1 | //===- ScriptLexer.cpp ----------------------------------------------------===// |
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 defines a lexer for the linker script. |
10 | // |
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 |
14 | // script processor. |
15 | // |
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. |
22 | // |
23 | // Overall, this lexer works fine for most linker scripts. There might |
24 | // be room for improving compatibility, but that's probably not at the |
25 | // top of our todo list. |
26 | // |
27 | //===----------------------------------------------------------------------===// |
28 | |
29 | #include "ScriptLexer.h" |
30 | #include "Config.h" |
31 | #include "llvm/ADT/Twine.h" |
32 | #include "llvm/Support/ErrorHandling.h" |
33 | #include "llvm/Support/FileSystem.h" |
34 | #include "llvm/Support/Path.h" |
35 | |
36 | using namespace llvm; |
37 | using namespace lld; |
38 | using namespace lld::elf; |
39 | |
40 | ScriptLexer::Buffer::Buffer(Ctx &ctx, MemoryBufferRef mb) |
41 | : s(mb.getBuffer()), filename(mb.getBufferIdentifier()), |
42 | begin(mb.getBufferStart()) { |
43 | if (ctx.arg.sysroot == "" ) |
44 | return; |
45 | StringRef path = filename; |
46 | for (; !path.empty(); path = sys::path::parent_path(path)) { |
47 | if (!sys::fs::equivalent(A: ctx.arg.sysroot, B: path)) |
48 | continue; |
49 | isUnderSysroot = true; |
50 | return; |
51 | } |
52 | } |
53 | |
54 | ScriptLexer::ScriptLexer(Ctx &ctx, MemoryBufferRef mb) |
55 | : ctx(ctx), curBuf(ctx, mb), mbs(1, mb) { |
56 | activeFilenames.insert(V: mb.getBufferIdentifier()); |
57 | } |
58 | |
59 | // Returns a whole line containing the current token. |
60 | StringRef ScriptLexer::getLine() { |
61 | StringRef s = getCurrentMB().getBuffer(); |
62 | |
63 | size_t pos = s.rfind(C: '\n', From: prevTok.data() - s.data()); |
64 | if (pos != StringRef::npos) |
65 | s = s.substr(Start: pos + 1); |
66 | return s.substr(Start: 0, N: s.find_first_of(Chars: "\r\n" )); |
67 | } |
68 | |
69 | // Returns 0-based column number of the current token. |
70 | size_t ScriptLexer::getColumnNumber() { |
71 | return prevTok.data() - getLine().data(); |
72 | } |
73 | |
74 | std::string ScriptLexer::getCurrentLocation() { |
75 | std::string filename = std::string(getCurrentMB().getBufferIdentifier()); |
76 | return (filename + ":" + Twine(prevTokLine)).str(); |
77 | } |
78 | |
79 | // We don't want to record cascading errors. Keep only the first one. |
80 | void ScriptLexer::setError(const Twine &msg) { |
81 | if (errCount(ctx)) |
82 | return; |
83 | |
84 | std::string s = (getCurrentLocation() + ": " + msg).str(); |
85 | if (prevTok.size()) |
86 | s += "\n>>> " + getLine().str() + "\n>>> " + |
87 | std::string(getColumnNumber(), ' ') + "^" ; |
88 | ErrAlways(ctx) << s; |
89 | } |
90 | |
91 | void ScriptLexer::lex() { |
92 | for (;;) { |
93 | StringRef &s = curBuf.s; |
94 | s = skipSpace(s); |
95 | if (s.empty()) { |
96 | // If this buffer is from an INCLUDE command, switch to the "return |
97 | // value"; otherwise, mark EOF. |
98 | if (buffers.empty()) { |
99 | eof = true; |
100 | return; |
101 | } |
102 | activeFilenames.erase(V: curBuf.filename); |
103 | curBuf = buffers.pop_back_val(); |
104 | continue; |
105 | } |
106 | curTokState = lexState; |
107 | |
108 | // Quoted token. Note that double-quote characters are parts of a token |
109 | // because, in a glob match context, only unquoted tokens are interpreted |
110 | // as glob patterns. Double-quoted tokens are literal patterns in that |
111 | // context. |
112 | if (s.starts_with(Prefix: "\"" )) { |
113 | size_t e = s.find(Str: "\"" , From: 1); |
114 | if (e == StringRef::npos) { |
115 | size_t lineno = |
116 | StringRef(curBuf.begin, s.data() - curBuf.begin).count(C: '\n'); |
117 | ErrAlways(ctx) << curBuf.filename << ":" << (lineno + 1) |
118 | << ": unclosed quote" ; |
119 | return; |
120 | } |
121 | |
122 | curTok = s.take_front(N: e + 1); |
123 | s = s.substr(Start: e + 1); |
124 | return; |
125 | } |
126 | |
127 | // Some operators form separate tokens. |
128 | if (s.starts_with(Prefix: "<<=" ) || s.starts_with(Prefix: ">>=" )) { |
129 | curTok = s.substr(Start: 0, N: 3); |
130 | s = s.substr(Start: 3); |
131 | return; |
132 | } |
133 | if (s.size() > 1 && (s[1] == '=' && strchr(s: "+-*/!&^|" , c: s[0]))) { |
134 | curTok = s.substr(Start: 0, N: 2); |
135 | s = s.substr(Start: 2); |
136 | return; |
137 | } |
138 | |
139 | // Unquoted token. The non-expression token is more relaxed than tokens in |
140 | // C-like languages, so that you can write "file-name.cpp" as one bare |
141 | // token. |
142 | size_t pos; |
143 | switch (lexState) { |
144 | case State::Script: |
145 | pos = s.find_first_not_of( |
146 | Chars: "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz" |
147 | "0123456789_.$/\\~=+[]*?-!^:" ); |
148 | break; |
149 | case State::Expr: |
150 | pos = s.find_first_not_of( |
151 | Chars: "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz" |
152 | "0123456789_.$" ); |
153 | if (pos == 0 && s.size() >= 2 && |
154 | ((s[0] == s[1] && strchr(s: "<>&|" , c: s[0])) || |
155 | is_contained(Set: {"==" , "!=" , "<=" , ">=" , "<<" , ">>" }, Element: s.substr(Start: 0, N: 2)))) |
156 | pos = 2; |
157 | break; |
158 | } |
159 | |
160 | if (pos == 0) |
161 | pos = 1; |
162 | curTok = s.substr(Start: 0, N: pos); |
163 | s = s.substr(Start: pos); |
164 | break; |
165 | } |
166 | } |
167 | |
168 | // Skip leading whitespace characters or comments. |
169 | StringRef ScriptLexer::skipSpace(StringRef s) { |
170 | for (;;) { |
171 | if (s.starts_with(Prefix: "/*" )) { |
172 | size_t e = s.find(Str: "*/" , From: 2); |
173 | if (e == StringRef::npos) { |
174 | setError("unclosed comment in a linker script" ); |
175 | return "" ; |
176 | } |
177 | curBuf.lineNumber += s.substr(Start: 0, N: e).count(C: '\n'); |
178 | s = s.substr(Start: e + 2); |
179 | continue; |
180 | } |
181 | if (s.starts_with(Prefix: "#" )) { |
182 | size_t e = s.find(C: '\n', From: 1); |
183 | if (e == StringRef::npos) |
184 | e = s.size() - 1; |
185 | else |
186 | ++curBuf.lineNumber; |
187 | s = s.substr(Start: e + 1); |
188 | continue; |
189 | } |
190 | StringRef saved = s; |
191 | s = s.ltrim(); |
192 | auto len = saved.size() - s.size(); |
193 | if (len == 0) |
194 | return s; |
195 | curBuf.lineNumber += saved.substr(Start: 0, N: len).count(C: '\n'); |
196 | } |
197 | } |
198 | |
199 | // Used to determine whether to stop parsing. Treat errors like EOF. |
200 | bool ScriptLexer::atEOF() { return eof || errCount(ctx); } |
201 | |
202 | StringRef ScriptLexer::next() { |
203 | prevTok = peek(); |
204 | // `prevTokLine` is not updated for EOF so that the line number in `setError` |
205 | // will be more useful. |
206 | if (prevTok.size()) |
207 | prevTokLine = curBuf.lineNumber; |
208 | return std::exchange(obj&: curTok, new_val: StringRef(curBuf.s.data(), 0)); |
209 | } |
210 | |
211 | StringRef ScriptLexer::peek() { |
212 | // curTok is invalid if curTokState and lexState mismatch. |
213 | if (curTok.size() && curTokState != lexState) { |
214 | curBuf.s = StringRef(curTok.data(), curBuf.s.end() - curTok.data()); |
215 | curTok = {}; |
216 | } |
217 | if (curTok.empty()) |
218 | lex(); |
219 | return curTok; |
220 | } |
221 | |
222 | bool ScriptLexer::consume(StringRef tok) { |
223 | if (peek() != tok) |
224 | return false; |
225 | next(); |
226 | return true; |
227 | } |
228 | |
229 | void ScriptLexer::skip() { (void)next(); } |
230 | |
231 | void ScriptLexer::expect(StringRef expect) { |
232 | if (errCount(ctx)) |
233 | return; |
234 | StringRef tok = next(); |
235 | if (tok != expect) { |
236 | if (atEOF()) |
237 | setError("unexpected EOF" ); |
238 | else |
239 | setError(expect + " expected, but got " + tok); |
240 | } |
241 | } |
242 | |
243 | ScriptLexer::Token ScriptLexer::till(StringRef tok) { |
244 | StringRef str = next(); |
245 | if (str == tok) |
246 | return {}; |
247 | if (!atEOF()) |
248 | return {.str: str}; |
249 | prevTok = {}; |
250 | setError("unexpected EOF" ); |
251 | return {}; |
252 | } |
253 | |
254 | // Returns true if S encloses T. |
255 | static bool encloses(StringRef s, StringRef t) { |
256 | return s.bytes_begin() <= t.bytes_begin() && t.bytes_end() <= s.bytes_end(); |
257 | } |
258 | |
259 | MemoryBufferRef ScriptLexer::getCurrentMB() { |
260 | // Find input buffer containing the current token. |
261 | assert(!mbs.empty()); |
262 | for (MemoryBufferRef mb : mbs) |
263 | if (encloses(s: mb.getBuffer(), t: curBuf.s)) |
264 | return mb; |
265 | llvm_unreachable("getCurrentMB: failed to find a token" ); |
266 | } |
267 | |