1 //===- ScriptLexer.cpp ----------------------------------------------------===//
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
7 //===----------------------------------------------------------------------===//
9 // This file defines a lexer for the linker script.
11 // The linker script's grammar is not complex but ambiguous due to the
12 // lack of the formal specification of the language. What we are trying to
13 // do in this and other files in LLD is to make a "reasonable" linker
16 // Among simplicity, compatibility and efficiency, we put the most
17 // emphasis on simplicity when we wrote this lexer. Compatibility with the
18 // GNU linkers is important, but we did not try to clone every tiny corner
19 // case of their lexers, as even ld.bfd and ld.gold are subtly different
20 // in various corner cases. We do not care much about efficiency because
21 // the time spent in parsing linker scripts is usually negligible.
23 // Our grammar of the linker script is LL(2), meaning that it needs at
24 // most two-token lookahead to parse. The only place we need two-token
25 // lookahead is labels in version scripts, where we need to parse "local :"
28 // Overall, this lexer works fine for most linker scripts. There might
29 // be room for improving compatibility, but that's probably not at the
30 // top of our todo list.
32 //===----------------------------------------------------------------------===//
34 #include "ScriptLexer.h"
35 #include "lld/Common/ErrorHandler.h"
36 #include "llvm/ADT/Twine.h"
42 // Returns a whole line containing the current token.
43 StringRef ScriptLexer::getLine() {
44 StringRef s = getCurrentMB().getBuffer();
45 StringRef tok = tokens[pos - 1];
47 size_t pos = s.rfind('\n', tok.data() - s.data());
48 if (pos != StringRef::npos)
49 s = s.substr(pos + 1);
50 return s.substr(0, s.find_first_of("\r\n"));
53 // Returns 1-based line number of the current token.
54 size_t ScriptLexer::getLineNumber() {
55 StringRef s = getCurrentMB().getBuffer();
56 StringRef tok = tokens[pos - 1];
57 return s.substr(0, tok.data() - s.data()).count('\n') + 1;
60 // Returns 0-based column number of the current token.
61 size_t ScriptLexer::getColumnNumber() {
62 StringRef tok = tokens[pos - 1];
63 return tok.data() - getLine().data();
66 std::string ScriptLexer::getCurrentLocation() {
67 std::string filename = getCurrentMB().getBufferIdentifier();
68 return (filename + ":" + Twine(getLineNumber())).str();
71 ScriptLexer::ScriptLexer(MemoryBufferRef mb) { tokenize(mb); }
73 // We don't want to record cascading errors. Keep only the first one.
74 void ScriptLexer::setError(const Twine &msg) {
78 std::string s = (getCurrentLocation() + ": " + msg).str();
80 s += "\n>>> " + getLine().str() + "\n>>> " +
81 std::string(getColumnNumber(), ' ') + "^";
85 // Split S into linker script tokens.
86 void ScriptLexer::tokenize(MemoryBufferRef mb) {
87 std::vector<StringRef> vec;
89 StringRef s = mb.getBuffer();
97 // Quoted token. Note that double-quote characters are parts of a token
98 // because, in a glob match context, only unquoted tokens are interpreted
99 // as glob patterns. Double-quoted tokens are literal patterns in that
101 if (s.startswith("\"")) {
102 size_t e = s.find("\"", 1);
103 if (e == StringRef::npos) {
104 StringRef filename = mb.getBufferIdentifier();
105 size_t lineno = begin.substr(0, s.data() - begin.data()).count('\n');
106 error(filename + ":" + Twine(lineno + 1) + ": unclosed quote");
110 vec.push_back(s.take_front(e + 1));
115 // ">foo" is parsed to ">" and "foo", but ">>" is parsed to ">>".
116 // "|", "||", "&" and "&&" are different operators.
117 if (s.startswith("<<") || s.startswith("<=") || s.startswith(">>") ||
118 s.startswith(">=") || s.startswith("||") || s.startswith("&&")) {
119 vec.push_back(s.substr(0, 2));
124 // Unquoted token. This is more relaxed than tokens in C-like language,
125 // so that you can write "file-name.cpp" as one bare token, for example.
126 size_t pos = s.find_first_not_of(
127 "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"
128 "0123456789_.$/\\~=+[]*?-!^:");
130 // A character that cannot start a word (which is usually a
131 // punctuation) forms a single character token.
134 vec.push_back(s.substr(0, pos));
138 tokens.insert(tokens.begin() + pos, vec.begin(), vec.end());
141 // Skip leading whitespace characters or comments.
142 StringRef ScriptLexer::skipSpace(StringRef s) {
144 if (s.startswith("/*")) {
145 size_t e = s.find("*/", 2);
146 if (e == StringRef::npos) {
147 error("unclosed comment in a linker script");
153 if (s.startswith("#")) {
154 size_t e = s.find('\n', 1);
155 if (e == StringRef::npos)
160 size_t size = s.size();
162 if (s.size() == size)
167 // An erroneous token is handled as if it were the last token before EOF.
168 bool ScriptLexer::atEOF() { return errorCount() || tokens.size() == pos; }
170 // Split a given string as an expression.
171 // This function returns "3", "*" and "5" for "3*5" for example.
172 static std::vector<StringRef> tokenizeExpr(StringRef s) {
173 StringRef ops = "+-*/:!~=<>"; // List of operators
175 // Quoted strings are literal strings, so we don't want to split it.
176 if (s.startswith("\""))
179 // Split S with operators as separators.
180 std::vector<StringRef> ret;
182 size_t e = s.find_first_of(ops);
184 // No need to split if there is no operator.
185 if (e == StringRef::npos) {
190 // Get a token before the opreator.
192 ret.push_back(s.substr(0, e));
194 // Get the operator as a token.
195 // Keep !=, ==, >=, <=, << and >> operators as a single tokens.
196 if (s.substr(e).startswith("!=") || s.substr(e).startswith("==") ||
197 s.substr(e).startswith(">=") || s.substr(e).startswith("<=") ||
198 s.substr(e).startswith("<<") || s.substr(e).startswith(">>")) {
199 ret.push_back(s.substr(e, 2));
202 ret.push_back(s.substr(e, 1));
209 // In contexts where expressions are expected, the lexer should apply
210 // different tokenization rules than the default one. By default,
211 // arithmetic operator characters are regular characters, but in the
212 // expression context, they should be independent tokens.
214 // For example, "foo*3" should be tokenized to "foo", "*" and "3" only
215 // in the expression context.
217 // This function may split the current token into multiple tokens.
218 void ScriptLexer::maybeSplitExpr() {
219 if (!inExpr || errorCount() || atEOF())
222 std::vector<StringRef> v = tokenizeExpr(tokens[pos]);
225 tokens.erase(tokens.begin() + pos);
226 tokens.insert(tokens.begin() + pos, v.begin(), v.end());
229 StringRef ScriptLexer::next() {
235 setError("unexpected EOF");
238 return tokens[pos++];
241 StringRef ScriptLexer::peek() {
242 StringRef tok = next();
249 StringRef ScriptLexer::peek2() {
251 StringRef tok = next();
258 bool ScriptLexer::consume(StringRef tok) {
266 // Consumes Tok followed by ":". Space is allowed between Tok and ":".
267 bool ScriptLexer::consumeLabel(StringRef tok) {
268 if (consume((tok + ":").str()))
270 if (tokens.size() >= pos + 2 && tokens[pos] == tok &&
271 tokens[pos + 1] == ":") {
278 void ScriptLexer::skip() { (void)next(); }
280 void ScriptLexer::expect(StringRef expect) {
283 StringRef tok = next();
285 setError(expect + " expected, but got " + tok);
288 // Returns true if S encloses T.
289 static bool encloses(StringRef s, StringRef t) {
290 return s.bytes_begin() <= t.bytes_begin() && t.bytes_end() <= s.bytes_end();
293 MemoryBufferRef ScriptLexer::getCurrentMB() {
294 // Find input buffer containing the current token.
295 assert(!mbs.empty() && pos > 0);
296 for (MemoryBufferRef mb : mbs)
297 if (encloses(mb.getBuffer(), tokens[pos - 1]))
299 llvm_unreachable("getCurrentMB: failed to find a token");