Bug Summary

File: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 x86_64-pc-linux-gnu -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 -analyzer-config-compatibility-mode=true -mrelocation-model pic -pic-level 2 -mframe-pointer=none -relaxed-aliasing -fmath-errno -fno-rounding-math -mconstructor-aliases -munwind-tables -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -ffunction-sections -fdata-sections -fcoverage-compilation-dir=/build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/build-llvm/tools/clang/lib/Lex -resource-dir /usr/lib/llvm-14/lib/clang/14.0.0 -D _GNU_SOURCE -D __STDC_CONSTANT_MACROS -D __STDC_FORMAT_MACROS -D __STDC_LIMIT_MACROS -I /build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/build-llvm/tools/clang/lib/Lex -I /build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/clang/lib/Lex -I /build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/clang/include -I /build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/build-llvm/tools/clang/include -I /build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/build-llvm/include -I /build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/llvm/include -D NDEBUG -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../include/c++/10 -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../include/x86_64-linux-gnu/c++/10 -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../include/c++/10/backward -internal-isystem /usr/lib/llvm-14/lib/clang/14.0.0/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -Wno-unused-parameter -Wwrite-strings -Wno-missing-field-initializers -Wno-long-long -Wno-maybe-uninitialized -Wno-class-memaccess -Wno-redundant-move -Wno-pessimizing-move -Wno-noexcept-type -Wno-comment -std=c++14 -fdeprecated-macro -fdebug-compilation-dir=/build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/build-llvm/tools/clang/lib/Lex -fdebug-prefix-map=/build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e=. -ferror-limit 19 -fvisibility-inlines-hidden -stack-protector 2 -fgnuc-version=4.2.1 -vectorize-loops -vectorize-slp -analyzer-output=html -analyzer-config stable-report-filename=true -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/scan-build-2021-09-04-040900-46481-1 -x c++ /build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/clang/lib/Lex/Lexer.cpp

/build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/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 &&(static_cast<void> (0))
86 "We assume that the input buffer has a null character at the end"(static_cast<void> (0))
87 " to simplify lexing!")(static_cast<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")(static_cast<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!")(static_cast<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")(static_cast<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() &&(static_cast<void> (0))
318 "NeedsCleaning flag set on token that didn't need cleaning!")(static_cast<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!")(static_cast<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!")(static_cast<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())(static_cast<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 &&(static_cast<void> (0))
572 FileLocInfo.second >= BeginFileLocInfo.second)(static_cast<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")(static_cast<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")(static_cast<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())(static_cast<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())(static_cast<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")(static_cast<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")(static_cast<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]))(static_cast<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")(static_cast<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 &&(static_cast<void> (0))
1147 "Location out of range for this buffer!")(static_cast<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")(static_cast<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);
36
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 isUnicodeWhitespace(uint32_t Codepoint) {
1450 static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
1451 UnicodeWhitespaceCharRanges);
1452 return UnicodeWhitespaceChars.contains(Codepoint);
1453}
1454
1455static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1456 if (LangOpts.AsmPreprocessor) {
1457 return false;
1458 } else if (LangOpts.DollarIdents && '$' == C) {
1459 return true;
1460 } else if (LangOpts.CPlusPlus) {
1461 // A non-leading codepoint must have the XID_Continue property.
1462 // XIDContinueRanges doesn't contains characters also in XIDStartRanges,
1463 // so we need to check both tables.
1464 // '_' doesn't have the XID_Continue property but is allowed in C++.
1465 static const llvm::sys::UnicodeCharSet XIDStartChars(XIDStartRanges);
1466 static const llvm::sys::UnicodeCharSet XIDContinueChars(XIDContinueRanges);
1467 return C == '_' || XIDStartChars.contains(C) ||
1468 XIDContinueChars.contains(C);
1469 } else if (LangOpts.C11) {
1470 static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1471 C11AllowedIDCharRanges);
1472 return C11AllowedIDChars.contains(C);
1473 } else {
1474 static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1475 C99AllowedIDCharRanges);
1476 return C99AllowedIDChars.contains(C);
1477 }
1478}
1479
1480static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1481 if (LangOpts.AsmPreprocessor) {
1482 return false;
1483 }
1484 if (LangOpts.CPlusPlus) {
1485 static const llvm::sys::UnicodeCharSet XIDStartChars(XIDStartRanges);
1486 // '_' doesn't have the XID_Start property but is allowed in C++.
1487 return C == '_' || XIDStartChars.contains(C);
1488 }
1489 if (!isAllowedIDChar(C, LangOpts))
1490 return false;
1491 if (LangOpts.C11) {
1492 static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1493 C11DisallowedInitialIDCharRanges);
1494 return !C11DisallowedInitialIDChars.contains(C);
1495 }
1496 static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1497 C99DisallowedInitialIDCharRanges);
1498 return !C99DisallowedInitialIDChars.contains(C);
1499}
1500
1501static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1502 const char *End) {
1503 return CharSourceRange::getCharRange(L.getSourceLocation(Begin),
1504 L.getSourceLocation(End));
1505}
1506
1507static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1508 CharSourceRange Range, bool IsFirst) {
1509 // Check C99 compatibility.
1510 if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1511 enum {
1512 CannotAppearInIdentifier = 0,
1513 CannotStartIdentifier
1514 };
1515
1516 static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1517 C99AllowedIDCharRanges);
1518 static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1519 C99DisallowedInitialIDCharRanges);
1520 if (!C99AllowedIDChars.contains(C)) {
1521 Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1522 << Range
1523 << CannotAppearInIdentifier;
1524 } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1525 Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1526 << Range
1527 << CannotStartIdentifier;
1528 }
1529 }
1530}
1531
1532/// After encountering UTF-8 character C and interpreting it as an identifier
1533/// character, check whether it's a homoglyph for a common non-identifier
1534/// source character that is unlikely to be an intentional identifier
1535/// character and warn if so.
1536static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C,
1537 CharSourceRange Range) {
1538 // FIXME: Handle Unicode quotation marks (smart quotes, fullwidth quotes).
1539 struct HomoglyphPair {
1540 uint32_t Character;
1541 char LooksLike;
1542 bool operator<(HomoglyphPair R) const { return Character < R.Character; }
1543 };
1544 static constexpr HomoglyphPair SortedHomoglyphs[] = {
1545 {U'\u00ad', 0}, // SOFT HYPHEN
1546 {U'\u01c3', '!'}, // LATIN LETTER RETROFLEX CLICK
1547 {U'\u037e', ';'}, // GREEK QUESTION MARK
1548 {U'\u200b', 0}, // ZERO WIDTH SPACE
1549 {U'\u200c', 0}, // ZERO WIDTH NON-JOINER
1550 {U'\u200d', 0}, // ZERO WIDTH JOINER
1551 {U'\u2060', 0}, // WORD JOINER
1552 {U'\u2061', 0}, // FUNCTION APPLICATION
1553 {U'\u2062', 0}, // INVISIBLE TIMES
1554 {U'\u2063', 0}, // INVISIBLE SEPARATOR
1555 {U'\u2064', 0}, // INVISIBLE PLUS
1556 {U'\u2212', '-'}, // MINUS SIGN
1557 {U'\u2215', '/'}, // DIVISION SLASH
1558 {U'\u2216', '\\'}, // SET MINUS
1559 {U'\u2217', '*'}, // ASTERISK OPERATOR
1560 {U'\u2223', '|'}, // DIVIDES
1561 {U'\u2227', '^'}, // LOGICAL AND
1562 {U'\u2236', ':'}, // RATIO
1563 {U'\u223c', '~'}, // TILDE OPERATOR
1564 {U'\ua789', ':'}, // MODIFIER LETTER COLON
1565 {U'\ufeff', 0}, // ZERO WIDTH NO-BREAK SPACE
1566 {U'\uff01', '!'}, // FULLWIDTH EXCLAMATION MARK
1567 {U'\uff03', '#'}, // FULLWIDTH NUMBER SIGN
1568 {U'\uff04', '$'}, // FULLWIDTH DOLLAR SIGN
1569 {U'\uff05', '%'}, // FULLWIDTH PERCENT SIGN
1570 {U'\uff06', '&'}, // FULLWIDTH AMPERSAND
1571 {U'\uff08', '('}, // FULLWIDTH LEFT PARENTHESIS
1572 {U'\uff09', ')'}, // FULLWIDTH RIGHT PARENTHESIS
1573 {U'\uff0a', '*'}, // FULLWIDTH ASTERISK
1574 {U'\uff0b', '+'}, // FULLWIDTH ASTERISK
1575 {U'\uff0c', ','}, // FULLWIDTH COMMA
1576 {U'\uff0d', '-'}, // FULLWIDTH HYPHEN-MINUS
1577 {U'\uff0e', '.'}, // FULLWIDTH FULL STOP
1578 {U'\uff0f', '/'}, // FULLWIDTH SOLIDUS
1579 {U'\uff1a', ':'}, // FULLWIDTH COLON
1580 {U'\uff1b', ';'}, // FULLWIDTH SEMICOLON
1581 {U'\uff1c', '<'}, // FULLWIDTH LESS-THAN SIGN
1582 {U'\uff1d', '='}, // FULLWIDTH EQUALS SIGN
1583 {U'\uff1e', '>'}, // FULLWIDTH GREATER-THAN SIGN
1584 {U'\uff1f', '?'}, // FULLWIDTH QUESTION MARK
1585 {U'\uff20', '@'}, // FULLWIDTH COMMERCIAL AT
1586 {U'\uff3b', '['}, // FULLWIDTH LEFT SQUARE BRACKET
1587 {U'\uff3c', '\\'}, // FULLWIDTH REVERSE SOLIDUS
1588 {U'\uff3d', ']'}, // FULLWIDTH RIGHT SQUARE BRACKET
1589 {U'\uff3e', '^'}, // FULLWIDTH CIRCUMFLEX ACCENT
1590 {U'\uff5b', '{'}, // FULLWIDTH LEFT CURLY BRACKET
1591 {U'\uff5c', '|'}, // FULLWIDTH VERTICAL LINE
1592 {U'\uff5d', '}'}, // FULLWIDTH RIGHT CURLY BRACKET
1593 {U'\uff5e', '~'}, // FULLWIDTH TILDE
1594 {0, 0}
1595 };
1596 auto Homoglyph =
1597 std::lower_bound(std::begin(SortedHomoglyphs),
1598 std::end(SortedHomoglyphs) - 1, HomoglyphPair{C, '\0'});
1599 if (Homoglyph->Character == C) {
1600 llvm::SmallString<5> CharBuf;
1601 {
1602 llvm::raw_svector_ostream CharOS(CharBuf);
1603 llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
1604 }
1605 if (Homoglyph->LooksLike) {
1606 const char LooksLikeStr[] = {Homoglyph->LooksLike, 0};
1607 Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_homoglyph)
1608 << Range << CharBuf << LooksLikeStr;
1609 } else {
1610 Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_zero_width)
1611 << Range << CharBuf;
1612 }
1613 }
1614}
1615
1616static void diagnoseInvalidUnicodeCodepointInIdentifier(
1617 DiagnosticsEngine &Diags, const LangOptions &LangOpts, uint32_t CodePoint,
1618 CharSourceRange Range, bool IsFirst) {
1619 if (isASCII(CodePoint))
1620 return;
1621
1622 bool IsIDStart = isAllowedInitiallyIDChar(CodePoint, LangOpts);
1623 bool IsIDContinue = IsIDStart || isAllowedIDChar(CodePoint, LangOpts);
1624
1625 if ((IsFirst && IsIDStart) || (!IsFirst && IsIDContinue))
1626 return;
1627
1628 bool InvalidOnlyAtStart = IsFirst && !IsIDStart && IsIDContinue;
1629
1630 llvm::SmallString<5> CharBuf;
1631 llvm::raw_svector_ostream CharOS(CharBuf);
1632 llvm::write_hex(CharOS, CodePoint, llvm::HexPrintStyle::Upper, 4);
1633
1634 if (!IsFirst || InvalidOnlyAtStart) {
1635 Diags.Report(Range.getBegin(), diag::err_character_not_allowed_identifier)
1636 << Range << CharBuf << int(InvalidOnlyAtStart)
1637 << FixItHint::CreateRemoval(Range);
1638 } else {
1639 Diags.Report(Range.getBegin(), diag::err_character_not_allowed)
1640 << Range << CharBuf << FixItHint::CreateRemoval(Range);
1641 }
1642}
1643
1644bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1645 Token &Result) {
1646 const char *UCNPtr = CurPtr + Size;
1647 uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1648 if (CodePoint == 0) {
1649 return false;
1650 }
1651
1652 if (!isAllowedIDChar(CodePoint, LangOpts)) {
1653 if (isASCII(CodePoint) || isUnicodeWhitespace(CodePoint))
1654 return false;
1655 if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
1656 !PP->isPreprocessedOutput())
1657 diagnoseInvalidUnicodeCodepointInIdentifier(
1658 PP->getDiagnostics(), LangOpts, CodePoint,
1659 makeCharRange(*this, CurPtr, UCNPtr),
1660 /*IsFirst=*/false);
1661
1662 // We got a unicode codepoint that is neither a space nor a
1663 // a valid identifier part.
1664 // Carry on as if the codepoint was valid for recovery purposes.
1665 } else if (!isLexingRawMode())
1666 maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1667 makeCharRange(*this, CurPtr, UCNPtr),
1668 /*IsFirst=*/false);
1669
1670 Result.setFlag(Token::HasUCN);
1671 if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1672 (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1673 CurPtr = UCNPtr;
1674 else
1675 while (CurPtr != UCNPtr)
1676 (void)getAndAdvanceChar(CurPtr, Result);
1677 return true;
1678}
1679
1680bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1681 const char *UnicodePtr = CurPtr;
1682 llvm::UTF32 CodePoint;
1683 llvm::ConversionResult Result =
1684 llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
1685 (const llvm::UTF8 *)BufferEnd,
1686 &CodePoint,
1687 llvm::strictConversion);
1688 if (Result != llvm::conversionOK)
1689 return false;
1690
1691 if (!isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts)) {
1692 if (isASCII(CodePoint) || isUnicodeWhitespace(CodePoint))
1693 return false;
1694
1695 if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
1696 !PP->isPreprocessedOutput())
1697 diagnoseInvalidUnicodeCodepointInIdentifier(
1698 PP->getDiagnostics(), LangOpts, CodePoint,
1699 makeCharRange(*this, CurPtr, UnicodePtr), /*IsFirst=*/false);
1700 // We got a unicode codepoint that is neither a space nor a
1701 // a valid identifier part. Carry on as if the codepoint was
1702 // valid for recovery purposes.
1703 } else if (!isLexingRawMode()) {
1704 maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1705 makeCharRange(*this, CurPtr, UnicodePtr),
1706 /*IsFirst=*/false);
1707 maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
1708 makeCharRange(*this, CurPtr, UnicodePtr));
1709 }
1710
1711 CurPtr = UnicodePtr;
1712 return true;
1713}
1714
1715bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1716 // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1717 unsigned Size;
1718 unsigned char C = *CurPtr++;
1719 while (isIdentifierBody(C))
1720 C = *CurPtr++;
1721
1722 --CurPtr; // Back up over the skipped character.
1723
1724 // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1725 // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1726 //
1727 // TODO: Could merge these checks into an InfoTable flag to make the
1728 // comparison cheaper
1729 if (isASCII(C) && C != '\\' && C != '?' &&
1730 (C != '$' || !LangOpts.DollarIdents)) {
1731FinishIdentifier:
1732 const char *IdStart = BufferPtr;
1733 FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1734 Result.setRawIdentifierData(IdStart);
1735
1736 // If we are in raw mode, return this identifier raw. There is no need to
1737 // look up identifier information or attempt to macro expand it.
1738 if (LexingRawMode)
1739 return true;
1740
1741 // Fill in Result.IdentifierInfo and update the token kind,
1742 // looking up the identifier in the identifier table.
1743 IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1744 // Note that we have to call PP->LookUpIdentifierInfo() even for code
1745 // completion, it writes IdentifierInfo into Result, and callers rely on it.
1746
1747 // If the completion point is at the end of an identifier, we want to treat
1748 // the identifier as incomplete even if it resolves to a macro or a keyword.
1749 // This allows e.g. 'class^' to complete to 'classifier'.
1750 if (isCodeCompletionPoint(CurPtr)) {
1751 // Return the code-completion token.
1752 Result.setKind(tok::code_completion);
1753 // Skip the code-completion char and all immediate identifier characters.
1754 // This ensures we get consistent behavior when completing at any point in
1755 // an identifier (i.e. at the start, in the middle, at the end). Note that
1756 // only simple cases (i.e. [a-zA-Z0-9_]) are supported to keep the code
1757 // simpler.
1758 assert(*CurPtr == 0 && "Completion character must be 0")(static_cast<void> (0));
1759 ++CurPtr;
1760 // Note that code completion token is not added as a separate character
1761 // when the completion point is at the end of the buffer. Therefore, we need
1762 // to check if the buffer has ended.
1763 if (CurPtr < BufferEnd) {
1764 while (isIdentifierBody(*CurPtr))
1765 ++CurPtr;
1766 }
1767 BufferPtr = CurPtr;
1768 return true;
1769 }
1770
1771 // Finally, now that we know we have an identifier, pass this off to the
1772 // preprocessor, which may macro expand it or something.
1773 if (II->isHandleIdentifierCase())
1774 return PP->HandleIdentifier(Result);
1775
1776 return true;
1777 }
1778
1779 // Otherwise, $,\,? in identifier found. Enter slower path.
1780
1781 C = getCharAndSize(CurPtr, Size);
1782 while (true) {
1783 if (C == '$') {
1784 // If we hit a $ and they are not supported in identifiers, we are done.
1785 if (!LangOpts.DollarIdents) goto FinishIdentifier;
1786
1787 // Otherwise, emit a diagnostic and continue.
1788 if (!isLexingRawMode())
1789 Diag(CurPtr, diag::ext_dollar_in_identifier);
1790 CurPtr = ConsumeChar(CurPtr, Size, Result);
1791 C = getCharAndSize(CurPtr, Size);
1792 continue;
1793 } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1794 C = getCharAndSize(CurPtr, Size);
1795 continue;
1796 } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1797 C = getCharAndSize(CurPtr, Size);
1798 continue;
1799 } else if (!isIdentifierBody(C)) {
1800 goto FinishIdentifier;
1801 }
1802
1803 // Otherwise, this character is good, consume it.
1804 CurPtr = ConsumeChar(CurPtr, Size, Result);
1805
1806 C = getCharAndSize(CurPtr, Size);
1807 while (isIdentifierBody(C)) {
1808 CurPtr = ConsumeChar(CurPtr, Size, Result);
1809 C = getCharAndSize(CurPtr, Size);
1810 }
1811 }
1812}
1813
1814/// isHexaLiteral - Return true if Start points to a hex constant.
1815/// in microsoft mode (where this is supposed to be several different tokens).
1816bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1817 unsigned Size;
1818 char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1819 if (C1 != '0')
1820 return false;
1821 char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1822 return (C2 == 'x' || C2 == 'X');
1823}
1824
1825/// LexNumericConstant - Lex the remainder of a integer or floating point
1826/// constant. From[-1] is the first character lexed. Return the end of the
1827/// constant.
1828bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1829 unsigned Size;
1830 char C = getCharAndSize(CurPtr, Size);
1831 char PrevCh = 0;
1832 while (isPreprocessingNumberBody(C)) {
1833 CurPtr = ConsumeChar(CurPtr, Size, Result);
1834 PrevCh = C;
1835 C = getCharAndSize(CurPtr, Size);
1836 }
1837
1838 // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1839 if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1840 // If we are in Microsoft mode, don't continue if the constant is hex.
1841 // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1842 if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1843 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1844 }
1845
1846 // If we have a hex FP constant, continue.
1847 if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1848 // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1849 // not-quite-conforming extension. Only do so if this looks like it's
1850 // actually meant to be a hexfloat, and not if it has a ud-suffix.
1851 bool IsHexFloat = true;
1852 if (!LangOpts.C99) {
1853 if (!isHexaLiteral(BufferPtr, LangOpts))
1854 IsHexFloat = false;
1855 else if (!getLangOpts().CPlusPlus17 &&
1856 std::find(BufferPtr, CurPtr, '_') != CurPtr)
1857 IsHexFloat = false;
1858 }
1859 if (IsHexFloat)
1860 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1861 }
1862
1863 // If we have a digit separator, continue.
1864 if (C == '\'' && (getLangOpts().CPlusPlus14 || getLangOpts().C2x)) {
1865 unsigned NextSize;
1866 char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1867 if (isIdentifierBody(Next)) {
1868 if (!isLexingRawMode())
1869 Diag(CurPtr, getLangOpts().CPlusPlus
1870 ? diag::warn_cxx11_compat_digit_separator
1871 : diag::warn_c2x_compat_digit_separator);
1872 CurPtr = ConsumeChar(CurPtr, Size, Result);
1873 CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1874 return LexNumericConstant(Result, CurPtr);
1875 }
1876 }
1877
1878 // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1879 if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1880 return LexNumericConstant(Result, CurPtr);
1881 if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1882 return LexNumericConstant(Result, CurPtr);
1883
1884 // Update the location of token as well as BufferPtr.
1885 const char *TokStart = BufferPtr;
1886 FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1887 Result.setLiteralData(TokStart);
1888 return true;
1889}
1890
1891/// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1892/// in C++11, or warn on a ud-suffix in C++98.
1893const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1894 bool IsStringLiteral) {
1895 assert(getLangOpts().CPlusPlus)(static_cast<void> (0));
1896
1897 // Maximally munch an identifier.
1898 unsigned Size;
1899 char C = getCharAndSize(CurPtr, Size);
1900 bool Consumed = false;
1901
1902 if (!isIdentifierHead(C)) {
1903 if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1904 Consumed = true;
1905 else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1906 Consumed = true;
1907 else
1908 return CurPtr;
1909 }
1910
1911 if (!getLangOpts().CPlusPlus11) {
1912 if (!isLexingRawMode())
1913 Diag(CurPtr,
1914 C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1915 : diag::warn_cxx11_compat_reserved_user_defined_literal)
1916 << FixItHint::CreateInsertion(getSourceLocation(CurPtr), " ");
1917 return CurPtr;
1918 }
1919
1920 // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1921 // that does not start with an underscore is ill-formed. As a conforming
1922 // extension, we treat all such suffixes as if they had whitespace before
1923 // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1924 // likely to be a ud-suffix than a macro, however, and accept that.
1925 if (!Consumed) {
1926 bool IsUDSuffix = false;
1927 if (C == '_')
1928 IsUDSuffix = true;
1929 else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1930 // In C++1y, we need to look ahead a few characters to see if this is a
1931 // valid suffix for a string literal or a numeric literal (this could be
1932 // the 'operator""if' defining a numeric literal operator).
1933 const unsigned MaxStandardSuffixLength = 3;
1934 char Buffer[MaxStandardSuffixLength] = { C };
1935 unsigned Consumed = Size;
1936 unsigned Chars = 1;
1937 while (true) {
1938 unsigned NextSize;
1939 char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1940 getLangOpts());
1941 if (!isIdentifierBody(Next)) {
1942 // End of suffix. Check whether this is on the allowed list.
1943 const StringRef CompleteSuffix(Buffer, Chars);
1944 IsUDSuffix = StringLiteralParser::isValidUDSuffix(getLangOpts(),
1945 CompleteSuffix);
1946 break;
1947 }
1948
1949 if (Chars == MaxStandardSuffixLength)
1950 // Too long: can't be a standard suffix.
1951 break;
1952
1953 Buffer[Chars++] = Next;
1954 Consumed += NextSize;
1955 }
1956 }
1957
1958 if (!IsUDSuffix) {
1959 if (!isLexingRawMode())
1960 Diag(CurPtr, getLangOpts().MSVCCompat
1961 ? diag::ext_ms_reserved_user_defined_literal
1962 : diag::ext_reserved_user_defined_literal)
1963 << FixItHint::CreateInsertion(getSourceLocation(CurPtr), " ");
1964 return CurPtr;
1965 }
1966
1967 CurPtr = ConsumeChar(CurPtr, Size, Result);
1968 }
1969
1970 Result.setFlag(Token::HasUDSuffix);
1971 while (true) {
1972 C = getCharAndSize(CurPtr, Size);
1973 if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1974 else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1975 else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1976 else break;
1977 }
1978
1979 return CurPtr;
1980}
1981
1982/// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1983/// either " or L" or u8" or u" or U".
1984bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1985 tok::TokenKind Kind) {
1986 const char *AfterQuote = CurPtr;
1987 // Does this string contain the \0 character?
1988 const char *NulCharacter = nullptr;
1989
1990 if (!isLexingRawMode() &&
1991 (Kind == tok::utf8_string_literal ||
1992 Kind == tok::utf16_string_literal ||
1993 Kind == tok::utf32_string_literal))
1994 Diag(BufferPtr, getLangOpts().CPlusPlus
1995 ? diag::warn_cxx98_compat_unicode_literal
1996 : diag::warn_c99_compat_unicode_literal);
1997
1998 char C = getAndAdvanceChar(CurPtr, Result);
1999 while (C != '"') {
2000 // Skip escaped characters. Escaped newlines will already be processed by
2001 // getAndAdvanceChar.
2002 if (C == '\\')
2003 C = getAndAdvanceChar(CurPtr, Result);
2004
2005 if (C == '\n' || C == '\r' || // Newline.
2006 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2007 if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2008 Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
2009 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2010 return true;
2011 }
2012
2013 if (C == 0) {
2014 if (isCodeCompletionPoint(CurPtr-1)) {
2015 if (ParsingFilename)
2016 codeCompleteIncludedFile(AfterQuote, CurPtr - 1, /*IsAngled=*/false);
2017 else
2018 PP->CodeCompleteNaturalLanguage();
2019 FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
2020 cutOffLexing();
2021 return true;
2022 }
2023
2024 NulCharacter = CurPtr-1;
2025 }
2026 C = getAndAdvanceChar(CurPtr, Result);
2027 }
2028
2029 // If we are in C++11, lex the optional ud-suffix.
2030 if (getLangOpts().CPlusPlus)
2031 CurPtr = LexUDSuffix(Result, CurPtr, true);
2032
2033 // If a nul character existed in the string, warn about it.
2034 if (NulCharacter && !isLexingRawMode())
2035 Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2036
2037 // Update the location of the token as well as the BufferPtr instance var.
2038 const char *TokStart = BufferPtr;
2039 FormTokenWithChars(Result, CurPtr, Kind);
2040 Result.setLiteralData(TokStart);
2041 return true;
2042}
2043
2044/// LexRawStringLiteral - Lex the remainder of a raw string literal, after
2045/// having lexed R", LR", u8R", uR", or UR".
2046bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
2047 tok::TokenKind Kind) {
2048 // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
2049 // Between the initial and final double quote characters of the raw string,
2050 // any transformations performed in phases 1 and 2 (trigraphs,
2051 // universal-character-names, and line splicing) are reverted.
2052
2053 if (!isLexingRawMode())
2054 Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
2055
2056 unsigned PrefixLen = 0;
2057
2058 while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
2059 ++PrefixLen;
2060
2061 // If the last character was not a '(', then we didn't lex a valid delimiter.
2062 if (CurPtr[PrefixLen] != '(') {
2063 if (!isLexingRawMode()) {
2064 const char *PrefixEnd = &CurPtr[PrefixLen];
2065 if (PrefixLen == 16) {
2066 Diag(PrefixEnd, diag::err_raw_delim_too_long);
2067 } else {
2068 Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
2069 << StringRef(PrefixEnd, 1);
2070 }
2071 }
2072
2073 // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
2074 // it's possible the '"' was intended to be part of the raw string, but
2075 // there's not much we can do about that.
2076 while (true) {
2077 char C = *CurPtr++;
2078
2079 if (C == '"')
2080 break;
2081 if (C == 0 && CurPtr-1 == BufferEnd) {
2082 --CurPtr;
2083 break;
2084 }
2085 }
2086
2087 FormTokenWithChars(Result, CurPtr, tok::unknown);
2088 return true;
2089 }
2090
2091 // Save prefix and move CurPtr past it
2092 const char *Prefix = CurPtr;
2093 CurPtr += PrefixLen + 1; // skip over prefix and '('
2094
2095 while (true) {
2096 char C = *CurPtr++;
2097
2098 if (C == ')') {
2099 // Check for prefix match and closing quote.
2100 if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
2101 CurPtr += PrefixLen + 1; // skip over prefix and '"'
2102 break;
2103 }
2104 } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
2105 if (!isLexingRawMode())
2106 Diag(BufferPtr, diag::err_unterminated_raw_string)
2107 << StringRef(Prefix, PrefixLen);
2108 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2109 return true;
2110 }
2111 }
2112
2113 // If we are in C++11, lex the optional ud-suffix.
2114 if (getLangOpts().CPlusPlus)
2115 CurPtr = LexUDSuffix(Result, CurPtr, true);
2116
2117 // Update the location of token as well as BufferPtr.
2118 const char *TokStart = BufferPtr;
2119 FormTokenWithChars(Result, CurPtr, Kind);
2120 Result.setLiteralData(TokStart);
2121 return true;
2122}
2123
2124/// LexAngledStringLiteral - Lex the remainder of an angled string literal,
2125/// after having lexed the '<' character. This is used for #include filenames.
2126bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
2127 // Does this string contain the \0 character?
2128 const char *NulCharacter = nullptr;
2129 const char *AfterLessPos = CurPtr;
2130 char C = getAndAdvanceChar(CurPtr, Result);
2131 while (C != '>') {
2132 // Skip escaped characters. Escaped newlines will already be processed by
2133 // getAndAdvanceChar.
2134 if (C == '\\')
2135 C = getAndAdvanceChar(CurPtr, Result);
2136
2137 if (isVerticalWhitespace(C) || // Newline.
2138 (C == 0 && (CurPtr - 1 == BufferEnd))) { // End of file.
2139 // If the filename is unterminated, then it must just be a lone <
2140 // character. Return this as such.
2141 FormTokenWithChars(Result, AfterLessPos, tok::less);
2142 return true;
2143 }
2144
2145 if (C == 0) {
2146 if (isCodeCompletionPoint(CurPtr - 1)) {
2147 codeCompleteIncludedFile(AfterLessPos, CurPtr - 1, /*IsAngled=*/true);
2148 cutOffLexing();
2149 FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
2150 return true;
2151 }
2152 NulCharacter = CurPtr-1;
2153 }
2154 C = getAndAdvanceChar(CurPtr, Result);
2155 }
2156
2157 // If a nul character existed in the string, warn about it.
2158 if (NulCharacter && !isLexingRawMode())
2159 Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2160
2161 // Update the location of token as well as BufferPtr.
2162 const char *TokStart = BufferPtr;
2163 FormTokenWithChars(Result, CurPtr, tok::header_name);
2164 Result.setLiteralData(TokStart);
2165 return true;
2166}
2167
2168void Lexer::codeCompleteIncludedFile(const char *PathStart,
2169 const char *CompletionPoint,
2170 bool IsAngled) {
2171 // Completion only applies to the filename, after the last slash.
2172 StringRef PartialPath(PathStart, CompletionPoint - PathStart);
2173 llvm::StringRef SlashChars = LangOpts.MSVCCompat ? "/\\" : "/";
2174 auto Slash = PartialPath.find_last_of(SlashChars);
2175 StringRef Dir =
2176 (Slash == StringRef::npos) ? "" : PartialPath.take_front(Slash);
2177 const char *StartOfFilename =
2178 (Slash == StringRef::npos) ? PathStart : PathStart + Slash + 1;
2179 // Code completion filter range is the filename only, up to completion point.
2180 PP->setCodeCompletionIdentifierInfo(&PP->getIdentifierTable().get(
2181 StringRef(StartOfFilename, CompletionPoint - StartOfFilename)));
2182 // We should replace the characters up to the closing quote or closest slash,
2183 // if any.
2184 while (CompletionPoint < BufferEnd) {
2185 char Next = *(CompletionPoint + 1);
2186 if (Next == 0 || Next == '\r' || Next == '\n')
2187 break;
2188 ++CompletionPoint;
2189 if (Next == (IsAngled ? '>' : '"'))
2190 break;
2191 if (llvm::is_contained(SlashChars, Next))
2192 break;
2193 }
2194
2195 PP->setCodeCompletionTokenRange(
2196 FileLoc.getLocWithOffset(StartOfFilename - BufferStart),
2197 FileLoc.getLocWithOffset(CompletionPoint - BufferStart));
2198 PP->CodeCompleteIncludedFile(Dir, IsAngled);
2199}
2200
2201/// LexCharConstant - Lex the remainder of a character constant, after having
2202/// lexed either ' or L' or u8' or u' or U'.
2203bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
2204 tok::TokenKind Kind) {
2205 // Does this character contain the \0 character?
2206 const char *NulCharacter = nullptr;
2207
2208 if (!isLexingRawMode()) {
2209 if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
2210 Diag(BufferPtr, getLangOpts().CPlusPlus
2211 ? diag::warn_cxx98_compat_unicode_literal
2212 : diag::warn_c99_compat_unicode_literal);
2213 else if (Kind == tok::utf8_char_constant)
2214 Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
2215 }
2216
2217 char C = getAndAdvanceChar(CurPtr, Result);
2218 if (C == '\'') {
2219 if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2220 Diag(BufferPtr, diag::ext_empty_character);
2221 FormTokenWithChars(Result, CurPtr, tok::unknown);
2222 return true;
2223 }
2224
2225 while (C != '\'') {
2226 // Skip escaped characters.
2227 if (C == '\\')
2228 C = getAndAdvanceChar(CurPtr, Result);
2229
2230 if (C == '\n' || C == '\r' || // Newline.
2231 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2232 if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2233 Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
2234 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2235 return true;
2236 }
2237
2238 if (C == 0) {
2239 if (isCodeCompletionPoint(CurPtr-1)) {
2240 PP->CodeCompleteNaturalLanguage();
2241 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2242 cutOffLexing();
2243 return true;
2244 }
2245
2246 NulCharacter = CurPtr-1;
2247 }
2248 C = getAndAdvanceChar(CurPtr, Result);
2249 }
2250
2251 // If we are in C++11, lex the optional ud-suffix.
2252 if (getLangOpts().CPlusPlus)
2253 CurPtr = LexUDSuffix(Result, CurPtr, false);
2254
2255 // If a nul character existed in the character, warn about it.
2256 if (NulCharacter && !isLexingRawMode())
2257 Diag(NulCharacter, diag::null_in_char_or_string) << 0;
2258
2259 // Update the location of token as well as BufferPtr.
2260 const char *TokStart = BufferPtr;
2261 FormTokenWithChars(Result, CurPtr, Kind);
2262 Result.setLiteralData(TokStart);
2263 return true;
2264}
2265
2266/// SkipWhitespace - Efficiently skip over a series of whitespace characters.
2267/// Update BufferPtr to point to the next non-whitespace character and return.
2268///
2269/// This method forms a token and returns true if KeepWhitespaceMode is enabled.
2270bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
2271 bool &TokAtPhysicalStartOfLine) {
2272 // Whitespace - Skip it, then return the token after the whitespace.
2273 bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
2274
2275 unsigned char Char = *CurPtr;
2276
2277 const char *lastNewLine = nullptr;
2278 auto setLastNewLine = [&](const char *Ptr) {
2279 lastNewLine = Ptr;
2280 if (!NewLinePtr)
2281 NewLinePtr = Ptr;
2282 };
2283 if (SawNewline)
2284 setLastNewLine(CurPtr - 1);
2285
2286 // Skip consecutive spaces efficiently.
2287 while (true) {
2288 // Skip horizontal whitespace very aggressively.
2289 while (isHorizontalWhitespace(Char))
2290 Char = *++CurPtr;
2291
2292 // Otherwise if we have something other than whitespace, we're done.
2293 if (!isVerticalWhitespace(Char))
2294 break;
2295
2296 if (ParsingPreprocessorDirective) {
2297 // End of preprocessor directive line, let LexTokenInternal handle this.
2298 BufferPtr = CurPtr;
2299 return false;
2300 }
2301
2302 // OK, but handle newline.
2303 if (*CurPtr == '\n')
2304 setLastNewLine(CurPtr);
2305 SawNewline = true;
2306 Char = *++CurPtr;
2307 }
2308
2309 // If the client wants us to return whitespace, return it now.
2310 if (isKeepWhitespaceMode()) {
2311 FormTokenWithChars(Result, CurPtr, tok::unknown);
2312 if (SawNewline) {
2313 IsAtStartOfLine = true;
2314 IsAtPhysicalStartOfLine = true;
2315 }
2316 // FIXME: The next token will not have LeadingSpace set.
2317 return true;
2318 }
2319
2320 // If this isn't immediately after a newline, there is leading space.
2321 char PrevChar = CurPtr[-1];
2322 bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2323
2324 Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2325 if (SawNewline) {
2326 Result.setFlag(Token::StartOfLine);
2327 TokAtPhysicalStartOfLine = true;
2328
2329 if (NewLinePtr && lastNewLine && NewLinePtr != lastNewLine && PP) {
2330 if (auto *Handler = PP->getEmptylineHandler())
2331 Handler->HandleEmptyline(SourceRange(getSourceLocation(NewLinePtr + 1),
2332 getSourceLocation(lastNewLine)));
2333 }
2334 }
2335
2336 BufferPtr = CurPtr;
2337 return false;
2338}
2339
2340/// We have just read the // characters from input. Skip until we find the
2341/// newline character that terminates the comment. Then update BufferPtr and
2342/// return.
2343///
2344/// If we're in KeepCommentMode or any CommentHandler has inserted
2345/// some tokens, this will store the first token and return true.
2346bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2347 bool &TokAtPhysicalStartOfLine) {
2348 // If Line comments aren't explicitly enabled for this language, emit an
2349 // extension warning.
2350 if (!LangOpts.LineComment && !isLexingRawMode()) {
2351 Diag(BufferPtr, diag::ext_line_comment);
2352
2353 // Mark them enabled so we only emit one warning for this translation
2354 // unit.
2355 LangOpts.LineComment = true;
2356 }
2357
2358 // Scan over the body of the comment. The common case, when scanning, is that
2359 // the comment contains normal ascii characters with nothing interesting in
2360 // them. As such, optimize for this case with the inner loop.
2361 //
2362 // This loop terminates with CurPtr pointing at the newline (or end of buffer)
2363 // character that ends the line comment.
2364 char C;
2365 while (true) {
2366 C = *CurPtr;
2367 // Skip over characters in the fast loop.
2368 while (C != 0 && // Potentially EOF.
2369 C != '\n' && C != '\r') // Newline or DOS-style newline.
2370 C = *++CurPtr;
2371
2372 const char *NextLine = CurPtr;
2373 if (C != 0) {
2374 // We found a newline, see if it's escaped.
2375 const char *EscapePtr = CurPtr-1;
2376 bool HasSpace = false;
2377 while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2378 --EscapePtr;
2379 HasSpace = true;
2380 }
2381
2382 if (*EscapePtr == '\\')
2383 // Escaped newline.
2384 CurPtr = EscapePtr;
2385 else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2386 EscapePtr[-2] == '?' && LangOpts.Trigraphs)
2387 // Trigraph-escaped newline.
2388 CurPtr = EscapePtr-2;
2389 else
2390 break; // This is a newline, we're done.
2391
2392 // If there was space between the backslash and newline, warn about it.
2393 if (HasSpace && !isLexingRawMode())
2394 Diag(EscapePtr, diag::backslash_newline_space);
2395 }
2396
2397 // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2398 // properly decode the character. Read it in raw mode to avoid emitting
2399 // diagnostics about things like trigraphs. If we see an escaped newline,
2400 // we'll handle it below.
2401 const char *OldPtr = CurPtr;
2402 bool OldRawMode = isLexingRawMode();
2403 LexingRawMode = true;
2404 C = getAndAdvanceChar(CurPtr, Result);
2405 LexingRawMode = OldRawMode;
2406
2407 // If we only read only one character, then no special handling is needed.
2408 // We're done and can skip forward to the newline.
2409 if (C != 0 && CurPtr == OldPtr+1) {
2410 CurPtr = NextLine;
2411 break;
2412 }
2413
2414 // If we read multiple characters, and one of those characters was a \r or
2415 // \n, then we had an escaped newline within the comment. Emit diagnostic
2416 // unless the next line is also a // comment.
2417 if (CurPtr != OldPtr + 1 && C != '/' &&
2418 (CurPtr == BufferEnd + 1 || CurPtr[0] != '/')) {
2419 for (; OldPtr != CurPtr; ++OldPtr)
2420 if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2421 // Okay, we found a // comment that ends in a newline, if the next
2422 // line is also a // comment, but has spaces, don't emit a diagnostic.
2423 if (isWhitespace(C)) {
2424 const char *ForwardPtr = CurPtr;
2425 while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2426 ++ForwardPtr;
2427 if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2428 break;
2429 }
2430
2431 if (!isLexingRawMode())
2432 Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2433 break;
2434 }
2435 }
2436
2437 if (C == '\r' || C == '\n' || CurPtr == BufferEnd + 1) {
2438 --CurPtr;
2439 break;
2440 }
2441
2442 if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2443 PP->CodeCompleteNaturalLanguage();
2444 cutOffLexing();
2445 return false;
2446 }
2447 }
2448
2449 // Found but did not consume the newline. Notify comment handlers about the
2450 // comment unless we're in a #if 0 block.
2451 if (PP && !isLexingRawMode() &&
2452 PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2453 getSourceLocation(CurPtr)))) {
2454 BufferPtr = CurPtr;
2455 return true; // A token has to be returned.
2456 }
2457
2458 // If we are returning comments as tokens, return this comment as a token.
2459 if (inKeepCommentMode())
2460 return SaveLineComment(Result, CurPtr);
2461
2462 // If we are inside a preprocessor directive and we see the end of line,
2463 // return immediately, so that the lexer can return this as an EOD token.
2464 if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2465 BufferPtr = CurPtr;
2466 return false;
2467 }
2468
2469 // Otherwise, eat the \n character. We don't care if this is a \n\r or
2470 // \r\n sequence. This is an efficiency hack (because we know the \n can't
2471 // contribute to another token), it isn't needed for correctness. Note that
2472 // this is ok even in KeepWhitespaceMode, because we would have returned the
2473 /// comment above in that mode.
2474 NewLinePtr = CurPtr++;
2475
2476 // The next returned token is at the start of the line.
2477 Result.setFlag(Token::StartOfLine);
2478 TokAtPhysicalStartOfLine = true;
2479 // No leading whitespace seen so far.
2480 Result.clearFlag(Token::LeadingSpace);
2481 BufferPtr = CurPtr;
2482 return false;
2483}
2484
2485/// If in save-comment mode, package up this Line comment in an appropriate
2486/// way and return it.
2487bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2488 // If we're not in a preprocessor directive, just return the // comment
2489 // directly.
2490 FormTokenWithChars(Result, CurPtr, tok::comment);
2491
2492 if (!ParsingPreprocessorDirective || LexingRawMode)
2493 return true;
2494
2495 // If this Line-style comment is in a macro definition, transmogrify it into
2496 // a C-style block comment.
2497 bool Invalid = false;
2498 std::string Spelling = PP->getSpelling(Result, &Invalid);
2499 if (Invalid)
2500 return true;
2501
2502 assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?")(static_cast<void> (0));
2503 Spelling[1] = '*'; // Change prefix to "/*".
2504 Spelling += "*/"; // add suffix.
2505
2506 Result.setKind(tok::comment);
2507 PP->CreateString(Spelling, Result,
2508 Result.getLocation(), Result.getLocation());
2509 return true;
2510}
2511
2512/// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2513/// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2514/// a diagnostic if so. We know that the newline is inside of a block comment.
2515static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2516 Lexer *L) {
2517 assert(CurPtr[0] == '\n' || CurPtr[0] == '\r')(static_cast<void> (0));
2518
2519 // Position of the first trigraph in the ending sequence.
2520 const char *TrigraphPos = 0;
2521 // Position of the first whitespace after a '\' in the ending sequence.
2522 const char *SpacePos = 0;
2523
2524 while (true) {
2525 // Back up off the newline.
2526 --CurPtr;
2527
2528 // If this is a two-character newline sequence, skip the other character.
2529 if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2530 // \n\n or \r\r -> not escaped newline.
2531 if (CurPtr[0] == CurPtr[1])
2532 return false;
2533 // \n\r or \r\n -> skip the newline.
2534 --CurPtr;
2535 }
2536
2537 // If we have horizontal whitespace, skip over it. We allow whitespace
2538 // between the slash and newline.
2539 while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2540 SpacePos = CurPtr;
2541 --CurPtr;
2542 }
2543
2544 // If we have a slash, this is an escaped newline.
2545 if (*CurPtr == '\\') {
2546 --CurPtr;
2547 } else if (CurPtr[0] == '/' && CurPtr[-1] == '?' && CurPtr[-2] == '?') {
2548 // This is a trigraph encoding of a slash.
2549 TrigraphPos = CurPtr - 2;
2550 CurPtr -= 3;
2551 } else {
2552 return false;
2553 }
2554
2555 // If the character preceding the escaped newline is a '*', then after line
2556 // splicing we have a '*/' ending the comment.
2557 if (*CurPtr == '*')
2558 break;
2559
2560 if (*CurPtr != '\n' && *CurPtr != '\r')
2561 return false;
2562 }
2563
2564 if (TrigraphPos) {
2565 // If no trigraphs are enabled, warn that we ignored this trigraph and
2566 // ignore this * character.
2567 if (!L->getLangOpts().Trigraphs) {
2568 if (!L->isLexingRawMode())
2569 L->Diag(TrigraphPos, diag::trigraph_ignored_block_comment);
2570 return false;
2571 }
2572 if (!L->isLexingRawMode())
2573 L->Diag(TrigraphPos, diag::trigraph_ends_block_comment);
2574 }
2575
2576 // Warn about having an escaped newline between the */ characters.
2577 if (!L->isLexingRawMode())
2578 L->Diag(CurPtr + 1, diag::escaped_newline_block_comment_end);
2579
2580 // If there was space between the backslash and newline, warn about it.
2581 if (SpacePos && !L->isLexingRawMode())
2582 L->Diag(SpacePos, diag::backslash_newline_space);
2583
2584 return true;
2585}
2586
2587#ifdef __SSE2__1
2588#include <emmintrin.h>
2589#elif __ALTIVEC__
2590#include <altivec.h>
2591#undef bool
2592#endif
2593
2594/// We have just read from input the / and * characters that started a comment.
2595/// Read until we find the * and / characters that terminate the comment.
2596/// Note that we don't bother decoding trigraphs or escaped newlines in block
2597/// comments, because they cannot cause the comment to end. The only thing
2598/// that can happen is the comment could end with an escaped newline between
2599/// the terminating * and /.
2600///
2601/// If we're in KeepCommentMode or any CommentHandler has inserted
2602/// some tokens, this will store the first token and return true.
2603bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2604 bool &TokAtPhysicalStartOfLine) {
2605 // Scan one character past where we should, looking for a '/' character. Once
2606 // we find it, check to see if it was preceded by a *. This common
2607 // optimization helps people who like to put a lot of * characters in their
2608 // comments.
2609
2610 // The first character we get with newlines and trigraphs skipped to handle
2611 // the degenerate /*/ case below correctly if the * has an escaped newline
2612 // after it.
2613 unsigned CharSize;
2614 unsigned char C = getCharAndSize(CurPtr, CharSize);
12
Calling 'Lexer::getCharAndSize'
15
Returning from 'Lexer::getCharAndSize'
2615 CurPtr += CharSize;
2616 if (C == 0 && CurPtr == BufferEnd+1) {
16
Assuming 'C' is not equal to 0
2617 if (!isLexingRawMode())
2618 Diag(BufferPtr, diag::err_unterminated_block_comment);
2619 --CurPtr;
2620
2621 // KeepWhitespaceMode should return this broken comment as a token. Since
2622 // it isn't a well formed comment, just return it as an 'unknown' token.
2623 if (isKeepWhitespaceMode()) {
2624 FormTokenWithChars(Result, CurPtr, tok::unknown);
2625 return true;
2626 }
2627
2628 BufferPtr = CurPtr;
2629 return false;
2630 }
2631
2632 // Check to see if the first character after the '/*' is another /. If so,
2633 // then this slash does not end the block comment, it is part of it.
2634 if (C == '/')
17
Assuming the condition is false
18
Taking false branch
2635 C = *CurPtr++;
2636
2637 while (true) {
2638 // Skip over all non-interesting characters until we find end of buffer or a
2639 // (probably ending) '/' character.
2640 if (CurPtr + 24 < BufferEnd &&
19
Assuming the condition is true
2641 // If there is a code-completion point avoid the fast scan because it
2642 // doesn't check for '\0'.
2643 !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
20
Assuming field 'PP' is null
2644 // While not aligned to a 16-byte boundary.
2645 while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
21
Loop condition is false. Execution continues on line 2648
2646 C = *CurPtr++;
2647
2648 if (C == '/') goto FoundSlash;
22
Taking false branch
2649
2650#ifdef __SSE2__1
2651 __m128i Slashes = _mm_set1_epi8('/');
2652 while (CurPtr+16 <= BufferEnd) {
23
Loop condition is true. Entering loop body
2653 int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2654 Slashes));
2655 if (cmp != 0) {
24
Assuming 'cmp' is not equal to 0
25
Taking true branch
2656 // Adjust the pointer to point directly after the first slash. It's
2657 // not necessary to set C here, it will be overwritten at the end of
2658 // the outer loop.
2659 CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2660 goto FoundSlash;
26
Control jumps to line 2693
2661 }
2662 CurPtr += 16;
2663 }
2664#elif __ALTIVEC__
2665 __vector unsigned char Slashes = {
2666 '/', '/', '/', '/', '/', '/', '/', '/',
2667 '/', '/', '/', '/', '/', '/', '/', '/'
2668 };
2669 while (CurPtr + 16 <= BufferEnd &&
2670 !vec_any_eq(*(const __vector unsigned char *)CurPtr, Slashes))
2671 CurPtr += 16;
2672#else
2673 // Scan for '/' quickly. Many block comments are very large.
2674 while (CurPtr[0] != '/' &&
2675 CurPtr[1] != '/' &&
2676 CurPtr[2] != '/' &&
2677 CurPtr[3] != '/' &&
2678 CurPtr+4 < BufferEnd) {
2679 CurPtr += 4;
2680 }
2681#endif
2682
2683 // It has to be one of the bytes scanned, increment to it and read one.
2684 C = *CurPtr++;
2685 }
2686
2687 // Loop to scan the remainder.
2688 while (C != '/' && C != '\0')
2689 C = *CurPtr++;
2690
2691 if (C == '/') {
2692 FoundSlash:
2693 if (CurPtr[-2] == '*') // We found the final */. We're done!
27
Assuming the condition is false
2694 break;
2695
2696 if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
28
Assuming the condition is false
29
Assuming the condition is false
2697 if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2698 // We found the final */, though it had an escaped newline between the
2699 // * and /. We're done!
2700 break;
2701 }
2702 }
2703 if (CurPtr[0] == '*' && CurPtr[1] != '/') {
30
Assuming the condition is true
31
Assuming the condition is true
32
Taking true branch
2704 // If this is a /* inside of the comment, emit a warning. Don't do this
2705 // if this is a /*/, which will end the comment. This misses cases with
2706 // embedded escaped newlines, but oh well.
2707 if (!isLexingRawMode())
33
Assuming the condition is true
34
Taking true branch
2708 Diag(CurPtr-1, diag::warn_nested_block_comment);
35
Calling 'Lexer::Diag'
2709 }
2710 } else if (C == 0 && CurPtr == BufferEnd+1) {
2711 if (!isLexingRawMode())
2712 Diag(BufferPtr, diag::err_unterminated_block_comment);
2713 // Note: the user probably forgot a */. We could continue immediately
2714 // after the /*, but this would involve lexing a lot of what really is the
2715 // comment, which surely would confuse the parser.
2716 --CurPtr;
2717
2718 // KeepWhitespaceMode should return this broken comment as a token. Since
2719 // it isn't a well formed comment, just return it as an 'unknown' token.
2720 if (isKeepWhitespaceMode()) {
2721 FormTokenWithChars(Result, CurPtr, tok::unknown);
2722 return true;
2723 }
2724
2725 BufferPtr = CurPtr;
2726 return false;
2727 } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2728 PP->CodeCompleteNaturalLanguage();
2729 cutOffLexing();
2730 return false;
2731 }
2732
2733 C = *CurPtr++;
2734 }
2735
2736 // Notify comment handlers about the comment unless we're in a #if 0 block.
2737 if (PP && !isLexingRawMode() &&
2738 PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2739 getSourceLocation(CurPtr)))) {
2740 BufferPtr = CurPtr;
2741 return true; // A token has to be returned.
2742 }
2743
2744 // If we are returning comments as tokens, return this comment as a token.
2745 if (inKeepCommentMode()) {
2746 FormTokenWithChars(Result, CurPtr, tok::comment);
2747 return true;
2748 }
2749
2750 // It is common for the tokens immediately after a /**/ comment to be
2751 // whitespace. Instead of going through the big switch, handle it
2752 // efficiently now. This is safe even in KeepWhitespaceMode because we would
2753 // have already returned above with the comment as a token.
2754 if (isHorizontalWhitespace(*CurPtr)) {
2755 SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2756 return false;
2757 }
2758
2759 // Otherwise, just return so that the next character will be lexed as a token.
2760 BufferPtr = CurPtr;
2761 Result.setFlag(Token::LeadingSpace);
2762 return false;
2763}
2764
2765//===----------------------------------------------------------------------===//
2766// Primary Lexing Entry Points
2767//===----------------------------------------------------------------------===//
2768
2769/// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2770/// uninterpreted string. This switches the lexer out of directive mode.
2771void Lexer::ReadToEndOfLine(SmallVectorImpl<char> *Result) {
2772 assert(ParsingPreprocessorDirective && ParsingFilename == false &&(static_cast<void> (0))
2773 "Must be in a preprocessing directive!")(static_cast<void> (0));
2774 Token Tmp;
2775 Tmp.startToken();
2776
2777 // CurPtr - Cache BufferPtr in an automatic variable.
2778 const char *CurPtr = BufferPtr;
2779 while (true) {
2780 char Char = getAndAdvanceChar(CurPtr, Tmp);
2781 switch (Char) {
2782 default:
2783 if (Result)
2784 Result->push_back(Char);
2785 break;
2786 case 0: // Null.
2787 // Found end of file?
2788 if (CurPtr-1 != BufferEnd) {
2789 if (isCodeCompletionPoint(CurPtr-1)) {
2790 PP->CodeCompleteNaturalLanguage();
2791 cutOffLexing();
2792 return;
2793 }
2794
2795 // Nope, normal character, continue.
2796 if (Result)
2797 Result->push_back(Char);
2798 break;
2799 }
2800 // FALL THROUGH.
2801 LLVM_FALLTHROUGH[[gnu::fallthrough]];
2802 case '\r':
2803 case '\n':
2804 // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2805 assert(CurPtr[-1] == Char && "Trigraphs for newline?")(static_cast<void> (0));
2806 BufferPtr = CurPtr-1;
2807
2808 // Next, lex the character, which should handle the EOD transition.
2809 Lex(Tmp);
2810 if (Tmp.is(tok::code_completion)) {
2811 if (PP)
2812 PP->CodeCompleteNaturalLanguage();
2813 Lex(Tmp);
2814 }
2815 assert(Tmp.is(tok::eod) && "Unexpected token!")(static_cast<void> (0));
2816
2817 // Finally, we're done;
2818 return;
2819 }
2820 }
2821}
2822
2823/// LexEndOfFile - CurPtr points to the end of this file. Handle this
2824/// condition, reporting diagnostics and handling other edge cases as required.
2825/// This returns true if Result contains a token, false if PP.Lex should be
2826/// called again.
2827bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2828 // If we hit the end of the file while parsing a preprocessor directive,
2829 // end the preprocessor directive first. The next token returned will
2830 // then be the end of file.
2831 if (ParsingPreprocessorDirective) {
2832 // Done parsing the "line".
2833 ParsingPreprocessorDirective = false;
2834 // Update the location of token as well as BufferPtr.
2835 FormTokenWithChars(Result, CurPtr, tok::eod);
2836
2837 // Restore comment saving mode, in case it was disabled for directive.
2838 if (PP)
2839 resetExtendedTokenMode();
2840 return true; // Have a token.
2841 }
2842
2843 // If we are in raw mode, return this event as an EOF token. Let the caller
2844 // that put us in raw mode handle the event.
2845 if (isLexingRawMode()) {
2846 Result.startToken();
2847 BufferPtr = BufferEnd;
2848 FormTokenWithChars(Result, BufferEnd, tok::eof);
2849 return true;
2850 }
2851
2852 if (PP->isRecordingPreamble() && PP->isInPrimaryFile()) {
2853 PP->setRecordedPreambleConditionalStack(ConditionalStack);
2854 // If the preamble cuts off the end of a header guard, consider it guarded.
2855 // The guard is valid for the preamble content itself, and for tools the
2856 // most useful answer is "yes, this file has a header guard".
2857 if (!ConditionalStack.empty())
2858 MIOpt.ExitTopLevelConditional();
2859 ConditionalStack.clear();
2860 }
2861
2862 // Issue diagnostics for unterminated #if and missing newline.
2863
2864 // If we are in a #if directive, emit an error.
2865 while (!ConditionalStack.empty()) {
2866 if (PP->getCodeCompletionFileLoc() != FileLoc)
2867 PP->Diag(ConditionalStack.back().IfLoc,
2868 diag::err_pp_unterminated_conditional);
2869 ConditionalStack.pop_back();
2870 }
2871
2872 SourceLocation EndLoc = getSourceLocation(BufferEnd);
2873 // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2874 // a pedwarn.
2875 if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2876 DiagnosticsEngine &Diags = PP->getDiagnostics();
2877 unsigned DiagID;
2878
2879 if (LangOpts.CPlusPlus11) {
2880 // C++11 [lex.phases] 2.2 p2
2881 // Prefer the C++98 pedantic compatibility warning over the generic,
2882 // non-extension, user-requested "missing newline at EOF" warning.
2883 if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2884 DiagID = diag::warn_cxx98_compat_no_newline_eof;
2885 } else {
2886 DiagID = diag::warn_no_newline_eof;
2887 }
2888 } else {
2889 DiagID = diag::ext_no_newline_eof;
2890 }
2891
2892 Diag(BufferEnd, DiagID)
2893 << FixItHint::CreateInsertion(EndLoc, "\n");
2894 }
2895
2896 BufferPtr = CurPtr;
2897
2898 // Finally, let the preprocessor handle this.
2899 return PP->HandleEndOfFile(Result, EndLoc, isPragmaLexer());
2900}
2901
2902/// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2903/// the specified lexer will return a tok::l_paren token, 0 if it is something
2904/// else and 2 if there are no more tokens in the buffer controlled by the
2905/// lexer.
2906unsigned Lexer::isNextPPTokenLParen() {
2907 assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?")(static_cast<void> (0));
2908
2909 // Switch to 'skipping' mode. This will ensure that we can lex a token
2910 // without emitting diagnostics, disables macro expansion, and will cause EOF
2911 // to return an EOF token instead of popping the include stack.
2912 LexingRawMode = true;
2913
2914 // Save state that can be changed while lexing so that we can restore it.
2915 const char *TmpBufferPtr = BufferPtr;
2916 bool inPPDirectiveMode = ParsingPreprocessorDirective;
2917 bool atStartOfLine = IsAtStartOfLine;
2918 bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2919 bool leadingSpace = HasLeadingSpace;
2920
2921 Token Tok;
2922 Lex(Tok);
2923
2924 // Restore state that may have changed.
2925 BufferPtr = TmpBufferPtr;
2926 ParsingPreprocessorDirective = inPPDirectiveMode;
2927 HasLeadingSpace = leadingSpace;
2928 IsAtStartOfLine = atStartOfLine;
2929 IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2930
2931 // Restore the lexer back to non-skipping mode.
2932 LexingRawMode = false;
2933
2934 if (Tok.is(tok::eof))
2935 return 2;
2936 return Tok.is(tok::l_paren);
2937}
2938
2939/// Find the end of a version control conflict marker.
2940static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2941 ConflictMarkerKind CMK) {
2942 const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2943 size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2944 auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2945 size_t Pos = RestOfBuffer.find(Terminator);
2946 while (Pos != StringRef::npos) {
2947 // Must occur at start of line.
2948 if (Pos == 0 ||
2949 (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2950 RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2951 Pos = RestOfBuffer.find(Terminator);
2952 continue;
2953 }
2954 return RestOfBuffer.data()+Pos;
2955 }
2956 return nullptr;
2957}
2958
2959/// IsStartOfConflictMarker - If the specified pointer is the start of a version
2960/// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2961/// and recover nicely. This returns true if it is a conflict marker and false
2962/// if not.
2963bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2964 // Only a conflict marker if it starts at the beginning of a line.
2965 if (CurPtr != BufferStart &&
2966 CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2967 return false;
2968
2969 // Check to see if we have <<<<<<< or >>>>.
2970 if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
2971 !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
2972 return false;
2973
2974 // If we have a situation where we don't care about conflict markers, ignore
2975 // it.
2976 if (CurrentConflictMarkerState || isLexingRawMode())
2977 return false;
2978
2979 ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2980
2981 // Check to see if there is an ending marker somewhere in the buffer at the
2982 // start of a line to terminate this conflict marker.
2983 if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2984 // We found a match. We are really in a conflict marker.
2985 // Diagnose this, and ignore to the end of line.
2986 Diag(CurPtr, diag::err_conflict_marker);
2987 CurrentConflictMarkerState = Kind;
2988
2989 // Skip ahead to the end of line. We know this exists because the
2990 // end-of-conflict marker starts with \r or \n.
2991 while (*CurPtr != '\r' && *CurPtr != '\n') {
2992 assert(CurPtr != BufferEnd && "Didn't find end of line")(static_cast<void> (0));
2993 ++CurPtr;
2994 }
2995 BufferPtr = CurPtr;
2996 return true;
2997 }
2998
2999 // No end of conflict marker found.
3000 return false;
3001}
3002
3003/// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
3004/// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
3005/// is the end of a conflict marker. Handle it by ignoring up until the end of
3006/// the line. This returns true if it is a conflict marker and false if not.
3007bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
3008 // Only a conflict marker if it starts at the beginning of a line.
3009 if (CurPtr != BufferStart &&
3010 CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
3011 return false;
3012
3013 // If we have a situation where we don't care about conflict markers, ignore
3014 // it.
3015 if (!CurrentConflictMarkerState || isLexingRawMode())
3016 return false;
3017
3018 // Check to see if we have the marker (4 characters in a row).
3019 for (unsigned i = 1; i != 4; ++i)
3020 if (CurPtr[i] != CurPtr[0])
3021 return false;
3022
3023 // If we do have it, search for the end of the conflict marker. This could
3024 // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
3025 // be the end of conflict marker.
3026 if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
3027 CurrentConflictMarkerState)) {
3028 CurPtr = End;
3029
3030 // Skip ahead to the end of line.
3031 while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
3032 ++CurPtr;
3033
3034 BufferPtr = CurPtr;
3035
3036 // No longer in the conflict marker.
3037 CurrentConflictMarkerState = CMK_None;
3038 return true;
3039 }
3040
3041 return false;
3042}
3043
3044static const char *findPlaceholderEnd(const char *CurPtr,
3045 const char *BufferEnd) {
3046 if (CurPtr == BufferEnd)
3047 return nullptr;
3048 BufferEnd -= 1; // Scan until the second last character.
3049 for (; CurPtr != BufferEnd; ++CurPtr) {
3050 if (CurPtr[0] == '#' && CurPtr[1] == '>')
3051 return CurPtr + 2;
3052 }
3053 return nullptr;
3054}
3055
3056bool Lexer::lexEditorPlaceholder(Token &Result, const char *CurPtr) {
3057 assert(CurPtr[-1] == '<' && CurPtr[0] == '#' && "Not a placeholder!")(static_cast<void> (0));
3058 if (!PP || !PP->getPreprocessorOpts().LexEditorPlaceholders || LexingRawMode)
3059 return false;
3060 const char *End = findPlaceholderEnd(CurPtr + 1, BufferEnd);
3061 if (!End)
3062 return false;
3063 const char *Start = CurPtr - 1;
3064 if (!LangOpts.AllowEditorPlaceholders)
3065 Diag(Start, diag::err_placeholder_in_source);
3066 Result.startToken();
3067 FormTokenWithChars(Result, End, tok::raw_identifier);
3068 Result.setRawIdentifierData(Start);
3069 PP->LookUpIdentifierInfo(Result);
3070 Result.setFlag(Token::IsEditorPlaceholder);
3071 BufferPtr = End;
3072 return true;
3073}
3074
3075bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
3076 if (PP && PP->isCodeCompletionEnabled()) {
3077 SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
3078 return Loc == PP->getCodeCompletionLoc();
3079 }
3080
3081 return false;
3082}
3083
3084uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
3085 Token *Result) {
3086 unsigned CharSize;
3087 char Kind = getCharAndSize(StartPtr, CharSize);
3088
3089 unsigned NumHexDigits;
3090 if (Kind == 'u')
3091 NumHexDigits = 4;
3092 else if (Kind == 'U')
3093 NumHexDigits = 8;
3094 else
3095 return 0;
3096
3097 if (!LangOpts.CPlusPlus && !LangOpts.C99) {
3098 if (Result && !isLexingRawMode())
3099 Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
3100 return 0;
3101 }
3102
3103 const char *CurPtr = StartPtr + CharSize;
3104 const char *KindLoc = &CurPtr[-1];
3105
3106 uint32_t CodePoint = 0;
3107 for (unsigned i = 0; i < NumHexDigits; ++i) {
3108 char C = getCharAndSize(CurPtr, CharSize);
3109
3110 unsigned Value = llvm::hexDigitValue(C);
3111 if (Value == -1U) {
3112 if (Result && !isLexingRawMode()) {
3113 if (i == 0) {
3114 Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
3115 << StringRef(KindLoc, 1);
3116 } else {
3117 Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
3118
3119 // If the user wrote \U1234, suggest a fixit to \u.
3120 if (i == 4 && NumHexDigits == 8) {
3121 CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
3122 Diag(KindLoc, diag::note_ucn_four_not_eight)
3123 << FixItHint::CreateReplacement(URange, "u");
3124 }
3125 }
3126 }
3127
3128 return 0;
3129 }
3130
3131 CodePoint <<= 4;
3132 CodePoint += Value;
3133
3134 CurPtr += CharSize;
3135 }
3136
3137 if (Result) {
3138 Result->setFlag(Token::HasUCN);
3139 if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
3140 StartPtr = CurPtr;
3141 else
3142 while (StartPtr != CurPtr)
3143 (void)getAndAdvanceChar(StartPtr, *Result);
3144 } else {
3145 StartPtr = CurPtr;
3146 }
3147
3148 // Don't apply C family restrictions to UCNs in assembly mode
3149 if (LangOpts.AsmPreprocessor)
3150 return CodePoint;
3151
3152 // C99 6.4.3p2: A universal character name shall not specify a character whose
3153 // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
3154 // 0060 (`), nor one in the range D800 through DFFF inclusive.)
3155 // C++11 [lex.charset]p2: If the hexadecimal value for a
3156 // universal-character-name corresponds to a surrogate code point (in the
3157 // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
3158 // if the hexadecimal value for a universal-character-name outside the
3159 // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
3160 // string literal corresponds to a control character (in either of the
3161 // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
3162 // basic source character set, the program is ill-formed.
3163 if (CodePoint < 0xA0) {
3164 if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
3165 return CodePoint;
3166
3167 // We don't use isLexingRawMode() here because we need to warn about bad
3168 // UCNs even when skipping preprocessing tokens in a #if block.
3169 if (Result && PP) {
3170 if (CodePoint < 0x20 || CodePoint >= 0x7F)
3171 Diag(BufferPtr, diag::err_ucn_control_character);
3172 else {
3173 char C = static_cast<char>(CodePoint);
3174 Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
3175 }
3176 }
3177
3178 return 0;
3179 } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
3180 // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
3181 // We don't use isLexingRawMode() here because we need to diagnose bad
3182 // UCNs even when skipping preprocessing tokens in a #if block.
3183 if (Result && PP) {
3184 if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
3185 Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
3186 else
3187 Diag(BufferPtr, diag::err_ucn_escape_invalid);
3188 }
3189 return 0;
3190 }
3191
3192 return CodePoint;
3193}
3194
3195bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
3196 const char *CurPtr) {
3197 if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
3198 isUnicodeWhitespace(C)) {
3199 Diag(BufferPtr, diag::ext_unicode_whitespace)
3200 << makeCharRange(*this, BufferPtr, CurPtr);
3201
3202 Result.setFlag(Token::LeadingSpace);
3203 return true;
3204 }
3205 return false;
3206}
3207
3208bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
3209 if (isAllowedInitiallyIDChar(C, LangOpts)) {
3210 if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
3211 !PP->isPreprocessedOutput()) {
3212 maybeDiagnoseIDCharCompat(PP->getDiagnostics(), C,
3213 makeCharRange(*this, BufferPtr, CurPtr),
3214 /*IsFirst=*/true);
3215 maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), C,
3216 makeCharRange(*this, BufferPtr, CurPtr));
3217 }
3218
3219 MIOpt.ReadToken();
3220 return LexIdentifier(Result, CurPtr);
3221 }
3222
3223 if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
3224 !PP->isPreprocessedOutput() && !isASCII(*BufferPtr) &&
3225 !isAllowedInitiallyIDChar(C, LangOpts) && !isUnicodeWhitespace(C)) {
3226 // Non-ASCII characters tend to creep into source code unintentionally.
3227 // Instead of letting the parser complain about the unknown token,
3228 // just drop the character.
3229 // Note that we can /only/ do this when the non-ASCII character is actually
3230 // spelled as Unicode, not written as a UCN. The standard requires that
3231 // we not throw away any possible preprocessor tokens, but there's a
3232 // loophole in the mapping of Unicode characters to basic character set
3233 // characters that allows us to map these particular characters to, say,
3234 // whitespace.
3235 diagnoseInvalidUnicodeCodepointInIdentifier(
3236 PP->getDiagnostics(), LangOpts, C,
3237 makeCharRange(*this, BufferPtr, CurPtr), /*IsStart*/ true);
3238 BufferPtr = CurPtr;
3239 return false;
3240 }
3241
3242 // Otherwise, we have an explicit UCN or a character that's unlikely to show
3243 // up by accident.
3244 MIOpt.ReadToken();
3245 FormTokenWithChars(Result, CurPtr, tok::unknown);
3246 return true;
3247}
3248
3249void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
3250 IsAtStartOfLine = Result.isAtStartOfLine();
3251 HasLeadingSpace = Result.hasLeadingSpace();
3252 HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
3253 // Note that this doesn't affect IsAtPhysicalStartOfLine.
3254}
3255
3256bool Lexer::Lex(Token &Result) {
3257 // Start a new token.
3258 Result.startToken();
3259
3260 // Set up misc whitespace flags for LexTokenInternal.
3261 if (IsAtStartOfLine) {
3262 Result.setFlag(Token::StartOfLine);
3263 IsAtStartOfLine = false;
3264 }
3265
3266 if (HasLeadingSpace) {
3267 Result.setFlag(Token::LeadingSpace);
3268 HasLeadingSpace = false;
3269 }
3270
3271 if (HasLeadingEmptyMacro) {
3272 Result.setFlag(Token::LeadingEmptyMacro);
3273 HasLeadingEmptyMacro = false;
3274 }
3275
3276 bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
3277 IsAtPhysicalStartOfLine = false;
3278 bool isRawLex = isLexingRawMode();
3279 (void) isRawLex;
3280 bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
3281 // (After the LexTokenInternal call, the lexer might be destroyed.)
3282 assert((returnedToken || !isRawLex) && "Raw lex must succeed")(static_cast<void> (0));
3283 return returnedToken;
3284}
3285
3286/// LexTokenInternal - This implements a simple C family lexer. It is an
3287/// extremely performance critical piece of code. This assumes that the buffer
3288/// has a null character at the end of the file. This returns a preprocessing
3289/// token, not a normal token, as such, it is an internal interface. It assumes
3290/// that the Flags of result have been cleared before calling this.
3291bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
3292LexNextToken:
3293 // New token, can't need cleaning yet.
3294 Result.clearFlag(Token::NeedsCleaning);
3295 Result.setIdentifierInfo(nullptr);
3296
3297 // CurPtr - Cache BufferPtr in an automatic variable.
3298 const char *CurPtr = BufferPtr;
3299
3300 // Small amounts of horizontal whitespace is very common between tokens.
3301 if (isHorizontalWhitespace(*CurPtr)) {
1
Assuming the condition is false
2
Taking false branch
3302 do {
3303 ++CurPtr;
3304 } while (isHorizontalWhitespace(*CurPtr));
3305
3306 // If we are keeping whitespace and other tokens, just return what we just
3307 // skipped. The next lexer invocation will return the token after the
3308 // whitespace.
3309 if (isKeepWhitespaceMode()) {
3310 FormTokenWithChars(Result, CurPtr, tok::unknown);
3311 // FIXME: The next token will not have LeadingSpace set.
3312 return true;
3313 }
3314
3315 BufferPtr = CurPtr;
3316 Result.setFlag(Token::LeadingSpace);
3317 }
3318
3319 unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
3320
3321 // Read a character, advancing over it.
3322 char Char = getAndAdvanceChar(CurPtr, Result);
3323 tok::TokenKind Kind;
3324
3325 if (!isVerticalWhitespace(Char))
3
Assuming the condition is false
4
Taking false branch
3326 NewLinePtr = nullptr;
3327
3328 switch (Char) {
5
Control jumps to 'case 32:' at line 3397
3329 case 0: // Null.
3330 // Found end of file?
3331 if (CurPtr-1 == BufferEnd)
3332 return LexEndOfFile(Result, CurPtr-1);
3333
3334 // Check if we are performing code completion.
3335 if (isCodeCompletionPoint(CurPtr-1)) {
3336 // Return the code-completion token.
3337 Result.startToken();
3338 FormTokenWithChars(Result, CurPtr, tok::code_completion);
3339 return true;
3340 }
3341
3342 if (!isLexingRawMode())
3343 Diag(CurPtr-1, diag::null_in_file);
3344 Result.setFlag(Token::LeadingSpace);
3345 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3346 return true; // KeepWhitespaceMode
3347
3348 // We know the lexer hasn't changed, so just try again with this lexer.
3349 // (We manually eliminate the tail call to avoid recursion.)
3350 goto LexNextToken;
3351
3352 case 26: // DOS & CP/M EOF: "^Z".
3353 // If we're in Microsoft extensions mode, treat this as end of file.
3354 if (LangOpts.MicrosoftExt) {
3355 if (!isLexingRawMode())
3356 Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
3357 return LexEndOfFile(Result, CurPtr-1);
3358 }
3359
3360 // If Microsoft extensions are disabled, this is just random garbage.
3361 Kind = tok::unknown;
3362 break;
3363
3364 case '\r':
3365 if (CurPtr[0] == '\n')
3366 (void)getAndAdvanceChar(CurPtr, Result);
3367 LLVM_FALLTHROUGH[[gnu::fallthrough]];
3368 case '\n':
3369 // If we are inside a preprocessor directive and we see the end of line,
3370 // we know we are done with the directive, so return an EOD token.
3371 if (ParsingPreprocessorDirective) {
3372 // Done parsing the "line".
3373 ParsingPreprocessorDirective = false;
3374
3375 // Restore comment saving mode, in case it was disabled for directive.
3376 if (PP)
3377 resetExtendedTokenMode();
3378
3379 // Since we consumed a newline, we are back at the start of a line.
3380 IsAtStartOfLine = true;
3381 IsAtPhysicalStartOfLine = true;
3382 NewLinePtr = CurPtr - 1;
3383
3384 Kind = tok::eod;
3385 break;
3386 }
3387
3388 // No leading whitespace seen so far.
3389 Result.clearFlag(Token::LeadingSpace);
3390
3391 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3392 return true; // KeepWhitespaceMode
3393
3394 // We only saw whitespace, so just try again with this lexer.
3395 // (We manually eliminate the tail call to avoid recursion.)
3396 goto LexNextToken;
3397 case ' ':
3398 case '\t':
3399 case '\f':
3400 case '\v':
3401 SkipHorizontalWhitespace:
3402 Result.setFlag(Token::LeadingSpace);
3403 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
6
Taking false branch
3404 return true; // KeepWhitespaceMode
3405
3406 SkipIgnoredUnits:
3407 CurPtr = BufferPtr;
3408
3409 // If the next token is obviously a // or /* */ comment, skip it efficiently
3410 // too (without going through the big switch stmt).
3411 if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
7
Assuming the condition is false
3412 LangOpts.LineComment &&
3413 (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3414 if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3415 return true; // There is a token to return.
3416 goto SkipIgnoredUnits;
3417 } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
8
Assuming the condition is true
9
Assuming the condition is true
10
Taking true branch
3418 if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
11
Calling 'Lexer::SkipBlockComment'
3419 return true; // There is a token to return.
3420 goto SkipIgnoredUnits;
3421 } else if (isHorizontalWhitespace(*CurPtr)) {
3422 goto SkipHorizontalWhitespace;
3423 }
3424 // We only saw whitespace, so just try again with this lexer.
3425 // (We manually eliminate the tail call to avoid recursion.)
3426 goto LexNextToken;
3427
3428 // C99 6.4.4.1: Integer Constants.
3429 // C99 6.4.4.2: Floating Constants.
3430 case '0': case '1': case '2': case '3': case '4':
3431 case '5': case '6': case '7': case '8': case '9':
3432 // Notify MIOpt that we read a non-whitespace/non-comment token.
3433 MIOpt.ReadToken();
3434 return LexNumericConstant(Result, CurPtr);
3435
3436 case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3437 // Notify MIOpt that we read a non-whitespace/non-comment token.
3438 MIOpt.ReadToken();
3439
3440 if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3441 Char = getCharAndSize(CurPtr, SizeTmp);
3442
3443 // UTF-16 string literal
3444 if (Char == '"')
3445 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3446 tok::utf16_string_literal);
3447
3448 // UTF-16 character constant
3449 if (Char == '\'')
3450 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3451 tok::utf16_char_constant);
3452
3453 // UTF-16 raw string literal
3454 if (Char == 'R' && LangOpts.CPlusPlus11 &&
3455 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3456 return LexRawStringLiteral(Result,
3457 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3458 SizeTmp2, Result),
3459 tok::utf16_string_literal);
3460
3461 if (Char == '8') {
3462 char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3463
3464 // UTF-8 string literal
3465 if (Char2 == '"')
3466 return LexStringLiteral(Result,
3467 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3468 SizeTmp2, Result),
3469 tok::utf8_string_literal);
3470 if (Char2 == '\'' && LangOpts.CPlusPlus17)
3471 return LexCharConstant(
3472 Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3473 SizeTmp2, Result),
3474 tok::utf8_char_constant);
3475
3476 if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3477 unsigned SizeTmp3;
3478 char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3479 // UTF-8 raw string literal
3480 if (Char3 == '"') {
3481 return LexRawStringLiteral(Result,
3482 ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3483 SizeTmp2, Result),
3484 SizeTmp3, Result),
3485 tok::utf8_string_literal);
3486 }
3487 }
3488 }
3489 }
3490
3491 // treat u like the start of an identifier.
3492 return LexIdentifier(Result, CurPtr);
3493
3494 case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3495 // Notify MIOpt that we read a non-whitespace/non-comment token.
3496 MIOpt.ReadToken();
3497
3498 if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3499 Char = getCharAndSize(CurPtr, SizeTmp);
3500
3501 // UTF-32 string literal
3502 if (Char == '"')
3503 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3504 tok::utf32_string_literal);
3505
3506 // UTF-32 character constant
3507 if (Char == '\'')
3508 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3509 tok::utf32_char_constant);
3510
3511 // UTF-32 raw string literal
3512 if (Char == 'R' && LangOpts.CPlusPlus11 &&
3513 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3514 return LexRawStringLiteral(Result,
3515 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3516 SizeTmp2, Result),
3517 tok::utf32_string_literal);
3518 }
3519
3520 // treat U like the start of an identifier.
3521 return LexIdentifier(Result, CurPtr);
3522
3523 case 'R': // Identifier or C++0x raw string literal
3524 // Notify MIOpt that we read a non-whitespace/non-comment token.
3525 MIOpt.ReadToken();
3526
3527 if (LangOpts.CPlusPlus11) {
3528 Char = getCharAndSize(CurPtr, SizeTmp);
3529
3530 if (Char == '"')
3531 return LexRawStringLiteral(Result,
3532 ConsumeChar(CurPtr, SizeTmp, Result),
3533 tok::string_literal);
3534 }
3535
3536 // treat R like the start of an identifier.
3537 return LexIdentifier(Result, CurPtr);
3538
3539 case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3540 // Notify MIOpt that we read a non-whitespace/non-comment token.
3541 MIOpt.ReadToken();
3542 Char = getCharAndSize(CurPtr, SizeTmp);
3543
3544 // Wide string literal.
3545 if (Char == '"')
3546 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3547 tok::wide_string_literal);
3548
3549 // Wide raw string literal.
3550 if (LangOpts.CPlusPlus11 && Char == 'R' &&
3551 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3552 return LexRawStringLiteral(Result,
3553 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3554 SizeTmp2, Result),
3555 tok::wide_string_literal);
3556
3557 // Wide character constant.
3558 if (Char == '\'')
3559 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3560 tok::wide_char_constant);
3561 // FALL THROUGH, treating L like the start of an identifier.
3562 LLVM_FALLTHROUGH[[gnu::fallthrough]];
3563
3564 // C99 6.4.2: Identifiers.
3565 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3566 case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3567 case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3568 case 'V': case 'W': case 'X': case 'Y': case 'Z':
3569 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3570 case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3571 case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3572 case 'v': case 'w': case 'x': case 'y': case 'z':
3573 case '_':
3574 // Notify MIOpt that we read a non-whitespace/non-comment token.
3575 MIOpt.ReadToken();
3576 return LexIdentifier(Result, CurPtr);
3577
3578 case '$': // $ in identifiers.
3579 if (LangOpts.DollarIdents) {
3580 if (!isLexingRawMode())
3581 Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3582 // Notify MIOpt that we read a non-whitespace/non-comment token.
3583 MIOpt.ReadToken();
3584 return LexIdentifier(Result, CurPtr);
3585 }
3586
3587 Kind = tok::unknown;
3588 break;
3589
3590 // C99 6.4.4: Character Constants.
3591 case '\'':
3592 // Notify MIOpt that we read a non-whitespace/non-comment token.
3593 MIOpt.ReadToken();
3594 return LexCharConstant(Result, CurPtr, tok::char_constant);
3595
3596 // C99 6.4.5: String Literals.
3597 case '"':
3598 // Notify MIOpt that we read a non-whitespace/non-comment token.
3599 MIOpt.ReadToken();
3600 return LexStringLiteral(Result, CurPtr,
3601 ParsingFilename ? tok::header_name
3602 : tok::string_literal);
3603
3604 // C99 6.4.6: Punctuators.
3605 case '?':
3606 Kind = tok::question;
3607 break;
3608 case '[':
3609 Kind = tok::l_square;
3610 break;
3611 case ']':
3612 Kind = tok::r_square;
3613 break;
3614 case '(':
3615 Kind = tok::l_paren;
3616 break;
3617 case ')':
3618 Kind = tok::r_paren;
3619 break;
3620 case '{':
3621 Kind = tok::l_brace;
3622 break;
3623 case '}':
3624 Kind = tok::r_brace;
3625 break;
3626 case '.':
3627 Char = getCharAndSize(CurPtr, SizeTmp);
3628 if (Char >= '0' && Char <= '9') {
3629 // Notify MIOpt that we read a non-whitespace/non-comment token.
3630 MIOpt.ReadToken();
3631
3632 return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3633 } else if (LangOpts.CPlusPlus && Char == '*') {
3634 Kind = tok::periodstar;
3635 CurPtr += SizeTmp;
3636 } else if (Char == '.' &&
3637 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3638 Kind = tok::ellipsis;
3639 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3640 SizeTmp2, Result);
3641 } else {
3642 Kind = tok::period;
3643 }
3644 break;
3645 case '&':
3646 Char = getCharAndSize(CurPtr, SizeTmp);
3647 if (Char == '&') {
3648 Kind = tok::ampamp;
3649 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3650 } else if (Char == '=') {
3651 Kind = tok::ampequal;
3652 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3653 } else {
3654 Kind = tok::amp;
3655 }
3656 break;
3657 case '*':
3658 if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3659 Kind = tok::starequal;
3660 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3661 } else {
3662 Kind = tok::star;
3663 }
3664 break;
3665 case '+':
3666 Char = getCharAndSize(CurPtr, SizeTmp);
3667 if (Char == '+') {
3668 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3669 Kind = tok::plusplus;
3670 } else if (Char == '=') {
3671 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3672 Kind = tok::plusequal;
3673 } else {
3674 Kind = tok::plus;
3675 }
3676 break;
3677 case '-':
3678 Char = getCharAndSize(CurPtr, SizeTmp);
3679 if (Char == '-') { // --
3680 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3681 Kind = tok::minusminus;
3682 } else if (Char == '>' && LangOpts.CPlusPlus &&
3683 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3684 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3685 SizeTmp2, Result);
3686 Kind = tok::arrowstar;
3687 } else if (Char == '>') { // ->
3688 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3689 Kind = tok::arrow;
3690 } else if (Char == '=') { // -=
3691 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3692 Kind = tok::minusequal;
3693 } else {
3694 Kind = tok::minus;
3695 }
3696 break;
3697 case '~':
3698 Kind = tok::tilde;
3699 break;
3700 case '!':
3701 if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3702 Kind = tok::exclaimequal;
3703 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3704 } else {
3705 Kind = tok::exclaim;
3706 }
3707 break;
3708 case '/':
3709 // 6.4.9: Comments
3710 Char = getCharAndSize(CurPtr, SizeTmp);
3711 if (Char == '/') { // Line comment.
3712 // Even if Line comments are disabled (e.g. in C89 mode), we generally
3713 // want to lex this as a comment. There is one problem with this though,
3714 // that in one particular corner case, this can change the behavior of the
3715 // resultant program. For example, In "foo //**/ bar", C89 would lex
3716 // this as "foo / bar" and languages with Line comments would lex it as
3717 // "foo". Check to see if the character after the second slash is a '*'.
3718 // If so, we will lex that as a "/" instead of the start of a comment.
3719 // However, we never do this if we are just preprocessing.
3720 bool TreatAsComment = LangOpts.LineComment &&
3721 (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3722 if (!TreatAsComment)
3723 if (!(PP && PP->isPreprocessedOutput()))
3724 TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3725
3726 if (TreatAsComment) {
3727 if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3728 TokAtPhysicalStartOfLine))
3729 return true; // There is a token to return.
3730
3731 // It is common for the tokens immediately after a // comment to be
3732 // whitespace (indentation for the next line). Instead of going through
3733 // the big switch, handle it efficiently now.
3734 goto SkipIgnoredUnits;
3735 }
3736 }
3737
3738 if (Char == '*') { // /**/ comment.
3739 if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3740 TokAtPhysicalStartOfLine))
3741 return true; // There is a token to return.
3742
3743 // We only saw whitespace, so just try again with this lexer.
3744 // (We manually eliminate the tail call to avoid recursion.)
3745 goto LexNextToken;
3746 }
3747
3748 if (Char == '=') {
3749 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3750 Kind = tok::slashequal;
3751 } else {
3752 Kind = tok::slash;
3753 }
3754 break;
3755 case '%':
3756 Char = getCharAndSize(CurPtr, SizeTmp);
3757 if (Char == '=') {
3758 Kind = tok::percentequal;
3759 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3760 } else if (LangOpts.Digraphs && Char == '>') {
3761 Kind = tok::r_brace; // '%>' -> '}'
3762 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3763 } else if (LangOpts.Digraphs && Char == ':') {
3764 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3765 Char = getCharAndSize(CurPtr, SizeTmp);
3766 if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3767 Kind = tok::hashhash; // '%:%:' -> '##'
3768 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3769 SizeTmp2, Result);
3770 } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3771 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3772 if (!isLexingRawMode())
3773 Diag(BufferPtr, diag::ext_charize_microsoft);
3774 Kind = tok::hashat;
3775 } else { // '%:' -> '#'
3776 // We parsed a # character. If this occurs at the start of the line,
3777 // it's actually the start of a preprocessing directive. Callback to
3778 // the preprocessor to handle it.
3779 // TODO: -fpreprocessed mode??
3780 if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3781 goto HandleDirective;
3782
3783 Kind = tok::hash;
3784 }
3785 } else {
3786 Kind = tok::percent;
3787 }
3788 break;
3789 case '<':
3790 Char = getCharAndSize(CurPtr, SizeTmp);
3791 if (ParsingFilename) {
3792 return LexAngledStringLiteral(Result, CurPtr);
3793 } else if (Char == '<') {
3794 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3795 if (After == '=') {
3796 Kind = tok::lesslessequal;
3797 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3798 SizeTmp2, Result);
3799 } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3800 // If this is actually a '<<<<<<<' version control conflict marker,
3801 // recognize it as such and recover nicely.
3802 goto LexNextToken;
3803 } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3804 // If this is '<<<<' and we're in a Perforce-style conflict marker,
3805 // ignore it.
3806 goto LexNextToken;
3807 } else if (LangOpts.CUDA && After == '<') {
3808 Kind = tok::lesslessless;
3809 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3810 SizeTmp2, Result);
3811 } else {
3812 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3813 Kind = tok::lessless;
3814 }
3815 } else if (Char == '=') {
3816 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3817 if (After == '>') {
3818 if (getLangOpts().CPlusPlus20) {
3819 if (!isLexingRawMode())
3820 Diag(BufferPtr, diag::warn_cxx17_compat_spaceship);
3821 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3822 SizeTmp2, Result);
3823 Kind = tok::spaceship;
3824 break;
3825 }
3826 // Suggest adding a space between the '<=' and the '>' to avoid a
3827 // change in semantics if this turns up in C++ <=17 mode.
3828 if (getLangOpts().CPlusPlus && !isLexingRawMode()) {
3829 Diag(BufferPtr, diag::warn_cxx20_compat_spaceship)
3830 << FixItHint::CreateInsertion(
3831 getSourceLocation(CurPtr + SizeTmp, SizeTmp2), " ");
3832 }
3833 }
3834 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3835 Kind = tok::lessequal;
3836 } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3837 if (LangOpts.CPlusPlus11 &&
3838 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3839 // C++0x [lex.pptoken]p3:
3840 // Otherwise, if the next three characters are <:: and the subsequent
3841 // character is neither : nor >, the < is treated as a preprocessor
3842 // token by itself and not as the first character of the alternative
3843 // token <:.
3844 unsigned SizeTmp3;
3845 char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3846 if (After != ':' && After != '>') {
3847 Kind = tok::less;
3848 if (!isLexingRawMode())
3849 Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3850 break;
3851 }
3852 }
3853
3854 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3855 Kind = tok::l_square;
3856 } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3857 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3858 Kind = tok::l_brace;
3859 } else if (Char == '#' && /*Not a trigraph*/ SizeTmp == 1 &&
3860 lexEditorPlaceholder(Result, CurPtr)) {
3861 return true;
3862 } else {
3863 Kind = tok::less;
3864 }
3865 break;
3866 case '>':
3867 Char = getCharAndSize(CurPtr, SizeTmp);
3868 if (Char == '=') {
3869 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3870 Kind = tok::greaterequal;
3871 } else if (Char == '>') {
3872 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3873 if (After == '=') {
3874 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3875 SizeTmp2, Result);
3876 Kind = tok::greatergreaterequal;
3877 } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3878 // If this is actually a '>>>>' conflict marker, recognize it as such
3879 // and recover nicely.
3880 goto LexNextToken;
3881 } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3882 // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3883 goto LexNextToken;
3884 } else if (LangOpts.CUDA && After == '>') {
3885 Kind = tok::greatergreatergreater;
3886 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3887 SizeTmp2, Result);
3888 } else {
3889 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3890 Kind = tok::greatergreater;
3891 }
3892 } else {
3893 Kind = tok::greater;
3894 }
3895 break;
3896 case '^':
3897 Char = getCharAndSize(CurPtr, SizeTmp);
3898 if (Char == '=') {
3899 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3900 Kind = tok::caretequal;
3901 } else if (LangOpts.OpenCL && Char == '^') {
3902 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3903 Kind = tok::caretcaret;
3904 } else {
3905 Kind = tok::caret;
3906 }
3907 break;
3908 case '|':
3909 Char = getCharAndSize(CurPtr, SizeTmp);
3910 if (Char == '=') {
3911 Kind = tok::pipeequal;
3912 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3913 } else if (Char == '|') {
3914 // If this is '|||||||' and we're in a conflict marker, ignore it.
3915 if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3916 goto LexNextToken;
3917 Kind = tok::pipepipe;
3918 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3919 } else {
3920 Kind = tok::pipe;
3921 }
3922 break;
3923 case ':':
3924 Char = getCharAndSize(CurPtr, SizeTmp);
3925 if (LangOpts.Digraphs && Char == '>') {
3926 Kind = tok::r_square; // ':>' -> ']'
3927 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3928 } else if ((LangOpts.CPlusPlus ||
3929 LangOpts.DoubleSquareBracketAttributes) &&
3930 Char == ':') {
3931 Kind = tok::coloncolon;
3932 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3933 } else {
3934 Kind = tok::colon;
3935 }
3936 break;
3937 case ';':
3938 Kind = tok::semi;
3939 break;
3940 case '=':
3941 Char = getCharAndSize(CurPtr, SizeTmp);
3942 if (Char == '=') {
3943 // If this is '====' and we're in a conflict marker, ignore it.
3944 if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3945 goto LexNextToken;
3946
3947 Kind = tok::equalequal;
3948 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3949 } else {
3950 Kind = tok::equal;
3951 }
3952 break;
3953 case ',':
3954 Kind = tok::comma;
3955 break;
3956 case '#':
3957 Char = getCharAndSize(CurPtr, SizeTmp);
3958 if (Char == '#') {
3959 Kind = tok::hashhash;
3960 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3961 } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3962 Kind = tok::hashat;
3963 if (!isLexingRawMode())
3964 Diag(BufferPtr, diag::ext_charize_microsoft);
3965 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3966 } else {
3967 // We parsed a # character. If this occurs at the start of the line,
3968 // it's actually the start of a preprocessing directive. Callback to
3969 // the preprocessor to handle it.
3970 // TODO: -fpreprocessed mode??
3971 if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3972 goto HandleDirective;
3973
3974 Kind = tok::hash;
3975 }
3976 break;
3977
3978 case '@':
3979 // Objective C support.
3980 if (CurPtr[-1] == '@' && LangOpts.ObjC)
3981 Kind = tok::at;
3982 else
3983 Kind = tok::unknown;
3984 break;
3985
3986 // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3987 case '\\':
3988 if (!LangOpts.AsmPreprocessor) {
3989 if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3990 if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3991 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3992 return true; // KeepWhitespaceMode
3993
3994 // We only saw whitespace, so just try again with this lexer.
3995 // (We manually eliminate the tail call to avoid recursion.)
3996 goto LexNextToken;
3997 }
3998
3999 return LexUnicode(Result, CodePoint, CurPtr);
4000 }
4001 }
4002
4003 Kind = tok::unknown;
4004 break;
4005
4006 default: {
4007 if (isASCII(Char)) {
4008 Kind = tok::unknown;
4009 break;
4010 }
4011
4012 llvm::UTF32 CodePoint;
4013
4014 // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
4015 // an escaped newline.
4016 --CurPtr;
4017 llvm::ConversionResult Status =
4018 llvm::convertUTF8Sequence((const llvm::UTF8 **)&CurPtr,
4019 (const llvm::UTF8 *)BufferEnd,
4020 &CodePoint,
4021 llvm::strictConversion);
4022 if (Status == llvm::conversionOK) {
4023 if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
4024 if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
4025 return true; // KeepWhitespaceMode
4026
4027 // We only saw whitespace, so just try again with this lexer.
4028 // (We manually eliminate the tail call to avoid recursion.)
4029 goto LexNextToken;
4030 }
4031 return LexUnicode(Result, CodePoint, CurPtr);
4032 }
4033
4034 if (isLexingRawMode() || ParsingPreprocessorDirective ||
4035 PP->isPreprocessedOutput()) {
4036 ++CurPtr;
4037 Kind = tok::unknown;
4038 break;
4039 }
4040
4041 // Non-ASCII characters tend to creep into source code unintentionally.
4042 // Instead of letting the parser complain about the unknown token,
4043 // just diagnose the invalid UTF-8, then drop the character.
4044 Diag(CurPtr, diag::err_invalid_utf8);
4045
4046 BufferPtr = CurPtr+1;
4047 // We're pretending the character didn't exist, so just try again with
4048 // this lexer.
4049 // (We manually eliminate the tail call to avoid recursion.)
4050 goto LexNextToken;
4051 }
4052 }
4053
4054 // Notify MIOpt that we read a non-whitespace/non-comment token.
4055 MIOpt.ReadToken();
4056
4057 // Update the location of token as well as BufferPtr.
4058 FormTokenWithChars(Result, CurPtr, Kind);
4059 return true;
4060
4061HandleDirective:
4062 // We parsed a # character and it's the start of a preprocessing directive.
4063
4064 FormTokenWithChars(Result, CurPtr, tok::hash);
4065 PP->HandleDirective(Result);
4066
4067 if (PP->hadModuleLoaderFatalFailure()) {
4068 // With a fatal failure in the module loader, we abort parsing.
4069 assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof")(static_cast<void> (0));
4070 return true;
4071 }
4072
4073 // We parsed the directive; lex a token with the new state.
4074 return false;
4075}

/build/llvm-toolchain-snapshot-14~++20210903100615+fd66b44ec19e/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!")(static_cast<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) &&(static_cast<void> (0))
218 "Can only retain whitespace in raw mode or -traditional-cpp")(static_cast<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() &&(static_cast<void> (0))
233 "Can't play with comment retention state when retaining whitespace")(static_cast<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")(static_cast<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