1 //===- ScriptLexer.cpp ----------------------------------------------------===//
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file defines a lexer for the linker script.
12 // The linker script's grammar is not complex but ambiguous due to the
13 // lack of the formal specification of the language. What we are trying to
14 // do in this and other files in LLD is to make a "reasonable" linker
17 // Among simplicity, compatibility and efficiency, we put the most
18 // emphasis on simplicity when we wrote this lexer. Compatibility with the
19 // GNU linkers is important, but we did not try to clone every tiny corner
20 // case of their lexers, as even ld.bfd and ld.gold are subtly different
21 // in various corner cases. We do not care much about efficiency because
22 // the time spent in parsing linker scripts is usually negligible.
24 // Our grammar of the linker script is LL(2), meaning that it needs at
25 // most two-token lookahead to parse. The only place we need two-token
26 // lookahead is labels in version scripts, where we need to parse "local :"
29 // Overall, this lexer works fine for most linker scripts. There might
30 // be room for improving compatibility, but that's probably not at the
31 // top of our todo list.
33 //===----------------------------------------------------------------------===//
35 #include "ScriptLexer.h"
37 #include "llvm/ADT/Twine.h"
41 using namespace lld::elf;
43 // Returns a whole line containing the current token.
44 StringRef ScriptLexer::getLine() {
45 StringRef S = getCurrentMB().getBuffer();
46 StringRef Tok = Tokens[Pos - 1];
48 size_t Pos = S.rfind('\n', Tok.data() - S.data());
49 if (Pos != StringRef::npos)
50 S = S.substr(Pos + 1);
51 return S.substr(0, S.find_first_of("\r\n"));
54 // Returns 1-based line number of the current token.
55 size_t ScriptLexer::getLineNumber() {
56 StringRef S = getCurrentMB().getBuffer();
57 StringRef Tok = Tokens[Pos - 1];
58 return S.substr(0, Tok.data() - S.data()).count('\n') + 1;
61 // Returns 0-based column number of the current token.
62 size_t ScriptLexer::getColumnNumber() {
63 StringRef Tok = Tokens[Pos - 1];
64 return Tok.data() - getLine().data();
67 std::string ScriptLexer::getCurrentLocation() {
68 std::string Filename = getCurrentMB().getBufferIdentifier();
71 return (Filename + ":" + Twine(getLineNumber())).str();
74 ScriptLexer::ScriptLexer(MemoryBufferRef MB) { tokenize(MB); }
76 // We don't want to record cascading errors. Keep only the first one.
77 void ScriptLexer::setError(const Twine &Msg) {
83 error(getCurrentLocation() + ": " + Msg);
87 std::string S = getCurrentLocation() + ": ";
90 error(S + std::string(getColumnNumber(), ' ') + "^");
93 // Split S into linker script tokens.
94 void ScriptLexer::tokenize(MemoryBufferRef MB) {
95 std::vector<StringRef> Vec;
97 StringRef S = MB.getBuffer();
105 // Quoted token. Note that double-quote characters are parts of a token
106 // because, in a glob match context, only unquoted tokens are interpreted
107 // as glob patterns. Double-quoted tokens are literal patterns in that
109 if (S.startswith("\"")) {
110 size_t E = S.find("\"", 1);
111 if (E == StringRef::npos) {
112 StringRef Filename = MB.getBufferIdentifier();
113 size_t Lineno = Begin.substr(0, S.data() - Begin.data()).count('\n');
114 error(Filename + ":" + Twine(Lineno + 1) + ": unclosed quote");
118 Vec.push_back(S.take_front(E + 1));
123 // Unquoted token. This is more relaxed than tokens in C-like language,
124 // so that you can write "file-name.cpp" as one bare token, for example.
125 size_t Pos = S.find_first_not_of(
126 "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"
127 "0123456789_.$/\\~=+[]*?-!<>^:");
129 // A character that cannot start a word (which is usually a
130 // punctuation) forms a single character token.
133 Vec.push_back(S.substr(0, Pos));
137 Tokens.insert(Tokens.begin() + Pos, Vec.begin(), Vec.end());
140 // Skip leading whitespace characters or comments.
141 StringRef ScriptLexer::skipSpace(StringRef S) {
143 if (S.startswith("/*")) {
144 size_t E = S.find("*/", 2);
145 if (E == StringRef::npos) {
146 error("unclosed comment in a linker script");
152 if (S.startswith("#")) {
153 size_t E = S.find('\n', 1);
154 if (E == StringRef::npos)
159 size_t Size = S.size();
161 if (S.size() == Size)
166 // An erroneous token is handled as if it were the last token before EOF.
167 bool ScriptLexer::atEOF() { return Error || Tokens.size() == Pos; }
169 // Split a given string as an expression.
170 // This function returns "3", "*" and "5" for "3*5" for example.
171 static std::vector<StringRef> tokenizeExpr(StringRef S) {
172 StringRef Ops = "+-*/:"; // List of operators
174 // Quoted strings are literal strings, so we don't want to split it.
175 if (S.startswith("\""))
178 // Split S with +-*/ as separators.
179 std::vector<StringRef> Ret;
181 size_t E = S.find_first_of(Ops);
183 // No need to split if there is no operator.
184 if (E == StringRef::npos) {
189 // Get a token before the opreator.
191 Ret.push_back(S.substr(0, E));
193 // Get the operator as a token.
194 Ret.push_back(S.substr(E, 1));
200 // In contexts where expressions are expected, the lexer should apply
201 // different tokenization rules than the default one. By default,
202 // arithmetic operator characters are regular characters, but in the
203 // expression context, they should be independent tokens.
205 // For example, "foo*3" should be tokenized to "foo", "*" and "3" only
206 // in the expression context.
208 // This function may split the current token into multiple tokens.
209 void ScriptLexer::maybeSplitExpr() {
210 if (!InExpr || Error || atEOF())
213 std::vector<StringRef> V = tokenizeExpr(Tokens[Pos]);
216 Tokens.erase(Tokens.begin() + Pos);
217 Tokens.insert(Tokens.begin() + Pos, V.begin(), V.end());
220 StringRef ScriptLexer::next() {
226 setError("unexpected EOF");
229 return Tokens[Pos++];
232 StringRef ScriptLexer::peek() {
233 StringRef Tok = next();
240 bool ScriptLexer::consume(StringRef Tok) {
248 // Consumes Tok followed by ":". Space is allowed between Tok and ":".
249 bool ScriptLexer::consumeLabel(StringRef Tok) {
250 if (consume((Tok + ":").str()))
252 if (Tokens.size() >= Pos + 2 && Tokens[Pos] == Tok &&
253 Tokens[Pos + 1] == ":") {
260 void ScriptLexer::skip() { (void)next(); }
262 void ScriptLexer::expect(StringRef Expect) {
265 StringRef Tok = next();
267 setError(Expect + " expected, but got " + Tok);
270 // Returns true if S encloses T.
271 static bool encloses(StringRef S, StringRef T) {
272 return S.bytes_begin() <= T.bytes_begin() && T.bytes_end() <= S.bytes_end();
275 MemoryBufferRef ScriptLexer::getCurrentMB() {
276 // Find input buffer containing the current token.
277 assert(!MBs.empty());
281 for (MemoryBufferRef MB : MBs)
282 if (encloses(MB.getBuffer(), Tokens[Pos - 1]))
284 llvm_unreachable("getCurrentMB: failed to find a token");