blob: 51f729334cb9c3c7f4f18f395e6e906cc44fe87b [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 Gregorb9e1b752010-03-19 17:12:43 +000057 CurPPLexer(0), CurDirLookup(0), Callbacks(0), MacroArgCache(0), Record(0) {
Reid Spencer5f016e22007-07-11 17:01:13 +000058 ScratchBuf = new ScratchBuffer(SourceMgr);
Chris Lattnerc1f9d822009-04-13 01:29:17 +000059 CounterValue = 0; // __COUNTER__ starts at 0.
Daniel Dunbar5814e652009-11-11 21:44:21 +000060 OwnsHeaderSearch = OwnsHeaders;
Mike Stump1eb44332009-09-09 15:08:12 +000061
Reid Spencer5f016e22007-07-11 17:01:13 +000062 // Clear stats.
63 NumDirectives = NumDefined = NumUndefined = NumPragma = 0;
64 NumIf = NumElse = NumEndif = 0;
65 NumEnteredSourceFiles = 0;
66 NumMacroExpanded = NumFnMacroExpanded = NumBuiltinMacroExpanded = 0;
67 NumFastMacroExpanded = NumTokenPaste = NumFastTokenPaste = 0;
Mike Stump1eb44332009-09-09 15:08:12 +000068 MaxIncludeStackDepth = 0;
Reid Spencer5f016e22007-07-11 17:01:13 +000069 NumSkipped = 0;
70
71 // Default to discarding comments.
72 KeepComments = false;
73 KeepMacroComments = false;
Mike Stump1eb44332009-09-09 15:08:12 +000074
Reid Spencer5f016e22007-07-11 17:01:13 +000075 // Macro expansion is enabled.
76 DisableMacroExpansion = false;
77 InMacroArgs = false;
Chris Lattner6cfe7592008-03-09 02:26:03 +000078 NumCachedTokenLexers = 0;
Reid Spencer5f016e22007-07-11 17:01:13 +000079
Argyrios Kyrtzidis03db1b32008-08-10 13:15:22 +000080 CachedLexPos = 0;
81
Douglas Gregor88a35862010-01-04 19:18:44 +000082 // We haven't read anything from the external source.
83 ReadMacrosFromExternalSource = false;
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +000084
Reid Spencer5f016e22007-07-11 17:01:13 +000085 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
86 // This gets unpoisoned where it is allowed.
87 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
Mike Stump1eb44332009-09-09 15:08:12 +000088
Reid Spencer5f016e22007-07-11 17:01:13 +000089 // Initialize the pragma handlers.
Argyrios Kyrtzidis9b36c3f2010-07-13 09:07:17 +000090 PragmaHandlers = new PragmaNamespace(llvm::StringRef());
Reid Spencer5f016e22007-07-11 17:01:13 +000091 RegisterBuiltinPragmas();
Mike Stump1eb44332009-09-09 15:08:12 +000092
Reid Spencer5f016e22007-07-11 17:01:13 +000093 // Initialize builtin macros like __LINE__ and friends.
94 RegisterBuiltinMacros();
95}
96
97Preprocessor::~Preprocessor() {
Argyrios Kyrtzidis2174a4f2008-08-23 12:12:06 +000098 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
99
Reid Spencer5f016e22007-07-11 17:01:13 +0000100 while (!IncludeMacroStack.empty()) {
101 delete IncludeMacroStack.back().TheLexer;
Chris Lattner6cfe7592008-03-09 02:26:03 +0000102 delete IncludeMacroStack.back().TheTokenLexer;
Reid Spencer5f016e22007-07-11 17:01:13 +0000103 IncludeMacroStack.pop_back();
104 }
Chris Lattnercc1a8752007-10-07 08:44:20 +0000105
106 // Free any macro definitions.
107 for (llvm::DenseMap<IdentifierInfo*, MacroInfo*>::iterator I =
108 Macros.begin(), E = Macros.end(); I != E; ++I) {
Ted Kremenek0ea76722008-12-15 19:56:42 +0000109 // We don't need to free the MacroInfo objects directly. These
110 // will be released when the BumpPtrAllocator 'BP' object gets
Chris Lattner2b9d6372009-12-14 22:02:43 +0000111 // destroyed. We still need to run the dtor, however, to free
Ted Kremenek9ee7d452009-01-19 07:45:44 +0000112 // memory alocated by MacroInfo.
Chris Lattner685befe2009-02-20 22:46:43 +0000113 I->second->Destroy(BP);
Chris Lattnercc1a8752007-10-07 08:44:20 +0000114 I->first->setHasMacroDefinition(false);
115 }
Ted Kremenek83c1a6f2010-06-08 23:00:53 +0000116 for (std::vector<MacroInfo*>::iterator I = MICache.begin(),
117 E = MICache.end(); I != E; ++I) {
118 // We don't need to free the MacroInfo objects directly. These
119 // will be released when the BumpPtrAllocator 'BP' object gets
120 // destroyed. We still need to run the dtor, however, to free
121 // memory alocated by MacroInfo.
122 (*I)->Destroy(BP);
123 }
Mike Stump1eb44332009-09-09 15:08:12 +0000124
Chris Lattner9594acf2007-07-15 00:25:26 +0000125 // Free any cached macro expanders.
Chris Lattner6cfe7592008-03-09 02:26:03 +0000126 for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i)
127 delete TokenLexerCache[i];
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000128
Chris Lattner23f77e52009-12-15 01:51:03 +0000129 // Free any cached MacroArgs.
130 for (MacroArgs *ArgList = MacroArgCache; ArgList; )
131 ArgList = ArgList->deallocate();
Mike Stump1eb44332009-09-09 15:08:12 +0000132
Reid Spencer5f016e22007-07-11 17:01:13 +0000133 // Release pragma information.
134 delete PragmaHandlers;
135
136 // Delete the scratch buffer info.
137 delete ScratchBuf;
Chris Lattnereb50ed82008-03-14 06:07:05 +0000138
Daniel Dunbar5814e652009-11-11 21:44:21 +0000139 // Delete the header search info, if we own it.
140 if (OwnsHeaderSearch)
141 delete &HeaderInfo;
142
Chris Lattnereb50ed82008-03-14 06:07:05 +0000143 delete Callbacks;
Reid Spencer5f016e22007-07-11 17:01:13 +0000144}
145
Ted Kremenek337edcd2009-02-12 03:26:59 +0000146void Preprocessor::setPTHManager(PTHManager* pm) {
147 PTH.reset(pm);
Douglas Gregor52e71082009-10-16 18:18:30 +0000148 FileMgr.addStatCache(PTH->createStatCache());
Ted Kremenek337edcd2009-02-12 03:26:59 +0000149}
150
Chris Lattnerd2177732007-07-20 16:59:19 +0000151void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000152 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
153 << getSpelling(Tok) << "'";
Mike Stump1eb44332009-09-09 15:08:12 +0000154
Reid Spencer5f016e22007-07-11 17:01:13 +0000155 if (!DumpFlags) return;
Mike Stump1eb44332009-09-09 15:08:12 +0000156
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000157 llvm::errs() << "\t";
Reid Spencer5f016e22007-07-11 17:01:13 +0000158 if (Tok.isAtStartOfLine())
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000159 llvm::errs() << " [StartOfLine]";
Reid Spencer5f016e22007-07-11 17:01:13 +0000160 if (Tok.hasLeadingSpace())
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000161 llvm::errs() << " [LeadingSpace]";
Reid Spencer5f016e22007-07-11 17:01:13 +0000162 if (Tok.isExpandDisabled())
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000163 llvm::errs() << " [ExpandDisabled]";
Reid Spencer5f016e22007-07-11 17:01:13 +0000164 if (Tok.needsCleaning()) {
165 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000166 llvm::errs() << " [UnClean='" << std::string(Start, Start+Tok.getLength())
167 << "']";
Reid Spencer5f016e22007-07-11 17:01:13 +0000168 }
Mike Stump1eb44332009-09-09 15:08:12 +0000169
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000170 llvm::errs() << "\tLoc=<";
Chris Lattnerc3d8d572007-12-09 20:31:55 +0000171 DumpLocation(Tok.getLocation());
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000172 llvm::errs() << ">";
Chris Lattnerc3d8d572007-12-09 20:31:55 +0000173}
174
175void Preprocessor::DumpLocation(SourceLocation Loc) const {
Chris Lattnerb9c3f962009-01-27 07:57:44 +0000176 Loc.dump(SourceMgr);
Reid Spencer5f016e22007-07-11 17:01:13 +0000177}
178
179void Preprocessor::DumpMacro(const MacroInfo &MI) const {
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000180 llvm::errs() << "MACRO: ";
Reid Spencer5f016e22007-07-11 17:01:13 +0000181 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
182 DumpToken(MI.getReplacementToken(i));
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000183 llvm::errs() << " ";
Reid Spencer5f016e22007-07-11 17:01:13 +0000184 }
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000185 llvm::errs() << "\n";
Reid Spencer5f016e22007-07-11 17:01:13 +0000186}
187
188void Preprocessor::PrintStats() {
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000189 llvm::errs() << "\n*** Preprocessor Stats:\n";
190 llvm::errs() << NumDirectives << " directives found:\n";
191 llvm::errs() << " " << NumDefined << " #define.\n";
192 llvm::errs() << " " << NumUndefined << " #undef.\n";
193 llvm::errs() << " #include/#include_next/#import:\n";
194 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
195 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
196 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
197 llvm::errs() << " " << NumElse << " #else/#elif.\n";
198 llvm::errs() << " " << NumEndif << " #endif.\n";
199 llvm::errs() << " " << NumPragma << " #pragma.\n";
200 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
Reid Spencer5f016e22007-07-11 17:01:13 +0000201
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000202 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
Ted Kremenekbdd30c22008-01-14 16:44:48 +0000203 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
204 << NumFastMacroExpanded << " on the fast path.\n";
Benjamin Kramer6cb7c1a2009-08-23 12:08:50 +0000205 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
Ted Kremenekbdd30c22008-01-14 16:44:48 +0000206 << " token paste (##) operations performed, "
207 << NumFastTokenPaste << " on the fast path.\n";
Reid Spencer5f016e22007-07-11 17:01:13 +0000208}
209
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000210Preprocessor::macro_iterator
211Preprocessor::macro_begin(bool IncludeExternalMacros) const {
212 if (IncludeExternalMacros && ExternalSource &&
Douglas Gregor88a35862010-01-04 19:18:44 +0000213 !ReadMacrosFromExternalSource) {
214 ReadMacrosFromExternalSource = true;
215 ExternalSource->ReadDefinedMacros();
216 }
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000217
218 return Macros.begin();
Douglas Gregor88a35862010-01-04 19:18:44 +0000219}
220
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000221Preprocessor::macro_iterator
222Preprocessor::macro_end(bool IncludeExternalMacros) const {
223 if (IncludeExternalMacros && ExternalSource &&
Douglas Gregor88a35862010-01-04 19:18:44 +0000224 !ReadMacrosFromExternalSource) {
225 ReadMacrosFromExternalSource = true;
226 ExternalSource->ReadDefinedMacros();
227 }
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000228
229 return Macros.end();
Douglas Gregor88a35862010-01-04 19:18:44 +0000230}
231
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000232bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
233 unsigned TruncateAtLine,
Douglas Gregor29684422009-12-02 06:49:09 +0000234 unsigned TruncateAtColumn) {
235 using llvm::MemoryBuffer;
236
237 CodeCompletionFile = File;
238
239 // Okay to clear out the code-completion point by passing NULL.
240 if (!CodeCompletionFile)
241 return false;
242
243 // Load the actual file's contents.
Douglas Gregoraa38c3d2010-03-16 19:49:24 +0000244 bool Invalid = false;
245 const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
246 if (Invalid)
Douglas Gregor29684422009-12-02 06:49:09 +0000247 return true;
248
249 // Find the byte position of the truncation point.
250 const char *Position = Buffer->getBufferStart();
251 for (unsigned Line = 1; Line < TruncateAtLine; ++Line) {
252 for (; *Position; ++Position) {
253 if (*Position != '\r' && *Position != '\n')
254 continue;
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000255
Douglas Gregor29684422009-12-02 06:49:09 +0000256 // Eat \r\n or \n\r as a single line.
257 if ((Position[1] == '\r' || Position[1] == '\n') &&
258 Position[0] != Position[1])
259 ++Position;
260 ++Position;
261 break;
262 }
263 }
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000264
Douglas Gregorb760fe82009-12-08 21:45:46 +0000265 Position += TruncateAtColumn - 1;
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000266
Douglas Gregor29684422009-12-02 06:49:09 +0000267 // Truncate the buffer.
Douglas Gregorb760fe82009-12-08 21:45:46 +0000268 if (Position < Buffer->getBufferEnd()) {
Chris Lattnera0a270c2010-04-05 22:42:27 +0000269 llvm::StringRef Data(Buffer->getBufferStart(),
270 Position-Buffer->getBufferStart());
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000271 MemoryBuffer *TruncatedBuffer
Chris Lattnera0a270c2010-04-05 22:42:27 +0000272 = MemoryBuffer::getMemBufferCopy(Data, Buffer->getBufferIdentifier());
Douglas Gregor29684422009-12-02 06:49:09 +0000273 SourceMgr.overrideFileContents(File, TruncatedBuffer);
274 }
275
276 return false;
277}
278
Douglas Gregor109ae732009-12-03 17:05:59 +0000279bool Preprocessor::isCodeCompletionFile(SourceLocation FileLoc) const {
Douglas Gregor29684422009-12-02 06:49:09 +0000280 return CodeCompletionFile && FileLoc.isFileID() &&
281 SourceMgr.getFileEntryForID(SourceMgr.getFileID(FileLoc))
282 == CodeCompletionFile;
283}
284
Reid Spencer5f016e22007-07-11 17:01:13 +0000285//===----------------------------------------------------------------------===//
286// Token Spelling
287//===----------------------------------------------------------------------===//
288
Reid Spencer5f016e22007-07-11 17:01:13 +0000289/// getSpelling() - Return the 'spelling' of this token. The spelling of a
290/// token are the characters used to represent the token in the source file
291/// after trigraph expansion and escaped-newline folding. In particular, this
292/// wants to get the true, uncanonicalized, spelling of things like digraphs
293/// UCNs, etc.
Daniel Dunbar0ff10422009-11-14 01:20:48 +0000294std::string Preprocessor::getSpelling(const Token &Tok,
295 const SourceManager &SourceMgr,
Douglas Gregor50f6af72010-03-16 05:20:39 +0000296 const LangOptions &Features,
297 bool *Invalid) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000298 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
Ted Kremenek277faca2009-01-27 00:01:05 +0000299
Reid Spencer5f016e22007-07-11 17:01:13 +0000300 // If this token contains nothing interesting, return it directly.
Douglas Gregor50f6af72010-03-16 05:20:39 +0000301 bool CharDataInvalid = false;
302 const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
303 &CharDataInvalid);
304 if (Invalid)
305 *Invalid = CharDataInvalid;
306 if (CharDataInvalid)
307 return std::string();
308
Reid Spencer5f016e22007-07-11 17:01:13 +0000309 if (!Tok.needsCleaning())
310 return std::string(TokStart, TokStart+Tok.getLength());
Mike Stump1eb44332009-09-09 15:08:12 +0000311
Reid Spencer5f016e22007-07-11 17:01:13 +0000312 std::string Result;
313 Result.reserve(Tok.getLength());
Mike Stump1eb44332009-09-09 15:08:12 +0000314
Reid Spencer5f016e22007-07-11 17:01:13 +0000315 // Otherwise, hard case, relex the characters into the string.
316 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
317 Ptr != End; ) {
318 unsigned CharSize;
319 Result.push_back(Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features));
320 Ptr += CharSize;
321 }
322 assert(Result.size() != unsigned(Tok.getLength()) &&
323 "NeedsCleaning flag set on something that didn't need cleaning!");
324 return Result;
325}
326
Daniel Dunbar0ff10422009-11-14 01:20:48 +0000327/// getSpelling() - Return the 'spelling' of this token. The spelling of a
328/// token are the characters used to represent the token in the source file
329/// after trigraph expansion and escaped-newline folding. In particular, this
330/// wants to get the true, uncanonicalized, spelling of things like digraphs
331/// UCNs, etc.
Douglas Gregor50f6af72010-03-16 05:20:39 +0000332std::string Preprocessor::getSpelling(const Token &Tok, bool *Invalid) const {
333 return getSpelling(Tok, SourceMgr, Features, Invalid);
Daniel Dunbar0ff10422009-11-14 01:20:48 +0000334}
335
Reid Spencer5f016e22007-07-11 17:01:13 +0000336/// getSpelling - This method is used to get the spelling of a token into a
337/// preallocated buffer, instead of as an std::string. The caller is required
338/// to allocate enough space for the token, which is guaranteed to be at least
339/// Tok.getLength() bytes long. The actual length of the token is returned.
340///
341/// Note that this method may do two possible things: it may either fill in
342/// the buffer specified with characters, or it may *change the input pointer*
343/// to point to a constant buffer with the data already in it (avoiding a
344/// copy). The caller is not allowed to modify the returned buffer pointer
345/// if an internal buffer is returned.
Chris Lattnerd2177732007-07-20 16:59:19 +0000346unsigned Preprocessor::getSpelling(const Token &Tok,
Douglas Gregor50f6af72010-03-16 05:20:39 +0000347 const char *&Buffer, bool *Invalid) const {
Reid Spencer5f016e22007-07-11 17:01:13 +0000348 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
Mike Stump1eb44332009-09-09 15:08:12 +0000349
Reid Spencer5f016e22007-07-11 17:01:13 +0000350 // If this token is an identifier, just return the string from the identifier
351 // table, which is very quick.
352 if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
Daniel Dunbare013d682009-10-18 20:26:12 +0000353 Buffer = II->getNameStart();
Chris Lattnere1dccae2009-01-05 19:44:41 +0000354 return II->getLength();
Reid Spencer5f016e22007-07-11 17:01:13 +0000355 }
Ted Kremenekb70e3da2009-01-08 02:47:16 +0000356
Reid Spencer5f016e22007-07-11 17:01:13 +0000357 // Otherwise, compute the start of the token in the input lexer buffer.
Chris Lattner47246be2009-01-26 19:29:26 +0000358 const char *TokStart = 0;
Mike Stump1eb44332009-09-09 15:08:12 +0000359
Chris Lattner47246be2009-01-26 19:29:26 +0000360 if (Tok.isLiteral())
361 TokStart = Tok.getLiteralData();
Mike Stump1eb44332009-09-09 15:08:12 +0000362
Douglas Gregor50f6af72010-03-16 05:20:39 +0000363 if (TokStart == 0) {
364 bool CharDataInvalid = false;
365 TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
366 if (Invalid)
367 *Invalid = CharDataInvalid;
368 if (CharDataInvalid) {
369 Buffer = "";
370 return 0;
371 }
372 }
Reid Spencer5f016e22007-07-11 17:01:13 +0000373
374 // If this token contains nothing interesting, return it directly.
375 if (!Tok.needsCleaning()) {
376 Buffer = TokStart;
377 return Tok.getLength();
378 }
Mike Stump1eb44332009-09-09 15:08:12 +0000379
Reid Spencer5f016e22007-07-11 17:01:13 +0000380 // Otherwise, hard case, relex the characters into the string.
381 char *OutBuf = const_cast<char*>(Buffer);
382 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
383 Ptr != End; ) {
384 unsigned CharSize;
385 *OutBuf++ = Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features);
386 Ptr += CharSize;
387 }
388 assert(unsigned(OutBuf-Buffer) != Tok.getLength() &&
389 "NeedsCleaning flag set on something that didn't need cleaning!");
Mike Stump1eb44332009-09-09 15:08:12 +0000390
Reid Spencer5f016e22007-07-11 17:01:13 +0000391 return OutBuf-Buffer;
392}
393
Benjamin Kramer51f5fe32010-02-27 17:05:45 +0000394/// getSpelling - This method is used to get the spelling of a token into a
395/// SmallVector. Note that the returned StringRef may not point to the
396/// supplied buffer if a copy can be avoided.
397llvm::StringRef Preprocessor::getSpelling(const Token &Tok,
Douglas Gregor50f6af72010-03-16 05:20:39 +0000398 llvm::SmallVectorImpl<char> &Buffer,
399 bool *Invalid) const {
Benjamin Kramer51f5fe32010-02-27 17:05:45 +0000400 // Try the fast path.
401 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
402 return II->getName();
403
404 // Resize the buffer if we need to copy into it.
405 if (Tok.needsCleaning())
406 Buffer.resize(Tok.getLength());
407
408 const char *Ptr = Buffer.data();
Douglas Gregor50f6af72010-03-16 05:20:39 +0000409 unsigned Len = getSpelling(Tok, Ptr, Invalid);
Benjamin Kramer51f5fe32010-02-27 17:05:45 +0000410 return llvm::StringRef(Ptr, Len);
411}
412
Reid Spencer5f016e22007-07-11 17:01:13 +0000413/// CreateString - Plop the specified string into a scratch buffer and return a
414/// location for it. If specified, the source location provides a source
415/// location for the token.
Chris Lattner47246be2009-01-26 19:29:26 +0000416void Preprocessor::CreateString(const char *Buf, unsigned Len, Token &Tok,
417 SourceLocation InstantiationLoc) {
418 Tok.setLength(Len);
Mike Stump1eb44332009-09-09 15:08:12 +0000419
Chris Lattner47246be2009-01-26 19:29:26 +0000420 const char *DestPtr;
421 SourceLocation Loc = ScratchBuf->getToken(Buf, Len, DestPtr);
Mike Stump1eb44332009-09-09 15:08:12 +0000422
Chris Lattner47246be2009-01-26 19:29:26 +0000423 if (InstantiationLoc.isValid())
Chris Lattnere7fb4842009-02-15 20:52:18 +0000424 Loc = SourceMgr.createInstantiationLoc(Loc, InstantiationLoc,
425 InstantiationLoc, Len);
Chris Lattner47246be2009-01-26 19:29:26 +0000426 Tok.setLocation(Loc);
Mike Stump1eb44332009-09-09 15:08:12 +0000427
Chris Lattner47246be2009-01-26 19:29:26 +0000428 // If this is a literal token, set the pointer data.
429 if (Tok.isLiteral())
430 Tok.setLiteralData(DestPtr);
Reid Spencer5f016e22007-07-11 17:01:13 +0000431}
432
433
Chris Lattner97ba77c2007-07-16 06:48:38 +0000434/// AdvanceToTokenCharacter - Given a location that specifies the start of a
435/// token, return a new location that specifies a character within the token.
Mike Stump1eb44332009-09-09 15:08:12 +0000436SourceLocation Preprocessor::AdvanceToTokenCharacter(SourceLocation TokStart,
Chris Lattner97ba77c2007-07-16 06:48:38 +0000437 unsigned CharNo) {
Chris Lattnerf7cf85b2009-01-16 07:36:28 +0000438 // Figure out how many physical characters away the specified instantiation
Chris Lattner97ba77c2007-07-16 06:48:38 +0000439 // character is. This needs to take into consideration newlines and
440 // trigraphs.
Douglas Gregora5430162010-03-16 20:46:42 +0000441 bool Invalid = false;
442 const char *TokPtr = SourceMgr.getCharacterData(TokStart, &Invalid);
Mike Stump1eb44332009-09-09 15:08:12 +0000443
Chris Lattner88e25242009-04-18 22:28:58 +0000444 // If they request the first char of the token, we're trivially done.
Douglas Gregora5430162010-03-16 20:46:42 +0000445 if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
Chris Lattner88e25242009-04-18 22:28:58 +0000446 return TokStart;
Mike Stump1eb44332009-09-09 15:08:12 +0000447
Chris Lattner9dc1f532007-07-20 16:37:10 +0000448 unsigned PhysOffset = 0;
Mike Stump1eb44332009-09-09 15:08:12 +0000449
Chris Lattner97ba77c2007-07-16 06:48:38 +0000450 // The usual case is that tokens don't contain anything interesting. Skip
451 // over the uninteresting characters. If a token only consists of simple
452 // chars, this method is extremely fast.
Chris Lattner88e25242009-04-18 22:28:58 +0000453 while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
454 if (CharNo == 0)
455 return TokStart.getFileLocWithOffset(PhysOffset);
Chris Lattner9dc1f532007-07-20 16:37:10 +0000456 ++TokPtr, --CharNo, ++PhysOffset;
Chris Lattner88e25242009-04-18 22:28:58 +0000457 }
Mike Stump1eb44332009-09-09 15:08:12 +0000458
Chris Lattner28c90ad2009-01-17 07:57:25 +0000459 // If we have a character that may be a trigraph or escaped newline, use a
Chris Lattner97ba77c2007-07-16 06:48:38 +0000460 // lexer to parse it correctly.
Chris Lattner88e25242009-04-18 22:28:58 +0000461 for (; CharNo; --CharNo) {
462 unsigned Size;
463 Lexer::getCharAndSizeNoWarn(TokPtr, Size, Features);
464 TokPtr += Size;
465 PhysOffset += Size;
Chris Lattner97ba77c2007-07-16 06:48:38 +0000466 }
Mike Stump1eb44332009-09-09 15:08:12 +0000467
Chris Lattner88e25242009-04-18 22:28:58 +0000468 // Final detail: if we end up on an escaped newline, we want to return the
469 // location of the actual byte of the token. For example foo\<newline>bar
470 // advanced by 3 should return the location of b, not of \\. One compounding
471 // detail of this is that the escape may be made by a trigraph.
472 if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
Ted Kremenekc0178e92010-01-29 19:38:24 +0000473 PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
Mike Stump1eb44332009-09-09 15:08:12 +0000474
Chris Lattner9dc1f532007-07-20 16:37:10 +0000475 return TokStart.getFileLocWithOffset(PhysOffset);
Chris Lattner97ba77c2007-07-16 06:48:38 +0000476}
477
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000478SourceLocation Preprocessor::getLocForEndOfToken(SourceLocation Loc,
Douglas Gregor33e9abd2010-01-22 19:49:59 +0000479 unsigned Offset) {
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000480 if (Loc.isInvalid() || !Loc.isFileID())
481 return SourceLocation();
482
Chris Lattner2c78b872009-04-14 23:22:57 +0000483 unsigned Len = Lexer::MeasureTokenLength(Loc, getSourceManager(), Features);
Douglas Gregor33e9abd2010-01-22 19:49:59 +0000484 if (Len > Offset)
485 Len = Len - Offset;
486 else
487 return Loc;
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000488
Douglas Gregorb2fb6de2009-02-27 17:53:17 +0000489 return AdvanceToTokenCharacter(Loc, Len);
490}
491
492
Chris Lattner97ba77c2007-07-16 06:48:38 +0000493
Chris Lattner53b0dab2007-10-09 22:10:18 +0000494//===----------------------------------------------------------------------===//
495// Preprocessor Initialization Methods
496//===----------------------------------------------------------------------===//
497
Chris Lattner53b0dab2007-10-09 22:10:18 +0000498
499/// EnterMainSourceFile - Enter the specified FileID as the main source file,
Nate Begeman6b616022008-01-07 04:01:26 +0000500/// which implicitly adds the builtin defines etc.
Chris Lattnere127a0d2010-04-20 20:35:58 +0000501void Preprocessor::EnterMainSourceFile() {
Chris Lattner05db4272009-02-13 19:33:24 +0000502 // We do not allow the preprocessor to reenter the main file. Doing so will
503 // cause FileID's to accumulate information from both runs (e.g. #line
504 // information) and predefined macros aren't guaranteed to be set properly.
505 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
Chris Lattner2b2453a2009-01-17 06:22:33 +0000506 FileID MainFileID = SourceMgr.getMainFileID();
Mike Stump1eb44332009-09-09 15:08:12 +0000507
Chris Lattner53b0dab2007-10-09 22:10:18 +0000508 // Enter the main file source buffer.
Chris Lattnere127a0d2010-04-20 20:35:58 +0000509 EnterSourceFile(MainFileID, 0, SourceLocation());
Mike Stump1eb44332009-09-09 15:08:12 +0000510
Chris Lattnerb2832982007-11-15 19:07:47 +0000511 // Tell the header info that the main file was entered. If the file is later
512 // #imported, it won't be re-entered.
Chris Lattner2b2453a2009-01-17 06:22:33 +0000513 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
Chris Lattnerb2832982007-11-15 19:07:47 +0000514 HeaderInfo.IncrementIncludeCount(FE);
Mike Stump1eb44332009-09-09 15:08:12 +0000515
Benjamin Kramerffd6e392009-12-31 15:33:09 +0000516 // Preprocess Predefines to populate the initial preprocessor state.
Mike Stump1eb44332009-09-09 15:08:12 +0000517 llvm::MemoryBuffer *SB =
Chris Lattnera0a270c2010-04-05 22:42:27 +0000518 llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
Chris Lattner53b0dab2007-10-09 22:10:18 +0000519 assert(SB && "Cannot fail to create predefined source buffer");
Chris Lattner2b2453a2009-01-17 06:22:33 +0000520 FileID FID = SourceMgr.createFileIDForMemBuffer(SB);
521 assert(!FID.isInvalid() && "Could not create FileID for predefines?");
Mike Stump1eb44332009-09-09 15:08:12 +0000522
Chris Lattner53b0dab2007-10-09 22:10:18 +0000523 // Start parsing the predefines.
Chris Lattnere127a0d2010-04-20 20:35:58 +0000524 EnterSourceFile(FID, 0, SourceLocation());
Chris Lattner53b0dab2007-10-09 22:10:18 +0000525}
Chris Lattner97ba77c2007-07-16 06:48:38 +0000526
Daniel Dunbardbd82092010-03-23 05:09:10 +0000527void Preprocessor::EndSourceFile() {
528 // Notify the client that we reached the end of the source file.
529 if (Callbacks)
530 Callbacks->EndOfMainFile();
531}
Reid Spencer5f016e22007-07-11 17:01:13 +0000532
533//===----------------------------------------------------------------------===//
534// Lexer Event Handling.
535//===----------------------------------------------------------------------===//
536
537/// LookUpIdentifierInfo - Given a tok::identifier token, look up the
538/// identifier information for the token and install it into the token.
Chris Lattnerd2177732007-07-20 16:59:19 +0000539IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier,
Daniel Dunbarc3222092009-11-05 01:53:52 +0000540 const char *BufPtr) const {
Chris Lattner22f6bbc2007-10-09 18:02:16 +0000541 assert(Identifier.is(tok::identifier) && "Not an identifier!");
Reid Spencer5f016e22007-07-11 17:01:13 +0000542 assert(Identifier.getIdentifierInfo() == 0 && "Identinfo already exists!");
Mike Stump1eb44332009-09-09 15:08:12 +0000543
Reid Spencer5f016e22007-07-11 17:01:13 +0000544 // Look up this token, see if it is a macro, or if it is a language keyword.
545 IdentifierInfo *II;
546 if (BufPtr && !Identifier.needsCleaning()) {
547 // No cleaning needed, just use the characters from the lexed buffer.
Daniel Dunbar3da736c2009-11-05 01:53:39 +0000548 II = getIdentifierInfo(llvm::StringRef(BufPtr, Identifier.getLength()));
Reid Spencer5f016e22007-07-11 17:01:13 +0000549 } else {
550 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
Kovarththanan Rajaratnam19357542010-03-13 10:17:05 +0000551 llvm::SmallString<64> IdentifierBuffer;
Benjamin Kramerddeea562010-02-27 13:44:12 +0000552 llvm::StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
553 II = getIdentifierInfo(CleanedStr);
Reid Spencer5f016e22007-07-11 17:01:13 +0000554 }
555 Identifier.setIdentifierInfo(II);
556 return II;
557}
558
559
560/// HandleIdentifier - This callback is invoked when the lexer reads an
561/// identifier. This callback looks up the identifier in the map and/or
562/// potentially macro expands it or turns it into a named token (like 'for').
Chris Lattner6a170eb2009-01-21 07:43:11 +0000563///
564/// Note that callers of this method are guarded by checking the
565/// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
566/// IdentifierInfo methods that compute these properties will need to change to
567/// match.
Chris Lattnerd2177732007-07-20 16:59:19 +0000568void Preprocessor::HandleIdentifier(Token &Identifier) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000569 assert(Identifier.getIdentifierInfo() &&
570 "Can't handle identifiers without identifier info!");
Mike Stump1eb44332009-09-09 15:08:12 +0000571
Reid Spencer5f016e22007-07-11 17:01:13 +0000572 IdentifierInfo &II = *Identifier.getIdentifierInfo();
573
574 // If this identifier was poisoned, and if it was not produced from a macro
575 // expansion, emit an error.
Ted Kremenek1a531572008-11-19 22:43:49 +0000576 if (II.isPoisoned() && CurPPLexer) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000577 if (&II != Ident__VA_ARGS__) // We warn about __VA_ARGS__ with poisoning.
578 Diag(Identifier, diag::err_pp_used_poisoned_id);
579 else
580 Diag(Identifier, diag::ext_pp_bad_vaargs_use);
581 }
Mike Stump1eb44332009-09-09 15:08:12 +0000582
Reid Spencer5f016e22007-07-11 17:01:13 +0000583 // If this is a macro to be expanded, do it.
Chris Lattnercc1a8752007-10-07 08:44:20 +0000584 if (MacroInfo *MI = getMacroInfo(&II)) {
Reid Spencer5f016e22007-07-11 17:01:13 +0000585 if (!DisableMacroExpansion && !Identifier.isExpandDisabled()) {
586 if (MI->isEnabled()) {
587 if (!HandleMacroExpandedIdentifier(Identifier, MI))
588 return;
589 } else {
590 // C99 6.10.3.4p2 says that a disabled macro may never again be
591 // expanded, even if it's in a context where it could be expanded in the
592 // future.
Chris Lattnerd2177732007-07-20 16:59:19 +0000593 Identifier.setFlag(Token::DisableExpand);
Reid Spencer5f016e22007-07-11 17:01:13 +0000594 }
595 }
Reid Spencer5f016e22007-07-11 17:01:13 +0000596 }
597
598 // C++ 2.11p2: If this is an alternative representation of a C++ operator,
599 // then we act as if it is the actual operator and not the textual
600 // representation of it.
601 if (II.isCPlusPlusOperatorKeyword())
602 Identifier.setIdentifierInfo(0);
603
Reid Spencer5f016e22007-07-11 17:01:13 +0000604 // If this is an extension token, diagnose its use.
Steve Naroffb4eaf9c2008-09-02 18:50:17 +0000605 // We avoid diagnosing tokens that originate from macro definitions.
Eli Friedman2962f4d2009-04-28 03:59:15 +0000606 // FIXME: This warning is disabled in cases where it shouldn't be,
607 // like "#define TY typeof", "TY(1) x".
608 if (II.isExtensionToken() && !DisableMacroExpansion)
Reid Spencer5f016e22007-07-11 17:01:13 +0000609 Diag(Identifier, diag::ext_token_used);
610}
Douglas Gregor2e222532009-07-02 17:08:52 +0000611
612void Preprocessor::AddCommentHandler(CommentHandler *Handler) {
613 assert(Handler && "NULL comment handler");
614 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
615 CommentHandlers.end() && "Comment handler already registered");
616 CommentHandlers.push_back(Handler);
617}
618
619void Preprocessor::RemoveCommentHandler(CommentHandler *Handler) {
620 std::vector<CommentHandler *>::iterator Pos
621 = std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
622 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
623 CommentHandlers.erase(Pos);
624}
625
Chris Lattner046c2272010-01-18 22:35:47 +0000626bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
627 bool AnyPendingTokens = false;
Douglas Gregor2e222532009-07-02 17:08:52 +0000628 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
629 HEnd = CommentHandlers.end();
Chris Lattner046c2272010-01-18 22:35:47 +0000630 H != HEnd; ++H) {
631 if ((*H)->HandleComment(*this, Comment))
632 AnyPendingTokens = true;
633 }
634 if (!AnyPendingTokens || getCommentRetentionState())
635 return false;
636 Lex(result);
637 return true;
Douglas Gregor2e222532009-07-02 17:08:52 +0000638}
639
640CommentHandler::~CommentHandler() { }
Douglas Gregor94dc8f62010-03-19 16:15:56 +0000641
Douglas Gregor94dc8f62010-03-19 16:15:56 +0000642void Preprocessor::createPreprocessingRecord() {
643 if (Record)
644 return;
645
Douglas Gregorb9e1b752010-03-19 17:12:43 +0000646 Record = new PreprocessingRecord;
647 addPPCallbacks(Record);
Douglas Gregor94dc8f62010-03-19 16:15:56 +0000648}