:51 +00001310 // Update the location of token as well as BufferPtr.
Chris Lattnerb11c3232008-10-12 04:51:35 +00001311 FormTokenWithChars(Result, CurPtr, tok::eom);
Mike Stump11289f42009-09-09 15:08:12 +00001312
Chris Lattner457fc152006-07-29 06:30:25 +00001313 // Restore comment saving mode, in case it was disabled for directive.
Chris Lattner097a8b82008-10-12 03:27:19 +00001314 SetCommentRetentionState(PP->getCommentRetentionState());
Chris Lattner2183a6e2006-07-18 06:36:12 +00001315 return true; // Have a token.
Mike Stump11289f42009-09-09 15:08:12 +00001316 }
Douglas Gregor3545ff42009-09-21 16:56:56 +00001317
Chris Lattner30a2fa12006-07-19 06:31:49 +00001318 // If we are in raw mode, return this event as an EOF token. Let the caller
1319 // that put us in raw mode handle the event.
Chris Lattner6d27a162008-11-22 02:02:22 +00001320 if (isLexingRawMode()) {
Chris Lattner8c204872006-10-14 05:19:21 +00001321 Result.startToken();
Chris Lattner30a2fa12006-07-19 06:31:49 +00001322 BufferPtr = BufferEnd;
Chris Lattnerb11c3232008-10-12 04:51:35 +00001323 FormTokenWithChars(Result, BufferEnd, tok::eof);
Chris Lattner30a2fa12006-07-19 06:31:49 +00001324 return true;
Chris Lattnerd8aee0e2006-07-11 05:04:55 +00001325 }
Mike Stump11289f42009-09-09 15:08:12 +00001326
Douglas Gregor3545ff42009-09-21 16:56:56 +00001327 // Otherwise, check if we are code-completing, then issue diagnostics for
1328 // unterminated #if and missing newline.
Chris Lattner30a2fa12006-07-19 06:31:49 +00001329
Douglas Gregor53ad6b92009-12-02 06:49:09 +00001330 if (PP && PP->isCodeCompletionFile(FileLoc)) {
1331 // We're at the end of the file, but we've been asked to consider the
1332 // end of the file to be a code-completion token. Return the
1333 // code-completion token.
1334 Result.startToken();
1335 FormTokenWithChars(Result, CurPtr, tok::code_completion);
Douglas Gregor3545ff42009-09-21 16:56:56 +00001336
Douglas Gregor53ad6b92009-12-02 06:49:09 +00001337 // Only do the eof -> code_completion translation once.
1338 PP->SetCodeCompletionPoint(0, 0, 0);
1339 return true;
Douglas Gregor3545ff42009-09-21 16:56:56 +00001340 }
1341
Chris Lattner30a2fa12006-07-19 06:31:49 +00001342 // If we are in a #if directive, emit an error.
1343 while (!ConditionalStack.empty()) {
Chris Lattner014156e2008-11-22 06:22:39 +00001344 PP->Diag(ConditionalStack.back().IfLoc,
1345 diag::err_pp_unterminated_conditional);
Chris Lattner30a2fa12006-07-19 06:31:49 +00001346 ConditionalStack.pop_back();
1347 }
Mike Stump11289f42009-09-09 15:08:12 +00001348
Chris Lattner8f96d042008-04-12 05:54:25 +00001349 // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
1350 // a pedwarn.
1351 if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r'))
Mike Stump0be88752009-04-02 02:29:42 +00001352 Diag(BufferEnd, diag::ext_no_newline_eof)
1353 << CodeModificationHint::CreateInsertion(getSourceLocation(BufferEnd),
1354 "\n");
Mike Stump11289f42009-09-09 15:08:12 +00001355
Chris Lattner22eb9722006-06-18 05:43:12 +00001356 BufferPtr = CurPtr;
Chris Lattner30a2fa12006-07-19 06:31:49 +00001357
1358 // Finally, let the preprocessor handle this.
Chris Lattner02b436a2007-10-17 20:41:00 +00001359 return PP->HandleEndOfFile(Result);
Chris Lattner22eb9722006-06-18 05:43:12 +00001360}
1361
Chris Lattner678c8802006-07-11 05:46:12 +00001362/// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
1363/// the specified lexer will return a tok::l_paren token, 0 if it is something
1364/// else and 2 if there are no more tokens in the buffer controlled by the
1365/// lexer.
1366unsigned Lexer::isNextPPTokenLParen() {
1367 assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
Mike Stump11289f42009-09-09 15:08:12 +00001368
Chris Lattner678c8802006-07-11 05:46:12 +00001369 // Switch to 'skipping' mode. This will ensure that we can lex a token
1370 // without emitting diagnostics, disables macro expansion, and will cause EOF
1371 // to return an EOF token instead of popping the include stack.
1372 LexingRawMode = true;
Mike Stump11289f42009-09-09 15:08:12 +00001373
Chris Lattner678c8802006-07-11 05:46:12 +00001374 // Save state that can be changed while lexing so that we can restore it.
1375 const char *TmpBufferPtr = BufferPtr;
Chris Lattner40493eb2009-04-24 07:15:46 +00001376 bool inPPDirectiveMode = ParsingPreprocessorDirective;
Mike Stump11289f42009-09-09 15:08:12 +00001377
Chris Lattner146762e2007-07-20 16:59:19 +00001378 Token Tok;
Chris Lattner8c204872006-10-14 05:19:21 +00001379 Tok.startToken();
Chris Lattner678c8802006-07-11 05:46:12 +00001380 LexTokenInternal(Tok);
Mike Stump11289f42009-09-09 15:08:12 +00001381
Chris Lattner678c8802006-07-11 05:46:12 +00001382 // Restore state that may have changed.
1383 BufferPtr = TmpBufferPtr;
Chris Lattner40493eb2009-04-24 07:15:46 +00001384 ParsingPreprocessorDirective = inPPDirectiveMode;
Mike Stump11289f42009-09-09 15:08:12 +00001385
Chris Lattner678c8802006-07-11 05:46:12 +00001386 // Restore the lexer back to non-skipping mode.
1387 LexingRawMode = false;
Mike Stump11289f42009-09-09 15:08:12 +00001388
Chris Lattner98c1f7c2007-10-09 18:02:16 +00001389 if (Tok.is(tok::eof))
Chris Lattner678c8802006-07-11 05:46:12 +00001390 return 2;
Chris Lattner98c1f7c2007-10-09 18:02:16 +00001391 return Tok.is(tok::l_paren);
Chris Lattner678c8802006-07-11 05:46:12 +00001392}
1393
Chris Lattner7c027ee2009-12-14 06:16:57 +00001394/// FindConflictEnd - Find the end of a version control conflict marker.
1395static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd) {
1396 llvm::StringRef RestOfBuffer(CurPtr+7, BufferEnd-CurPtr-7);
1397 size_t Pos = RestOfBuffer.find(">>>>>>>");
1398 while (Pos != llvm::StringRef::npos) {
1399 // Must occur at start of line.
1400 if (RestOfBuffer[Pos-1] != '\r' &&
1401 RestOfBuffer[Pos-1] != '\n') {
1402 RestOfBuffer = RestOfBuffer.substr(Pos+7);
1403 continue;
1404 }
1405 return RestOfBuffer.data()+Pos;
1406 }
1407 return 0;
1408}
1409
1410/// IsStartOfConflictMarker - If the specified pointer is the start of a version
1411/// control conflict marker like '<<<<<<<', recognize it as such, emit an error
1412/// and recover nicely. This returns true if it is a conflict marker and false
1413/// if not.
1414bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
1415 // Only a conflict marker if it starts at the beginning of a line.
1416 if (CurPtr != BufferStart &&
1417 CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
1418 return false;
1419
1420 // Check to see if we have <<<<<<<.
1421 if (BufferEnd-CurPtr < 8 ||
1422 llvm::StringRef(CurPtr, 7) != "<<<<<<<")
1423 return false;
1424
1425 // If we have a situation where we don't care about conflict markers, ignore
1426 // it.
1427 if (IsInConflictMarker || isLexingRawMode())
1428 return false;
1429
1430 // Check to see if there is a >>>>>>> somewhere in the buffer at the start of
1431 // a line to terminate this conflict marker.
1432 if (FindConflictEnd(CurPtr+7, BufferEnd)) {
1433 // We found a match. We are really in a conflict marker.
1434 // Diagnose this, and ignore to the end of line.
1435 Diag(CurPtr, diag::err_conflict_marker);
1436 IsInConflictMarker = true;
1437
1438 // Skip ahead to the end of line. We know this exists because the
1439 // end-of-conflict marker starts with \r or \n.
1440 while (*CurPtr != '\r' && *CurPtr != '\n') {
1441 assert(CurPtr != BufferEnd && "Didn't find end of line");
1442 ++CurPtr;
1443 }
1444 BufferPtr = CurPtr;
1445 return true;
1446 }
1447
1448 // No end of conflict marker found.
1449 return false;
1450}
1451
1452
1453/// HandleEndOfConflictMarker - If this is a '=======' or '|||||||' or '>>>>>>>'
1454/// marker, then it is the end of a conflict marker. Handle it by ignoring up
1455/// until the end of the line. This returns true if it is a conflict marker and
1456/// false if not.
1457bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
1458 // Only a conflict marker if it starts at the beginning of a line.
1459 if (CurPtr != BufferStart &&
1460 CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
1461 return false;
1462
1463 // If we have a situation where we don't care about conflict markers, ignore
1464 // it.
1465 if (!IsInConflictMarker || isLexingRawMode())
1466 return false;
1467
1468 // Check to see if we have the marker (7 characters in a row).
1469 for (unsigned i = 1; i != 7; ++i)
1470 if (CurPtr[i] != CurPtr[0])
1471 return false;
1472
1473 // If we do have it, search for the end of the conflict marker. This could
1474 // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
1475 // be the end of conflict marker.
1476 if (const char *End = FindConflictEnd(CurPtr, BufferEnd)) {
1477 CurPtr = End;
1478
1479 // Skip ahead to the end of line.
1480 while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
1481 ++CurPtr;
1482
1483 BufferPtr = CurPtr;
1484
1485 // No longer in the conflict marker.
1486 IsInConflictMarker = false;
1487 return true;
1488 }
1489
1490 return false;
1491}
1492
Chris Lattner22eb9722006-06-18 05:43:12 +00001493
1494/// LexTokenInternal - This implements a simple C family lexer. It is an
1495/// extremely performance critical piece of code. This assumes that the buffer
Chris Lattner5c349382009-07-07 05:05:42 +00001496/// has a null character at the end of the file. This returns a preprocessing
1497/// token, not a normal token, as such, it is an internal interface. It assumes
1498/// that the Flags of result have been cleared before calling this.
Chris Lattner146762e2007-07-20 16:59:19 +00001499void Lexer::LexTokenInternal(Token &Result) {
Chris Lattner22eb9722006-06-18 05:43:12 +00001500LexNextToken:
1501 // New token, can't need cleaning yet.
Chris Lattner146762e2007-07-20 16:59:19 +00001502 Result.clearFlag(Token::NeedsCleaning);
Chris Lattner8c204872006-10-14 05:19:21 +00001503 Result.setIdentifierInfo(0);
Mike Stump11289f42009-09-09 15:08:12 +00001504
Chris Lattner22eb9722006-06-18 05:43:12 +00001505 // CurPtr - Cache BufferPtr in an automatic variable.
1506 const char *CurPtr = BufferPtr;
Chris Lattner22eb9722006-06-18 05:43:12 +00001507
Chris Lattnereb54b592006-07-10 06:34:27 +00001508 // Small amounts of horizontal whitespace is very common between tokens.
1509 if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
1510 ++CurPtr;
1511 while ((*CurPtr == ' ') || (*CurPtr == '\t'))
1512 ++CurPtr;
Mike Stump11289f42009-09-09 15:08:12 +00001513
Chris Lattner4d963442008-10-12 04:05:48 +00001514 // If we are keeping whitespace and other tokens, just return what we just
1515 // skipped. The next lexer invocation will return the token after the
1516 // whitespace.
1517 if (isKeepWhitespaceMode()) {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001518 FormTokenWithChars(Result, CurPtr, tok::unknown);
Chris Lattner4d963442008-10-12 04:05:48 +00001519 return;
1520 }
Mike Stump11289f42009-09-09 15:08:12 +00001521
Chris Lattnereb54b592006-07-10 06:34:27 +00001522 BufferPtr = CurPtr;
Chris Lattner146762e2007-07-20 16:59:19 +00001523 Result.setFlag(Token::LeadingSpace);
Chris Lattnereb54b592006-07-10 06:34:27 +00001524 }
Mike Stump11289f42009-09-09 15:08:12 +00001525
Chris Lattner22eb9722006-06-18 05:43:12 +00001526 unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
Mike Stump11289f42009-09-09 15:08:12 +00001527
Chris Lattner22eb9722006-06-18 05:43:12 +00001528 // Read a character, advancing over it.
1529 char Char = getAndAdvanceChar(CurPtr, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001530 tok::TokenKind Kind;
Mike Stump11289f42009-09-09 15:08:12 +00001531
Chris Lattner22eb9722006-06-18 05:43:12 +00001532 switch (Char) {
1533 case 0: // Null.
1534 // Found end of file?
Chris Lattner2183a6e2006-07-18 06:36:12 +00001535 if (CurPtr-1 == BufferEnd) {
1536 // Read the PP instance variable into an automatic variable, because
1537 // LexEndOfFile will often delete 'this'.
Chris Lattner02b436a2007-10-17 20:41:00 +00001538 Preprocessor *PPCache = PP;
Chris Lattner2183a6e2006-07-18 06:36:12 +00001539 if (LexEndOfFile(Result, CurPtr-1)) // Retreat back into the file.
1540 return; // Got a token to return.
Chris Lattner02b436a2007-10-17 20:41:00 +00001541 assert(PPCache && "Raw buffer::LexEndOfFile should return a token");
1542 return PPCache->Lex(Result);
Chris Lattner2183a6e2006-07-18 06:36:12 +00001543 }
Mike Stump11289f42009-09-09 15:08:12 +00001544
Chris Lattner6d27a162008-11-22 02:02:22 +00001545 if (!isLexingRawMode())
1546 Diag(CurPtr-1, diag::null_in_file);
Chris Lattner146762e2007-07-20 16:59:19 +00001547 Result.setFlag(Token::LeadingSpace);
Chris Lattner4d963442008-10-12 04:05:48 +00001548 if (SkipWhitespace(Result, CurPtr))
1549 return; // KeepWhitespaceMode
Mike Stump11289f42009-09-09 15:08:12 +00001550
Chris Lattner22eb9722006-06-18 05:43:12 +00001551 goto LexNextToken; // GCC isn't tail call eliminating.
Chris Lattner3dfff972009-12-17 05:29:40 +00001552
1553 case 26: // DOS & CP/M EOF: "^Z".
1554 // If we're in Microsoft extensions mode, treat this as end of file.
1555 if (Features.Microsoft) {
1556 // Read the PP instance variable into an automatic variable, because
1557 // LexEndOfFile will often delete 'this'.
1558 Preprocessor *PPCache = PP;
1559 if (LexEndOfFile(Result, CurPtr-1)) // Retreat back into the file.
1560 return; // Got a token to return.
1561 assert(PPCache && "Raw buffer::LexEndOfFile should return a token");
1562 return PPCache->Lex(Result);
1563 }
1564 // If Microsoft extensions are disabled, this is just random garbage.
1565 Kind = tok::unknown;
1566 break;
1567
Chris Lattner22eb9722006-06-18 05:43:12 +00001568 case '\n':
1569 case '\r':
1570 // If we are inside a preprocessor directive and we see the end of line,
1571 // we know we are done with the directive, so return an EOM token.
1572 if (ParsingPreprocessorDirective) {
1573 // Done parsing the "line".
1574 ParsingPreprocessorDirective = false;
Mike Stump11289f42009-09-09 15:08:12 +00001575
Chris Lattner457fc152006-07-29 06:30:25 +00001576 // Restore comment saving mode, in case it was disabled for directive.
Chris Lattner097a8b82008-10-12 03:27:19 +00001577 SetCommentRetentionState(PP->getCommentRetentionState());
Mike Stump11289f42009-09-09 15:08:12 +00001578
Chris Lattner22eb9722006-06-18 05:43:12 +00001579 // Since we consumed a newline, we are back at the start of a line.
1580 IsAtStartOfLine = true;
Mike Stump11289f42009-09-09 15:08:12 +00001581
Chris Lattnerb11c3232008-10-12 04:51:35 +00001582 Kind = tok::eom;
Chris Lattner22eb9722006-06-18 05:43:12 +00001583 break;
1584 }
1585 // The returned token is at the start of the line.
Chris Lattner146762e2007-07-20 16:59:19 +00001586 Result.setFlag(Token::StartOfLine);
Chris Lattner22eb9722006-06-18 05:43:12 +00001587 // No leading whitespace seen so far.
Chris Lattner146762e2007-07-20 16:59:19 +00001588 Result.clearFlag(Token::LeadingSpace);
Mike Stump11289f42009-09-09 15:08:12 +00001589
Chris Lattner4d963442008-10-12 04:05:48 +00001590 if (SkipWhitespace(Result, CurPtr))
1591 return; // KeepWhitespaceMode
Chris Lattner22eb9722006-06-18 05:43:12 +00001592 goto LexNextToken; // GCC isn't tail call eliminating.
1593 case ' ':
1594 case '\t':
1595 case '\f':
1596 case '\v':
Chris Lattnerb9b85972007-07-22 06:29:05 +00001597 SkipHorizontalWhitespace:
Chris Lattner146762e2007-07-20 16:59:19 +00001598 Result.setFlag(Token::LeadingSpace);
Chris Lattner4d963442008-10-12 04:05:48 +00001599 if (SkipWhitespace(Result, CurPtr))
1600 return; // KeepWhitespaceMode
Chris Lattnerb9b85972007-07-22 06:29:05 +00001601
1602 SkipIgnoredUnits:
1603 CurPtr = BufferPtr;
Mike Stump11289f42009-09-09 15:08:12 +00001604
Chris Lattnerb9b85972007-07-22 06:29:05 +00001605 // If the next token is obviously a // or /* */ comment, skip it efficiently
1606 // too (without going through the big switch stmt).
Chris Lattner58827712009-01-16 22:39:25 +00001607 if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
1608 Features.BCPLComment) {
Chris Lattnerb9b85972007-07-22 06:29:05 +00001609 SkipBCPLComment(Result, CurPtr+2);
1610 goto SkipIgnoredUnits;
Chris Lattner8637abd2008-10-12 03:22:02 +00001611 } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
Chris Lattnerb9b85972007-07-22 06:29:05 +00001612 SkipBlockComment(Result, CurPtr+2);
1613 goto SkipIgnoredUnits;
1614 } else if (isHorizontalWhitespace(*CurPtr)) {
1615 goto SkipHorizontalWhitespace;
1616 }
Chris Lattner22eb9722006-06-18 05:43:12 +00001617 goto LexNextToken; // GCC isn't tail call eliminating.
Chris Lattner3dfff972009-12-17 05:29:40 +00001618
Chris Lattner2b15cf72008-01-03 17:58:54 +00001619 // C99 6.4.4.1: Integer Constants.
1620 // C99 6.4.4.2: Floating Constants.
1621 case '0': case '1': case '2': case '3': case '4':
1622 case '5': case '6': case '7': case '8': case '9':
1623 // Notify MIOpt that we read a non-whitespace/non-comment token.
1624 MIOpt.ReadToken();
1625 return LexNumericConstant(Result, CurPtr);
Mike Stump11289f42009-09-09 15:08:12 +00001626
Chris Lattner2b15cf72008-01-03 17:58:54 +00001627 case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
Chris Lattner371ac8a2006-07-04 07:11:10 +00001628 // Notify MIOpt that we read a non-whitespace/non-comment token.
1629 MIOpt.ReadToken();
Chris Lattner22eb9722006-06-18 05:43:12 +00001630 Char = getCharAndSize(CurPtr, SizeTmp);
1631
1632 // Wide string literal.
1633 if (Char == '"')
Chris Lattnerd3e98952006-10-06 05:22:26 +00001634 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
1635 true);
Chris Lattner22eb9722006-06-18 05:43:12 +00001636
1637 // Wide character constant.
1638 if (Char == '\'')
1639 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
1640 // FALL THROUGH, treating L like the start of an identifier.
Mike Stump11289f42009-09-09 15:08:12 +00001641
Chris Lattner22eb9722006-06-18 05:43:12 +00001642 // C99 6.4.2: Identifiers.
1643 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
1644 case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
1645 case 'O': case 'P': case 'Q': case 'R': case 'S': case 'T': case 'U':
1646 case 'V': case 'W': case 'X': case 'Y': case 'Z':
1647 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
1648 case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
1649 case 'o': case 'p': case 'q': case 'r': case 's': case 't': case 'u':
1650 case 'v': case 'w': case 'x': case 'y': case 'z':
1651 case '_':
Chris Lattner371ac8a2006-07-04 07:11:10 +00001652 // Notify MIOpt that we read a non-whitespace/non-comment token.
1653 MIOpt.ReadToken();
Chris Lattner22eb9722006-06-18 05:43:12 +00001654 return LexIdentifier(Result, CurPtr);
Chris Lattner2b15cf72008-01-03 17:58:54 +00001655
1656 case '$': // $ in identifiers.
1657 if (Features.DollarIdents) {
Chris Lattner6d27a162008-11-22 02:02:22 +00001658 if (!isLexingRawMode())
1659 Diag(CurPtr-1, diag::ext_dollar_in_identifier);
Chris Lattner2b15cf72008-01-03 17:58:54 +00001660 // Notify MIOpt that we read a non-whitespace/non-comment token.
1661 MIOpt.ReadToken();
1662 return LexIdentifier(Result, CurPtr);
1663 }
Mike Stump11289f42009-09-09 15:08:12 +00001664
Chris Lattnerb11c3232008-10-12 04:51:35 +00001665 Kind = tok::unknown;
Chris Lattner2b15cf72008-01-03 17:58:54 +00001666 break;
Mike Stump11289f42009-09-09 15:08:12 +00001667
Chris Lattner22eb9722006-06-18 05:43:12 +00001668 // C99 6.4.4: Character Constants.
1669 case '\'':
Chris Lattner371ac8a2006-07-04 07:11:10 +00001670 // Notify MIOpt that we read a non-whitespace/non-comment token.
1671 MIOpt.ReadToken();
Chris Lattner22eb9722006-06-18 05:43:12 +00001672 return LexCharConstant(Result, CurPtr);
1673
1674 // C99 6.4.5: String Literals.
1675 case '"':
Chris Lattner371ac8a2006-07-04 07:11:10 +00001676 // Notify MIOpt that we read a non-whitespace/non-comment token.
1677 MIOpt.ReadToken();
Chris Lattnerd3e98952006-10-06 05:22:26 +00001678 return LexStringLiteral(Result, CurPtr, false);
Chris Lattner22eb9722006-06-18 05:43:12 +00001679
1680 // C99 6.4.6: Punctuators.
1681 case '?':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001682 Kind = tok::question;
Chris Lattner22eb9722006-06-18 05:43:12 +00001683 break;
1684 case '[':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001685 Kind = tok::l_square;
Chris Lattner22eb9722006-06-18 05:43:12 +00001686 break;
1687 case ']':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001688 Kind = tok::r_square;
Chris Lattner22eb9722006-06-18 05:43:12 +00001689 break;
1690 case '(':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001691 Kind = tok::l_paren;
Chris Lattner22eb9722006-06-18 05:43:12 +00001692 break;
1693 case ')':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001694 Kind = tok::r_paren;
Chris Lattner22eb9722006-06-18 05:43:12 +00001695 break;
1696 case '{':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001697 Kind = tok::l_brace;
Chris Lattner22eb9722006-06-18 05:43:12 +00001698 break;
1699 case '}':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001700 Kind = tok::r_brace;
Chris Lattner22eb9722006-06-18 05:43:12 +00001701 break;
1702 case '.':
1703 Char = getCharAndSize(CurPtr, SizeTmp);
1704 if (Char >= '0' && Char <= '9') {
Chris Lattner371ac8a2006-07-04 07:11:10 +00001705 // Notify MIOpt that we read a non-whitespace/non-comment token.
1706 MIOpt.ReadToken();
1707
Chris Lattner22eb9722006-06-18 05:43:12 +00001708 return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
1709 } else if (Features.CPlusPlus && Char == '*') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001710 Kind = tok::periodstar;
Chris Lattner22eb9722006-06-18 05:43:12 +00001711 CurPtr += SizeTmp;
1712 } else if (Char == '.' &&
1713 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001714 Kind = tok::ellipsis;
Chris Lattner22eb9722006-06-18 05:43:12 +00001715 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
1716 SizeTmp2, Result);
1717 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001718 Kind = tok::period;
Chris Lattner22eb9722006-06-18 05:43:12 +00001719 }
1720 break;
1721 case '&':
1722 Char = getCharAndSize(CurPtr, SizeTmp);
1723 if (Char == '&') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001724 Kind = tok::ampamp;
Chris Lattner22eb9722006-06-18 05:43:12 +00001725 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1726 } else if (Char == '=') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001727 Kind = tok::ampequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001728 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1729 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001730 Kind = tok::amp;
Chris Lattner22eb9722006-06-18 05:43:12 +00001731 }
1732 break;
Mike Stump11289f42009-09-09 15:08:12 +00001733 case '*':
Chris Lattner22eb9722006-06-18 05:43:12 +00001734 if (getCharAndSize(CurPtr, SizeTmp) == '=') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001735 Kind = tok::starequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001736 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1737 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001738 Kind = tok::star;
Chris Lattner22eb9722006-06-18 05:43:12 +00001739 }
1740 break;
1741 case '+':
1742 Char = getCharAndSize(CurPtr, SizeTmp);
1743 if (Char == '+') {
Chris Lattner22eb9722006-06-18 05:43:12 +00001744 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001745 Kind = tok::plusplus;
Chris Lattner22eb9722006-06-18 05:43:12 +00001746 } else if (Char == '=') {
Chris Lattner22eb9722006-06-18 05:43:12 +00001747 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001748 Kind = tok::plusequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001749 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001750 Kind = tok::plus;
Chris Lattner22eb9722006-06-18 05:43:12 +00001751 }
1752 break;
1753 case '-':
1754 Char = getCharAndSize(CurPtr, SizeTmp);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001755 if (Char == '-') { // --
Chris Lattner22eb9722006-06-18 05:43:12 +00001756 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001757 Kind = tok::minusminus;
Mike Stump11289f42009-09-09 15:08:12 +00001758 } else if (Char == '>' && Features.CPlusPlus &&
Chris Lattnerb11c3232008-10-12 04:51:35 +00001759 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
Chris Lattner22eb9722006-06-18 05:43:12 +00001760 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
1761 SizeTmp2, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001762 Kind = tok::arrowstar;
1763 } else if (Char == '>') { // ->
Chris Lattner22eb9722006-06-18 05:43:12 +00001764 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001765 Kind = tok::arrow;
1766 } else if (Char == '=') { // -=
Chris Lattner22eb9722006-06-18 05:43:12 +00001767 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001768 Kind = tok::minusequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001769 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001770 Kind = tok::minus;
Chris Lattner22eb9722006-06-18 05:43:12 +00001771 }
1772 break;
1773 case '~':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001774 Kind = tok::tilde;
Chris Lattner22eb9722006-06-18 05:43:12 +00001775 break;
1776 case '!':
1777 if (getCharAndSize(CurPtr, SizeTmp) == '=') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001778 Kind = tok::exclaimequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001779 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1780 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001781 Kind = tok::exclaim;
Chris Lattner22eb9722006-06-18 05:43:12 +00001782 }
1783 break;
1784 case '/':
1785 // 6.4.9: Comments
1786 Char = getCharAndSize(CurPtr, SizeTmp);
1787 if (Char == '/') { // BCPL comment.
Chris Lattner58827712009-01-16 22:39:25 +00001788 // Even if BCPL comments are disabled (e.g. in C89 mode), we generally
1789 // want to lex this as a comment. There is one problem with this though,
1790 // that in one particular corner case, this can change the behavior of the
1791 // resultant program. For example, In "foo //**/ bar", C89 would lex
1792 // this as "foo / bar" and langauges with BCPL comments would lex it as
1793 // "foo". Check to see if the character after the second slash is a '*'.
1794 // If so, we will lex that as a "/" instead of the start of a comment.
1795 if (Features.BCPLComment ||
1796 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*') {
1797 if (SkipBCPLComment(Result, ConsumeChar(CurPtr, SizeTmp, Result)))
1798 return; // KeepCommentMode
Mike Stump11289f42009-09-09 15:08:12 +00001799
Chris Lattner58827712009-01-16 22:39:25 +00001800 // It is common for the tokens immediately after a // comment to be
1801 // whitespace (indentation for the next line). Instead of going through
1802 // the big switch, handle it efficiently now.
1803 goto SkipIgnoredUnits;
1804 }
1805 }
Mike Stump11289f42009-09-09 15:08:12 +00001806
Chris Lattner58827712009-01-16 22:39:25 +00001807 if (Char == '*') { // /**/ comment.
Chris Lattner457fc152006-07-29 06:30:25 +00001808 if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result)))
Chris Lattnere01e7582008-10-12 04:15:42 +00001809 return; // KeepCommentMode
1810 goto LexNextToken; // GCC isn't tail call eliminating.
Chris Lattner58827712009-01-16 22:39:25 +00001811 }
Mike Stump11289f42009-09-09 15:08:12 +00001812
Chris Lattner58827712009-01-16 22:39:25 +00001813 if (Char == '=') {
Chris Lattner22eb9722006-06-18 05:43:12 +00001814 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001815 Kind = tok::slashequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001816 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001817 Kind = tok::slash;
Chris Lattner22eb9722006-06-18 05:43:12 +00001818 }
1819 break;
1820 case '%':
1821 Char = getCharAndSize(CurPtr, SizeTmp);
1822 if (Char == '=') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001823 Kind = tok::percentequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001824 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1825 } else if (Features.Digraphs && Char == '>') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001826 Kind = tok::r_brace; // '%>' -> '}'
Chris Lattner22eb9722006-06-18 05:43:12 +00001827 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1828 } else if (Features.Digraphs && Char == ':') {
1829 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattner2b271db2006-07-15 05:41:09 +00001830 Char = getCharAndSize(CurPtr, SizeTmp);
1831 if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001832 Kind = tok::hashhash; // '%:%:' -> '##'
Chris Lattner22eb9722006-06-18 05:43:12 +00001833 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
1834 SizeTmp2, Result);
Chris Lattner2b271db2006-07-15 05:41:09 +00001835 } else if (Char == '@' && Features.Microsoft) { // %:@ -> #@ -> Charize
Chris Lattner2b271db2006-07-15 05:41:09 +00001836 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattner6d27a162008-11-22 02:02:22 +00001837 if (!isLexingRawMode())
1838 Diag(BufferPtr, diag::charize_microsoft_ext);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001839 Kind = tok::hashat;
Chris Lattner2534324a2009-03-18 20:58:27 +00001840 } else { // '%:' -> '#'
Chris Lattner22eb9722006-06-18 05:43:12 +00001841 // We parsed a # character. If this occurs at the start of the line,
1842 // it's actually the start of a preprocessing directive. Callback to
1843 // the preprocessor to handle it.
1844 // FIXME: -fpreprocessed mode??
Chris Lattnerff96dd02009-05-13 06:10:29 +00001845 if (Result.isAtStartOfLine() && !LexingRawMode && !Is_PragmaLexer) {
Chris Lattner2534324a2009-03-18 20:58:27 +00001846 FormTokenWithChars(Result, CurPtr, tok::hash);
Chris Lattner02b436a2007-10-17 20:41:00 +00001847 PP->HandleDirective(Result);
Mike Stump11289f42009-09-09 15:08:12 +00001848
Chris Lattner22eb9722006-06-18 05:43:12 +00001849 // As an optimization, if the preprocessor didn't switch lexers, tail
1850 // recurse.
Chris Lattner02b436a2007-10-17 20:41:00 +00001851 if (PP->isCurrentLexer(this)) {
Chris Lattner22eb9722006-06-18 05:43:12 +00001852 // Start a new token. If this is a #include or something, the PP may
1853 // want us starting at the beginning of the line again. If so, set
1854 // the StartOfLine flag.
1855 if (IsAtStartOfLine) {
Chris Lattner146762e2007-07-20 16:59:19 +00001856 Result.setFlag(Token::StartOfLine);
Chris Lattner22eb9722006-06-18 05:43:12 +00001857 IsAtStartOfLine = false;
1858 }
1859 goto LexNextToken; // GCC isn't tail call eliminating.
1860 }
Mike Stump11289f42009-09-09 15:08:12 +00001861
Chris Lattner02b436a2007-10-17 20:41:00 +00001862 return PP->Lex(Result);
Chris Lattner22eb9722006-06-18 05:43:12 +00001863 }
Mike Stump11289f42009-09-09 15:08:12 +00001864
Chris Lattner2534324a2009-03-18 20:58:27 +00001865 Kind = tok::hash;
Chris Lattner22eb9722006-06-18 05:43:12 +00001866 }
1867 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001868 Kind = tok::percent;
Chris Lattner22eb9722006-06-18 05:43:12 +00001869 }
1870 break;
1871 case '<':
1872 Char = getCharAndSize(CurPtr, SizeTmp);
1873 if (ParsingFilename) {
Chris Lattnerb40289b2009-04-17 23:56:52 +00001874 return LexAngledStringLiteral(Result, CurPtr);
Chris Lattner22eb9722006-06-18 05:43:12 +00001875 } else if (Char == '<') {
Chris Lattner7c027ee2009-12-14 06:16:57 +00001876 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
1877 if (After == '=') {
1878 Kind = tok::lesslessequal;
1879 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
1880 SizeTmp2, Result);
1881 } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
1882 // If this is actually a '<<<<<<<' version control conflict marker,
1883 // recognize it as such and recover nicely.
1884 goto LexNextToken;
1885 } else {
1886 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1887 Kind = tok::lessless;
1888 }
Chris Lattner22eb9722006-06-18 05:43:12 +00001889 } else if (Char == '=') {
Chris Lattner22eb9722006-06-18 05:43:12 +00001890 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001891 Kind = tok::lessequal;
1892 } else if (Features.Digraphs && Char == ':') { // '<:' -> '['
Chris Lattner22eb9722006-06-18 05:43:12 +00001893 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001894 Kind = tok::l_square;
1895 } else if (Features.Digraphs && Char == '%') { // '<%' -> '{'
Chris Lattner22eb9722006-06-18 05:43:12 +00001896 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001897 Kind = tok::l_brace;
Chris Lattner22eb9722006-06-18 05:43:12 +00001898 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001899 Kind = tok::less;
Chris Lattner22eb9722006-06-18 05:43:12 +00001900 }
1901 break;
1902 case '>':
1903 Char = getCharAndSize(CurPtr, SizeTmp);
1904 if (Char == '=') {
Chris Lattner22eb9722006-06-18 05:43:12 +00001905 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001906 Kind = tok::greaterequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001907 } else if (Char == '>') {
Chris Lattner7c027ee2009-12-14 06:16:57 +00001908 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
1909 if (After == '=') {
1910 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
1911 SizeTmp2, Result);
1912 Kind = tok::greatergreaterequal;
1913 } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
1914 // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
1915 goto LexNextToken;
1916 } else {
1917 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1918 Kind = tok::greatergreater;
1919 }
1920
Chris Lattner22eb9722006-06-18 05:43:12 +00001921 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001922 Kind = tok::greater;
Chris Lattner22eb9722006-06-18 05:43:12 +00001923 }
1924 break;
1925 case '^':
1926 Char = getCharAndSize(CurPtr, SizeTmp);
1927 if (Char == '=') {
Chris Lattner22eb9722006-06-18 05:43:12 +00001928 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattnerb11c3232008-10-12 04:51:35 +00001929 Kind = tok::caretequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001930 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001931 Kind = tok::caret;
Chris Lattner22eb9722006-06-18 05:43:12 +00001932 }
1933 break;
1934 case '|':
1935 Char = getCharAndSize(CurPtr, SizeTmp);
1936 if (Char == '=') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001937 Kind = tok::pipeequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001938 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1939 } else if (Char == '|') {
Chris Lattner7c027ee2009-12-14 06:16:57 +00001940 // If this is '|||||||' and we're in a conflict marker, ignore it.
1941 if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
1942 goto LexNextToken;
Chris Lattnerb11c3232008-10-12 04:51:35 +00001943 Kind = tok::pipepipe;
Chris Lattner22eb9722006-06-18 05:43:12 +00001944 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1945 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001946 Kind = tok::pipe;
Chris Lattner22eb9722006-06-18 05:43:12 +00001947 }
1948 break;
1949 case ':':
1950 Char = getCharAndSize(CurPtr, SizeTmp);
1951 if (Features.Digraphs && Char == '>') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001952 Kind = tok::r_square; // ':>' -> ']'
Chris Lattner22eb9722006-06-18 05:43:12 +00001953 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
1954 } else if (Features.CPlusPlus && Char == ':') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001955 Kind = tok::coloncolon;
Chris Lattner22eb9722006-06-18 05:43:12 +00001956 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Mike Stump11289f42009-09-09 15:08:12 +00001957 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001958 Kind = tok::colon;
Chris Lattner22eb9722006-06-18 05:43:12 +00001959 }
1960 break;
1961 case ';':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001962 Kind = tok::semi;
Chris Lattner22eb9722006-06-18 05:43:12 +00001963 break;
1964 case '=':
1965 Char = getCharAndSize(CurPtr, SizeTmp);
1966 if (Char == '=') {
Chris Lattner7c027ee2009-12-14 06:16:57 +00001967 // If this is '=======' and we're in a conflict marker, ignore it.
1968 if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
1969 goto LexNextToken;
1970
Chris Lattnerb11c3232008-10-12 04:51:35 +00001971 Kind = tok::equalequal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001972 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Mike Stump11289f42009-09-09 15:08:12 +00001973 } else {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001974 Kind = tok::equal;
Chris Lattner22eb9722006-06-18 05:43:12 +00001975 }
1976 break;
1977 case ',':
Chris Lattnerb11c3232008-10-12 04:51:35 +00001978 Kind = tok::comma;
Chris Lattner22eb9722006-06-18 05:43:12 +00001979 break;
1980 case '#':
1981 Char = getCharAndSize(CurPtr, SizeTmp);
1982 if (Char == '#') {
Chris Lattnerb11c3232008-10-12 04:51:35 +00001983 Kind = tok::hashhash;
Chris Lattner22eb9722006-06-18 05:43:12 +00001984 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattner2b271db2006-07-15 05:41:09 +00001985 } else if (Char == '@' && Features.Microsoft) { // #@ -> Charize
Chris Lattnerb11c3232008-10-12 04:51:35 +00001986 Kind = tok::hashat;
Chris Lattner6d27a162008-11-22 02:02:22 +00001987 if (!isLexingRawMode())
1988 Diag(BufferPtr, diag::charize_microsoft_ext);
Chris Lattner2b271db2006-07-15 05:41:09 +00001989 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
Chris Lattner22eb9722006-06-18 05:43:12 +00001990 } else {
Chris Lattner22eb9722006-06-18 05:43:12 +00001991 // We parsed a # character. If this occurs at the start of the line,
1992 // it's actually the start of a preprocessing directive. Callback to
1993 // the preprocessor to handle it.
Chris Lattner505c5472006-07-03 00:55:48 +00001994 // FIXME: -fpreprocessed mode??
Chris Lattnerff96dd02009-05-13 06:10:29 +00001995 if (Result.isAtStartOfLine() && !LexingRawMode && !Is_PragmaLexer) {
Chris Lattner2534324a2009-03-18 20:58:27 +00001996 FormTokenWithChars(Result, CurPtr, tok::hash);
Chris Lattner02b436a2007-10-17 20:41:00 +00001997 PP->HandleDirective(Result);
Mike Stump11289f42009-09-09 15:08:12 +00001998
Chris Lattner22eb9722006-06-18 05:43:12 +00001999 // As an optimization, if the preprocessor didn't switch lexers, tail
2000 // recurse.
Chris Lattner02b436a2007-10-17 20:41:00 +00002001 if (PP->isCurrentLexer(this)) {
Chris Lattner22eb9722006-06-18 05:43:12 +00002002 // Start a new token. If this is a #include or something, the PP may
2003 // want us starting at the beginning of the line again. If so, set
2004 // the StartOfLine flag.
2005 if (IsAtStartOfLine) {
Chris Lattner146762e2007-07-20 16:59:19 +00002006 Result.setFlag(Token::StartOfLine);
Chris Lattner22eb9722006-06-18 05:43:12 +00002007 IsAtStartOfLine = false;
2008 }
2009 goto LexNextToken; // GCC isn't tail call eliminating.
2010 }
Chris Lattner02b436a2007-10-17 20:41:00 +00002011 return PP->Lex(Result);
Chris Lattner22eb9722006-06-18 05:43:12 +00002012 }
Mike Stump11289f42009-09-09 15:08:12 +00002013
Chris Lattner2534324a2009-03-18 20:58:27 +00002014 Kind = tok::hash;
Chris Lattner22eb9722006-06-18 05:43:12 +00002015 }
2016 break;
2017
Chris Lattner2b15cf72008-01-03 17:58:54 +00002018 case '@':
2019 // Objective C support.
2020 if (CurPtr[-1] == '@' && Features.ObjC1)
Chris Lattnerb11c3232008-10-12 04:51:35 +00002021 Kind = tok::at;
Chris Lattner2b15cf72008-01-03 17:58:54 +00002022 else
Chris Lattnerb11c3232008-10-12 04:51:35 +00002023 Kind = tok::unknown;
Chris Lattner2b15cf72008-01-03 17:58:54 +00002024 break;
Mike Stump11289f42009-09-09 15:08:12 +00002025
Chris Lattner22eb9722006-06-18 05:43:12 +00002026 case '\\':
Chris Lattner505c5472006-07-03 00:55:48 +00002027 // FIXME: UCN's.
Chris Lattner22eb9722006-06-18 05:43:12 +00002028 // FALL THROUGH.
2029 default:
Chris Lattnerb11c3232008-10-12 04:51:35 +00002030 Kind = tok::unknown;
Chris Lattner041bef82006-07-11 05:52:53 +00002031 break;
Chris Lattner22eb9722006-06-18 05:43:12 +00002032 }
Mike Stump11289f42009-09-09 15:08:12 +00002033
Chris Lattner371ac8a2006-07-04 07:11:10 +00002034 // Notify MIOpt that we read a non-whitespace/non-comment token.
2035 MIOpt.ReadToken();
2036
Chris Lattnerd01e2912006-06-18 16:22:51 +00002037 // Update the location of token as well as BufferPtr.
Chris Lattnerb11c3232008-10-12 04:51:35 +00002038 FormTokenWithChars(Result, CurPtr, Kind);
Chris Lattner22eb9722006-06-18 05:43:12 +00002039}