Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 1 | //===--- PrintPreprocessedOutput.cpp - Implement the -E mode --------------===// |
| 2 | // |
| 3 | // The LLVM Compiler Infrastructure |
| 4 | // |
Chris Lattner | 0bc735f | 2007-12-29 19:59:25 +0000 | [diff] [blame] | 5 | // This file is distributed under the University of Illinois Open Source |
| 6 | // License. See LICENSE.TXT for details. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 7 | // |
| 8 | //===----------------------------------------------------------------------===// |
| 9 | // |
| 10 | // This code simply runs the preprocessor on the input file and prints out the |
| 11 | // result. This is the traditional behavior of the -E option. |
| 12 | // |
| 13 | //===----------------------------------------------------------------------===// |
| 14 | |
| 15 | #include "clang.h" |
Chris Lattner | f73903a | 2009-02-06 06:45:26 +0000 | [diff] [blame^] | 16 | #include "clang/Lex/MacroInfo.h" |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 17 | #include "clang/Lex/PPCallbacks.h" |
| 18 | #include "clang/Lex/Preprocessor.h" |
| 19 | #include "clang/Lex/Pragma.h" |
| 20 | #include "clang/Basic/SourceManager.h" |
Chris Lattner | 5db17c9 | 2008-04-08 04:16:20 +0000 | [diff] [blame] | 21 | #include "clang/Basic/Diagnostic.h" |
Chris Lattner | d8e3083 | 2007-07-24 06:57:14 +0000 | [diff] [blame] | 22 | #include "llvm/ADT/SmallString.h" |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 23 | #include "llvm/ADT/StringExtras.h" |
Chris Lattner | 5db17c9 | 2008-04-08 04:16:20 +0000 | [diff] [blame] | 24 | #include "llvm/System/Path.h" |
| 25 | #include "llvm/Support/CommandLine.h" |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 26 | #include "llvm/Config/config.h" |
Chris Lattner | dceb6a7 | 2008-08-17 01:47:12 +0000 | [diff] [blame] | 27 | #include "llvm/Support/raw_ostream.h" |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 28 | #include <cstdio> |
| 29 | using namespace clang; |
| 30 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 31 | //===----------------------------------------------------------------------===// |
| 32 | // Preprocessed token printer |
| 33 | //===----------------------------------------------------------------------===// |
| 34 | |
| 35 | static llvm::cl::opt<bool> |
| 36 | DisableLineMarkers("P", llvm::cl::desc("Disable linemarker output in -E mode")); |
| 37 | static llvm::cl::opt<bool> |
| 38 | EnableCommentOutput("C", llvm::cl::desc("Enable comment output in -E mode")); |
| 39 | static llvm::cl::opt<bool> |
| 40 | EnableMacroCommentOutput("CC", |
| 41 | llvm::cl::desc("Enable comment output in -E mode, " |
| 42 | "even from macro expansions")); |
Chris Lattner | f73903a | 2009-02-06 06:45:26 +0000 | [diff] [blame^] | 43 | static llvm::cl::opt<bool> |
| 44 | DumpMacros("dM", llvm::cl::desc("Print macro definitions in -E mode instead of" |
| 45 | " normal output")); |
| 46 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 47 | |
| 48 | namespace { |
| 49 | class PrintPPOutputPPCallbacks : public PPCallbacks { |
| 50 | Preprocessor &PP; |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 51 | public: |
| 52 | llvm::raw_ostream &OS; |
| 53 | private: |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 54 | unsigned CurLine; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 55 | bool EmittedTokensOnThisLine; |
Chris Lattner | 9d72851 | 2008-10-27 01:19:25 +0000 | [diff] [blame] | 56 | SrcMgr::CharacteristicKind FileType; |
Chris Lattner | d8e3083 | 2007-07-24 06:57:14 +0000 | [diff] [blame] | 57 | llvm::SmallString<512> CurFilename; |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 58 | bool Initialized; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 59 | public: |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 60 | PrintPPOutputPPCallbacks(Preprocessor &pp, llvm::raw_ostream &os) |
| 61 | : PP(pp), OS(os) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 62 | CurLine = 0; |
Chris Lattner | d8e3083 | 2007-07-24 06:57:14 +0000 | [diff] [blame] | 63 | CurFilename += "<uninit>"; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 64 | EmittedTokensOnThisLine = false; |
Chris Lattner | 0b9e736 | 2008-09-26 21:18:42 +0000 | [diff] [blame] | 65 | FileType = SrcMgr::C_User; |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 66 | Initialized = false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 67 | } |
| 68 | |
| 69 | void SetEmittedTokensOnThisLine() { EmittedTokensOnThisLine = true; } |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 70 | bool hasEmittedTokensOnThisLine() const { return EmittedTokensOnThisLine; } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 71 | |
| 72 | virtual void FileChanged(SourceLocation Loc, FileChangeReason Reason, |
Chris Lattner | 9d72851 | 2008-10-27 01:19:25 +0000 | [diff] [blame] | 73 | SrcMgr::CharacteristicKind FileType); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 74 | virtual void Ident(SourceLocation Loc, const std::string &str); |
Chris Lattner | c7d945d | 2009-01-16 19:25:54 +0000 | [diff] [blame] | 75 | virtual void PragmaComment(SourceLocation Loc, const IdentifierInfo *Kind, |
| 76 | const std::string &Str); |
| 77 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 78 | |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 79 | bool HandleFirstTokOnLine(Token &Tok); |
| 80 | bool MoveToLine(SourceLocation Loc); |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 81 | bool AvoidConcat(const Token &PrevTok, const Token &Tok); |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 82 | void WriteLineInfo(unsigned LineNo, const char *Extra=0, unsigned ExtraLen=0); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 83 | }; |
Chris Lattner | 5db17c9 | 2008-04-08 04:16:20 +0000 | [diff] [blame] | 84 | } // end anonymous namespace |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 85 | |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 86 | void PrintPPOutputPPCallbacks::WriteLineInfo(unsigned LineNo, |
| 87 | const char *Extra, |
| 88 | unsigned ExtraLen) { |
| 89 | if (EmittedTokensOnThisLine) { |
| 90 | OS << '\n'; |
| 91 | EmittedTokensOnThisLine = false; |
| 92 | } |
| 93 | |
| 94 | OS << '#' << ' ' << LineNo << ' ' << '"'; |
| 95 | OS.write(&CurFilename[0], CurFilename.size()); |
| 96 | OS << '"'; |
| 97 | |
| 98 | if (ExtraLen) |
| 99 | OS.write(Extra, ExtraLen); |
| 100 | |
Chris Lattner | 0b9e736 | 2008-09-26 21:18:42 +0000 | [diff] [blame] | 101 | if (FileType == SrcMgr::C_System) |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 102 | OS.write(" 3", 2); |
Chris Lattner | 0b9e736 | 2008-09-26 21:18:42 +0000 | [diff] [blame] | 103 | else if (FileType == SrcMgr::C_ExternCSystem) |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 104 | OS.write(" 3 4", 4); |
| 105 | OS << '\n'; |
| 106 | } |
| 107 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 108 | /// MoveToLine - Move the output to the source line specified by the location |
| 109 | /// object. We can do this by emitting some number of \n's, or be emitting a |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 110 | /// #line directive. This returns false if already at the specified line, true |
| 111 | /// if some newlines were emitted. |
| 112 | bool PrintPPOutputPPCallbacks::MoveToLine(SourceLocation Loc) { |
Chris Lattner | f7cf85b | 2009-01-16 07:36:28 +0000 | [diff] [blame] | 113 | unsigned LineNo = PP.getSourceManager().getInstantiationLineNumber(Loc); |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 114 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 115 | if (DisableLineMarkers) { |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 116 | if (LineNo == CurLine) return false; |
| 117 | |
| 118 | CurLine = LineNo; |
| 119 | |
| 120 | if (!EmittedTokensOnThisLine) |
| 121 | return true; |
| 122 | |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 123 | OS << '\n'; |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 124 | EmittedTokensOnThisLine = false; |
| 125 | return true; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 126 | } |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 127 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 128 | // If this line is "close enough" to the original line, just print newlines, |
| 129 | // otherwise print a #line directive. |
Daniel Dunbar | fd96684 | 2008-09-26 01:13:35 +0000 | [diff] [blame] | 130 | if (LineNo-CurLine <= 8) { |
Chris Lattner | 822f940 | 2007-07-23 05:14:05 +0000 | [diff] [blame] | 131 | if (LineNo-CurLine == 1) |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 132 | OS << '\n'; |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 133 | else if (LineNo == CurLine) |
Chris Lattner | f7cf85b | 2009-01-16 07:36:28 +0000 | [diff] [blame] | 134 | return false; // Spelling line moved, but instantiation line didn't. |
Chris Lattner | 822f940 | 2007-07-23 05:14:05 +0000 | [diff] [blame] | 135 | else { |
| 136 | const char *NewLines = "\n\n\n\n\n\n\n\n"; |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 137 | OS.write(NewLines, LineNo-CurLine); |
Chris Lattner | 822f940 | 2007-07-23 05:14:05 +0000 | [diff] [blame] | 138 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 139 | } else { |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 140 | WriteLineInfo(LineNo, 0, 0); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 141 | } |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 142 | |
| 143 | CurLine = LineNo; |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 144 | return true; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 145 | } |
| 146 | |
| 147 | |
| 148 | /// FileChanged - Whenever the preprocessor enters or exits a #include file |
| 149 | /// it invokes this handler. Update our conception of the current source |
| 150 | /// position. |
| 151 | void PrintPPOutputPPCallbacks::FileChanged(SourceLocation Loc, |
| 152 | FileChangeReason Reason, |
Chris Lattner | 9d72851 | 2008-10-27 01:19:25 +0000 | [diff] [blame] | 153 | SrcMgr::CharacteristicKind NewFileType) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 154 | // Unless we are exiting a #include, make sure to skip ahead to the line the |
| 155 | // #include directive was at. |
| 156 | SourceManager &SourceMgr = PP.getSourceManager(); |
| 157 | if (Reason == PPCallbacks::EnterFile) { |
Chris Lattner | 71d8bfb | 2009-01-30 18:44:17 +0000 | [diff] [blame] | 158 | SourceLocation IncludeLoc = SourceMgr.getPresumedLoc(Loc).getIncludeLoc(); |
| 159 | if (IncludeLoc.isValid()) |
| 160 | MoveToLine(IncludeLoc); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 161 | } else if (Reason == PPCallbacks::SystemHeaderPragma) { |
| 162 | MoveToLine(Loc); |
| 163 | |
| 164 | // TODO GCC emits the # directive for this directive on the line AFTER the |
| 165 | // directive and emits a bunch of spaces that aren't needed. Emulate this |
| 166 | // strange behavior. |
| 167 | } |
| 168 | |
Chris Lattner | f7cf85b | 2009-01-16 07:36:28 +0000 | [diff] [blame] | 169 | Loc = SourceMgr.getInstantiationLoc(Loc); |
Chris Lattner | 30fc933 | 2009-02-04 01:06:56 +0000 | [diff] [blame] | 170 | CurLine = SourceMgr.getInstantiationLineNumber(Loc); |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 171 | |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 172 | if (DisableLineMarkers) return; |
| 173 | |
Chris Lattner | d8e3083 | 2007-07-24 06:57:14 +0000 | [diff] [blame] | 174 | CurFilename.clear(); |
Chris Lattner | b9c3f96 | 2009-01-27 07:57:44 +0000 | [diff] [blame] | 175 | CurFilename += SourceMgr.getPresumedLoc(Loc).getFilename(); |
Chris Lattner | d8e3083 | 2007-07-24 06:57:14 +0000 | [diff] [blame] | 176 | Lexer::Stringify(CurFilename); |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 177 | FileType = NewFileType; |
| 178 | |
| 179 | if (!Initialized) { |
| 180 | WriteLineInfo(CurLine); |
| 181 | Initialized = true; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 182 | } |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 183 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 184 | switch (Reason) { |
| 185 | case PPCallbacks::EnterFile: |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 186 | WriteLineInfo(CurLine, " 1", 2); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 187 | break; |
| 188 | case PPCallbacks::ExitFile: |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 189 | WriteLineInfo(CurLine, " 2", 2); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 190 | break; |
Daniel Dunbar | 737bdb4 | 2008-09-05 03:22:57 +0000 | [diff] [blame] | 191 | case PPCallbacks::SystemHeaderPragma: |
| 192 | case PPCallbacks::RenameFile: |
| 193 | WriteLineInfo(CurLine); |
| 194 | break; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 195 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 196 | } |
| 197 | |
Chris Lattner | c7d945d | 2009-01-16 19:25:54 +0000 | [diff] [blame] | 198 | /// Ident - Handle #ident directives when read by the preprocessor. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 199 | /// |
| 200 | void PrintPPOutputPPCallbacks::Ident(SourceLocation Loc, const std::string &S) { |
| 201 | MoveToLine(Loc); |
| 202 | |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 203 | OS.write("#ident ", strlen("#ident ")); |
| 204 | OS.write(&S[0], S.size()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 205 | EmittedTokensOnThisLine = true; |
| 206 | } |
| 207 | |
Chris Lattner | c7d945d | 2009-01-16 19:25:54 +0000 | [diff] [blame] | 208 | void PrintPPOutputPPCallbacks::PragmaComment(SourceLocation Loc, |
| 209 | const IdentifierInfo *Kind, |
| 210 | const std::string &Str) { |
| 211 | MoveToLine(Loc); |
| 212 | OS << "#pragma comment(" << Kind->getName(); |
| 213 | |
| 214 | if (!Str.empty()) { |
| 215 | OS << ", \""; |
| 216 | |
| 217 | for (unsigned i = 0, e = Str.size(); i != e; ++i) { |
| 218 | unsigned char Char = Str[i]; |
Chris Lattner | 52a3e9e | 2009-01-16 22:13:37 +0000 | [diff] [blame] | 219 | if (isprint(Char) && Char != '\\' && Char != '"') |
Chris Lattner | c7d945d | 2009-01-16 19:25:54 +0000 | [diff] [blame] | 220 | OS << (char)Char; |
| 221 | else // Output anything hard as an octal escape. |
| 222 | OS << '\\' |
| 223 | << (char)('0'+ ((Char >> 6) & 7)) |
| 224 | << (char)('0'+ ((Char >> 3) & 7)) |
| 225 | << (char)('0'+ ((Char >> 0) & 7)); |
| 226 | } |
| 227 | OS << '"'; |
| 228 | } |
| 229 | |
| 230 | OS << ')'; |
| 231 | EmittedTokensOnThisLine = true; |
| 232 | } |
| 233 | |
| 234 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 235 | /// HandleFirstTokOnLine - When emitting a preprocessed file in -E mode, this |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 236 | /// is called for the first token on each new line. If this really is the start |
| 237 | /// of a new logical line, handle it and return true, otherwise return false. |
| 238 | /// This may not be the start of a logical line because the "start of line" |
Chris Lattner | f7cf85b | 2009-01-16 07:36:28 +0000 | [diff] [blame] | 239 | /// marker is set for spelling lines, not instantiation ones. |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 240 | bool PrintPPOutputPPCallbacks::HandleFirstTokOnLine(Token &Tok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 241 | // Figure out what line we went to and insert the appropriate number of |
| 242 | // newline characters. |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 243 | if (!MoveToLine(Tok.getLocation())) |
| 244 | return false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 245 | |
| 246 | // Print out space characters so that the first token on a line is |
| 247 | // indented for easy reading. |
Chris Lattner | 9dc1f53 | 2007-07-20 16:37:10 +0000 | [diff] [blame] | 248 | const SourceManager &SourceMgr = PP.getSourceManager(); |
Chris Lattner | f7cf85b | 2009-01-16 07:36:28 +0000 | [diff] [blame] | 249 | unsigned ColNo = SourceMgr.getInstantiationColumnNumber(Tok.getLocation()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 250 | |
| 251 | // This hack prevents stuff like: |
| 252 | // #define HASH # |
| 253 | // HASH define foo bar |
| 254 | // From having the # character end up at column 1, which makes it so it |
| 255 | // is not handled as a #define next time through the preprocessor if in |
| 256 | // -fpreprocessed mode. |
Chris Lattner | 057aaf6 | 2007-10-09 18:03:42 +0000 | [diff] [blame] | 257 | if (ColNo <= 1 && Tok.is(tok::hash)) |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 258 | OS << ' '; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 259 | |
| 260 | // Otherwise, indent the appropriate number of spaces. |
| 261 | for (; ColNo > 1; --ColNo) |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 262 | OS << ' '; |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 263 | |
| 264 | return true; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 265 | } |
| 266 | |
| 267 | namespace { |
| 268 | struct UnknownPragmaHandler : public PragmaHandler { |
| 269 | const char *Prefix; |
| 270 | PrintPPOutputPPCallbacks *Callbacks; |
| 271 | |
| 272 | UnknownPragmaHandler(const char *prefix, PrintPPOutputPPCallbacks *callbacks) |
| 273 | : PragmaHandler(0), Prefix(prefix), Callbacks(callbacks) {} |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 274 | virtual void HandlePragma(Preprocessor &PP, Token &PragmaTok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 275 | // Figure out what line we went to and insert the appropriate number of |
| 276 | // newline characters. |
| 277 | Callbacks->MoveToLine(PragmaTok.getLocation()); |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 278 | Callbacks->OS.write(Prefix, strlen(Prefix)); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 279 | |
| 280 | // Read and print all of the pragma tokens. |
Chris Lattner | 057aaf6 | 2007-10-09 18:03:42 +0000 | [diff] [blame] | 281 | while (PragmaTok.isNot(tok::eom)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 282 | if (PragmaTok.hasLeadingSpace()) |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 283 | Callbacks->OS << ' '; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 284 | std::string TokSpell = PP.getSpelling(PragmaTok); |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 285 | Callbacks->OS.write(&TokSpell[0], TokSpell.size()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 286 | PP.LexUnexpandedToken(PragmaTok); |
| 287 | } |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 288 | Callbacks->OS << '\n'; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 289 | } |
| 290 | }; |
| 291 | } // end anonymous namespace |
| 292 | |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 293 | |
| 294 | enum AvoidConcatInfo { |
| 295 | /// By default, a token never needs to avoid concatenation. Most tokens (e.g. |
| 296 | /// ',', ')', etc) don't cause a problem when concatenated. |
| 297 | aci_never_avoid_concat = 0, |
| 298 | |
| 299 | /// aci_custom_firstchar - AvoidConcat contains custom code to handle this |
| 300 | /// token's requirements, and it needs to know the first character of the |
| 301 | /// token. |
| 302 | aci_custom_firstchar = 1, |
| 303 | |
| 304 | /// aci_custom - AvoidConcat contains custom code to handle this token's |
| 305 | /// requirements, but it doesn't need to know the first character of the |
| 306 | /// token. |
| 307 | aci_custom = 2, |
| 308 | |
| 309 | /// aci_avoid_equal - Many tokens cannot be safely followed by an '=' |
| 310 | /// character. For example, "<<" turns into "<<=" when followed by an =. |
| 311 | aci_avoid_equal = 4 |
| 312 | }; |
| 313 | |
| 314 | /// This array contains information for each token on what action to take when |
| 315 | /// avoiding concatenation of tokens in the AvoidConcat method. |
| 316 | static char TokenInfo[tok::NUM_TOKENS]; |
| 317 | |
| 318 | /// InitAvoidConcatTokenInfo - Tokens that must avoid concatenation should be |
| 319 | /// marked by this function. |
| 320 | static void InitAvoidConcatTokenInfo() { |
| 321 | // These tokens have custom code in AvoidConcat. |
| 322 | TokenInfo[tok::identifier ] |= aci_custom; |
| 323 | TokenInfo[tok::numeric_constant] |= aci_custom_firstchar; |
| 324 | TokenInfo[tok::period ] |= aci_custom_firstchar; |
| 325 | TokenInfo[tok::amp ] |= aci_custom_firstchar; |
| 326 | TokenInfo[tok::plus ] |= aci_custom_firstchar; |
| 327 | TokenInfo[tok::minus ] |= aci_custom_firstchar; |
| 328 | TokenInfo[tok::slash ] |= aci_custom_firstchar; |
| 329 | TokenInfo[tok::less ] |= aci_custom_firstchar; |
| 330 | TokenInfo[tok::greater ] |= aci_custom_firstchar; |
| 331 | TokenInfo[tok::pipe ] |= aci_custom_firstchar; |
| 332 | TokenInfo[tok::percent ] |= aci_custom_firstchar; |
| 333 | TokenInfo[tok::colon ] |= aci_custom_firstchar; |
| 334 | TokenInfo[tok::hash ] |= aci_custom_firstchar; |
| 335 | TokenInfo[tok::arrow ] |= aci_custom_firstchar; |
| 336 | |
| 337 | // These tokens change behavior if followed by an '='. |
| 338 | TokenInfo[tok::amp ] |= aci_avoid_equal; // &= |
| 339 | TokenInfo[tok::plus ] |= aci_avoid_equal; // += |
| 340 | TokenInfo[tok::minus ] |= aci_avoid_equal; // -= |
| 341 | TokenInfo[tok::slash ] |= aci_avoid_equal; // /= |
| 342 | TokenInfo[tok::less ] |= aci_avoid_equal; // <= |
| 343 | TokenInfo[tok::greater ] |= aci_avoid_equal; // >= |
| 344 | TokenInfo[tok::pipe ] |= aci_avoid_equal; // |= |
| 345 | TokenInfo[tok::percent ] |= aci_avoid_equal; // %= |
| 346 | TokenInfo[tok::star ] |= aci_avoid_equal; // *= |
| 347 | TokenInfo[tok::exclaim ] |= aci_avoid_equal; // != |
| 348 | TokenInfo[tok::lessless ] |= aci_avoid_equal; // <<= |
| 349 | TokenInfo[tok::greaterequal] |= aci_avoid_equal; // >>= |
| 350 | TokenInfo[tok::caret ] |= aci_avoid_equal; // ^= |
| 351 | TokenInfo[tok::equal ] |= aci_avoid_equal; // == |
| 352 | } |
| 353 | |
Chris Lattner | b1a17ae | 2008-01-15 05:22:14 +0000 | [diff] [blame] | 354 | /// StartsWithL - Return true if the spelling of this token starts with 'L'. |
Chris Lattner | fdc0d3c | 2008-01-15 05:14:19 +0000 | [diff] [blame] | 355 | static bool StartsWithL(const Token &Tok, Preprocessor &PP) { |
Chris Lattner | fdc0d3c | 2008-01-15 05:14:19 +0000 | [diff] [blame] | 356 | if (!Tok.needsCleaning()) { |
| 357 | SourceManager &SrcMgr = PP.getSourceManager(); |
Chris Lattner | df7c17a | 2009-01-16 07:00:02 +0000 | [diff] [blame] | 358 | return *SrcMgr.getCharacterData(SrcMgr.getSpellingLoc(Tok.getLocation())) |
Chris Lattner | fdc0d3c | 2008-01-15 05:14:19 +0000 | [diff] [blame] | 359 | == 'L'; |
| 360 | } |
| 361 | |
| 362 | if (Tok.getLength() < 256) { |
Chris Lattner | b1a17ae | 2008-01-15 05:22:14 +0000 | [diff] [blame] | 363 | char Buffer[256]; |
Chris Lattner | fdc0d3c | 2008-01-15 05:14:19 +0000 | [diff] [blame] | 364 | const char *TokPtr = Buffer; |
| 365 | PP.getSpelling(Tok, TokPtr); |
| 366 | return TokPtr[0] == 'L'; |
| 367 | } |
| 368 | |
| 369 | return PP.getSpelling(Tok)[0] == 'L'; |
| 370 | } |
| 371 | |
Chris Lattner | b1a17ae | 2008-01-15 05:22:14 +0000 | [diff] [blame] | 372 | /// IsIdentifierL - Return true if the spelling of this token is literally 'L'. |
| 373 | static bool IsIdentifierL(const Token &Tok, Preprocessor &PP) { |
| 374 | if (!Tok.needsCleaning()) { |
| 375 | if (Tok.getLength() != 1) |
| 376 | return false; |
| 377 | SourceManager &SrcMgr = PP.getSourceManager(); |
Chris Lattner | df7c17a | 2009-01-16 07:00:02 +0000 | [diff] [blame] | 378 | return *SrcMgr.getCharacterData(SrcMgr.getSpellingLoc(Tok.getLocation())) |
Chris Lattner | b1a17ae | 2008-01-15 05:22:14 +0000 | [diff] [blame] | 379 | == 'L'; |
| 380 | } |
| 381 | |
| 382 | if (Tok.getLength() < 256) { |
| 383 | char Buffer[256]; |
| 384 | const char *TokPtr = Buffer; |
| 385 | if (PP.getSpelling(Tok, TokPtr) != 1) |
| 386 | return false; |
| 387 | return TokPtr[0] == 'L'; |
| 388 | } |
| 389 | |
| 390 | return PP.getSpelling(Tok) == "L"; |
| 391 | } |
| 392 | |
| 393 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 394 | /// AvoidConcat - If printing PrevTok immediately followed by Tok would cause |
| 395 | /// the two individual tokens to be lexed as a single token, return true (which |
| 396 | /// causes a space to be printed between them). This allows the output of -E |
| 397 | /// mode to be lexed to the same token stream as lexing the input directly |
| 398 | /// would. |
| 399 | /// |
| 400 | /// This code must conservatively return true if it doesn't want to be 100% |
| 401 | /// accurate. This will cause the output to include extra space characters, but |
| 402 | /// the resulting output won't have incorrect concatenations going on. Examples |
| 403 | /// include "..", which we print with a space between, because we don't want to |
| 404 | /// track enough to tell "x.." from "...". |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 405 | bool PrintPPOutputPPCallbacks::AvoidConcat(const Token &PrevTok, |
| 406 | const Token &Tok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 407 | char Buffer[256]; |
| 408 | |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 409 | tok::TokenKind PrevKind = PrevTok.getKind(); |
| 410 | if (PrevTok.getIdentifierInfo()) // Language keyword or named operator. |
| 411 | PrevKind = tok::identifier; |
| 412 | |
| 413 | // Look up information on when we should avoid concatenation with prevtok. |
| 414 | unsigned ConcatInfo = TokenInfo[PrevKind]; |
| 415 | |
| 416 | // If prevtok never causes a problem for anything after it, return quickly. |
| 417 | if (ConcatInfo == 0) return false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 418 | |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 419 | if (ConcatInfo & aci_avoid_equal) { |
| 420 | // If the next token is '=' or '==', avoid concatenation. |
Chris Lattner | 057aaf6 | 2007-10-09 18:03:42 +0000 | [diff] [blame] | 421 | if (Tok.is(tok::equal) || Tok.is(tok::equalequal)) |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 422 | return true; |
Chris Lattner | b638a30 | 2007-07-23 23:21:34 +0000 | [diff] [blame] | 423 | ConcatInfo &= ~aci_avoid_equal; |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 424 | } |
| 425 | |
| 426 | if (ConcatInfo == 0) return false; |
| 427 | |
| 428 | |
| 429 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 430 | // Basic algorithm: we look at the first character of the second token, and |
| 431 | // determine whether it, if appended to the first token, would form (or would |
| 432 | // contribute) to a larger token if concatenated. |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 433 | char FirstChar = 0; |
| 434 | if (ConcatInfo & aci_custom) { |
| 435 | // If the token does not need to know the first character, don't get it. |
| 436 | } else if (IdentifierInfo *II = Tok.getIdentifierInfo()) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 437 | // Avoid spelling identifiers, the most common form of token. |
| 438 | FirstChar = II->getName()[0]; |
Chris Lattner | b19f5e8 | 2007-07-23 05:18:42 +0000 | [diff] [blame] | 439 | } else if (!Tok.needsCleaning()) { |
Chris Lattner | 33116d6 | 2009-01-26 19:33:54 +0000 | [diff] [blame] | 440 | if (Tok.isLiteral() && Tok.getLiteralData()) { |
| 441 | FirstChar = *Tok.getLiteralData(); |
| 442 | } else { |
| 443 | SourceManager &SrcMgr = PP.getSourceManager(); |
| 444 | FirstChar = |
| 445 | *SrcMgr.getCharacterData(SrcMgr.getSpellingLoc(Tok.getLocation())); |
| 446 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 447 | } else if (Tok.getLength() < 256) { |
| 448 | const char *TokPtr = Buffer; |
| 449 | PP.getSpelling(Tok, TokPtr); |
| 450 | FirstChar = TokPtr[0]; |
| 451 | } else { |
| 452 | FirstChar = PP.getSpelling(Tok)[0]; |
| 453 | } |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 454 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 455 | switch (PrevKind) { |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 456 | default: assert(0 && "InitAvoidConcatTokenInfo built wrong"); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 457 | case tok::identifier: // id+id or id+number or id+L"foo". |
Chris Lattner | 057aaf6 | 2007-10-09 18:03:42 +0000 | [diff] [blame] | 458 | if (Tok.is(tok::numeric_constant) || Tok.getIdentifierInfo() || |
| 459 | Tok.is(tok::wide_string_literal) /* || |
| 460 | Tok.is(tok::wide_char_literal)*/) |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 461 | return true; |
Chris Lattner | fdc0d3c | 2008-01-15 05:14:19 +0000 | [diff] [blame] | 462 | |
| 463 | // If this isn't identifier + string, we're done. |
| 464 | if (Tok.isNot(tok::char_constant) && Tok.isNot(tok::string_literal)) |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 465 | return false; |
| 466 | |
| 467 | // FIXME: need a wide_char_constant! |
Chris Lattner | fdc0d3c | 2008-01-15 05:14:19 +0000 | [diff] [blame] | 468 | |
| 469 | // If the string was a wide string L"foo" or wide char L'f', it would concat |
| 470 | // with the previous identifier into fooL"bar". Avoid this. |
| 471 | if (StartsWithL(Tok, PP)) |
| 472 | return true; |
| 473 | |
Chris Lattner | b1a17ae | 2008-01-15 05:22:14 +0000 | [diff] [blame] | 474 | // Otherwise, this is a narrow character or string. If the *identifier* is |
| 475 | // a literal 'L', avoid pasting L "foo" -> L"foo". |
| 476 | return IsIdentifierL(PrevTok, PP); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 477 | case tok::numeric_constant: |
Chris Lattner | 057aaf6 | 2007-10-09 18:03:42 +0000 | [diff] [blame] | 478 | return isalnum(FirstChar) || Tok.is(tok::numeric_constant) || |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 479 | FirstChar == '+' || FirstChar == '-' || FirstChar == '.'; |
| 480 | case tok::period: // ..., .*, .1234 |
Chris Lattner | d7a7c00 | 2009-01-11 19:48:19 +0000 | [diff] [blame] | 481 | return FirstChar == '.' || isdigit(FirstChar) || |
| 482 | (FirstChar == '*' && PP.getLangOptions().CPlusPlus); |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 483 | case tok::amp: // && |
| 484 | return FirstChar == '&'; |
| 485 | case tok::plus: // ++ |
| 486 | return FirstChar == '+'; |
| 487 | case tok::minus: // --, ->, ->* |
| 488 | return FirstChar == '-' || FirstChar == '>'; |
| 489 | case tok::slash: //, /*, // |
| 490 | return FirstChar == '*' || FirstChar == '/'; |
| 491 | case tok::less: // <<, <<=, <:, <% |
| 492 | return FirstChar == '<' || FirstChar == ':' || FirstChar == '%'; |
| 493 | case tok::greater: // >>, >>= |
| 494 | return FirstChar == '>'; |
| 495 | case tok::pipe: // || |
| 496 | return FirstChar == '|'; |
| 497 | case tok::percent: // %>, %: |
Chris Lattner | d7a7c00 | 2009-01-11 19:48:19 +0000 | [diff] [blame] | 498 | return (FirstChar == '>' || FirstChar == ':') && |
| 499 | PP.getLangOptions().Digraphs; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 500 | case tok::colon: // ::, :> |
Chris Lattner | d7a7c00 | 2009-01-11 19:48:19 +0000 | [diff] [blame] | 501 | return (FirstChar == ':' && PP.getLangOptions().CPlusPlus) || |
| 502 | (FirstChar == '>' && PP.getLangOptions().Digraphs); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 503 | case tok::hash: // ##, #@, %:%: |
| 504 | return FirstChar == '#' || FirstChar == '@' || FirstChar == '%'; |
| 505 | case tok::arrow: // ->* |
| 506 | return FirstChar == '*'; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 507 | } |
| 508 | } |
| 509 | |
Chris Lattner | 59076ab | 2009-02-06 05:56:11 +0000 | [diff] [blame] | 510 | static void PrintPreprocessedTokens(Preprocessor &PP, Token &Tok, |
| 511 | PrintPPOutputPPCallbacks *Callbacks, |
| 512 | llvm::raw_ostream &OS) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 513 | char Buffer[256]; |
Chris Lattner | 59076ab | 2009-02-06 05:56:11 +0000 | [diff] [blame] | 514 | Token PrevTok; |
Chris Lattner | 6f688e1 | 2007-10-10 20:45:16 +0000 | [diff] [blame] | 515 | while (1) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 516 | |
| 517 | // If this token is at the start of a line, emit newlines if needed. |
Chris Lattner | 5f18032 | 2007-12-09 21:11:08 +0000 | [diff] [blame] | 518 | if (Tok.isAtStartOfLine() && Callbacks->HandleFirstTokOnLine(Tok)) { |
| 519 | // done. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 520 | } else if (Tok.hasLeadingSpace() || |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 521 | // If we haven't emitted a token on this line yet, PrevTok isn't |
| 522 | // useful to look at and no concatenation could happen anyway. |
Chris Lattner | b638a30 | 2007-07-23 23:21:34 +0000 | [diff] [blame] | 523 | (Callbacks->hasEmittedTokensOnThisLine() && |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 524 | // Don't print "-" next to "-", it would form "--". |
| 525 | Callbacks->AvoidConcat(PrevTok, Tok))) { |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 526 | OS << ' '; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 527 | } |
| 528 | |
Chris Lattner | 2933f41 | 2007-07-23 06:14:36 +0000 | [diff] [blame] | 529 | if (IdentifierInfo *II = Tok.getIdentifierInfo()) { |
Chris Lattner | 33116d6 | 2009-01-26 19:33:54 +0000 | [diff] [blame] | 530 | OS.write(II->getName(), II->getLength()); |
| 531 | } else if (Tok.isLiteral() && !Tok.needsCleaning() && |
| 532 | Tok.getLiteralData()) { |
| 533 | OS.write(Tok.getLiteralData(), Tok.getLength()); |
Chris Lattner | 2933f41 | 2007-07-23 06:14:36 +0000 | [diff] [blame] | 534 | } else if (Tok.getLength() < 256) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 535 | const char *TokPtr = Buffer; |
| 536 | unsigned Len = PP.getSpelling(Tok, TokPtr); |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 537 | OS.write(TokPtr, Len); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 538 | } else { |
| 539 | std::string S = PP.getSpelling(Tok); |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 540 | OS.write(&S[0], S.size()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 541 | } |
| 542 | Callbacks->SetEmittedTokensOnThisLine(); |
Chris Lattner | 6f688e1 | 2007-10-10 20:45:16 +0000 | [diff] [blame] | 543 | |
| 544 | if (Tok.is(tok::eof)) break; |
Chris Lattner | 59076ab | 2009-02-06 05:56:11 +0000 | [diff] [blame] | 545 | |
Chris Lattner | 6f688e1 | 2007-10-10 20:45:16 +0000 | [diff] [blame] | 546 | PrevTok = Tok; |
| 547 | PP.Lex(Tok); |
| 548 | } |
Chris Lattner | 59076ab | 2009-02-06 05:56:11 +0000 | [diff] [blame] | 549 | } |
| 550 | |
Chris Lattner | f73903a | 2009-02-06 06:45:26 +0000 | [diff] [blame^] | 551 | /// PrintMacroDefinition - Print a macro definition in a form that will be |
| 552 | /// properly accepted back as a definition. |
| 553 | static void PrintMacroDefinition(IdentifierInfo &II, const MacroInfo &MI, |
| 554 | Preprocessor &PP, llvm::raw_ostream &OS) { |
| 555 | // Ignore computed macros like __LINE__ and friends. |
| 556 | if (MI.isBuiltinMacro()) return; |
| 557 | OS << "#define " << II.getName(); |
| 558 | |
| 559 | if (MI.isFunctionLike()) { |
| 560 | OS << '('; |
| 561 | if (MI.arg_empty()) |
| 562 | ; |
| 563 | else if (MI.getNumArgs() == 1) |
| 564 | OS << (*MI.arg_begin())->getName(); |
| 565 | else { |
| 566 | MacroInfo::arg_iterator AI = MI.arg_begin(), E = MI.arg_end(); |
| 567 | OS << (*AI++)->getName(); |
| 568 | while (AI != E) |
| 569 | OS << ',' << (*AI++)->getName(); |
| 570 | } |
| 571 | |
| 572 | if (MI.isVariadic()) { |
| 573 | if (!MI.arg_empty()) |
| 574 | OS << ','; |
| 575 | OS << "..."; |
| 576 | } |
| 577 | OS << ')'; |
| 578 | } |
| 579 | |
| 580 | // GCC always emits a space, even if the macro body is empty. However, do not |
| 581 | // want to emit two spaces if the first token has a leading space. |
| 582 | if (MI.tokens_empty() || !MI.tokens_begin()->hasLeadingSpace()) |
| 583 | OS << ' '; |
| 584 | |
| 585 | for (MacroInfo::tokens_iterator I = MI.tokens_begin(), E = MI.tokens_end(); |
| 586 | I != E; ++I) { |
| 587 | if (I->hasLeadingSpace()) |
| 588 | OS << ' '; |
| 589 | OS << PP.getSpelling(*I); |
| 590 | } |
| 591 | OS << "\n"; |
| 592 | } |
| 593 | |
Chris Lattner | 59076ab | 2009-02-06 05:56:11 +0000 | [diff] [blame] | 594 | |
| 595 | /// DoPrintPreprocessedInput - This implements -E mode. |
| 596 | /// |
| 597 | void clang::DoPrintPreprocessedInput(Preprocessor &PP, |
| 598 | const std::string &OutFile) { |
| 599 | // Inform the preprocessor whether we want it to retain comments or not, due |
| 600 | // to -C or -CC. |
| 601 | PP.SetCommentRetentionState(EnableCommentOutput, EnableMacroCommentOutput); |
| 602 | InitAvoidConcatTokenInfo(); |
| 603 | |
| 604 | |
| 605 | // Open the output buffer. |
| 606 | std::string Err; |
| 607 | llvm::raw_fd_ostream OS(OutFile.empty() ? "-" : OutFile.c_str(), false, Err); |
| 608 | if (!Err.empty()) { |
| 609 | fprintf(stderr, "%s\n", Err.c_str()); |
| 610 | exit(1); |
| 611 | } |
| 612 | |
| 613 | OS.SetBufferSize(64*1024); |
| 614 | |
Chris Lattner | f73903a | 2009-02-06 06:45:26 +0000 | [diff] [blame^] | 615 | if (DumpMacros) { |
| 616 | // -dM mode just scans and ignores all tokens in the files, then dumps out |
| 617 | // the macro table at the end. |
| 618 | PP.EnterMainSourceFile(); |
| 619 | |
| 620 | Token Tok; |
| 621 | do PP.Lex(Tok); |
| 622 | while (Tok.isNot(tok::eof)); |
| 623 | |
| 624 | for (Preprocessor::macro_iterator I = PP.macro_begin(), E = PP.macro_end(); |
| 625 | I != E; ++I) |
| 626 | PrintMacroDefinition(*I->first, *I->second, PP, OS); |
| 627 | |
| 628 | } else { |
| 629 | PrintPPOutputPPCallbacks *Callbacks; |
| 630 | Callbacks = new PrintPPOutputPPCallbacks(PP, OS); |
| 631 | PP.AddPragmaHandler(0, new UnknownPragmaHandler("#pragma", Callbacks)); |
| 632 | PP.AddPragmaHandler("GCC", new UnknownPragmaHandler("#pragma GCC", |
| 633 | Callbacks)); |
Chris Lattner | 59076ab | 2009-02-06 05:56:11 +0000 | [diff] [blame] | 634 | |
Chris Lattner | f73903a | 2009-02-06 06:45:26 +0000 | [diff] [blame^] | 635 | PP.setPPCallbacks(Callbacks); |
Chris Lattner | 59076ab | 2009-02-06 05:56:11 +0000 | [diff] [blame] | 636 | |
Chris Lattner | f73903a | 2009-02-06 06:45:26 +0000 | [diff] [blame^] | 637 | // After we have configured the preprocessor, enter the main file. |
| 638 | PP.EnterMainSourceFile(); |
Chris Lattner | 59076ab | 2009-02-06 05:56:11 +0000 | [diff] [blame] | 639 | |
Chris Lattner | f73903a | 2009-02-06 06:45:26 +0000 | [diff] [blame^] | 640 | // Consume all of the tokens that come from the predefines buffer. Those |
| 641 | // should not be emitted into the output and are guaranteed to be at the |
| 642 | // start. |
| 643 | const SourceManager &SourceMgr = PP.getSourceManager(); |
| 644 | Token Tok; |
| 645 | do PP.Lex(Tok); |
| 646 | while (Tok.isNot(tok::eof) && Tok.getLocation().isFileID() && |
| 647 | !strcmp(SourceMgr.getPresumedLoc(Tok.getLocation()).getFilename(), |
| 648 | "<predefines>")); |
Chris Lattner | 59076ab | 2009-02-06 05:56:11 +0000 | [diff] [blame] | 649 | |
Chris Lattner | f73903a | 2009-02-06 06:45:26 +0000 | [diff] [blame^] | 650 | // Read all the preprocessed tokens, printing them out to the stream. |
| 651 | PrintPreprocessedTokens(PP, Tok, Callbacks, OS); |
| 652 | OS << '\n'; |
| 653 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 654 | |
Chris Lattner | 76b3a72 | 2008-08-17 07:07:01 +0000 | [diff] [blame] | 655 | // Flush the ostream. |
| 656 | OS.flush(); |
Chris Lattner | e96de3e | 2008-08-17 03:12:02 +0000 | [diff] [blame] | 657 | |
| 658 | // If an error occurred, remove the output file. |
| 659 | if (PP.getDiagnostics().hasErrorOccurred() && !OutFile.empty()) |
| 660 | llvm::sys::Path(OutFile).eraseFromDisk(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 661 | } |
| 662 | |