1 //===--- MacroArgs.cpp - Formal argument info for Macros ------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the MacroArgs interface.
12 //===----------------------------------------------------------------------===//
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"
22 using namespace clang;
24 /// MacroArgs ctor function - This destroys the vector passed in.
25 MacroArgs *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;
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) {
41 // If we have an exact match, use it.
42 if ((*Entry)->NumUnexpArgTokens == UnexpArgTokens.size())
44 // Otherwise, use the best fit.
45 ClosestMatch = (*Entry)->NumUnexpArgTokens;
50 // Allocate memory for a MacroArgs object with the lexer tokens at the end,
51 // and construct the MacroArgs object.
52 Result = new (std::malloc(totalSizeToAlloc<Token>(UnexpArgTokens.size())))
53 MacroArgs(UnexpArgTokens.size(), VarargsElided, MI->getNumParams());
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();
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 "
69 std::copy(UnexpArgTokens.begin(), UnexpArgTokens.end(),
70 Result->getTrailingObjects<Token>());
76 /// destroy - Destroy and deallocate the memory for this object.
78 void MacroArgs::destroy(Preprocessor &PP) {
79 StringifiedArgs.clear();
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();
86 // Add this to the preprocessor's free list.
87 ArgCache = PP.MacroArgCache;
88 PP.MacroArgCache = this;
91 /// deallocate - This should only be called by the Preprocessor when managing
93 MacroArgs *MacroArgs::deallocate() {
94 MacroArgs *Next = ArgCache;
96 // Run the dtor to deallocate the vectors.
98 // Release the memory for the object.
99 static_assert(std::is_trivially_destructible<Token>::value,
100 "assume trivially destructible and forego destructors");
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
110 unsigned MacroArgs::getArgLength(const Token *ArgPtr) {
111 unsigned NumArgTokens = 0;
112 for (; ArgPtr->isNot(tok::eof); ++ArgPtr)
118 /// getUnexpArgument - Return the unexpanded tokens for the specified formal.
120 const Token *MacroArgs::getUnexpArgument(unsigned Arg) const {
122 assert(Arg < getNumMacroArguments() && "Invalid arg #");
123 // The unexpanded argument tokens start immediately after the MacroArgs object
125 const Token *Start = getTrailingObjects<Token>();
126 const Token *Result = Start;
129 for (; Arg; ++Result) {
130 assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
131 if (Result->is(tok::eof))
134 assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
138 // This function assumes that the variadic arguments are the tokens
139 // corresponding to the last parameter (ellipsis) - and since tokens are
140 // separated by the 'eof' token, if that is the only token corresponding to that
141 // last parameter, we know no variadic arguments were supplied.
142 bool MacroArgs::invokedWithVariadicArgument(const MacroInfo *const MI) const {
143 if (!MI->isVariadic())
145 const int VariadicArgIndex = getNumMacroArguments() - 1;
146 return getUnexpArgument(VariadicArgIndex)->isNot(tok::eof);
149 /// ArgNeedsPreexpansion - If we can prove that the argument won't be affected
150 /// by pre-expansion, return false. Otherwise, conservatively return true.
151 bool MacroArgs::ArgNeedsPreexpansion(const Token *ArgTok,
152 Preprocessor &PP) const {
153 // If there are no identifiers in the argument list, or if the identifiers are
154 // known to not be macros, pre-expansion won't modify it.
155 for (; ArgTok->isNot(tok::eof); ++ArgTok)
156 if (IdentifierInfo *II = ArgTok->getIdentifierInfo())
157 if (II->hasMacroDefinition())
158 // Return true even though the macro could be a function-like macro
159 // without a following '(' token, or could be disabled, or not visible.
164 /// getPreExpArgument - Return the pre-expanded form of the specified
166 const std::vector<Token> &MacroArgs::getPreExpArgument(unsigned Arg,
168 assert(Arg < getNumMacroArguments() && "Invalid argument number!");
170 // If we have already computed this, return it.
171 if (PreExpArgTokens.size() < getNumMacroArguments())
172 PreExpArgTokens.resize(getNumMacroArguments());
174 std::vector<Token> &Result = PreExpArgTokens[Arg];
175 if (!Result.empty()) return Result;
177 SaveAndRestore<bool> PreExpandingMacroArgs(PP.InMacroArgPreExpansion, true);
179 const Token *AT = getUnexpArgument(Arg);
180 unsigned NumToks = getArgLength(AT)+1; // Include the EOF.
182 // Otherwise, we have to pre-expand this argument, populating Result. To do
183 // this, we set up a fake TokenLexer to lex from the unexpanded argument
184 // list. With this installed, we lex expanded tokens until we hit the EOF
185 // token at the end of the unexp list.
186 PP.EnterTokenStream(AT, NumToks, false /*disable expand*/,
187 false /*owns tokens*/);
189 // Lex all of the macro-expanded tokens into Result.
191 Result.push_back(Token());
192 Token &Tok = Result.back();
194 } while (Result.back().isNot(tok::eof));
196 // Pop the token stream off the top of the stack. We know that the internal
197 // pointer inside of it is to the "end" of the token stream, but the stack
198 // will not otherwise be popped until the next token is lexed. The problem is
199 // that the token may be lexed sometime after the vector of tokens itself is
200 // destroyed, which would be badness.
201 if (PP.InCachingLexMode())
202 PP.ExitCachingLexMode();
203 PP.RemoveTopOfLexerStack();
208 /// StringifyArgument - Implement C99 6.10.3.2p2, converting a sequence of
209 /// tokens into the literal string token that should be produced by the C #
210 /// preprocessor operator. If Charify is true, then it should be turned into
211 /// a character literal for the Microsoft charize (#@) extension.
213 Token MacroArgs::StringifyArgument(const Token *ArgToks,
214 Preprocessor &PP, bool Charify,
215 SourceLocation ExpansionLocStart,
216 SourceLocation ExpansionLocEnd) {
219 Tok.setKind(Charify ? tok::char_constant : tok::string_literal);
221 const Token *ArgTokStart = ArgToks;
223 // Stringify all the tokens.
224 SmallString<128> Result;
228 for (; ArgToks->isNot(tok::eof); ++ArgToks) {
229 const Token &Tok = *ArgToks;
230 if (!isFirst && (Tok.hasLeadingSpace() || Tok.isAtStartOfLine()))
234 // If this is a string or character constant, escape the token as specified
236 if (tok::isStringLiteral(Tok.getKind()) || // "foo", u8R"x(foo)x"_bar, etc.
237 Tok.is(tok::char_constant) || // 'x'
238 Tok.is(tok::wide_char_constant) || // L'x'.
239 Tok.is(tok::utf8_char_constant) || // u8'x'.
240 Tok.is(tok::utf16_char_constant) || // u'x'.
241 Tok.is(tok::utf32_char_constant)) { // U'x'.
242 bool Invalid = false;
243 std::string TokStr = PP.getSpelling(Tok, &Invalid);
245 std::string Str = Lexer::Stringify(TokStr);
246 Result.append(Str.begin(), Str.end());
248 } else if (Tok.is(tok::code_completion)) {
249 PP.CodeCompleteNaturalLanguage();
251 // Otherwise, just append the token. Do some gymnastics to get the token
252 // in place and avoid copies where possible.
253 unsigned CurStrLen = Result.size();
254 Result.resize(CurStrLen+Tok.getLength());
255 const char *BufPtr = Result.data() + CurStrLen;
256 bool Invalid = false;
257 unsigned ActualTokLen = PP.getSpelling(Tok, BufPtr, &Invalid);
260 // If getSpelling returned a pointer to an already uniqued version of
261 // the string instead of filling in BufPtr, memcpy it onto our string.
262 if (ActualTokLen && BufPtr != &Result[CurStrLen])
263 memcpy(&Result[CurStrLen], BufPtr, ActualTokLen);
265 // If the token was dirty, the spelling may be shorter than the token.
266 if (ActualTokLen != Tok.getLength())
267 Result.resize(CurStrLen+ActualTokLen);
272 // If the last character of the string is a \, and if it isn't escaped, this
273 // is an invalid string literal, diagnose it as specified in C99.
274 if (Result.back() == '\\') {
275 // Count the number of consequtive \ characters. If even, then they are
276 // just escaped backslashes, otherwise it's an error.
277 unsigned FirstNonSlash = Result.size()-2;
278 // Guaranteed to find the starting " if nothing else.
279 while (Result[FirstNonSlash] == '\\')
281 if ((Result.size()-1-FirstNonSlash) & 1) {
282 // Diagnose errors for things like: #define F(X) #X / F(\)
283 PP.Diag(ArgToks[-1], diag::pp_invalid_string_literal);
284 Result.pop_back(); // remove one of the \'s.
289 // If this is the charify operation and the result is not a legal character
290 // constant, diagnose it.
292 // First step, turn double quotes into single quotes:
294 Result[Result.size()-1] = '\'';
296 // Check for bogus character.
298 if (Result.size() == 3)
299 isBad = Result[1] == '\''; // ''' is not legal. '\' already fixed above.
301 isBad = (Result.size() != 4 || Result[1] != '\\'); // Not '\x'
304 PP.Diag(ArgTokStart[0], diag::err_invalid_character_to_charify);
305 Result = "' '"; // Use something arbitrary, but legal.
309 PP.CreateString(Result, Tok,
310 ExpansionLocStart, ExpansionLocEnd);
314 /// getStringifiedArgument - Compute, cache, and return the specified argument
315 /// that has been 'stringified' as required by the # operator.
316 const Token &MacroArgs::getStringifiedArgument(unsigned ArgNo,
318 SourceLocation ExpansionLocStart,
319 SourceLocation ExpansionLocEnd) {
320 assert(ArgNo < getNumMacroArguments() && "Invalid argument number!");
321 if (StringifiedArgs.empty())
322 StringifiedArgs.resize(getNumMacroArguments(), {});
324 if (StringifiedArgs[ArgNo].isNot(tok::string_literal))
325 StringifiedArgs[ArgNo] = StringifyArgument(getUnexpArgument(ArgNo), PP,
329 return StringifiedArgs[ArgNo];