MacroArgs.cpp revision 321369
1//===--- MacroArgs.cpp - Formal argument info for Macros ------------------===// 2// 3// The LLVM Compiler Infrastructure 4// 5// This file is distributed under the University of Illinois Open Source 6// License. See LICENSE.TXT for details. 7// 8//===----------------------------------------------------------------------===// 9// 10// This file implements the MacroArgs interface. 11// 12//===----------------------------------------------------------------------===// 13 14#include "clang/Lex/MacroArgs.h" 15#include "clang/Lex/LexDiagnostic.h" 16#include "clang/Lex/MacroInfo.h" 17#include "clang/Lex/Preprocessor.h" 18#include "llvm/ADT/SmallString.h" 19#include "llvm/Support/SaveAndRestore.h" 20#include <algorithm> 21 22using namespace clang; 23 24/// MacroArgs ctor function - This destroys the vector passed in. 25MacroArgs *MacroArgs::create(const MacroInfo *MI, 26 ArrayRef<Token> UnexpArgTokens, 27 bool VarargsElided, Preprocessor &PP) { 28 assert(MI->isFunctionLike() && 29 "Can't have args for an object-like macro!"); 30 MacroArgs **ResultEnt = nullptr; 31 unsigned ClosestMatch = ~0U; 32 33 // See if we have an entry with a big enough argument list to reuse on the 34 // free list. If so, reuse it. 35 for (MacroArgs **Entry = &PP.MacroArgCache; *Entry; 36 Entry = &(*Entry)->ArgCache) 37 if ((*Entry)->NumUnexpArgTokens >= UnexpArgTokens.size() && 38 (*Entry)->NumUnexpArgTokens < ClosestMatch) { 39 ResultEnt = Entry; 40 41 // If we have an exact match, use it. 42 if ((*Entry)->NumUnexpArgTokens == UnexpArgTokens.size()) 43 break; 44 // Otherwise, use the best fit. 45 ClosestMatch = (*Entry)->NumUnexpArgTokens; 46 } 47 48 MacroArgs *Result; 49 if (!ResultEnt) { 50 // Allocate memory for a MacroArgs object with the lexer tokens at the end. 51 Result = (MacroArgs *)malloc(sizeof(MacroArgs) + 52 UnexpArgTokens.size() * sizeof(Token)); 53 // Construct the MacroArgs object. 54 new (Result) 55 MacroArgs(UnexpArgTokens.size(), VarargsElided, MI->getNumParams()); 56 } else { 57 Result = *ResultEnt; 58 // Unlink this node from the preprocessors singly linked list. 59 *ResultEnt = Result->ArgCache; 60 Result->NumUnexpArgTokens = UnexpArgTokens.size(); 61 Result->VarargsElided = VarargsElided; 62 Result->NumMacroArgs = MI->getNumParams(); 63 } 64 65 // Copy the actual unexpanded tokens to immediately after the result ptr. 66 if (!UnexpArgTokens.empty()) 67 std::copy(UnexpArgTokens.begin(), UnexpArgTokens.end(), 68 const_cast<Token*>(Result->getUnexpArgument(0))); 69 70 return Result; 71} 72 73/// destroy - Destroy and deallocate the memory for this object. 74/// 75void MacroArgs::destroy(Preprocessor &PP) { 76 StringifiedArgs.clear(); 77 78 // Don't clear PreExpArgTokens, just clear the entries. Clearing the entries 79 // would deallocate the element vectors. 80 for (unsigned i = 0, e = PreExpArgTokens.size(); i != e; ++i) 81 PreExpArgTokens[i].clear(); 82 83 // Add this to the preprocessor's free list. 84 ArgCache = PP.MacroArgCache; 85 PP.MacroArgCache = this; 86} 87 88/// deallocate - This should only be called by the Preprocessor when managing 89/// its freelist. 90MacroArgs *MacroArgs::deallocate() { 91 MacroArgs *Next = ArgCache; 92 93 // Run the dtor to deallocate the vectors. 94 this->~MacroArgs(); 95 // Release the memory for the object. 96 free(this); 97 98 return Next; 99} 100 101 102/// getArgLength - Given a pointer to an expanded or unexpanded argument, 103/// return the number of tokens, not counting the EOF, that make up the 104/// argument. 105unsigned MacroArgs::getArgLength(const Token *ArgPtr) { 106 unsigned NumArgTokens = 0; 107 for (; ArgPtr->isNot(tok::eof); ++ArgPtr) 108 ++NumArgTokens; 109 return NumArgTokens; 110} 111 112 113/// getUnexpArgument - Return the unexpanded tokens for the specified formal. 114/// 115const Token *MacroArgs::getUnexpArgument(unsigned Arg) const { 116 // The unexpanded argument tokens start immediately after the MacroArgs object 117 // in memory. 118 const Token *Start = (const Token *)(this+1); 119 const Token *Result = Start; 120 // Scan to find Arg. 121 for (; Arg; ++Result) { 122 assert(Result < Start+NumUnexpArgTokens && "Invalid arg #"); 123 if (Result->is(tok::eof)) 124 --Arg; 125 } 126 assert(Result < Start+NumUnexpArgTokens && "Invalid arg #"); 127 return Result; 128} 129 130 131/// ArgNeedsPreexpansion - If we can prove that the argument won't be affected 132/// by pre-expansion, return false. Otherwise, conservatively return true. 133bool MacroArgs::ArgNeedsPreexpansion(const Token *ArgTok, 134 Preprocessor &PP) const { 135 // If there are no identifiers in the argument list, or if the identifiers are 136 // known to not be macros, pre-expansion won't modify it. 137 for (; ArgTok->isNot(tok::eof); ++ArgTok) 138 if (IdentifierInfo *II = ArgTok->getIdentifierInfo()) 139 if (II->hasMacroDefinition()) 140 // Return true even though the macro could be a function-like macro 141 // without a following '(' token, or could be disabled, or not visible. 142 return true; 143 return false; 144} 145 146/// getPreExpArgument - Return the pre-expanded form of the specified 147/// argument. 148const std::vector<Token> & 149MacroArgs::getPreExpArgument(unsigned Arg, const MacroInfo *MI, 150 Preprocessor &PP) { 151 assert(Arg < MI->getNumParams() && "Invalid argument number!"); 152 153 // If we have already computed this, return it. 154 if (PreExpArgTokens.size() < MI->getNumParams()) 155 PreExpArgTokens.resize(MI->getNumParams()); 156 157 std::vector<Token> &Result = PreExpArgTokens[Arg]; 158 if (!Result.empty()) return Result; 159 160 SaveAndRestore<bool> PreExpandingMacroArgs(PP.InMacroArgPreExpansion, true); 161 162 const Token *AT = getUnexpArgument(Arg); 163 unsigned NumToks = getArgLength(AT)+1; // Include the EOF. 164 165 // Otherwise, we have to pre-expand this argument, populating Result. To do 166 // this, we set up a fake TokenLexer to lex from the unexpanded argument 167 // list. With this installed, we lex expanded tokens until we hit the EOF 168 // token at the end of the unexp list. 169 PP.EnterTokenStream(AT, NumToks, false /*disable expand*/, 170 false /*owns tokens*/); 171 172 // Lex all of the macro-expanded tokens into Result. 173 do { 174 Result.push_back(Token()); 175 Token &Tok = Result.back(); 176 PP.Lex(Tok); 177 } while (Result.back().isNot(tok::eof)); 178 179 // Pop the token stream off the top of the stack. We know that the internal 180 // pointer inside of it is to the "end" of the token stream, but the stack 181 // will not otherwise be popped until the next token is lexed. The problem is 182 // that the token may be lexed sometime after the vector of tokens itself is 183 // destroyed, which would be badness. 184 if (PP.InCachingLexMode()) 185 PP.ExitCachingLexMode(); 186 PP.RemoveTopOfLexerStack(); 187 return Result; 188} 189 190 191/// StringifyArgument - Implement C99 6.10.3.2p2, converting a sequence of 192/// tokens into the literal string token that should be produced by the C # 193/// preprocessor operator. If Charify is true, then it should be turned into 194/// a character literal for the Microsoft charize (#@) extension. 195/// 196Token MacroArgs::StringifyArgument(const Token *ArgToks, 197 Preprocessor &PP, bool Charify, 198 SourceLocation ExpansionLocStart, 199 SourceLocation ExpansionLocEnd) { 200 Token Tok; 201 Tok.startToken(); 202 Tok.setKind(Charify ? tok::char_constant : tok::string_literal); 203 204 const Token *ArgTokStart = ArgToks; 205 206 // Stringify all the tokens. 207 SmallString<128> Result; 208 Result += "\""; 209 210 bool isFirst = true; 211 for (; ArgToks->isNot(tok::eof); ++ArgToks) { 212 const Token &Tok = *ArgToks; 213 if (!isFirst && (Tok.hasLeadingSpace() || Tok.isAtStartOfLine())) 214 Result += ' '; 215 isFirst = false; 216 217 // If this is a string or character constant, escape the token as specified 218 // by 6.10.3.2p2. 219 if (tok::isStringLiteral(Tok.getKind()) || // "foo", u8R"x(foo)x"_bar, etc. 220 Tok.is(tok::char_constant) || // 'x' 221 Tok.is(tok::wide_char_constant) || // L'x'. 222 Tok.is(tok::utf8_char_constant) || // u8'x'. 223 Tok.is(tok::utf16_char_constant) || // u'x'. 224 Tok.is(tok::utf32_char_constant)) { // U'x'. 225 bool Invalid = false; 226 std::string TokStr = PP.getSpelling(Tok, &Invalid); 227 if (!Invalid) { 228 std::string Str = Lexer::Stringify(TokStr); 229 Result.append(Str.begin(), Str.end()); 230 } 231 } else if (Tok.is(tok::code_completion)) { 232 PP.CodeCompleteNaturalLanguage(); 233 } else { 234 // Otherwise, just append the token. Do some gymnastics to get the token 235 // in place and avoid copies where possible. 236 unsigned CurStrLen = Result.size(); 237 Result.resize(CurStrLen+Tok.getLength()); 238 const char *BufPtr = Result.data() + CurStrLen; 239 bool Invalid = false; 240 unsigned ActualTokLen = PP.getSpelling(Tok, BufPtr, &Invalid); 241 242 if (!Invalid) { 243 // If getSpelling returned a pointer to an already uniqued version of 244 // the string instead of filling in BufPtr, memcpy it onto our string. 245 if (ActualTokLen && BufPtr != &Result[CurStrLen]) 246 memcpy(&Result[CurStrLen], BufPtr, ActualTokLen); 247 248 // If the token was dirty, the spelling may be shorter than the token. 249 if (ActualTokLen != Tok.getLength()) 250 Result.resize(CurStrLen+ActualTokLen); 251 } 252 } 253 } 254 255 // If the last character of the string is a \, and if it isn't escaped, this 256 // is an invalid string literal, diagnose it as specified in C99. 257 if (Result.back() == '\\') { 258 // Count the number of consequtive \ characters. If even, then they are 259 // just escaped backslashes, otherwise it's an error. 260 unsigned FirstNonSlash = Result.size()-2; 261 // Guaranteed to find the starting " if nothing else. 262 while (Result[FirstNonSlash] == '\\') 263 --FirstNonSlash; 264 if ((Result.size()-1-FirstNonSlash) & 1) { 265 // Diagnose errors for things like: #define F(X) #X / F(\) 266 PP.Diag(ArgToks[-1], diag::pp_invalid_string_literal); 267 Result.pop_back(); // remove one of the \'s. 268 } 269 } 270 Result += '"'; 271 272 // If this is the charify operation and the result is not a legal character 273 // constant, diagnose it. 274 if (Charify) { 275 // First step, turn double quotes into single quotes: 276 Result[0] = '\''; 277 Result[Result.size()-1] = '\''; 278 279 // Check for bogus character. 280 bool isBad = false; 281 if (Result.size() == 3) 282 isBad = Result[1] == '\''; // ''' is not legal. '\' already fixed above. 283 else 284 isBad = (Result.size() != 4 || Result[1] != '\\'); // Not '\x' 285 286 if (isBad) { 287 PP.Diag(ArgTokStart[0], diag::err_invalid_character_to_charify); 288 Result = "' '"; // Use something arbitrary, but legal. 289 } 290 } 291 292 PP.CreateString(Result, Tok, 293 ExpansionLocStart, ExpansionLocEnd); 294 return Tok; 295} 296 297/// getStringifiedArgument - Compute, cache, and return the specified argument 298/// that has been 'stringified' as required by the # operator. 299const Token &MacroArgs::getStringifiedArgument(unsigned ArgNo, 300 Preprocessor &PP, 301 SourceLocation ExpansionLocStart, 302 SourceLocation ExpansionLocEnd) { 303 assert(ArgNo < getNumMacroArguments() && "Invalid argument number!"); 304 if (StringifiedArgs.empty()) 305 StringifiedArgs.resize(getNumMacroArguments(), {}); 306 307 if (StringifiedArgs[ArgNo].isNot(tok::string_literal)) 308 StringifiedArgs[ArgNo] = StringifyArgument(getUnexpArgument(ArgNo), PP, 309 /*Charify=*/false, 310 ExpansionLocStart, 311 ExpansionLocEnd); 312 return StringifiedArgs[ArgNo]; 313} 314