xref: /llvm-project/clang/lib/Lex/MacroArgs.cpp (revision 764275949897533a4be0728250e69a94d228fbc5)
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/Support/SaveAndRestore.h"
18 #include <algorithm>
19 
20 using namespace clang;
21 
22 /// MacroArgs ctor function - This destroys the vector passed in.
23 MacroArgs *MacroArgs::create(const MacroInfo *MI,
24                              ArrayRef<Token> UnexpArgTokens,
25                              bool VarargsElided, Preprocessor &PP) {
26   assert(MI->isFunctionLike() &&
27          "Can't have args for an object-like macro!");
28   MacroArgs **ResultEnt = nullptr;
29   unsigned ClosestMatch = ~0U;
30 
31   // See if we have an entry with a big enough argument list to reuse on the
32   // free list.  If so, reuse it.
33   for (MacroArgs **Entry = &PP.MacroArgCache; *Entry;
34        Entry = &(*Entry)->ArgCache) {
35     if ((*Entry)->NumUnexpArgTokens >= UnexpArgTokens.size() &&
36         (*Entry)->NumUnexpArgTokens < ClosestMatch) {
37       ResultEnt = Entry;
38 
39       // If we have an exact match, use it.
40       if ((*Entry)->NumUnexpArgTokens == UnexpArgTokens.size())
41         break;
42       // Otherwise, use the best fit.
43       ClosestMatch = (*Entry)->NumUnexpArgTokens;
44     }
45   }
46   MacroArgs *Result;
47   if (!ResultEnt) {
48     // Allocate memory for a MacroArgs object with the lexer tokens at the end,
49     // and construct the MacroArgs object.
50     Result = new (
51         llvm::safe_malloc(totalSizeToAlloc<Token>(UnexpArgTokens.size())))
52         MacroArgs(UnexpArgTokens.size(), VarargsElided, MI->getNumParams());
53   } else {
54     Result = *ResultEnt;
55     // Unlink this node from the preprocessors singly linked list.
56     *ResultEnt = Result->ArgCache;
57     Result->NumUnexpArgTokens = UnexpArgTokens.size();
58     Result->VarargsElided = VarargsElided;
59     Result->NumMacroArgs = MI->getNumParams();
60   }
61 
62   // Copy the actual unexpanded tokens to immediately after the result ptr.
63   if (!UnexpArgTokens.empty()) {
64     static_assert(std::is_trivial_v<Token>,
65                   "assume trivial copyability if copying into the "
66                   "uninitialized array (as opposed to reusing a cached "
67                   "MacroArgs)");
68     std::copy(UnexpArgTokens.begin(), UnexpArgTokens.end(),
69               Result->getTrailingObjects<Token>());
70   }
71 
72   return Result;
73 }
74 
75 /// destroy - Destroy and deallocate the memory for this object.
76 ///
77 void MacroArgs::destroy(Preprocessor &PP) {
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.
90 MacroArgs *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   static_assert(std::is_trivially_destructible_v<Token>,
97                 "assume trivially destructible and forego destructors");
98   free(this);
99 
100   return Next;
101 }
102 
103 
104 /// getArgLength - Given a pointer to an expanded or unexpanded argument,
105 /// return the number of tokens, not counting the EOF, that make up the
106 /// argument.
107 unsigned MacroArgs::getArgLength(const Token *ArgPtr) {
108   unsigned NumArgTokens = 0;
109   for (; ArgPtr->isNot(tok::eof); ++ArgPtr)
110     ++NumArgTokens;
111   return NumArgTokens;
112 }
113 
114 
115 /// getUnexpArgument - Return the unexpanded tokens for the specified formal.
116 ///
117 const Token *MacroArgs::getUnexpArgument(unsigned Arg) const {
118 
119   assert(Arg < getNumMacroArguments() && "Invalid arg #");
120   // The unexpanded argument tokens start immediately after the MacroArgs object
121   // in memory.
122   const Token *Start = getTrailingObjects<Token>();
123   const Token *Result = Start;
124 
125   // Scan to find Arg.
126   for (; Arg; ++Result) {
127     assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
128     if (Result->is(tok::eof))
129       --Arg;
130   }
131   assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
132   return Result;
133 }
134 
135 bool MacroArgs::invokedWithVariadicArgument(const MacroInfo *const MI,
136                                             Preprocessor &PP) {
137   if (!MI->isVariadic())
138     return false;
139   const int VariadicArgIndex = getNumMacroArguments() - 1;
140   return getPreExpArgument(VariadicArgIndex, PP).front().isNot(tok::eof);
141 }
142 
143 /// ArgNeedsPreexpansion - If we can prove that the argument won't be affected
144 /// by pre-expansion, return false.  Otherwise, conservatively return true.
145 bool MacroArgs::ArgNeedsPreexpansion(const Token *ArgTok,
146                                      Preprocessor &PP) const {
147   // If there are no identifiers in the argument list, or if the identifiers are
148   // known to not be macros, pre-expansion won't modify it.
149   for (; ArgTok->isNot(tok::eof); ++ArgTok)
150     if (IdentifierInfo *II = ArgTok->getIdentifierInfo())
151       if (II->hasMacroDefinition())
152         // Return true even though the macro could be a function-like macro
153         // without a following '(' token, or could be disabled, or not visible.
154         return true;
155   return false;
156 }
157 
158 /// getPreExpArgument - Return the pre-expanded form of the specified
159 /// argument.
160 const std::vector<Token> &MacroArgs::getPreExpArgument(unsigned Arg,
161                                                        Preprocessor &PP) {
162   assert(Arg < getNumMacroArguments() && "Invalid argument number!");
163 
164   // If we have already computed this, return it.
165   if (PreExpArgTokens.size() < getNumMacroArguments())
166     PreExpArgTokens.resize(getNumMacroArguments());
167 
168   std::vector<Token> &Result = PreExpArgTokens[Arg];
169   if (!Result.empty()) return Result;
170 
171   SaveAndRestore PreExpandingMacroArgs(PP.InMacroArgPreExpansion, true);
172 
173   const Token *AT = getUnexpArgument(Arg);
174   unsigned NumToks = getArgLength(AT)+1;  // Include the EOF.
175 
176   // Otherwise, we have to pre-expand this argument, populating Result.  To do
177   // this, we set up a fake TokenLexer to lex from the unexpanded argument
178   // list.  With this installed, we lex expanded tokens until we hit the EOF
179   // token at the end of the unexp list.
180   PP.EnterTokenStream(AT, NumToks, false /*disable expand*/,
181                       false /*owns tokens*/, false /*is reinject*/);
182 
183   // Lex all of the macro-expanded tokens into Result.
184   do {
185     Result.push_back(Token());
186     Token &Tok = Result.back();
187     PP.Lex(Tok);
188   } while (Result.back().isNot(tok::eof));
189 
190   // Pop the token stream off the top of the stack.  We know that the internal
191   // pointer inside of it is to the "end" of the token stream, but the stack
192   // will not otherwise be popped until the next token is lexed.  The problem is
193   // that the token may be lexed sometime after the vector of tokens itself is
194   // destroyed, which would be badness.
195   if (PP.InCachingLexMode())
196     PP.ExitCachingLexMode();
197   PP.RemoveTopOfLexerStack();
198   return Result;
199 }
200 
201 
202 /// StringifyArgument - Implement C99 6.10.3.2p2, converting a sequence of
203 /// tokens into the literal string token that should be produced by the C #
204 /// preprocessor operator.  If Charify is true, then it should be turned into
205 /// a character literal for the Microsoft charize (#@) extension.
206 ///
207 Token MacroArgs::StringifyArgument(const Token *ArgToks,
208                                    Preprocessor &PP, bool Charify,
209                                    SourceLocation ExpansionLocStart,
210                                    SourceLocation ExpansionLocEnd) {
211   Token Tok;
212   Tok.startToken();
213   Tok.setKind(Charify ? tok::char_constant : tok::string_literal);
214 
215   const Token *ArgTokStart = ArgToks;
216 
217   // Stringify all the tokens.
218   SmallString<128> Result;
219   Result += "\"";
220 
221   bool isFirst = true;
222   for (; ArgToks->isNot(tok::eof); ++ArgToks) {
223     const Token &Tok = *ArgToks;
224     if (!isFirst && (Tok.hasLeadingSpace() || Tok.isAtStartOfLine()))
225       Result += ' ';
226     isFirst = false;
227 
228     // If this is a string or character constant, escape the token as specified
229     // by 6.10.3.2p2.
230     if (tok::isStringLiteral(Tok.getKind()) || // "foo", u8R"x(foo)x"_bar, etc.
231         Tok.is(tok::char_constant) ||          // 'x'
232         Tok.is(tok::wide_char_constant) ||     // L'x'.
233         Tok.is(tok::utf8_char_constant) ||     // u8'x'.
234         Tok.is(tok::utf16_char_constant) ||    // u'x'.
235         Tok.is(tok::utf32_char_constant)) {    // U'x'.
236       bool Invalid = false;
237       std::string TokStr = PP.getSpelling(Tok, &Invalid);
238       if (!Invalid) {
239         std::string Str = Lexer::Stringify(TokStr);
240         Result.append(Str.begin(), Str.end());
241       }
242     } else if (Tok.is(tok::code_completion)) {
243       PP.CodeCompleteNaturalLanguage();
244     } else {
245       // Otherwise, just append the token.  Do some gymnastics to get the token
246       // in place and avoid copies where possible.
247       unsigned CurStrLen = Result.size();
248       Result.resize(CurStrLen+Tok.getLength());
249       const char *BufPtr = Result.data() + CurStrLen;
250       bool Invalid = false;
251       unsigned ActualTokLen = PP.getSpelling(Tok, BufPtr, &Invalid);
252 
253       if (!Invalid) {
254         // If getSpelling returned a pointer to an already uniqued version of
255         // the string instead of filling in BufPtr, memcpy it onto our string.
256         if (ActualTokLen && BufPtr != &Result[CurStrLen])
257           memcpy(&Result[CurStrLen], BufPtr, ActualTokLen);
258 
259         // If the token was dirty, the spelling may be shorter than the token.
260         if (ActualTokLen != Tok.getLength())
261           Result.resize(CurStrLen+ActualTokLen);
262       }
263     }
264   }
265 
266   // If the last character of the string is a \, and if it isn't escaped, this
267   // is an invalid string literal, diagnose it as specified in C99.
268   if (Result.back() == '\\') {
269     // Count the number of consecutive \ characters.  If even, then they are
270     // just escaped backslashes, otherwise it's an error.
271     unsigned FirstNonSlash = Result.size()-2;
272     // Guaranteed to find the starting " if nothing else.
273     while (Result[FirstNonSlash] == '\\')
274       --FirstNonSlash;
275     if ((Result.size()-1-FirstNonSlash) & 1) {
276       // Diagnose errors for things like: #define F(X) #X   /   F(\)
277       PP.Diag(ArgToks[-1], diag::pp_invalid_string_literal);
278       Result.pop_back();  // remove one of the \'s.
279     }
280   }
281   Result += '"';
282 
283   // If this is the charify operation and the result is not a legal character
284   // constant, diagnose it.
285   if (Charify) {
286     // First step, turn double quotes into single quotes:
287     Result[0] = '\'';
288     Result[Result.size()-1] = '\'';
289 
290     // Check for bogus character.
291     bool isBad = false;
292     if (Result.size() == 3)
293       isBad = Result[1] == '\'';   // ''' is not legal. '\' already fixed above.
294     else
295       isBad = (Result.size() != 4 || Result[1] != '\\');  // Not '\x'
296 
297     if (isBad) {
298       PP.Diag(ArgTokStart[0], diag::err_invalid_character_to_charify);
299       Result = "' '";  // Use something arbitrary, but legal.
300     }
301   }
302 
303   PP.CreateString(Result, Tok,
304                   ExpansionLocStart, ExpansionLocEnd);
305   return Tok;
306 }
307