| 1 |
//===--- RewriteMacros.cpp - Rewrite macros into their expansions ---------===// |
| 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 code rewrites macro invocations into their expansions. This gives you |
| 10 |
// a macro expanded file that retains comments and #includes. |
| 11 |
// |
| 12 |
//===----------------------------------------------------------------------===// |
| 13 |
|
| 14 |
#include "clang/Rewrite/Frontend/Rewriters.h" |
| 15 |
#include "clang/Basic/SourceManager.h" |
| 16 |
#include "clang/Lex/Preprocessor.h" |
| 17 |
#include "clang/Rewrite/Core/Rewriter.h" |
| 18 |
#include "llvm/Support/Path.h" |
| 19 |
#include "llvm/Support/raw_ostream.h" |
| 20 |
#include <cstdio> |
| 21 |
#include <memory> |
| 22 |
|
| 23 |
using namespace clang; |
| 24 |
|
| 25 |
/// isSameToken - Return true if the two specified tokens start have the same |
| 26 |
/// content. |
| 27 |
static bool isSameToken(Token &RawTok, Token &PPTok) { |
| 28 |
// If two tokens have the same kind and the same identifier info, they are |
| 29 |
// obviously the same. |
| 30 |
if (PPTok.getKind() == RawTok.getKind() && |
| 31 |
PPTok.getIdentifierInfo() == RawTok.getIdentifierInfo()) |
| 32 |
return true; |
| 33 |
|
| 34 |
// Otherwise, if they are different but have the same identifier info, they |
| 35 |
// are also considered to be the same. This allows keywords and raw lexed |
| 36 |
// identifiers with the same name to be treated the same. |
| 37 |
if (PPTok.getIdentifierInfo() && |
| 38 |
PPTok.getIdentifierInfo() == RawTok.getIdentifierInfo()) |
| 39 |
return true; |
| 40 |
|
| 41 |
return false; |
| 42 |
} |
| 43 |
|
| 44 |
|
| 45 |
/// GetNextRawTok - Return the next raw token in the stream, skipping over |
| 46 |
/// comments if ReturnComment is false. |
| 47 |
static const Token &GetNextRawTok(const std::vector<Token> &RawTokens, |
| 48 |
unsigned &CurTok, bool ReturnComment) { |
| 49 |
assert(CurTok < RawTokens.size() && "Overran eof!"); |
| 50 |
|
| 51 |
// If the client doesn't want comments and we have one, skip it. |
| 52 |
if (!ReturnComment && RawTokens[CurTok].is(tok::comment)) |
| 53 |
++CurTok; |
| 54 |
|
| 55 |
return RawTokens[CurTok++]; |
| 56 |
} |
| 57 |
|
| 58 |
|
| 59 |
/// LexRawTokensFromMainFile - Lets all the raw tokens from the main file into |
| 60 |
/// the specified vector. |
| 61 |
static void LexRawTokensFromMainFile(Preprocessor &PP, |
| 62 |
std::vector<Token> &RawTokens) { |
| 63 |
SourceManager &SM = PP.getSourceManager(); |
| 64 |
|
| 65 |
// Create a lexer to lex all the tokens of the main file in raw mode. Even |
| 66 |
// though it is in raw mode, it will not return comments. |
| 67 |
llvm::MemoryBufferRef FromFile = SM.getBufferOrFake(SM.getMainFileID()); |
| 68 |
Lexer RawLex(SM.getMainFileID(), FromFile, SM, PP.getLangOpts()); |
| 69 |
|
| 70 |
// Switch on comment lexing because we really do want them. |
| 71 |
RawLex.SetCommentRetentionState(true); |
| 72 |
|
| 73 |
Token RawTok; |
| 74 |
do { |
| 75 |
RawLex.LexFromRawLexer(RawTok); |
| 76 |
|
| 77 |
// If we have an identifier with no identifier info for our raw token, look |
| 78 |
// up the identifier info. This is important for equality comparison of |
| 79 |
// identifier tokens. |
| 80 |
if (RawTok.is(tok::raw_identifier)) |
| 81 |
PP.LookUpIdentifierInfo(RawTok); |
| 82 |
|
| 83 |
RawTokens.push_back(RawTok); |
| 84 |
} while (RawTok.isNot(tok::eof)); |
| 85 |
} |
| 86 |
|
| 87 |
|
| 88 |
/// RewriteMacrosInInput - Implement -rewrite-macros mode. |
| 89 |
void clang::RewriteMacrosInInput(Preprocessor &PP, raw_ostream *OS) { |
| 90 |
SourceManager &SM = PP.getSourceManager(); |
| 91 |
|
| 92 |
Rewriter Rewrite; |
| 93 |
Rewrite.setSourceMgr(SM, PP.getLangOpts()); |
| 94 |
RewriteBuffer &RB = Rewrite.getEditBuffer(SM.getMainFileID()); |
| 95 |
|
| 96 |
std::vector<Token> RawTokens; |
| 97 |
LexRawTokensFromMainFile(PP, RawTokens); |
| 98 |
unsigned CurRawTok = 0; |
| 99 |
Token RawTok = GetNextRawTok(RawTokens, CurRawTok, false); |
| 100 |
|
| 101 |
|
| 102 |
// Get the first preprocessing token. |
| 103 |
PP.EnterMainSourceFile(); |
| 104 |
Token PPTok; |
| 105 |
PP.Lex(PPTok); |
| 106 |
|
| 107 |
// Preprocess the input file in parallel with raw lexing the main file. Ignore |
| 108 |
// all tokens that are preprocessed from a file other than the main file (e.g. |
| 109 |
// a header). If we see tokens that are in the preprocessed file but not the |
| 110 |
// lexed file, we have a macro expansion. If we see tokens in the lexed file |
| 111 |
// that aren't in the preprocessed view, we have macros that expand to no |
| 112 |
// tokens, or macro arguments etc. |
| 113 |
while (RawTok.isNot(tok::eof) || PPTok.isNot(tok::eof)) { |
| 114 |
SourceLocation PPLoc = SM.getExpansionLoc(PPTok.getLocation()); |
| 115 |
|
| 116 |
// If PPTok is from a different source file, ignore it. |
| 117 |
if (!SM.isWrittenInMainFile(PPLoc)) { |
| 118 |
PP.Lex(PPTok); |
| 119 |
continue; |
| 120 |
} |
| 121 |
|
| 122 |
// If the raw file hits a preprocessor directive, they will be extra tokens |
| 123 |
// in the raw file that don't exist in the preprocsesed file. However, we |
| 124 |
// choose to preserve them in the output file and otherwise handle them |
| 125 |
// specially. |
| 126 |
if (RawTok.is(tok::hash) && RawTok.isAtStartOfLine()) { |
| 127 |
// If this is a #warning directive or #pragma mark (GNU extensions), |
| 128 |
// comment the line out. |
| 129 |
if (RawTokens[CurRawTok].is(tok::identifier)) { |
| 130 |
const IdentifierInfo *II = RawTokens[CurRawTok].getIdentifierInfo(); |
| 131 |
if (II->getName() == "warning") { |
| 132 |
// Comment out #warning. |
| 133 |
RB.InsertTextAfter(SM.getFileOffset(RawTok.getLocation()), "//"); |
| 134 |
} else if (II->getName() == "pragma" && |
| 135 |
RawTokens[CurRawTok+1].is(tok::identifier) && |
| 136 |
(RawTokens[CurRawTok+1].getIdentifierInfo()->getName() == |
| 137 |
"mark")) { |
| 138 |
// Comment out #pragma mark. |
| 139 |
RB.InsertTextAfter(SM.getFileOffset(RawTok.getLocation()), "//"); |
| 140 |
} |
| 141 |
} |
| 142 |
|
| 143 |
// Otherwise, if this is a #include or some other directive, just leave it |
| 144 |
// in the file by skipping over the line. |
| 145 |
RawTok = GetNextRawTok(RawTokens, CurRawTok, false); |
| 146 |
while (!RawTok.isAtStartOfLine() && RawTok.isNot(tok::eof)) |
| 147 |
RawTok = GetNextRawTok(RawTokens, CurRawTok, false); |
| 148 |
continue; |
| 149 |
} |
| 150 |
|
| 151 |
// Okay, both tokens are from the same file. Get their offsets from the |
| 152 |
// start of the file. |
| 153 |
unsigned PPOffs = SM.getFileOffset(PPLoc); |
| 154 |
unsigned RawOffs = SM.getFileOffset(RawTok.getLocation()); |
| 155 |
|
| 156 |
// If the offsets are the same and the token kind is the same, ignore them. |
| 157 |
if (PPOffs == RawOffs && isSameToken(RawTok, PPTok)) { |
| 158 |
RawTok = GetNextRawTok(RawTokens, CurRawTok, false); |
| 159 |
PP.Lex(PPTok); |
| 160 |
continue; |
| 161 |
} |
| 162 |
|
| 163 |
// If the PP token is farther along than the raw token, something was |
| 164 |
// deleted. Comment out the raw token. |
| 165 |
if (RawOffs <= PPOffs) { |
| 166 |
// Comment out a whole run of tokens instead of bracketing each one with |
| 167 |
// comments. Add a leading space if RawTok didn't have one. |
| 168 |
bool HasSpace = RawTok.hasLeadingSpace(); |
| 169 |
RB.InsertTextAfter(RawOffs, &" /*"[HasSpace]); |
| 170 |
unsigned EndPos; |
| 171 |
|
| 172 |
do { |
| 173 |
EndPos = RawOffs+RawTok.getLength(); |
| 174 |
|
| 175 |
RawTok = GetNextRawTok(RawTokens, CurRawTok, true); |
| 176 |
RawOffs = SM.getFileOffset(RawTok.getLocation()); |
| 177 |
|
| 178 |
if (RawTok.is(tok::comment)) { |
| 179 |
// Skip past the comment. |
| 180 |
RawTok = GetNextRawTok(RawTokens, CurRawTok, false); |
| 181 |
break; |
| 182 |
} |
| 183 |
|
| 184 |
} while (RawOffs <= PPOffs && !RawTok.isAtStartOfLine() && |
| 185 |
(PPOffs != RawOffs || !isSameToken(RawTok, PPTok))); |
| 186 |
|
| 187 |
RB.InsertTextBefore(EndPos, "*/"); |
| 188 |
continue; |
| 189 |
} |
| 190 |
|
| 191 |
// Otherwise, there was a replacement an expansion. Insert the new token |
| 192 |
// in the output buffer. Insert the whole run of new tokens at once to get |
| 193 |
// them in the right order. |
| 194 |
unsigned InsertPos = PPOffs; |
| 195 |
std::string Expansion; |
| 196 |
while (PPOffs < RawOffs) { |
| 197 |
Expansion += ' ' + PP.getSpelling(PPTok); |
| 198 |
PP.Lex(PPTok); |
| 199 |
PPLoc = SM.getExpansionLoc(PPTok.getLocation()); |
| 200 |
PPOffs = SM.getFileOffset(PPLoc); |
| 201 |
} |
| 202 |
Expansion += ' '; |
| 203 |
RB.InsertTextBefore(InsertPos, Expansion); |
| 204 |
} |
| 205 |
|
| 206 |
// Get the buffer corresponding to MainFileID. If we haven't changed it, then |
| 207 |
// we are done. |
| 208 |
if (const RewriteBuffer *RewriteBuf = |
| 209 |
Rewrite.getRewriteBufferFor(SM.getMainFileID())) { |
| 210 |
//printf("Changed:\n"); |
| 211 |
*OS << std::string(RewriteBuf->begin(), RewriteBuf->end()); |
| 212 |
} else { |
| 213 |
fprintf(stderr, "No changes\n"); |
| 214 |
} |
| 215 |
OS->flush(); |
| 216 |
} |