Commit | Line | Data |
---|---|---|
dbe717ef ILT |
1 | // script.cc -- handle linker scripts for gold. |
2 | ||
62dfdd4d | 3 | // Copyright 2006, 2007, 2008, 2009, 2010 Free Software Foundation, Inc. |
6cb15b7f ILT |
4 | // Written by Ian Lance Taylor <iant@google.com>. |
5 | ||
6 | // This file is part of gold. | |
7 | ||
8 | // This program is free software; you can redistribute it and/or modify | |
9 | // it under the terms of the GNU General Public License as published by | |
10 | // the Free Software Foundation; either version 3 of the License, or | |
11 | // (at your option) any later version. | |
12 | ||
13 | // This program is distributed in the hope that it will be useful, | |
14 | // but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | // GNU General Public License for more details. | |
17 | ||
18 | // You should have received a copy of the GNU General Public License | |
19 | // along with this program; if not, write to the Free Software | |
20 | // Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, | |
21 | // MA 02110-1301, USA. | |
22 | ||
dbe717ef ILT |
23 | #include "gold.h" |
24 | ||
04bf7072 ILT |
25 | #include <cstdio> |
26 | #include <cstdlib> | |
27 | #include <cstring> | |
09124467 | 28 | #include <fnmatch.h> |
dbe717ef ILT |
29 | #include <string> |
30 | #include <vector> | |
ad2d6943 | 31 | #include "filenames.h" |
dbe717ef | 32 | |
e5756efb | 33 | #include "elfcpp.h" |
09124467 | 34 | #include "demangle.h" |
3c2fafa5 | 35 | #include "dirsearch.h" |
dbe717ef ILT |
36 | #include "options.h" |
37 | #include "fileread.h" | |
38 | #include "workqueue.h" | |
39 | #include "readsyms.h" | |
ad2d6943 | 40 | #include "parameters.h" |
d391083d | 41 | #include "layout.h" |
e5756efb | 42 | #include "symtab.h" |
15f8229b | 43 | #include "target-select.h" |
dbe717ef ILT |
44 | #include "script.h" |
45 | #include "script-c.h" | |
072fe7ce | 46 | #include "incremental.h" |
dbe717ef ILT |
47 | |
48 | namespace gold | |
49 | { | |
50 | ||
51 | // A token read from a script file. We don't implement keywords here; | |
52 | // all keywords are simply represented as a string. | |
53 | ||
54 | class Token | |
55 | { | |
56 | public: | |
57 | // Token classification. | |
58 | enum Classification | |
59 | { | |
60 | // Token is invalid. | |
61 | TOKEN_INVALID, | |
62 | // Token indicates end of input. | |
63 | TOKEN_EOF, | |
64 | // Token is a string of characters. | |
65 | TOKEN_STRING, | |
e5756efb ILT |
66 | // Token is a quoted string of characters. |
67 | TOKEN_QUOTED_STRING, | |
dbe717ef ILT |
68 | // Token is an operator. |
69 | TOKEN_OPERATOR, | |
70 | // Token is a number (an integer). | |
71 | TOKEN_INTEGER | |
72 | }; | |
73 | ||
74 | // We need an empty constructor so that we can put this STL objects. | |
75 | Token() | |
e5756efb ILT |
76 | : classification_(TOKEN_INVALID), value_(NULL), value_length_(0), |
77 | opcode_(0), lineno_(0), charpos_(0) | |
dbe717ef ILT |
78 | { } |
79 | ||
80 | // A general token with no value. | |
2ea97941 ILT |
81 | Token(Classification classification, int lineno, int charpos) |
82 | : classification_(classification), value_(NULL), value_length_(0), | |
83 | opcode_(0), lineno_(lineno), charpos_(charpos) | |
a3ad94ed | 84 | { |
2ea97941 ILT |
85 | gold_assert(classification == TOKEN_INVALID |
86 | || classification == TOKEN_EOF); | |
a3ad94ed | 87 | } |
dbe717ef ILT |
88 | |
89 | // A general token with a value. | |
2ea97941 ILT |
90 | Token(Classification classification, const char* value, size_t length, |
91 | int lineno, int charpos) | |
92 | : classification_(classification), value_(value), value_length_(length), | |
93 | opcode_(0), lineno_(lineno), charpos_(charpos) | |
a3ad94ed | 94 | { |
2ea97941 ILT |
95 | gold_assert(classification != TOKEN_INVALID |
96 | && classification != TOKEN_EOF); | |
a3ad94ed | 97 | } |
dbe717ef | 98 | |
dbe717ef | 99 | // A token representing an operator. |
2ea97941 | 100 | Token(int opcode, int lineno, int charpos) |
e5756efb | 101 | : classification_(TOKEN_OPERATOR), value_(NULL), value_length_(0), |
2ea97941 | 102 | opcode_(opcode), lineno_(lineno), charpos_(charpos) |
dbe717ef ILT |
103 | { } |
104 | ||
105 | // Return whether the token is invalid. | |
106 | bool | |
107 | is_invalid() const | |
108 | { return this->classification_ == TOKEN_INVALID; } | |
109 | ||
110 | // Return whether this is an EOF token. | |
111 | bool | |
112 | is_eof() const | |
113 | { return this->classification_ == TOKEN_EOF; } | |
114 | ||
115 | // Return the token classification. | |
116 | Classification | |
117 | classification() const | |
118 | { return this->classification_; } | |
119 | ||
120 | // Return the line number at which the token starts. | |
121 | int | |
122 | lineno() const | |
123 | { return this->lineno_; } | |
124 | ||
125 | // Return the character position at this the token starts. | |
126 | int | |
127 | charpos() const | |
128 | { return this->charpos_; } | |
129 | ||
130 | // Get the value of a token. | |
131 | ||
e5756efb ILT |
132 | const char* |
133 | string_value(size_t* length) const | |
dbe717ef | 134 | { |
e5756efb ILT |
135 | gold_assert(this->classification_ == TOKEN_STRING |
136 | || this->classification_ == TOKEN_QUOTED_STRING); | |
137 | *length = this->value_length_; | |
dbe717ef ILT |
138 | return this->value_; |
139 | } | |
140 | ||
141 | int | |
142 | operator_value() const | |
143 | { | |
a3ad94ed | 144 | gold_assert(this->classification_ == TOKEN_OPERATOR); |
dbe717ef ILT |
145 | return this->opcode_; |
146 | } | |
147 | ||
e5756efb | 148 | uint64_t |
dbe717ef ILT |
149 | integer_value() const |
150 | { | |
a3ad94ed | 151 | gold_assert(this->classification_ == TOKEN_INTEGER); |
e5756efb ILT |
152 | // Null terminate. |
153 | std::string s(this->value_, this->value_length_); | |
154 | return strtoull(s.c_str(), NULL, 0); | |
dbe717ef ILT |
155 | } |
156 | ||
157 | private: | |
158 | // The token classification. | |
159 | Classification classification_; | |
e5756efb ILT |
160 | // The token value, for TOKEN_STRING or TOKEN_QUOTED_STRING or |
161 | // TOKEN_INTEGER. | |
162 | const char* value_; | |
163 | // The length of the token value. | |
164 | size_t value_length_; | |
dbe717ef ILT |
165 | // The token value, for TOKEN_OPERATOR. |
166 | int opcode_; | |
167 | // The line number where this token started (one based). | |
168 | int lineno_; | |
169 | // The character position within the line where this token started | |
170 | // (one based). | |
171 | int charpos_; | |
172 | }; | |
173 | ||
e5756efb | 174 | // This class handles lexing a file into a sequence of tokens. |
dbe717ef ILT |
175 | |
176 | class Lex | |
177 | { | |
178 | public: | |
e5756efb ILT |
179 | // We unfortunately have to support different lexing modes, because |
180 | // when reading different parts of a linker script we need to parse | |
181 | // things differently. | |
182 | enum Mode | |
183 | { | |
184 | // Reading an ordinary linker script. | |
185 | LINKER_SCRIPT, | |
186 | // Reading an expression in a linker script. | |
187 | EXPRESSION, | |
188 | // Reading a version script. | |
c82fbeee CS |
189 | VERSION_SCRIPT, |
190 | // Reading a --dynamic-list file. | |
191 | DYNAMIC_LIST | |
e5756efb ILT |
192 | }; |
193 | ||
194 | Lex(const char* input_string, size_t input_length, int parsing_token) | |
195 | : input_string_(input_string), input_length_(input_length), | |
196 | current_(input_string), mode_(LINKER_SCRIPT), | |
197 | first_token_(parsing_token), token_(), | |
198 | lineno_(1), linestart_(input_string) | |
dbe717ef ILT |
199 | { } |
200 | ||
e5756efb ILT |
201 | // Read a file into a string. |
202 | static void | |
203 | read_file(Input_file*, std::string*); | |
204 | ||
205 | // Return the next token. | |
206 | const Token* | |
207 | next_token(); | |
dbe717ef | 208 | |
e5756efb ILT |
209 | // Return the current lexing mode. |
210 | Lex::Mode | |
211 | mode() const | |
212 | { return this->mode_; } | |
dbe717ef | 213 | |
e5756efb ILT |
214 | // Set the lexing mode. |
215 | void | |
2ea97941 ILT |
216 | set_mode(Mode mode) |
217 | { this->mode_ = mode; } | |
dbe717ef ILT |
218 | |
219 | private: | |
220 | Lex(const Lex&); | |
221 | Lex& operator=(const Lex&); | |
222 | ||
dbe717ef ILT |
223 | // Make a general token with no value at the current location. |
224 | Token | |
e5756efb ILT |
225 | make_token(Token::Classification c, const char* start) const |
226 | { return Token(c, this->lineno_, start - this->linestart_ + 1); } | |
dbe717ef ILT |
227 | |
228 | // Make a general token with a value at the current location. | |
229 | Token | |
e5756efb ILT |
230 | make_token(Token::Classification c, const char* v, size_t len, |
231 | const char* start) | |
dbe717ef | 232 | const |
e5756efb | 233 | { return Token(c, v, len, this->lineno_, start - this->linestart_ + 1); } |
dbe717ef ILT |
234 | |
235 | // Make an operator token at the current location. | |
236 | Token | |
e5756efb ILT |
237 | make_token(int opcode, const char* start) const |
238 | { return Token(opcode, this->lineno_, start - this->linestart_ + 1); } | |
dbe717ef ILT |
239 | |
240 | // Make an invalid token at the current location. | |
241 | Token | |
e5756efb ILT |
242 | make_invalid_token(const char* start) |
243 | { return this->make_token(Token::TOKEN_INVALID, start); } | |
dbe717ef ILT |
244 | |
245 | // Make an EOF token at the current location. | |
246 | Token | |
e5756efb ILT |
247 | make_eof_token(const char* start) |
248 | { return this->make_token(Token::TOKEN_EOF, start); } | |
dbe717ef ILT |
249 | |
250 | // Return whether C can be the first character in a name. C2 is the | |
251 | // next character, since we sometimes need that. | |
e5756efb | 252 | inline bool |
dbe717ef ILT |
253 | can_start_name(char c, char c2); |
254 | ||
09124467 ILT |
255 | // If C can appear in a name which has already started, return a |
256 | // pointer to a character later in the token or just past | |
257 | // it. Otherwise, return NULL. | |
258 | inline const char* | |
259 | can_continue_name(const char* c); | |
dbe717ef ILT |
260 | |
261 | // Return whether C, C2, C3 can start a hex number. | |
e5756efb | 262 | inline bool |
dbe717ef ILT |
263 | can_start_hex(char c, char c2, char c3); |
264 | ||
09124467 ILT |
265 | // If C can appear in a hex number which has already started, return |
266 | // a pointer to a character later in the token or just past | |
267 | // it. Otherwise, return NULL. | |
268 | inline const char* | |
269 | can_continue_hex(const char* c); | |
dbe717ef ILT |
270 | |
271 | // Return whether C can start a non-hex number. | |
272 | static inline bool | |
273 | can_start_number(char c); | |
274 | ||
09124467 ILT |
275 | // If C can appear in a decimal number which has already started, |
276 | // return a pointer to a character later in the token or just past | |
277 | // it. Otherwise, return NULL. | |
278 | inline const char* | |
279 | can_continue_number(const char* c) | |
280 | { return Lex::can_start_number(*c) ? c + 1 : NULL; } | |
dbe717ef ILT |
281 | |
282 | // If C1 C2 C3 form a valid three character operator, return the | |
283 | // opcode. Otherwise return 0. | |
284 | static inline int | |
285 | three_char_operator(char c1, char c2, char c3); | |
286 | ||
287 | // If C1 C2 form a valid two character operator, return the opcode. | |
288 | // Otherwise return 0. | |
289 | static inline int | |
290 | two_char_operator(char c1, char c2); | |
291 | ||
292 | // If C1 is a valid one character operator, return the opcode. | |
293 | // Otherwise return 0. | |
294 | static inline int | |
295 | one_char_operator(char c1); | |
296 | ||
297 | // Read the next token. | |
298 | Token | |
299 | get_token(const char**); | |
300 | ||
301 | // Skip a C style /* */ comment. Return false if the comment did | |
302 | // not end. | |
303 | bool | |
304 | skip_c_comment(const char**); | |
305 | ||
306 | // Skip a line # comment. Return false if there was no newline. | |
307 | bool | |
308 | skip_line_comment(const char**); | |
309 | ||
310 | // Build a token CLASSIFICATION from all characters that match | |
311 | // CAN_CONTINUE_FN. The token starts at START. Start matching from | |
312 | // MATCH. Set *PP to the character following the token. | |
313 | inline Token | |
e5756efb | 314 | gather_token(Token::Classification, |
09124467 | 315 | const char* (Lex::*can_continue_fn)(const char*), |
dbe717ef ILT |
316 | const char* start, const char* match, const char** pp); |
317 | ||
318 | // Build a token from a quoted string. | |
319 | Token | |
320 | gather_quoted_string(const char** pp); | |
321 | ||
e5756efb ILT |
322 | // The string we are tokenizing. |
323 | const char* input_string_; | |
324 | // The length of the string. | |
325 | size_t input_length_; | |
326 | // The current offset into the string. | |
327 | const char* current_; | |
328 | // The current lexing mode. | |
329 | Mode mode_; | |
330 | // The code to use for the first token. This is set to 0 after it | |
331 | // is used. | |
332 | int first_token_; | |
333 | // The current token. | |
334 | Token token_; | |
dbe717ef ILT |
335 | // The current line number. |
336 | int lineno_; | |
e5756efb | 337 | // The start of the current line in the string. |
dbe717ef ILT |
338 | const char* linestart_; |
339 | }; | |
340 | ||
341 | // Read the whole file into memory. We don't expect linker scripts to | |
342 | // be large, so we just use a std::string as a buffer. We ignore the | |
343 | // data we've already read, so that we read aligned buffers. | |
344 | ||
345 | void | |
e5756efb | 346 | Lex::read_file(Input_file* input_file, std::string* contents) |
dbe717ef | 347 | { |
e5756efb | 348 | off_t filesize = input_file->file().filesize(); |
dbe717ef | 349 | contents->clear(); |
82dcae9d ILT |
350 | contents->reserve(filesize); |
351 | ||
dbe717ef | 352 | off_t off = 0; |
dbe717ef | 353 | unsigned char buf[BUFSIZ]; |
82dcae9d | 354 | while (off < filesize) |
dbe717ef | 355 | { |
82dcae9d ILT |
356 | off_t get = BUFSIZ; |
357 | if (get > filesize - off) | |
358 | get = filesize - off; | |
e5756efb | 359 | input_file->file().read(off, get, buf); |
82dcae9d ILT |
360 | contents->append(reinterpret_cast<char*>(&buf[0]), get); |
361 | off += get; | |
dbe717ef | 362 | } |
dbe717ef ILT |
363 | } |
364 | ||
365 | // Return whether C can be the start of a name, if the next character | |
366 | // is C2. A name can being with a letter, underscore, period, or | |
367 | // dollar sign. Because a name can be a file name, we also permit | |
368 | // forward slash, backslash, and tilde. Tilde is the tricky case | |
369 | // here; GNU ld also uses it as a bitwise not operator. It is only | |
370 | // recognized as the operator if it is not immediately followed by | |
e5756efb ILT |
371 | // some character which can appear in a symbol. That is, when we |
372 | // don't know that we are looking at an expression, "~0" is a file | |
373 | // name, and "~ 0" is an expression using bitwise not. We are | |
dbe717ef ILT |
374 | // compatible. |
375 | ||
376 | inline bool | |
377 | Lex::can_start_name(char c, char c2) | |
378 | { | |
379 | switch (c) | |
380 | { | |
381 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
382 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
383 | case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R': | |
384 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
385 | case 'Y': case 'Z': | |
386 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
387 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
388 | case 'm': case 'n': case 'o': case 'q': case 'p': case 'r': | |
389 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
390 | case 'y': case 'z': | |
e5756efb | 391 | case '_': case '.': case '$': |
dbe717ef ILT |
392 | return true; |
393 | ||
e5756efb ILT |
394 | case '/': case '\\': |
395 | return this->mode_ == LINKER_SCRIPT; | |
396 | ||
dbe717ef | 397 | case '~': |
09124467 ILT |
398 | return this->mode_ == LINKER_SCRIPT && can_continue_name(&c2); |
399 | ||
c82fbeee | 400 | case '*': case '[': |
3802b2dd | 401 | return (this->mode_ == VERSION_SCRIPT |
c82fbeee | 402 | || this->mode_ == DYNAMIC_LIST |
3802b2dd ILT |
403 | || (this->mode_ == LINKER_SCRIPT |
404 | && can_continue_name(&c2))); | |
dbe717ef ILT |
405 | |
406 | default: | |
407 | return false; | |
408 | } | |
409 | } | |
410 | ||
411 | // Return whether C can continue a name which has already started. | |
412 | // Subsequent characters in a name are the same as the leading | |
413 | // characters, plus digits and "=+-:[],?*". So in general the linker | |
e5756efb ILT |
414 | // script language requires spaces around operators, unless we know |
415 | // that we are parsing an expression. | |
dbe717ef | 416 | |
09124467 ILT |
417 | inline const char* |
418 | Lex::can_continue_name(const char* c) | |
dbe717ef | 419 | { |
09124467 | 420 | switch (*c) |
dbe717ef ILT |
421 | { |
422 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
423 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
424 | case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R': | |
425 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
426 | case 'Y': case 'Z': | |
427 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
428 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
429 | case 'm': case 'n': case 'o': case 'q': case 'p': case 'r': | |
430 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
431 | case 'y': case 'z': | |
e5756efb | 432 | case '_': case '.': case '$': |
dbe717ef ILT |
433 | case '0': case '1': case '2': case '3': case '4': |
434 | case '5': case '6': case '7': case '8': case '9': | |
09124467 | 435 | return c + 1; |
dbe717ef | 436 | |
c82fbeee | 437 | // TODO(csilvers): why not allow ~ in names for version-scripts? |
e5756efb | 438 | case '/': case '\\': case '~': |
09124467 | 439 | case '=': case '+': |
afe47622 | 440 | case ',': |
09124467 ILT |
441 | if (this->mode_ == LINKER_SCRIPT) |
442 | return c + 1; | |
443 | return NULL; | |
444 | ||
afe47622 | 445 | case '[': case ']': case '*': case '?': case '-': |
c82fbeee CS |
446 | if (this->mode_ == LINKER_SCRIPT || this->mode_ == VERSION_SCRIPT |
447 | || this->mode_ == DYNAMIC_LIST) | |
09124467 ILT |
448 | return c + 1; |
449 | return NULL; | |
450 | ||
c82fbeee | 451 | // TODO(csilvers): why allow this? ^ is meaningless in version scripts. |
09124467 | 452 | case '^': |
c82fbeee | 453 | if (this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST) |
09124467 ILT |
454 | return c + 1; |
455 | return NULL; | |
456 | ||
457 | case ':': | |
458 | if (this->mode_ == LINKER_SCRIPT) | |
459 | return c + 1; | |
c82fbeee CS |
460 | else if ((this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST) |
461 | && (c[1] == ':')) | |
09124467 ILT |
462 | { |
463 | // A name can have '::' in it, as that's a c++ namespace | |
464 | // separator. But a single colon is not part of a name. | |
465 | return c + 2; | |
466 | } | |
467 | return NULL; | |
e5756efb | 468 | |
dbe717ef | 469 | default: |
09124467 | 470 | return NULL; |
dbe717ef ILT |
471 | } |
472 | } | |
473 | ||
474 | // For a number we accept 0x followed by hex digits, or any sequence | |
475 | // of digits. The old linker accepts leading '$' for hex, and | |
476 | // trailing HXBOD. Those are for MRI compatibility and we don't | |
477 | // accept them. The old linker also accepts trailing MK for mega or | |
e5756efb ILT |
478 | // kilo. FIXME: Those are mentioned in the documentation, and we |
479 | // should accept them. | |
dbe717ef ILT |
480 | |
481 | // Return whether C1 C2 C3 can start a hex number. | |
482 | ||
483 | inline bool | |
484 | Lex::can_start_hex(char c1, char c2, char c3) | |
485 | { | |
486 | if (c1 == '0' && (c2 == 'x' || c2 == 'X')) | |
09124467 | 487 | return this->can_continue_hex(&c3); |
dbe717ef ILT |
488 | return false; |
489 | } | |
490 | ||
491 | // Return whether C can appear in a hex number. | |
492 | ||
09124467 ILT |
493 | inline const char* |
494 | Lex::can_continue_hex(const char* c) | |
dbe717ef | 495 | { |
09124467 | 496 | switch (*c) |
dbe717ef ILT |
497 | { |
498 | case '0': case '1': case '2': case '3': case '4': | |
499 | case '5': case '6': case '7': case '8': case '9': | |
500 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
501 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
09124467 | 502 | return c + 1; |
dbe717ef ILT |
503 | |
504 | default: | |
09124467 | 505 | return NULL; |
dbe717ef ILT |
506 | } |
507 | } | |
508 | ||
509 | // Return whether C can start a non-hex number. | |
510 | ||
511 | inline bool | |
512 | Lex::can_start_number(char c) | |
513 | { | |
514 | switch (c) | |
515 | { | |
516 | case '0': case '1': case '2': case '3': case '4': | |
517 | case '5': case '6': case '7': case '8': case '9': | |
518 | return true; | |
519 | ||
520 | default: | |
521 | return false; | |
522 | } | |
523 | } | |
524 | ||
525 | // If C1 C2 C3 form a valid three character operator, return the | |
526 | // opcode (defined in the yyscript.h file generated from yyscript.y). | |
527 | // Otherwise return 0. | |
528 | ||
529 | inline int | |
530 | Lex::three_char_operator(char c1, char c2, char c3) | |
531 | { | |
532 | switch (c1) | |
533 | { | |
534 | case '<': | |
535 | if (c2 == '<' && c3 == '=') | |
536 | return LSHIFTEQ; | |
537 | break; | |
538 | case '>': | |
539 | if (c2 == '>' && c3 == '=') | |
540 | return RSHIFTEQ; | |
541 | break; | |
542 | default: | |
543 | break; | |
544 | } | |
545 | return 0; | |
546 | } | |
547 | ||
548 | // If C1 C2 form a valid two character operator, return the opcode | |
549 | // (defined in the yyscript.h file generated from yyscript.y). | |
550 | // Otherwise return 0. | |
551 | ||
552 | inline int | |
553 | Lex::two_char_operator(char c1, char c2) | |
554 | { | |
555 | switch (c1) | |
556 | { | |
557 | case '=': | |
558 | if (c2 == '=') | |
559 | return EQ; | |
560 | break; | |
561 | case '!': | |
562 | if (c2 == '=') | |
563 | return NE; | |
564 | break; | |
565 | case '+': | |
566 | if (c2 == '=') | |
567 | return PLUSEQ; | |
568 | break; | |
569 | case '-': | |
570 | if (c2 == '=') | |
571 | return MINUSEQ; | |
572 | break; | |
573 | case '*': | |
574 | if (c2 == '=') | |
575 | return MULTEQ; | |
576 | break; | |
577 | case '/': | |
578 | if (c2 == '=') | |
579 | return DIVEQ; | |
580 | break; | |
581 | case '|': | |
582 | if (c2 == '=') | |
583 | return OREQ; | |
584 | if (c2 == '|') | |
585 | return OROR; | |
586 | break; | |
587 | case '&': | |
588 | if (c2 == '=') | |
589 | return ANDEQ; | |
590 | if (c2 == '&') | |
591 | return ANDAND; | |
592 | break; | |
593 | case '>': | |
594 | if (c2 == '=') | |
595 | return GE; | |
596 | if (c2 == '>') | |
597 | return RSHIFT; | |
598 | break; | |
599 | case '<': | |
600 | if (c2 == '=') | |
601 | return LE; | |
602 | if (c2 == '<') | |
603 | return LSHIFT; | |
604 | break; | |
605 | default: | |
606 | break; | |
607 | } | |
608 | return 0; | |
609 | } | |
610 | ||
611 | // If C1 is a valid operator, return the opcode. Otherwise return 0. | |
612 | ||
613 | inline int | |
614 | Lex::one_char_operator(char c1) | |
615 | { | |
616 | switch (c1) | |
617 | { | |
618 | case '+': | |
619 | case '-': | |
620 | case '*': | |
621 | case '/': | |
622 | case '%': | |
623 | case '!': | |
624 | case '&': | |
625 | case '|': | |
626 | case '^': | |
627 | case '~': | |
628 | case '<': | |
629 | case '>': | |
630 | case '=': | |
631 | case '?': | |
632 | case ',': | |
633 | case '(': | |
634 | case ')': | |
635 | case '{': | |
636 | case '}': | |
637 | case '[': | |
638 | case ']': | |
639 | case ':': | |
640 | case ';': | |
641 | return c1; | |
642 | default: | |
643 | return 0; | |
644 | } | |
645 | } | |
646 | ||
647 | // Skip a C style comment. *PP points to just after the "/*". Return | |
648 | // false if the comment did not end. | |
649 | ||
650 | bool | |
651 | Lex::skip_c_comment(const char** pp) | |
652 | { | |
653 | const char* p = *pp; | |
654 | while (p[0] != '*' || p[1] != '/') | |
655 | { | |
656 | if (*p == '\0') | |
657 | { | |
658 | *pp = p; | |
659 | return false; | |
660 | } | |
661 | ||
662 | if (*p == '\n') | |
663 | { | |
664 | ++this->lineno_; | |
665 | this->linestart_ = p + 1; | |
666 | } | |
667 | ++p; | |
668 | } | |
669 | ||
670 | *pp = p + 2; | |
671 | return true; | |
672 | } | |
673 | ||
674 | // Skip a line # comment. Return false if there was no newline. | |
675 | ||
676 | bool | |
677 | Lex::skip_line_comment(const char** pp) | |
678 | { | |
679 | const char* p = *pp; | |
680 | size_t skip = strcspn(p, "\n"); | |
681 | if (p[skip] == '\0') | |
682 | { | |
683 | *pp = p + skip; | |
684 | return false; | |
685 | } | |
686 | ||
687 | p += skip + 1; | |
688 | ++this->lineno_; | |
689 | this->linestart_ = p; | |
690 | *pp = p; | |
691 | ||
692 | return true; | |
693 | } | |
694 | ||
695 | // Build a token CLASSIFICATION from all characters that match | |
696 | // CAN_CONTINUE_FN. Update *PP. | |
697 | ||
698 | inline Token | |
699 | Lex::gather_token(Token::Classification classification, | |
09124467 | 700 | const char* (Lex::*can_continue_fn)(const char*), |
dbe717ef ILT |
701 | const char* start, |
702 | const char* match, | |
703 | const char **pp) | |
704 | { | |
09124467 ILT |
705 | const char* new_match = NULL; |
706 | while ((new_match = (this->*can_continue_fn)(match))) | |
707 | match = new_match; | |
dbe717ef | 708 | *pp = match; |
e5756efb | 709 | return this->make_token(classification, start, match - start, start); |
dbe717ef ILT |
710 | } |
711 | ||
712 | // Build a token from a quoted string. | |
713 | ||
714 | Token | |
715 | Lex::gather_quoted_string(const char** pp) | |
716 | { | |
717 | const char* start = *pp; | |
718 | const char* p = start; | |
719 | ++p; | |
720 | size_t skip = strcspn(p, "\"\n"); | |
721 | if (p[skip] != '"') | |
722 | return this->make_invalid_token(start); | |
723 | *pp = p + skip + 1; | |
e5756efb | 724 | return this->make_token(Token::TOKEN_QUOTED_STRING, p, skip, start); |
dbe717ef ILT |
725 | } |
726 | ||
727 | // Return the next token at *PP. Update *PP. General guideline: we | |
728 | // require linker scripts to be simple ASCII. No unicode linker | |
729 | // scripts. In particular we can assume that any '\0' is the end of | |
730 | // the input. | |
731 | ||
732 | Token | |
733 | Lex::get_token(const char** pp) | |
734 | { | |
735 | const char* p = *pp; | |
736 | ||
737 | while (true) | |
738 | { | |
739 | if (*p == '\0') | |
740 | { | |
741 | *pp = p; | |
742 | return this->make_eof_token(p); | |
743 | } | |
744 | ||
745 | // Skip whitespace quickly. | |
f3048a1d | 746 | while (*p == ' ' || *p == '\t' || *p == '\r') |
dbe717ef ILT |
747 | ++p; |
748 | ||
749 | if (*p == '\n') | |
750 | { | |
751 | ++p; | |
752 | ++this->lineno_; | |
753 | this->linestart_ = p; | |
754 | continue; | |
755 | } | |
756 | ||
757 | // Skip C style comments. | |
758 | if (p[0] == '/' && p[1] == '*') | |
759 | { | |
760 | int lineno = this->lineno_; | |
761 | int charpos = p - this->linestart_ + 1; | |
762 | ||
763 | *pp = p + 2; | |
764 | if (!this->skip_c_comment(pp)) | |
765 | return Token(Token::TOKEN_INVALID, lineno, charpos); | |
766 | p = *pp; | |
767 | ||
768 | continue; | |
769 | } | |
770 | ||
771 | // Skip line comments. | |
772 | if (*p == '#') | |
773 | { | |
774 | *pp = p + 1; | |
775 | if (!this->skip_line_comment(pp)) | |
776 | return this->make_eof_token(p); | |
777 | p = *pp; | |
778 | continue; | |
779 | } | |
780 | ||
781 | // Check for a name. | |
e5756efb | 782 | if (this->can_start_name(p[0], p[1])) |
dbe717ef | 783 | return this->gather_token(Token::TOKEN_STRING, |
e5756efb ILT |
784 | &Lex::can_continue_name, |
785 | p, p + 1, pp); | |
dbe717ef ILT |
786 | |
787 | // We accept any arbitrary name in double quotes, as long as it | |
788 | // does not cross a line boundary. | |
789 | if (*p == '"') | |
790 | { | |
791 | *pp = p; | |
792 | return this->gather_quoted_string(pp); | |
793 | } | |
794 | ||
795 | // Check for a number. | |
796 | ||
e5756efb | 797 | if (this->can_start_hex(p[0], p[1], p[2])) |
dbe717ef | 798 | return this->gather_token(Token::TOKEN_INTEGER, |
e5756efb | 799 | &Lex::can_continue_hex, |
dbe717ef ILT |
800 | p, p + 3, pp); |
801 | ||
802 | if (Lex::can_start_number(p[0])) | |
803 | return this->gather_token(Token::TOKEN_INTEGER, | |
e5756efb | 804 | &Lex::can_continue_number, |
dbe717ef ILT |
805 | p, p + 1, pp); |
806 | ||
807 | // Check for operators. | |
808 | ||
809 | int opcode = Lex::three_char_operator(p[0], p[1], p[2]); | |
810 | if (opcode != 0) | |
811 | { | |
812 | *pp = p + 3; | |
813 | return this->make_token(opcode, p); | |
814 | } | |
815 | ||
816 | opcode = Lex::two_char_operator(p[0], p[1]); | |
817 | if (opcode != 0) | |
818 | { | |
819 | *pp = p + 2; | |
820 | return this->make_token(opcode, p); | |
821 | } | |
822 | ||
823 | opcode = Lex::one_char_operator(p[0]); | |
824 | if (opcode != 0) | |
825 | { | |
826 | *pp = p + 1; | |
827 | return this->make_token(opcode, p); | |
828 | } | |
829 | ||
830 | return this->make_token(Token::TOKEN_INVALID, p); | |
831 | } | |
832 | } | |
833 | ||
e5756efb | 834 | // Return the next token. |
dbe717ef | 835 | |
e5756efb ILT |
836 | const Token* |
837 | Lex::next_token() | |
dbe717ef | 838 | { |
e5756efb ILT |
839 | // The first token is special. |
840 | if (this->first_token_ != 0) | |
dbe717ef | 841 | { |
e5756efb ILT |
842 | this->token_ = Token(this->first_token_, 0, 0); |
843 | this->first_token_ = 0; | |
844 | return &this->token_; | |
845 | } | |
dbe717ef | 846 | |
e5756efb | 847 | this->token_ = this->get_token(&this->current_); |
dbe717ef | 848 | |
e5756efb ILT |
849 | // Don't let an early null byte fool us into thinking that we've |
850 | // reached the end of the file. | |
851 | if (this->token_.is_eof() | |
852 | && (static_cast<size_t>(this->current_ - this->input_string_) | |
853 | < this->input_length_)) | |
854 | this->token_ = this->make_invalid_token(this->current_); | |
dbe717ef | 855 | |
e5756efb | 856 | return &this->token_; |
dbe717ef ILT |
857 | } |
858 | ||
494e05f4 | 859 | // class Symbol_assignment. |
e5756efb | 860 | |
494e05f4 ILT |
861 | // Add the symbol to the symbol table. This makes sure the symbol is |
862 | // there and defined. The actual value is stored later. We can't | |
863 | // determine the actual value at this point, because we can't | |
864 | // necessarily evaluate the expression until all ordinary symbols have | |
865 | // been finalized. | |
e5756efb | 866 | |
caa9d5d9 ILT |
867 | // The GNU linker lets symbol assignments in the linker script |
868 | // silently override defined symbols in object files. We are | |
869 | // compatible. FIXME: Should we issue a warning? | |
870 | ||
e5756efb | 871 | void |
9b07f471 | 872 | Symbol_assignment::add_to_table(Symbol_table* symtab) |
e5756efb | 873 | { |
494e05f4 | 874 | elfcpp::STV vis = this->hidden_ ? elfcpp::STV_HIDDEN : elfcpp::STV_DEFAULT; |
9b07f471 | 875 | this->sym_ = symtab->define_as_constant(this->name_.c_str(), |
e5756efb | 876 | NULL, // version |
99fff23b ILT |
877 | (this->is_defsym_ |
878 | ? Symbol_table::DEFSYM | |
879 | : Symbol_table::SCRIPT), | |
e5756efb ILT |
880 | 0, // value |
881 | 0, // size | |
882 | elfcpp::STT_NOTYPE, | |
883 | elfcpp::STB_GLOBAL, | |
884 | vis, | |
885 | 0, // nonvis | |
caa9d5d9 ILT |
886 | this->provide_, |
887 | true); // force_override | |
e5756efb ILT |
888 | } |
889 | ||
494e05f4 | 890 | // Finalize a symbol value. |
e5756efb ILT |
891 | |
892 | void | |
494e05f4 | 893 | Symbol_assignment::finalize(Symbol_table* symtab, const Layout* layout) |
a445fddf | 894 | { |
77e65537 | 895 | this->finalize_maybe_dot(symtab, layout, false, 0, NULL); |
a445fddf ILT |
896 | } |
897 | ||
898 | // Finalize a symbol value which can refer to the dot symbol. | |
899 | ||
900 | void | |
901 | Symbol_assignment::finalize_with_dot(Symbol_table* symtab, | |
902 | const Layout* layout, | |
77e65537 ILT |
903 | uint64_t dot_value, |
904 | Output_section* dot_section) | |
a445fddf | 905 | { |
77e65537 | 906 | this->finalize_maybe_dot(symtab, layout, true, dot_value, dot_section); |
a445fddf ILT |
907 | } |
908 | ||
909 | // Finalize a symbol value, internal version. | |
910 | ||
911 | void | |
912 | Symbol_assignment::finalize_maybe_dot(Symbol_table* symtab, | |
913 | const Layout* layout, | |
914 | bool is_dot_available, | |
77e65537 ILT |
915 | uint64_t dot_value, |
916 | Output_section* dot_section) | |
e5756efb | 917 | { |
494e05f4 ILT |
918 | // If we were only supposed to provide this symbol, the sym_ field |
919 | // will be NULL if the symbol was not referenced. | |
920 | if (this->sym_ == NULL) | |
921 | { | |
922 | gold_assert(this->provide_); | |
923 | return; | |
924 | } | |
925 | ||
8851ecca | 926 | if (parameters->target().get_size() == 32) |
e5756efb ILT |
927 | { |
928 | #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG) | |
77e65537 ILT |
929 | this->sized_finalize<32>(symtab, layout, is_dot_available, dot_value, |
930 | dot_section); | |
e5756efb ILT |
931 | #else |
932 | gold_unreachable(); | |
933 | #endif | |
934 | } | |
8851ecca | 935 | else if (parameters->target().get_size() == 64) |
e5756efb ILT |
936 | { |
937 | #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG) | |
77e65537 ILT |
938 | this->sized_finalize<64>(symtab, layout, is_dot_available, dot_value, |
939 | dot_section); | |
e5756efb ILT |
940 | #else |
941 | gold_unreachable(); | |
942 | #endif | |
943 | } | |
944 | else | |
945 | gold_unreachable(); | |
946 | } | |
947 | ||
948 | template<int size> | |
949 | void | |
a445fddf | 950 | Symbol_assignment::sized_finalize(Symbol_table* symtab, const Layout* layout, |
77e65537 ILT |
951 | bool is_dot_available, uint64_t dot_value, |
952 | Output_section* dot_section) | |
a445fddf | 953 | { |
77e65537 | 954 | Output_section* section; |
919ed24c | 955 | uint64_t final_val = this->val_->eval_maybe_dot(symtab, layout, true, |
a445fddf | 956 | is_dot_available, |
77e65537 ILT |
957 | dot_value, dot_section, |
958 | §ion); | |
494e05f4 | 959 | Sized_symbol<size>* ssym = symtab->get_sized_symbol<size>(this->sym_); |
a445fddf | 960 | ssym->set_value(final_val); |
77e65537 ILT |
961 | if (section != NULL) |
962 | ssym->set_output_section(section); | |
a445fddf ILT |
963 | } |
964 | ||
965 | // Set the symbol value if the expression yields an absolute value. | |
966 | ||
967 | void | |
968 | Symbol_assignment::set_if_absolute(Symbol_table* symtab, const Layout* layout, | |
77e65537 | 969 | bool is_dot_available, uint64_t dot_value) |
a445fddf ILT |
970 | { |
971 | if (this->sym_ == NULL) | |
972 | return; | |
973 | ||
77e65537 | 974 | Output_section* val_section; |
919ed24c ILT |
975 | uint64_t val = this->val_->eval_maybe_dot(symtab, layout, false, |
976 | is_dot_available, dot_value, | |
977 | NULL, &val_section); | |
77e65537 | 978 | if (val_section != NULL) |
a445fddf ILT |
979 | return; |
980 | ||
8851ecca | 981 | if (parameters->target().get_size() == 32) |
a445fddf ILT |
982 | { |
983 | #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG) | |
984 | Sized_symbol<32>* ssym = symtab->get_sized_symbol<32>(this->sym_); | |
985 | ssym->set_value(val); | |
986 | #else | |
987 | gold_unreachable(); | |
988 | #endif | |
989 | } | |
8851ecca | 990 | else if (parameters->target().get_size() == 64) |
a445fddf ILT |
991 | { |
992 | #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG) | |
993 | Sized_symbol<64>* ssym = symtab->get_sized_symbol<64>(this->sym_); | |
994 | ssym->set_value(val); | |
995 | #else | |
996 | gold_unreachable(); | |
997 | #endif | |
998 | } | |
999 | else | |
1000 | gold_unreachable(); | |
494e05f4 ILT |
1001 | } |
1002 | ||
1003 | // Print for debugging. | |
1004 | ||
1005 | void | |
1006 | Symbol_assignment::print(FILE* f) const | |
1007 | { | |
1008 | if (this->provide_ && this->hidden_) | |
1009 | fprintf(f, "PROVIDE_HIDDEN("); | |
1010 | else if (this->provide_) | |
1011 | fprintf(f, "PROVIDE("); | |
1012 | else if (this->hidden_) | |
1013 | gold_unreachable(); | |
1014 | ||
1015 | fprintf(f, "%s = ", this->name_.c_str()); | |
1016 | this->val_->print(f); | |
1017 | ||
1018 | if (this->provide_ || this->hidden_) | |
1019 | fprintf(f, ")"); | |
1020 | ||
1021 | fprintf(f, "\n"); | |
1022 | } | |
1023 | ||
1024 | // Class Script_assertion. | |
1025 | ||
1026 | // Check the assertion. | |
1027 | ||
1028 | void | |
1029 | Script_assertion::check(const Symbol_table* symtab, const Layout* layout) | |
1030 | { | |
919ed24c | 1031 | if (!this->check_->eval(symtab, layout, true)) |
494e05f4 ILT |
1032 | gold_error("%s", this->message_.c_str()); |
1033 | } | |
1034 | ||
1035 | // Print for debugging. | |
1036 | ||
1037 | void | |
1038 | Script_assertion::print(FILE* f) const | |
1039 | { | |
1040 | fprintf(f, "ASSERT("); | |
1041 | this->check_->print(f); | |
1042 | fprintf(f, ", \"%s\")\n", this->message_.c_str()); | |
1043 | } | |
1044 | ||
1045 | // Class Script_options. | |
1046 | ||
1047 | Script_options::Script_options() | |
1048 | : entry_(), symbol_assignments_(), version_script_info_(), | |
1049 | script_sections_() | |
1050 | { | |
1051 | } | |
1052 | ||
1053 | // Add a symbol to be defined. | |
1054 | ||
1055 | void | |
1056 | Script_options::add_symbol_assignment(const char* name, size_t length, | |
99fff23b ILT |
1057 | bool is_defsym, Expression* value, |
1058 | bool provide, bool hidden) | |
494e05f4 | 1059 | { |
a445fddf ILT |
1060 | if (length != 1 || name[0] != '.') |
1061 | { | |
1062 | if (this->script_sections_.in_sections_clause()) | |
a445fddf | 1063 | { |
99fff23b ILT |
1064 | gold_assert(!is_defsym); |
1065 | this->script_sections_.add_symbol_assignment(name, length, value, | |
a445fddf | 1066 | provide, hidden); |
99fff23b ILT |
1067 | } |
1068 | else | |
1069 | { | |
1070 | Symbol_assignment* p = new Symbol_assignment(name, length, is_defsym, | |
1071 | value, provide, hidden); | |
a445fddf ILT |
1072 | this->symbol_assignments_.push_back(p); |
1073 | } | |
1074 | } | |
494e05f4 ILT |
1075 | else |
1076 | { | |
a445fddf ILT |
1077 | if (provide || hidden) |
1078 | gold_error(_("invalid use of PROVIDE for dot symbol")); | |
12edd763 ILT |
1079 | |
1080 | // The GNU linker permits assignments to dot outside of SECTIONS | |
1081 | // clauses and treats them as occurring inside, so we don't | |
1082 | // check in_sections_clause here. | |
1083 | this->script_sections_.add_dot_assignment(value); | |
494e05f4 ILT |
1084 | } |
1085 | } | |
1086 | ||
1087 | // Add an assertion. | |
1088 | ||
1089 | void | |
1090 | Script_options::add_assertion(Expression* check, const char* message, | |
1091 | size_t messagelen) | |
1092 | { | |
1093 | if (this->script_sections_.in_sections_clause()) | |
1094 | this->script_sections_.add_assertion(check, message, messagelen); | |
1095 | else | |
1096 | { | |
1097 | Script_assertion* p = new Script_assertion(check, message, messagelen); | |
1098 | this->assertions_.push_back(p); | |
1099 | } | |
1100 | } | |
1101 | ||
919ed24c ILT |
1102 | // Create sections required by any linker scripts. |
1103 | ||
1104 | void | |
1105 | Script_options::create_script_sections(Layout* layout) | |
1106 | { | |
1107 | if (this->saw_sections_clause()) | |
1108 | this->script_sections_.create_sections(layout); | |
1109 | } | |
1110 | ||
494e05f4 ILT |
1111 | // Add any symbols we are defining to the symbol table. |
1112 | ||
1113 | void | |
9b07f471 | 1114 | Script_options::add_symbols_to_table(Symbol_table* symtab) |
e5756efb ILT |
1115 | { |
1116 | for (Symbol_assignments::iterator p = this->symbol_assignments_.begin(); | |
1117 | p != this->symbol_assignments_.end(); | |
1118 | ++p) | |
9b07f471 | 1119 | (*p)->add_to_table(symtab); |
a445fddf | 1120 | this->script_sections_.add_symbols_to_table(symtab); |
494e05f4 ILT |
1121 | } |
1122 | ||
a445fddf | 1123 | // Finalize symbol values. Also check assertions. |
494e05f4 ILT |
1124 | |
1125 | void | |
1126 | Script_options::finalize_symbols(Symbol_table* symtab, const Layout* layout) | |
1127 | { | |
7c07ecec ILT |
1128 | // We finalize the symbols defined in SECTIONS first, because they |
1129 | // are the ones which may have changed. This way if symbol outside | |
1130 | // SECTIONS are defined in terms of symbols inside SECTIONS, they | |
1131 | // will get the right value. | |
1132 | this->script_sections_.finalize_symbols(symtab, layout); | |
1133 | ||
494e05f4 ILT |
1134 | for (Symbol_assignments::iterator p = this->symbol_assignments_.begin(); |
1135 | p != this->symbol_assignments_.end(); | |
1136 | ++p) | |
1137 | (*p)->finalize(symtab, layout); | |
a445fddf ILT |
1138 | |
1139 | for (Assertions::iterator p = this->assertions_.begin(); | |
1140 | p != this->assertions_.end(); | |
1141 | ++p) | |
1142 | (*p)->check(symtab, layout); | |
a445fddf ILT |
1143 | } |
1144 | ||
1145 | // Set section addresses. We set all the symbols which have absolute | |
1146 | // values. Then we let the SECTIONS clause do its thing. This | |
1147 | // returns the segment which holds the file header and segment | |
1148 | // headers, if any. | |
1149 | ||
1150 | Output_segment* | |
1151 | Script_options::set_section_addresses(Symbol_table* symtab, Layout* layout) | |
1152 | { | |
1153 | for (Symbol_assignments::iterator p = this->symbol_assignments_.begin(); | |
1154 | p != this->symbol_assignments_.end(); | |
1155 | ++p) | |
77e65537 | 1156 | (*p)->set_if_absolute(symtab, layout, false, 0); |
a445fddf ILT |
1157 | |
1158 | return this->script_sections_.set_section_addresses(symtab, layout); | |
e5756efb ILT |
1159 | } |
1160 | ||
dbe717ef ILT |
1161 | // This class holds data passed through the parser to the lexer and to |
1162 | // the parser support functions. This avoids global variables. We | |
17a1d0a9 ILT |
1163 | // can't use global variables because we need not be called by a |
1164 | // singleton thread. | |
dbe717ef ILT |
1165 | |
1166 | class Parser_closure | |
1167 | { | |
1168 | public: | |
2ea97941 | 1169 | Parser_closure(const char* filename, |
dbe717ef | 1170 | const Position_dependent_options& posdep_options, |
99fff23b | 1171 | bool parsing_defsym, bool in_group, bool is_in_sysroot, |
2ea97941 ILT |
1172 | Command_line* command_line, |
1173 | Script_options* script_options, | |
15f8229b | 1174 | Lex* lex, |
2ea97941 ILT |
1175 | bool skip_on_incompatible_target) |
1176 | : filename_(filename), posdep_options_(posdep_options), | |
99fff23b ILT |
1177 | parsing_defsym_(parsing_defsym), in_group_(in_group), |
1178 | is_in_sysroot_(is_in_sysroot), | |
2ea97941 | 1179 | skip_on_incompatible_target_(skip_on_incompatible_target), |
15f8229b | 1180 | found_incompatible_target_(false), |
2ea97941 ILT |
1181 | command_line_(command_line), script_options_(script_options), |
1182 | version_script_info_(script_options->version_script_info()), | |
e5756efb | 1183 | lex_(lex), lineno_(0), charpos_(0), lex_mode_stack_(), inputs_(NULL) |
c82fbeee | 1184 | { |
09124467 | 1185 | // We start out processing C symbols in the default lex mode. |
6affe781 ILT |
1186 | this->language_stack_.push_back(Version_script_info::LANGUAGE_C); |
1187 | this->lex_mode_stack_.push_back(lex->mode()); | |
09124467 | 1188 | } |
dbe717ef ILT |
1189 | |
1190 | // Return the file name. | |
1191 | const char* | |
1192 | filename() const | |
1193 | { return this->filename_; } | |
1194 | ||
1195 | // Return the position dependent options. The caller may modify | |
1196 | // this. | |
1197 | Position_dependent_options& | |
1198 | position_dependent_options() | |
1199 | { return this->posdep_options_; } | |
1200 | ||
99fff23b ILT |
1201 | // Whether we are parsing a --defsym. |
1202 | bool | |
1203 | parsing_defsym() const | |
1204 | { return this->parsing_defsym_; } | |
1205 | ||
dbe717ef ILT |
1206 | // Return whether this script is being run in a group. |
1207 | bool | |
1208 | in_group() const | |
1209 | { return this->in_group_; } | |
1210 | ||
ad2d6943 ILT |
1211 | // Return whether this script was found using a directory in the |
1212 | // sysroot. | |
1213 | bool | |
1214 | is_in_sysroot() const | |
1215 | { return this->is_in_sysroot_; } | |
1216 | ||
15f8229b ILT |
1217 | // Whether to skip to the next file with the same name if we find an |
1218 | // incompatible target in an OUTPUT_FORMAT statement. | |
1219 | bool | |
1220 | skip_on_incompatible_target() const | |
1221 | { return this->skip_on_incompatible_target_; } | |
1222 | ||
e6a307ba | 1223 | // Stop skipping to the next file on an incompatible target. This |
15f8229b ILT |
1224 | // is called when we make some unrevocable change to the data |
1225 | // structures. | |
1226 | void | |
1227 | clear_skip_on_incompatible_target() | |
1228 | { this->skip_on_incompatible_target_ = false; } | |
1229 | ||
1230 | // Whether we found an incompatible target in an OUTPUT_FORMAT | |
1231 | // statement. | |
1232 | bool | |
1233 | found_incompatible_target() const | |
1234 | { return this->found_incompatible_target_; } | |
1235 | ||
1236 | // Note that we found an incompatible target. | |
1237 | void | |
1238 | set_found_incompatible_target() | |
1239 | { this->found_incompatible_target_ = true; } | |
1240 | ||
a0451b38 ILT |
1241 | // Returns the Command_line structure passed in at constructor time. |
1242 | // This value may be NULL. The caller may modify this, which modifies | |
1243 | // the passed-in Command_line object (not a copy). | |
e5756efb ILT |
1244 | Command_line* |
1245 | command_line() | |
a0451b38 ILT |
1246 | { return this->command_line_; } |
1247 | ||
e5756efb ILT |
1248 | // Return the options which may be set by a script. |
1249 | Script_options* | |
1250 | script_options() | |
1251 | { return this->script_options_; } | |
dbe717ef | 1252 | |
09124467 ILT |
1253 | // Return the object in which version script information should be stored. |
1254 | Version_script_info* | |
1255 | version_script() | |
1256 | { return this->version_script_info_; } | |
1257 | ||
2dd3e587 | 1258 | // Return the next token, and advance. |
dbe717ef ILT |
1259 | const Token* |
1260 | next_token() | |
1261 | { | |
e5756efb ILT |
1262 | const Token* token = this->lex_->next_token(); |
1263 | this->lineno_ = token->lineno(); | |
1264 | this->charpos_ = token->charpos(); | |
1265 | return token; | |
dbe717ef ILT |
1266 | } |
1267 | ||
e5756efb ILT |
1268 | // Set a new lexer mode, pushing the current one. |
1269 | void | |
1270 | push_lex_mode(Lex::Mode mode) | |
1271 | { | |
1272 | this->lex_mode_stack_.push_back(this->lex_->mode()); | |
1273 | this->lex_->set_mode(mode); | |
1274 | } | |
1275 | ||
1276 | // Pop the lexer mode. | |
1277 | void | |
1278 | pop_lex_mode() | |
2dd3e587 | 1279 | { |
e5756efb ILT |
1280 | gold_assert(!this->lex_mode_stack_.empty()); |
1281 | this->lex_->set_mode(this->lex_mode_stack_.back()); | |
1282 | this->lex_mode_stack_.pop_back(); | |
2dd3e587 ILT |
1283 | } |
1284 | ||
09124467 ILT |
1285 | // Return the current lexer mode. |
1286 | Lex::Mode | |
1287 | lex_mode() const | |
1288 | { return this->lex_mode_stack_.back(); } | |
1289 | ||
e5756efb ILT |
1290 | // Return the line number of the last token. |
1291 | int | |
1292 | lineno() const | |
1293 | { return this->lineno_; } | |
1294 | ||
1295 | // Return the character position in the line of the last token. | |
1296 | int | |
1297 | charpos() const | |
1298 | { return this->charpos_; } | |
1299 | ||
dbe717ef ILT |
1300 | // Return the list of input files, creating it if necessary. This |
1301 | // is a space leak--we never free the INPUTS_ pointer. | |
1302 | Input_arguments* | |
1303 | inputs() | |
1304 | { | |
1305 | if (this->inputs_ == NULL) | |
1306 | this->inputs_ = new Input_arguments(); | |
1307 | return this->inputs_; | |
1308 | } | |
1309 | ||
1310 | // Return whether we saw any input files. | |
1311 | bool | |
1312 | saw_inputs() const | |
1313 | { return this->inputs_ != NULL && !this->inputs_->empty(); } | |
1314 | ||
09124467 ILT |
1315 | // Return the current language being processed in a version script |
1316 | // (eg, "C++"). The empty string represents unmangled C names. | |
6affe781 | 1317 | Version_script_info::Language |
09124467 ILT |
1318 | get_current_language() const |
1319 | { return this->language_stack_.back(); } | |
1320 | ||
1321 | // Push a language onto the stack when entering an extern block. | |
6affe781 ILT |
1322 | void |
1323 | push_language(Version_script_info::Language lang) | |
09124467 ILT |
1324 | { this->language_stack_.push_back(lang); } |
1325 | ||
1326 | // Pop a language off of the stack when exiting an extern block. | |
6affe781 ILT |
1327 | void |
1328 | pop_language() | |
09124467 ILT |
1329 | { |
1330 | gold_assert(!this->language_stack_.empty()); | |
1331 | this->language_stack_.pop_back(); | |
1332 | } | |
1333 | ||
dbe717ef ILT |
1334 | private: |
1335 | // The name of the file we are reading. | |
1336 | const char* filename_; | |
1337 | // The position dependent options. | |
1338 | Position_dependent_options posdep_options_; | |
99fff23b ILT |
1339 | // True if we are parsing a --defsym. |
1340 | bool parsing_defsym_; | |
dbe717ef ILT |
1341 | // Whether we are currently in a --start-group/--end-group. |
1342 | bool in_group_; | |
ad2d6943 ILT |
1343 | // Whether the script was found in a sysrooted directory. |
1344 | bool is_in_sysroot_; | |
15f8229b ILT |
1345 | // If this is true, then if we find an OUTPUT_FORMAT with an |
1346 | // incompatible target, then we tell the parser to abort so that we | |
1347 | // can search for the next file with the same name. | |
1348 | bool skip_on_incompatible_target_; | |
1349 | // True if we found an OUTPUT_FORMAT with an incompatible target. | |
1350 | bool found_incompatible_target_; | |
a0451b38 ILT |
1351 | // May be NULL if the user chooses not to pass one in. |
1352 | Command_line* command_line_; | |
e5756efb ILT |
1353 | // Options which may be set from any linker script. |
1354 | Script_options* script_options_; | |
09124467 ILT |
1355 | // Information parsed from a version script. |
1356 | Version_script_info* version_script_info_; | |
e5756efb ILT |
1357 | // The lexer. |
1358 | Lex* lex_; | |
1359 | // The line number of the last token returned by next_token. | |
1360 | int lineno_; | |
1361 | // The column number of the last token returned by next_token. | |
1362 | int charpos_; | |
1363 | // A stack of lexer modes. | |
1364 | std::vector<Lex::Mode> lex_mode_stack_; | |
09124467 ILT |
1365 | // A stack of which extern/language block we're inside. Can be C++, |
1366 | // java, or empty for C. | |
6affe781 | 1367 | std::vector<Version_script_info::Language> language_stack_; |
dbe717ef ILT |
1368 | // New input files found to add to the link. |
1369 | Input_arguments* inputs_; | |
1370 | }; | |
1371 | ||
1372 | // FILE was found as an argument on the command line. Try to read it | |
da769d56 | 1373 | // as a script. Return true if the file was handled. |
dbe717ef ILT |
1374 | |
1375 | bool | |
f1ed28fb | 1376 | read_input_script(Workqueue* workqueue, Symbol_table* symtab, Layout* layout, |
15f8229b ILT |
1377 | Dirsearch* dirsearch, int dirindex, |
1378 | Input_objects* input_objects, Mapfile* mapfile, | |
1379 | Input_group* input_group, | |
dbe717ef | 1380 | const Input_argument* input_argument, |
da769d56 ILT |
1381 | Input_file* input_file, Task_token* next_blocker, |
1382 | bool* used_next_blocker) | |
dbe717ef | 1383 | { |
da769d56 ILT |
1384 | *used_next_blocker = false; |
1385 | ||
e5756efb ILT |
1386 | std::string input_string; |
1387 | Lex::read_file(input_file, &input_string); | |
1388 | ||
1389 | Lex lex(input_string.c_str(), input_string.length(), PARSING_LINKER_SCRIPT); | |
dbe717ef ILT |
1390 | |
1391 | Parser_closure closure(input_file->filename().c_str(), | |
1392 | input_argument->file().options(), | |
99fff23b | 1393 | false, |
dbe717ef | 1394 | input_group != NULL, |
ad2d6943 | 1395 | input_file->is_in_sysroot(), |
a0451b38 | 1396 | NULL, |
e5756efb | 1397 | layout->script_options(), |
15f8229b ILT |
1398 | &lex, |
1399 | input_file->will_search_for()); | |
dbe717ef | 1400 | |
d7bb5745 ILT |
1401 | bool old_saw_sections_clause = |
1402 | layout->script_options()->saw_sections_clause(); | |
1403 | ||
dbe717ef | 1404 | if (yyparse(&closure) != 0) |
15f8229b ILT |
1405 | { |
1406 | if (closure.found_incompatible_target()) | |
1407 | { | |
1408 | Read_symbols::incompatible_warning(input_argument, input_file); | |
1409 | Read_symbols::requeue(workqueue, input_objects, symtab, layout, | |
1410 | dirsearch, dirindex, mapfile, input_argument, | |
1411 | input_group, next_blocker); | |
1412 | return true; | |
1413 | } | |
1414 | return false; | |
1415 | } | |
dbe717ef | 1416 | |
d7bb5745 ILT |
1417 | if (!old_saw_sections_clause |
1418 | && layout->script_options()->saw_sections_clause() | |
1419 | && layout->have_added_input_section()) | |
1420 | gold_error(_("%s: SECTIONS seen after other input files; try -T/--script"), | |
1421 | input_file->filename().c_str()); | |
1422 | ||
dbe717ef | 1423 | if (!closure.saw_inputs()) |
da769d56 | 1424 | return true; |
dbe717ef | 1425 | |
da769d56 | 1426 | Task_token* this_blocker = NULL; |
dbe717ef ILT |
1427 | for (Input_arguments::const_iterator p = closure.inputs()->begin(); |
1428 | p != closure.inputs()->end(); | |
1429 | ++p) | |
1430 | { | |
1431 | Task_token* nb; | |
1432 | if (p + 1 == closure.inputs()->end()) | |
1433 | nb = next_blocker; | |
1434 | else | |
1435 | { | |
17a1d0a9 | 1436 | nb = new Task_token(true); |
dbe717ef ILT |
1437 | nb->add_blocker(); |
1438 | } | |
f1ed28fb | 1439 | workqueue->queue_soon(new Read_symbols(input_objects, symtab, |
15f8229b | 1440 | layout, dirsearch, 0, mapfile, &*p, |
da769d56 | 1441 | input_group, this_blocker, nb)); |
dbe717ef ILT |
1442 | this_blocker = nb; |
1443 | } | |
1444 | ||
072fe7ce ILT |
1445 | if (layout->incremental_inputs()) |
1446 | { | |
1447 | // Like new Read_symbols(...) above, we rely on close.inputs() | |
1448 | // getting leaked by closure. | |
1449 | Script_info* info = new Script_info(closure.inputs()); | |
98fa85cb ILT |
1450 | layout->incremental_inputs()->report_script( |
1451 | input_argument, | |
1452 | input_file->file().get_mtime(), | |
1453 | info); | |
072fe7ce | 1454 | } |
da769d56 ILT |
1455 | *used_next_blocker = true; |
1456 | ||
dbe717ef ILT |
1457 | return true; |
1458 | } | |
1459 | ||
09124467 ILT |
1460 | // Helper function for read_version_script() and |
1461 | // read_commandline_script(). Processes the given file in the mode | |
1462 | // indicated by first_token and lex_mode. | |
3c2fafa5 | 1463 | |
09124467 ILT |
1464 | static bool |
1465 | read_script_file(const char* filename, Command_line* cmdline, | |
c82fbeee | 1466 | Script_options* script_options, |
09124467 | 1467 | int first_token, Lex::Mode lex_mode) |
3c2fafa5 | 1468 | { |
a0451b38 ILT |
1469 | // TODO: if filename is a relative filename, search for it manually |
1470 | // using "." + cmdline->options()->search_path() -- not dirsearch. | |
3c2fafa5 ILT |
1471 | Dirsearch dirsearch; |
1472 | ||
17a1d0a9 ILT |
1473 | // The file locking code wants to record a Task, but we haven't |
1474 | // started the workqueue yet. This is only for debugging purposes, | |
1475 | // so we invent a fake value. | |
1476 | const Task* task = reinterpret_cast<const Task*>(-1); | |
1477 | ||
b0d8593d ILT |
1478 | // We don't want this file to be opened in binary mode. |
1479 | Position_dependent_options posdep = cmdline->position_dependent_options(); | |
7cc619c3 ILT |
1480 | if (posdep.format_enum() == General_options::OBJECT_FORMAT_BINARY) |
1481 | posdep.set_format_enum(General_options::OBJECT_FORMAT_ELF); | |
ae3b5189 CD |
1482 | Input_file_argument input_argument(filename, |
1483 | Input_file_argument::INPUT_FILE_TYPE_FILE, | |
1484 | "", false, posdep); | |
3c2fafa5 | 1485 | Input_file input_file(&input_argument); |
15f8229b ILT |
1486 | int dummy = 0; |
1487 | if (!input_file.open(dirsearch, task, &dummy)) | |
3c2fafa5 ILT |
1488 | return false; |
1489 | ||
e5756efb ILT |
1490 | std::string input_string; |
1491 | Lex::read_file(&input_file, &input_string); | |
1492 | ||
09124467 ILT |
1493 | Lex lex(input_string.c_str(), input_string.length(), first_token); |
1494 | lex.set_mode(lex_mode); | |
3c2fafa5 ILT |
1495 | |
1496 | Parser_closure closure(filename, | |
1497 | cmdline->position_dependent_options(), | |
99fff23b | 1498 | first_token == Lex::DYNAMIC_LIST, |
3c2fafa5 ILT |
1499 | false, |
1500 | input_file.is_in_sysroot(), | |
a0451b38 | 1501 | cmdline, |
c82fbeee | 1502 | script_options, |
15f8229b ILT |
1503 | &lex, |
1504 | false); | |
3c2fafa5 ILT |
1505 | if (yyparse(&closure) != 0) |
1506 | { | |
17a1d0a9 | 1507 | input_file.file().unlock(task); |
3c2fafa5 ILT |
1508 | return false; |
1509 | } | |
1510 | ||
17a1d0a9 | 1511 | input_file.file().unlock(task); |
d391083d ILT |
1512 | |
1513 | gold_assert(!closure.saw_inputs()); | |
1514 | ||
3c2fafa5 ILT |
1515 | return true; |
1516 | } | |
1517 | ||
09124467 ILT |
1518 | // FILENAME was found as an argument to --script (-T). |
1519 | // Read it as a script, and execute its contents immediately. | |
1520 | ||
1521 | bool | |
1522 | read_commandline_script(const char* filename, Command_line* cmdline) | |
1523 | { | |
c82fbeee | 1524 | return read_script_file(filename, cmdline, &cmdline->script_options(), |
09124467 ILT |
1525 | PARSING_LINKER_SCRIPT, Lex::LINKER_SCRIPT); |
1526 | } | |
1527 | ||
c82fbeee CS |
1528 | // FILENAME was found as an argument to --version-script. Read it as |
1529 | // a version script, and store its contents in | |
09124467 ILT |
1530 | // cmdline->script_options()->version_script_info(). |
1531 | ||
1532 | bool | |
1533 | read_version_script(const char* filename, Command_line* cmdline) | |
1534 | { | |
c82fbeee | 1535 | return read_script_file(filename, cmdline, &cmdline->script_options(), |
09124467 ILT |
1536 | PARSING_VERSION_SCRIPT, Lex::VERSION_SCRIPT); |
1537 | } | |
1538 | ||
c82fbeee CS |
1539 | // FILENAME was found as an argument to --dynamic-list. Read it as a |
1540 | // list of symbols, and store its contents in DYNAMIC_LIST. | |
1541 | ||
1542 | bool | |
1543 | read_dynamic_list(const char* filename, Command_line* cmdline, | |
1544 | Script_options* dynamic_list) | |
1545 | { | |
1546 | return read_script_file(filename, cmdline, dynamic_list, | |
1547 | PARSING_DYNAMIC_LIST, Lex::DYNAMIC_LIST); | |
1548 | } | |
1549 | ||
e5756efb ILT |
1550 | // Implement the --defsym option on the command line. Return true if |
1551 | // all is well. | |
1552 | ||
1553 | bool | |
1554 | Script_options::define_symbol(const char* definition) | |
1555 | { | |
1556 | Lex lex(definition, strlen(definition), PARSING_DEFSYM); | |
1557 | lex.set_mode(Lex::EXPRESSION); | |
1558 | ||
1559 | // Dummy value. | |
1560 | Position_dependent_options posdep_options; | |
1561 | ||
99fff23b ILT |
1562 | Parser_closure closure("command line", posdep_options, true, |
1563 | false, false, NULL, this, &lex, false); | |
e5756efb ILT |
1564 | |
1565 | if (yyparse(&closure) != 0) | |
1566 | return false; | |
1567 | ||
1568 | gold_assert(!closure.saw_inputs()); | |
1569 | ||
1570 | return true; | |
1571 | } | |
1572 | ||
494e05f4 ILT |
1573 | // Print the script to F for debugging. |
1574 | ||
1575 | void | |
1576 | Script_options::print(FILE* f) const | |
1577 | { | |
1578 | fprintf(f, "%s: Dumping linker script\n", program_name); | |
1579 | ||
1580 | if (!this->entry_.empty()) | |
1581 | fprintf(f, "ENTRY(%s)\n", this->entry_.c_str()); | |
1582 | ||
1583 | for (Symbol_assignments::const_iterator p = | |
1584 | this->symbol_assignments_.begin(); | |
1585 | p != this->symbol_assignments_.end(); | |
1586 | ++p) | |
1587 | (*p)->print(f); | |
1588 | ||
1589 | for (Assertions::const_iterator p = this->assertions_.begin(); | |
1590 | p != this->assertions_.end(); | |
1591 | ++p) | |
1592 | (*p)->print(f); | |
1593 | ||
1594 | this->script_sections_.print(f); | |
1595 | ||
1596 | this->version_script_info_.print(f); | |
1597 | } | |
1598 | ||
dbe717ef | 1599 | // Manage mapping from keywords to the codes expected by the bison |
09124467 ILT |
1600 | // parser. We construct one global object for each lex mode with |
1601 | // keywords. | |
dbe717ef ILT |
1602 | |
1603 | class Keyword_to_parsecode | |
1604 | { | |
1605 | public: | |
1606 | // The structure which maps keywords to parsecodes. | |
1607 | struct Keyword_parsecode | |
1608 | { | |
1609 | // Keyword. | |
1610 | const char* keyword; | |
1611 | // Corresponding parsecode. | |
1612 | int parsecode; | |
1613 | }; | |
1614 | ||
09124467 ILT |
1615 | Keyword_to_parsecode(const Keyword_parsecode* keywords, |
1616 | int keyword_count) | |
1617 | : keyword_parsecodes_(keywords), keyword_count_(keyword_count) | |
1618 | { } | |
1619 | ||
dbe717ef ILT |
1620 | // Return the parsecode corresponding KEYWORD, or 0 if it is not a |
1621 | // keyword. | |
09124467 ILT |
1622 | int |
1623 | keyword_to_parsecode(const char* keyword, size_t len) const; | |
dbe717ef ILT |
1624 | |
1625 | private: | |
09124467 ILT |
1626 | const Keyword_parsecode* keyword_parsecodes_; |
1627 | const int keyword_count_; | |
dbe717ef ILT |
1628 | }; |
1629 | ||
1630 | // Mapping from keyword string to keyword parsecode. This array must | |
1631 | // be kept in sorted order. Parsecodes are looked up using bsearch. | |
1632 | // This array must correspond to the list of parsecodes in yyscript.y. | |
1633 | ||
09124467 ILT |
1634 | static const Keyword_to_parsecode::Keyword_parsecode |
1635 | script_keyword_parsecodes[] = | |
dbe717ef ILT |
1636 | { |
1637 | { "ABSOLUTE", ABSOLUTE }, | |
1638 | { "ADDR", ADDR }, | |
1639 | { "ALIGN", ALIGN_K }, | |
e5756efb | 1640 | { "ALIGNOF", ALIGNOF }, |
dbe717ef ILT |
1641 | { "ASSERT", ASSERT_K }, |
1642 | { "AS_NEEDED", AS_NEEDED }, | |
1643 | { "AT", AT }, | |
1644 | { "BIND", BIND }, | |
1645 | { "BLOCK", BLOCK }, | |
1646 | { "BYTE", BYTE }, | |
1647 | { "CONSTANT", CONSTANT }, | |
1648 | { "CONSTRUCTORS", CONSTRUCTORS }, | |
dbe717ef ILT |
1649 | { "CREATE_OBJECT_SYMBOLS", CREATE_OBJECT_SYMBOLS }, |
1650 | { "DATA_SEGMENT_ALIGN", DATA_SEGMENT_ALIGN }, | |
1651 | { "DATA_SEGMENT_END", DATA_SEGMENT_END }, | |
1652 | { "DATA_SEGMENT_RELRO_END", DATA_SEGMENT_RELRO_END }, | |
1653 | { "DEFINED", DEFINED }, | |
dbe717ef ILT |
1654 | { "ENTRY", ENTRY }, |
1655 | { "EXCLUDE_FILE", EXCLUDE_FILE }, | |
1656 | { "EXTERN", EXTERN }, | |
1657 | { "FILL", FILL }, | |
1658 | { "FLOAT", FLOAT }, | |
1659 | { "FORCE_COMMON_ALLOCATION", FORCE_COMMON_ALLOCATION }, | |
1660 | { "GROUP", GROUP }, | |
1661 | { "HLL", HLL }, | |
1662 | { "INCLUDE", INCLUDE }, | |
dbe717ef ILT |
1663 | { "INHIBIT_COMMON_ALLOCATION", INHIBIT_COMMON_ALLOCATION }, |
1664 | { "INPUT", INPUT }, | |
1665 | { "KEEP", KEEP }, | |
1666 | { "LENGTH", LENGTH }, | |
1667 | { "LOADADDR", LOADADDR }, | |
1668 | { "LONG", LONG }, | |
1669 | { "MAP", MAP }, | |
1670 | { "MAX", MAX_K }, | |
1671 | { "MEMORY", MEMORY }, | |
1672 | { "MIN", MIN_K }, | |
1673 | { "NEXT", NEXT }, | |
1674 | { "NOCROSSREFS", NOCROSSREFS }, | |
1675 | { "NOFLOAT", NOFLOAT }, | |
dbe717ef ILT |
1676 | { "ONLY_IF_RO", ONLY_IF_RO }, |
1677 | { "ONLY_IF_RW", ONLY_IF_RW }, | |
195e7dc6 | 1678 | { "OPTION", OPTION }, |
dbe717ef ILT |
1679 | { "ORIGIN", ORIGIN }, |
1680 | { "OUTPUT", OUTPUT }, | |
1681 | { "OUTPUT_ARCH", OUTPUT_ARCH }, | |
1682 | { "OUTPUT_FORMAT", OUTPUT_FORMAT }, | |
1683 | { "OVERLAY", OVERLAY }, | |
1684 | { "PHDRS", PHDRS }, | |
1685 | { "PROVIDE", PROVIDE }, | |
1686 | { "PROVIDE_HIDDEN", PROVIDE_HIDDEN }, | |
1687 | { "QUAD", QUAD }, | |
1688 | { "SEARCH_DIR", SEARCH_DIR }, | |
1689 | { "SECTIONS", SECTIONS }, | |
1690 | { "SEGMENT_START", SEGMENT_START }, | |
1691 | { "SHORT", SHORT }, | |
1692 | { "SIZEOF", SIZEOF }, | |
1693 | { "SIZEOF_HEADERS", SIZEOF_HEADERS }, | |
3802b2dd | 1694 | { "SORT", SORT_BY_NAME }, |
dbe717ef ILT |
1695 | { "SORT_BY_ALIGNMENT", SORT_BY_ALIGNMENT }, |
1696 | { "SORT_BY_NAME", SORT_BY_NAME }, | |
1697 | { "SPECIAL", SPECIAL }, | |
1698 | { "SQUAD", SQUAD }, | |
1699 | { "STARTUP", STARTUP }, | |
1700 | { "SUBALIGN", SUBALIGN }, | |
1701 | { "SYSLIB", SYSLIB }, | |
1702 | { "TARGET", TARGET_K }, | |
1703 | { "TRUNCATE", TRUNCATE }, | |
1704 | { "VERSION", VERSIONK }, | |
1705 | { "global", GLOBAL }, | |
1706 | { "l", LENGTH }, | |
1707 | { "len", LENGTH }, | |
1708 | { "local", LOCAL }, | |
1709 | { "o", ORIGIN }, | |
1710 | { "org", ORIGIN }, | |
1711 | { "sizeof_headers", SIZEOF_HEADERS }, | |
1712 | }; | |
1713 | ||
09124467 ILT |
1714 | static const Keyword_to_parsecode |
1715 | script_keywords(&script_keyword_parsecodes[0], | |
1716 | (sizeof(script_keyword_parsecodes) | |
1717 | / sizeof(script_keyword_parsecodes[0]))); | |
1718 | ||
1719 | static const Keyword_to_parsecode::Keyword_parsecode | |
1720 | version_script_keyword_parsecodes[] = | |
1721 | { | |
1722 | { "extern", EXTERN }, | |
1723 | { "global", GLOBAL }, | |
1724 | { "local", LOCAL }, | |
1725 | }; | |
1726 | ||
1727 | static const Keyword_to_parsecode | |
1728 | version_script_keywords(&version_script_keyword_parsecodes[0], | |
1729 | (sizeof(version_script_keyword_parsecodes) | |
1730 | / sizeof(version_script_keyword_parsecodes[0]))); | |
dbe717ef | 1731 | |
c82fbeee CS |
1732 | static const Keyword_to_parsecode::Keyword_parsecode |
1733 | dynamic_list_keyword_parsecodes[] = | |
1734 | { | |
1735 | { "extern", EXTERN }, | |
1736 | }; | |
1737 | ||
1738 | static const Keyword_to_parsecode | |
1739 | dynamic_list_keywords(&dynamic_list_keyword_parsecodes[0], | |
1740 | (sizeof(dynamic_list_keyword_parsecodes) | |
1741 | / sizeof(dynamic_list_keyword_parsecodes[0]))); | |
1742 | ||
1743 | ||
1744 | ||
dbe717ef ILT |
1745 | // Comparison function passed to bsearch. |
1746 | ||
1747 | extern "C" | |
1748 | { | |
1749 | ||
e5756efb ILT |
1750 | struct Ktt_key |
1751 | { | |
1752 | const char* str; | |
1753 | size_t len; | |
1754 | }; | |
1755 | ||
dbe717ef ILT |
1756 | static int |
1757 | ktt_compare(const void* keyv, const void* kttv) | |
1758 | { | |
e5756efb | 1759 | const Ktt_key* key = static_cast<const Ktt_key*>(keyv); |
dbe717ef ILT |
1760 | const Keyword_to_parsecode::Keyword_parsecode* ktt = |
1761 | static_cast<const Keyword_to_parsecode::Keyword_parsecode*>(kttv); | |
e5756efb ILT |
1762 | int i = strncmp(key->str, ktt->keyword, key->len); |
1763 | if (i != 0) | |
1764 | return i; | |
1765 | if (ktt->keyword[key->len] != '\0') | |
1766 | return -1; | |
1767 | return 0; | |
dbe717ef ILT |
1768 | } |
1769 | ||
1770 | } // End extern "C". | |
1771 | ||
1772 | int | |
09124467 ILT |
1773 | Keyword_to_parsecode::keyword_to_parsecode(const char* keyword, |
1774 | size_t len) const | |
dbe717ef | 1775 | { |
e5756efb ILT |
1776 | Ktt_key key; |
1777 | key.str = keyword; | |
1778 | key.len = len; | |
1779 | void* kttv = bsearch(&key, | |
09124467 ILT |
1780 | this->keyword_parsecodes_, |
1781 | this->keyword_count_, | |
1782 | sizeof(this->keyword_parsecodes_[0]), | |
1783 | ktt_compare); | |
dbe717ef ILT |
1784 | if (kttv == NULL) |
1785 | return 0; | |
1786 | Keyword_parsecode* ktt = static_cast<Keyword_parsecode*>(kttv); | |
1787 | return ktt->parsecode; | |
1788 | } | |
1789 | ||
494e05f4 ILT |
1790 | // The following structs are used within the VersionInfo class as well |
1791 | // as in the bison helper functions. They store the information | |
1792 | // parsed from the version script. | |
dbe717ef | 1793 | |
494e05f4 ILT |
1794 | // A single version expression. |
1795 | // For example, pattern="std::map*" and language="C++". | |
62dfdd4d | 1796 | struct Version_expression |
6affe781 ILT |
1797 | { |
1798 | Version_expression(const std::string& a_pattern, | |
1799 | Version_script_info::Language a_language, | |
1800 | bool a_exact_match) | |
62dfdd4d ILT |
1801 | : pattern(a_pattern), language(a_language), exact_match(a_exact_match), |
1802 | was_matched_by_symbol(false) | |
6affe781 | 1803 | { } |
dbe717ef | 1804 | |
494e05f4 | 1805 | std::string pattern; |
6affe781 | 1806 | Version_script_info::Language language; |
494e05f4 ILT |
1807 | // If false, we use glob() to match pattern. If true, we use strcmp(). |
1808 | bool exact_match; | |
62dfdd4d ILT |
1809 | // True if --no-undefined-version is in effect and we found this |
1810 | // version in get_symbol_version. We use mutable because this | |
1811 | // struct is generally not modifiable after it has been created. | |
1812 | mutable bool was_matched_by_symbol; | |
494e05f4 | 1813 | }; |
dbe717ef | 1814 | |
494e05f4 | 1815 | // A list of expressions. |
6affe781 ILT |
1816 | struct Version_expression_list |
1817 | { | |
494e05f4 ILT |
1818 | std::vector<struct Version_expression> expressions; |
1819 | }; | |
e5756efb | 1820 | |
494e05f4 ILT |
1821 | // A list of which versions upon which another version depends. |
1822 | // Strings should be from the Stringpool. | |
6affe781 ILT |
1823 | struct Version_dependency_list |
1824 | { | |
494e05f4 ILT |
1825 | std::vector<std::string> dependencies; |
1826 | }; | |
dbe717ef | 1827 | |
494e05f4 ILT |
1828 | // The total definition of a version. It includes the tag for the |
1829 | // version, its global and local expressions, and any dependencies. | |
6affe781 ILT |
1830 | struct Version_tree |
1831 | { | |
494e05f4 | 1832 | Version_tree() |
6affe781 ILT |
1833 | : tag(), global(NULL), local(NULL), dependencies(NULL) |
1834 | { } | |
e5756efb | 1835 | |
494e05f4 ILT |
1836 | std::string tag; |
1837 | const struct Version_expression_list* global; | |
1838 | const struct Version_expression_list* local; | |
1839 | const struct Version_dependency_list* dependencies; | |
1840 | }; | |
dbe717ef | 1841 | |
98e090bd ILT |
1842 | // Helper class that calls cplus_demangle when needed and takes care of freeing |
1843 | // the result. | |
1844 | ||
1845 | class Lazy_demangler | |
1846 | { | |
1847 | public: | |
1848 | Lazy_demangler(const char* symbol, int options) | |
1849 | : symbol_(symbol), options_(options), demangled_(NULL), did_demangle_(false) | |
1850 | { } | |
1851 | ||
1852 | ~Lazy_demangler() | |
1853 | { free(this->demangled_); } | |
1854 | ||
1855 | // Return the demangled name. The actual demangling happens on the first call, | |
1856 | // and the result is later cached. | |
1857 | inline char* | |
1858 | get(); | |
1859 | ||
1860 | private: | |
1861 | // The symbol to demangle. | |
1862 | const char *symbol_; | |
1863 | // Option flags to pass to cplus_demagle. | |
1864 | const int options_; | |
1865 | // The cached demangled value, or NULL if demangling didn't happen yet or | |
1866 | // failed. | |
1867 | char *demangled_; | |
1868 | // Whether we already called cplus_demangle | |
1869 | bool did_demangle_; | |
1870 | }; | |
1871 | ||
1872 | // Return the demangled name. The actual demangling happens on the first call, | |
1873 | // and the result is later cached. Returns NULL if the symbol cannot be | |
1874 | // demangled. | |
1875 | ||
1876 | inline char* | |
1877 | Lazy_demangler::get() | |
1878 | { | |
1879 | if (!this->did_demangle_) | |
1880 | { | |
1881 | this->demangled_ = cplus_demangle(this->symbol_, this->options_); | |
1882 | this->did_demangle_ = true; | |
1883 | } | |
1884 | return this->demangled_; | |
1885 | } | |
1886 | ||
6affe781 ILT |
1887 | // Class Version_script_info. |
1888 | ||
1889 | Version_script_info::Version_script_info() | |
98e090bd ILT |
1890 | : dependency_lists_(), expression_lists_(), version_trees_(), globs_(), |
1891 | default_version_(NULL), default_is_global_(false), is_finalized_(false) | |
6affe781 ILT |
1892 | { |
1893 | for (int i = 0; i < LANGUAGE_COUNT; ++i) | |
98e090bd | 1894 | this->exact_[i] = NULL; |
6affe781 ILT |
1895 | } |
1896 | ||
494e05f4 | 1897 | Version_script_info::~Version_script_info() |
1ef1f3d3 | 1898 | { |
1ef1f3d3 ILT |
1899 | } |
1900 | ||
6affe781 ILT |
1901 | // Forget all the known version script information. |
1902 | ||
1ef1f3d3 ILT |
1903 | void |
1904 | Version_script_info::clear() | |
494e05f4 | 1905 | { |
6affe781 ILT |
1906 | for (size_t k = 0; k < this->dependency_lists_.size(); ++k) |
1907 | delete this->dependency_lists_[k]; | |
1ef1f3d3 | 1908 | this->dependency_lists_.clear(); |
6affe781 ILT |
1909 | for (size_t k = 0; k < this->version_trees_.size(); ++k) |
1910 | delete this->version_trees_[k]; | |
1ef1f3d3 | 1911 | this->version_trees_.clear(); |
6affe781 ILT |
1912 | for (size_t k = 0; k < this->expression_lists_.size(); ++k) |
1913 | delete this->expression_lists_[k]; | |
1ef1f3d3 | 1914 | this->expression_lists_.clear(); |
dbe717ef ILT |
1915 | } |
1916 | ||
6affe781 ILT |
1917 | // Finalize the version script information. |
1918 | ||
1919 | void | |
1920 | Version_script_info::finalize() | |
1921 | { | |
1922 | if (!this->is_finalized_) | |
1923 | { | |
1924 | this->build_lookup_tables(); | |
1925 | this->is_finalized_ = true; | |
1926 | } | |
1927 | } | |
1928 | ||
1929 | // Return all the versions. | |
1930 | ||
494e05f4 ILT |
1931 | std::vector<std::string> |
1932 | Version_script_info::get_versions() const | |
dbe717ef | 1933 | { |
494e05f4 | 1934 | std::vector<std::string> ret; |
6affe781 | 1935 | for (size_t j = 0; j < this->version_trees_.size(); ++j) |
057ead22 ILT |
1936 | if (!this->version_trees_[j]->tag.empty()) |
1937 | ret.push_back(this->version_trees_[j]->tag); | |
494e05f4 | 1938 | return ret; |
dbe717ef ILT |
1939 | } |
1940 | ||
6affe781 ILT |
1941 | // Return the dependencies of VERSION. |
1942 | ||
494e05f4 ILT |
1943 | std::vector<std::string> |
1944 | Version_script_info::get_dependencies(const char* version) const | |
dbe717ef | 1945 | { |
494e05f4 | 1946 | std::vector<std::string> ret; |
6affe781 ILT |
1947 | for (size_t j = 0; j < this->version_trees_.size(); ++j) |
1948 | if (this->version_trees_[j]->tag == version) | |
494e05f4 ILT |
1949 | { |
1950 | const struct Version_dependency_list* deps = | |
6affe781 | 1951 | this->version_trees_[j]->dependencies; |
494e05f4 ILT |
1952 | if (deps != NULL) |
1953 | for (size_t k = 0; k < deps->dependencies.size(); ++k) | |
1954 | ret.push_back(deps->dependencies[k]); | |
1955 | return ret; | |
1956 | } | |
1957 | return ret; | |
1958 | } | |
1959 | ||
98e090bd ILT |
1960 | // A version script essentially maps a symbol name to a version tag |
1961 | // and an indication of whether symbol is global or local within that | |
1962 | // version tag. Each symbol maps to at most one version tag. | |
1963 | // Unfortunately, in practice, version scripts are ambiguous, and list | |
1964 | // symbols multiple times. Thus, we have to document the matching | |
1965 | // process. | |
1966 | ||
1967 | // This is a description of what the GNU linker does as of 2010-01-11. | |
1968 | // It walks through the version tags in the order in which they appear | |
1969 | // in the version script. For each tag, it first walks through the | |
1970 | // global patterns for that tag, then the local patterns. When | |
1971 | // looking at a single pattern, it first applies any language specific | |
1972 | // demangling as specified for the pattern, and then matches the | |
1973 | // resulting symbol name to the pattern. If it finds an exact match | |
1974 | // for a literal pattern (a pattern enclosed in quotes or with no | |
1975 | // wildcard characters), then that is the match that it uses. If | |
1976 | // finds a match with a wildcard pattern, then it saves it and | |
1977 | // continues searching. Wildcard patterns that are exactly "*" are | |
1978 | // saved separately. | |
1979 | ||
1980 | // If no exact match with a literal pattern is ever found, then if a | |
1981 | // wildcard match with a global pattern was found it is used, | |
1982 | // otherwise if a wildcard match with a local pattern was found it is | |
1983 | // used. | |
1984 | ||
1985 | // This is the result: | |
1986 | // * If there is an exact match, then we use the first tag in the | |
1987 | // version script where it matches. | |
1988 | // + If the exact match in that tag is global, it is used. | |
1989 | // + Otherwise the exact match in that tag is local, and is used. | |
1990 | // * Otherwise, if there is any match with a global wildcard pattern: | |
1991 | // + If there is any match with a wildcard pattern which is not | |
1992 | // "*", then we use the tag in which the *last* such pattern | |
1993 | // appears. | |
1994 | // + Otherwise, we matched "*". If there is no match with a local | |
1995 | // wildcard pattern which is not "*", then we use the *last* | |
1996 | // match with a global "*". Otherwise, continue. | |
1997 | // * Otherwise, if there is any match with a local wildcard pattern: | |
1998 | // + If there is any match with a wildcard pattern which is not | |
1999 | // "*", then we use the tag in which the *last* such pattern | |
2000 | // appears. | |
2001 | // + Otherwise, we matched "*", and we use the tag in which the | |
2002 | // *last* such match occurred. | |
2003 | ||
2004 | // There is an additional wrinkle. When the GNU linker finds a symbol | |
2005 | // with a version defined in an object file due to a .symver | |
2006 | // directive, it looks up that symbol name in that version tag. If it | |
2007 | // finds it, it matches the symbol name against the patterns for that | |
2008 | // version. If there is no match with a global pattern, but there is | |
2009 | // a match with a local pattern, then the GNU linker marks the symbol | |
2010 | // as local. | |
2011 | ||
2012 | // We want gold to be generally compatible, but we also want gold to | |
2013 | // be fast. These are the rules that gold implements: | |
2014 | // * If there is an exact match for the mangled name, we use it. | |
2015 | // + If there is more than one exact match, we give a warning, and | |
2016 | // we use the first tag in the script which matches. | |
2017 | // + If a symbol has an exact match as both global and local for | |
2018 | // the same version tag, we give an error. | |
2019 | // * Otherwise, we look for an extern C++ or an extern Java exact | |
2020 | // match. If we find an exact match, we use it. | |
2021 | // + If there is more than one exact match, we give a warning, and | |
2022 | // we use the first tag in the script which matches. | |
2023 | // + If a symbol has an exact match as both global and local for | |
2024 | // the same version tag, we give an error. | |
2025 | // * Otherwise, we look through the wildcard patterns, ignoring "*" | |
2026 | // patterns. We look through the version tags in reverse order. | |
2027 | // For each version tag, we look through the global patterns and | |
2028 | // then the local patterns. We use the first match we find (i.e., | |
2029 | // the last matching version tag in the file). | |
2030 | // * Otherwise, we use the "*" pattern if there is one. We give an | |
2031 | // error if there are multiple "*" patterns. | |
2032 | ||
2033 | // At least for now, gold does not look up the version tag for a | |
2034 | // symbol version found in an object file to see if it should be | |
2035 | // forced local. There are other ways to force a symbol to be local, | |
2036 | // and I don't understand why this one is useful. | |
2037 | ||
6affe781 ILT |
2038 | // Build a set of fast lookup tables for a version script. |
2039 | ||
2040 | void | |
2041 | Version_script_info::build_lookup_tables() | |
2042 | { | |
2043 | size_t size = this->version_trees_.size(); | |
2044 | for (size_t j = 0; j < size; ++j) | |
2045 | { | |
2046 | const Version_tree* v = this->version_trees_[j]; | |
98e090bd ILT |
2047 | this->build_expression_list_lookup(v->local, v, false); |
2048 | this->build_expression_list_lookup(v->global, v, true); | |
2049 | } | |
2050 | } | |
2051 | ||
2052 | // If a pattern has backlashes but no unquoted wildcard characters, | |
2053 | // then we apply backslash unquoting and look for an exact match. | |
2054 | // Otherwise we treat it as a wildcard pattern. This function returns | |
2055 | // true for a wildcard pattern. Otherwise, it does backslash | |
2056 | // unquoting on *PATTERN and returns false. If this returns true, | |
2057 | // *PATTERN may have been partially unquoted. | |
2058 | ||
2059 | bool | |
2060 | Version_script_info::unquote(std::string* pattern) const | |
2061 | { | |
2062 | bool saw_backslash = false; | |
2063 | size_t len = pattern->length(); | |
2064 | size_t j = 0; | |
2065 | for (size_t i = 0; i < len; ++i) | |
2066 | { | |
2067 | if (saw_backslash) | |
2068 | saw_backslash = false; | |
2069 | else | |
2070 | { | |
2071 | switch ((*pattern)[i]) | |
2072 | { | |
2073 | case '?': case '[': case '*': | |
2074 | return true; | |
2075 | case '\\': | |
2076 | saw_backslash = true; | |
2077 | continue; | |
2078 | default: | |
2079 | break; | |
2080 | } | |
2081 | } | |
2082 | ||
2083 | if (i != j) | |
2084 | (*pattern)[j] = (*pattern)[i]; | |
2085 | ++j; | |
2086 | } | |
2087 | return false; | |
2088 | } | |
2089 | ||
2090 | // Add an exact match for MATCH to *PE. The result of the match is | |
2091 | // V/IS_GLOBAL. | |
2092 | ||
2093 | void | |
2094 | Version_script_info::add_exact_match(const std::string& match, | |
2095 | const Version_tree* v, bool is_global, | |
2096 | const Version_expression* ve, | |
2097 | Exact *pe) | |
2098 | { | |
2099 | std::pair<Exact::iterator, bool> ins = | |
2100 | pe->insert(std::make_pair(match, Version_tree_match(v, is_global, ve))); | |
2101 | if (ins.second) | |
2102 | { | |
2103 | // This is the first time we have seen this match. | |
2104 | return; | |
2105 | } | |
2106 | ||
2107 | Version_tree_match& vtm(ins.first->second); | |
2108 | if (vtm.real->tag != v->tag) | |
2109 | { | |
2110 | // This is an ambiguous match. We still return the | |
2111 | // first version that we found in the script, but we | |
2112 | // record the new version to issue a warning if we | |
2113 | // wind up looking up this symbol. | |
2114 | if (vtm.ambiguous == NULL) | |
2115 | vtm.ambiguous = v; | |
2116 | } | |
2117 | else if (is_global != vtm.is_global) | |
2118 | { | |
2119 | // We have a match for both the global and local entries for a | |
2120 | // version tag. That's got to be wrong. | |
2121 | gold_error(_("'%s' appears as both a global and a local symbol " | |
2122 | "for version '%s' in script"), | |
2123 | match.c_str(), v->tag.c_str()); | |
6affe781 ILT |
2124 | } |
2125 | } | |
2126 | ||
2127 | // Build fast lookup information for EXPLIST and store it in LOOKUP. | |
98e090bd ILT |
2128 | // All matches go to V, and IS_GLOBAL is true if they are global |
2129 | // matches. | |
6affe781 ILT |
2130 | |
2131 | void | |
2132 | Version_script_info::build_expression_list_lookup( | |
2133 | const Version_expression_list* explist, | |
2134 | const Version_tree* v, | |
98e090bd | 2135 | bool is_global) |
6affe781 ILT |
2136 | { |
2137 | if (explist == NULL) | |
2138 | return; | |
2139 | size_t size = explist->expressions.size(); | |
98e090bd | 2140 | for (size_t i = 0; i < size; ++i) |
6affe781 | 2141 | { |
98e090bd ILT |
2142 | const Version_expression& exp(explist->expressions[i]); |
2143 | ||
2144 | if (exp.pattern.length() == 1 && exp.pattern[0] == '*') | |
6affe781 | 2145 | { |
98e090bd ILT |
2146 | if (this->default_version_ != NULL |
2147 | && this->default_version_->tag != v->tag) | |
d0a91bd8 ILT |
2148 | gold_warning(_("wildcard match appears in both version '%s' " |
2149 | "and '%s' in script"), | |
2150 | this->default_version_->tag.c_str(), v->tag.c_str()); | |
98e090bd ILT |
2151 | else if (this->default_version_ != NULL |
2152 | && this->default_is_global_ != is_global) | |
2153 | gold_error(_("wildcard match appears as both global and local " | |
2154 | "in version '%s' in script"), | |
2155 | v->tag.c_str()); | |
2156 | this->default_version_ = v; | |
2157 | this->default_is_global_ = is_global; | |
2158 | continue; | |
2159 | } | |
2160 | ||
2161 | std::string pattern = exp.pattern; | |
2162 | if (!exp.exact_match) | |
2163 | { | |
2164 | if (this->unquote(&pattern)) | |
6affe781 | 2165 | { |
98e090bd ILT |
2166 | this->globs_.push_back(Glob(&exp, v, is_global)); |
2167 | continue; | |
6affe781 ILT |
2168 | } |
2169 | } | |
98e090bd ILT |
2170 | |
2171 | if (this->exact_[exp.language] == NULL) | |
2172 | this->exact_[exp.language] = new Exact(); | |
2173 | this->add_exact_match(pattern, v, is_global, &exp, | |
2174 | this->exact_[exp.language]); | |
6affe781 ILT |
2175 | } |
2176 | } | |
2177 | ||
98e090bd ILT |
2178 | // Return the name to match given a name, a language code, and two |
2179 | // lazy demanglers. | |
62dfdd4d | 2180 | |
98e090bd ILT |
2181 | const char* |
2182 | Version_script_info::get_name_to_match(const char* name, | |
2183 | int language, | |
2184 | Lazy_demangler* cpp_demangler, | |
2185 | Lazy_demangler* java_demangler) const | |
62dfdd4d | 2186 | { |
98e090bd | 2187 | switch (language) |
62dfdd4d | 2188 | { |
98e090bd ILT |
2189 | case LANGUAGE_C: |
2190 | return name; | |
2191 | case LANGUAGE_CXX: | |
2192 | return cpp_demangler->get(); | |
2193 | case LANGUAGE_JAVA: | |
2194 | return java_demangler->get(); | |
2195 | default: | |
2196 | gold_unreachable(); | |
62dfdd4d | 2197 | } |
62dfdd4d ILT |
2198 | } |
2199 | ||
98e090bd ILT |
2200 | // Look up SYMBOL_NAME in the list of versions. Return true if the |
2201 | // symbol is found, false if not. If the symbol is found, then if | |
2202 | // PVERSION is not NULL, set *PVERSION to the version tag, and if | |
2203 | // P_IS_GLOBAL is not NULL, set *P_IS_GLOBAL according to whether the | |
2204 | // symbol is global or not. | |
057ead22 ILT |
2205 | |
2206 | bool | |
98e090bd ILT |
2207 | Version_script_info::get_symbol_version(const char* symbol_name, |
2208 | std::string* pversion, | |
2209 | bool* p_is_global) const | |
494e05f4 | 2210 | { |
98e090bd ILT |
2211 | Lazy_demangler cpp_demangled_name(symbol_name, DMGL_ANSI | DMGL_PARAMS); |
2212 | Lazy_demangler java_demangled_name(symbol_name, | |
2213 | DMGL_ANSI | DMGL_PARAMS | DMGL_JAVA); | |
2214 | ||
6affe781 | 2215 | gold_assert(this->is_finalized_); |
6affe781 | 2216 | for (int i = 0; i < LANGUAGE_COUNT; ++i) |
494e05f4 | 2217 | { |
98e090bd ILT |
2218 | Exact* exact = this->exact_[i]; |
2219 | if (exact == NULL) | |
6affe781 ILT |
2220 | continue; |
2221 | ||
98e090bd ILT |
2222 | const char* name_to_match = this->get_name_to_match(symbol_name, i, |
2223 | &cpp_demangled_name, | |
2224 | &java_demangled_name); | |
2225 | if (name_to_match == NULL) | |
6affe781 | 2226 | { |
98e090bd ILT |
2227 | // If the name can not be demangled, the GNU linker goes |
2228 | // ahead and tries to match it anyhow. That does not | |
2229 | // make sense to me and I have not implemented it. | |
2230 | continue; | |
6affe781 ILT |
2231 | } |
2232 | ||
98e090bd ILT |
2233 | Exact::const_iterator pe = exact->find(name_to_match); |
2234 | if (pe != exact->end()) | |
6affe781 | 2235 | { |
98e090bd ILT |
2236 | const Version_tree_match& vtm(pe->second); |
2237 | if (vtm.ambiguous != NULL) | |
2238 | gold_warning(_("using '%s' as version for '%s' which is also " | |
2239 | "named in version '%s' in script"), | |
2240 | vtm.real->tag.c_str(), name_to_match, | |
2241 | vtm.ambiguous->tag.c_str()); | |
2242 | ||
6affe781 | 2243 | if (pversion != NULL) |
98e090bd ILT |
2244 | *pversion = vtm.real->tag; |
2245 | if (p_is_global != NULL) | |
2246 | *p_is_global = vtm.is_global; | |
62dfdd4d ILT |
2247 | |
2248 | // If we are using --no-undefined-version, and this is a | |
2249 | // global symbol, we have to record that we have found this | |
2250 | // symbol, so that we don't warn about it. We have to do | |
2251 | // this now, because otherwise we have no way to get from a | |
2252 | // non-C language back to the demangled name that we | |
2253 | // matched. | |
98e090bd ILT |
2254 | if (p_is_global != NULL && vtm.is_global) |
2255 | vtm.expression->was_matched_by_symbol = true; | |
62dfdd4d | 2256 | |
6affe781 ILT |
2257 | return true; |
2258 | } | |
98e090bd ILT |
2259 | } |
2260 | ||
2261 | // Look through the glob patterns in reverse order. | |
6affe781 | 2262 | |
98e090bd ILT |
2263 | for (Globs::const_reverse_iterator p = this->globs_.rbegin(); |
2264 | p != this->globs_.rend(); | |
2265 | ++p) | |
2266 | { | |
2267 | int language = p->expression->language; | |
2268 | const char* name_to_match = this->get_name_to_match(symbol_name, | |
2269 | language, | |
2270 | &cpp_demangled_name, | |
2271 | &java_demangled_name); | |
2272 | if (name_to_match == NULL) | |
2273 | continue; | |
2274 | ||
2275 | if (fnmatch(p->expression->pattern.c_str(), name_to_match, | |
2276 | FNM_NOESCAPE) == 0) | |
6affe781 | 2277 | { |
98e090bd ILT |
2278 | if (pversion != NULL) |
2279 | *pversion = p->version->tag; | |
2280 | if (p_is_global != NULL) | |
2281 | *p_is_global = p->is_global; | |
2282 | return true; | |
6affe781 | 2283 | } |
98e090bd | 2284 | } |
6affe781 | 2285 | |
98e090bd ILT |
2286 | // Finally, there may be a wildcard. |
2287 | if (this->default_version_ != NULL) | |
2288 | { | |
2289 | if (pversion != NULL) | |
2290 | *pversion = this->default_version_->tag; | |
2291 | if (p_is_global != NULL) | |
2292 | *p_is_global = this->default_is_global_; | |
2293 | return true; | |
494e05f4 | 2294 | } |
6affe781 | 2295 | |
057ead22 | 2296 | return false; |
494e05f4 ILT |
2297 | } |
2298 | ||
62dfdd4d ILT |
2299 | // Give an error if any exact symbol names (not wildcards) appear in a |
2300 | // version script, but there is no such symbol. | |
2301 | ||
2302 | void | |
2303 | Version_script_info::check_unmatched_names(const Symbol_table* symtab) const | |
2304 | { | |
2305 | for (size_t i = 0; i < this->version_trees_.size(); ++i) | |
2306 | { | |
2307 | const Version_tree* vt = this->version_trees_[i]; | |
2308 | if (vt->global == NULL) | |
2309 | continue; | |
2310 | for (size_t j = 0; j < vt->global->expressions.size(); ++j) | |
2311 | { | |
2312 | const Version_expression& expression(vt->global->expressions[j]); | |
2313 | ||
2314 | // Ignore cases where we used the version because we saw a | |
2315 | // symbol that we looked up. Note that | |
2316 | // WAS_MATCHED_BY_SYMBOL will be true even if the symbol was | |
2317 | // not a definition. That's OK as in that case we most | |
2318 | // likely gave an undefined symbol error anyhow. | |
2319 | if (expression.was_matched_by_symbol) | |
2320 | continue; | |
2321 | ||
2322 | // Just ignore names which are in languages other than C. | |
2323 | // We have no way to look them up in the symbol table. | |
2324 | if (expression.language != LANGUAGE_C) | |
2325 | continue; | |
2326 | ||
98e090bd ILT |
2327 | // Remove backslash quoting, and ignore wildcard patterns. |
2328 | std::string pattern = expression.pattern; | |
2329 | if (!expression.exact_match) | |
62dfdd4d | 2330 | { |
98e090bd ILT |
2331 | if (this->unquote(&pattern)) |
2332 | continue; | |
62dfdd4d | 2333 | } |
98e090bd ILT |
2334 | |
2335 | if (symtab->lookup(pattern.c_str(), vt->tag.c_str()) == NULL) | |
2336 | gold_error(_("version script assignment of %s to symbol %s " | |
2337 | "failed: symbol not defined"), | |
2338 | vt->tag.c_str(), pattern.c_str()); | |
62dfdd4d ILT |
2339 | } |
2340 | } | |
2341 | } | |
2342 | ||
494e05f4 ILT |
2343 | struct Version_dependency_list* |
2344 | Version_script_info::allocate_dependency_list() | |
2345 | { | |
2346 | dependency_lists_.push_back(new Version_dependency_list); | |
2347 | return dependency_lists_.back(); | |
2348 | } | |
2349 | ||
2350 | struct Version_expression_list* | |
2351 | Version_script_info::allocate_expression_list() | |
2352 | { | |
2353 | expression_lists_.push_back(new Version_expression_list); | |
2354 | return expression_lists_.back(); | |
2355 | } | |
2356 | ||
2357 | struct Version_tree* | |
2358 | Version_script_info::allocate_version_tree() | |
2359 | { | |
2360 | version_trees_.push_back(new Version_tree); | |
2361 | return version_trees_.back(); | |
2362 | } | |
2363 | ||
2364 | // Print for debugging. | |
2365 | ||
2366 | void | |
2367 | Version_script_info::print(FILE* f) const | |
2368 | { | |
2369 | if (this->empty()) | |
2370 | return; | |
2371 | ||
2372 | fprintf(f, "VERSION {"); | |
2373 | ||
2374 | for (size_t i = 0; i < this->version_trees_.size(); ++i) | |
2375 | { | |
2376 | const Version_tree* vt = this->version_trees_[i]; | |
2377 | ||
2378 | if (vt->tag.empty()) | |
2379 | fprintf(f, " {\n"); | |
2380 | else | |
2381 | fprintf(f, " %s {\n", vt->tag.c_str()); | |
2382 | ||
2383 | if (vt->global != NULL) | |
2384 | { | |
2385 | fprintf(f, " global :\n"); | |
2386 | this->print_expression_list(f, vt->global); | |
2387 | } | |
2388 | ||
2389 | if (vt->local != NULL) | |
2390 | { | |
2391 | fprintf(f, " local :\n"); | |
2392 | this->print_expression_list(f, vt->local); | |
2393 | } | |
2394 | ||
2395 | fprintf(f, " }"); | |
2396 | if (vt->dependencies != NULL) | |
2397 | { | |
2398 | const Version_dependency_list* deps = vt->dependencies; | |
2399 | for (size_t j = 0; j < deps->dependencies.size(); ++j) | |
2400 | { | |
2401 | if (j < deps->dependencies.size() - 1) | |
2402 | fprintf(f, "\n"); | |
2403 | fprintf(f, " %s", deps->dependencies[j].c_str()); | |
2404 | } | |
2405 | } | |
2406 | fprintf(f, ";\n"); | |
2407 | } | |
2408 | ||
2409 | fprintf(f, "}\n"); | |
2410 | } | |
2411 | ||
2412 | void | |
2413 | Version_script_info::print_expression_list( | |
2414 | FILE* f, | |
2415 | const Version_expression_list* vel) const | |
2416 | { | |
6affe781 | 2417 | Version_script_info::Language current_language = LANGUAGE_C; |
494e05f4 ILT |
2418 | for (size_t i = 0; i < vel->expressions.size(); ++i) |
2419 | { | |
2420 | const Version_expression& ve(vel->expressions[i]); | |
2421 | ||
2422 | if (ve.language != current_language) | |
2423 | { | |
6affe781 | 2424 | if (current_language != LANGUAGE_C) |
494e05f4 | 2425 | fprintf(f, " }\n"); |
6affe781 ILT |
2426 | switch (ve.language) |
2427 | { | |
2428 | case LANGUAGE_C: | |
2429 | break; | |
2430 | case LANGUAGE_CXX: | |
2431 | fprintf(f, " extern \"C++\" {\n"); | |
2432 | break; | |
2433 | case LANGUAGE_JAVA: | |
2434 | fprintf(f, " extern \"Java\" {\n"); | |
2435 | break; | |
2436 | default: | |
2437 | gold_unreachable(); | |
2438 | } | |
494e05f4 ILT |
2439 | current_language = ve.language; |
2440 | } | |
2441 | ||
2442 | fprintf(f, " "); | |
6affe781 | 2443 | if (current_language != LANGUAGE_C) |
494e05f4 ILT |
2444 | fprintf(f, " "); |
2445 | ||
2446 | if (ve.exact_match) | |
2447 | fprintf(f, "\""); | |
2448 | fprintf(f, "%s", ve.pattern.c_str()); | |
2449 | if (ve.exact_match) | |
2450 | fprintf(f, "\""); | |
2451 | ||
2452 | fprintf(f, "\n"); | |
2453 | } | |
2454 | ||
6affe781 | 2455 | if (current_language != LANGUAGE_C) |
494e05f4 ILT |
2456 | fprintf(f, " }\n"); |
2457 | } | |
2458 | ||
2459 | } // End namespace gold. | |
2460 | ||
2461 | // The remaining functions are extern "C", so it's clearer to not put | |
2462 | // them in namespace gold. | |
2463 | ||
2464 | using namespace gold; | |
2465 | ||
2466 | // This function is called by the bison parser to return the next | |
2467 | // token. | |
2468 | ||
2469 | extern "C" int | |
2470 | yylex(YYSTYPE* lvalp, void* closurev) | |
2471 | { | |
2472 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2473 | const Token* token = closure->next_token(); | |
2474 | switch (token->classification()) | |
2475 | { | |
2476 | default: | |
2477 | gold_unreachable(); | |
2478 | ||
2479 | case Token::TOKEN_INVALID: | |
2480 | yyerror(closurev, "invalid character"); | |
2481 | return 0; | |
2482 | ||
2483 | case Token::TOKEN_EOF: | |
2484 | return 0; | |
2485 | ||
2486 | case Token::TOKEN_STRING: | |
2487 | { | |
2488 | // This is either a keyword or a STRING. | |
2489 | size_t len; | |
2490 | const char* str = token->string_value(&len); | |
2491 | int parsecode = 0; | |
2492 | switch (closure->lex_mode()) | |
2493 | { | |
2494 | case Lex::LINKER_SCRIPT: | |
2495 | parsecode = script_keywords.keyword_to_parsecode(str, len); | |
2496 | break; | |
2497 | case Lex::VERSION_SCRIPT: | |
2498 | parsecode = version_script_keywords.keyword_to_parsecode(str, len); | |
2499 | break; | |
c82fbeee CS |
2500 | case Lex::DYNAMIC_LIST: |
2501 | parsecode = dynamic_list_keywords.keyword_to_parsecode(str, len); | |
2502 | break; | |
494e05f4 ILT |
2503 | default: |
2504 | break; | |
2505 | } | |
2506 | if (parsecode != 0) | |
2507 | return parsecode; | |
2508 | lvalp->string.value = str; | |
2509 | lvalp->string.length = len; | |
2510 | return STRING; | |
2511 | } | |
2512 | ||
2513 | case Token::TOKEN_QUOTED_STRING: | |
2514 | lvalp->string.value = token->string_value(&lvalp->string.length); | |
2515 | return QUOTED_STRING; | |
2516 | ||
2517 | case Token::TOKEN_OPERATOR: | |
2518 | return token->operator_value(); | |
2519 | ||
2520 | case Token::TOKEN_INTEGER: | |
2521 | lvalp->integer = token->integer_value(); | |
2522 | return INTEGER; | |
2523 | } | |
2524 | } | |
2525 | ||
2526 | // This function is called by the bison parser to report an error. | |
2527 | ||
2528 | extern "C" void | |
2529 | yyerror(void* closurev, const char* message) | |
2530 | { | |
2531 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2532 | gold_error(_("%s:%d:%d: %s"), closure->filename(), closure->lineno(), | |
2533 | closure->charpos(), message); | |
2534 | } | |
2535 | ||
afc06bb8 ILT |
2536 | // Called by the bison parser to add an external symbol to the link. |
2537 | ||
2538 | extern "C" void | |
2539 | script_add_extern(void* closurev, const char* name, size_t length) | |
2540 | { | |
2541 | // We treat exactly like -u NAME. FIXME: If it seems useful, we | |
2542 | // could handle this after the command line has been read, by adding | |
2543 | // entries to the symbol table directly. | |
2544 | std::string arg("--undefined="); | |
2545 | arg.append(name, length); | |
2546 | script_parse_option(closurev, arg.c_str(), arg.size()); | |
2547 | } | |
2548 | ||
494e05f4 ILT |
2549 | // Called by the bison parser to add a file to the link. |
2550 | ||
2551 | extern "C" void | |
2552 | script_add_file(void* closurev, const char* name, size_t length) | |
2553 | { | |
2554 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2555 | ||
2556 | // If this is an absolute path, and we found the script in the | |
2557 | // sysroot, then we want to prepend the sysroot to the file name. | |
2558 | // For example, this is how we handle a cross link to the x86_64 | |
2559 | // libc.so, which refers to /lib/libc.so.6. | |
2560 | std::string name_string(name, length); | |
2561 | const char* extra_search_path = "."; | |
2562 | std::string script_directory; | |
2563 | if (IS_ABSOLUTE_PATH(name_string.c_str())) | |
2564 | { | |
2565 | if (closure->is_in_sysroot()) | |
2566 | { | |
8851ecca | 2567 | const std::string& sysroot(parameters->options().sysroot()); |
494e05f4 ILT |
2568 | gold_assert(!sysroot.empty()); |
2569 | name_string = sysroot + name_string; | |
2570 | } | |
2571 | } | |
2572 | else | |
2573 | { | |
2574 | // In addition to checking the normal library search path, we | |
2575 | // also want to check in the script-directory. | |
2576 | const char *slash = strrchr(closure->filename(), '/'); | |
2577 | if (slash != NULL) | |
2578 | { | |
2579 | script_directory.assign(closure->filename(), | |
2580 | slash - closure->filename() + 1); | |
2581 | extra_search_path = script_directory.c_str(); | |
2582 | } | |
2583 | } | |
2584 | ||
ae3b5189 CD |
2585 | Input_file_argument file(name_string.c_str(), |
2586 | Input_file_argument::INPUT_FILE_TYPE_FILE, | |
2587 | extra_search_path, false, | |
2588 | closure->position_dependent_options()); | |
494e05f4 | 2589 | closure->inputs()->add_file(file); |
dbe717ef ILT |
2590 | } |
2591 | ||
2592 | // Called by the bison parser to start a group. If we are already in | |
2593 | // a group, that means that this script was invoked within a | |
2594 | // --start-group --end-group sequence on the command line, or that | |
2595 | // this script was found in a GROUP of another script. In that case, | |
2596 | // we simply continue the existing group, rather than starting a new | |
2597 | // one. It is possible to construct a case in which this will do | |
2598 | // something other than what would happen if we did a recursive group, | |
2599 | // but it's hard to imagine why the different behaviour would be | |
2600 | // useful for a real program. Avoiding recursive groups is simpler | |
2601 | // and more efficient. | |
2602 | ||
2603 | extern "C" void | |
2604 | script_start_group(void* closurev) | |
2605 | { | |
2606 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2607 | if (!closure->in_group()) | |
2608 | closure->inputs()->start_group(); | |
2609 | } | |
2610 | ||
2611 | // Called by the bison parser at the end of a group. | |
2612 | ||
2613 | extern "C" void | |
2614 | script_end_group(void* closurev) | |
2615 | { | |
2616 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2617 | if (!closure->in_group()) | |
2618 | closure->inputs()->end_group(); | |
2619 | } | |
2620 | ||
2621 | // Called by the bison parser to start an AS_NEEDED list. | |
2622 | ||
2623 | extern "C" void | |
2624 | script_start_as_needed(void* closurev) | |
2625 | { | |
2626 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
45aa233b | 2627 | closure->position_dependent_options().set_as_needed(true); |
dbe717ef ILT |
2628 | } |
2629 | ||
2630 | // Called by the bison parser at the end of an AS_NEEDED list. | |
2631 | ||
2632 | extern "C" void | |
2633 | script_end_as_needed(void* closurev) | |
2634 | { | |
2635 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
45aa233b | 2636 | closure->position_dependent_options().set_as_needed(false); |
dbe717ef | 2637 | } |
195e7dc6 | 2638 | |
d391083d ILT |
2639 | // Called by the bison parser to set the entry symbol. |
2640 | ||
2641 | extern "C" void | |
e5756efb | 2642 | script_set_entry(void* closurev, const char* entry, size_t length) |
d391083d | 2643 | { |
a5dc0706 ILT |
2644 | // We'll parse this exactly the same as --entry=ENTRY on the commandline |
2645 | // TODO(csilvers): FIXME -- call set_entry directly. | |
1890b465 | 2646 | std::string arg("--entry="); |
a5dc0706 ILT |
2647 | arg.append(entry, length); |
2648 | script_parse_option(closurev, arg.c_str(), arg.size()); | |
e5756efb ILT |
2649 | } |
2650 | ||
0dfbdef4 ILT |
2651 | // Called by the bison parser to set whether to define common symbols. |
2652 | ||
2653 | extern "C" void | |
2654 | script_set_common_allocation(void* closurev, int set) | |
2655 | { | |
2656 | const char* arg = set != 0 ? "--define-common" : "--no-define-common"; | |
2657 | script_parse_option(closurev, arg, strlen(arg)); | |
2658 | } | |
2659 | ||
e5756efb ILT |
2660 | // Called by the bison parser to define a symbol. |
2661 | ||
2662 | extern "C" void | |
2663 | script_set_symbol(void* closurev, const char* name, size_t length, | |
2664 | Expression* value, int providei, int hiddeni) | |
2665 | { | |
2666 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2667 | const bool provide = providei != 0; | |
2668 | const bool hidden = hiddeni != 0; | |
99fff23b ILT |
2669 | closure->script_options()->add_symbol_assignment(name, length, |
2670 | closure->parsing_defsym(), | |
2671 | value, provide, hidden); | |
15f8229b | 2672 | closure->clear_skip_on_incompatible_target(); |
d391083d ILT |
2673 | } |
2674 | ||
494e05f4 ILT |
2675 | // Called by the bison parser to add an assertion. |
2676 | ||
2677 | extern "C" void | |
2678 | script_add_assertion(void* closurev, Expression* check, const char* message, | |
2679 | size_t messagelen) | |
2680 | { | |
2681 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2682 | closure->script_options()->add_assertion(check, message, messagelen); | |
15f8229b | 2683 | closure->clear_skip_on_incompatible_target(); |
494e05f4 ILT |
2684 | } |
2685 | ||
195e7dc6 ILT |
2686 | // Called by the bison parser to parse an OPTION. |
2687 | ||
2688 | extern "C" void | |
e5756efb | 2689 | script_parse_option(void* closurev, const char* option, size_t length) |
195e7dc6 ILT |
2690 | { |
2691 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
a0451b38 ILT |
2692 | // We treat the option as a single command-line option, even if |
2693 | // it has internal whitespace. | |
2694 | if (closure->command_line() == NULL) | |
2695 | { | |
2696 | // There are some options that we could handle here--e.g., | |
2697 | // -lLIBRARY. Should we bother? | |
e5756efb | 2698 | gold_warning(_("%s:%d:%d: ignoring command OPTION; OPTION is only valid" |
d391083d | 2699 | " for scripts specified via -T/--script"), |
e5756efb | 2700 | closure->filename(), closure->lineno(), closure->charpos()); |
a0451b38 ILT |
2701 | } |
2702 | else | |
2703 | { | |
2704 | bool past_a_double_dash_option = false; | |
ee1fe73e | 2705 | const char* mutable_option = strndup(option, length); |
e5756efb | 2706 | gold_assert(mutable_option != NULL); |
a0451b38 ILT |
2707 | closure->command_line()->process_one_option(1, &mutable_option, 0, |
2708 | &past_a_double_dash_option); | |
a5dc0706 ILT |
2709 | // The General_options class will quite possibly store a pointer |
2710 | // into mutable_option, so we can't free it. In cases the class | |
2711 | // does not store such a pointer, this is a memory leak. Alas. :( | |
a0451b38 | 2712 | } |
15f8229b ILT |
2713 | closure->clear_skip_on_incompatible_target(); |
2714 | } | |
2715 | ||
2716 | // Called by the bison parser to handle OUTPUT_FORMAT. OUTPUT_FORMAT | |
2717 | // takes either one or three arguments. In the three argument case, | |
2718 | // the format depends on the endianness option, which we don't | |
2719 | // currently support (FIXME). If we see an OUTPUT_FORMAT for the | |
2720 | // wrong format, then we want to search for a new file. Returning 0 | |
2721 | // here will cause the parser to immediately abort. | |
2722 | ||
2723 | extern "C" int | |
2724 | script_check_output_format(void* closurev, | |
2725 | const char* default_name, size_t default_length, | |
2726 | const char*, size_t, const char*, size_t) | |
2727 | { | |
2728 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2729 | std::string name(default_name, default_length); | |
2730 | Target* target = select_target_by_name(name.c_str()); | |
2731 | if (target == NULL || !parameters->is_compatible_target(target)) | |
2732 | { | |
2733 | if (closure->skip_on_incompatible_target()) | |
2734 | { | |
2735 | closure->set_found_incompatible_target(); | |
2736 | return 0; | |
2737 | } | |
2738 | // FIXME: Should we warn about the unknown target? | |
2739 | } | |
2740 | return 1; | |
195e7dc6 | 2741 | } |
e5756efb | 2742 | |
e6a307ba ILT |
2743 | // Called by the bison parser to handle TARGET. |
2744 | ||
2745 | extern "C" void | |
2746 | script_set_target(void* closurev, const char* target, size_t len) | |
2747 | { | |
2748 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2749 | std::string s(target, len); | |
2750 | General_options::Object_format format_enum; | |
2751 | format_enum = General_options::string_to_object_format(s.c_str()); | |
2752 | closure->position_dependent_options().set_format_enum(format_enum); | |
2753 | } | |
2754 | ||
3802b2dd ILT |
2755 | // Called by the bison parser to handle SEARCH_DIR. This is handled |
2756 | // exactly like a -L option. | |
2757 | ||
2758 | extern "C" void | |
2759 | script_add_search_dir(void* closurev, const char* option, size_t length) | |
2760 | { | |
2761 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2762 | if (closure->command_line() == NULL) | |
2763 | gold_warning(_("%s:%d:%d: ignoring SEARCH_DIR; SEARCH_DIR is only valid" | |
2764 | " for scripts specified via -T/--script"), | |
2765 | closure->filename(), closure->lineno(), closure->charpos()); | |
2766 | else | |
2767 | { | |
2768 | std::string s = "-L" + std::string(option, length); | |
2769 | script_parse_option(closurev, s.c_str(), s.size()); | |
2770 | } | |
2771 | } | |
2772 | ||
e5756efb ILT |
2773 | /* Called by the bison parser to push the lexer into expression |
2774 | mode. */ | |
2775 | ||
494e05f4 | 2776 | extern "C" void |
e5756efb ILT |
2777 | script_push_lex_into_expression_mode(void* closurev) |
2778 | { | |
2779 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2780 | closure->push_lex_mode(Lex::EXPRESSION); | |
2781 | } | |
2782 | ||
09124467 ILT |
2783 | /* Called by the bison parser to push the lexer into version |
2784 | mode. */ | |
2785 | ||
494e05f4 | 2786 | extern "C" void |
09124467 ILT |
2787 | script_push_lex_into_version_mode(void* closurev) |
2788 | { | |
2789 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
6affe781 ILT |
2790 | if (closure->version_script()->is_finalized()) |
2791 | gold_error(_("%s:%d:%d: invalid use of VERSION in input file"), | |
2792 | closure->filename(), closure->lineno(), closure->charpos()); | |
09124467 ILT |
2793 | closure->push_lex_mode(Lex::VERSION_SCRIPT); |
2794 | } | |
2795 | ||
e5756efb ILT |
2796 | /* Called by the bison parser to pop the lexer mode. */ |
2797 | ||
494e05f4 | 2798 | extern "C" void |
e5756efb ILT |
2799 | script_pop_lex_mode(void* closurev) |
2800 | { | |
2801 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2802 | closure->pop_lex_mode(); | |
2803 | } | |
09124467 | 2804 | |
09124467 ILT |
2805 | // Register an entire version node. For example: |
2806 | // | |
2807 | // GLIBC_2.1 { | |
2808 | // global: foo; | |
2809 | // } GLIBC_2.0; | |
2810 | // | |
2811 | // - tag is "GLIBC_2.1" | |
2812 | // - tree contains the information "global: foo" | |
2813 | // - deps contains "GLIBC_2.0" | |
2814 | ||
2815 | extern "C" void | |
2816 | script_register_vers_node(void*, | |
2817 | const char* tag, | |
2818 | int taglen, | |
2819 | struct Version_tree *tree, | |
2820 | struct Version_dependency_list *deps) | |
2821 | { | |
2822 | gold_assert(tree != NULL); | |
09124467 | 2823 | tree->dependencies = deps; |
057ead22 ILT |
2824 | if (tag != NULL) |
2825 | tree->tag = std::string(tag, taglen); | |
09124467 ILT |
2826 | } |
2827 | ||
2828 | // Add a dependencies to the list of existing dependencies, if any, | |
2829 | // and return the expanded list. | |
2830 | ||
2831 | extern "C" struct Version_dependency_list * | |
2832 | script_add_vers_depend(void* closurev, | |
2833 | struct Version_dependency_list *all_deps, | |
2834 | const char *depend_to_add, int deplen) | |
2835 | { | |
2836 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2837 | if (all_deps == NULL) | |
2838 | all_deps = closure->version_script()->allocate_dependency_list(); | |
2839 | all_deps->dependencies.push_back(std::string(depend_to_add, deplen)); | |
2840 | return all_deps; | |
2841 | } | |
2842 | ||
2843 | // Add a pattern expression to an existing list of expressions, if any. | |
09124467 ILT |
2844 | |
2845 | extern "C" struct Version_expression_list * | |
2846 | script_new_vers_pattern(void* closurev, | |
2847 | struct Version_expression_list *expressions, | |
10600224 | 2848 | const char *pattern, int patlen, int exact_match) |
09124467 ILT |
2849 | { |
2850 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2851 | if (expressions == NULL) | |
2852 | expressions = closure->version_script()->allocate_expression_list(); | |
2853 | expressions->expressions.push_back( | |
2854 | Version_expression(std::string(pattern, patlen), | |
10600224 ILT |
2855 | closure->get_current_language(), |
2856 | static_cast<bool>(exact_match))); | |
09124467 ILT |
2857 | return expressions; |
2858 | } | |
2859 | ||
10600224 ILT |
2860 | // Attaches b to the end of a, and clears b. So a = a + b and b = {}. |
2861 | ||
2862 | extern "C" struct Version_expression_list* | |
2863 | script_merge_expressions(struct Version_expression_list *a, | |
2864 | struct Version_expression_list *b) | |
2865 | { | |
2866 | a->expressions.insert(a->expressions.end(), | |
2867 | b->expressions.begin(), b->expressions.end()); | |
2868 | // We could delete b and remove it from expressions_lists_, but | |
2869 | // that's a lot of work. This works just as well. | |
2870 | b->expressions.clear(); | |
2871 | return a; | |
2872 | } | |
2873 | ||
09124467 ILT |
2874 | // Combine the global and local expressions into a a Version_tree. |
2875 | ||
2876 | extern "C" struct Version_tree * | |
2877 | script_new_vers_node(void* closurev, | |
2878 | struct Version_expression_list *global, | |
2879 | struct Version_expression_list *local) | |
2880 | { | |
2881 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2882 | Version_tree* tree = closure->version_script()->allocate_version_tree(); | |
2883 | tree->global = global; | |
2884 | tree->local = local; | |
2885 | return tree; | |
2886 | } | |
2887 | ||
10600224 | 2888 | // Handle a transition in language, such as at the |
09124467 ILT |
2889 | // start or end of 'extern "C++"' |
2890 | ||
2891 | extern "C" void | |
2892 | version_script_push_lang(void* closurev, const char* lang, int langlen) | |
2893 | { | |
2894 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
6affe781 ILT |
2895 | std::string language(lang, langlen); |
2896 | Version_script_info::Language code; | |
2897 | if (language.empty() || language == "C") | |
2898 | code = Version_script_info::LANGUAGE_C; | |
2899 | else if (language == "C++") | |
2900 | code = Version_script_info::LANGUAGE_CXX; | |
2901 | else if (language == "Java") | |
2902 | code = Version_script_info::LANGUAGE_JAVA; | |
2903 | else | |
2904 | { | |
2905 | char* buf = new char[langlen + 100]; | |
2906 | snprintf(buf, langlen + 100, | |
2907 | _("unrecognized version script language '%s'"), | |
2908 | language.c_str()); | |
2909 | yyerror(closurev, buf); | |
2910 | delete[] buf; | |
2911 | code = Version_script_info::LANGUAGE_C; | |
2912 | } | |
2913 | closure->push_language(code); | |
09124467 ILT |
2914 | } |
2915 | ||
2916 | extern "C" void | |
2917 | version_script_pop_lang(void* closurev) | |
2918 | { | |
2919 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2920 | closure->pop_language(); | |
2921 | } | |
494e05f4 ILT |
2922 | |
2923 | // Called by the bison parser to start a SECTIONS clause. | |
2924 | ||
2925 | extern "C" void | |
2926 | script_start_sections(void* closurev) | |
2927 | { | |
2928 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2929 | closure->script_options()->script_sections()->start_sections(); | |
15f8229b | 2930 | closure->clear_skip_on_incompatible_target(); |
494e05f4 ILT |
2931 | } |
2932 | ||
2933 | // Called by the bison parser to finish a SECTIONS clause. | |
2934 | ||
2935 | extern "C" void | |
2936 | script_finish_sections(void* closurev) | |
2937 | { | |
2938 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2939 | closure->script_options()->script_sections()->finish_sections(); | |
2940 | } | |
2941 | ||
2942 | // Start processing entries for an output section. | |
2943 | ||
2944 | extern "C" void | |
2945 | script_start_output_section(void* closurev, const char* name, size_t namelen, | |
2946 | const struct Parser_output_section_header* header) | |
2947 | { | |
2948 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2949 | closure->script_options()->script_sections()->start_output_section(name, | |
2950 | namelen, | |
2951 | header); | |
2952 | } | |
2953 | ||
2954 | // Finish processing entries for an output section. | |
2955 | ||
2956 | extern "C" void | |
c82fbeee | 2957 | script_finish_output_section(void* closurev, |
494e05f4 ILT |
2958 | const struct Parser_output_section_trailer* trail) |
2959 | { | |
2960 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2961 | closure->script_options()->script_sections()->finish_output_section(trail); | |
2962 | } | |
2963 | ||
2964 | // Add a data item (e.g., "WORD (0)") to the current output section. | |
2965 | ||
2966 | extern "C" void | |
2967 | script_add_data(void* closurev, int data_token, Expression* val) | |
2968 | { | |
2969 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
2970 | int size; | |
2971 | bool is_signed = true; | |
2972 | switch (data_token) | |
2973 | { | |
2974 | case QUAD: | |
2975 | size = 8; | |
2976 | is_signed = false; | |
2977 | break; | |
2978 | case SQUAD: | |
2979 | size = 8; | |
2980 | break; | |
2981 | case LONG: | |
2982 | size = 4; | |
2983 | break; | |
2984 | case SHORT: | |
2985 | size = 2; | |
2986 | break; | |
2987 | case BYTE: | |
2988 | size = 1; | |
2989 | break; | |
2990 | default: | |
2991 | gold_unreachable(); | |
2992 | } | |
2993 | closure->script_options()->script_sections()->add_data(size, is_signed, val); | |
2994 | } | |
2995 | ||
2996 | // Add a clause setting the fill value to the current output section. | |
2997 | ||
2998 | extern "C" void | |
2999 | script_add_fill(void* closurev, Expression* val) | |
3000 | { | |
3001 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
3002 | closure->script_options()->script_sections()->add_fill(val); | |
3003 | } | |
3004 | ||
3005 | // Add a new input section specification to the current output | |
3006 | // section. | |
3007 | ||
3008 | extern "C" void | |
3009 | script_add_input_section(void* closurev, | |
3010 | const struct Input_section_spec* spec, | |
3011 | int keepi) | |
3012 | { | |
3013 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
3014 | bool keep = keepi != 0; | |
3015 | closure->script_options()->script_sections()->add_input_section(spec, keep); | |
3016 | } | |
3017 | ||
2d924fd9 ILT |
3018 | // When we see DATA_SEGMENT_ALIGN we record that following output |
3019 | // sections may be relro. | |
3020 | ||
3021 | extern "C" void | |
3022 | script_data_segment_align(void* closurev) | |
3023 | { | |
3024 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
3025 | if (!closure->script_options()->saw_sections_clause()) | |
3026 | gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"), | |
3027 | closure->filename(), closure->lineno(), closure->charpos()); | |
3028 | else | |
3029 | closure->script_options()->script_sections()->data_segment_align(); | |
3030 | } | |
3031 | ||
3032 | // When we see DATA_SEGMENT_RELRO_END we know that all output sections | |
3033 | // since DATA_SEGMENT_ALIGN should be relro. | |
3034 | ||
3035 | extern "C" void | |
3036 | script_data_segment_relro_end(void* closurev) | |
3037 | { | |
3038 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
3039 | if (!closure->script_options()->saw_sections_clause()) | |
3040 | gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"), | |
3041 | closure->filename(), closure->lineno(), closure->charpos()); | |
3042 | else | |
3043 | closure->script_options()->script_sections()->data_segment_relro_end(); | |
3044 | } | |
3045 | ||
494e05f4 ILT |
3046 | // Create a new list of string/sort pairs. |
3047 | ||
3048 | extern "C" String_sort_list_ptr | |
3049 | script_new_string_sort_list(const struct Wildcard_section* string_sort) | |
3050 | { | |
3051 | return new String_sort_list(1, *string_sort); | |
3052 | } | |
3053 | ||
3054 | // Add an entry to a list of string/sort pairs. The way the parser | |
3055 | // works permits us to simply modify the first parameter, rather than | |
3056 | // copy the vector. | |
3057 | ||
3058 | extern "C" String_sort_list_ptr | |
3059 | script_string_sort_list_add(String_sort_list_ptr pv, | |
3060 | const struct Wildcard_section* string_sort) | |
3061 | { | |
a445fddf ILT |
3062 | if (pv == NULL) |
3063 | return script_new_string_sort_list(string_sort); | |
3064 | else | |
3065 | { | |
3066 | pv->push_back(*string_sort); | |
3067 | return pv; | |
3068 | } | |
494e05f4 ILT |
3069 | } |
3070 | ||
3071 | // Create a new list of strings. | |
3072 | ||
3073 | extern "C" String_list_ptr | |
3074 | script_new_string_list(const char* str, size_t len) | |
3075 | { | |
3076 | return new String_list(1, std::string(str, len)); | |
3077 | } | |
3078 | ||
3079 | // Add an element to a list of strings. The way the parser works | |
3080 | // permits us to simply modify the first parameter, rather than copy | |
3081 | // the vector. | |
3082 | ||
3083 | extern "C" String_list_ptr | |
3084 | script_string_list_push_back(String_list_ptr pv, const char* str, size_t len) | |
3085 | { | |
1c4f3631 ILT |
3086 | if (pv == NULL) |
3087 | return script_new_string_list(str, len); | |
3088 | else | |
3089 | { | |
3090 | pv->push_back(std::string(str, len)); | |
3091 | return pv; | |
3092 | } | |
494e05f4 ILT |
3093 | } |
3094 | ||
3095 | // Concatenate two string lists. Either or both may be NULL. The way | |
3096 | // the parser works permits us to modify the parameters, rather than | |
3097 | // copy the vector. | |
3098 | ||
3099 | extern "C" String_list_ptr | |
3100 | script_string_list_append(String_list_ptr pv1, String_list_ptr pv2) | |
3101 | { | |
3102 | if (pv1 == NULL) | |
3103 | return pv2; | |
3104 | if (pv2 == NULL) | |
3105 | return pv1; | |
3106 | pv1->insert(pv1->end(), pv2->begin(), pv2->end()); | |
3107 | return pv1; | |
3108 | } | |
1c4f3631 ILT |
3109 | |
3110 | // Add a new program header. | |
3111 | ||
3112 | extern "C" void | |
3113 | script_add_phdr(void* closurev, const char* name, size_t namelen, | |
3114 | unsigned int type, const Phdr_info* info) | |
3115 | { | |
3116 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
3117 | bool includes_filehdr = info->includes_filehdr != 0; | |
3118 | bool includes_phdrs = info->includes_phdrs != 0; | |
3119 | bool is_flags_valid = info->is_flags_valid != 0; | |
3120 | Script_sections* ss = closure->script_options()->script_sections(); | |
3121 | ss->add_phdr(name, namelen, type, includes_filehdr, includes_phdrs, | |
3122 | is_flags_valid, info->flags, info->load_address); | |
15f8229b | 3123 | closure->clear_skip_on_incompatible_target(); |
1c4f3631 ILT |
3124 | } |
3125 | ||
3126 | // Convert a program header string to a type. | |
3127 | ||
3128 | #define PHDR_TYPE(NAME) { #NAME, sizeof(#NAME) - 1, elfcpp::NAME } | |
3129 | ||
3130 | static struct | |
3131 | { | |
3132 | const char* name; | |
3133 | size_t namelen; | |
3134 | unsigned int val; | |
3135 | } phdr_type_names[] = | |
3136 | { | |
3137 | PHDR_TYPE(PT_NULL), | |
3138 | PHDR_TYPE(PT_LOAD), | |
3139 | PHDR_TYPE(PT_DYNAMIC), | |
3140 | PHDR_TYPE(PT_INTERP), | |
3141 | PHDR_TYPE(PT_NOTE), | |
3142 | PHDR_TYPE(PT_SHLIB), | |
3143 | PHDR_TYPE(PT_PHDR), | |
3144 | PHDR_TYPE(PT_TLS), | |
3145 | PHDR_TYPE(PT_GNU_EH_FRAME), | |
3146 | PHDR_TYPE(PT_GNU_STACK), | |
3147 | PHDR_TYPE(PT_GNU_RELRO) | |
3148 | }; | |
3149 | ||
3150 | extern "C" unsigned int | |
3151 | script_phdr_string_to_type(void* closurev, const char* name, size_t namelen) | |
3152 | { | |
3153 | for (unsigned int i = 0; | |
3154 | i < sizeof(phdr_type_names) / sizeof(phdr_type_names[0]); | |
3155 | ++i) | |
3156 | if (namelen == phdr_type_names[i].namelen | |
3157 | && strncmp(name, phdr_type_names[i].name, namelen) == 0) | |
3158 | return phdr_type_names[i].val; | |
3159 | yyerror(closurev, _("unknown PHDR type (try integer)")); | |
3160 | return elfcpp::PT_NULL; | |
3161 | } | |
3c12dcdb DK |
3162 | |
3163 | extern "C" void | |
3164 | script_saw_segment_start_expression(void* closurev) | |
3165 | { | |
3166 | Parser_closure* closure = static_cast<Parser_closure*>(closurev); | |
3167 | Script_sections* ss = closure->script_options()->script_sections(); | |
3168 | ss->set_saw_segment_start_expression(true); | |
3169 | } |