blob: fc1418ffef63d7eaaea6d49aa623ca82c6b0c338 [file] [log] [blame]
Reid Spencer5f016e22007-07-11 17:01:13 +00001//===--- Preprocess.cpp - C Language Family Preprocessor Implementation ---===//
2//
3// The LLVM Compiler Infrastructure
4//
Chris Lattner0bc735f2007-12-29 19:59:25 +00005// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
Reid Spencer5f016e22007-07-11 17:01:13 +00007//
8//===----------------------------------------------------------------------===//
9//
10// This file implements the Preprocessor interface.
11//
12//===----------------------------------------------------------------------===//
13//
14// Options to support:
15// -H - Print the name of each header file used.
Chris Lattnerf73903a2009-02-06 06:45:26 +000016// -d[DNI] - Dump various things.
Reid Spencer5f016e22007-07-11 17:01:13 +000017// -fworking-directory - #line's with preprocessor's working dir.
18// -fpreprocessed
19// -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD
20// -W*
21// -w
22//
23// Messages to emit:
24// "Multiple include guards may be useful for:\n"
25//
26//===----------------------------------------------------------------------===//
27
28#include "clang/Lex/Preprocessor.h"
Chris Lattner23f77e52009-12-15 01:51:03 +000029#include "MacroArgs.h"
Douglas Gregor88a35862010-01-04 19:18:44 +000030#include "clang/Lex/ExternalPreprocessorSource.h"
Reid Spencer5f016e22007-07-11 17:01:13 +000031#include "clang/Lex/HeaderSearch.h"
32#include "clang/Lex/MacroInfo.h"
Reid Spencer5f016e22007-07-11 17:01:13 +000033#include "clang/Lex/Pragma.h"
Douglas Gregor94dc8f62010-03-19 16:15:56 +000034#include "clang/Lex/PreprocessingRecord.h"
Reid Spencer5f016e22007-07-11 17:01:13 +000035#include "clang/Lex/ScratchBuffer.h"
Chris Lattner500d3292009-01-29 05:15:15 +000036#include "clang/Lex/LexDiagnostic.h"
Reid Spencer5f016e22007-07-11 17:01:13 +000037#include "clang/Basic/SourceManager.h"
Ted Kremenek337edcd2009-02-12 03:26:59 +000038#include "clang/Basic/FileManager.h"
Reid Spencer5f016e22007-07-11 17:01:13 +000039#include "clang/Basic/TargetInfo.h"
Chris Lattner2db78dd2008-10-05 20:40:30 +000040#include "llvm/ADT/APFloat.h"
Reid Spencer5f016e22007-07-11 17:01:13 +000041#include "llvm/ADT/SmallVector.h"
Chris Lattner97ba77c2007-07-16 06:48:38 +000042#include "llvm/Support/MemoryBuffer.h"
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +000043#include "llvm/Support/raw_ostream.h"
Reid Spencer5f016e22007-07-11 17:01:13 +000044using namespace clang;
45
46//===----------------------------------------------------------------------===//
Douglas Gregor88a35862010-01-04 19:18:44 +000047ExternalPreprocessorSource::~ExternalPreprocessorSource() { }
Reid Spencer5f016e22007-07-11 17:01:13 +000048
49Preprocessor::Preprocessor(Diagnostic &diags, const LangOptions &opts,
Daniel Dunbar444be732009-11-13 05:51:54 +000050 const TargetInfo &target, SourceManager &SM,
Ted Kremenek72b1b152009-01-15 18:47:46 +000051 HeaderSearch &Headers,
Daniel Dunbar5814e652009-11-11 21:44:21 +000052 IdentifierInfoLookup* IILookup,
53 bool OwnsHeaders)
Chris Lattner836040f2009-03-13 21:17:43 +000054 : Diags(&diags), Features(opts), Target(target),FileMgr(Headers.getFileMgr()),
Douglas Gregor88a35862010-01-04 19:18:44 +000055 SourceMgr(SM), HeaderInfo(Headers), ExternalSource(0),
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +000056 Identifiers(opts, IILookup), BuiltinInfo(Target), CodeCompletionFile(0),
Douglas Gregorf4f6c9d2010-07-26 21:36:20 +000057 SkipMainFilePreamble(0, true), CurPPLexer(0), CurDirLookup(0), Callbacks(0),
58 MacroArgCache(0), Record(0) {
Reid Spencer5f016e22007-07-11 17:01:13 +000059 ScratchBuf = new ScratchBuffer(SourceMgr);
Chris Lattnerc1f9d822009-04-13 01:29:17 +000060 CounterValue = 0; // __COUNTER__ starts at 0.
Daniel Dunbar5814e652009-11-11 21:44:21 +000061 OwnsHeaderSearch = OwnsHeaders;
Mike Stump1eb44332009-09-09 15:08:12 +000062
Reid Spencer5f016e22007-07-11 17:01:13 +000063 // Clear stats.
64 NumDirectives = NumDefined = NumUndefined = NumPragma = 0;
65 NumIf = NumElse = NumEndif = 0;
66 NumEnteredSourceFiles = 0;
67 NumMacroExpanded = NumFnMacroExpanded = NumBuiltinMacroExpanded = 0;
68 NumFastMacroExpanded = NumTokenPaste = NumFastTokenPaste = 0;
Mike Stump1eb44332009-09-09 15:08:12 +000069 MaxIncludeStackDepth = 0;
Reid Spencer5f016e22007-07-11 17:01:13 +000070 NumSkipped = 0;
71
72 // Default to discarding comments.
73 KeepComments = false;
74 KeepMacroComments = false;
Mike Stump1eb44332009-09-09 15:08:12 +000075
Reid Spencer5f016e22007-07-11 17:01:13 +000076 // Macro expansion is enabled.
77 DisableMacroExpansion = false;
78 InMacroArgs = false;
Chris Lattner6cfe7592008-03-09 02:26:03 +000079 NumCachedTokenLexers = 0;
Reid Spencer5f016e22007-07-11 17:01:13 +000080
Argyrios Kyrtzidis03db1b32008-08-10 13:15:22 +000081 CachedLexPos = 0;
82
Douglas Gregor88a35862010-01-04 19:18:44 +000083 // We haven't read anything from the external source.
84 ReadMacrosFromExternalSource = false;
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +000085
Reid Spencer5f016e22007-07-11 17:01:13 +000086 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
87 // This gets unpoisoned where it is allowed.
88 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
Mike Stump1eb44332009-09-09 15:08:12 +000089
Reid Spencer5f016e22007-07-11 17:01:13 +000090 // Initialize the pragma handlers.
Argyrios Kyrtzidis9b36c3f2010-07-13 09:07:17 +000091 PragmaHandlers = new PragmaNamespace(llvm::StringRef());
Reid Spencer5f016e22007-07-11 17:01:13 +000092 RegisterBuiltinPragmas();
Mike Stump1eb44332009-09-09 15:08:12 +000093
Reid Spencer5f016e22007-07-11 17:01:13 +000094 // Initialize builtin macros like __LINE__ and friends.
95 RegisterBuiltinMacros();
96}
97
98Preprocessor::~Preprocessor() {
Argyrios Kyrtzidis2174a4f2008-08-23 12:12:06 +000099 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
100
Reid Spencer5f016e22007-07-11 17:01:13 +0000101 while (!IncludeMacroStack.empty()) {
102 delete IncludeMacroStack.back().TheLexer;
Chris Lattner6cfe7592008-03-09 02:26:03 +0000103 delete IncludeMacroStack.back().TheTokenLexer;
Reid Spencer5f016e22007-07-11 17:01:13 +0000104 IncludeMacroStack.pop_back();
105 }
Chris Lattnercc1a8752007-10-07 08:44:20 +0000106
107 // Free any macro definitions.
108 for (llvm::DenseMap<IdentifierInfo*, MacroInfo*>::iterator I =
109 Macros.begin(), E = Macros.end(); I != E; ++I) {
Ted Kremenek0ea76722008-12-15 19:56:42 +0000110 // We don't need to free the MacroInfo objects directly. These
111 // will be released when the BumpPtrAllocator 'BP' object gets
Chris Lattner2b9d6372009-12-14 22:02:43 +0000112 // destroyed. We still need to run the dtor, however, to free
Ted Kremenek9ee7d452009-01-19 07:45:44 +0000113 // memory alocated by MacroInfo.
Chris Lattner685befe2009-02-20 22:46:43 +0000114 I->second->Destroy(BP);
Chris Lattnercc1a8752007-10-07 08:44:20 +0000115 I->first->setHasMacroDefinition(false);
116 }
Ted Kremenek83c1a6f2010-06-08 23:00:53 +0000117 for (std::vector<MacroInfo*>::iterator I = MICache.begin(),
118 E = MICache.end(); I != E; ++I) {
119 // We don't need to free the MacroInfo objects directly. These
120 // will be released when the BumpPtrAllocator 'BP' object gets
121 // destroyed. We still need to run the dtor, however, to free
122 // memory alocated by MacroInfo.
123 (*I)->Destroy(BP);
124 }
Mike Stump1eb44332009-09-09 15:08:12 +0000125
Chris Lattner9594acf2007-07-15 00:25:26 +0000126 // Free any cached macro expanders.
Chris Lattner6cfe7592008-03-09 02:26:03 +0000127 for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i)
128 delete TokenLexerCache[i];
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000129
Chris Lattner23f77e52009-12-15 01:51:03 +0000130 // Free any cached MacroArgs.
131 for (MacroArgs *ArgList = MacroArgCache; ArgList; )
132 ArgList = ArgList->deallocate();
Mike Stump1eb44332009-09-09 15:08:12 +0000133
Reid Spencer5f016e22007-07-11 17:01:13 +0000134 // Release pragma information.
135 delete PragmaHandlers;
136
137 // Delete the scratch buffer info.
138 delete ScratchBuf;
Chris Lattnereb50ed82008-03-14 06:07:05 +0000139
Daniel Dunbar5814e652009-11-11 21:44:21 +0000140 // Delete the header search info, if we own it.
141 if (OwnsHeaderSearch)
142 delete &HeaderInfo;
143
Chris Lattnereb50ed82008-03-14 06:07:05 +0000144 delete Callbacks;
Reid Spencer5f016e22007-07-11 17:01:13 +0000145}
146
Ted Kremenek337edcd2009-02-12 03:26:59 +0000147void Preprocessor::setPTHManager(PTHManager* pm) {
148 PTH.reset(pm);
Douglas Gregor52e71082009-10-16 18:18:30 +0000149 FileMgr.addStatCache(PTH->createStatCache());
Ted Kremenek337edcd2009-02-12 03:26:59 +0000150}
151
Chris Lattnerd2177732007-07-20 16:59:19 +0000152void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000153 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
154 << getSpelling(Tok) << "'";
Mike Stump1eb44332009-09-09 15:08:12 +0000155
Reid Spencer5f016e22007-07-11 17:01:13 +0000156 if (!DumpFlags) return;
Mike Stump1eb44332009-09-09 15:08:12 +0000157
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000158 llvm::errs() << "\t";
Reid Spencer5f016e22007-07-11 17:01:13 +0000159 if (Tok.isAtStartOfLine())
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000160 llvm::errs() << " [StartOfLine]";
Reid Spencer5f016e22007-07-11 17:01:13 +0000161 if (Tok.hasLeadingSpace())
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000162 llvm::errs() << " [LeadingSpace]";
Reid Spencer5f016e22007-07-11 17:01:13 +0000163 if (Tok.isExpandDisabled())
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000164 llvm::errs() << " [ExpandDisabled]";
Reid Spencer5f016e22007-07-11 17:01:13 +0000165 if (Tok.needsCleaning()) {
166 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000167 llvm::errs() << " [UnClean='" << std::string(Start, Start+Tok.getLength())
168 << "']";
Reid Spencer5f016e22007-07-11 17:01:13 +0000169 }
Mike Stump1eb44332009-09-09 15:08:12 +0000170
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000171 llvm::errs() << "\tLoc=<";
Chris Lattnerc3d8d572007-12-09 20:31:55 +0000172 DumpLocation(Tok.getLocation());
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000173 llvm::errs() << ">";
Chris Lattnerc3d8d572007-12-09 20:31:55 +0000174}
175
176void Preprocessor::DumpLocation(SourceLocation Loc) const {
Chris Lattnerb9c3f962009-01-27 07:57:44 +0000177 Loc.dump(SourceMgr);
Reid Spencer5f016e22007-07-11 17:01:13 +0000178}
179
180void Preprocessor::DumpMacro(const MacroInfo &MI) const {
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000181 llvm::errs() << "MACRO: ";
Reid Spencer5f016e22007-07-11 17:01:13 +0000182 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
183 DumpToken(MI.getReplacementToken(i));
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000184 llvm::errs() << " ";
Reid Spencer5f016e22007-07-11 17:01:13 +0000185 }
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000186 llvm::errs() << "\n";
Reid Spencer5f016e22007-07-11 17:01:13 +0000187}
188
189void Preprocessor::PrintStats() {
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000190 llvm::errs() << "\n*** Preprocessor Stats:\n";
191 llvm::errs() << NumDirectives << " directives found:\n";
192 llvm::errs() << " " << NumDefined << " #define.\n";
193 llvm::errs() << " " << NumUndefined << " #undef.\n";
194 llvm::errs() << " #include/#include_next/#import:\n";
195 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
196 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
197 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
198 llvm::errs() << " " << NumElse << " #else/#elif.\n";
199 llvm::errs() << " " << NumEndif << " #endif.\n";
200 llvm::errs() << " " << NumPragma << " #pragma.\n";
201 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
Reid Spencer5f016e22007-07-11 17:01:13 +0000202
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000203 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
Ted Kremenekbdd30c22008-01-14 16:44:48 +0000204 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
205 << NumFastMacroExpanded << " on the fast path.\n";
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000206 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
Ted Kremenekbdd30c22008-01-14 16:44:48 +0000207 << " token paste (##) operations performed, "
208 << NumFastTokenPaste << " on the fast path.\n";
Reid Spencer5f016e22007-07-11 17:01:13 +0000209}
210
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000211Preprocessor::macro_iterator
212Preprocessor::macro_begin(bool IncludeExternalMacros) const {
213 if (IncludeExternalMacros && ExternalSource &&
Douglas Gregor88a35862010-01-04 19:18:44 +0000214 !ReadMacrosFromExternalSource) {
215 ReadMacrosFromExternalSource = true;
216 ExternalSource->ReadDefinedMacros();
217 }
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000218
219 return Macros.begin();
Douglas Gregor88a35862010-01-04 19:18:44 +0000220}
221
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000222Preprocessor::macro_iterator
223Preprocessor::macro_end(bool IncludeExternalMacros) const {
224 if (IncludeExternalMacros && ExternalSource &&
Douglas Gregor88a35862010-01-04 19:18:44 +0000225 !ReadMacrosFromExternalSource) {
226 ReadMacrosFromExternalSource = true;
227 ExternalSource->ReadDefinedMacros();
228 }
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000229
230 return Macros.end();
Douglas Gregor88a35862010-01-04 19:18:44 +0000231}
232
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000233bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
234 unsigned TruncateAtLine,
Douglas Gregor29684422009-12-02 06:49:09 +0000235 unsigned TruncateAtColumn) {
236 using llvm::MemoryBuffer;
237
238 CodeCompletionFile = File;
239
240 // Okay to clear out the code-completion point by passing NULL.
241 if (!CodeCompletionFile)
242 return false;
243
244 // Load the actual file's contents.
Douglas Gregoraa38c3d2010-03-16 19:49:24 +0000245 bool Invalid = false;
246 const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
247 if (Invalid)
Douglas Gregor29684422009-12-02 06:49:09 +0000248 return true;
249
250 // Find the byte position of the truncation point.
251 const char *Position = Buffer->getBufferStart();
252 for (unsigned Line = 1; Line < TruncateAtLine; ++Line) {
253 for (; *Position; ++Position) {
254 if (*Position != '\r' && *Position != '\n')
255 continue;
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000256
Douglas Gregor29684422009-12-02 06:49:09 +0000257 // Eat \r\n or \n\r as a single line.
258 if ((Position[1] == '\r' || Position[1] == '\n') &&
259 Position[0] != Position[1])
260 ++Position;
261 ++Position;
262 break;
263 }
264 }
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000265
Douglas Gregorb760fe82009-12-08 21:45:46 +0000266 Position += TruncateAtColumn - 1;
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000267
Douglas Gregor29684422009-12-02 06:49:09 +0000268 // Truncate the buffer.
Douglas Gregorb760fe82009-12-08 21:45:46 +0000269 if (Position < Buffer->getBufferEnd()) {
Chris Lattnera0a270c2010-04-05 22:42:27 +0000270 llvm::StringRef Data(Buffer->getBufferStart(),
271 Position-Buffer->getBufferStart());
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000272 MemoryBuffer *TruncatedBuffer
Chris Lattnera0a270c2010-04-05 22:42:27 +0000273 = MemoryBuffer::getMemBufferCopy(Data, Buffer->getBufferIdentifier());
Douglas Gregor29684422009-12-02 06:49:09 +0000274 SourceMgr.overrideFileContents(File, TruncatedBuffer);
275 }
276
277 return false;
278}
279
Douglas Gregor109ae732009-12-03 17:05:59 +0000280bool Preprocessor::isCodeCompletionFile(SourceLocation FileLoc) const {
Douglas Gregor29684422009-12-02 06:49:09 +0000281 return CodeCompletionFile && FileLoc.isFileID() &&
282 SourceMgr.getFileEntryForID(SourceMgr.getFileID(FileLoc))
283 == CodeCompletionFile;
284}
285
Reid Spencer5f016e22007-07-11 17:01:13 +0000286//===----------------------------------------------------------------------===//
287// Token Spelling
288//===----------------------------------------------------------------------===//
289
Reid Spencer5f016e22007-07-11 17:01:13 +0000290/// getSpelling() - Return the 'spelling' of this token. The spelling of a
291/// token are the characters used to represent the token in the source file
292/// after trigraph expansion and escaped-newline folding. In particular, this
293/// wants to get the true, uncanonicalized, spelling of things like digraphs
294/// UCNs, etc.
Daniel Dunbar0ff10422009-11-14 01:20:48 +0000295std::string Preprocessor::getSpelling(const Token &Tok,
296 const SourceManager &SourceMgr,
Douglas Gregor50f6af72010-03-16 05:20:39 +0000297 const LangOptions &Features,
298 bool *Invalid) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000299 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
Ted Kremenek277faca2009-01-27 00:01:05 +0000300
Reid Spencer5f016e22007-07-11 17:01:13 +0000301 // If this token contains nothing interesting, return it directly.
Douglas Gregor50f6af72010-03-16 05:20:39 +0000302 bool CharDataInvalid = false;
303 const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
304 &CharDataInvalid);
305 if (Invalid)
306 *Invalid = CharDataInvalid;
307 if (CharDataInvalid)
308 return std::string();
309
Reid Spencer5f016e22007-07-11 17:01:13 +0000310 if (!Tok.needsCleaning())
311 return std::string(TokStart, TokStart+Tok.getLength());
Mike Stump1eb44332009-09-09 15:08:12 +0000312
Reid Spencer5f016e22007-07-11 17:01:13 +0000313 std::string Result;
314 Result.reserve(Tok.getLength());
Mike Stump1eb44332009-09-09 15:08:12 +0000315
Reid Spencer5f016e22007-07-11 17:01:13 +0000316 // Otherwise, hard case, relex the characters into the string.
317 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
318 Ptr != End; ) {
319 unsigned CharSize;
320 Result.push_back(Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features));
321 Ptr += CharSize;
322 }
323 assert(Result.size() != unsigned(Tok.getLength()) &&
324 "NeedsCleaning flag set on something that didn't need cleaning!");
325 return Result;
326}
327
Daniel Dunbar0ff10422009-11-14 01:20:48 +0000328/// getSpelling() - Return the 'spelling' of this token. The spelling of a
329/// token are the characters used to represent the token in the source file
330/// after trigraph expansion and escaped-newline folding. In particular, this
331/// wants to get the true, uncanonicalized, spelling of things like digraphs
332/// UCNs, etc.
Douglas Gregor50f6af72010-03-16 05:20:39 +0000333std::string Preprocessor::getSpelling(const Token &Tok, bool *Invalid) const {
334 return getSpelling(Tok, SourceMgr, Features, Invalid);
Daniel Dunbar0ff10422009-11-14 01:20:48 +0000335}
336
Reid Spencer5f016e22007-07-11 17:01:13 +0000337/// getSpelling - This method is used to get the spelling of a token into a
338/// preallocated buffer, instead of as an std::string. The caller is required
339/// to allocate enough space for the token, which is guaranteed to be at least
340/// Tok.getLength() bytes long. The actual length of the token is returned.
341///
342/// Note that this method may do two possible things: it may either fill in
343/// the buffer specified with characters, or it may *change the input pointer*
344/// to point to a constant buffer with the data already in it (avoiding a
345/// copy). The caller is not allowed to modify the returned buffer pointer
346/// if an internal buffer is returned.
Chris Lattnerd2177732007-07-20 16:59:19 +0000347unsigned Preprocessor::getSpelling(const Token &Tok,
Douglas Gregor50f6af72010-03-16 05:20:39 +0000348 const char *&Buffer, bool *Invalid) const {
Reid Spencer5f016e22007-07-11 17:01:13 +0000349 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
Mike Stump1eb44332009-09-09 15:08:12 +0000350
Reid Spencer5f016e22007-07-11 17:01:13 +0000351 // If this token is an identifier, just return the string from the identifier
352 // table, which is very quick.
353 if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
Daniel Dunbare013d682009-10-18 20:26:12 +0000354 Buffer = II->getNameStart();
Chris Lattnere1dccae2009-01-05 19:44:41 +0000355 return II->getLength();
Reid Spencer5f016e22007-07-11 17:01:13 +0000356 }
Ted Kremenekb70e3da2009-01-08 02:47:16 +0000357
Reid Spencer5f016e22007-07-11 17:01:13 +0000358 // Otherwise, compute the start of the token in the input lexer buffer.
Chris Lattner47246be2009-01-26 19:29:26 +0000359 const char *TokStart = 0;
Mike Stump1eb44332009-09-09 15:08:12 +0000360
Chris Lattner47246be2009-01-26 19:29:26 +0000361 if (Tok.isLiteral())
362 TokStart = Tok.getLiteralData();
Mike Stump1eb44332009-09-09 15:08:12 +0000363
Douglas Gregor50f6af72010-03-16 05:20:39 +0000364 if (TokStart == 0) {
365 bool CharDataInvalid = false;
366 TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
367 if (Invalid)
368 *Invalid = CharDataInvalid;
369 if (CharDataInvalid) {
370 Buffer = "";
371 return 0;
372 }
373 }
Reid Spencer5f016e22007-07-11 17:01:13 +0000374
375 // If this token contains nothing interesting, return it directly.
376 if (!Tok.needsCleaning()) {
377 Buffer = TokStart;
378 return Tok.getLength();
379 }
Mike Stump1eb44332009-09-09 15:08:12 +0000380
Reid Spencer5f016e22007-07-11 17:01:13 +0000381 // Otherwise, hard case, relex the characters into the string.
382 char *OutBuf = const_cast<char*>(Buffer);
383 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
384 Ptr != End; ) {
385 unsigned CharSize;
386 *OutBuf++ = Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features);
387 Ptr += CharSize;
388 }
389 assert(unsigned(OutBuf-Buffer) != Tok.getLength() &&
390 "NeedsCleaning flag set on something that didn't need cleaning!");
Mike Stump1eb44332009-09-09 15:08:12 +0000391
Reid Spencer5f016e22007-07-11 17:01:13 +0000392 return OutBuf-Buffer;
393}
394
Benjamin Kramer51f5fe32010-02-27 17:05:45 +0000395/// getSpelling - This method is used to get the spelling of a token into a
396/// SmallVector. Note that the returned StringRef may not point to the
397/// supplied buffer if a copy can be avoided.
398llvm::StringRef Preprocessor::getSpelling(const Token &Tok,
Douglas Gregor50f6af72010-03-16 05:20:39 +0000399 llvm::SmallVectorImpl<char> &Buffer,
400 bool *Invalid) const {
Benjamin Kramer51f5fe32010-02-27 17:05:45 +0000401 // Try the fast path.
402 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
403 return II->getName();
404
405 // Resize the buffer if we need to copy into it.
406 if (Tok.needsCleaning())
407 Buffer.resize(Tok.getLength());
408
409 const char *Ptr = Buffer.data();
Douglas Gregor50f6af72010-03-16 05:20:39 +0000410 unsigned Len = getSpelling(Tok, Ptr, Invalid);
Benjamin Kramer51f5fe32010-02-27 17:05:45 +0000411 return llvm::StringRef(Ptr, Len);
412}
413
Reid Spencer5f016e22007-07-11 17:01:13 +0000414/// CreateString - Plop the specified string into a scratch buffer and return a
415/// location for it. If specified, the source location provides a source
416/// location for the token.
Chris Lattner47246be2009-01-26 19:29:26 +0000417void Preprocessor::CreateString(const char *Buf, unsigned Len, Token &Tok,
418 SourceLocation InstantiationLoc) {
419 Tok.setLength(Len);
Mike Stump1eb44332009-09-09 15:08:12 +0000420
Chris Lattner47246be2009-01-26 19:29:26 +0000421 const char *DestPtr;
422 SourceLocation Loc = ScratchBuf->getToken(Buf, Len, DestPtr);
Mike Stump1eb44332009-09-09 15:08:12 +0000423
Chris Lattner47246be2009-01-26 19:29:26 +0000424 if (InstantiationLoc.isValid())
Chris Lattnere7fb4842009-02-15 20:52:18 +0000425 Loc = SourceMgr.createInstantiationLoc(Loc, InstantiationLoc,
426 InstantiationLoc, Len);
Chris Lattner47246be2009-01-26 19:29:26 +0000427 Tok.setLocation(Loc);
Mike Stump1eb44332009-09-09 15:08:12 +0000428
Chris Lattner47246be2009-01-26 19:29:26 +0000429 // If this is a literal token, set the pointer data.
430 if (Tok.isLiteral())
431 Tok.setLiteralData(DestPtr);
Reid Spencer5f016e22007-07-11 17:01:13 +0000432}
433
434
Chris Lattner97ba77c2007-07-16 06:48:38 +0000435/// AdvanceToTokenCharacter - Given a location that specifies the start of a
436/// token, return a new location that specifies a character within the token.
Mike Stump1eb44332009-09-09 15:08:12 +0000437SourceLocation Preprocessor::AdvanceToTokenCharacter(SourceLocation TokStart,
Chris Lattner97ba77c2007-07-16 06:48:38 +0000438 unsigned CharNo) {
Chris Lattnerf7cf85b2009-01-16 07:36:28 +0000439 // Figure out how many physical characters away the specified instantiation
Chris Lattner97ba77c2007-07-16 06:48:38 +0000440 // character is. This needs to take into consideration newlines and
441 // trigraphs.
Douglas Gregora5430162010-03-16 20:46:42 +0000442 bool Invalid = false;
443 const char *TokPtr = SourceMgr.getCharacterData(TokStart, &Invalid);
Mike Stump1eb44332009-09-09 15:08:12 +0000444
Chris Lattner88e25242009-04-18 22:28:58 +0000445 // If they request the first char of the token, we're trivially done.
Douglas Gregora5430162010-03-16 20:46:42 +0000446 if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
Chris Lattner88e25242009-04-18 22:28:58 +0000447 return TokStart;
Mike Stump1eb44332009-09-09 15:08:12 +0000448
Chris Lattner9dc1f532007-07-20 16:37:10 +0000449 unsigned PhysOffset = 0;
Mike Stump1eb44332009-09-09 15:08:12 +0000450
Chris Lattner97ba77c2007-07-16 06:48:38 +0000451 // The usual case is that tokens don't contain anything interesting. Skip
452 // over the uninteresting characters. If a token only consists of simple
453 // chars, this method is extremely fast.
Chris Lattner88e25242009-04-18 22:28:58 +0000454 while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
455 if (CharNo == 0)
456 return TokStart.getFileLocWithOffset(PhysOffset);
Chris Lattner9dc1f532007-07-20 16:37:10 +0000457 ++TokPtr, --CharNo, ++PhysOffset;
Chris Lattner88e25242009-04-18 22:28:58 +0000458 }
Mike Stump1eb44332009-09-09 15:08:12 +0000459
Chris Lattner28c90ad2009-01-17 07:57:25 +0000460 // If we have a character that may be a trigraph or escaped newline, use a
Chris Lattner97ba77c2007-07-16 06:48:38 +0000461 // lexer to parse it correctly.
Chris Lattner88e25242009-04-18 22:28:58 +0000462 for (; CharNo; --CharNo) {
463 unsigned Size;
464 Lexer::getCharAndSizeNoWarn(TokPtr, Size, Features);
465 TokPtr += Size;
466 PhysOffset += Size;
Chris Lattner97ba77c2007-07-16 06:48:38 +0000467 }
Mike Stump1eb44332009-09-09 15:08:12 +0000468
Chris Lattner88e25242009-04-18 22:28:58 +0000469 // Final detail: if we end up on an escaped newline, we want to return the
470 // location of the actual byte of the token. For example foo\<newline>bar
471 // advanced by 3 should return the location of b, not of \\. One compounding
472 // detail of this is that the escape may be made by a trigraph.
473 if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
Ted Kremenekc0178e92010-01-29 19:38:24 +0000474 PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
Mike Stump1eb44332009-09-09 15:08:12 +0000475
Chris Lattner9dc1f532007-07-20 16:37:10 +0000476 return TokStart.getFileLocWithOffset(PhysOffset);
Chris Lattner97ba77c2007-07-16 06:48:38 +0000477}
478
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000479SourceLocation Preprocessor::getLocForEndOfToken(SourceLocation Loc,
Douglas Gregor33e9abd2010-01-22 19:49:59 +0000480 unsigned Offset) {
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000481 if (Loc.isInvalid() || !Loc.isFileID())
482 return SourceLocation();
483
Chris Lattner2c78b872009-04-14 23:22:57 +0000484 unsigned Len = Lexer::MeasureTokenLength(Loc, getSourceManager(), Features);
Douglas Gregor33e9abd2010-01-22 19:49:59 +0000485 if (Len > Offset)
486 Len = Len - Offset;
487 else
488 return Loc;
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000489
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000490 return AdvanceToTokenCharacter(Loc, Len);
491}
492
493
Chris Lattner97ba77c2007-07-16 06:48:38 +0000494
Chris Lattner53b0dab2007-10-09 22:10:18 +0000495//===----------------------------------------------------------------------===//
496// Preprocessor Initialization Methods
497//===----------------------------------------------------------------------===//
498
Chris Lattner53b0dab2007-10-09 22:10:18 +0000499
500/// EnterMainSourceFile - Enter the specified FileID as the main source file,
Nate Begeman6b616022008-01-07 04:01:26 +0000501/// which implicitly adds the builtin defines etc.
Chris Lattnere127a0d2010-04-20 20:35:58 +0000502void Preprocessor::EnterMainSourceFile() {
Chris Lattner05db4272009-02-13 19:33:24 +0000503 // We do not allow the preprocessor to reenter the main file. Doing so will
504 // cause FileID's to accumulate information from both runs (e.g. #line
505 // information) and predefined macros aren't guaranteed to be set properly.
506 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
Chris Lattner2b2453a2009-01-17 06:22:33 +0000507 FileID MainFileID = SourceMgr.getMainFileID();
Mike Stump1eb44332009-09-09 15:08:12 +0000508
Chris Lattner53b0dab2007-10-09 22:10:18 +0000509 // Enter the main file source buffer.
Chris Lattnere127a0d2010-04-20 20:35:58 +0000510 EnterSourceFile(MainFileID, 0, SourceLocation());
Mike Stump1eb44332009-09-09 15:08:12 +0000511
Douglas Gregorf4f6c9d2010-07-26 21:36:20 +0000512 // If we've been asked to skip bytes in the main file (e.g., as part of a
513 // precompiled preamble), do so now.
514 if (SkipMainFilePreamble.first > 0)
515 CurLexer->SkipBytes(SkipMainFilePreamble.first,
516 SkipMainFilePreamble.second);
517
Chris Lattnerb2832982007-11-15 19:07:47 +0000518 // Tell the header info that the main file was entered. If the file is later
519 // #imported, it won't be re-entered.
Chris Lattner2b2453a2009-01-17 06:22:33 +0000520 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
Chris Lattnerb2832982007-11-15 19:07:47 +0000521 HeaderInfo.IncrementIncludeCount(FE);
Mike Stump1eb44332009-09-09 15:08:12 +0000522
Benjamin Kramerffd6e392009-12-31 15:33:09 +0000523 // Preprocess Predefines to populate the initial preprocessor state.
Mike Stump1eb44332009-09-09 15:08:12 +0000524 llvm::MemoryBuffer *SB =
Chris Lattnera0a270c2010-04-05 22:42:27 +0000525 llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
Chris Lattner53b0dab2007-10-09 22:10:18 +0000526 assert(SB && "Cannot fail to create predefined source buffer");
Chris Lattner2b2453a2009-01-17 06:22:33 +0000527 FileID FID = SourceMgr.createFileIDForMemBuffer(SB);
528 assert(!FID.isInvalid() && "Could not create FileID for predefines?");
Mike Stump1eb44332009-09-09 15:08:12 +0000529
Chris Lattner53b0dab2007-10-09 22:10:18 +0000530 // Start parsing the predefines.
Chris Lattnere127a0d2010-04-20 20:35:58 +0000531 EnterSourceFile(FID, 0, SourceLocation());
Chris Lattner53b0dab2007-10-09 22:10:18 +0000532}
Chris Lattner97ba77c2007-07-16 06:48:38 +0000533
Daniel Dunbardbd82092010-03-23 05:09:10 +0000534void Preprocessor::EndSourceFile() {
535 // Notify the client that we reached the end of the source file.
536 if (Callbacks)
537 Callbacks->EndOfMainFile();
538}
Reid Spencer5f016e22007-07-11 17:01:13 +0000539
540//===----------------------------------------------------------------------===//
541// Lexer Event Handling.
542//===----------------------------------------------------------------------===//
543
544/// LookUpIdentifierInfo - Given a tok::identifier token, look up the
545/// identifier information for the token and install it into the token.
Chris Lattnerd2177732007-07-20 16:59:19 +0000546IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier,
Daniel Dunbarc3222092009-11-05 01:53:52 +0000547 const char *BufPtr) const {
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000548 assert(Identifier.is(tok::identifier) && "Not an identifier!");
Reid Spencer5f016e22007-07-11 17:01:13 +0000549 assert(Identifier.getIdentifierInfo() == 0 && "Identinfo already exists!");
Mike Stump1eb44332009-09-09 15:08:12 +0000550
Reid Spencer5f016e22007-07-11 17:01:13 +0000551 // Look up this token, see if it is a macro, or if it is a language keyword.
552 IdentifierInfo *II;
553 if (BufPtr && !Identifier.needsCleaning()) {
554 // No cleaning needed, just use the characters from the lexed buffer.
Daniel Dunbar3da736c2009-11-05 01:53:39 +0000555 II = getIdentifierInfo(llvm::StringRef(BufPtr, Identifier.getLength()));
Reid Spencer5f016e22007-07-11 17:01:13 +0000556 } else {
557 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000558 llvm::SmallString<64> IdentifierBuffer;
Benjamin Kramerddeea562010-02-27 13:44:12 +0000559 llvm::StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
560 II = getIdentifierInfo(CleanedStr);
Reid Spencer5f016e22007-07-11 17:01:13 +0000561 }
562 Identifier.setIdentifierInfo(II);
563 return II;
564}
565
566
567/// HandleIdentifier - This callback is invoked when the lexer reads an
568/// identifier. This callback looks up the identifier in the map and/or
569/// potentially macro expands it or turns it into a named token (like 'for').
Chris Lattner6a170eb2009-01-21 07:43:11 +0000570///
571/// Note that callers of this method are guarded by checking the
572/// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
573/// IdentifierInfo methods that compute these properties will need to change to
574/// match.
Chris Lattnerd2177732007-07-20 16:59:19 +0000575void Preprocessor::HandleIdentifier(Token &Identifier) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000576 assert(Identifier.getIdentifierInfo() &&
577 "Can't handle identifiers without identifier info!");
Mike Stump1eb44332009-09-09 15:08:12 +0000578
Reid Spencer5f016e22007-07-11 17:01:13 +0000579 IdentifierInfo &II = *Identifier.getIdentifierInfo();
580
581 // If this identifier was poisoned, and if it was not produced from a macro
582 // expansion, emit an error.
Ted Kremenek1a531572008-11-19 22:43:49 +0000583 if (II.isPoisoned() && CurPPLexer) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000584 if (&II != Ident__VA_ARGS__) // We warn about __VA_ARGS__ with poisoning.
585 Diag(Identifier, diag::err_pp_used_poisoned_id);
586 else
587 Diag(Identifier, diag::ext_pp_bad_vaargs_use);
588 }
Mike Stump1eb44332009-09-09 15:08:12 +0000589
Reid Spencer5f016e22007-07-11 17:01:13 +0000590 // If this is a macro to be expanded, do it.
Chris Lattnercc1a8752007-10-07 08:44:20 +0000591 if (MacroInfo *MI = getMacroInfo(&II)) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000592 if (!DisableMacroExpansion && !Identifier.isExpandDisabled()) {
593 if (MI->isEnabled()) {
594 if (!HandleMacroExpandedIdentifier(Identifier, MI))
595 return;
596 } else {
597 // C99 6.10.3.4p2 says that a disabled macro may never again be
598 // expanded, even if it's in a context where it could be expanded in the
599 // future.
Chris Lattnerd2177732007-07-20 16:59:19 +0000600 Identifier.setFlag(Token::DisableExpand);
Reid Spencer5f016e22007-07-11 17:01:13 +0000601 }
602 }
Reid Spencer5f016e22007-07-11 17:01:13 +0000603 }
604
605 // C++ 2.11p2: If this is an alternative representation of a C++ operator,
606 // then we act as if it is the actual operator and not the textual
607 // representation of it.
608 if (II.isCPlusPlusOperatorKeyword())
609 Identifier.setIdentifierInfo(0);
610
Reid Spencer5f016e22007-07-11 17:01:13 +0000611 // If this is an extension token, diagnose its use.
Steve Naroffb4eaf9c2008-09-02 18:50:17 +0000612 // We avoid diagnosing tokens that originate from macro definitions.
Eli Friedman2962f4d2009-04-28 03:59:15 +0000613 // FIXME: This warning is disabled in cases where it shouldn't be,
614 // like "#define TY typeof", "TY(1) x".
615 if (II.isExtensionToken() && !DisableMacroExpansion)
Reid Spencer5f016e22007-07-11 17:01:13 +0000616 Diag(Identifier, diag::ext_token_used);
617}
Douglas Gregor2e222532009-07-02 17:08:52 +0000618
619void Preprocessor::AddCommentHandler(CommentHandler *Handler) {
620 assert(Handler && "NULL comment handler");
621 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
622 CommentHandlers.end() && "Comment handler already registered");
623 CommentHandlers.push_back(Handler);
624}
625
626void Preprocessor::RemoveCommentHandler(CommentHandler *Handler) {
627 std::vector<CommentHandler *>::iterator Pos
628 = std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
629 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
630 CommentHandlers.erase(Pos);
631}
632
Chris Lattner046c2272010-01-18 22:35:47 +0000633bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
634 bool AnyPendingTokens = false;
Douglas Gregor2e222532009-07-02 17:08:52 +0000635 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
636 HEnd = CommentHandlers.end();
Chris Lattner046c2272010-01-18 22:35:47 +0000637 H != HEnd; ++H) {
638 if ((*H)->HandleComment(*this, Comment))
639 AnyPendingTokens = true;
640 }
641 if (!AnyPendingTokens || getCommentRetentionState())
642 return false;
643 Lex(result);
644 return true;
Douglas Gregor2e222532009-07-02 17:08:52 +0000645}
646
647CommentHandler::~CommentHandler() { }
Douglas Gregor94dc8f62010-03-19 16:15:56 +0000648
Douglas Gregor94dc8f62010-03-19 16:15:56 +0000649void Preprocessor::createPreprocessingRecord() {
650 if (Record)
651 return;
652
Douglas Gregorb9e1b752010-03-19 17:12:43 +0000653 Record = new PreprocessingRecord;
654 addPPCallbacks(Record);
Douglas Gregor94dc8f62010-03-19 16:15:56 +0000655}