Bug Summary

File:src/gnu/usr.bin/clang/libclangLex/../../../llvm/clang/lib/Lex/Lexer.cpp
Warning:line 1164, column 10
Called C++ object pointer is null

Annotated Source Code

Press '?' to see keyboard shortcuts

clang -cc1 -cc1 -triple amd64-unknown-openbsd7.0 -analyze -disable-free -disable-llvm-verifier -discard-value-names -main-file-name Lexer.cpp -analyzer-store=region -analyzer-opt-analyze-nested-blocks -analyzer-checker=core -analyzer-checker=apiModeling -analyzer-checker=unix -analyzer-checker=deadcode -analyzer-checker=cplusplus -analyzer-checker=security.insecureAPI.UncheckedReturn -analyzer-checker=security.insecureAPI.getpw -analyzer-checker=security.insecureAPI.gets -analyzer-checker=security.insecureAPI.mktemp -analyzer-checker=security.insecureAPI.mkstemp -analyzer-checker=security.insecureAPI.vfork -analyzer-checker=nullability.NullPassedToNonnull -analyzer-checker=nullability.NullReturnedFromNonnull -analyzer-output plist -w -setup-static-analyzer -mrelocation-model static -mframe-pointer=all -relaxed-aliasing -fno-rounding-math -mconstructor-aliases -munwind-tables -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fcoverage-compilation-dir=/usr/src/gnu/usr.bin/clang/libclangLex/obj -resource-dir /usr/local/lib/clang/13.0.0 -I /usr/src/gnu/usr.bin/clang/libclangLex/../../../llvm/clang/include -I /usr/src/gnu/usr.bin/clang/libclangLex/../../../llvm/llvm/include -I /usr/src/gnu/usr.bin/clang/libclangLex/../include -I /usr/src/gnu/usr.bin/clang/libclangLex/obj -I /usr/src/gnu/usr.bin/clang/libclangLex/obj/../include -D NDEBUG -D __STDC_LIMIT_MACROS -D __STDC_CONSTANT_MACROS -D __STDC_FORMAT_MACROS -D LLVM_PREFIX="/usr" -internal-isystem /usr/include/c++/v1 -internal-isystem /usr/local/lib/clang/13.0.0/include -internal-externc-isystem /usr/include -O2 -Wno-unused-parameter -Wwrite-strings -Wno-missing-field-initializers -Wno-long-long -Wno-comment -std=c++14 -fdeprecated-macro -fdebug-compilation-dir=/usr/src/gnu/usr.bin/clang/libclangLex/obj -ferror-limit 19 -fvisibility-inlines-hidden -fwrapv -stack-protector 2 -fno-rtti -fgnuc-version=4.2.1 -vectorize-loops -vectorize-slp -fno-builtin-malloc -fno-builtin-calloc -fno-builtin-realloc -fno-builtin-valloc -fno-builtin-free -fno-builtin-strdup -fno-builtin-strndup -analyzer-output=html -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /home/ben/Projects/vmm/scan-build/2022-01-12-194120-40624-1 -x c++ /usr/src/gnu/usr.bin/clang/libclangLex/../../../llvm/clang/lib/Lex/Lexer.cpp

/usr/src/gnu/usr.bin/clang/libclangLex/../../../llvm/clang/lib/Lex/Lexer.cpp

1//===- Lexer.cpp - C Language Family Lexer --------------------------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This file implements the Lexer and Token interfaces.
10//
11//===----------------------------------------------------------------------===//
12
13#include "clang/Lex/Lexer.h"
14#include "UnicodeCharSets.h"
15#include "clang/Basic/CharInfo.h"
16#include "clang/Basic/Diagnostic.h"
17#include "clang/Basic/IdentifierTable.h"
18#include "clang/Basic/LLVM.h"
19#include "clang/Basic/LangOptions.h"
20#include "clang/Basic/SourceLocation.h"
21#include "clang/Basic/SourceManager.h"
22#include "clang/Basic/TokenKinds.h"
23#include "clang/Lex/LexDiagnostic.h"
24#include "clang/Lex/LiteralSupport.h"
25#include "clang/Lex/MultipleIncludeOpt.h"
26#include "clang/Lex/Preprocessor.h"
27#include "clang/Lex/PreprocessorOptions.h"
28#include "clang/Lex/Token.h"
29#include "llvm/ADT/None.h"
30#include "llvm/ADT/Optional.h"
31#include "llvm/ADT/STLExtras.h"
32#include "llvm/ADT/StringExtras.h"
33#include "llvm/ADT/StringRef.h"
34#include "llvm/ADT/StringSwitch.h"
35#include "llvm/Support/Compiler.h"
36#include "llvm/Support/ConvertUTF.h"
37#include "llvm/Support/MathExtras.h"
38#include "llvm/Support/MemoryBufferRef.h"
39#include "llvm/Support/NativeFormatting.h"
40#include "llvm/Support/UnicodeCharRanges.h"
41#include <algorithm>
42#include <cassert>
43#include <cstddef>
44#include <cstdint>
45#include <cstring>
46#include <string>
47#include <tuple>
48#include <utility>
49
50using namespace clang;
51
52//===----------------------------------------------------------------------===//
53// Token Class Implementation
54//===----------------------------------------------------------------------===//
55
56/// isObjCAtKeyword - Return true if we have an ObjC keyword identifier.
57bool Token::isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const {
58 if (isAnnotation())
59 return false;
60 if (IdentifierInfo *II = getIdentifierInfo())
61 return II->getObjCKeywordID() == objcKey;
62 return false;
63}
64
65/// getObjCKeywordID - Return the ObjC keyword kind.
66tok::ObjCKeywordKind Token::getObjCKeywordID() const {
67 if (isAnnotation())
68 return tok::objc_not_keyword;
69 IdentifierInfo *specId = getIdentifierInfo();
70 return specId ? specId->getObjCKeywordID() : tok::objc_not_keyword;
71}
72
73//===----------------------------------------------------------------------===//
74// Lexer Class Implementation
75//===----------------------------------------------------------------------===//
76
77void Lexer::anchor() {}
78
79void Lexer::InitLexer(const char *BufStart, const char *BufPtr,
80 const char *BufEnd) {
81 BufferStart = BufStart;
82 BufferPtr = BufPtr;
83 BufferEnd = BufEnd;
84
85 assert(BufEnd[0] == 0 &&((void)0)
86 "We assume that the input buffer has a null character at the end"((void)0)
87 " to simplify lexing!")((void)0);
88
89 // Check whether we have a BOM in the beginning of the buffer. If yes - act
90 // accordingly. Right now we support only UTF-8 with and without BOM, so, just
91 // skip the UTF-8 BOM if it's present.
92 if (BufferStart == BufferPtr) {
93 // Determine the size of the BOM.
94 StringRef Buf(BufferStart, BufferEnd - BufferStart);
95 size_t BOMLength = llvm::StringSwitch<size_t>(Buf)
96 .StartsWith("\xEF\xBB\xBF", 3) // UTF-8 BOM
97 .Default(0);
98
99 // Skip the BOM.
100 BufferPtr += BOMLength;
101 }
102
103 Is_PragmaLexer = false;
104 CurrentConflictMarkerState = CMK_None;
105
106 // Start of the file is a start of line.
107 IsAtStartOfLine = true;
108 IsAtPhysicalStartOfLine = true;
109
110 HasLeadingSpace = false;
111 HasLeadingEmptyMacro = false;
112
113 // We are not after parsing a #.
114 ParsingPreprocessorDirective = false;
115
116 // We are not after parsing #include.
117 ParsingFilename = false;
118
119 // We are not in raw mode. Raw mode disables diagnostics and interpretation
120 // of tokens (e.g. identifiers, thus disabling macro expansion). It is used
121 // to quickly lex the tokens of the buffer, e.g. when handling a "#if 0" block
122 // or otherwise skipping over tokens.
123 LexingRawMode = false;
124
125 // Default to not keeping comments.
126 ExtendedTokenMode = 0;
127
128 NewLinePtr = nullptr;
129}
130
131/// Lexer constructor - Create a new lexer object for the specified buffer
132/// with the specified preprocessor managing the lexing process. This lexer
133/// assumes that the associated file buffer and Preprocessor objects will
134/// outlive it, so it doesn't take ownership of either of them.
135Lexer::Lexer(FileID FID, const llvm::MemoryBufferRef &InputFile,
136 Preprocessor &PP)
137 : PreprocessorLexer(&PP, FID),
138 FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
139 LangOpts(PP.getLangOpts()) {
140 InitLexer(InputFile.getBufferStart(), InputFile.getBufferStart(),
141 InputFile.getBufferEnd());
142
143 resetExtendedTokenMode();
144}
145
146/// Lexer constructor - Create a new raw lexer object. This object is only
147/// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
148/// range will outlive it, so it doesn't take ownership of it.
149Lexer::Lexer(SourceLocation fileloc, const LangOptions &langOpts,
150 const char *BufStart, const char *BufPtr, const char *BufEnd)
151 : FileLoc(fileloc), LangOpts(langOpts) {
152 InitLexer(BufStart, BufPtr, BufEnd);
153
154 // We *are* in raw mode.
155 LexingRawMode = true;
156}
157
158/// Lexer constructor - Create a new raw lexer object. This object is only
159/// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
160/// range will outlive it, so it doesn't take ownership of it.
161Lexer::Lexer(FileID FID, const llvm::MemoryBufferRef &FromFile,
162 const SourceManager &SM, const LangOptions &langOpts)
163 : Lexer(SM.getLocForStartOfFile(FID), langOpts, FromFile.getBufferStart(),
164 FromFile.getBufferStart(), FromFile.getBufferEnd()) {}
165
166void Lexer::resetExtendedTokenMode() {
167 assert(PP && "Cannot reset token mode without a preprocessor")((void)0);
168 if (LangOpts.TraditionalCPP)
169 SetKeepWhitespaceMode(true);
170 else
171 SetCommentRetentionState(PP->getCommentRetentionState());
172}
173
174/// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
175/// _Pragma expansion. This has a variety of magic semantics that this method
176/// sets up. It returns a new'd Lexer that must be delete'd when done.
177///
178/// On entrance to this routine, TokStartLoc is a macro location which has a
179/// spelling loc that indicates the bytes to be lexed for the token and an
180/// expansion location that indicates where all lexed tokens should be
181/// "expanded from".
182///
183/// TODO: It would really be nice to make _Pragma just be a wrapper around a
184/// normal lexer that remaps tokens as they fly by. This would require making
185/// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer
186/// interface that could handle this stuff. This would pull GetMappedTokenLoc
187/// out of the critical path of the lexer!
188///
189Lexer *Lexer::Create_PragmaLexer(SourceLocation SpellingLoc,
190 SourceLocation ExpansionLocStart,
191 SourceLocation ExpansionLocEnd,
192 unsigned TokLen, Preprocessor &PP) {
193 SourceManager &SM = PP.getSourceManager();
194
195 // Create the lexer as if we were going to lex the file normally.
196 FileID SpellingFID = SM.getFileID(SpellingLoc);
197 llvm::MemoryBufferRef InputFile = SM.getBufferOrFake(SpellingFID);
198 Lexer *L = new Lexer(SpellingFID, InputFile, PP);
199
200 // Now that the lexer is created, change the start/end locations so that we
201 // just lex the subsection of the file that we want. This is lexing from a
202 // scratch buffer.
203 const char *StrData = SM.getCharacterData(SpellingLoc);
204
205 L->BufferPtr = StrData;
206 L->BufferEnd = StrData+TokLen;
207 assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!")((void)0);
208
209 // Set the SourceLocation with the remapping information. This ensures that
210 // GetMappedTokenLoc will remap the tokens as they are lexed.
211 L->FileLoc = SM.createExpansionLoc(SM.getLocForStartOfFile(SpellingFID),
212 ExpansionLocStart,
213 ExpansionLocEnd, TokLen);
214
215 // Ensure that the lexer thinks it is inside a directive, so that end \n will
216 // return an EOD token.
217 L->ParsingPreprocessorDirective = true;
218
219 // This lexer really is for _Pragma.
220 L->Is_PragmaLexer = true;
221 return L;
222}
223
224bool Lexer::skipOver(unsigned NumBytes) {
225 IsAtPhysicalStartOfLine = true;
226 IsAtStartOfLine = true;
227 if ((BufferPtr + NumBytes) > BufferEnd)
228 return true;
229 BufferPtr += NumBytes;
230 return false;
231}
232
233template <typename T> static void StringifyImpl(T &Str, char Quote) {
234 typename T::size_type i = 0, e = Str.size();
235 while (i < e) {
236 if (Str[i] == '\\' || Str[i] == Quote) {
237 Str.insert(Str.begin() + i, '\\');
238 i += 2;
239 ++e;
240 } else if (Str[i] == '\n' || Str[i] == '\r') {
241 // Replace '\r\n' and '\n\r' to '\\' followed by 'n'.
242 if ((i < e - 1) && (Str[i + 1] == '\n' || Str[i + 1] == '\r') &&
243 Str[i] != Str[i + 1]) {
244 Str[i] = '\\';
245 Str[i + 1] = 'n';
246 } else {
247 // Replace '\n' and '\r' to '\\' followed by 'n'.
248 Str[i] = '\\';
249 Str.insert(Str.begin() + i + 1, 'n');
250 ++e;
251 }
252 i += 2;
253 } else
254 ++i;
255 }
256}
257
258std::string Lexer::Stringify(StringRef Str, bool Charify) {
259 std::string Result = std::string(Str);
260 char Quote = Charify ? '\'' : '"';
261 StringifyImpl(Result, Quote);
262 return Result;
263}
264
265void Lexer::Stringify(SmallVectorImpl<char> &Str) { StringifyImpl(Str, '"'); }
266
267//===----------------------------------------------------------------------===//
268// Token Spelling
269//===----------------------------------------------------------------------===//
270
271/// Slow case of getSpelling. Extract the characters comprising the
272/// spelling of this token from the provided input buffer.
273static size_t getSpellingSlow(const Token &Tok, const char *BufPtr,
274 const LangOptions &LangOpts, char *Spelling) {
275 assert(Tok.needsCleaning() && "getSpellingSlow called on simple token")((void)0);
276
277 size_t Length = 0;
278 const char *BufEnd = BufPtr + Tok.getLength();
279
280 if (tok::isStringLiteral(Tok.getKind())) {
281 // Munch the encoding-prefix and opening double-quote.
282 while (BufPtr < BufEnd) {
283 unsigned Size;
284 Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
285 BufPtr += Size;
286
287 if (Spelling[Length - 1] == '"')
288 break;
289 }
290
291 // Raw string literals need special handling; trigraph expansion and line
292 // splicing do not occur within their d-char-sequence nor within their
293 // r-char-sequence.
294 if (Length >= 2 &&
295 Spelling[Length - 2] == 'R' && Spelling[Length - 1] == '"') {
296 // Search backwards from the end of the token to find the matching closing
297 // quote.
298 const char *RawEnd = BufEnd;
299 do --RawEnd; while (*RawEnd != '"');
300 size_t RawLength = RawEnd - BufPtr + 1;
301
302 // Everything between the quotes is included verbatim in the spelling.
303 memcpy(Spelling + Length, BufPtr, RawLength);
304 Length += RawLength;
305 BufPtr += RawLength;
306
307 // The rest of the token is lexed normally.
308 }
309 }
310
311 while (BufPtr < BufEnd) {
312 unsigned Size;
313 Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
314 BufPtr += Size;
315 }
316
317 assert(Length < Tok.getLength() &&((void)0)
318 "NeedsCleaning flag set on token that didn't need cleaning!")((void)0);
319 return Length;
320}
321
322/// getSpelling() - Return the 'spelling' of this token. The spelling of a
323/// token are the characters used to represent the token in the source file
324/// after trigraph expansion and escaped-newline folding. In particular, this
325/// wants to get the true, uncanonicalized, spelling of things like digraphs
326/// UCNs, etc.
327StringRef Lexer::getSpelling(SourceLocation loc,
328 SmallVectorImpl<char> &buffer,
329 const SourceManager &SM,
330 const LangOptions &options,
331 bool *invalid) {
332 // Break down the source location.
333 std::pair<FileID, unsigned> locInfo = SM.getDecomposedLoc(loc);
334
335 // Try to the load the file buffer.
336 bool invalidTemp = false;
337 StringRef file = SM.getBufferData(locInfo.first, &invalidTemp);
338 if (invalidTemp) {
339 if (invalid) *invalid = true;
340 return {};
341 }
342
343 const char *tokenBegin = file.data() + locInfo.second;
344
345 // Lex from the start of the given location.
346 Lexer lexer(SM.getLocForStartOfFile(locInfo.first), options,
347 file.begin(), tokenBegin, file.end());
348 Token token;
349 lexer.LexFromRawLexer(token);
350
351 unsigned length = token.getLength();
352
353 // Common case: no need for cleaning.
354 if (!token.needsCleaning())
355 return StringRef(tokenBegin, length);
356
357 // Hard case, we need to relex the characters into the string.
358 buffer.resize(length);
359 buffer.resize(getSpellingSlow(token, tokenBegin, options, buffer.data()));
360 return StringRef(buffer.data(), buffer.size());
361}
362
363/// getSpelling() - Return the 'spelling' of this token. The spelling of a
364/// token are the characters used to represent the token in the source file
365/// after trigraph expansion and escaped-newline folding. In particular, this
366/// wants to get the true, uncanonicalized, spelling of things like digraphs
367/// UCNs, etc.
368std::string Lexer::getSpelling(const Token &Tok, const SourceManager &SourceMgr,
369 const LangOptions &LangOpts, bool *Invalid) {
370 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!")((void)0);
371
372 bool CharDataInvalid = false;
373 const char *TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
374 &CharDataInvalid);
375 if (Invalid)
376 *Invalid = CharDataInvalid;
377 if (CharDataInvalid)
378 return {};
379
380 // If this token contains nothing interesting, return it directly.
381 if (!Tok.needsCleaning())
382 return std::string(TokStart, TokStart + Tok.getLength());
383
384 std::string Result;
385 Result.resize(Tok.getLength());
386 Result.resize(getSpellingSlow(Tok, TokStart, LangOpts, &*Result.begin()));
387 return Result;
388}
389
390/// getSpelling - This method is used to get the spelling of a token into a
391/// preallocated buffer, instead of as an std::string. The caller is required
392/// to allocate enough space for the token, which is guaranteed to be at least
393/// Tok.getLength() bytes long. The actual length of the token is returned.
394///
395/// Note that this method may do two possible things: it may either fill in
396/// the buffer specified with characters, or it may *change the input pointer*
397/// to point to a constant buffer with the data already in it (avoiding a
398/// copy). The caller is not allowed to modify the returned buffer pointer
399/// if an internal buffer is returned.
400unsigned Lexer::getSpelling(const Token &Tok, const char *&Buffer,
401 const SourceManager &SourceMgr,
402 const LangOptions &LangOpts, bool *Invalid) {
403 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!")((void)0);
404
405 const char *TokStart = nullptr;
406 // NOTE: this has to be checked *before* testing for an IdentifierInfo.
407 if (Tok.is(tok::raw_identifier))
408 TokStart = Tok.getRawIdentifier().data();
409 else if (!Tok.hasUCN()) {
410 if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
411 // Just return the string from the identifier table, which is very quick.
412 Buffer = II->getNameStart();
413 return II->getLength();
414 }
415 }
416
417 // NOTE: this can be checked even after testing for an IdentifierInfo.
418 if (Tok.isLiteral())
419 TokStart = Tok.getLiteralData();
420
421 if (!TokStart) {
422 // Compute the start of the token in the input lexer buffer.
423 bool CharDataInvalid = false;
424 TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
425 if (Invalid)
426 *Invalid = CharDataInvalid;
427 if (CharDataInvalid) {
428 Buffer = "";
429 return 0;
430 }
431 }
432
433 // If this token contains nothing interesting, return it directly.
434 if (!Tok.needsCleaning()) {
435 Buffer = TokStart;
436 return Tok.getLength();
437 }
438
439 // Otherwise, hard case, relex the characters into the string.
440 return getSpellingSlow(Tok, TokStart, LangOpts, const_cast<char*>(Buffer));
441}
442
443/// MeasureTokenLength - Relex the token at the specified location and return
444/// its length in bytes in the input file. If the token needs cleaning (e.g.
445/// includes a trigraph or an escaped newline) then this count includes bytes
446/// that are part of that.
447unsigned Lexer::MeasureTokenLength(SourceLocation Loc,
448 const SourceManager &SM,
449 const LangOptions &LangOpts) {
450 Token TheTok;
451 if (getRawToken(Loc, TheTok, SM, LangOpts))
452 return 0;
453 return TheTok.getLength();
454}
455
456/// Relex the token at the specified location.
457/// \returns true if there was a failure, false on success.
458bool Lexer::getRawToken(SourceLocation Loc, Token &Result,
459 const SourceManager &SM,
460 const LangOptions &LangOpts,
461 bool IgnoreWhiteSpace) {
462 // TODO: this could be special cased for common tokens like identifiers, ')',
463 // etc to make this faster, if it mattered. Just look at StrData[0] to handle
464 // all obviously single-char tokens. This could use
465 // Lexer::isObviouslySimpleCharacter for example to handle identifiers or
466 // something.
467
468 // If this comes from a macro expansion, we really do want the macro name, not
469 // the token this macro expanded to.
470 Loc = SM.getExpansionLoc(Loc);
471 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
472 bool Invalid = false;
473 StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
474 if (Invalid)
475 return true;
476
477 const char *StrData = Buffer.data()+LocInfo.second;
478
479 if (!IgnoreWhiteSpace && isWhitespace(StrData[0]))
480 return true;
481
482 // Create a lexer starting at the beginning of this token.
483 Lexer TheLexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts,
484 Buffer.begin(), StrData, Buffer.end());
485 TheLexer.SetCommentRetentionState(true);
486 TheLexer.LexFromRawLexer(Result);
487 return false;
488}
489
490/// Returns the pointer that points to the beginning of line that contains
491/// the given offset, or null if the offset if invalid.
492static const char *findBeginningOfLine(StringRef Buffer, unsigned Offset) {
493 const char *BufStart = Buffer.data();
494 if (Offset >= Buffer.size())
495 return nullptr;
496
497 const char *LexStart = BufStart + Offset;
498 for (; LexStart != BufStart; --LexStart) {
499 if (isVerticalWhitespace(LexStart[0]) &&
500 !Lexer::isNewLineEscaped(BufStart, LexStart)) {
501 // LexStart should point at first character of logical line.
502 ++LexStart;
503 break;
504 }
505 }
506 return LexStart;
507}
508
509static SourceLocation getBeginningOfFileToken(SourceLocation Loc,
510 const SourceManager &SM,
511 const LangOptions &LangOpts) {
512 assert(Loc.isFileID())((void)0);
513 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
514 if (LocInfo.first.isInvalid())
515 return Loc;
516
517 bool Invalid = false;
518 StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
519 if (Invalid)
520 return Loc;
521
522 // Back up from the current location until we hit the beginning of a line
523 // (or the buffer). We'll relex from that point.
524 const char *StrData = Buffer.data() + LocInfo.second;
525 const char *LexStart = findBeginningOfLine(Buffer, LocInfo.second);
526 if (!LexStart || LexStart == StrData)
527 return Loc;
528
529 // Create a lexer starting at the beginning of this token.
530 SourceLocation LexerStartLoc = Loc.getLocWithOffset(-LocInfo.second);
531 Lexer TheLexer(LexerStartLoc, LangOpts, Buffer.data(), LexStart,
532 Buffer.end());
533 TheLexer.SetCommentRetentionState(true);
534
535 // Lex tokens until we find the token that contains the source location.
536 Token TheTok;
537 do {
538 TheLexer.LexFromRawLexer(TheTok);
539
540 if (TheLexer.getBufferLocation() > StrData) {
541 // Lexing this token has taken the lexer past the source location we're
542 // looking for. If the current token encompasses our source location,
543 // return the beginning of that token.
544 if (TheLexer.getBufferLocation() - TheTok.getLength() <= StrData)
545 return TheTok.getLocation();
546
547 // We ended up skipping over the source location entirely, which means
548 // that it points into whitespace. We're done here.
549 break;
550 }
551 } while (TheTok.getKind() != tok::eof);
552
553 // We've passed our source location; just return the original source location.
554 return Loc;
555}
556
557SourceLocation Lexer::GetBeginningOfToken(SourceLocation Loc,
558 const SourceManager &SM,
559 const LangOptions &LangOpts) {
560 if (Loc.isFileID())
561 return getBeginningOfFileToken(Loc, SM, LangOpts);
562
563 if (!SM.isMacroArgExpansion(Loc))
564 return Loc;
565
566 SourceLocation FileLoc = SM.getSpellingLoc(Loc);
567 SourceLocation BeginFileLoc = getBeginningOfFileToken(FileLoc, SM, LangOpts);
568 std::pair<FileID, unsigned> FileLocInfo = SM.getDecomposedLoc(FileLoc);
569 std::pair<FileID, unsigned> BeginFileLocInfo =
570 SM.getDecomposedLoc(BeginFileLoc);
571 assert(FileLocInfo.first == BeginFileLocInfo.first &&((void)0)
572 FileLocInfo.second >= BeginFileLocInfo.second)((void)0);
573 return Loc.getLocWithOffset(BeginFileLocInfo.second - FileLocInfo.second);
574}
575
576namespace {
577
578enum PreambleDirectiveKind {
579 PDK_Skipped,
580 PDK_Unknown
581};
582
583} // namespace
584
585PreambleBounds Lexer::ComputePreamble(StringRef Buffer,
586 const LangOptions &LangOpts,
587 unsigned MaxLines) {
588 // Create a lexer starting at the beginning of the file. Note that we use a
589 // "fake" file source location at offset 1 so that the lexer will track our
590 // position within the file.
591 const SourceLocation::UIntTy StartOffset = 1;
592 SourceLocation FileLoc = SourceLocation::getFromRawEncoding(StartOffset);
593 Lexer TheLexer(FileLoc, LangOpts, Buffer.begin(), Buffer.begin(),
594 Buffer.end());
595 TheLexer.SetCommentRetentionState(true);
596
597 bool InPreprocessorDirective = false;
598 Token TheTok;
599 SourceLocation ActiveCommentLoc;
600
601 unsigned MaxLineOffset = 0;
602 if (MaxLines) {
603 const char *CurPtr = Buffer.begin();
604 unsigned CurLine = 0;
605 while (CurPtr != Buffer.end()) {
606 char ch = *CurPtr++;
607 if (ch == '\n') {
608 ++CurLine;
609 if (CurLine == MaxLines)
610 break;
611 }
612 }
613 if (CurPtr != Buffer.end())
614 MaxLineOffset = CurPtr - Buffer.begin();
615 }
616
617 do {
618 TheLexer.LexFromRawLexer(TheTok);
619
620 if (InPreprocessorDirective) {
621 // If we've hit the end of the file, we're done.
622 if (TheTok.getKind() == tok::eof) {
623 break;
624 }
625
626 // If we haven't hit the end of the preprocessor directive, skip this
627 // token.
628 if (!TheTok.isAtStartOfLine())
629 continue;
630
631 // We've passed the end of the preprocessor directive, and will look
632 // at this token again below.
633 InPreprocessorDirective = false;
634 }
635
636 // Keep track of the # of lines in the preamble.
637 if (TheTok.isAtStartOfLine()) {
638 unsigned TokOffset = TheTok.getLocation().getRawEncoding() - StartOffset;
639
640 // If we were asked to limit the number of lines in the preamble,
641 // and we're about to exceed that limit, we're done.
642 if (MaxLineOffset && TokOffset >= MaxLineOffset)
643 break;
644 }
645
646 // Comments are okay; skip over them.
647 if (TheTok.getKind() == tok::comment) {
648 if (ActiveCommentLoc.isInvalid())
649 ActiveCommentLoc = TheTok.getLocation();
650 continue;
651 }
652
653 if (TheTok.isAtStartOfLine() && TheTok.getKind() == tok::hash) {
654 // This is the start of a preprocessor directive.
655 Token HashTok = TheTok;
656 InPreprocessorDirective = true;
657 ActiveCommentLoc = SourceLocation();
658
659 // Figure out which directive this is. Since we're lexing raw tokens,
660 // we don't have an identifier table available. Instead, just look at
661 // the raw identifier to recognize and categorize preprocessor directives.
662 TheLexer.LexFromRawLexer(TheTok);
663 if (TheTok.getKind() == tok::raw_identifier && !TheTok.needsCleaning()) {
664 StringRef Keyword = TheTok.getRawIdentifier();
665 PreambleDirectiveKind PDK
666 = llvm::StringSwitch<PreambleDirectiveKind>(Keyword)
667 .Case("include", PDK_Skipped)
668 .Case("__include_macros", PDK_Skipped)
669 .Case("define", PDK_Skipped)
670 .Case("undef", PDK_Skipped)
671 .Case("line", PDK_Skipped)
672 .Case("error", PDK_Skipped)
673 .Case("pragma", PDK_Skipped)
674 .Case("import", PDK_Skipped)
675 .Case("include_next", PDK_Skipped)
676 .Case("warning", PDK_Skipped)
677 .Case("ident", PDK_Skipped)
678 .Case("sccs", PDK_Skipped)
679 .Case("assert", PDK_Skipped)
680 .Case("unassert", PDK_Skipped)
681 .Case("if", PDK_Skipped)
682 .Case("ifdef", PDK_Skipped)
683 .Case("ifndef", PDK_Skipped)
684 .Case("elif", PDK_Skipped)
685 .Case("elifdef", PDK_Skipped)
686 .Case("elifndef", PDK_Skipped)
687 .Case("else", PDK_Skipped)
688 .Case("endif", PDK_Skipped)
689 .Default(PDK_Unknown);
690
691 switch (PDK) {
692 case PDK_Skipped:
693 continue;
694
695 case PDK_Unknown:
696 // We don't know what this directive is; stop at the '#'.
697 break;
698 }
699 }
700
701 // We only end up here if we didn't recognize the preprocessor
702 // directive or it was one that can't occur in the preamble at this
703 // point. Roll back the current token to the location of the '#'.
704 TheTok = HashTok;
705 }
706
707 // We hit a token that we don't recognize as being in the
708 // "preprocessing only" part of the file, so we're no longer in
709 // the preamble.
710 break;
711 } while (true);
712
713 SourceLocation End;
714 if (ActiveCommentLoc.isValid())
715 End = ActiveCommentLoc; // don't truncate a decl comment.
716 else
717 End = TheTok.getLocation();
718
719 return PreambleBounds(End.getRawEncoding() - FileLoc.getRawEncoding(),
720 TheTok.isAtStartOfLine());
721}
722
723unsigned Lexer::getTokenPrefixLength(SourceLocation TokStart, unsigned CharNo,
724 const SourceManager &SM,
725 const LangOptions &LangOpts) {
726 // Figure out how many physical characters away the specified expansion
727 // character is. This needs to take into consideration newlines and
728 // trigraphs.
729 bool Invalid = false;
730 const char *TokPtr = SM.getCharacterData(TokStart, &Invalid);
731
732 // If they request the first char of the token, we're trivially done.
733 if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
734 return 0;
735
736 unsigned PhysOffset = 0;
737
738 // The usual case is that tokens don't contain anything interesting. Skip
739 // over the uninteresting characters. If a token only consists of simple
740 // chars, this method is extremely fast.
741 while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
742 if (CharNo == 0)
743 return PhysOffset;
744 ++TokPtr;
745 --CharNo;
746 ++PhysOffset;
747 }
748
749 // If we have a character that may be a trigraph or escaped newline, use a
750 // lexer to parse it correctly.
751 for (; CharNo; --CharNo) {
752 unsigned Size;
753 Lexer::getCharAndSizeNoWarn(TokPtr, Size, LangOpts);
754 TokPtr += Size;
755 PhysOffset += Size;
756 }
757
758 // Final detail: if we end up on an escaped newline, we want to return the
759 // location of the actual byte of the token. For example foo\<newline>bar
760 // advanced by 3 should return the location of b, not of \\. One compounding
761 // detail of this is that the escape may be made by a trigraph.
762 if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
763 PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
764
765 return PhysOffset;
766}
767
768/// Computes the source location just past the end of the
769/// token at this source location.
770///
771/// This routine can be used to produce a source location that
772/// points just past the end of the token referenced by \p Loc, and
773/// is generally used when a diagnostic needs to point just after a
774/// token where it expected something different that it received. If
775/// the returned source location would not be meaningful (e.g., if
776/// it points into a macro), this routine returns an invalid
777/// source location.
778///
779/// \param Offset an offset from the end of the token, where the source
780/// location should refer to. The default offset (0) produces a source
781/// location pointing just past the end of the token; an offset of 1 produces
782/// a source location pointing to the last character in the token, etc.
783SourceLocation Lexer::getLocForEndOfToken(SourceLocation Loc, unsigned Offset,
784 const SourceManager &SM,
785 const LangOptions &LangOpts) {
786 if (Loc.isInvalid())
787 return {};
788
789 if (Loc.isMacroID()) {
790 if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
791 return {}; // Points inside the macro expansion.
792 }
793
794 unsigned Len = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
795 if (Len > Offset)
796 Len = Len - Offset;
797 else
798 return Loc;
799
800 return Loc.getLocWithOffset(Len);
801}
802
803/// Returns true if the given MacroID location points at the first
804/// token of the macro expansion.
805bool Lexer::isAtStartOfMacroExpansion(SourceLocation loc,
806 const SourceManager &SM,
807 const LangOptions &LangOpts,
808 SourceLocation *MacroBegin) {
809 assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc")((void)0);
810
811 SourceLocation expansionLoc;
812 if (!SM.isAtStartOfImmediateMacroExpansion(loc, &expansionLoc))
813 return false;
814
815 if (expansionLoc.isFileID()) {
816 // No other macro expansions, this is the first.
817 if (MacroBegin)
818 *MacroBegin = expansionLoc;
819 return true;
820 }
821
822 return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts, MacroBegin);
823}
824
825/// Returns true if the given MacroID location points at the last
826/// token of the macro expansion.
827bool Lexer::isAtEndOfMacroExpansion(SourceLocation loc,
828 const SourceManager &SM,
829 const LangOptions &LangOpts,
830 SourceLocation *MacroEnd) {
831 assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc")((void)0);
832
833 SourceLocation spellLoc = SM.getSpellingLoc(loc);
834 unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
835 if (tokLen == 0)
836 return false;
837
838 SourceLocation afterLoc = loc.getLocWithOffset(tokLen);
839 SourceLocation expansionLoc;
840 if (!SM.isAtEndOfImmediateMacroExpansion(afterLoc, &expansionLoc))
841 return false;
842
843 if (expansionLoc.isFileID()) {
844 // No other macro expansions.
845 if (MacroEnd)
846 *MacroEnd = expansionLoc;
847 return true;
848 }
849
850 return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts, MacroEnd);
851}
852
853static CharSourceRange makeRangeFromFileLocs(CharSourceRange Range,
854 const SourceManager &SM,
855 const LangOptions &LangOpts) {
856 SourceLocation Begin = Range.getBegin();
857 SourceLocation End = Range.getEnd();
858 assert(Begin.isFileID() && End.isFileID())((void)0);
859 if (Range.isTokenRange()) {
860 End = Lexer::getLocForEndOfToken(End, 0, SM,LangOpts);
861 if (End.isInvalid())
862 return {};
863 }
864
865 // Break down the source locations.
866 FileID FID;
867 unsigned BeginOffs;
868 std::tie(FID, BeginOffs) = SM.getDecomposedLoc(Begin);
869 if (FID.isInvalid())
870 return {};
871
872 unsigned EndOffs;
873 if (!SM.isInFileID(End, FID, &EndOffs) ||
874 BeginOffs > EndOffs)
875 return {};
876
877 return CharSourceRange::getCharRange(Begin, End);
878}
879
880// Assumes that `Loc` is in an expansion.
881static bool isInExpansionTokenRange(const SourceLocation Loc,
882 const SourceManager &SM) {
883 return SM.getSLocEntry(SM.getFileID(Loc))
884 .getExpansion()
885 .isExpansionTokenRange();
886}
887
888CharSourceRange Lexer::makeFileCharRange(CharSourceRange Range,
889 const SourceManager &SM,
890 const LangOptions &LangOpts) {
891 SourceLocation Begin = Range.getBegin();
892 SourceLocation End = Range.getEnd();
893 if (Begin.isInvalid() || End.isInvalid())
894 return {};
895
896 if (Begin.isFileID() && End.isFileID())
897 return makeRangeFromFileLocs(Range, SM, LangOpts);
898
899 if (Begin.isMacroID() && End.isFileID()) {
900 if (!isAtStartOfMacroExpansion(Begin, SM, LangOpts, &Begin))
901 return {};
902 Range.setBegin(Begin);
903 return makeRangeFromFileLocs(Range, SM, LangOpts);
904 }
905
906 if (Begin.isFileID() && End.isMacroID()) {
907 if (Range.isTokenRange()) {
908 if (!isAtEndOfMacroExpansion(End, SM, LangOpts, &End))
909 return {};
910 // Use the *original* end, not the expanded one in `End`.
911 Range.setTokenRange(isInExpansionTokenRange(Range.getEnd(), SM));
912 } else if (!isAtStartOfMacroExpansion(End, SM, LangOpts, &End))
913 return {};
914 Range.setEnd(End);
915 return makeRangeFromFileLocs(Range, SM, LangOpts);
916 }
917
918 assert(Begin.isMacroID() && End.isMacroID())((void)0);
919 SourceLocation MacroBegin, MacroEnd;
920 if (isAtStartOfMacroExpansion(Begin, SM, LangOpts, &MacroBegin) &&
921 ((Range.isTokenRange() && isAtEndOfMacroExpansion(End, SM, LangOpts,
922 &MacroEnd)) ||
923 (Range.isCharRange() && isAtStartOfMacroExpansion(End, SM, LangOpts,
924 &MacroEnd)))) {
925 Range.setBegin(MacroBegin);
926 Range.setEnd(MacroEnd);
927 // Use the *original* `End`, not the expanded one in `MacroEnd`.
928 if (Range.isTokenRange())
929 Range.setTokenRange(isInExpansionTokenRange(End, SM));
930 return makeRangeFromFileLocs(Range, SM, LangOpts);
931 }
932
933 bool Invalid = false;
934 const SrcMgr::SLocEntry &BeginEntry = SM.getSLocEntry(SM.getFileID(Begin),
935 &Invalid);
936 if (Invalid)
937 return {};
938
939 if (BeginEntry.getExpansion().isMacroArgExpansion()) {
940 const SrcMgr::SLocEntry &EndEntry = SM.getSLocEntry(SM.getFileID(End),
941 &Invalid);
942 if (Invalid)
943 return {};
944
945 if (EndEntry.getExpansion().isMacroArgExpansion() &&
946 BeginEntry.getExpansion().getExpansionLocStart() ==
947 EndEntry.getExpansion().getExpansionLocStart()) {
948 Range.setBegin(SM.getImmediateSpellingLoc(Begin));
949 Range.setEnd(SM.getImmediateSpellingLoc(End));
950 return makeFileCharRange(Range, SM, LangOpts);
951 }
952 }
953
954 return {};
955}
956
957StringRef Lexer::getSourceText(CharSourceRange Range,
958 const SourceManager &SM,
959 const LangOptions &LangOpts,
960 bool *Invalid) {
961 Range = makeFileCharRange(Range, SM, LangOpts);
962 if (Range.isInvalid()) {
963 if (Invalid) *Invalid = true;
964 return {};
965 }
966
967 // Break down the source location.
968 std::pair<FileID, unsigned> beginInfo = SM.getDecomposedLoc(Range.getBegin());
969 if (beginInfo.first.isInvalid()) {
970 if (Invalid) *Invalid = true;
971 return {};
972 }
973
974 unsigned EndOffs;
975 if (!SM.isInFileID(Range.getEnd(), beginInfo.first, &EndOffs) ||
976 beginInfo.second > EndOffs) {
977 if (Invalid) *Invalid = true;
978 return {};
979 }
980
981 // Try to the load the file buffer.
982 bool invalidTemp = false;
983 StringRef file = SM.getBufferData(beginInfo.first, &invalidTemp);
984 if (invalidTemp) {
985 if (Invalid) *Invalid = true;
986 return {};
987 }
988
989 if (Invalid) *Invalid = false;
990 return file.substr(beginInfo.second, EndOffs - beginInfo.second);
991}
992
993StringRef Lexer::getImmediateMacroName(SourceLocation Loc,
994 const SourceManager &SM,
995 const LangOptions &LangOpts) {
996 assert(Loc.isMacroID() && "Only reasonable to call this on macros")((void)0);
997
998 // Find the location of the immediate macro expansion.
999 while (true) {
1000 FileID FID = SM.getFileID(Loc);
1001 const SrcMgr::SLocEntry *E = &SM.getSLocEntry(FID);
1002 const SrcMgr::ExpansionInfo &Expansion = E->getExpansion();
1003 Loc = Expansion.getExpansionLocStart();
1004 if (!Expansion.isMacroArgExpansion())
1005 break;
1006
1007 // For macro arguments we need to check that the argument did not come
1008 // from an inner macro, e.g: "MAC1( MAC2(foo) )"
1009
1010 // Loc points to the argument id of the macro definition, move to the
1011 // macro expansion.
1012 Loc = SM.getImmediateExpansionRange(Loc).getBegin();
1013 SourceLocation SpellLoc = Expansion.getSpellingLoc();
1014 if (SpellLoc.isFileID())
1015 break; // No inner macro.
1016
1017 // If spelling location resides in the same FileID as macro expansion
1018 // location, it means there is no inner macro.
1019 FileID MacroFID = SM.getFileID(Loc);
1020 if (SM.isInFileID(SpellLoc, MacroFID))
1021 break;
1022
1023 // Argument came from inner macro.
1024 Loc = SpellLoc;
1025 }
1026
1027 // Find the spelling location of the start of the non-argument expansion
1028 // range. This is where the macro name was spelled in order to begin
1029 // expanding this macro.
1030 Loc = SM.getSpellingLoc(Loc);
1031
1032 // Dig out the buffer where the macro name was spelled and the extents of the
1033 // name so that we can render it into the expansion note.
1034 std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1035 unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1036 StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1037 return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1038}
1039
1040StringRef Lexer::getImmediateMacroNameForDiagnostics(
1041 SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts) {
1042 assert(Loc.isMacroID() && "Only reasonable to call this on macros")((void)0);
1043 // Walk past macro argument expansions.
1044 while (SM.isMacroArgExpansion(Loc))
1045 Loc = SM.getImmediateExpansionRange(Loc).getBegin();
1046
1047 // If the macro's spelling has no FileID, then it's actually a token paste
1048 // or stringization (or similar) and not a macro at all.
1049 if (!SM.getFileEntryForID(SM.getFileID(SM.getSpellingLoc(Loc))))
1050 return {};
1051
1052 // Find the spelling location of the start of the non-argument expansion
1053 // range. This is where the macro name was spelled in order to begin
1054 // expanding this macro.
1055 Loc = SM.getSpellingLoc(SM.getImmediateExpansionRange(Loc).getBegin());
1056
1057 // Dig out the buffer where the macro name was spelled and the extents of the
1058 // name so that we can render it into the expansion note.
1059 std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1060 unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1061 StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1062 return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1063}
1064
1065bool Lexer::isIdentifierBodyChar(char c, const LangOptions &LangOpts) {
1066 return isIdentifierBody(c, LangOpts.DollarIdents);
1067}
1068
1069bool Lexer::isNewLineEscaped(const char *BufferStart, const char *Str) {
1070 assert(isVerticalWhitespace(Str[0]))((void)0);
1071 if (Str - 1 < BufferStart)
1072 return false;
1073
1074 if ((Str[0] == '\n' && Str[-1] == '\r') ||
1075 (Str[0] == '\r' && Str[-1] == '\n')) {
1076 if (Str - 2 < BufferStart)
1077 return false;
1078 --Str;
1079 }
1080 --Str;
1081
1082 // Rewind to first non-space character:
1083 while (Str > BufferStart && isHorizontalWhitespace(*Str))
1084 --Str;
1085
1086 return *Str == '\\';
1087}
1088
1089StringRef Lexer::getIndentationForLine(SourceLocation Loc,
1090 const SourceManager &SM) {
1091 if (Loc.isInvalid() || Loc.isMacroID())
1092 return {};
1093 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1094 if (LocInfo.first.isInvalid())
1095 return {};
1096 bool Invalid = false;
1097 StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
1098 if (Invalid)
1099 return {};
1100 const char *Line = findBeginningOfLine(Buffer, LocInfo.second);
1101 if (!Line)
1102 return {};
1103 StringRef Rest = Buffer.substr(Line - Buffer.data());
1104 size_t NumWhitespaceChars = Rest.find_first_not_of(" \t");
1105 return NumWhitespaceChars == StringRef::npos
1106 ? ""
1107 : Rest.take_front(NumWhitespaceChars);
1108}
1109
1110//===----------------------------------------------------------------------===//
1111// Diagnostics forwarding code.
1112//===----------------------------------------------------------------------===//
1113
1114/// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
1115/// lexer buffer was all expanded at a single point, perform the mapping.
1116/// This is currently only used for _Pragma implementation, so it is the slow
1117/// path of the hot getSourceLocation method. Do not allow it to be inlined.
1118static LLVM_ATTRIBUTE_NOINLINE__attribute__((noinline)) SourceLocation GetMappedTokenLoc(
1119 Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
1120static SourceLocation GetMappedTokenLoc(Preprocessor &PP,
1121 SourceLocation FileLoc,
1122 unsigned CharNo, unsigned TokLen) {
1123 assert(FileLoc.isMacroID() && "Must be a macro expansion")((void)0);
1124
1125 // Otherwise, we're lexing "mapped tokens". This is used for things like
1126 // _Pragma handling. Combine the expansion location of FileLoc with the
1127 // spelling location.
1128 SourceManager &SM = PP.getSourceManager();
1129
1130 // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
1131 // characters come from spelling(FileLoc)+Offset.
1132 SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
1133 SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
1134
1135 // Figure out the expansion loc range, which is the range covered by the
1136 // original _Pragma(...) sequence.
1137 CharSourceRange II = SM.getImmediateExpansionRange(FileLoc);
1138
1139 return SM.createExpansionLoc(SpellingLoc, II.getBegin(), II.getEnd(), TokLen);
1140}
1141
1142/// getSourceLocation - Return a source location identifier for the specified
1143/// offset in the current file.
1144SourceLocation Lexer::getSourceLocation(const char *Loc,
1145 unsigned TokLen) const {
1146 assert(Loc >= BufferStart && Loc <= BufferEnd &&((void)0)
1147 "Location out of range for this buffer!")((void)0);
1148
1149 // In the normal case, we're just lexing from a simple file buffer, return
1150 // the file id from FileLoc with the offset specified.
1151 unsigned CharNo = Loc-BufferStart;
1152 if (FileLoc.isFileID())
1153 return FileLoc.getLocWithOffset(CharNo);
1154
1155 // Otherwise, this is the _Pragma lexer case, which pretends that all of the
1156 // tokens are lexed from where the _Pragma was defined.
1157 assert(PP && "This doesn't work on raw lexers")((void)0);
1158 return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
1159}
1160
1161/// Diag - Forwarding function for diagnostics. This translate a source
1162/// position in the current buffer into a SourceLocation object for rendering.
1163DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
1164 return PP->Diag(getSourceLocation(Loc), DiagID);
40
Called C++ object pointer is null
1165}
1166
1167//===----------------------------------------------------------------------===//
1168// Trigraph and Escaped Newline Handling Code.
1169//===----------------------------------------------------------------------===//
1170
1171/// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
1172/// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
1173static char GetTrigraphCharForLetter(char Letter) {
1174 switch (Letter) {
1175 default: return 0;
1176 case '=': return '#';
1177 case ')': return ']';
1178 case '(': return '[';
1179 case '!': return '|';
1180 case '\'': return '^';
1181 case '>': return '}';
1182 case '/': return '\\';
1183 case '<': return '{';
1184 case '-': return '~';
1185 }
1186}
1187
1188/// DecodeTrigraphChar - If the specified character is a legal trigraph when
1189/// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1190/// return the result character. Finally, emit a warning about trigraph use
1191/// whether trigraphs are enabled or not.
1192static char DecodeTrigraphChar(const char *CP, Lexer *L) {
1193 char Res = GetTrigraphCharForLetter(*CP);
1194 if (!Res || !L) return Res;
1195
1196 if (!L->getLangOpts().Trigraphs) {
1197 if (!L->isLexingRawMode())
1198 L->Diag(CP-2, diag::trigraph_ignored);
1199 return 0;
1200 }
1201
1202 if (!L->isLexingRawMode())
1203 L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1204 return Res;
1205}
1206
1207/// getEscapedNewLineSize - Return the size of the specified escaped newline,
1208/// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1209/// trigraph equivalent on entry to this function.
1210unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1211 unsigned Size = 0;
1212 while (isWhitespace(Ptr[Size])) {
1213 ++Size;
1214
1215 if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1216 continue;
1217
1218 // If this is a \r\n or \n\r, skip the other half.
1219 if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1220 Ptr[Size-1] != Ptr[Size])
1221 ++Size;
1222
1223 return Size;
1224 }
1225
1226 // Not an escaped newline, must be a \t or something else.
1227 return 0;
1228}
1229
1230/// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1231/// them), skip over them and return the first non-escaped-newline found,
1232/// otherwise return P.
1233const char *Lexer::SkipEscapedNewLines(const char *P) {
1234 while (true) {
1235 const char *AfterEscape;
1236 if (*P == '\\') {
1237 AfterEscape = P+1;
1238 } else if (*P == '?') {
1239 // If not a trigraph for escape, bail out.
1240 if (P[1] != '?' || P[2] != '/')
1241 return P;
1242 // FIXME: Take LangOpts into account; the language might not
1243 // support trigraphs.
1244 AfterEscape = P+3;
1245 } else {
1246 return P;
1247 }
1248
1249 unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1250 if (NewLineSize == 0) return P;
1251 P = AfterEscape+NewLineSize;
1252 }
1253}
1254
1255Optional<Token> Lexer::findNextToken(SourceLocation Loc,
1256 const SourceManager &SM,
1257 const LangOptions &LangOpts) {
1258 if (Loc.isMacroID()) {
1259 if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
1260 return None;
1261 }
1262 Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1263
1264 // Break down the source location.
1265 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1266
1267 // Try to load the file buffer.
1268 bool InvalidTemp = false;
1269 StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1270 if (InvalidTemp)
1271 return None;
1272
1273 const char *TokenBegin = File.data() + LocInfo.second;
1274
1275 // Lex from the start of the given location.
1276 Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1277 TokenBegin, File.end());
1278 // Find the token.
1279 Token Tok;
1280 lexer.LexFromRawLexer(Tok);
1281 return Tok;
1282}
1283
1284/// Checks that the given token is the first token that occurs after the
1285/// given location (this excludes comments and whitespace). Returns the location
1286/// immediately after the specified token. If the token is not found or the
1287/// location is inside a macro, the returned source location will be invalid.
1288SourceLocation Lexer::findLocationAfterToken(
1289 SourceLocation Loc, tok::TokenKind TKind, const SourceManager &SM,
1290 const LangOptions &LangOpts, bool SkipTrailingWhitespaceAndNewLine) {
1291 Optional<Token> Tok = findNextToken(Loc, SM, LangOpts);
1292 if (!Tok || Tok->isNot(TKind))
1293 return {};
1294 SourceLocation TokenLoc = Tok->getLocation();
1295
1296 // Calculate how much whitespace needs to be skipped if any.
1297 unsigned NumWhitespaceChars = 0;
1298 if (SkipTrailingWhitespaceAndNewLine) {
1299 const char *TokenEnd = SM.getCharacterData(TokenLoc) + Tok->getLength();
1300 unsigned char C = *TokenEnd;
1301 while (isHorizontalWhitespace(C)) {
1302 C = *(++TokenEnd);
1303 NumWhitespaceChars++;
1304 }
1305
1306 // Skip \r, \n, \r\n, or \n\r
1307 if (C == '\n' || C == '\r') {
1308 char PrevC = C;
1309 C = *(++TokenEnd);
1310 NumWhitespaceChars++;
1311 if ((C == '\n' || C == '\r') && C != PrevC)
1312 NumWhitespaceChars++;
1313 }
1314 }
1315
1316 return TokenLoc.getLocWithOffset(Tok->getLength() + NumWhitespaceChars);
1317}
1318
1319/// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1320/// get its size, and return it. This is tricky in several cases:
1321/// 1. If currently at the start of a trigraph, we warn about the trigraph,
1322/// then either return the trigraph (skipping 3 chars) or the '?',
1323/// depending on whether trigraphs are enabled or not.
1324/// 2. If this is an escaped newline (potentially with whitespace between
1325/// the backslash and newline), implicitly skip the newline and return
1326/// the char after it.
1327///
1328/// This handles the slow/uncommon case of the getCharAndSize method. Here we
1329/// know that we can accumulate into Size, and that we have already incremented
1330/// Ptr by Size bytes.
1331///
1332/// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1333/// be updated to match.
1334char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1335 Token *Tok) {
1336 // If we have a slash, look for an escaped newline.
1337 if (Ptr[0] == '\\') {
1338 ++Size;
1339 ++Ptr;
1340Slash:
1341 // Common case, backslash-char where the char is not whitespace.
1342 if (!isWhitespace(Ptr[0])) return '\\';
1343
1344 // See if we have optional whitespace characters between the slash and
1345 // newline.
1346 if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1347 // Remember that this token needs to be cleaned.
1348 if (Tok) Tok->setFlag(Token::NeedsCleaning);
1349
1350 // Warn if there was whitespace between the backslash and newline.
1351 if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1352 Diag(Ptr, diag::backslash_newline_space);
1353
1354 // Found backslash<whitespace><newline>. Parse the char after it.
1355 Size += EscapedNewLineSize;
1356 Ptr += EscapedNewLineSize;
1357
1358 // Use slow version to accumulate a correct size field.
1359 return getCharAndSizeSlow(Ptr, Size, Tok);
1360 }
1361
1362 // Otherwise, this is not an escaped newline, just return the slash.
1363 return '\\';
1364 }
1365
1366 // If this is a trigraph, process it.
1367 if (Ptr[0] == '?' && Ptr[1] == '?') {
1368 // If this is actually a legal trigraph (not something like "??x"), emit
1369 // a trigraph warning. If so, and if trigraphs are enabled, return it.
1370 if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : nullptr)) {
1371 // Remember that this token needs to be cleaned.
1372 if (Tok) Tok->setFlag(Token::NeedsCleaning);
1373
1374 Ptr += 3;
1375 Size += 3;
1376 if (C == '\\') goto Slash;
1377 return C;
1378 }
1379 }
1380
1381 // If this is neither, return a single character.
1382 ++Size;
1383 return *Ptr;
1384}
1385
1386/// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1387/// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1388/// and that we have already incremented Ptr by Size bytes.
1389///
1390/// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1391/// be updated to match.
1392char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1393 const LangOptions &LangOpts) {
1394 // If we have a slash, look for an escaped newline.
1395 if (Ptr[0] == '\\') {
1396 ++Size;
1397 ++Ptr;
1398Slash:
1399 // Common case, backslash-char where the char is not whitespace.
1400 if (!isWhitespace(Ptr[0])) return '\\';
1401
1402 // See if we have optional whitespace characters followed by a newline.
1403 if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1404 // Found backslash<whitespace><newline>. Parse the char after it.
1405 Size += EscapedNewLineSize;
1406 Ptr += EscapedNewLineSize;
1407
1408 // Use slow version to accumulate a correct size field.
1409 return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
1410 }
1411
1412 // Otherwise, this is not an escaped newline, just return the slash.
1413 return '\\';
1414 }
1415
1416 // If this is a trigraph, process it.
1417 if (LangOpts.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1418 // If this is actually a legal trigraph (not something like "??x"), return
1419 // it.
1420 if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1421 Ptr += 3;
1422 Size += 3;
1423 if (C == '\\') goto Slash;
1424 return C;
1425 }
1426 }
1427
1428 // If this is neither, return a single character.
1429 ++Size;
1430 return *Ptr;
1431}
1432
1433//===----------------------------------------------------------------------===//
1434// Helper methods for lexing.
1435//===----------------------------------------------------------------------===//
1436
1437/// Routine that indiscriminately sets the offset into the source file.
1438void Lexer::SetByteOffset(unsigned Offset, bool StartOfLine) {
1439 BufferPtr = BufferStart + Offset;
1440 if (BufferPtr > BufferEnd)
1441 BufferPtr = BufferEnd;
1442 // FIXME: What exactly does the StartOfLine bit mean? There are two
1443 // possible meanings for the "start" of the line: the first token on the
1444 // unexpanded line, or the first token on the expanded line.
1445 IsAtStartOfLine = StartOfLine;
1446 IsAtPhysicalStartOfLine = StartOfLine;
1447}
1448
1449static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1450 if (LangOpts.AsmPreprocessor) {
1451 return false;
1452 } else if (LangOpts.DollarIdents && '$' == C) {
1453 return true;
1454 } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1455 static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1456 C11AllowedIDCharRanges);
1457 return C11AllowedIDChars.contains(C);
1458 } else if (LangOpts.CPlusPlus) {
1459 static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1460 CXX03AllowedIDCharRanges);
1461 return CXX03AllowedIDChars.contains(C);
1462 } else {
1463 static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1464 C99AllowedIDCharRanges);
1465 return C99AllowedIDChars.contains(C);
1466 }
1467}
1468
1469static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1470 assert(isAllowedIDChar(C, LangOpts))((void)0);
1471 if (LangOpts.AsmPreprocessor) {
1472 return false;
1473 } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1474 static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1475 C11DisallowedInitialIDCharRanges);
1476 return !C11DisallowedInitialIDChars.contains(C);
1477 } else if (LangOpts.CPlusPlus) {
1478 return true;
1479 } else {
1480 static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1481 C99DisallowedInitialIDCharRanges);
1482 return !C99DisallowedInitialIDChars.contains(C);
1483 }
1484}
1485
1486static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1487 const char *End) {
1488 return CharSourceRange::getCharRange(L.getSourceLocation(Begin),
1489 L.getSourceLocation(End));
1490}
1491
1492static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1493 CharSourceRange Range, bool IsFirst) {
1494 // Check C99 compatibility.
1495 if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1496 enum {
1497 CannotAppearInIdentifier = 0,
1498 CannotStartIdentifier
1499 };
1500
1501 static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1502 C99AllowedIDCharRanges);
1503 static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1504 C99DisallowedInitialIDCharRanges);
1505 if (!C99AllowedIDChars.contains(C)) {
1506 Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1507 << Range
1508 << CannotAppearInIdentifier;
1509 } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1510 Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1511 << Range
1512 << CannotStartIdentifier;
1513 }
1514 }
1515
1516 // Check C++98 compatibility.
1517 if (!Diags.isIgnored(diag::warn_cxx98_compat_unicode_id, Range.getBegin())) {
1518 static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1519 CXX03AllowedIDCharRanges);
1520 if (!CXX03AllowedIDChars.contains(C)) {
1521 Diags.Report(Range.getBegin(), diag::warn_cxx98_compat_unicode_id)
1522 << Range;
1523 }
1524 }
1525}
1526
1527/// After encountering UTF-8 character C and interpreting it as an identifier
1528/// character, check whether it's a homoglyph for a common non-identifier
1529/// source character that is unlikely to be an intentional identifier
1530/// character and warn if so.
1531static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C,
1532 CharSourceRange Range) {
1533 // FIXME: Handle Unicode quotation marks (smart quotes, fullwidth quotes).
1534 struct HomoglyphPair {
1535 uint32_t Character;
1536 char LooksLike;
1537 bool operator<(HomoglyphPair R) const { return Character < R.Character; }
1538 };
1539 static constexpr HomoglyphPair SortedHomoglyphs[] = {
1540 {U'\u00ad', 0}, // SOFT HYPHEN
1541 {U'\u01c3', '!'}, // LATIN LETTER RETROFLEX CLICK
1542 {U'\u037e', ';'}, // GREEK QUESTION MARK
1543 {U'\u200b', 0}, // ZERO WIDTH SPACE
1544 {U'\u200c', 0}, // ZERO WIDTH NON-JOINER
1545 {U'\u200d', 0}, // ZERO WIDTH JOINER
1546 {U'\u2060', 0}, // WORD JOINER
1547 {U'\u2061', 0}, // FUNCTION APPLICATION
1548 {U'\u2062', 0}, // INVISIBLE TIMES
1549 {U'\u2063', 0}, // INVISIBLE SEPARATOR
1550 {U'\u2064', 0}, // INVISIBLE PLUS
1551 {U'\u2212', '-'}, // MINUS SIGN
1552 {U'\u2215', '/'}, // DIVISION SLASH
1553 {U'\u2216', '\\'}, // SET MINUS
1554 {U'\u2217', '*'}, // ASTERISK OPERATOR
1555 {U'\u2223', '|'}, // DIVIDES
1556 {U'\u2227', '^'}, // LOGICAL AND
1557 {U'\u2236', ':'}, // RATIO
1558 {U'\u223c', '~'}, // TILDE OPERATOR
1559 {U'\ua789', ':'}, // MODIFIER LETTER COLON
1560 {U'\ufeff', 0}, // ZERO WIDTH NO-BREAK SPACE
1561 {U'\uff01', '!'}, // FULLWIDTH EXCLAMATION MARK
1562 {U'\uff03', '#'}, // FULLWIDTH NUMBER SIGN
1563 {U'\uff04', '$'}, // FULLWIDTH DOLLAR SIGN
1564 {U'\uff05', '%'}, // FULLWIDTH PERCENT SIGN
1565 {U'\uff06', '&'}, // FULLWIDTH AMPERSAND
1566 {U'\uff08', '('}, // FULLWIDTH LEFT PARENTHESIS
1567 {U'\uff09', ')'}, // FULLWIDTH RIGHT PARENTHESIS
1568 {U'\uff0a', '*'}, // FULLWIDTH ASTERISK
1569 {U'\uff0b', '+'}, // FULLWIDTH ASTERISK
1570 {U'\uff0c', ','}, // FULLWIDTH COMMA
1571 {U'\uff0d', '-'}, // FULLWIDTH HYPHEN-MINUS
1572 {U'\uff0e', '.'}, // FULLWIDTH FULL STOP
1573 {U'\uff0f', '/'}, // FULLWIDTH SOLIDUS
1574 {U'\uff1a', ':'}, // FULLWIDTH COLON
1575 {U'\uff1b', ';'}, // FULLWIDTH SEMICOLON
1576 {U'\uff1c', '<'}, // FULLWIDTH LESS-THAN SIGN
1577 {U'\uff1d', '='}, // FULLWIDTH EQUALS SIGN
1578 {U'\uff1e', '>'}, // FULLWIDTH GREATER-THAN SIGN
1579 {U'\uff1f', '?'}, // FULLWIDTH QUESTION MARK
1580 {U'\uff20', '@'}, // FULLWIDTH COMMERCIAL AT
1581 {U'\uff3b', '['}, // FULLWIDTH LEFT SQUARE BRACKET
1582 {U'\uff3c', '\\'}, // FULLWIDTH REVERSE SOLIDUS
1583 {U'\uff3d', ']'}, // FULLWIDTH RIGHT SQUARE BRACKET
1584 {U'\uff3e', '^'}, // FULLWIDTH CIRCUMFLEX ACCENT
1585 {U'\uff5b', '{'}, // FULLWIDTH LEFT CURLY BRACKET
1586 {U'\uff5c', '|'}, // FULLWIDTH VERTICAL LINE
1587 {U'\uff5d', '}'}, // FULLWIDTH RIGHT CURLY BRACKET
1588 {U'\uff5e', '~'}, // FULLWIDTH TILDE
1589 {0, 0}
1590 };
1591 auto Homoglyph =
1592 std::lower_bound(std::begin(SortedHomoglyphs),
1593 std::end(SortedHomoglyphs) - 1, HomoglyphPair{C, '\0'});
1594 if (Homoglyph->Character == C) {
1595 llvm::SmallString<5> CharBuf;
1596 {
1597 llvm::raw_svector_ostream CharOS(CharBuf);
1598 llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
1599 }
1600 if (Homoglyph->LooksLike) {
1601 const char LooksLikeStr[] = {Homoglyph->LooksLike, 0};
1602 Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_homoglyph)
1603 << Range << CharBuf << LooksLikeStr;
1604 } else {
1605 Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_zero_width)
1606 << Range << CharBuf;
1607 }
1608 }
1609}
1610
1611bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1612 Token &Result) {
1613 const char *UCNPtr = CurPtr + Size;
1614 uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1615 if (CodePoint == 0 || !isAllowedIDChar(CodePoint, LangOpts))
1616 return false;
1617
1618 if (!isLexingRawMode())
1619 maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1620 makeCharRange(*this, CurPtr, UCNPtr),
1621 /*IsFirst=*/false);
1622
1623 Result.setFlag(Token::HasUCN);
1624 if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1625 (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1626 CurPtr = UCNPtr;
1627 else
1628 while (CurPtr != UCNPtr)
1629 (void)getAndAdvanceChar(CurPtr, Result);
1630 return true;
1631}
1632
1633bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1634 const char *UnicodePtr = CurPtr;
1635 llvm::UTF32 CodePoint;
1636 llvm::ConversionResult Result =
1637 llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
1638 (const llvm::UTF8 *)BufferEnd,
1639 &CodePoint,
1640 llvm::strictConversion);
1641 if (Result != llvm::conversionOK ||
1642 !isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts))
1643 return false;
1644
1645 if (!isLexingRawMode()) {
1646 maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1647 makeCharRange(*this, CurPtr, UnicodePtr),
1648 /*IsFirst=*/false);
1649 maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
1650 makeCharRange(*this, CurPtr, UnicodePtr));
1651 }
1652
1653 CurPtr = UnicodePtr;
1654 return true;
1655}
1656
1657bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1658 // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1659 unsigned Size;
1660 unsigned char C = *CurPtr++;
1661 while (isIdentifierBody(C))
1662 C = *CurPtr++;
1663
1664 --CurPtr; // Back up over the skipped character.
1665
1666 // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1667 // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1668 //
1669 // TODO: Could merge these checks into an InfoTable flag to make the
1670 // comparison cheaper
1671 if (isASCII(C) && C != '\\' && C != '?' &&
1672 (C != '$' || !LangOpts.DollarIdents)) {
1673FinishIdentifier:
1674 const char *IdStart = BufferPtr;
1675 FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1676 Result.setRawIdentifierData(IdStart);
1677
1678 // If we are in raw mode, return this identifier raw. There is no need to
1679 // look up identifier information or attempt to macro expand it.
1680 if (LexingRawMode)
1681 return true;
1682
1683 // Fill in Result.IdentifierInfo and update the token kind,
1684 // looking up the identifier in the identifier table.
1685 IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1686 // Note that we have to call PP->LookUpIdentifierInfo() even for code
1687 // completion, it writes IdentifierInfo into Result, and callers rely on it.
1688
1689 // If the completion point is at the end of an identifier, we want to treat
1690 // the identifier as incomplete even if it resolves to a macro or a keyword.
1691 // This allows e.g. 'class^' to complete to 'classifier'.
1692 if (isCodeCompletionPoint(CurPtr)) {
1693 // Return the code-completion token.
1694 Result.setKind(tok::code_completion);
1695 // Skip the code-completion char and all immediate identifier characters.
1696 // This ensures we get consistent behavior when completing at any point in
1697 // an identifier (i.e. at the start, in the middle, at the end). Note that
1698 // only simple cases (i.e. [a-zA-Z0-9_]) are supported to keep the code
1699 // simpler.
1700 assert(*CurPtr == 0 && "Completion character must be 0")((void)0);
1701 ++CurPtr;
1702 // Note that code completion token is not added as a separate character
1703 // when the completion point is at the end of the buffer. Therefore, we need
1704 // to check if the buffer has ended.
1705 if (CurPtr < BufferEnd) {
1706 while (isIdentifierBody(*CurPtr))
1707 ++CurPtr;
1708 }
1709 BufferPtr = CurPtr;
1710 return true;
1711 }
1712
1713 // Finally, now that we know we have an identifier, pass this off to the
1714 // preprocessor, which may macro expand it or something.
1715 if (II->isHandleIdentifierCase())
1716 return PP->HandleIdentifier(Result);
1717
1718 return true;
1719 }
1720
1721 // Otherwise, $,\,? in identifier found. Enter slower path.
1722
1723 C = getCharAndSize(CurPtr, Size);
1724 while (true) {
1725 if (C == '$') {
1726 // If we hit a $ and they are not supported in identifiers, we are done.
1727 if (!LangOpts.DollarIdents) goto FinishIdentifier;
1728
1729 // Otherwise, emit a diagnostic and continue.
1730 if (!isLexingRawMode())
1731 Diag(CurPtr, diag::ext_dollar_in_identifier);
1732 CurPtr = ConsumeChar(CurPtr, Size, Result);
1733 C = getCharAndSize(CurPtr, Size);
1734 continue;
1735 } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1736 C = getCharAndSize(CurPtr, Size);
1737 continue;
1738 } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1739 C = getCharAndSize(CurPtr, Size);
1740 continue;
1741 } else if (!isIdentifierBody(C)) {
1742 goto FinishIdentifier;
1743 }
1744
1745 // Otherwise, this character is good, consume it.
1746 CurPtr = ConsumeChar(CurPtr, Size, Result);
1747
1748 C = getCharAndSize(CurPtr, Size);
1749 while (isIdentifierBody(C)) {
1750 CurPtr = ConsumeChar(CurPtr, Size, Result);
1751 C = getCharAndSize(CurPtr, Size);
1752 }
1753 }
1754}
1755
1756/// isHexaLiteral - Return true if Start points to a hex constant.
1757/// in microsoft mode (where this is supposed to be several different tokens).
1758bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1759 unsigned Size;
1760 char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1761 if (C1 != '0')
1762 return false;
1763 char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1764 return (C2 == 'x' || C2 == 'X');
1765}
1766
1767/// LexNumericConstant - Lex the remainder of a integer or floating point
1768/// constant. From[-1] is the first character lexed. Return the end of the
1769/// constant.
1770bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1771 unsigned Size;
1772 char C = getCharAndSize(CurPtr, Size);
1773 char PrevCh = 0;
1774 while (isPreprocessingNumberBody(C)) {
1775 CurPtr = ConsumeChar(CurPtr, Size, Result);
1776 PrevCh = C;
1777 C = getCharAndSize(CurPtr, Size);
1778 }
1779
1780 // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1781 if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1782 // If we are in Microsoft mode, don't continue if the constant is hex.
1783 // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1784 if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1785 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1786 }
1787
1788 // If we have a hex FP constant, continue.
1789 if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1790 // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1791 // not-quite-conforming extension. Only do so if this looks like it's
1792 // actually meant to be a hexfloat, and not if it has a ud-suffix.
1793 bool IsHexFloat = true;
1794 if (!LangOpts.C99) {
1795 if (!isHexaLiteral(BufferPtr, LangOpts))
1796 IsHexFloat = false;
1797 else if (!getLangOpts().CPlusPlus17 &&
1798 std::find(BufferPtr, CurPtr, '_') != CurPtr)
1799 IsHexFloat = false;
1800 }
1801 if (IsHexFloat)
1802 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1803 }
1804
1805 // If we have a digit separator, continue.
1806 if (C == '\'' && (getLangOpts().CPlusPlus14 || getLangOpts().C2x)) {
1807 unsigned NextSize;
1808 char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1809 if (isIdentifierBody(Next)) {
1810 if (!isLexingRawMode())
1811 Diag(CurPtr, getLangOpts().CPlusPlus
1812 ? diag::warn_cxx11_compat_digit_separator
1813 : diag::warn_c2x_compat_digit_separator);
1814 CurPtr = ConsumeChar(CurPtr, Size, Result);
1815 CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1816 return LexNumericConstant(Result, CurPtr);
1817 }
1818 }
1819
1820 // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1821 if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1822 return LexNumericConstant(Result, CurPtr);
1823 if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1824 return LexNumericConstant(Result, CurPtr);
1825
1826 // Update the location of token as well as BufferPtr.
1827 const char *TokStart = BufferPtr;
1828 FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1829 Result.setLiteralData(TokStart);
1830 return true;
1831}
1832
1833/// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1834/// in C++11, or warn on a ud-suffix in C++98.
1835const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1836 bool IsStringLiteral) {
1837 assert(getLangOpts().CPlusPlus)((void)0);
1838
1839 // Maximally munch an identifier.
1840 unsigned Size;
1841 char C = getCharAndSize(CurPtr, Size);
1842 bool Consumed = false;
1843
1844 if (!isIdentifierHead(C)) {
1845 if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1846 Consumed = true;
1847 else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1848 Consumed = true;
1849 else
1850 return CurPtr;
1851 }
1852
1853 if (!getLangOpts().CPlusPlus11) {
1854 if (!isLexingRawMode())
1855 Diag(CurPtr,
1856 C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1857 : diag::warn_cxx11_compat_reserved_user_defined_literal)
1858 << FixItHint::CreateInsertion(getSourceLocation(CurPtr), " ");
1859 return CurPtr;
1860 }
1861
1862 // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1863 // that does not start with an underscore is ill-formed. As a conforming
1864 // extension, we treat all such suffixes as if they had whitespace before
1865 // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1866 // likely to be a ud-suffix than a macro, however, and accept that.
1867 if (!Consumed) {
1868 bool IsUDSuffix = false;
1869 if (C == '_')
1870 IsUDSuffix = true;
1871 else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1872 // In C++1y, we need to look ahead a few characters to see if this is a
1873 // valid suffix for a string literal or a numeric literal (this could be
1874 // the 'operator""if' defining a numeric literal operator).
1875 const unsigned MaxStandardSuffixLength = 3;
1876 char Buffer[MaxStandardSuffixLength] = { C };
1877 unsigned Consumed = Size;
1878 unsigned Chars = 1;
1879 while (true) {
1880 unsigned NextSize;
1881 char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1882 getLangOpts());
1883 if (!isIdentifierBody(Next)) {
1884 // End of suffix. Check whether this is on the allowed list.
1885 const StringRef CompleteSuffix(Buffer, Chars);
1886 IsUDSuffix = StringLiteralParser::isValidUDSuffix(getLangOpts(),
1887 CompleteSuffix);
1888 break;
1889 }
1890
1891 if (Chars == MaxStandardSuffixLength)
1892 // Too long: can't be a standard suffix.
1893 break;
1894
1895 Buffer[Chars++] = Next;
1896 Consumed += NextSize;
1897 }
1898 }
1899
1900 if (!IsUDSuffix) {
1901 if (!isLexingRawMode())
1902 Diag(CurPtr, getLangOpts().MSVCCompat
1903 ? diag::ext_ms_reserved_user_defined_literal
1904 : diag::ext_reserved_user_defined_literal)
1905 << FixItHint::CreateInsertion(getSourceLocation(CurPtr), " ");
1906 return CurPtr;
1907 }
1908
1909 CurPtr = ConsumeChar(CurPtr, Size, Result);
1910 }
1911
1912 Result.setFlag(Token::HasUDSuffix);
1913 while (true) {
1914 C = getCharAndSize(CurPtr, Size);
1915 if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1916 else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1917 else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1918 else break;
1919 }
1920
1921 return CurPtr;
1922}
1923
1924/// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1925/// either " or L" or u8" or u" or U".
1926bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1927 tok::TokenKind Kind) {
1928 const char *AfterQuote = CurPtr;
1929 // Does this string contain the \0 character?
1930 const char *NulCharacter = nullptr;
1931
1932 if (!isLexingRawMode() &&
1933 (Kind == tok::utf8_string_literal ||
1934 Kind == tok::utf16_string_literal ||
1935 Kind == tok::utf32_string_literal))
1936 Diag(BufferPtr, getLangOpts().CPlusPlus
1937 ? diag::warn_cxx98_compat_unicode_literal
1938 : diag::warn_c99_compat_unicode_literal);
1939
1940 char C = getAndAdvanceChar(CurPtr, Result);
1941 while (C != '"') {
1942 // Skip escaped characters. Escaped newlines will already be processed by
1943 // getAndAdvanceChar.
1944 if (C == '\\')
1945 C = getAndAdvanceChar(CurPtr, Result);
1946
1947 if (C == '\n' || C == '\r' || // Newline.
1948 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1949 if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1950 Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
1951 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1952 return true;
1953 }
1954
1955 if (C == 0) {
1956 if (isCodeCompletionPoint(CurPtr-1)) {
1957 if (ParsingFilename)
1958 codeCompleteIncludedFile(AfterQuote, CurPtr - 1, /*IsAngled=*/false);
1959 else
1960 PP->CodeCompleteNaturalLanguage();
1961 FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
1962 cutOffLexing();
1963 return true;
1964 }
1965
1966 NulCharacter = CurPtr-1;
1967 }
1968 C = getAndAdvanceChar(CurPtr, Result);
1969 }
1970
1971 // If we are in C++11, lex the optional ud-suffix.
1972 if (getLangOpts().CPlusPlus)
1973 CurPtr = LexUDSuffix(Result, CurPtr, true);
1974
1975 // If a nul character existed in the string, warn about it.
1976 if (NulCharacter && !isLexingRawMode())
1977 Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1978
1979 // Update the location of the token as well as the BufferPtr instance var.
1980 const char *TokStart = BufferPtr;
1981 FormTokenWithChars(Result, CurPtr, Kind);
1982 Result.setLiteralData(TokStart);
1983 return true;
1984}
1985
1986/// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1987/// having lexed R", LR", u8R", uR", or UR".
1988bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1989 tok::TokenKind Kind) {
1990 // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1991 // Between the initial and final double quote characters of the raw string,
1992 // any transformations performed in phases 1 and 2 (trigraphs,
1993 // universal-character-names, and line splicing) are reverted.
1994
1995 if (!isLexingRawMode())
1996 Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1997
1998 unsigned PrefixLen = 0;
1999
2000 while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
2001 ++PrefixLen;
2002
2003 // If the last character was not a '(', then we didn't lex a valid delimiter.
2004 if (CurPtr[PrefixLen] != '(') {
2005 if (!isLexingRawMode()) {
2006 const char *PrefixEnd = &CurPtr[PrefixLen];
2007 if (PrefixLen == 16) {
2008 Diag(PrefixEnd, diag::err_raw_delim_too_long);
2009 } else {
2010 Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
2011 << StringRef(PrefixEnd, 1);
2012 }
2013 }
2014
2015 // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
2016 // it's possible the '"' was intended to be part of the raw string, but
2017 // there's not much we can do about that.
2018 while (true) {
2019 char C = *CurPtr++;
2020
2021 if (C == '"')
2022 break;
2023 if (C == 0 && CurPtr-1 == BufferEnd) {
2024 --CurPtr;
2025 break;
2026 }
2027 }
2028
2029 FormTokenWithChars(Result, CurPtr, tok::unknown);
2030 return true;
2031 }
2032
2033 // Save prefix and move CurPtr past it
2034 const char *Prefix = CurPtr;
2035 CurPtr += PrefixLen + 1; // skip over prefix and '('
2036
2037 while (true) {
2038 char C = *CurPtr++;
2039
2040 if (C == ')') {
2041 // Check for prefix match and closing quote.
2042 if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
2043 CurPtr += PrefixLen + 1; // skip over prefix and '"'
2044 break;
2045 }
2046 } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
2047 if (!isLexingRawMode())
2048 Diag(BufferPtr, diag::err_unterminated_raw_string)
2049 << StringRef(Prefix, PrefixLen);
2050 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2051 return true;
2052 }
2053 }
2054
2055 // If we are in C++11, lex the optional ud-suffix.
2056 if (getLangOpts().CPlusPlus)
2057 CurPtr = LexUDSuffix(Result, CurPtr, true);
2058
2059 // Update the location of token as well as BufferPtr.
2060 const char *TokStart = BufferPtr;
2061 FormTokenWithChars(Result, CurPtr, Kind);
2062 Result.setLiteralData(TokStart);
2063 return true;
2064}
2065
2066/// LexAngledStringLiteral - Lex the remainder of an angled string literal,
2067/// after having lexed the '<' character. This is used for #include filenames.
2068bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
2069 // Does this string contain the \0 character?
2070 const char *NulCharacter = nullptr;
2071 const char *AfterLessPos = CurPtr;
2072 char C = getAndAdvanceChar(CurPtr, Result);
2073 while (C != '>') {
2074 // Skip escaped characters. Escaped newlines will already be processed by
2075 // getAndAdvanceChar.
2076 if (C == '\\')
2077 C = getAndAdvanceChar(CurPtr, Result);
2078
2079 if (isVerticalWhitespace(C) || // Newline.
2080 (C == 0 && (CurPtr - 1 == BufferEnd))) { // End of file.
2081 // If the filename is unterminated, then it must just be a lone <
2082 // character. Return this as such.
2083 FormTokenWithChars(Result, AfterLessPos, tok::less);
2084 return true;
2085 }
2086
2087 if (C == 0) {
2088 if (isCodeCompletionPoint(CurPtr - 1)) {
2089 codeCompleteIncludedFile(AfterLessPos, CurPtr - 1, /*IsAngled=*/true);
2090 cutOffLexing();
2091 FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
2092 return true;
2093 }
2094 NulCharacter = CurPtr-1;
2095 }
2096 C = getAndAdvanceChar(CurPtr, Result);
2097 }
2098
2099 // If a nul character existed in the string, warn about it.
2100 if (NulCharacter && !isLexingRawMode())
2101 Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2102
2103 // Update the location of token as well as BufferPtr.
2104 const char *TokStart = BufferPtr;
2105 FormTokenWithChars(Result, CurPtr, tok::header_name);
2106 Result.setLiteralData(TokStart);
2107 return true;
2108}
2109
2110void Lexer::codeCompleteIncludedFile(const char *PathStart,
2111 const char *CompletionPoint,
2112 bool IsAngled) {
2113 // Completion only applies to the filename, after the last slash.
2114 StringRef PartialPath(PathStart, CompletionPoint - PathStart);
2115 llvm::StringRef SlashChars = LangOpts.MSVCCompat ? "/\\" : "/";
2116 auto Slash = PartialPath.find_last_of(SlashChars);
2117 StringRef Dir =
2118 (Slash == StringRef::npos) ? "" : PartialPath.take_front(Slash);
2119 const char *StartOfFilename =
2120 (Slash == StringRef::npos) ? PathStart : PathStart + Slash + 1;
2121 // Code completion filter range is the filename only, up to completion point.
2122 PP->setCodeCompletionIdentifierInfo(&PP->getIdentifierTable().get(
2123 StringRef(StartOfFilename, CompletionPoint - StartOfFilename)));
2124 // We should replace the characters up to the closing quote or closest slash,
2125 // if any.
2126 while (CompletionPoint < BufferEnd) {
2127 char Next = *(CompletionPoint + 1);
2128 if (Next == 0 || Next == '\r' || Next == '\n')
2129 break;
2130 ++CompletionPoint;
2131 if (Next == (IsAngled ? '>' : '"'))
2132 break;
2133 if (llvm::is_contained(SlashChars, Next))
2134 break;
2135 }
2136
2137 PP->setCodeCompletionTokenRange(
2138 FileLoc.getLocWithOffset(StartOfFilename - BufferStart),
2139 FileLoc.getLocWithOffset(CompletionPoint - BufferStart));
2140 PP->CodeCompleteIncludedFile(Dir, IsAngled);
2141}
2142
2143/// LexCharConstant - Lex the remainder of a character constant, after having
2144/// lexed either ' or L' or u8' or u' or U'.
2145bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
2146 tok::TokenKind Kind) {
2147 // Does this character contain the \0 character?
2148 const char *NulCharacter = nullptr;
2149
2150 if (!isLexingRawMode()) {
2151 if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
2152 Diag(BufferPtr, getLangOpts().CPlusPlus
2153 ? diag::warn_cxx98_compat_unicode_literal
2154 : diag::warn_c99_compat_unicode_literal);
2155 else if (Kind == tok::utf8_char_constant)
2156 Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
2157 }
2158
2159 char C = getAndAdvanceChar(CurPtr, Result);
2160 if (C == '\'') {
2161 if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2162 Diag(BufferPtr, diag::ext_empty_character);
2163 FormTokenWithChars(Result, CurPtr, tok::unknown);
2164 return true;
2165 }
2166
2167 while (C != '\'') {
2168 // Skip escaped characters.
2169 if (C == '\\')
2170 C = getAndAdvanceChar(CurPtr, Result);
2171
2172 if (C == '\n' || C == '\r' || // Newline.
2173 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2174 if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2175 Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
2176 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2177 return true;
2178 }
2179
2180 if (C == 0) {
2181 if (isCodeCompletionPoint(CurPtr-1)) {
2182 PP->CodeCompleteNaturalLanguage();
2183 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2184 cutOffLexing();
2185 return true;
2186 }
2187
2188 NulCharacter = CurPtr-1;
2189 }
2190 C = getAndAdvanceChar(CurPtr, Result);
2191 }
2192
2193 // If we are in C++11, lex the optional ud-suffix.
2194 if (getLangOpts().CPlusPlus)
2195 CurPtr = LexUDSuffix(Result, CurPtr, false);
2196
2197 // If a nul character existed in the character, warn about it.
2198 if (NulCharacter && !isLexingRawMode())
2199 Diag(NulCharacter, diag::null_in_char_or_string) << 0;
2200
2201 // Update the location of token as well as BufferPtr.
2202 const char *TokStart = BufferPtr;
2203 FormTokenWithChars(Result, CurPtr, Kind);
2204 Result.setLiteralData(TokStart);
2205 return true;
2206}
2207
2208/// SkipWhitespace - Efficiently skip over a series of whitespace characters.
2209/// Update BufferPtr to point to the next non-whitespace character and return.
2210///
2211/// This method forms a token and returns true if KeepWhitespaceMode is enabled.
2212bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
2213 bool &TokAtPhysicalStartOfLine) {
2214 // Whitespace - Skip it, then return the token after the whitespace.
2215 bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
2216
2217 unsigned char Char = *CurPtr;
2218
2219 const char *lastNewLine = nullptr;
2220 auto setLastNewLine = [&](const char *Ptr) {
2221 lastNewLine = Ptr;
2222 if (!NewLinePtr)
2223 NewLinePtr = Ptr;
2224 };
2225 if (SawNewline)
2226 setLastNewLine(CurPtr - 1);
2227
2228 // Skip consecutive spaces efficiently.
2229 while (true) {
2230 // Skip horizontal whitespace very aggressively.
2231 while (isHorizontalWhitespace(Char))
2232 Char = *++CurPtr;
2233
2234 // Otherwise if we have something other than whitespace, we're done.
2235 if (!isVerticalWhitespace(Char))
2236 break;
2237
2238 if (ParsingPreprocessorDirective) {
2239 // End of preprocessor directive line, let LexTokenInternal handle this.
2240 BufferPtr = CurPtr;
2241 return false;
2242 }
2243
2244 // OK, but handle newline.
2245 if (*CurPtr == '\n')
2246 setLastNewLine(CurPtr);
2247 SawNewline = true;
2248 Char = *++CurPtr;
2249 }
2250
2251 // If the client wants us to return whitespace, return it now.
2252 if (isKeepWhitespaceMode()) {
2253 FormTokenWithChars(Result, CurPtr, tok::unknown);
2254 if (SawNewline) {
2255 IsAtStartOfLine = true;
2256 IsAtPhysicalStartOfLine = true;
2257 }
2258 // FIXME: The next token will not have LeadingSpace set.
2259 return true;
2260 }
2261
2262 // If this isn't immediately after a newline, there is leading space.
2263 char PrevChar = CurPtr[-1];
2264 bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2265
2266 Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2267 if (SawNewline) {
2268 Result.setFlag(Token::StartOfLine);
2269 TokAtPhysicalStartOfLine = true;
2270
2271 if (NewLinePtr && lastNewLine && NewLinePtr != lastNewLine && PP) {
2272 if (auto *Handler = PP->getEmptylineHandler())
2273 Handler->HandleEmptyline(SourceRange(getSourceLocation(NewLinePtr + 1),
2274 getSourceLocation(lastNewLine)));
2275 }
2276 }
2277
2278 BufferPtr = CurPtr;
2279 return false;
2280}
2281
2282/// We have just read the // characters from input. Skip until we find the
2283/// newline character that terminates the comment. Then update BufferPtr and
2284/// return.
2285///
2286/// If we're in KeepCommentMode or any CommentHandler has inserted
2287/// some tokens, this will store the first token and return true.
2288bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2289 bool &TokAtPhysicalStartOfLine) {
2290 // If Line comments aren't explicitly enabled for this language, emit an
2291 // extension warning.
2292 if (!LangOpts.LineComment && !isLexingRawMode()) {
2293 Diag(BufferPtr, diag::ext_line_comment);
2294
2295 // Mark them enabled so we only emit one warning for this translation
2296 // unit.
2297 LangOpts.LineComment = true;
2298 }
2299
2300 // Scan over the body of the comment. The common case, when scanning, is that
2301 // the comment contains normal ascii characters with nothing interesting in
2302 // them. As such, optimize for this case with the inner loop.
2303 //
2304 // This loop terminates with CurPtr pointing at the newline (or end of buffer)
2305 // character that ends the line comment.
2306 char C;
2307 while (true) {
2308 C = *CurPtr;
2309 // Skip over characters in the fast loop.
2310 while (C != 0 && // Potentially EOF.
2311 C != '\n' && C != '\r') // Newline or DOS-style newline.
2312 C = *++CurPtr;
2313
2314 const char *NextLine = CurPtr;
2315 if (C != 0) {
2316 // We found a newline, see if it's escaped.
2317 const char *EscapePtr = CurPtr-1;
2318 bool HasSpace = false;
2319 while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2320 --EscapePtr;
2321 HasSpace = true;
2322 }
2323
2324 if (*EscapePtr == '\\')
2325 // Escaped newline.
2326 CurPtr = EscapePtr;
2327 else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2328 EscapePtr[-2] == '?' && LangOpts.Trigraphs)
2329 // Trigraph-escaped newline.
2330 CurPtr = EscapePtr-2;
2331 else
2332 break; // This is a newline, we're done.
2333
2334 // If there was space between the backslash and newline, warn about it.
2335 if (HasSpace && !isLexingRawMode())
2336 Diag(EscapePtr, diag::backslash_newline_space);
2337 }
2338
2339 // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2340 // properly decode the character. Read it in raw mode to avoid emitting
2341 // diagnostics about things like trigraphs. If we see an escaped newline,
2342 // we'll handle it below.
2343 const char *OldPtr = CurPtr;
2344 bool OldRawMode = isLexingRawMode();
2345 LexingRawMode = true;
2346 C = getAndAdvanceChar(CurPtr, Result);
2347 LexingRawMode = OldRawMode;
2348
2349 // If we only read only one character, then no special handling is needed.
2350 // We're done and can skip forward to the newline.
2351 if (C != 0 && CurPtr == OldPtr+1) {
2352 CurPtr = NextLine;
2353 break;
2354 }
2355
2356 // If we read multiple characters, and one of those characters was a \r or
2357 // \n, then we had an escaped newline within the comment. Emit diagnostic
2358 // unless the next line is also a // comment.
2359 if (CurPtr != OldPtr + 1 && C != '/' &&
2360 (CurPtr == BufferEnd + 1 || CurPtr[0] != '/')) {
2361 for (; OldPtr != CurPtr; ++OldPtr)
2362 if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2363 // Okay, we found a // comment that ends in a newline, if the next
2364 // line is also a // comment, but has spaces, don't emit a diagnostic.
2365 if (isWhitespace(C)) {
2366 const char *ForwardPtr = CurPtr;
2367 while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2368 ++ForwardPtr;
2369 if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2370 break;
2371 }
2372
2373 if (!isLexingRawMode())
2374 Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2375 break;
2376 }
2377 }
2378
2379 if (C == '\r' || C == '\n' || CurPtr == BufferEnd + 1) {
2380 --CurPtr;
2381 break;
2382 }
2383
2384 if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2385 PP->CodeCompleteNaturalLanguage();
2386 cutOffLexing();
2387 return false;
2388 }
2389 }
2390
2391 // Found but did not consume the newline. Notify comment handlers about the
2392 // comment unless we're in a #if 0 block.
2393 if (PP && !isLexingRawMode() &&
2394 PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2395 getSourceLocation(CurPtr)))) {
2396 BufferPtr = CurPtr;
2397 return true; // A token has to be returned.
2398 }
2399
2400 // If we are returning comments as tokens, return this comment as a token.
2401 if (inKeepCommentMode())
2402 return SaveLineComment(Result, CurPtr);
2403
2404 // If we are inside a preprocessor directive and we see the end of line,
2405 // return immediately, so that the lexer can return this as an EOD token.
2406 if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2407 BufferPtr = CurPtr;
2408 return false;
2409 }
2410
2411 // Otherwise, eat the \n character. We don't care if this is a \n\r or
2412 // \r\n sequence. This is an efficiency hack (because we know the \n can't
2413 // contribute to another token), it isn't needed for correctness. Note that
2414 // this is ok even in KeepWhitespaceMode, because we would have returned the
2415 /// comment above in that mode.
2416 NewLinePtr = CurPtr++;
2417
2418 // The next returned token is at the start of the line.
2419 Result.setFlag(Token::StartOfLine);
2420 TokAtPhysicalStartOfLine = true;
2421 // No leading whitespace seen so far.
2422 Result.clearFlag(Token::LeadingSpace);
2423 BufferPtr = CurPtr;
2424 return false;
2425}
2426
2427/// If in save-comment mode, package up this Line comment in an appropriate
2428/// way and return it.
2429bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2430 // If we're not in a preprocessor directive, just return the // comment
2431 // directly.
2432 FormTokenWithChars(Result, CurPtr, tok::comment);
2433
2434 if (!ParsingPreprocessorDirective || LexingRawMode)
2435 return true;
2436
2437 // If this Line-style comment is in a macro definition, transmogrify it into
2438 // a C-style block comment.
2439 bool Invalid = false;
2440 std::string Spelling = PP->getSpelling(Result, &Invalid);
2441 if (Invalid)
2442 return true;
2443
2444 assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?")((void)0);
2445 Spelling[1] = '*'; // Change prefix to "/*".
2446 Spelling += "*/"; // add suffix.
2447
2448 Result.setKind(tok::comment);
2449 PP->CreateString(Spelling, Result,
2450 Result.getLocation(), Result.getLocation());
2451 return true;
2452}
2453
2454/// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2455/// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2456/// a diagnostic if so. We know that the newline is inside of a block comment.
2457static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2458 Lexer *L) {
2459 assert(CurPtr[0] == '\n' || CurPtr[0] == '\r')((void)0);
2460
2461 // Position of the first trigraph in the ending sequence.
2462 const char *TrigraphPos = 0;
2463 // Position of the first whitespace after a '\' in the ending sequence.
2464 const char *SpacePos = 0;
2465
2466 while (true) {
2467 // Back up off the newline.
2468 --CurPtr;
2469
2470 // If this is a two-character newline sequence, skip the other character.
2471 if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2472 // \n\n or \r\r -> not escaped newline.
2473 if (CurPtr[0] == CurPtr[1])
2474 return false;
2475 // \n\r or \r\n -> skip the newline.
2476 --CurPtr;
2477 }
2478
2479 // If we have horizontal whitespace, skip over it. We allow whitespace
2480 // between the slash and newline.
2481 while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2482 SpacePos = CurPtr;
2483 --CurPtr;
2484 }
2485
2486 // If we have a slash, this is an escaped newline.
2487 if (*CurPtr == '\\') {
2488 --CurPtr;
2489 } else if (CurPtr[0] == '/' && CurPtr[-1] == '?' && CurPtr[-2] == '?') {
2490 // This is a trigraph encoding of a slash.
2491 TrigraphPos = CurPtr - 2;
2492 CurPtr -= 3;
2493 } else {
2494 return false;
2495 }
2496
2497 // If the character preceding the escaped newline is a '*', then after line
2498 // splicing we have a '*/' ending the comment.
2499 if (*CurPtr == '*')
2500 break;
2501
2502 if (*CurPtr != '\n' && *CurPtr != '\r')
2503 return false;
2504 }
2505
2506 if (TrigraphPos) {
2507 // If no trigraphs are enabled, warn that we ignored this trigraph and
2508 // ignore this * character.
2509 if (!L->getLangOpts().Trigraphs) {
2510 if (!L->isLexingRawMode())
2511 L->Diag(TrigraphPos, diag::trigraph_ignored_block_comment);
2512 return false;
2513 }
2514 if (!L->isLexingRawMode())
2515 L->Diag(TrigraphPos, diag::trigraph_ends_block_comment);
2516 }
2517
2518 // Warn about having an escaped newline between the */ characters.
2519 if (!L->isLexingRawMode())
2520 L->Diag(CurPtr + 1, diag::escaped_newline_block_comment_end);
2521
2522 // If there was space between the backslash and newline, warn about it.
2523 if (SpacePos && !L->isLexingRawMode())
2524 L->Diag(SpacePos, diag::backslash_newline_space);
2525
2526 return true;
2527}
2528
2529#ifdef __SSE2__1
2530#include <emmintrin.h>
2531#elif __ALTIVEC__
2532#include <altivec.h>
2533#undef bool
2534#endif
2535
2536/// We have just read from input the / and * characters that started a comment.
2537/// Read until we find the * and / characters that terminate the comment.
2538/// Note that we don't bother decoding trigraphs or escaped newlines in block
2539/// comments, because they cannot cause the comment to end. The only thing
2540/// that can happen is the comment could end with an escaped newline between
2541/// the terminating * and /.
2542///
2543/// If we're in KeepCommentMode or any CommentHandler has inserted
2544/// some tokens, this will store the first token and return true.
2545bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2546 bool &TokAtPhysicalStartOfLine) {
2547 // Scan one character past where we should, looking for a '/' character. Once
2548 // we find it, check to see if it was preceded by a *. This common
2549 // optimization helps people who like to put a lot of * characters in their
2550 // comments.
2551
2552 // The first character we get with newlines and trigraphs skipped to handle
2553 // the degenerate /*/ case below correctly if the * has an escaped newline
2554 // after it.
2555 unsigned CharSize;
2556 unsigned char C = getCharAndSize(CurPtr, CharSize);
12
Calling 'Lexer::getCharAndSize'
15
Returning from 'Lexer::getCharAndSize'
2557 CurPtr += CharSize;
2558 if (C == 0 && CurPtr == BufferEnd+1) {
16
Assuming 'C' is not equal to 0
2559 if (!isLexingRawMode())
2560 Diag(BufferPtr, diag::err_unterminated_block_comment);
2561 --CurPtr;
2562
2563 // KeepWhitespaceMode should return this broken comment as a token. Since
2564 // it isn't a well formed comment, just return it as an 'unknown' token.
2565 if (isKeepWhitespaceMode()) {
2566 FormTokenWithChars(Result, CurPtr, tok::unknown);
2567 return true;
2568 }
2569
2570 BufferPtr = CurPtr;
2571 return false;
2572 }
2573
2574 // Check to see if the first character after the '/*' is another /. If so,
2575 // then this slash does not end the block comment, it is part of it.
2576 if (C == '/')
17
Assuming the condition is false
18
Taking false branch
2577 C = *CurPtr++;
2578
2579 while (true) {
19
Loop condition is true. Entering loop body
2580 // Skip over all non-interesting characters until we find end of buffer or a
2581 // (probably ending) '/' character.
2582 if (CurPtr + 24 < BufferEnd &&
20
Assuming the condition is true
22
Taking true branch
2583 // If there is a code-completion point avoid the fast scan because it
2584 // doesn't check for '\0'.
2585 !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
21
Assuming field 'PP' is null
2586 // While not aligned to a 16-byte boundary.
2587 while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
23
Loop condition is false. Execution continues on line 2590
2588 C = *CurPtr++;
2589
2590 if (C == '/') goto FoundSlash;
24
Taking false branch
2591
2592#ifdef __SSE2__1
2593 __m128i Slashes = _mm_set1_epi8('/');
2594 while (CurPtr+16 <= BufferEnd) {
25
Loop condition is true. Entering loop body
2595 int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2596 Slashes));
2597 if (cmp != 0) {
26
Assuming 'cmp' is not equal to 0
27
Taking true branch
2598 // Adjust the pointer to point directly after the first slash. It's
2599 // not necessary to set C here, it will be overwritten at the end of
2600 // the outer loop.
2601 CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2602 goto FoundSlash;
28
Control jumps to line 2635
2603 }
2604 CurPtr += 16;
2605 }
2606#elif __ALTIVEC__
2607 __vector unsigned char Slashes = {
2608 '/', '/', '/', '/', '/', '/', '/', '/',
2609 '/', '/', '/', '/', '/', '/', '/', '/'
2610 };
2611 while (CurPtr + 16 <= BufferEnd &&
2612 !vec_any_eq(*(const __vector unsigned char *)CurPtr, Slashes))
2613 CurPtr += 16;
2614#else
2615 // Scan for '/' quickly. Many block comments are very large.
2616 while (CurPtr[0] != '/' &&
2617 CurPtr[1] != '/' &&
2618 CurPtr[2] != '/' &&
2619 CurPtr[3] != '/' &&
2620 CurPtr+4 < BufferEnd) {
2621 CurPtr += 4;
2622 }
2623#endif
2624
2625 // It has to be one of the bytes scanned, increment to it and read one.
2626 C = *CurPtr++;
2627 }
2628
2629 // Loop to scan the remainder.
2630 while (C != '/' && C != '\0')
2631 C = *CurPtr++;
2632
2633 if (C == '/') {
2634 FoundSlash:
2635 if (CurPtr[-2] == '*') // We found the final */. We're done!
29
Assuming the condition is false
30
Taking false branch
2636 break;
2637
2638 if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
31
Assuming the condition is false
32
Assuming the condition is false
33
Taking false branch
2639 if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2640 // We found the final */, though it had an escaped newline between the
2641 // * and /. We're done!
2642 break;
2643 }
2644 }
2645 if (CurPtr[0] == '*' && CurPtr[1] != '/') {
34
Assuming the condition is true
35
Assuming the condition is true
36
Taking true branch
2646 // If this is a /* inside of the comment, emit a warning. Don't do this
2647 // if this is a /*/, which will end the comment. This misses cases with
2648 // embedded escaped newlines, but oh well.
2649 if (!isLexingRawMode())
37
Assuming the condition is true
38
Taking true branch
2650 Diag(CurPtr-1, diag::warn_nested_block_comment);
39
Calling 'Lexer::Diag'
2651 }
2652 } else if (C == 0 && CurPtr == BufferEnd+1) {
2653 if (!isLexingRawMode())
2654 Diag(BufferPtr, diag::err_unterminated_block_comment);
2655 // Note: the user probably forgot a */. We could continue immediately
2656 // after the /*, but this would involve lexing a lot of what really is the
2657 // comment, which surely would confuse the parser.
2658 --CurPtr;
2659
2660 // KeepWhitespaceMode should return this broken comment as a token. Since
2661 // it isn't a well formed comment, just return it as an 'unknown' token.
2662 if (isKeepWhitespaceMode()) {
2663 FormTokenWithChars(Result, CurPtr, tok::unknown);
2664 return true;
2665 }
2666
2667 BufferPtr = CurPtr;
2668 return false;
2669 } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2670 PP->CodeCompleteNaturalLanguage();
2671 cutOffLexing();
2672 return false;
2673 }
2674
2675 C = *CurPtr++;
2676 }
2677
2678 // Notify comment handlers about the comment unless we're in a #if 0 block.
2679 if (PP && !isLexingRawMode() &&
2680 PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2681 getSourceLocation(CurPtr)))) {
2682 BufferPtr = CurPtr;
2683 return true; // A token has to be returned.
2684 }
2685
2686 // If we are returning comments as tokens, return this comment as a token.
2687 if (inKeepCommentMode()) {
2688 FormTokenWithChars(Result, CurPtr, tok::comment);
2689 return true;
2690 }
2691
2692 // It is common for the tokens immediately after a /**/ comment to be
2693 // whitespace. Instead of going through the big switch, handle it
2694 // efficiently now. This is safe even in KeepWhitespaceMode because we would
2695 // have already returned above with the comment as a token.
2696 if (isHorizontalWhitespace(*CurPtr)) {
2697 SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2698 return false;
2699 }
2700
2701 // Otherwise, just return so that the next character will be lexed as a token.
2702 BufferPtr = CurPtr;
2703 Result.setFlag(Token::LeadingSpace);
2704 return false;
2705}
2706
2707//===----------------------------------------------------------------------===//
2708// Primary Lexing Entry Points
2709//===----------------------------------------------------------------------===//
2710
2711/// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2712/// uninterpreted string. This switches the lexer out of directive mode.
2713void Lexer::ReadToEndOfLine(SmallVectorImpl<char> *Result) {
2714 assert(ParsingPreprocessorDirective && ParsingFilename == false &&((void)0)
2715 "Must be in a preprocessing directive!")((void)0);
2716 Token Tmp;
2717 Tmp.startToken();
2718
2719 // CurPtr - Cache BufferPtr in an automatic variable.
2720 const char *CurPtr = BufferPtr;
2721 while (true) {
2722 char Char = getAndAdvanceChar(CurPtr, Tmp);
2723 switch (Char) {
2724 default:
2725 if (Result)
2726 Result->push_back(Char);
2727 break;
2728 case 0: // Null.
2729 // Found end of file?
2730 if (CurPtr-1 != BufferEnd) {
2731 if (isCodeCompletionPoint(CurPtr-1)) {
2732 PP->CodeCompleteNaturalLanguage();
2733 cutOffLexing();
2734 return;
2735 }
2736
2737 // Nope, normal character, continue.
2738 if (Result)
2739 Result->push_back(Char);
2740 break;
2741 }
2742 // FALL THROUGH.
2743 LLVM_FALLTHROUGH[[gnu::fallthrough]];
2744 case '\r':
2745 case '\n':
2746 // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2747 assert(CurPtr[-1] == Char && "Trigraphs for newline?")((void)0);
2748 BufferPtr = CurPtr-1;
2749
2750 // Next, lex the character, which should handle the EOD transition.
2751 Lex(Tmp);
2752 if (Tmp.is(tok::code_completion)) {
2753 if (PP)
2754 PP->CodeCompleteNaturalLanguage();
2755 Lex(Tmp);
2756 }
2757 assert(Tmp.is(tok::eod) && "Unexpected token!")((void)0);
2758
2759 // Finally, we're done;
2760 return;
2761 }
2762 }
2763}
2764
2765/// LexEndOfFile - CurPtr points to the end of this file. Handle this
2766/// condition, reporting diagnostics and handling other edge cases as required.
2767/// This returns true if Result contains a token, false if PP.Lex should be
2768/// called again.
2769bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2770 // If we hit the end of the file while parsing a preprocessor directive,
2771 // end the preprocessor directive first. The next token returned will
2772 // then be the end of file.
2773 if (ParsingPreprocessorDirective) {
2774 // Done parsing the "line".
2775 ParsingPreprocessorDirective = false;
2776 // Update the location of token as well as BufferPtr.
2777 FormTokenWithChars(Result, CurPtr, tok::eod);
2778
2779 // Restore comment saving mode, in case it was disabled for directive.
2780 if (PP)
2781 resetExtendedTokenMode();
2782 return true; // Have a token.
2783 }
2784
2785 // If we are in raw mode, return this event as an EOF token. Let the caller
2786 // that put us in raw mode handle the event.
2787 if (isLexingRawMode()) {
2788 Result.startToken();
2789 BufferPtr = BufferEnd;
2790 FormTokenWithChars(Result, BufferEnd, tok::eof);
2791 return true;
2792 }
2793
2794 if (PP->isRecordingPreamble() && PP->isInPrimaryFile()) {
2795 PP->setRecordedPreambleConditionalStack(ConditionalStack);
2796 // If the preamble cuts off the end of a header guard, consider it guarded.
2797 // The guard is valid for the preamble content itself, and for tools the
2798 // most useful answer is "yes, this file has a header guard".
2799 if (!ConditionalStack.empty())
2800 MIOpt.ExitTopLevelConditional();
2801 ConditionalStack.clear();
2802 }
2803
2804 // Issue diagnostics for unterminated #if and missing newline.
2805
2806 // If we are in a #if directive, emit an error.
2807 while (!ConditionalStack.empty()) {
2808 if (PP->getCodeCompletionFileLoc() != FileLoc)
2809 PP->Diag(ConditionalStack.back().IfLoc,
2810 diag::err_pp_unterminated_conditional);
2811 ConditionalStack.pop_back();
2812 }
2813
2814 SourceLocation EndLoc = getSourceLocation(BufferEnd);
2815 // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2816 // a pedwarn.
2817 if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2818 DiagnosticsEngine &Diags = PP->getDiagnostics();
2819 unsigned DiagID;
2820
2821 if (LangOpts.CPlusPlus11) {
2822 // C++11 [lex.phases] 2.2 p2
2823 // Prefer the C++98 pedantic compatibility warning over the generic,
2824 // non-extension, user-requested "missing newline at EOF" warning.
2825 if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2826 DiagID = diag::warn_cxx98_compat_no_newline_eof;
2827 } else {
2828 DiagID = diag::warn_no_newline_eof;
2829 }
2830 } else {
2831 DiagID = diag::ext_no_newline_eof;
2832 }
2833
2834 Diag(BufferEnd, DiagID)
2835 << FixItHint::CreateInsertion(EndLoc, "\n");
2836 }
2837
2838 BufferPtr = CurPtr;
2839
2840 // Finally, let the preprocessor handle this.
2841 return PP->HandleEndOfFile(Result, EndLoc, isPragmaLexer());
2842}
2843
2844/// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2845/// the specified lexer will return a tok::l_paren token, 0 if it is something
2846/// else and 2 if there are no more tokens in the buffer controlled by the
2847/// lexer.
2848unsigned Lexer::isNextPPTokenLParen() {
2849 assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?")((void)0);
2850
2851 // Switch to 'skipping' mode. This will ensure that we can lex a token
2852 // without emitting diagnostics, disables macro expansion, and will cause EOF
2853 // to return an EOF token instead of popping the include stack.
2854 LexingRawMode = true;
2855
2856 // Save state that can be changed while lexing so that we can restore it.
2857 const char *TmpBufferPtr = BufferPtr;
2858 bool inPPDirectiveMode = ParsingPreprocessorDirective;
2859 bool atStartOfLine = IsAtStartOfLine;
2860 bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2861 bool leadingSpace = HasLeadingSpace;
2862
2863 Token Tok;
2864 Lex(Tok);
2865
2866 // Restore state that may have changed.
2867 BufferPtr = TmpBufferPtr;
2868 ParsingPreprocessorDirective = inPPDirectiveMode;
2869 HasLeadingSpace = leadingSpace;
2870 IsAtStartOfLine = atStartOfLine;
2871 IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2872
2873 // Restore the lexer back to non-skipping mode.
2874 LexingRawMode = false;
2875
2876 if (Tok.is(tok::eof))
2877 return 2;
2878 return Tok.is(tok::l_paren);
2879}
2880
2881/// Find the end of a version control conflict marker.
2882static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2883 ConflictMarkerKind CMK) {
2884 const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2885 size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2886 auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2887 size_t Pos = RestOfBuffer.find(Terminator);
2888 while (Pos != StringRef::npos) {
2889 // Must occur at start of line.
2890 if (Pos == 0 ||
2891 (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2892 RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2893 Pos = RestOfBuffer.find(Terminator);
2894 continue;
2895 }
2896 return RestOfBuffer.data()+Pos;
2897 }
2898 return nullptr;
2899}
2900
2901/// IsStartOfConflictMarker - If the specified pointer is the start of a version
2902/// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2903/// and recover nicely. This returns true if it is a conflict marker and false
2904/// if not.
2905bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2906 // Only a conflict marker if it starts at the beginning of a line.
2907 if (CurPtr != BufferStart &&
2908 CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2909 return false;
2910
2911 // Check to see if we have <<<<<<< or >>>>.
2912 if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
2913 !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
2914 return false;
2915
2916 // If we have a situation where we don't care about conflict markers, ignore
2917 // it.
2918 if (CurrentConflictMarkerState || isLexingRawMode())
2919 return false;
2920
2921 ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2922
2923 // Check to see if there is an ending marker somewhere in the buffer at the
2924 // start of a line to terminate this conflict marker.
2925 if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2926 // We found a match. We are really in a conflict marker.
2927 // Diagnose this, and ignore to the end of line.
2928 Diag(CurPtr, diag::err_conflict_marker);
2929 CurrentConflictMarkerState = Kind;
2930
2931 // Skip ahead to the end of line. We know this exists because the
2932 // end-of-conflict marker starts with \r or \n.
2933 while (*CurPtr != '\r' && *CurPtr != '\n') {
2934 assert(CurPtr != BufferEnd && "Didn't find end of line")((void)0);
2935 ++CurPtr;
2936 }
2937 BufferPtr = CurPtr;
2938 return true;
2939 }
2940
2941 // No end of conflict marker found.
2942 return false;
2943}
2944
2945/// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2946/// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2947/// is the end of a conflict marker. Handle it by ignoring up until the end of
2948/// the line. This returns true if it is a conflict marker and false if not.
2949bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2950 // Only a conflict marker if it starts at the beginning of a line.
2951 if (CurPtr != BufferStart &&
2952 CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2953 return false;
2954
2955 // If we have a situation where we don't care about conflict markers, ignore
2956 // it.
2957 if (!CurrentConflictMarkerState || isLexingRawMode())
2958 return false;
2959
2960 // Check to see if we have the marker (4 characters in a row).
2961 for (unsigned i = 1; i != 4; ++i)
2962 if (CurPtr[i] != CurPtr[0])
2963 return false;
2964
2965 // If we do have it, search for the end of the conflict marker. This could
2966 // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2967 // be the end of conflict marker.
2968 if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2969 CurrentConflictMarkerState)) {
2970 CurPtr = End;
2971
2972 // Skip ahead to the end of line.
2973 while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2974 ++CurPtr;
2975
2976 BufferPtr = CurPtr;
2977
2978 // No longer in the conflict marker.
2979 CurrentConflictMarkerState = CMK_None;
2980 return true;
2981 }
2982
2983 return false;
2984}
2985
2986static const char *findPlaceholderEnd(const char *CurPtr,
2987 const char *BufferEnd) {
2988 if (CurPtr == BufferEnd)
2989 return nullptr;
2990 BufferEnd -= 1; // Scan until the second last character.
2991 for (; CurPtr != BufferEnd; ++CurPtr) {
2992 if (CurPtr[0] == '#' && CurPtr[1] == '>')
2993 return CurPtr + 2;
2994 }
2995 return nullptr;
2996}
2997
2998bool Lexer::lexEditorPlaceholder(Token &Result, const char *CurPtr) {
2999 assert(CurPtr[-1] == '<' && CurPtr[0] == '#' && "Not a placeholder!")((void)0);
3000 if (!PP || !PP->getPreprocessorOpts().LexEditorPlaceholders || LexingRawMode)
3001 return false;
3002 const char *End = findPlaceholderEnd(CurPtr + 1, BufferEnd);
3003 if (!End)
3004 return false;
3005 const char *Start = CurPtr - 1;
3006 if (!LangOpts.AllowEditorPlaceholders)
3007 Diag(Start, diag::err_placeholder_in_source);
3008 Result.startToken();
3009 FormTokenWithChars(Result, End, tok::raw_identifier);
3010 Result.setRawIdentifierData(Start);
3011 PP->LookUpIdentifierInfo(Result);
3012 Result.setFlag(Token::IsEditorPlaceholder);
3013 BufferPtr = End;
3014 return true;
3015}
3016
3017bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
3018 if (PP && PP->isCodeCompletionEnabled()) {
3019 SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
3020 return Loc == PP->getCodeCompletionLoc();
3021 }
3022
3023 return false;
3024}
3025
3026uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
3027 Token *Result) {
3028 unsigned CharSize;
3029 char Kind = getCharAndSize(StartPtr, CharSize);
3030
3031 unsigned NumHexDigits;
3032 if (Kind == 'u')
3033 NumHexDigits = 4;
3034 else if (Kind == 'U')
3035 NumHexDigits = 8;
3036 else
3037 return 0;
3038
3039 if (!LangOpts.CPlusPlus && !LangOpts.C99) {
3040 if (Result && !isLexingRawMode())
3041 Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
3042 return 0;
3043 }
3044
3045 const char *CurPtr = StartPtr + CharSize;
3046 const char *KindLoc = &CurPtr[-1];
3047
3048 uint32_t CodePoint = 0;
3049 for (unsigned i = 0; i < NumHexDigits; ++i) {
3050 char C = getCharAndSize(CurPtr, CharSize);
3051
3052 unsigned Value = llvm::hexDigitValue(C);
3053 if (Value == -1U) {
3054 if (Result && !isLexingRawMode()) {
3055 if (i == 0) {
3056 Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
3057 << StringRef(KindLoc, 1);
3058 } else {
3059 Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
3060
3061 // If the user wrote \U1234, suggest a fixit to \u.
3062 if (i == 4 && NumHexDigits == 8) {
3063 CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
3064 Diag(KindLoc, diag::note_ucn_four_not_eight)
3065 << FixItHint::CreateReplacement(URange, "u");
3066 }
3067 }
3068 }
3069
3070 return 0;
3071 }
3072
3073 CodePoint <<= 4;
3074 CodePoint += Value;
3075
3076 CurPtr += CharSize;
3077 }
3078
3079 if (Result) {
3080 Result->setFlag(Token::HasUCN);
3081 if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
3082 StartPtr = CurPtr;
3083 else
3084 while (StartPtr != CurPtr)
3085 (void)getAndAdvanceChar(StartPtr, *Result);
3086 } else {
3087 StartPtr = CurPtr;
3088 }
3089
3090 // Don't apply C family restrictions to UCNs in assembly mode
3091 if (LangOpts.AsmPreprocessor)
3092 return CodePoint;
3093
3094 // C99 6.4.3p2: A universal character name shall not specify a character whose
3095 // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
3096 // 0060 (`), nor one in the range D800 through DFFF inclusive.)
3097 // C++11 [lex.charset]p2: If the hexadecimal value for a
3098 // universal-character-name corresponds to a surrogate code point (in the
3099 // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
3100 // if the hexadecimal value for a universal-character-name outside the
3101 // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
3102 // string literal corresponds to a control character (in either of the
3103 // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
3104 // basic source character set, the program is ill-formed.
3105 if (CodePoint < 0xA0) {
3106 if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
3107 return CodePoint;
3108
3109 // We don't use isLexingRawMode() here because we need to warn about bad
3110 // UCNs even when skipping preprocessing tokens in a #if block.
3111 if (Result && PP) {
3112 if (CodePoint < 0x20 || CodePoint >= 0x7F)
3113 Diag(BufferPtr, diag::err_ucn_control_character);
3114 else {
3115 char C = static_cast<char>(CodePoint);
3116 Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
3117 }
3118 }
3119
3120 return 0;
3121 } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
3122 // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
3123 // We don't use isLexingRawMode() here because we need to diagnose bad
3124 // UCNs even when skipping preprocessing tokens in a #if block.
3125 if (Result && PP) {
3126 if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
3127 Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
3128 else
3129 Diag(BufferPtr, diag::err_ucn_escape_invalid);
3130 }
3131 return 0;
3132 }
3133
3134 return CodePoint;
3135}
3136
3137bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
3138 const char *CurPtr) {
3139 static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
3140 UnicodeWhitespaceCharRanges);
3141 if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
3142 UnicodeWhitespaceChars.contains(C)) {
3143 Diag(BufferPtr, diag::ext_unicode_whitespace)
3144 << makeCharRange(*this, BufferPtr, CurPtr);
3145
3146 Result.setFlag(Token::LeadingSpace);
3147 return true;
3148 }
3149 return false;
3150}
3151
3152bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
3153 if (isAllowedIDChar(C, LangOpts) && isAllowedInitiallyIDChar(C, LangOpts)) {
3154 if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
3155 !PP->isPreprocessedOutput()) {
3156 maybeDiagnoseIDCharCompat(PP->getDiagnostics(), C,
3157 makeCharRange(*this, BufferPtr, CurPtr),
3158 /*IsFirst=*/true);
3159 maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), C,
3160 makeCharRange(*this, BufferPtr, CurPtr));
3161 }
3162
3163 MIOpt.ReadToken();
3164 return LexIdentifier(Result, CurPtr);
3165 }
3166
3167 if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
3168 !PP->isPreprocessedOutput() &&
3169 !isASCII(*BufferPtr) && !isAllowedIDChar(C, LangOpts)) {
3170 // Non-ASCII characters tend to creep into source code unintentionally.
3171 // Instead of letting the parser complain about the unknown token,
3172 // just drop the character.
3173 // Note that we can /only/ do this when the non-ASCII character is actually
3174 // spelled as Unicode, not written as a UCN. The standard requires that
3175 // we not throw away any possible preprocessor tokens, but there's a
3176 // loophole in the mapping of Unicode characters to basic character set
3177 // characters that allows us to map these particular characters to, say,
3178 // whitespace.
3179 Diag(BufferPtr, diag::err_non_ascii)
3180 << FixItHint::CreateRemoval(makeCharRange(*this, BufferPtr, CurPtr));
3181
3182 BufferPtr = CurPtr;
3183 return false;
3184 }
3185
3186 // Otherwise, we have an explicit UCN or a character that's unlikely to show
3187 // up by accident.
3188 MIOpt.ReadToken();
3189 FormTokenWithChars(Result, CurPtr, tok::unknown);
3190 return true;
3191}
3192
3193void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
3194 IsAtStartOfLine = Result.isAtStartOfLine();
3195 HasLeadingSpace = Result.hasLeadingSpace();
3196 HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
3197 // Note that this doesn't affect IsAtPhysicalStartOfLine.
3198}
3199
3200bool Lexer::Lex(Token &Result) {
3201 // Start a new token.
3202 Result.startToken();
3203
3204 // Set up misc whitespace flags for LexTokenInternal.
3205 if (IsAtStartOfLine) {
3206 Result.setFlag(Token::StartOfLine);
3207 IsAtStartOfLine = false;
3208 }
3209
3210 if (HasLeadingSpace) {
3211 Result.setFlag(Token::LeadingSpace);
3212 HasLeadingSpace = false;
3213 }
3214
3215 if (HasLeadingEmptyMacro) {
3216 Result.setFlag(Token::LeadingEmptyMacro);
3217 HasLeadingEmptyMacro = false;
3218 }
3219
3220 bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
3221 IsAtPhysicalStartOfLine = false;
3222 bool isRawLex = isLexingRawMode();
3223 (void) isRawLex;
3224 bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
3225 // (After the LexTokenInternal call, the lexer might be destroyed.)
3226 assert((returnedToken || !isRawLex) && "Raw lex must succeed")((void)0);
3227 return returnedToken;
3228}
3229
3230/// LexTokenInternal - This implements a simple C family lexer. It is an
3231/// extremely performance critical piece of code. This assumes that the buffer
3232/// has a null character at the end of the file. This returns a preprocessing
3233/// token, not a normal token, as such, it is an internal interface. It assumes
3234/// that the Flags of result have been cleared before calling this.
3235bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
3236LexNextToken:
3237 // New token, can't need cleaning yet.
3238 Result.clearFlag(Token::NeedsCleaning);
3239 Result.setIdentifierInfo(nullptr);
3240
3241 // CurPtr - Cache BufferPtr in an automatic variable.
3242 const char *CurPtr = BufferPtr;
3243
3244 // Small amounts of horizontal whitespace is very common between tokens.
3245 if (isHorizontalWhitespace(*CurPtr)) {
1
Assuming the condition is false
2
Taking false branch
3246 do {
3247 ++CurPtr;
3248 } while (isHorizontalWhitespace(*CurPtr));
3249
3250 // If we are keeping whitespace and other tokens, just return what we just
3251 // skipped. The next lexer invocation will return the token after the
3252 // whitespace.
3253 if (isKeepWhitespaceMode()) {
3254 FormTokenWithChars(Result, CurPtr, tok::unknown);
3255 // FIXME: The next token will not have LeadingSpace set.
3256 return true;
3257 }
3258
3259 BufferPtr = CurPtr;
3260 Result.setFlag(Token::LeadingSpace);
3261 }
3262
3263 unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
3264
3265 // Read a character, advancing over it.
3266 char Char = getAndAdvanceChar(CurPtr, Result);
3267 tok::TokenKind Kind;
3268
3269 if (!isVerticalWhitespace(Char))
3
Assuming the condition is false
4
Taking false branch
3270 NewLinePtr = nullptr;
3271
3272 switch (Char) {
5
Control jumps to 'case 32:' at line 3341
3273 case 0: // Null.
3274 // Found end of file?
3275 if (CurPtr-1 == BufferEnd)
3276 return LexEndOfFile(Result, CurPtr-1);
3277
3278 // Check if we are performing code completion.
3279 if (isCodeCompletionPoint(CurPtr-1)) {
3280 // Return the code-completion token.
3281 Result.startToken();
3282 FormTokenWithChars(Result, CurPtr, tok::code_completion);
3283 return true;
3284 }
3285
3286 if (!isLexingRawMode())
3287 Diag(CurPtr-1, diag::null_in_file);
3288 Result.setFlag(Token::LeadingSpace);
3289 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3290 return true; // KeepWhitespaceMode
3291
3292 // We know the lexer hasn't changed, so just try again with this lexer.
3293 // (We manually eliminate the tail call to avoid recursion.)
3294 goto LexNextToken;
3295
3296 case 26: // DOS & CP/M EOF: "^Z".
3297 // If we're in Microsoft extensions mode, treat this as end of file.
3298 if (LangOpts.MicrosoftExt) {
3299 if (!isLexingRawMode())
3300 Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
3301 return LexEndOfFile(Result, CurPtr-1);
3302 }
3303
3304 // If Microsoft extensions are disabled, this is just random garbage.
3305 Kind = tok::unknown;
3306 break;
3307
3308 case '\r':
3309 if (CurPtr[0] == '\n')
3310 (void)getAndAdvanceChar(CurPtr, Result);
3311 LLVM_FALLTHROUGH[[gnu::fallthrough]];
3312 case '\n':
3313 // If we are inside a preprocessor directive and we see the end of line,
3314 // we know we are done with the directive, so return an EOD token.
3315 if (ParsingPreprocessorDirective) {
3316 // Done parsing the "line".
3317 ParsingPreprocessorDirective = false;
3318
3319 // Restore comment saving mode, in case it was disabled for directive.
3320 if (PP)
3321 resetExtendedTokenMode();
3322
3323 // Since we consumed a newline, we are back at the start of a line.
3324 IsAtStartOfLine = true;
3325 IsAtPhysicalStartOfLine = true;
3326 NewLinePtr = CurPtr - 1;
3327
3328 Kind = tok::eod;
3329 break;
3330 }
3331
3332 // No leading whitespace seen so far.
3333 Result.clearFlag(Token::LeadingSpace);
3334
3335 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3336 return true; // KeepWhitespaceMode
3337
3338 // We only saw whitespace, so just try again with this lexer.
3339 // (We manually eliminate the tail call to avoid recursion.)
3340 goto LexNextToken;
3341 case ' ':
3342 case '\t':
3343 case '\f':
3344 case '\v':
3345 SkipHorizontalWhitespace:
3346 Result.setFlag(Token::LeadingSpace);
3347 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
6
Taking false branch
3348 return true; // KeepWhitespaceMode
3349
3350 SkipIgnoredUnits:
3351 CurPtr = BufferPtr;
3352
3353 // If the next token is obviously a // or /* */ comment, skip it efficiently
3354 // too (without going through the big switch stmt).
3355 if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
7
Assuming the condition is false
3356 LangOpts.LineComment &&
3357 (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3358 if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3359 return true; // There is a token to return.
3360 goto SkipIgnoredUnits;
3361 } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
8
Assuming the condition is true
9
Assuming the condition is true
10
Taking true branch
3362 if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
11
Calling 'Lexer::SkipBlockComment'
3363 return true; // There is a token to return.
3364 goto SkipIgnoredUnits;
3365 } else if (isHorizontalWhitespace(*CurPtr)) {
3366 goto SkipHorizontalWhitespace;
3367 }
3368 // We only saw whitespace, so just try again with this lexer.
3369 // (We manually eliminate the tail call to avoid recursion.)
3370 goto LexNextToken;
3371
3372 // C99 6.4.4.1: Integer Constants.
3373 // C99 6.4.4.2: Floating Constants.
3374 case '0': case '1': case '2': case '3': case '4':
3375 case '5': case '6': case '7': case '8': case '9':
3376 // Notify MIOpt that we read a non-whitespace/non-comment token.
3377 MIOpt.ReadToken();
3378 return LexNumericConstant(Result, CurPtr);
3379
3380 case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3381 // Notify MIOpt that we read a non-whitespace/non-comment token.
3382 MIOpt.ReadToken();
3383
3384 if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3385 Char = getCharAndSize(CurPtr, SizeTmp);
3386
3387 // UTF-16 string literal
3388 if (Char == '"')
3389 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3390 tok::utf16_string_literal);
3391
3392 // UTF-16 character constant
3393 if (Char == '\'')
3394 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3395 tok::utf16_char_constant);
3396
3397 // UTF-16 raw string literal
3398 if (Char == 'R' && LangOpts.CPlusPlus11 &&
3399 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3400 return LexRawStringLiteral(Result,
3401 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3402 SizeTmp2, Result),
3403 tok::utf16_string_literal);
3404
3405 if (Char == '8') {
3406 char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3407
3408 // UTF-8 string literal
3409 if (Char2 == '"')
3410 return LexStringLiteral(Result,
3411 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3412 SizeTmp2, Result),
3413 tok::utf8_string_literal);
3414 if (Char2 == '\'' && LangOpts.CPlusPlus17)
3415 return LexCharConstant(
3416 Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3417 SizeTmp2, Result),
3418 tok::utf8_char_constant);
3419
3420 if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3421 unsigned SizeTmp3;
3422 char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3423 // UTF-8 raw string literal
3424 if (Char3 == '"') {
3425 return LexRawStringLiteral(Result,
3426 ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3427 SizeTmp2, Result),
3428 SizeTmp3, Result),
3429 tok::utf8_string_literal);
3430 }
3431 }
3432 }
3433 }
3434
3435 // treat u like the start of an identifier.
3436 return LexIdentifier(Result, CurPtr);
3437
3438 case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3439 // Notify MIOpt that we read a non-whitespace/non-comment token.
3440 MIOpt.ReadToken();
3441
3442 if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3443 Char = getCharAndSize(CurPtr, SizeTmp);
3444
3445 // UTF-32 string literal
3446 if (Char == '"')
3447 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3448 tok::utf32_string_literal);
3449
3450 // UTF-32 character constant
3451 if (Char == '\'')
3452 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3453 tok::utf32_char_constant);
3454
3455 // UTF-32 raw string literal
3456 if (Char == 'R' && LangOpts.CPlusPlus11 &&
3457 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3458 return LexRawStringLiteral(Result,
3459 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3460 SizeTmp2, Result),
3461 tok::utf32_string_literal);
3462 }
3463
3464 // treat U like the start of an identifier.
3465 return LexIdentifier(Result, CurPtr);
3466
3467 case 'R': // Identifier or C++0x raw string literal
3468 // Notify MIOpt that we read a non-whitespace/non-comment token.
3469 MIOpt.ReadToken();
3470
3471 if (LangOpts.CPlusPlus11) {
3472 Char = getCharAndSize(CurPtr, SizeTmp);
3473
3474 if (Char == '"')
3475 return LexRawStringLiteral(Result,
3476 ConsumeChar(CurPtr, SizeTmp, Result),
3477 tok::string_literal);
3478 }
3479
3480 // treat R like the start of an identifier.
3481 return LexIdentifier(Result, CurPtr);
3482
3483 case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3484 // Notify MIOpt that we read a non-whitespace/non-comment token.
3485 MIOpt.ReadToken();
3486 Char = getCharAndSize(CurPtr, SizeTmp);
3487
3488 // Wide string literal.
3489 if (Char == '"')
3490 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3491 tok::wide_string_literal);
3492
3493 // Wide raw string literal.
3494 if (LangOpts.CPlusPlus11 && Char == 'R' &&
3495 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3496 return LexRawStringLiteral(Result,
3497 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3498 SizeTmp2, Result),
3499 tok::wide_string_literal);
3500
3501 // Wide character constant.
3502 if (Char == '\'')
3503 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3504 tok::wide_char_constant);
3505 // FALL THROUGH, treating L like the start of an identifier.
3506 LLVM_FALLTHROUGH[[gnu::fallthrough]];
3507
3508 // C99 6.4.2: Identifiers.
3509 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3510 case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3511 case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3512 case 'V': case 'W': case 'X': case 'Y': case 'Z':
3513 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3514 case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3515 case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3516 case 'v': case 'w': case 'x': case 'y': case 'z':
3517 case '_':
3518 // Notify MIOpt that we read a non-whitespace/non-comment token.
3519 MIOpt.ReadToken();
3520 return LexIdentifier(Result, CurPtr);
3521
3522 case '$': // $ in identifiers.
3523 if (LangOpts.DollarIdents) {
3524 if (!isLexingRawMode())
3525 Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3526 // Notify MIOpt that we read a non-whitespace/non-comment token.
3527 MIOpt.ReadToken();
3528 return LexIdentifier(Result, CurPtr);
3529 }
3530
3531 Kind = tok::unknown;
3532 break;
3533
3534 // C99 6.4.4: Character Constants.
3535 case '\'':
3536 // Notify MIOpt that we read a non-whitespace/non-comment token.
3537 MIOpt.ReadToken();
3538 return LexCharConstant(Result, CurPtr, tok::char_constant);
3539
3540 // C99 6.4.5: String Literals.
3541 case '"':
3542 // Notify MIOpt that we read a non-whitespace/non-comment token.
3543 MIOpt.ReadToken();
3544 return LexStringLiteral(Result, CurPtr,
3545 ParsingFilename ? tok::header_name
3546 : tok::string_literal);
3547
3548 // C99 6.4.6: Punctuators.
3549 case '?':
3550 Kind = tok::question;
3551 break;
3552 case '[':
3553 Kind = tok::l_square;
3554 break;
3555 case ']':
3556 Kind = tok::r_square;
3557 break;
3558 case '(':
3559 Kind = tok::l_paren;
3560 break;
3561 case ')':
3562 Kind = tok::r_paren;
3563 break;
3564 case '{':
3565 Kind = tok::l_brace;
3566 break;
3567 case '}':
3568 Kind = tok::r_brace;
3569 break;
3570 case '.':
3571 Char = getCharAndSize(CurPtr, SizeTmp);
3572 if (Char >= '0' && Char <= '9') {
3573 // Notify MIOpt that we read a non-whitespace/non-comment token.
3574 MIOpt.ReadToken();
3575
3576 return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3577 } else if (LangOpts.CPlusPlus && Char == '*') {
3578 Kind = tok::periodstar;
3579 CurPtr += SizeTmp;
3580 } else if (Char == '.' &&
3581 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3582 Kind = tok::ellipsis;
3583 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3584 SizeTmp2, Result);
3585 } else {
3586 Kind = tok::period;
3587 }
3588 break;
3589 case '&':
3590 Char = getCharAndSize(CurPtr, SizeTmp);
3591 if (Char == '&') {
3592 Kind = tok::ampamp;
3593 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3594 } else if (Char == '=') {
3595 Kind = tok::ampequal;
3596 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3597 } else {
3598 Kind = tok::amp;
3599 }
3600 break;
3601 case '*':
3602 if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3603 Kind = tok::starequal;
3604 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3605 } else {
3606 Kind = tok::star;
3607 }
3608 break;
3609 case '+':
3610 Char = getCharAndSize(CurPtr, SizeTmp);
3611 if (Char == '+') {
3612 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3613 Kind = tok::plusplus;
3614 } else if (Char == '=') {
3615 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3616 Kind = tok::plusequal;
3617 } else {
3618 Kind = tok::plus;
3619 }
3620 break;
3621 case '-':
3622 Char = getCharAndSize(CurPtr, SizeTmp);
3623 if (Char == '-') { // --
3624 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3625 Kind = tok::minusminus;
3626 } else if (Char == '>' && LangOpts.CPlusPlus &&
3627 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3628 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3629 SizeTmp2, Result);
3630 Kind = tok::arrowstar;
3631 } else if (Char == '>') { // ->
3632 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3633 Kind = tok::arrow;
3634 } else if (Char == '=') { // -=
3635 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3636 Kind = tok::minusequal;
3637 } else {
3638 Kind = tok::minus;
3639 }
3640 break;
3641 case '~':
3642 Kind = tok::tilde;
3643 break;
3644 case '!':
3645 if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3646 Kind = tok::exclaimequal;
3647 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3648 } else {
3649 Kind = tok::exclaim;
3650 }
3651 break;
3652 case '/':
3653 // 6.4.9: Comments
3654 Char = getCharAndSize(CurPtr, SizeTmp);
3655 if (Char == '/') { // Line comment.
3656 // Even if Line comments are disabled (e.g. in C89 mode), we generally
3657 // want to lex this as a comment. There is one problem with this though,
3658 // that in one particular corner case, this can change the behavior of the
3659 // resultant program. For example, In "foo //**/ bar", C89 would lex
3660 // this as "foo / bar" and languages with Line comments would lex it as
3661 // "foo". Check to see if the character after the second slash is a '*'.
3662 // If so, we will lex that as a "/" instead of the start of a comment.
3663 // However, we never do this if we are just preprocessing.
3664 bool TreatAsComment = LangOpts.LineComment &&
3665 (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3666 if (!TreatAsComment)
3667 if (!(PP && PP->isPreprocessedOutput()))
3668 TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3669
3670 if (TreatAsComment) {
3671 if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3672 TokAtPhysicalStartOfLine))
3673 return true; // There is a token to return.
3674
3675 // It is common for the tokens immediately after a // comment to be
3676 // whitespace (indentation for the next line). Instead of going through
3677 // the big switch, handle it efficiently now.
3678 goto SkipIgnoredUnits;
3679 }
3680 }
3681
3682 if (Char == '*') { // /**/ comment.
3683 if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3684 TokAtPhysicalStartOfLine))
3685 return true; // There is a token to return.
3686
3687 // We only saw whitespace, so just try again with this lexer.
3688 // (We manually eliminate the tail call to avoid recursion.)
3689 goto LexNextToken;
3690 }
3691
3692 if (Char == '=') {
3693 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3694 Kind = tok::slashequal;
3695 } else {
3696 Kind = tok::slash;
3697 }
3698 break;
3699 case '%':
3700 Char = getCharAndSize(CurPtr, SizeTmp);
3701 if (Char == '=') {
3702 Kind = tok::percentequal;
3703 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3704 } else if (LangOpts.Digraphs && Char == '>') {
3705 Kind = tok::r_brace; // '%>' -> '}'
3706 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3707 } else if (LangOpts.Digraphs && Char == ':') {
3708 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3709 Char = getCharAndSize(CurPtr, SizeTmp);
3710 if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3711 Kind = tok::hashhash; // '%:%:' -> '##'
3712 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3713 SizeTmp2, Result);
3714 } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3715 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3716 if (!isLexingRawMode())
3717 Diag(BufferPtr, diag::ext_charize_microsoft);
3718 Kind = tok::hashat;
3719 } else { // '%:' -> '#'
3720 // We parsed a # character. If this occurs at the start of the line,
3721 // it's actually the start of a preprocessing directive. Callback to
3722 // the preprocessor to handle it.
3723 // TODO: -fpreprocessed mode??
3724 if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3725 goto HandleDirective;
3726
3727 Kind = tok::hash;
3728 }
3729 } else {
3730 Kind = tok::percent;
3731 }
3732 break;
3733 case '<':
3734 Char = getCharAndSize(CurPtr, SizeTmp);
3735 if (ParsingFilename) {
3736 return LexAngledStringLiteral(Result, CurPtr);
3737 } else if (Char == '<') {
3738 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3739 if (After == '=') {
3740 Kind = tok::lesslessequal;
3741 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3742 SizeTmp2, Result);
3743 } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3744 // If this is actually a '<<<<<<<' version control conflict marker,
3745 // recognize it as such and recover nicely.
3746 goto LexNextToken;
3747 } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3748 // If this is '<<<<' and we're in a Perforce-style conflict marker,
3749 // ignore it.
3750 goto LexNextToken;
3751 } else if (LangOpts.CUDA && After == '<') {
3752 Kind = tok::lesslessless;
3753 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3754 SizeTmp2, Result);
3755 } else {
3756 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3757 Kind = tok::lessless;
3758 }
3759 } else if (Char == '=') {
3760 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3761 if (After == '>') {
3762 if (getLangOpts().CPlusPlus20) {
3763 if (!isLexingRawMode())
3764 Diag(BufferPtr, diag::warn_cxx17_compat_spaceship);
3765 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3766 SizeTmp2, Result);
3767 Kind = tok::spaceship;
3768 break;
3769 }
3770 // Suggest adding a space between the '<=' and the '>' to avoid a
3771 // change in semantics if this turns up in C++ <=17 mode.
3772 if (getLangOpts().CPlusPlus && !isLexingRawMode()) {
3773 Diag(BufferPtr, diag::warn_cxx20_compat_spaceship)
3774 << FixItHint::CreateInsertion(
3775 getSourceLocation(CurPtr + SizeTmp, SizeTmp2), " ");
3776 }
3777 }
3778 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3779 Kind = tok::lessequal;
3780 } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3781 if (LangOpts.CPlusPlus11 &&
3782 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3783 // C++0x [lex.pptoken]p3:
3784 // Otherwise, if the next three characters are <:: and the subsequent
3785 // character is neither : nor >, the < is treated as a preprocessor
3786 // token by itself and not as the first character of the alternative
3787 // token <:.
3788 unsigned SizeTmp3;
3789 char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3790 if (After != ':' && After != '>') {
3791 Kind = tok::less;
3792 if (!isLexingRawMode())
3793 Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3794 break;
3795 }
3796 }
3797
3798 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3799 Kind = tok::l_square;
3800 } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3801 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3802 Kind = tok::l_brace;
3803 } else if (Char == '#' && /*Not a trigraph*/ SizeTmp == 1 &&
3804 lexEditorPlaceholder(Result, CurPtr)) {
3805 return true;
3806 } else {
3807 Kind = tok::less;
3808 }
3809 break;
3810 case '>':
3811 Char = getCharAndSize(CurPtr, SizeTmp);
3812 if (Char == '=') {
3813 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3814 Kind = tok::greaterequal;
3815 } else if (Char == '>') {
3816 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3817 if (After == '=') {
3818 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3819 SizeTmp2, Result);
3820 Kind = tok::greatergreaterequal;
3821 } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3822 // If this is actually a '>>>>' conflict marker, recognize it as such
3823 // and recover nicely.
3824 goto LexNextToken;
3825 } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3826 // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3827 goto LexNextToken;
3828 } else if (LangOpts.CUDA && After == '>') {
3829 Kind = tok::greatergreatergreater;
3830 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3831 SizeTmp2, Result);
3832 } else {
3833 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3834 Kind = tok::greatergreater;
3835 }
3836 } else {
3837 Kind = tok::greater;
3838 }
3839 break;
3840 case '^':
3841 Char = getCharAndSize(CurPtr, SizeTmp);
3842 if (Char == '=') {
3843 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3844 Kind = tok::caretequal;
3845 } else if (LangOpts.OpenCL && Char == '^') {
3846 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3847 Kind = tok::caretcaret;
3848 } else {
3849 Kind = tok::caret;
3850 }
3851 break;
3852 case '|':
3853 Char = getCharAndSize(CurPtr, SizeTmp);
3854 if (Char == '=') {
3855 Kind = tok::pipeequal;
3856 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3857 } else if (Char == '|') {
3858 // If this is '|||||||' and we're in a conflict marker, ignore it.
3859 if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3860 goto LexNextToken;
3861 Kind = tok::pipepipe;
3862 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3863 } else {
3864 Kind = tok::pipe;
3865 }
3866 break;
3867 case ':':
3868 Char = getCharAndSize(CurPtr, SizeTmp);
3869 if (LangOpts.Digraphs && Char == '>') {
3870 Kind = tok::r_square; // ':>' -> ']'
3871 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3872 } else if ((LangOpts.CPlusPlus ||
3873 LangOpts.DoubleSquareBracketAttributes) &&
3874 Char == ':') {
3875 Kind = tok::coloncolon;
3876 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3877 } else {
3878 Kind = tok::colon;
3879 }
3880 break;
3881 case ';':
3882 Kind = tok::semi;
3883 break;
3884 case '=':
3885 Char = getCharAndSize(CurPtr, SizeTmp);
3886 if (Char == '=') {
3887 // If this is '====' and we're in a conflict marker, ignore it.
3888 if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3889 goto LexNextToken;
3890
3891 Kind = tok::equalequal;
3892 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3893 } else {
3894 Kind = tok::equal;
3895 }
3896 break;
3897 case ',':
3898 Kind = tok::comma;
3899 break;
3900 case '#':
3901 Char = getCharAndSize(CurPtr, SizeTmp);
3902 if (Char == '#') {
3903 Kind = tok::hashhash;
3904 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3905 } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3906 Kind = tok::hashat;
3907 if (!isLexingRawMode())
3908 Diag(BufferPtr, diag::ext_charize_microsoft);
3909 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3910 } else {
3911 // We parsed a # character. If this occurs at the start of the line,
3912 // it's actually the start of a preprocessing directive. Callback to
3913 // the preprocessor to handle it.
3914 // TODO: -fpreprocessed mode??
3915 if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3916 goto HandleDirective;
3917
3918 Kind = tok::hash;
3919 }
3920 break;
3921
3922 case '@':
3923 // Objective C support.
3924 if (CurPtr[-1] == '@' && LangOpts.ObjC)
3925 Kind = tok::at;
3926 else
3927 Kind = tok::unknown;
3928 break;
3929
3930 // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3931 case '\\':
3932 if (!LangOpts.AsmPreprocessor) {
3933 if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3934 if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3935 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3936 return true; // KeepWhitespaceMode
3937
3938 // We only saw whitespace, so just try again with this lexer.
3939 // (We manually eliminate the tail call to avoid recursion.)
3940 goto LexNextToken;
3941 }
3942
3943 return LexUnicode(Result, CodePoint, CurPtr);
3944 }
3945 }
3946
3947 Kind = tok::unknown;
3948 break;
3949
3950 default: {
3951 if (isASCII(Char)) {
3952 Kind = tok::unknown;
3953 break;
3954 }
3955
3956 llvm::UTF32 CodePoint;
3957
3958 // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
3959 // an escaped newline.
3960 --CurPtr;
3961 llvm::ConversionResult Status =
3962 llvm::convertUTF8Sequence((const llvm::UTF8 **)&CurPtr,
3963 (const llvm::UTF8 *)BufferEnd,
3964 &CodePoint,
3965 llvm::strictConversion);
3966 if (Status == llvm::conversionOK) {
3967 if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3968 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3969 return true; // KeepWhitespaceMode
3970
3971 // We only saw whitespace, so just try again with this lexer.
3972 // (We manually eliminate the tail call to avoid recursion.)
3973 goto LexNextToken;
3974 }
3975 return LexUnicode(Result, CodePoint, CurPtr);
3976 }
3977
3978 if (isLexingRawMode() || ParsingPreprocessorDirective ||
3979 PP->isPreprocessedOutput()) {
3980 ++CurPtr;
3981 Kind = tok::unknown;
3982 break;
3983 }
3984
3985 // Non-ASCII characters tend to creep into source code unintentionally.
3986 // Instead of letting the parser complain about the unknown token,
3987 // just diagnose the invalid UTF-8, then drop the character.
3988 Diag(CurPtr, diag::err_invalid_utf8);
3989
3990 BufferPtr = CurPtr+1;
3991 // We're pretending the character didn't exist, so just try again with
3992 // this lexer.
3993 // (We manually eliminate the tail call to avoid recursion.)
3994 goto LexNextToken;
3995 }
3996 }
3997
3998 // Notify MIOpt that we read a non-whitespace/non-comment token.
3999 MIOpt.ReadToken();
4000
4001 // Update the location of token as well as BufferPtr.
4002 FormTokenWithChars(Result, CurPtr, Kind);
4003 return true;
4004
4005HandleDirective:
4006 // We parsed a # character and it's the start of a preprocessing directive.
4007
4008 FormTokenWithChars(Result, CurPtr, tok::hash);
4009 PP->HandleDirective(Result);
4010
4011 if (PP->hadModuleLoaderFatalFailure()) {
4012 // With a fatal failure in the module loader, we abort parsing.
4013 assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof")((void)0);
4014 return true;
4015 }
4016
4017 // We parsed the directive; lex a token with the new state.
4018 return false;
4019}

/usr/src/gnu/usr.bin/clang/libclangLex/../../../llvm/clang/include/clang/Lex/Lexer.h

1//===- Lexer.h - C Language Family Lexer ------------------------*- C++ -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This file defines the Lexer interface.
10//
11//===----------------------------------------------------------------------===//
12
13#ifndef LLVM_CLANG_LEX_LEXER_H
14#define LLVM_CLANG_LEX_LEXER_H
15
16#include "clang/Basic/LangOptions.h"
17#include "clang/Basic/SourceLocation.h"
18#include "clang/Basic/TokenKinds.h"
19#include "clang/Lex/PreprocessorLexer.h"
20#include "clang/Lex/Token.h"
21#include "llvm/ADT/Optional.h"
22#include "llvm/ADT/SmallVector.h"
23#include "llvm/ADT/StringRef.h"
24#include <cassert>
25#include <cstdint>
26#include <string>
27
28namespace llvm {
29
30class MemoryBufferRef;
31
32} // namespace llvm
33
34namespace clang {
35
36class DiagnosticBuilder;
37class Preprocessor;
38class SourceManager;
39
40/// ConflictMarkerKind - Kinds of conflict marker which the lexer might be
41/// recovering from.
42enum ConflictMarkerKind {
43 /// Not within a conflict marker.
44 CMK_None,
45
46 /// A normal or diff3 conflict marker, initiated by at least 7 "<"s,
47 /// separated by at least 7 "="s or "|"s, and terminated by at least 7 ">"s.
48 CMK_Normal,
49
50 /// A Perforce-style conflict marker, initiated by 4 ">"s,
51 /// separated by 4 "="s, and terminated by 4 "<"s.
52 CMK_Perforce
53};
54
55/// Describes the bounds (start, size) of the preamble and a flag required by
56/// PreprocessorOptions::PrecompiledPreambleBytes.
57/// The preamble includes the BOM, if any.
58struct PreambleBounds {
59 /// Size of the preamble in bytes.
60 unsigned Size;
61
62 /// Whether the preamble ends at the start of a new line.
63 ///
64 /// Used to inform the lexer as to whether it's starting at the beginning of
65 /// a line after skipping the preamble.
66 bool PreambleEndsAtStartOfLine;
67
68 PreambleBounds(unsigned Size, bool PreambleEndsAtStartOfLine)
69 : Size(Size), PreambleEndsAtStartOfLine(PreambleEndsAtStartOfLine) {}
70};
71
72/// Lexer - This provides a simple interface that turns a text buffer into a
73/// stream of tokens. This provides no support for file reading or buffering,
74/// or buffering/seeking of tokens, only forward lexing is supported. It relies
75/// on the specified Preprocessor object to handle preprocessor directives, etc.
76class Lexer : public PreprocessorLexer {
77 friend class Preprocessor;
78
79 void anchor() override;
80
81 //===--------------------------------------------------------------------===//
82 // Constant configuration values for this lexer.
83
84 // Start of the buffer.
85 const char *BufferStart;
86
87 // End of the buffer.
88 const char *BufferEnd;
89
90 // Location for start of file.
91 SourceLocation FileLoc;
92
93 // LangOpts enabled by this language (cache).
94 LangOptions LangOpts;
95
96 // True if lexer for _Pragma handling.
97 bool Is_PragmaLexer;
98
99 //===--------------------------------------------------------------------===//
100 // Context-specific lexing flags set by the preprocessor.
101 //
102
103 /// ExtendedTokenMode - The lexer can optionally keep comments and whitespace
104 /// and return them as tokens. This is used for -C and -CC modes, and
105 /// whitespace preservation can be useful for some clients that want to lex
106 /// the file in raw mode and get every character from the file.
107 ///
108 /// When this is set to 2 it returns comments and whitespace. When set to 1
109 /// it returns comments, when it is set to 0 it returns normal tokens only.
110 unsigned char ExtendedTokenMode;
111
112 //===--------------------------------------------------------------------===//
113 // Context that changes as the file is lexed.
114 // NOTE: any state that mutates when in raw mode must have save/restore code
115 // in Lexer::isNextPPTokenLParen.
116
117 // BufferPtr - Current pointer into the buffer. This is the next character
118 // to be lexed.
119 const char *BufferPtr;
120
121 // IsAtStartOfLine - True if the next lexed token should get the "start of
122 // line" flag set on it.
123 bool IsAtStartOfLine;
124
125 bool IsAtPhysicalStartOfLine;
126
127 bool HasLeadingSpace;
128
129 bool HasLeadingEmptyMacro;
130
131 // NewLinePtr - A pointer to new line character '\n' being lexed. For '\r\n',
132 // it also points to '\n.'
133 const char *NewLinePtr;
134
135 // CurrentConflictMarkerState - The kind of conflict marker we are handling.
136 ConflictMarkerKind CurrentConflictMarkerState;
137
138 void InitLexer(const char *BufStart, const char *BufPtr, const char *BufEnd);
139
140public:
141 /// Lexer constructor - Create a new lexer object for the specified buffer
142 /// with the specified preprocessor managing the lexing process. This lexer
143 /// assumes that the associated file buffer and Preprocessor objects will
144 /// outlive it, so it doesn't take ownership of either of them.
145 Lexer(FileID FID, const llvm::MemoryBufferRef &InputFile, Preprocessor &PP);
146
147 /// Lexer constructor - Create a new raw lexer object. This object is only
148 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the
149 /// text range will outlive it, so it doesn't take ownership of it.
150 Lexer(SourceLocation FileLoc, const LangOptions &LangOpts,
151 const char *BufStart, const char *BufPtr, const char *BufEnd);
152
153 /// Lexer constructor - Create a new raw lexer object. This object is only
154 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the
155 /// text range will outlive it, so it doesn't take ownership of it.
156 Lexer(FileID FID, const llvm::MemoryBufferRef &FromFile,
157 const SourceManager &SM, const LangOptions &LangOpts);
158
159 Lexer(const Lexer &) = delete;
160 Lexer &operator=(const Lexer &) = delete;
161
162 /// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
163 /// _Pragma expansion. This has a variety of magic semantics that this method
164 /// sets up. It returns a new'd Lexer that must be delete'd when done.
165 static Lexer *Create_PragmaLexer(SourceLocation SpellingLoc,
166 SourceLocation ExpansionLocStart,
167 SourceLocation ExpansionLocEnd,
168 unsigned TokLen, Preprocessor &PP);
169
170 /// getLangOpts - Return the language features currently enabled.
171 /// NOTE: this lexer modifies features as a file is parsed!
172 const LangOptions &getLangOpts() const { return LangOpts; }
173
174 /// getFileLoc - Return the File Location for the file we are lexing out of.
175 /// The physical location encodes the location where the characters come from,
176 /// the virtual location encodes where we should *claim* the characters came
177 /// from. Currently this is only used by _Pragma handling.
178 SourceLocation getFileLoc() const { return FileLoc; }
179
180private:
181 /// Lex - Return the next token in the file. If this is the end of file, it
182 /// return the tok::eof token. This implicitly involves the preprocessor.
183 bool Lex(Token &Result);
184
185public:
186 /// isPragmaLexer - Returns true if this Lexer is being used to lex a pragma.
187 bool isPragmaLexer() const { return Is_PragmaLexer; }
188
189private:
190 /// IndirectLex - An indirect call to 'Lex' that can be invoked via
191 /// the PreprocessorLexer interface.
192 void IndirectLex(Token &Result) override { Lex(Result); }
193
194public:
195 /// LexFromRawLexer - Lex a token from a designated raw lexer (one with no
196 /// associated preprocessor object. Return true if the 'next character to
197 /// read' pointer points at the end of the lexer buffer, false otherwise.
198 bool LexFromRawLexer(Token &Result) {
199 assert(LexingRawMode && "Not already in raw mode!")((void)0);
200 Lex(Result);
201 // Note that lexing to the end of the buffer doesn't implicitly delete the
202 // lexer when in raw mode.
203 return BufferPtr == BufferEnd;
204 }
205
206 /// isKeepWhitespaceMode - Return true if the lexer should return tokens for
207 /// every character in the file, including whitespace and comments. This
208 /// should only be used in raw mode, as the preprocessor is not prepared to
209 /// deal with the excess tokens.
210 bool isKeepWhitespaceMode() const {
211 return ExtendedTokenMode > 1;
212 }
213
214 /// SetKeepWhitespaceMode - This method lets clients enable or disable
215 /// whitespace retention mode.
216 void SetKeepWhitespaceMode(bool Val) {
217 assert((!Val || LexingRawMode || LangOpts.TraditionalCPP) &&((void)0)
218 "Can only retain whitespace in raw mode or -traditional-cpp")((void)0);
219 ExtendedTokenMode = Val ? 2 : 0;
220 }
221
222 /// inKeepCommentMode - Return true if the lexer should return comments as
223 /// tokens.
224 bool inKeepCommentMode() const {
225 return ExtendedTokenMode > 0;
226 }
227
228 /// SetCommentRetentionMode - Change the comment retention mode of the lexer
229 /// to the specified mode. This is really only useful when lexing in raw
230 /// mode, because otherwise the lexer needs to manage this.
231 void SetCommentRetentionState(bool Mode) {
232 assert(!isKeepWhitespaceMode() &&((void)0)
233 "Can't play with comment retention state when retaining whitespace")((void)0);
234 ExtendedTokenMode = Mode ? 1 : 0;
235 }
236
237 /// Sets the extended token mode back to its initial value, according to the
238 /// language options and preprocessor. This controls whether the lexer
239 /// produces comment and whitespace tokens.
240 ///
241 /// This requires the lexer to have an associated preprocessor. A standalone
242 /// lexer has nothing to reset to.
243 void resetExtendedTokenMode();
244
245 /// Gets source code buffer.
246 StringRef getBuffer() const {
247 return StringRef(BufferStart, BufferEnd - BufferStart);
248 }
249
250 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
251 /// uninterpreted string. This switches the lexer out of directive mode.
252 void ReadToEndOfLine(SmallVectorImpl<char> *Result = nullptr);
253
254
255 /// Diag - Forwarding function for diagnostics. This translate a source
256 /// position in the current buffer into a SourceLocation object for rendering.
257 DiagnosticBuilder Diag(const char *Loc, unsigned DiagID) const;
258
259 /// getSourceLocation - Return a source location identifier for the specified
260 /// offset in the current file.
261 SourceLocation getSourceLocation(const char *Loc, unsigned TokLen = 1) const;
262
263 /// getSourceLocation - Return a source location for the next character in
264 /// the current file.
265 SourceLocation getSourceLocation() override {
266 return getSourceLocation(BufferPtr);
267 }
268
269 /// Return the current location in the buffer.
270 const char *getBufferLocation() const { return BufferPtr; }
271
272 /// Returns the current lexing offset.
273 unsigned getCurrentBufferOffset() {
274 assert(BufferPtr >= BufferStart && "Invalid buffer state")((void)0);
275 return BufferPtr - BufferStart;
276 }
277
278 /// Skip over \p NumBytes bytes.
279 ///
280 /// If the skip is successful, the next token will be lexed from the new
281 /// offset. The lexer also assumes that we skipped to the start of the line.
282 ///
283 /// \returns true if the skip failed (new offset would have been past the
284 /// end of the buffer), false otherwise.
285 bool skipOver(unsigned NumBytes);
286
287 /// Stringify - Convert the specified string into a C string by i) escaping
288 /// '\\' and " characters and ii) replacing newline character(s) with "\\n".
289 /// If Charify is true, this escapes the ' character instead of ".
290 static std::string Stringify(StringRef Str, bool Charify = false);
291
292 /// Stringify - Convert the specified string into a C string by i) escaping
293 /// '\\' and " characters and ii) replacing newline character(s) with "\\n".
294 static void Stringify(SmallVectorImpl<char> &Str);
295
296 /// getSpelling - This method is used to get the spelling of a token into a
297 /// preallocated buffer, instead of as an std::string. The caller is required
298 /// to allocate enough space for the token, which is guaranteed to be at least
299 /// Tok.getLength() bytes long. The length of the actual result is returned.
300 ///
301 /// Note that this method may do two possible things: it may either fill in
302 /// the buffer specified with characters, or it may *change the input pointer*
303 /// to point to a constant buffer with the data already in it (avoiding a
304 /// copy). The caller is not allowed to modify the returned buffer pointer
305 /// if an internal buffer is returned.
306 static unsigned getSpelling(const Token &Tok, const char *&Buffer,
307 const SourceManager &SourceMgr,
308 const LangOptions &LangOpts,
309 bool *Invalid = nullptr);
310
311 /// getSpelling() - Return the 'spelling' of the Tok token. The spelling of a
312 /// token is the characters used to represent the token in the source file
313 /// after trigraph expansion and escaped-newline folding. In particular, this
314 /// wants to get the true, uncanonicalized, spelling of things like digraphs
315 /// UCNs, etc.
316 static std::string getSpelling(const Token &Tok,
317 const SourceManager &SourceMgr,
318 const LangOptions &LangOpts,
319 bool *Invalid = nullptr);
320
321 /// getSpelling - This method is used to get the spelling of the
322 /// token at the given source location. If, as is usually true, it
323 /// is not necessary to copy any data, then the returned string may
324 /// not point into the provided buffer.
325 ///
326 /// This method lexes at the expansion depth of the given
327 /// location and does not jump to the expansion or spelling
328 /// location.
329 static StringRef getSpelling(SourceLocation loc,
330 SmallVectorImpl<char> &buffer,
331 const SourceManager &SM,
332 const LangOptions &options,
333 bool *invalid = nullptr);
334
335 /// MeasureTokenLength - Relex the token at the specified location and return
336 /// its length in bytes in the input file. If the token needs cleaning (e.g.
337 /// includes a trigraph or an escaped newline) then this count includes bytes
338 /// that are part of that.
339 static unsigned MeasureTokenLength(SourceLocation Loc,
340 const SourceManager &SM,
341 const LangOptions &LangOpts);
342
343 /// Relex the token at the specified location.
344 /// \returns true if there was a failure, false on success.
345 static bool getRawToken(SourceLocation Loc, Token &Result,
346 const SourceManager &SM,
347 const LangOptions &LangOpts,
348 bool IgnoreWhiteSpace = false);
349
350 /// Given a location any where in a source buffer, find the location
351 /// that corresponds to the beginning of the token in which the original
352 /// source location lands.
353 static SourceLocation GetBeginningOfToken(SourceLocation Loc,
354 const SourceManager &SM,
355 const LangOptions &LangOpts);
356
357 /// Get the physical length (including trigraphs and escaped newlines) of the
358 /// first \p Characters characters of the token starting at TokStart.
359 static unsigned getTokenPrefixLength(SourceLocation TokStart,
360 unsigned CharNo,
361 const SourceManager &SM,
362 const LangOptions &LangOpts);
363
364 /// AdvanceToTokenCharacter - If the current SourceLocation specifies a
365 /// location at the start of a token, return a new location that specifies a
366 /// character within the token. This handles trigraphs and escaped newlines.
367 static SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart,
368 unsigned Characters,
369 const SourceManager &SM,
370 const LangOptions &LangOpts) {
371 return TokStart.getLocWithOffset(
372 getTokenPrefixLength(TokStart, Characters, SM, LangOpts));
373 }
374
375 /// Computes the source location just past the end of the
376 /// token at this source location.
377 ///
378 /// This routine can be used to produce a source location that
379 /// points just past the end of the token referenced by \p Loc, and
380 /// is generally used when a diagnostic needs to point just after a
381 /// token where it expected something different that it received. If
382 /// the returned source location would not be meaningful (e.g., if
383 /// it points into a macro), this routine returns an invalid
384 /// source location.
385 ///
386 /// \param Offset an offset from the end of the token, where the source
387 /// location should refer to. The default offset (0) produces a source
388 /// location pointing just past the end of the token; an offset of 1 produces
389 /// a source location pointing to the last character in the token, etc.
390 static SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset,
391 const SourceManager &SM,
392 const LangOptions &LangOpts);
393
394 /// Given a token range, produce a corresponding CharSourceRange that
395 /// is not a token range. This allows the source range to be used by
396 /// components that don't have access to the lexer and thus can't find the
397 /// end of the range for themselves.
398 static CharSourceRange getAsCharRange(SourceRange Range,
399 const SourceManager &SM,
400 const LangOptions &LangOpts) {
401 SourceLocation End = getLocForEndOfToken(Range.getEnd(), 0, SM, LangOpts);
402 return End.isInvalid() ? CharSourceRange()
403 : CharSourceRange::getCharRange(
404 Range.getBegin(), End);
405 }
406 static CharSourceRange getAsCharRange(CharSourceRange Range,
407 const SourceManager &SM,
408 const LangOptions &LangOpts) {
409 return Range.isTokenRange()
410 ? getAsCharRange(Range.getAsRange(), SM, LangOpts)
411 : Range;
412 }
413
414 /// Returns true if the given MacroID location points at the first
415 /// token of the macro expansion.
416 ///
417 /// \param MacroBegin If non-null and function returns true, it is set to
418 /// begin location of the macro.
419 static bool isAtStartOfMacroExpansion(SourceLocation loc,
420 const SourceManager &SM,
421 const LangOptions &LangOpts,
422 SourceLocation *MacroBegin = nullptr);
423
424 /// Returns true if the given MacroID location points at the last
425 /// token of the macro expansion.
426 ///
427 /// \param MacroEnd If non-null and function returns true, it is set to
428 /// end location of the macro.
429 static bool isAtEndOfMacroExpansion(SourceLocation loc,
430 const SourceManager &SM,
431 const LangOptions &LangOpts,
432 SourceLocation *MacroEnd = nullptr);
433
434 /// Accepts a range and returns a character range with file locations.
435 ///
436 /// Returns a null range if a part of the range resides inside a macro
437 /// expansion or the range does not reside on the same FileID.
438 ///
439 /// This function is trying to deal with macros and return a range based on
440 /// file locations. The cases where it can successfully handle macros are:
441 ///
442 /// -begin or end range lies at the start or end of a macro expansion, in
443 /// which case the location will be set to the expansion point, e.g:
444 /// \#define M 1 2
445 /// a M
446 /// If you have a range [a, 2] (where 2 came from the macro), the function
447 /// will return a range for "a M"
448 /// if you have range [a, 1], the function will fail because the range
449 /// overlaps with only a part of the macro
450 ///
451 /// -The macro is a function macro and the range can be mapped to the macro
452 /// arguments, e.g:
453 /// \#define M 1 2
454 /// \#define FM(x) x
455 /// FM(a b M)
456 /// if you have range [b, 2], the function will return the file range "b M"
457 /// inside the macro arguments.
458 /// if you have range [a, 2], the function will return the file range
459 /// "FM(a b M)" since the range includes all of the macro expansion.
460 static CharSourceRange makeFileCharRange(CharSourceRange Range,
461 const SourceManager &SM,
462 const LangOptions &LangOpts);
463
464 /// Returns a string for the source that the range encompasses.
465 static StringRef getSourceText(CharSourceRange Range,
466 const SourceManager &SM,
467 const LangOptions &LangOpts,
468 bool *Invalid = nullptr);
469
470 /// Retrieve the name of the immediate macro expansion.
471 ///
472 /// This routine starts from a source location, and finds the name of the macro
473 /// responsible for its immediate expansion. It looks through any intervening
474 /// macro argument expansions to compute this. It returns a StringRef which
475 /// refers to the SourceManager-owned buffer of the source where that macro
476 /// name is spelled. Thus, the result shouldn't out-live that SourceManager.
477 static StringRef getImmediateMacroName(SourceLocation Loc,
478 const SourceManager &SM,
479 const LangOptions &LangOpts);
480
481 /// Retrieve the name of the immediate macro expansion.
482 ///
483 /// This routine starts from a source location, and finds the name of the
484 /// macro responsible for its immediate expansion. It looks through any
485 /// intervening macro argument expansions to compute this. It returns a
486 /// StringRef which refers to the SourceManager-owned buffer of the source
487 /// where that macro name is spelled. Thus, the result shouldn't out-live
488 /// that SourceManager.
489 ///
490 /// This differs from Lexer::getImmediateMacroName in that any macro argument
491 /// location will result in the topmost function macro that accepted it.
492 /// e.g.
493 /// \code
494 /// MAC1( MAC2(foo) )
495 /// \endcode
496 /// for location of 'foo' token, this function will return "MAC1" while
497 /// Lexer::getImmediateMacroName will return "MAC2".
498 static StringRef getImmediateMacroNameForDiagnostics(
499 SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts);
500
501 /// Compute the preamble of the given file.
502 ///
503 /// The preamble of a file contains the initial comments, include directives,
504 /// and other preprocessor directives that occur before the code in this
505 /// particular file actually begins. The preamble of the main source file is
506 /// a potential prefix header.
507 ///
508 /// \param Buffer The memory buffer containing the file's contents.
509 ///
510 /// \param MaxLines If non-zero, restrict the length of the preamble
511 /// to fewer than this number of lines.
512 ///
513 /// \returns The offset into the file where the preamble ends and the rest
514 /// of the file begins along with a boolean value indicating whether
515 /// the preamble ends at the beginning of a new line.
516 static PreambleBounds ComputePreamble(StringRef Buffer,
517 const LangOptions &LangOpts,
518 unsigned MaxLines = 0);
519
520 /// Finds the token that comes right after the given location.
521 ///
522 /// Returns the next token, or none if the location is inside a macro.
523 static Optional<Token> findNextToken(SourceLocation Loc,
524 const SourceManager &SM,
525 const LangOptions &LangOpts);
526
527 /// Checks that the given token is the first token that occurs after
528 /// the given location (this excludes comments and whitespace). Returns the
529 /// location immediately after the specified token. If the token is not found
530 /// or the location is inside a macro, the returned source location will be
531 /// invalid.
532 static SourceLocation findLocationAfterToken(SourceLocation loc,
533 tok::TokenKind TKind,
534 const SourceManager &SM,
535 const LangOptions &LangOpts,
536 bool SkipTrailingWhitespaceAndNewLine);
537
538 /// Returns true if the given character could appear in an identifier.
539 static bool isIdentifierBodyChar(char c, const LangOptions &LangOpts);
540
541 /// Checks whether new line pointed by Str is preceded by escape
542 /// sequence.
543 static bool isNewLineEscaped(const char *BufferStart, const char *Str);
544
545 /// getCharAndSizeNoWarn - Like the getCharAndSize method, but does not ever
546 /// emit a warning.
547 static inline char getCharAndSizeNoWarn(const char *Ptr, unsigned &Size,
548 const LangOptions &LangOpts) {
549 // If this is not a trigraph and not a UCN or escaped newline, return
550 // quickly.
551 if (isObviouslySimpleCharacter(Ptr[0])) {
552 Size = 1;
553 return *Ptr;
554 }
555
556 Size = 0;
557 return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
558 }
559
560 /// Returns the leading whitespace for line that corresponds to the given
561 /// location \p Loc.
562 static StringRef getIndentationForLine(SourceLocation Loc,
563 const SourceManager &SM);
564
565private:
566 //===--------------------------------------------------------------------===//
567 // Internal implementation interfaces.
568
569 /// LexTokenInternal - Internal interface to lex a preprocessing token. Called
570 /// by Lex.
571 ///
572 bool LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine);
573
574 bool CheckUnicodeWhitespace(Token &Result, uint32_t C, const char *CurPtr);
575
576 /// Given that a token begins with the Unicode character \p C, figure out
577 /// what kind of token it is and dispatch to the appropriate lexing helper
578 /// function.
579 bool LexUnicode(Token &Result, uint32_t C, const char *CurPtr);
580
581 /// FormTokenWithChars - When we lex a token, we have identified a span
582 /// starting at BufferPtr, going to TokEnd that forms the token. This method
583 /// takes that range and assigns it to the token as its location and size. In
584 /// addition, since tokens cannot overlap, this also updates BufferPtr to be
585 /// TokEnd.
586 void FormTokenWithChars(Token &Result, const char *TokEnd,
587 tok::TokenKind Kind) {
588 unsigned TokLen = TokEnd-BufferPtr;
589 Result.setLength(TokLen);
590 Result.setLocation(getSourceLocation(BufferPtr, TokLen));
591 Result.setKind(Kind);
592 BufferPtr = TokEnd;
593 }
594
595 /// isNextPPTokenLParen - Return 1 if the next unexpanded token will return a
596 /// tok::l_paren token, 0 if it is something else and 2 if there are no more
597 /// tokens in the buffer controlled by this lexer.
598 unsigned isNextPPTokenLParen();
599
600 //===--------------------------------------------------------------------===//
601 // Lexer character reading interfaces.
602
603 // This lexer is built on two interfaces for reading characters, both of which
604 // automatically provide phase 1/2 translation. getAndAdvanceChar is used
605 // when we know that we will be reading a character from the input buffer and
606 // that this character will be part of the result token. This occurs in (f.e.)
607 // string processing, because we know we need to read until we find the
608 // closing '"' character.
609 //
610 // The second interface is the combination of getCharAndSize with
611 // ConsumeChar. getCharAndSize reads a phase 1/2 translated character,
612 // returning it and its size. If the lexer decides that this character is
613 // part of the current token, it calls ConsumeChar on it. This two stage
614 // approach allows us to emit diagnostics for characters (e.g. warnings about
615 // trigraphs), knowing that they only are emitted if the character is
616 // consumed.
617
618 /// isObviouslySimpleCharacter - Return true if the specified character is
619 /// obviously the same in translation phase 1 and translation phase 3. This
620 /// can return false for characters that end up being the same, but it will
621 /// never return true for something that needs to be mapped.
622 static bool isObviouslySimpleCharacter(char C) {
623 return C != '?' && C != '\\';
624 }
625
626 /// getAndAdvanceChar - Read a single 'character' from the specified buffer,
627 /// advance over it, and return it. This is tricky in several cases. Here we
628 /// just handle the trivial case and fall-back to the non-inlined
629 /// getCharAndSizeSlow method to handle the hard case.
630 inline char getAndAdvanceChar(const char *&Ptr, Token &Tok) {
631 // If this is not a trigraph and not a UCN or escaped newline, return
632 // quickly.
633 if (isObviouslySimpleCharacter(Ptr[0])) return *Ptr++;
634
635 unsigned Size = 0;
636 char C = getCharAndSizeSlow(Ptr, Size, &Tok);
637 Ptr += Size;
638 return C;
639 }
640
641 /// ConsumeChar - When a character (identified by getCharAndSize) is consumed
642 /// and added to a given token, check to see if there are diagnostics that
643 /// need to be emitted or flags that need to be set on the token. If so, do
644 /// it.
645 const char *ConsumeChar(const char *Ptr, unsigned Size, Token &Tok) {
646 // Normal case, we consumed exactly one token. Just return it.
647 if (Size == 1)
648 return Ptr+Size;
649
650 // Otherwise, re-lex the character with a current token, allowing
651 // diagnostics to be emitted and flags to be set.
652 Size = 0;
653 getCharAndSizeSlow(Ptr, Size, &Tok);
654 return Ptr+Size;
655 }
656
657 /// getCharAndSize - Peek a single 'character' from the specified buffer,
658 /// get its size, and return it. This is tricky in several cases. Here we
659 /// just handle the trivial case and fall-back to the non-inlined
660 /// getCharAndSizeSlow method to handle the hard case.
661 inline char getCharAndSize(const char *Ptr, unsigned &Size) {
662 // If this is not a trigraph and not a UCN or escaped newline, return
663 // quickly.
664 if (isObviouslySimpleCharacter(Ptr[0])) {
13
Taking false branch
665 Size = 1;
666 return *Ptr;
667 }
668
669 Size = 0;
670 return getCharAndSizeSlow(Ptr, Size);
14
Value assigned to field 'PP'
671 }
672
673 /// getCharAndSizeSlow - Handle the slow/uncommon case of the getCharAndSize
674 /// method.
675 char getCharAndSizeSlow(const char *Ptr, unsigned &Size,
676 Token *Tok = nullptr);
677
678 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
679 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" on entry
680 /// to this function.
681 static unsigned getEscapedNewLineSize(const char *P);
682
683 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
684 /// them), skip over them and return the first non-escaped-newline found,
685 /// otherwise return P.
686 static const char *SkipEscapedNewLines(const char *P);
687
688 /// getCharAndSizeSlowNoWarn - Same as getCharAndSizeSlow, but never emits a
689 /// diagnostic.
690 static char getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
691 const LangOptions &LangOpts);
692
693 //===--------------------------------------------------------------------===//
694 // Other lexer functions.
695
696 void SetByteOffset(unsigned Offset, bool StartOfLine);
697
698 void PropagateLineStartLeadingSpaceInfo(Token &Result);
699
700 const char *LexUDSuffix(Token &Result, const char *CurPtr,
701 bool IsStringLiteral);
702
703 // Helper functions to lex the remainder of a token of the specific type.
704 bool LexIdentifier (Token &Result, const char *CurPtr);
705 bool LexNumericConstant (Token &Result, const char *CurPtr);
706 bool LexStringLiteral (Token &Result, const char *CurPtr,
707 tok::TokenKind Kind);
708 bool LexRawStringLiteral (Token &Result, const char *CurPtr,
709 tok::TokenKind Kind);
710 bool LexAngledStringLiteral(Token &Result, const char *CurPtr);
711 bool LexCharConstant (Token &Result, const char *CurPtr,
712 tok::TokenKind Kind);
713 bool LexEndOfFile (Token &Result, const char *CurPtr);
714 bool SkipWhitespace (Token &Result, const char *CurPtr,
715 bool &TokAtPhysicalStartOfLine);
716 bool SkipLineComment (Token &Result, const char *CurPtr,
717 bool &TokAtPhysicalStartOfLine);
718 bool SkipBlockComment (Token &Result, const char *CurPtr,
719 bool &TokAtPhysicalStartOfLine);
720 bool SaveLineComment (Token &Result, const char *CurPtr);
721
722 bool IsStartOfConflictMarker(const char *CurPtr);
723 bool HandleEndOfConflictMarker(const char *CurPtr);
724
725 bool lexEditorPlaceholder(Token &Result, const char *CurPtr);
726
727 bool isCodeCompletionPoint(const char *CurPtr) const;
728 void cutOffLexing() { BufferPtr = BufferEnd; }
729
730 bool isHexaLiteral(const char *Start, const LangOptions &LangOpts);
731
732 void codeCompleteIncludedFile(const char *PathStart,
733 const char *CompletionPoint, bool IsAngled);
734
735 /// Read a universal character name.
736 ///
737 /// \param StartPtr The position in the source buffer after the initial '\'.
738 /// If the UCN is syntactically well-formed (but not
739 /// necessarily valid), this parameter will be updated to
740 /// point to the character after the UCN.
741 /// \param SlashLoc The position in the source buffer of the '\'.
742 /// \param Result The token being formed. Pass \c nullptr to suppress
743 /// diagnostics and handle token formation in the caller.
744 ///
745 /// \return The Unicode codepoint specified by the UCN, or 0 if the UCN is
746 /// invalid.
747 uint32_t tryReadUCN(const char *&StartPtr, const char *SlashLoc, Token *Result);
748
749 /// Try to consume a UCN as part of an identifier at the current
750 /// location.
751 /// \param CurPtr Initially points to the range of characters in the source
752 /// buffer containing the '\'. Updated to point past the end of
753 /// the UCN on success.
754 /// \param Size The number of characters occupied by the '\' (including
755 /// trigraphs and escaped newlines).
756 /// \param Result The token being produced. Marked as containing a UCN on
757 /// success.
758 /// \return \c true if a UCN was lexed and it produced an acceptable
759 /// identifier character, \c false otherwise.
760 bool tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
761 Token &Result);
762
763 /// Try to consume an identifier character encoded in UTF-8.
764 /// \param CurPtr Points to the start of the (potential) UTF-8 code unit
765 /// sequence. On success, updated to point past the end of it.
766 /// \return \c true if a UTF-8 sequence mapping to an acceptable identifier
767 /// character was lexed, \c false otherwise.
768 bool tryConsumeIdentifierUTF8Char(const char *&CurPtr);
769};
770
771} // namespace clang
772
773#endif // LLVM_CLANG_LEX_LEXER_H