1 //===--- MacroArgs.cpp - Formal argument info for Macros ------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This file implements the MacroArgs interface. 10 // 11 //===----------------------------------------------------------------------===// 12 13 #include "clang/Lex/MacroArgs.h" 14 #include "clang/Lex/LexDiagnostic.h" 15 #include "clang/Lex/MacroInfo.h" 16 #include "clang/Lex/Preprocessor.h" 17 #include "llvm/ADT/SmallString.h" 18 #include "llvm/Support/SaveAndRestore.h" 19 #include <algorithm> 20 21 using namespace clang; 22 23 /// MacroArgs ctor function - This destroys the vector passed in. 24 MacroArgs *MacroArgs::create(const MacroInfo *MI, 25 ArrayRef<Token> UnexpArgTokens, 26 bool VarargsElided, Preprocessor &PP) { 27 assert(MI->isFunctionLike() && 28 "Can't have args for an object-like macro!"); 29 MacroArgs **ResultEnt = nullptr; 30 unsigned ClosestMatch = ~0U; 31 32 // See if we have an entry with a big enough argument list to reuse on the 33 // free list. If so, reuse it. 34 for (MacroArgs **Entry = &PP.MacroArgCache; *Entry; 35 Entry = &(*Entry)->ArgCache) { 36 if ((*Entry)->NumUnexpArgTokens >= UnexpArgTokens.size() && 37 (*Entry)->NumUnexpArgTokens < ClosestMatch) { 38 ResultEnt = Entry; 39 40 // If we have an exact match, use it. 41 if ((*Entry)->NumUnexpArgTokens == UnexpArgTokens.size()) 42 break; 43 // Otherwise, use the best fit. 44 ClosestMatch = (*Entry)->NumUnexpArgTokens; 45 } 46 } 47 MacroArgs *Result; 48 if (!ResultEnt) { 49 // Allocate memory for a MacroArgs object with the lexer tokens at the end, 50 // and construct the MacroArgs object. 51 Result = new ( 52 llvm::safe_malloc(totalSizeToAlloc<Token>(UnexpArgTokens.size()))) 53 MacroArgs(UnexpArgTokens.size(), VarargsElided, MI->getNumParams()); 54 } else { 55 Result = *ResultEnt; 56 // Unlink this node from the preprocessors singly linked list. 57 *ResultEnt = Result->ArgCache; 58 Result->NumUnexpArgTokens = UnexpArgTokens.size(); 59 Result->VarargsElided = VarargsElided; 60 Result->NumMacroArgs = MI->getNumParams(); 61 } 62 63 // Copy the actual unexpanded tokens to immediately after the result ptr. 64 if (!UnexpArgTokens.empty()) { 65 static_assert(std::is_trivial<Token>::value, 66 "assume trivial copyability if copying into the " 67 "uninitialized array (as opposed to reusing a cached " 68 "MacroArgs)"); 69 std::copy(UnexpArgTokens.begin(), UnexpArgTokens.end(), 70 Result->getTrailingObjects<Token>()); 71 } 72 73 return Result; 74 } 75 76 /// destroy - Destroy and deallocate the memory for this object. 77 /// 78 void MacroArgs::destroy(Preprocessor &PP) { 79 StringifiedArgs.clear(); 80 81 // Don't clear PreExpArgTokens, just clear the entries. Clearing the entries 82 // would deallocate the element vectors. 83 for (unsigned i = 0, e = PreExpArgTokens.size(); i != e; ++i) 84 PreExpArgTokens[i].clear(); 85 86 // Add this to the preprocessor's free list. 87 ArgCache = PP.MacroArgCache; 88 PP.MacroArgCache = this; 89 } 90 91 /// deallocate - This should only be called by the Preprocessor when managing 92 /// its freelist. 93 MacroArgs *MacroArgs::deallocate() { 94 MacroArgs *Next = ArgCache; 95 96 // Run the dtor to deallocate the vectors. 97 this->~MacroArgs(); 98 // Release the memory for the object. 99 static_assert(std::is_trivially_destructible<Token>::value, 100 "assume trivially destructible and forego destructors"); 101 free(this); 102 103 return Next; 104 } 105 106 107 /// getArgLength - Given a pointer to an expanded or unexpanded argument, 108 /// return the number of tokens, not counting the EOF, that make up the 109 /// argument. 110 unsigned MacroArgs::getArgLength(const Token *ArgPtr) { 111 unsigned NumArgTokens = 0; 112 for (; ArgPtr->isNot(tok::eof); ++ArgPtr) 113 ++NumArgTokens; 114 return NumArgTokens; 115 } 116 117 118 /// getUnexpArgument - Return the unexpanded tokens for the specified formal. 119 /// 120 const Token *MacroArgs::getUnexpArgument(unsigned Arg) const { 121 122 assert(Arg < getNumMacroArguments() && "Invalid arg #"); 123 // The unexpanded argument tokens start immediately after the MacroArgs object 124 // in memory. 125 const Token *Start = getTrailingObjects<Token>(); 126 const Token *Result = Start; 127 128 // Scan to find Arg. 129 for (; Arg; ++Result) { 130 assert(Result < Start+NumUnexpArgTokens && "Invalid arg #"); 131 if (Result->is(tok::eof)) 132 --Arg; 133 } 134 assert(Result < Start+NumUnexpArgTokens && "Invalid arg #"); 135 return Result; 136 } 137 138 bool MacroArgs::invokedWithVariadicArgument(const MacroInfo *const MI, 139 Preprocessor &PP) { 140 if (!MI->isVariadic()) 141 return false; 142 const int VariadicArgIndex = getNumMacroArguments() - 1; 143 return getPreExpArgument(VariadicArgIndex, PP).front().isNot(tok::eof); 144 } 145 146 /// ArgNeedsPreexpansion - If we can prove that the argument won't be affected 147 /// by pre-expansion, return false. Otherwise, conservatively return true. 148 bool MacroArgs::ArgNeedsPreexpansion(const Token *ArgTok, 149 Preprocessor &PP) const { 150 // If there are no identifiers in the argument list, or if the identifiers are 151 // known to not be macros, pre-expansion won't modify it. 152 for (; ArgTok->isNot(tok::eof); ++ArgTok) 153 if (IdentifierInfo *II = ArgTok->getIdentifierInfo()) 154 if (II->hasMacroDefinition()) 155 // Return true even though the macro could be a function-like macro 156 // without a following '(' token, or could be disabled, or not visible. 157 return true; 158 return false; 159 } 160 161 /// getPreExpArgument - Return the pre-expanded form of the specified 162 /// argument. 163 const std::vector<Token> &MacroArgs::getPreExpArgument(unsigned Arg, 164 Preprocessor &PP) { 165 assert(Arg < getNumMacroArguments() && "Invalid argument number!"); 166 167 // If we have already computed this, return it. 168 if (PreExpArgTokens.size() < getNumMacroArguments()) 169 PreExpArgTokens.resize(getNumMacroArguments()); 170 171 std::vector<Token> &Result = PreExpArgTokens[Arg]; 172 if (!Result.empty()) return Result; 173 174 SaveAndRestore<bool> PreExpandingMacroArgs(PP.InMacroArgPreExpansion, true); 175 176 const Token *AT = getUnexpArgument(Arg); 177 unsigned NumToks = getArgLength(AT)+1; // Include the EOF. 178 179 // Otherwise, we have to pre-expand this argument, populating Result. To do 180 // this, we set up a fake TokenLexer to lex from the unexpanded argument 181 // list. With this installed, we lex expanded tokens until we hit the EOF 182 // token at the end of the unexp list. 183 PP.EnterTokenStream(AT, NumToks, false /*disable expand*/, 184 false /*owns tokens*/, false /*is reinject*/); 185 186 // Lex all of the macro-expanded tokens into Result. 187 do { 188 Result.push_back(Token()); 189 Token &Tok = Result.back(); 190 PP.Lex(Tok); 191 } while (Result.back().isNot(tok::eof)); 192 193 // Pop the token stream off the top of the stack. We know that the internal 194 // pointer inside of it is to the "end" of the token stream, but the stack 195 // will not otherwise be popped until the next token is lexed. The problem is 196 // that the token may be lexed sometime after the vector of tokens itself is 197 // destroyed, which would be badness. 198 if (PP.InCachingLexMode()) 199 PP.ExitCachingLexMode(); 200 PP.RemoveTopOfLexerStack(); 201 return Result; 202 } 203 204 205 /// StringifyArgument - Implement C99 6.10.3.2p2, converting a sequence of 206 /// tokens into the literal string token that should be produced by the C # 207 /// preprocessor operator. If Charify is true, then it should be turned into 208 /// a character literal for the Microsoft charize (#@) extension. 209 /// 210 Token MacroArgs::StringifyArgument(const Token *ArgToks, 211 Preprocessor &PP, bool Charify, 212 SourceLocation ExpansionLocStart, 213 SourceLocation ExpansionLocEnd) { 214 Token Tok; 215 Tok.startToken(); 216 Tok.setKind(Charify ? tok::char_constant : tok::string_literal); 217 218 const Token *ArgTokStart = ArgToks; 219 220 // Stringify all the tokens. 221 SmallString<128> Result; 222 Result += "\""; 223 224 bool isFirst = true; 225 for (; ArgToks->isNot(tok::eof); ++ArgToks) { 226 const Token &Tok = *ArgToks; 227 if (!isFirst && (Tok.hasLeadingSpace() || Tok.isAtStartOfLine())) 228 Result += ' '; 229 isFirst = false; 230 231 // If this is a string or character constant, escape the token as specified 232 // by 6.10.3.2p2. 233 if (tok::isStringLiteral(Tok.getKind()) || // "foo", u8R"x(foo)x"_bar, etc. 234 Tok.is(tok::char_constant) || // 'x' 235 Tok.is(tok::wide_char_constant) || // L'x'. 236 Tok.is(tok::utf8_char_constant) || // u8'x'. 237 Tok.is(tok::utf16_char_constant) || // u'x'. 238 Tok.is(tok::utf32_char_constant)) { // U'x'. 239 bool Invalid = false; 240 std::string TokStr = PP.getSpelling(Tok, &Invalid); 241 if (!Invalid) { 242 std::string Str = Lexer::Stringify(TokStr); 243 Result.append(Str.begin(), Str.end()); 244 } 245 } else if (Tok.is(tok::code_completion)) { 246 PP.CodeCompleteNaturalLanguage(); 247 } else { 248 // Otherwise, just append the token. Do some gymnastics to get the token 249 // in place and avoid copies where possible. 250 unsigned CurStrLen = Result.size(); 251 Result.resize(CurStrLen+Tok.getLength()); 252 const char *BufPtr = Result.data() + CurStrLen; 253 bool Invalid = false; 254 unsigned ActualTokLen = PP.getSpelling(Tok, BufPtr, &Invalid); 255 256 if (!Invalid) { 257 // If getSpelling returned a pointer to an already uniqued version of 258 // the string instead of filling in BufPtr, memcpy it onto our string. 259 if (ActualTokLen && BufPtr != &Result[CurStrLen]) 260 memcpy(&Result[CurStrLen], BufPtr, ActualTokLen); 261 262 // If the token was dirty, the spelling may be shorter than the token. 263 if (ActualTokLen != Tok.getLength()) 264 Result.resize(CurStrLen+ActualTokLen); 265 } 266 } 267 } 268 269 // If the last character of the string is a \, and if it isn't escaped, this 270 // is an invalid string literal, diagnose it as specified in C99. 271 if (Result.back() == '\\') { 272 // Count the number of consecutive \ characters. If even, then they are 273 // just escaped backslashes, otherwise it's an error. 274 unsigned FirstNonSlash = Result.size()-2; 275 // Guaranteed to find the starting " if nothing else. 276 while (Result[FirstNonSlash] == '\\') 277 --FirstNonSlash; 278 if ((Result.size()-1-FirstNonSlash) & 1) { 279 // Diagnose errors for things like: #define F(X) #X / F(\) 280 PP.Diag(ArgToks[-1], diag::pp_invalid_string_literal); 281 Result.pop_back(); // remove one of the \'s. 282 } 283 } 284 Result += '"'; 285 286 // If this is the charify operation and the result is not a legal character 287 // constant, diagnose it. 288 if (Charify) { 289 // First step, turn double quotes into single quotes: 290 Result[0] = '\''; 291 Result[Result.size()-1] = '\''; 292 293 // Check for bogus character. 294 bool isBad = false; 295 if (Result.size() == 3) 296 isBad = Result[1] == '\''; // ''' is not legal. '\' already fixed above. 297 else 298 isBad = (Result.size() != 4 || Result[1] != '\\'); // Not '\x' 299 300 if (isBad) { 301 PP.Diag(ArgTokStart[0], diag::err_invalid_character_to_charify); 302 Result = "' '"; // Use something arbitrary, but legal. 303 } 304 } 305 306 PP.CreateString(Result, Tok, 307 ExpansionLocStart, ExpansionLocEnd); 308 return Tok; 309 } 310 311 /// getStringifiedArgument - Compute, cache, and return the specified argument 312 /// that has been 'stringified' as required by the # operator. 313 const Token &MacroArgs::getStringifiedArgument(unsigned ArgNo, 314 Preprocessor &PP, 315 SourceLocation ExpansionLocStart, 316 SourceLocation ExpansionLocEnd) { 317 assert(ArgNo < getNumMacroArguments() && "Invalid argument number!"); 318 if (StringifiedArgs.empty()) 319 StringifiedArgs.resize(getNumMacroArguments(), {}); 320 321 if (StringifiedArgs[ArgNo].isNot(tok::string_literal)) 322 StringifiedArgs[ArgNo] = StringifyArgument(getUnexpArgument(ArgNo), PP, 323 /*Charify=*/false, 324 ExpansionLocStart, 325 ExpansionLocEnd); 326 return StringifiedArgs[ArgNo]; 327 } 328