OpenCores
URL https://opencores.org/ocsvn/open8_urisc/open8_urisc/trunk

Subversion Repositories open8_urisc

[/] [open8_urisc/] [trunk/] [gnu/] [binutils/] [gold/] [script.cc] - Blame information for rev 46

Go to most recent revision | Details | Compare with Previous | View Log

Line No. Rev Author Line
1 27 khays
// script.cc -- handle linker scripts for gold.
2
 
3
// Copyright 2006, 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
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
 
23
#include "gold.h"
24
 
25
#include <cstdio>
26
#include <cstdlib>
27
#include <cstring>
28
#include <fnmatch.h>
29
#include <string>
30
#include <vector>
31
#include "filenames.h"
32
 
33
#include "elfcpp.h"
34
#include "demangle.h"
35
#include "dirsearch.h"
36
#include "options.h"
37
#include "fileread.h"
38
#include "workqueue.h"
39
#include "readsyms.h"
40
#include "parameters.h"
41
#include "layout.h"
42
#include "symtab.h"
43
#include "target-select.h"
44
#include "script.h"
45
#include "script-c.h"
46
#include "incremental.h"
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,
66
    // Token is a quoted string of characters.
67
    TOKEN_QUOTED_STRING,
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()
76
    : classification_(TOKEN_INVALID), value_(NULL), value_length_(0),
77
      opcode_(0), lineno_(0), charpos_(0)
78
  { }
79
 
80
  // A general token with no value.
81
  Token(Classification classification, int lineno, int charpos)
82
    : classification_(classification), value_(NULL), value_length_(0),
83
      opcode_(0), lineno_(lineno), charpos_(charpos)
84
  {
85
    gold_assert(classification == TOKEN_INVALID
86
                || classification == TOKEN_EOF);
87
  }
88
 
89
  // A general token with a value.
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)
94
  {
95
    gold_assert(classification != TOKEN_INVALID
96
                && classification != TOKEN_EOF);
97
  }
98
 
99
  // A token representing an operator.
100
  Token(int opcode, int lineno, int charpos)
101
    : classification_(TOKEN_OPERATOR), value_(NULL), value_length_(0),
102
      opcode_(opcode), lineno_(lineno), charpos_(charpos)
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
 
132
  const char*
133
  string_value(size_t* length) const
134
  {
135
    gold_assert(this->classification_ == TOKEN_STRING
136
                || this->classification_ == TOKEN_QUOTED_STRING);
137
    *length = this->value_length_;
138
    return this->value_;
139
  }
140
 
141
  int
142
  operator_value() const
143
  {
144
    gold_assert(this->classification_ == TOKEN_OPERATOR);
145
    return this->opcode_;
146
  }
147
 
148
  uint64_t
149
  integer_value() const
150
  {
151
    gold_assert(this->classification_ == TOKEN_INTEGER);
152
    // Null terminate.
153
    std::string s(this->value_, this->value_length_);
154
    return strtoull(s.c_str(), NULL, 0);
155
  }
156
 
157
 private:
158
  // The token classification.
159
  Classification classification_;
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_;
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
 
174
// This class handles lexing a file into a sequence of tokens.
175
 
176
class Lex
177
{
178
 public:
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.
189
    VERSION_SCRIPT,
190
    // Reading a --dynamic-list file.
191
    DYNAMIC_LIST
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)
199
  { }
200
 
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();
208
 
209
  // Return the current lexing mode.
210
  Lex::Mode
211
  mode() const
212
  { return this->mode_; }
213
 
214
  // Set the lexing mode.
215
  void
216
  set_mode(Mode mode)
217
  { this->mode_ = mode; }
218
 
219
 private:
220
  Lex(const Lex&);
221
  Lex& operator=(const Lex&);
222
 
223
  // Make a general token with no value at the current location.
224
  Token
225
  make_token(Token::Classification c, const char* start) const
226
  { return Token(c, this->lineno_, start - this->linestart_ + 1); }
227
 
228
  // Make a general token with a value at the current location.
229
  Token
230
  make_token(Token::Classification c, const char* v, size_t len,
231
             const char* start)
232
    const
233
  { return Token(c, v, len, this->lineno_, start - this->linestart_ + 1); }
234
 
235
  // Make an operator token at the current location.
236
  Token
237
  make_token(int opcode, const char* start) const
238
  { return Token(opcode, this->lineno_, start - this->linestart_ + 1); }
239
 
240
  // Make an invalid token at the current location.
241
  Token
242
  make_invalid_token(const char* start)
243
  { return this->make_token(Token::TOKEN_INVALID, start); }
244
 
245
  // Make an EOF token at the current location.
246
  Token
247
  make_eof_token(const char* start)
248
  { return this->make_token(Token::TOKEN_EOF, start); }
249
 
250
  // Return whether C can be the first character in a name.  C2 is the
251
  // next character, since we sometimes need that.
252
  inline bool
253
  can_start_name(char c, char c2);
254
 
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);
260
 
261
  // Return whether C, C2, C3 can start a hex number.
262
  inline bool
263
  can_start_hex(char c, char c2, char c3);
264
 
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);
270
 
271
  // Return whether C can start a non-hex number.
272
  static inline bool
273
  can_start_number(char c);
274
 
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; }
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
314
  gather_token(Token::Classification,
315
               const char* (Lex::*can_continue_fn)(const char*),
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
 
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_;
335
  // The current line number.
336
  int lineno_;
337
  // The start of the current line in the string.
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
346
Lex::read_file(Input_file* input_file, std::string* contents)
347
{
348
  off_t filesize = input_file->file().filesize();
349
  contents->clear();
350
  contents->reserve(filesize);
351
 
352
  off_t off = 0;
353
  unsigned char buf[BUFSIZ];
354
  while (off < filesize)
355
    {
356
      off_t get = BUFSIZ;
357
      if (get > filesize - off)
358
        get = filesize - off;
359
      input_file->file().read(off, get, buf);
360
      contents->append(reinterpret_cast<char*>(&buf[0]), get);
361
      off += get;
362
    }
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
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
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':
391
    case '_': case '.': case '$':
392
      return true;
393
 
394
    case '/': case '\\':
395
      return this->mode_ == LINKER_SCRIPT;
396
 
397
    case '~':
398
      return this->mode_ == LINKER_SCRIPT && can_continue_name(&c2);
399
 
400
    case '*': case '[':
401
      return (this->mode_ == VERSION_SCRIPT
402
              || this->mode_ == DYNAMIC_LIST
403
              || (this->mode_ == LINKER_SCRIPT
404
                  && can_continue_name(&c2)));
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
414
// script language requires spaces around operators, unless we know
415
// that we are parsing an expression.
416
 
417
inline const char*
418
Lex::can_continue_name(const char* c)
419
{
420
  switch (*c)
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':
432
    case '_': case '.': case '$':
433
    case '0': case '1': case '2': case '3': case '4':
434
    case '5': case '6': case '7': case '8': case '9':
435
      return c + 1;
436
 
437
    // TODO(csilvers): why not allow ~ in names for version-scripts?
438
    case '/': case '\\': case '~':
439
    case '=': case '+':
440
    case ',':
441
      if (this->mode_ == LINKER_SCRIPT)
442
        return c + 1;
443
      return NULL;
444
 
445
    case '[': case ']': case '*': case '?': case '-':
446
      if (this->mode_ == LINKER_SCRIPT || this->mode_ == VERSION_SCRIPT
447
          || this->mode_ == DYNAMIC_LIST)
448
        return c + 1;
449
      return NULL;
450
 
451
    // TODO(csilvers): why allow this?  ^ is meaningless in version scripts.
452
    case '^':
453
      if (this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST)
454
        return c + 1;
455
      return NULL;
456
 
457
    case ':':
458
      if (this->mode_ == LINKER_SCRIPT)
459
        return c + 1;
460
      else if ((this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST)
461
               && (c[1] == ':'))
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;
468
 
469
    default:
470
      return NULL;
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
478
// kilo.  FIXME: Those are mentioned in the documentation, and we
479
// should accept them.
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'))
487
    return this->can_continue_hex(&c3);
488
  return false;
489
}
490
 
491
// Return whether C can appear in a hex number.
492
 
493
inline const char*
494
Lex::can_continue_hex(const char* c)
495
{
496
  switch (*c)
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':
502
      return c + 1;
503
 
504
    default:
505
      return NULL;
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,
700
                  const char* (Lex::*can_continue_fn)(const char*),
701
                  const char* start,
702
                  const char* match,
703
                  const char** pp)
704
{
705
  const char* new_match = NULL;
706
  while ((new_match = (this->*can_continue_fn)(match)))
707
    match = new_match;
708
  *pp = match;
709
  return this->make_token(classification, start, match - start, start);
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;
724
  return this->make_token(Token::TOKEN_QUOTED_STRING, p, skip, start);
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.
746
      while (*p == ' ' || *p == '\t' || *p == '\r')
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.
782
      if (this->can_start_name(p[0], p[1]))
783
        return this->gather_token(Token::TOKEN_STRING,
784
                                  &Lex::can_continue_name,
785
                                  p, p + 1, pp);
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
 
797
      if (this->can_start_hex(p[0], p[1], p[2]))
798
        return this->gather_token(Token::TOKEN_INTEGER,
799
                                  &Lex::can_continue_hex,
800
                                  p, p + 3, pp);
801
 
802
      if (Lex::can_start_number(p[0]))
803
        return this->gather_token(Token::TOKEN_INTEGER,
804
                                  &Lex::can_continue_number,
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
 
834
// Return the next token.
835
 
836
const Token*
837
Lex::next_token()
838
{
839
  // The first token is special.
840
  if (this->first_token_ != 0)
841
    {
842
      this->token_ = Token(this->first_token_, 0, 0);
843
      this->first_token_ = 0;
844
      return &this->token_;
845
    }
846
 
847
  this->token_ = this->get_token(&this->current_);
848
 
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_);
855
 
856
  return &this->token_;
857
}
858
 
859
// class Symbol_assignment.
860
 
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.
866
 
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
 
871
void
872
Symbol_assignment::add_to_table(Symbol_table* symtab)
873
{
874
  elfcpp::STV vis = this->hidden_ ? elfcpp::STV_HIDDEN : elfcpp::STV_DEFAULT;
875
  this->sym_ = symtab->define_as_constant(this->name_.c_str(),
876
                                          NULL, // version
877
                                          (this->is_defsym_
878
                                           ? Symbol_table::DEFSYM
879
                                           : Symbol_table::SCRIPT),
880
                                          0, // value
881
                                          0, // size
882
                                          elfcpp::STT_NOTYPE,
883
                                          elfcpp::STB_GLOBAL,
884
                                          vis,
885
                                          0, // nonvis
886
                                          this->provide_,
887
                                          true); // force_override
888
}
889
 
890
// Finalize a symbol value.
891
 
892
void
893
Symbol_assignment::finalize(Symbol_table* symtab, const Layout* layout)
894
{
895
  this->finalize_maybe_dot(symtab, layout, false, 0, NULL);
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,
903
                                     uint64_t dot_value,
904
                                     Output_section* dot_section)
905
{
906
  this->finalize_maybe_dot(symtab, layout, true, dot_value, dot_section);
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,
915
                                      uint64_t dot_value,
916
                                      Output_section* dot_section)
917
{
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
 
926
  if (parameters->target().get_size() == 32)
927
    {
928
#if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
929
      this->sized_finalize<32>(symtab, layout, is_dot_available, dot_value,
930
                               dot_section);
931
#else
932
      gold_unreachable();
933
#endif
934
    }
935
  else if (parameters->target().get_size() == 64)
936
    {
937
#if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
938
      this->sized_finalize<64>(symtab, layout, is_dot_available, dot_value,
939
                               dot_section);
940
#else
941
      gold_unreachable();
942
#endif
943
    }
944
  else
945
    gold_unreachable();
946
}
947
 
948
template<int size>
949
void
950
Symbol_assignment::sized_finalize(Symbol_table* symtab, const Layout* layout,
951
                                  bool is_dot_available, uint64_t dot_value,
952
                                  Output_section* dot_section)
953
{
954
  Output_section* section;
955
  uint64_t final_val = this->val_->eval_maybe_dot(symtab, layout, true,
956
                                                  is_dot_available,
957
                                                  dot_value, dot_section,
958
                                                  &section, NULL);
959
  Sized_symbol<size>* ssym = symtab->get_sized_symbol<size>(this->sym_);
960
  ssym->set_value(final_val);
961
  if (section != NULL)
962
    ssym->set_output_section(section);
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,
969
                                   bool is_dot_available, uint64_t dot_value)
970
{
971
  if (this->sym_ == NULL)
972
    return;
973
 
974
  Output_section* val_section;
975
  uint64_t val = this->val_->eval_maybe_dot(symtab, layout, false,
976
                                            is_dot_available, dot_value,
977
                                            NULL, &val_section, NULL);
978
  if (val_section != NULL)
979
    return;
980
 
981
  if (parameters->target().get_size() == 32)
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
    }
990
  else if (parameters->target().get_size() == 64)
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();
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
{
1031
  if (!this->check_->eval(symtab, layout, true))
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_(), symbol_definitions_(),
1049
    symbol_references_(), version_script_info_(), script_sections_()
1050
{
1051
}
1052
 
1053
// Returns true if NAME is on the list of symbol assignments waiting
1054
// to be processed.
1055
 
1056
bool
1057
Script_options::is_pending_assignment(const char* name)
1058
{
1059
  for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1060
       p != this->symbol_assignments_.end();
1061
       ++p)
1062
    if ((*p)->name() == name)
1063
      return true;
1064
  return false;
1065
}
1066
 
1067
// Add a symbol to be defined.
1068
 
1069
void
1070
Script_options::add_symbol_assignment(const char* name, size_t length,
1071
                                      bool is_defsym, Expression* value,
1072
                                      bool provide, bool hidden)
1073
{
1074
  if (length != 1 || name[0] != '.')
1075
    {
1076
      if (this->script_sections_.in_sections_clause())
1077
        {
1078
          gold_assert(!is_defsym);
1079
          this->script_sections_.add_symbol_assignment(name, length, value,
1080
                                                       provide, hidden);
1081
        }
1082
      else
1083
        {
1084
          Symbol_assignment* p = new Symbol_assignment(name, length, is_defsym,
1085
                                                       value, provide, hidden);
1086
          this->symbol_assignments_.push_back(p);
1087
        }
1088
 
1089
      if (!provide)
1090
        {
1091
          std::string n(name, length);
1092
          this->symbol_definitions_.insert(n);
1093
          this->symbol_references_.erase(n);
1094
        }
1095
    }
1096
  else
1097
    {
1098
      if (provide || hidden)
1099
        gold_error(_("invalid use of PROVIDE for dot symbol"));
1100
 
1101
      // The GNU linker permits assignments to dot outside of SECTIONS
1102
      // clauses and treats them as occurring inside, so we don't
1103
      // check in_sections_clause here.
1104
      this->script_sections_.add_dot_assignment(value);
1105
    }
1106
}
1107
 
1108
// Add a reference to a symbol.
1109
 
1110
void
1111
Script_options::add_symbol_reference(const char* name, size_t length)
1112
{
1113
  if (length != 1 || name[0] != '.')
1114
    {
1115
      std::string n(name, length);
1116
      if (this->symbol_definitions_.find(n) == this->symbol_definitions_.end())
1117
        this->symbol_references_.insert(n);
1118
    }
1119
}
1120
 
1121
// Add an assertion.
1122
 
1123
void
1124
Script_options::add_assertion(Expression* check, const char* message,
1125
                              size_t messagelen)
1126
{
1127
  if (this->script_sections_.in_sections_clause())
1128
    this->script_sections_.add_assertion(check, message, messagelen);
1129
  else
1130
    {
1131
      Script_assertion* p = new Script_assertion(check, message, messagelen);
1132
      this->assertions_.push_back(p);
1133
    }
1134
}
1135
 
1136
// Create sections required by any linker scripts.
1137
 
1138
void
1139
Script_options::create_script_sections(Layout* layout)
1140
{
1141
  if (this->saw_sections_clause())
1142
    this->script_sections_.create_sections(layout);
1143
}
1144
 
1145
// Add any symbols we are defining to the symbol table.
1146
 
1147
void
1148
Script_options::add_symbols_to_table(Symbol_table* symtab)
1149
{
1150
  for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1151
       p != this->symbol_assignments_.end();
1152
       ++p)
1153
    (*p)->add_to_table(symtab);
1154
  this->script_sections_.add_symbols_to_table(symtab);
1155
}
1156
 
1157
// Finalize symbol values.  Also check assertions.
1158
 
1159
void
1160
Script_options::finalize_symbols(Symbol_table* symtab, const Layout* layout)
1161
{
1162
  // We finalize the symbols defined in SECTIONS first, because they
1163
  // are the ones which may have changed.  This way if symbol outside
1164
  // SECTIONS are defined in terms of symbols inside SECTIONS, they
1165
  // will get the right value.
1166
  this->script_sections_.finalize_symbols(symtab, layout);
1167
 
1168
  for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1169
       p != this->symbol_assignments_.end();
1170
       ++p)
1171
    (*p)->finalize(symtab, layout);
1172
 
1173
  for (Assertions::iterator p = this->assertions_.begin();
1174
       p != this->assertions_.end();
1175
       ++p)
1176
    (*p)->check(symtab, layout);
1177
}
1178
 
1179
// Set section addresses.  We set all the symbols which have absolute
1180
// values.  Then we let the SECTIONS clause do its thing.  This
1181
// returns the segment which holds the file header and segment
1182
// headers, if any.
1183
 
1184
Output_segment*
1185
Script_options::set_section_addresses(Symbol_table* symtab, Layout* layout)
1186
{
1187
  for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1188
       p != this->symbol_assignments_.end();
1189
       ++p)
1190
    (*p)->set_if_absolute(symtab, layout, false, 0);
1191
 
1192
  return this->script_sections_.set_section_addresses(symtab, layout);
1193
}
1194
 
1195
// This class holds data passed through the parser to the lexer and to
1196
// the parser support functions.  This avoids global variables.  We
1197
// can't use global variables because we need not be called by a
1198
// singleton thread.
1199
 
1200
class Parser_closure
1201
{
1202
 public:
1203
  Parser_closure(const char* filename,
1204
                 const Position_dependent_options& posdep_options,
1205
                 bool parsing_defsym, bool in_group, bool is_in_sysroot,
1206
                 Command_line* command_line,
1207
                 Script_options* script_options,
1208
                 Lex* lex,
1209
                 bool skip_on_incompatible_target,
1210
                 Script_info* script_info)
1211
    : filename_(filename), posdep_options_(posdep_options),
1212
      parsing_defsym_(parsing_defsym), in_group_(in_group),
1213
      is_in_sysroot_(is_in_sysroot),
1214
      skip_on_incompatible_target_(skip_on_incompatible_target),
1215
      found_incompatible_target_(false),
1216
      command_line_(command_line), script_options_(script_options),
1217
      version_script_info_(script_options->version_script_info()),
1218
      lex_(lex), lineno_(0), charpos_(0), lex_mode_stack_(), inputs_(NULL),
1219
      script_info_(script_info)
1220
  {
1221
    // We start out processing C symbols in the default lex mode.
1222
    this->language_stack_.push_back(Version_script_info::LANGUAGE_C);
1223
    this->lex_mode_stack_.push_back(lex->mode());
1224
  }
1225
 
1226
  // Return the file name.
1227
  const char*
1228
  filename() const
1229
  { return this->filename_; }
1230
 
1231
  // Return the position dependent options.  The caller may modify
1232
  // this.
1233
  Position_dependent_options&
1234
  position_dependent_options()
1235
  { return this->posdep_options_; }
1236
 
1237
  // Whether we are parsing a --defsym.
1238
  bool
1239
  parsing_defsym() const
1240
  { return this->parsing_defsym_; }
1241
 
1242
  // Return whether this script is being run in a group.
1243
  bool
1244
  in_group() const
1245
  { return this->in_group_; }
1246
 
1247
  // Return whether this script was found using a directory in the
1248
  // sysroot.
1249
  bool
1250
  is_in_sysroot() const
1251
  { return this->is_in_sysroot_; }
1252
 
1253
  // Whether to skip to the next file with the same name if we find an
1254
  // incompatible target in an OUTPUT_FORMAT statement.
1255
  bool
1256
  skip_on_incompatible_target() const
1257
  { return this->skip_on_incompatible_target_; }
1258
 
1259
  // Stop skipping to the next file on an incompatible target.  This
1260
  // is called when we make some unrevocable change to the data
1261
  // structures.
1262
  void
1263
  clear_skip_on_incompatible_target()
1264
  { this->skip_on_incompatible_target_ = false; }
1265
 
1266
  // Whether we found an incompatible target in an OUTPUT_FORMAT
1267
  // statement.
1268
  bool
1269
  found_incompatible_target() const
1270
  { return this->found_incompatible_target_; }
1271
 
1272
  // Note that we found an incompatible target.
1273
  void
1274
  set_found_incompatible_target()
1275
  { this->found_incompatible_target_ = true; }
1276
 
1277
  // Returns the Command_line structure passed in at constructor time.
1278
  // This value may be NULL.  The caller may modify this, which modifies
1279
  // the passed-in Command_line object (not a copy).
1280
  Command_line*
1281
  command_line()
1282
  { return this->command_line_; }
1283
 
1284
  // Return the options which may be set by a script.
1285
  Script_options*
1286
  script_options()
1287
  { return this->script_options_; }
1288
 
1289
  // Return the object in which version script information should be stored.
1290
  Version_script_info*
1291
  version_script()
1292
  { return this->version_script_info_; }
1293
 
1294
  // Return the next token, and advance.
1295
  const Token*
1296
  next_token()
1297
  {
1298
    const Token* token = this->lex_->next_token();
1299
    this->lineno_ = token->lineno();
1300
    this->charpos_ = token->charpos();
1301
    return token;
1302
  }
1303
 
1304
  // Set a new lexer mode, pushing the current one.
1305
  void
1306
  push_lex_mode(Lex::Mode mode)
1307
  {
1308
    this->lex_mode_stack_.push_back(this->lex_->mode());
1309
    this->lex_->set_mode(mode);
1310
  }
1311
 
1312
  // Pop the lexer mode.
1313
  void
1314
  pop_lex_mode()
1315
  {
1316
    gold_assert(!this->lex_mode_stack_.empty());
1317
    this->lex_->set_mode(this->lex_mode_stack_.back());
1318
    this->lex_mode_stack_.pop_back();
1319
  }
1320
 
1321
  // Return the current lexer mode.
1322
  Lex::Mode
1323
  lex_mode() const
1324
  { return this->lex_mode_stack_.back(); }
1325
 
1326
  // Return the line number of the last token.
1327
  int
1328
  lineno() const
1329
  { return this->lineno_; }
1330
 
1331
  // Return the character position in the line of the last token.
1332
  int
1333
  charpos() const
1334
  { return this->charpos_; }
1335
 
1336
  // Return the list of input files, creating it if necessary.  This
1337
  // is a space leak--we never free the INPUTS_ pointer.
1338
  Input_arguments*
1339
  inputs()
1340
  {
1341
    if (this->inputs_ == NULL)
1342
      this->inputs_ = new Input_arguments();
1343
    return this->inputs_;
1344
  }
1345
 
1346
  // Return whether we saw any input files.
1347
  bool
1348
  saw_inputs() const
1349
  { return this->inputs_ != NULL && !this->inputs_->empty(); }
1350
 
1351
  // Return the current language being processed in a version script
1352
  // (eg, "C++").  The empty string represents unmangled C names.
1353
  Version_script_info::Language
1354
  get_current_language() const
1355
  { return this->language_stack_.back(); }
1356
 
1357
  // Push a language onto the stack when entering an extern block.
1358
  void
1359
  push_language(Version_script_info::Language lang)
1360
  { this->language_stack_.push_back(lang); }
1361
 
1362
  // Pop a language off of the stack when exiting an extern block.
1363
  void
1364
  pop_language()
1365
  {
1366
    gold_assert(!this->language_stack_.empty());
1367
    this->language_stack_.pop_back();
1368
  }
1369
 
1370
  // Return a pointer to the incremental info.
1371
  Script_info*
1372
  script_info()
1373
  { return this->script_info_; }
1374
 
1375
 private:
1376
  // The name of the file we are reading.
1377
  const char* filename_;
1378
  // The position dependent options.
1379
  Position_dependent_options posdep_options_;
1380
  // True if we are parsing a --defsym.
1381
  bool parsing_defsym_;
1382
  // Whether we are currently in a --start-group/--end-group.
1383
  bool in_group_;
1384
  // Whether the script was found in a sysrooted directory.
1385
  bool is_in_sysroot_;
1386
  // If this is true, then if we find an OUTPUT_FORMAT with an
1387
  // incompatible target, then we tell the parser to abort so that we
1388
  // can search for the next file with the same name.
1389
  bool skip_on_incompatible_target_;
1390
  // True if we found an OUTPUT_FORMAT with an incompatible target.
1391
  bool found_incompatible_target_;
1392
  // May be NULL if the user chooses not to pass one in.
1393
  Command_line* command_line_;
1394
  // Options which may be set from any linker script.
1395
  Script_options* script_options_;
1396
  // Information parsed from a version script.
1397
  Version_script_info* version_script_info_;
1398
  // The lexer.
1399
  Lex* lex_;
1400
  // The line number of the last token returned by next_token.
1401
  int lineno_;
1402
  // The column number of the last token returned by next_token.
1403
  int charpos_;
1404
  // A stack of lexer modes.
1405
  std::vector<Lex::Mode> lex_mode_stack_;
1406
  // A stack of which extern/language block we're inside. Can be C++,
1407
  // java, or empty for C.
1408
  std::vector<Version_script_info::Language> language_stack_;
1409
  // New input files found to add to the link.
1410
  Input_arguments* inputs_;
1411
  // Pointer to incremental linking info.
1412
  Script_info* script_info_;
1413
};
1414
 
1415
// FILE was found as an argument on the command line.  Try to read it
1416
// as a script.  Return true if the file was handled.
1417
 
1418
bool
1419
read_input_script(Workqueue* workqueue, Symbol_table* symtab, Layout* layout,
1420
                  Dirsearch* dirsearch, int dirindex,
1421
                  Input_objects* input_objects, Mapfile* mapfile,
1422
                  Input_group* input_group,
1423
                  const Input_argument* input_argument,
1424
                  Input_file* input_file, Task_token* next_blocker,
1425
                  bool* used_next_blocker)
1426
{
1427
  *used_next_blocker = false;
1428
 
1429
  std::string input_string;
1430
  Lex::read_file(input_file, &input_string);
1431
 
1432
  Lex lex(input_string.c_str(), input_string.length(), PARSING_LINKER_SCRIPT);
1433
 
1434
  Script_info* script_info = NULL;
1435
  if (layout->incremental_inputs() != NULL)
1436
    {
1437
      const std::string& filename = input_file->filename();
1438
      Timespec mtime = input_file->file().get_mtime();
1439
      unsigned int arg_serial = input_argument->file().arg_serial();
1440
      script_info = new Script_info(filename);
1441
      layout->incremental_inputs()->report_script(script_info, arg_serial,
1442
                                                  mtime);
1443
    }
1444
 
1445
  Parser_closure closure(input_file->filename().c_str(),
1446
                         input_argument->file().options(),
1447
                         false,
1448
                         input_group != NULL,
1449
                         input_file->is_in_sysroot(),
1450
                         NULL,
1451
                         layout->script_options(),
1452
                         &lex,
1453
                         input_file->will_search_for(),
1454
                         script_info);
1455
 
1456
  bool old_saw_sections_clause =
1457
    layout->script_options()->saw_sections_clause();
1458
 
1459
  if (yyparse(&closure) != 0)
1460
    {
1461
      if (closure.found_incompatible_target())
1462
        {
1463
          Read_symbols::incompatible_warning(input_argument, input_file);
1464
          Read_symbols::requeue(workqueue, input_objects, symtab, layout,
1465
                                dirsearch, dirindex, mapfile, input_argument,
1466
                                input_group, next_blocker);
1467
          return true;
1468
        }
1469
      return false;
1470
    }
1471
 
1472
  if (!old_saw_sections_clause
1473
      && layout->script_options()->saw_sections_clause()
1474
      && layout->have_added_input_section())
1475
    gold_error(_("%s: SECTIONS seen after other input files; try -T/--script"),
1476
               input_file->filename().c_str());
1477
 
1478
  if (!closure.saw_inputs())
1479
    return true;
1480
 
1481
  Task_token* this_blocker = NULL;
1482
  for (Input_arguments::const_iterator p = closure.inputs()->begin();
1483
       p != closure.inputs()->end();
1484
       ++p)
1485
    {
1486
      Task_token* nb;
1487
      if (p + 1 == closure.inputs()->end())
1488
        nb = next_blocker;
1489
      else
1490
        {
1491
          nb = new Task_token(true);
1492
          nb->add_blocker();
1493
        }
1494
      workqueue->queue_soon(new Read_symbols(input_objects, symtab,
1495
                                             layout, dirsearch, 0, mapfile, &*p,
1496
                                             input_group, NULL, this_blocker, nb));
1497
      this_blocker = nb;
1498
    }
1499
 
1500
  *used_next_blocker = true;
1501
 
1502
  return true;
1503
}
1504
 
1505
// Helper function for read_version_script() and
1506
// read_commandline_script().  Processes the given file in the mode
1507
// indicated by first_token and lex_mode.
1508
 
1509
static bool
1510
read_script_file(const char* filename, Command_line* cmdline,
1511
                 Script_options* script_options,
1512
                 int first_token, Lex::Mode lex_mode)
1513
{
1514
  // TODO: if filename is a relative filename, search for it manually
1515
  // using "." + cmdline->options()->search_path() -- not dirsearch.
1516
  Dirsearch dirsearch;
1517
 
1518
  // The file locking code wants to record a Task, but we haven't
1519
  // started the workqueue yet.  This is only for debugging purposes,
1520
  // so we invent a fake value.
1521
  const Task* task = reinterpret_cast<const Task*>(-1);
1522
 
1523
  // We don't want this file to be opened in binary mode.
1524
  Position_dependent_options posdep = cmdline->position_dependent_options();
1525
  if (posdep.format_enum() == General_options::OBJECT_FORMAT_BINARY)
1526
    posdep.set_format_enum(General_options::OBJECT_FORMAT_ELF);
1527
  Input_file_argument input_argument(filename,
1528
                                     Input_file_argument::INPUT_FILE_TYPE_FILE,
1529
                                     "", false, posdep);
1530
  Input_file input_file(&input_argument);
1531
  int dummy = 0;
1532
  if (!input_file.open(dirsearch, task, &dummy))
1533
    return false;
1534
 
1535
  std::string input_string;
1536
  Lex::read_file(&input_file, &input_string);
1537
 
1538
  Lex lex(input_string.c_str(), input_string.length(), first_token);
1539
  lex.set_mode(lex_mode);
1540
 
1541
  Parser_closure closure(filename,
1542
                         cmdline->position_dependent_options(),
1543
                         first_token == Lex::DYNAMIC_LIST,
1544
                         false,
1545
                         input_file.is_in_sysroot(),
1546
                         cmdline,
1547
                         script_options,
1548
                         &lex,
1549
                         false,
1550
                         NULL);
1551
  if (yyparse(&closure) != 0)
1552
    {
1553
      input_file.file().unlock(task);
1554
      return false;
1555
    }
1556
 
1557
  input_file.file().unlock(task);
1558
 
1559
  gold_assert(!closure.saw_inputs());
1560
 
1561
  return true;
1562
}
1563
 
1564
// FILENAME was found as an argument to --script (-T).
1565
// Read it as a script, and execute its contents immediately.
1566
 
1567
bool
1568
read_commandline_script(const char* filename, Command_line* cmdline)
1569
{
1570
  return read_script_file(filename, cmdline, &cmdline->script_options(),
1571
                          PARSING_LINKER_SCRIPT, Lex::LINKER_SCRIPT);
1572
}
1573
 
1574
// FILENAME was found as an argument to --version-script.  Read it as
1575
// a version script, and store its contents in
1576
// cmdline->script_options()->version_script_info().
1577
 
1578
bool
1579
read_version_script(const char* filename, Command_line* cmdline)
1580
{
1581
  return read_script_file(filename, cmdline, &cmdline->script_options(),
1582
                          PARSING_VERSION_SCRIPT, Lex::VERSION_SCRIPT);
1583
}
1584
 
1585
// FILENAME was found as an argument to --dynamic-list.  Read it as a
1586
// list of symbols, and store its contents in DYNAMIC_LIST.
1587
 
1588
bool
1589
read_dynamic_list(const char* filename, Command_line* cmdline,
1590
                  Script_options* dynamic_list)
1591
{
1592
  return read_script_file(filename, cmdline, dynamic_list,
1593
                          PARSING_DYNAMIC_LIST, Lex::DYNAMIC_LIST);
1594
}
1595
 
1596
// Implement the --defsym option on the command line.  Return true if
1597
// all is well.
1598
 
1599
bool
1600
Script_options::define_symbol(const char* definition)
1601
{
1602
  Lex lex(definition, strlen(definition), PARSING_DEFSYM);
1603
  lex.set_mode(Lex::EXPRESSION);
1604
 
1605
  // Dummy value.
1606
  Position_dependent_options posdep_options;
1607
 
1608
  Parser_closure closure("command line", posdep_options, true,
1609
                         false, false, NULL, this, &lex, false, NULL);
1610
 
1611
  if (yyparse(&closure) != 0)
1612
    return false;
1613
 
1614
  gold_assert(!closure.saw_inputs());
1615
 
1616
  return true;
1617
}
1618
 
1619
// Print the script to F for debugging.
1620
 
1621
void
1622
Script_options::print(FILE* f) const
1623
{
1624
  fprintf(f, "%s: Dumping linker script\n", program_name);
1625
 
1626
  if (!this->entry_.empty())
1627
    fprintf(f, "ENTRY(%s)\n", this->entry_.c_str());
1628
 
1629
  for (Symbol_assignments::const_iterator p =
1630
         this->symbol_assignments_.begin();
1631
       p != this->symbol_assignments_.end();
1632
       ++p)
1633
    (*p)->print(f);
1634
 
1635
  for (Assertions::const_iterator p = this->assertions_.begin();
1636
       p != this->assertions_.end();
1637
       ++p)
1638
    (*p)->print(f);
1639
 
1640
  this->script_sections_.print(f);
1641
 
1642
  this->version_script_info_.print(f);
1643
}
1644
 
1645
// Manage mapping from keywords to the codes expected by the bison
1646
// parser.  We construct one global object for each lex mode with
1647
// keywords.
1648
 
1649
class Keyword_to_parsecode
1650
{
1651
 public:
1652
  // The structure which maps keywords to parsecodes.
1653
  struct Keyword_parsecode
1654
  {
1655
    // Keyword.
1656
    const char* keyword;
1657
    // Corresponding parsecode.
1658
    int parsecode;
1659
  };
1660
 
1661
  Keyword_to_parsecode(const Keyword_parsecode* keywords,
1662
                       int keyword_count)
1663
      : keyword_parsecodes_(keywords), keyword_count_(keyword_count)
1664
  { }
1665
 
1666
  // Return the parsecode corresponding KEYWORD, or 0 if it is not a
1667
  // keyword.
1668
  int
1669
  keyword_to_parsecode(const char* keyword, size_t len) const;
1670
 
1671
 private:
1672
  const Keyword_parsecode* keyword_parsecodes_;
1673
  const int keyword_count_;
1674
};
1675
 
1676
// Mapping from keyword string to keyword parsecode.  This array must
1677
// be kept in sorted order.  Parsecodes are looked up using bsearch.
1678
// This array must correspond to the list of parsecodes in yyscript.y.
1679
 
1680
static const Keyword_to_parsecode::Keyword_parsecode
1681
script_keyword_parsecodes[] =
1682
{
1683
  { "ABSOLUTE", ABSOLUTE },
1684
  { "ADDR", ADDR },
1685
  { "ALIGN", ALIGN_K },
1686
  { "ALIGNOF", ALIGNOF },
1687
  { "ASSERT", ASSERT_K },
1688
  { "AS_NEEDED", AS_NEEDED },
1689
  { "AT", AT },
1690
  { "BIND", BIND },
1691
  { "BLOCK", BLOCK },
1692
  { "BYTE", BYTE },
1693
  { "CONSTANT", CONSTANT },
1694
  { "CONSTRUCTORS", CONSTRUCTORS },
1695
  { "COPY", COPY },
1696
  { "CREATE_OBJECT_SYMBOLS", CREATE_OBJECT_SYMBOLS },
1697
  { "DATA_SEGMENT_ALIGN", DATA_SEGMENT_ALIGN },
1698
  { "DATA_SEGMENT_END", DATA_SEGMENT_END },
1699
  { "DATA_SEGMENT_RELRO_END", DATA_SEGMENT_RELRO_END },
1700
  { "DEFINED", DEFINED },
1701
  { "DSECT", DSECT },
1702
  { "ENTRY", ENTRY },
1703
  { "EXCLUDE_FILE", EXCLUDE_FILE },
1704
  { "EXTERN", EXTERN },
1705
  { "FILL", FILL },
1706
  { "FLOAT", FLOAT },
1707
  { "FORCE_COMMON_ALLOCATION", FORCE_COMMON_ALLOCATION },
1708
  { "GROUP", GROUP },
1709
  { "HLL", HLL },
1710
  { "INCLUDE", INCLUDE },
1711
  { "INFO", INFO },
1712
  { "INHIBIT_COMMON_ALLOCATION", INHIBIT_COMMON_ALLOCATION },
1713
  { "INPUT", INPUT },
1714
  { "KEEP", KEEP },
1715
  { "LENGTH", LENGTH },
1716
  { "LOADADDR", LOADADDR },
1717
  { "LONG", LONG },
1718
  { "MAP", MAP },
1719
  { "MAX", MAX_K },
1720
  { "MEMORY", MEMORY },
1721
  { "MIN", MIN_K },
1722
  { "NEXT", NEXT },
1723
  { "NOCROSSREFS", NOCROSSREFS },
1724
  { "NOFLOAT", NOFLOAT },
1725
  { "NOLOAD", NOLOAD },
1726
  { "ONLY_IF_RO", ONLY_IF_RO },
1727
  { "ONLY_IF_RW", ONLY_IF_RW },
1728
  { "OPTION", OPTION },
1729
  { "ORIGIN", ORIGIN },
1730
  { "OUTPUT", OUTPUT },
1731
  { "OUTPUT_ARCH", OUTPUT_ARCH },
1732
  { "OUTPUT_FORMAT", OUTPUT_FORMAT },
1733
  { "OVERLAY", OVERLAY },
1734
  { "PHDRS", PHDRS },
1735
  { "PROVIDE", PROVIDE },
1736
  { "PROVIDE_HIDDEN", PROVIDE_HIDDEN },
1737
  { "QUAD", QUAD },
1738
  { "SEARCH_DIR", SEARCH_DIR },
1739
  { "SECTIONS", SECTIONS },
1740
  { "SEGMENT_START", SEGMENT_START },
1741
  { "SHORT", SHORT },
1742
  { "SIZEOF", SIZEOF },
1743
  { "SIZEOF_HEADERS", SIZEOF_HEADERS },
1744
  { "SORT", SORT_BY_NAME },
1745
  { "SORT_BY_ALIGNMENT", SORT_BY_ALIGNMENT },
1746
  { "SORT_BY_NAME", SORT_BY_NAME },
1747
  { "SPECIAL", SPECIAL },
1748
  { "SQUAD", SQUAD },
1749
  { "STARTUP", STARTUP },
1750
  { "SUBALIGN", SUBALIGN },
1751
  { "SYSLIB", SYSLIB },
1752
  { "TARGET", TARGET_K },
1753
  { "TRUNCATE", TRUNCATE },
1754
  { "VERSION", VERSIONK },
1755
  { "global", GLOBAL },
1756
  { "l", LENGTH },
1757
  { "len", LENGTH },
1758
  { "local", LOCAL },
1759
  { "o", ORIGIN },
1760
  { "org", ORIGIN },
1761
  { "sizeof_headers", SIZEOF_HEADERS },
1762
};
1763
 
1764
static const Keyword_to_parsecode
1765
script_keywords(&script_keyword_parsecodes[0],
1766
                (sizeof(script_keyword_parsecodes)
1767
                 / sizeof(script_keyword_parsecodes[0])));
1768
 
1769
static const Keyword_to_parsecode::Keyword_parsecode
1770
version_script_keyword_parsecodes[] =
1771
{
1772
  { "extern", EXTERN },
1773
  { "global", GLOBAL },
1774
  { "local", LOCAL },
1775
};
1776
 
1777
static const Keyword_to_parsecode
1778
version_script_keywords(&version_script_keyword_parsecodes[0],
1779
                        (sizeof(version_script_keyword_parsecodes)
1780
                         / sizeof(version_script_keyword_parsecodes[0])));
1781
 
1782
static const Keyword_to_parsecode::Keyword_parsecode
1783
dynamic_list_keyword_parsecodes[] =
1784
{
1785
  { "extern", EXTERN },
1786
};
1787
 
1788
static const Keyword_to_parsecode
1789
dynamic_list_keywords(&dynamic_list_keyword_parsecodes[0],
1790
                      (sizeof(dynamic_list_keyword_parsecodes)
1791
                       / sizeof(dynamic_list_keyword_parsecodes[0])));
1792
 
1793
 
1794
 
1795
// Comparison function passed to bsearch.
1796
 
1797
extern "C"
1798
{
1799
 
1800
struct Ktt_key
1801
{
1802
  const char* str;
1803
  size_t len;
1804
};
1805
 
1806
static int
1807
ktt_compare(const void* keyv, const void* kttv)
1808
{
1809
  const Ktt_key* key = static_cast<const Ktt_key*>(keyv);
1810
  const Keyword_to_parsecode::Keyword_parsecode* ktt =
1811
    static_cast<const Keyword_to_parsecode::Keyword_parsecode*>(kttv);
1812
  int i = strncmp(key->str, ktt->keyword, key->len);
1813
  if (i != 0)
1814
    return i;
1815
  if (ktt->keyword[key->len] != '\0')
1816
    return -1;
1817
  return 0;
1818
}
1819
 
1820
} // End extern "C".
1821
 
1822
int
1823
Keyword_to_parsecode::keyword_to_parsecode(const char* keyword,
1824
                                           size_t len) const
1825
{
1826
  Ktt_key key;
1827
  key.str = keyword;
1828
  key.len = len;
1829
  void* kttv = bsearch(&key,
1830
                       this->keyword_parsecodes_,
1831
                       this->keyword_count_,
1832
                       sizeof(this->keyword_parsecodes_[0]),
1833
                       ktt_compare);
1834
  if (kttv == NULL)
1835
    return 0;
1836
  Keyword_parsecode* ktt = static_cast<Keyword_parsecode*>(kttv);
1837
  return ktt->parsecode;
1838
}
1839
 
1840
// The following structs are used within the VersionInfo class as well
1841
// as in the bison helper functions.  They store the information
1842
// parsed from the version script.
1843
 
1844
// A single version expression.
1845
// For example, pattern="std::map*" and language="C++".
1846
struct Version_expression
1847
{
1848
  Version_expression(const std::string& a_pattern,
1849
                     Version_script_info::Language a_language,
1850
                     bool a_exact_match)
1851
    : pattern(a_pattern), language(a_language), exact_match(a_exact_match),
1852
      was_matched_by_symbol(false)
1853
  { }
1854
 
1855
  std::string pattern;
1856
  Version_script_info::Language language;
1857
  // If false, we use glob() to match pattern.  If true, we use strcmp().
1858
  bool exact_match;
1859
  // True if --no-undefined-version is in effect and we found this
1860
  // version in get_symbol_version.  We use mutable because this
1861
  // struct is generally not modifiable after it has been created.
1862
  mutable bool was_matched_by_symbol;
1863
};
1864
 
1865
// A list of expressions.
1866
struct Version_expression_list
1867
{
1868
  std::vector<struct Version_expression> expressions;
1869
};
1870
 
1871
// A list of which versions upon which another version depends.
1872
// Strings should be from the Stringpool.
1873
struct Version_dependency_list
1874
{
1875
  std::vector<std::string> dependencies;
1876
};
1877
 
1878
// The total definition of a version.  It includes the tag for the
1879
// version, its global and local expressions, and any dependencies.
1880
struct Version_tree
1881
{
1882
  Version_tree()
1883
      : tag(), global(NULL), local(NULL), dependencies(NULL)
1884
  { }
1885
 
1886
  std::string tag;
1887
  const struct Version_expression_list* global;
1888
  const struct Version_expression_list* local;
1889
  const struct Version_dependency_list* dependencies;
1890
};
1891
 
1892
// Helper class that calls cplus_demangle when needed and takes care of freeing
1893
// the result.
1894
 
1895
class Lazy_demangler
1896
{
1897
 public:
1898
  Lazy_demangler(const char* symbol, int options)
1899
    : symbol_(symbol), options_(options), demangled_(NULL), did_demangle_(false)
1900
  { }
1901
 
1902
  ~Lazy_demangler()
1903
  { free(this->demangled_); }
1904
 
1905
  // Return the demangled name. The actual demangling happens on the first call,
1906
  // and the result is later cached.
1907
  inline char*
1908
  get();
1909
 
1910
 private:
1911
  // The symbol to demangle.
1912
  const char* symbol_;
1913
  // Option flags to pass to cplus_demagle.
1914
  const int options_;
1915
  // The cached demangled value, or NULL if demangling didn't happen yet or
1916
  // failed.
1917
  char* demangled_;
1918
  // Whether we already called cplus_demangle
1919
  bool did_demangle_;
1920
};
1921
 
1922
// Return the demangled name. The actual demangling happens on the first call,
1923
// and the result is later cached. Returns NULL if the symbol cannot be
1924
// demangled.
1925
 
1926
inline char*
1927
Lazy_demangler::get()
1928
{
1929
  if (!this->did_demangle_)
1930
    {
1931
      this->demangled_ = cplus_demangle(this->symbol_, this->options_);
1932
      this->did_demangle_ = true;
1933
    }
1934
  return this->demangled_;
1935
}
1936
 
1937
// Class Version_script_info.
1938
 
1939
Version_script_info::Version_script_info()
1940
  : dependency_lists_(), expression_lists_(), version_trees_(), globs_(),
1941
    default_version_(NULL), default_is_global_(false), is_finalized_(false)
1942
{
1943
  for (int i = 0; i < LANGUAGE_COUNT; ++i)
1944
    this->exact_[i] = NULL;
1945
}
1946
 
1947
Version_script_info::~Version_script_info()
1948
{
1949
}
1950
 
1951
// Forget all the known version script information.
1952
 
1953
void
1954
Version_script_info::clear()
1955
{
1956
  for (size_t k = 0; k < this->dependency_lists_.size(); ++k)
1957
    delete this->dependency_lists_[k];
1958
  this->dependency_lists_.clear();
1959
  for (size_t k = 0; k < this->version_trees_.size(); ++k)
1960
    delete this->version_trees_[k];
1961
  this->version_trees_.clear();
1962
  for (size_t k = 0; k < this->expression_lists_.size(); ++k)
1963
    delete this->expression_lists_[k];
1964
  this->expression_lists_.clear();
1965
}
1966
 
1967
// Finalize the version script information.
1968
 
1969
void
1970
Version_script_info::finalize()
1971
{
1972
  if (!this->is_finalized_)
1973
    {
1974
      this->build_lookup_tables();
1975
      this->is_finalized_ = true;
1976
    }
1977
}
1978
 
1979
// Return all the versions.
1980
 
1981
std::vector<std::string>
1982
Version_script_info::get_versions() const
1983
{
1984
  std::vector<std::string> ret;
1985
  for (size_t j = 0; j < this->version_trees_.size(); ++j)
1986
    if (!this->version_trees_[j]->tag.empty())
1987
      ret.push_back(this->version_trees_[j]->tag);
1988
  return ret;
1989
}
1990
 
1991
// Return the dependencies of VERSION.
1992
 
1993
std::vector<std::string>
1994
Version_script_info::get_dependencies(const char* version) const
1995
{
1996
  std::vector<std::string> ret;
1997
  for (size_t j = 0; j < this->version_trees_.size(); ++j)
1998
    if (this->version_trees_[j]->tag == version)
1999
      {
2000
        const struct Version_dependency_list* deps =
2001
          this->version_trees_[j]->dependencies;
2002
        if (deps != NULL)
2003
          for (size_t k = 0; k < deps->dependencies.size(); ++k)
2004
            ret.push_back(deps->dependencies[k]);
2005
        return ret;
2006
      }
2007
  return ret;
2008
}
2009
 
2010
// A version script essentially maps a symbol name to a version tag
2011
// and an indication of whether symbol is global or local within that
2012
// version tag.  Each symbol maps to at most one version tag.
2013
// Unfortunately, in practice, version scripts are ambiguous, and list
2014
// symbols multiple times.  Thus, we have to document the matching
2015
// process.
2016
 
2017
// This is a description of what the GNU linker does as of 2010-01-11.
2018
// It walks through the version tags in the order in which they appear
2019
// in the version script.  For each tag, it first walks through the
2020
// global patterns for that tag, then the local patterns.  When
2021
// looking at a single pattern, it first applies any language specific
2022
// demangling as specified for the pattern, and then matches the
2023
// resulting symbol name to the pattern.  If it finds an exact match
2024
// for a literal pattern (a pattern enclosed in quotes or with no
2025
// wildcard characters), then that is the match that it uses.  If
2026
// finds a match with a wildcard pattern, then it saves it and
2027
// continues searching.  Wildcard patterns that are exactly "*" are
2028
// saved separately.
2029
 
2030
// If no exact match with a literal pattern is ever found, then if a
2031
// wildcard match with a global pattern was found it is used,
2032
// otherwise if a wildcard match with a local pattern was found it is
2033
// used.
2034
 
2035
// This is the result:
2036
//   * If there is an exact match, then we use the first tag in the
2037
//     version script where it matches.
2038
//     + If the exact match in that tag is global, it is used.
2039
//     + Otherwise the exact match in that tag is local, and is used.
2040
//   * Otherwise, if there is any match with a global wildcard pattern:
2041
//     + If there is any match with a wildcard pattern which is not
2042
//       "*", then we use the tag in which the *last* such pattern
2043
//       appears.
2044
//     + Otherwise, we matched "*".  If there is no match with a local
2045
//       wildcard pattern which is not "*", then we use the *last*
2046
//       match with a global "*".  Otherwise, continue.
2047
//   * Otherwise, if there is any match with a local wildcard pattern:
2048
//     + If there is any match with a wildcard pattern which is not
2049
//       "*", then we use the tag in which the *last* such pattern
2050
//       appears.
2051
//     + Otherwise, we matched "*", and we use the tag in which the
2052
//       *last* such match occurred.
2053
 
2054
// There is an additional wrinkle.  When the GNU linker finds a symbol
2055
// with a version defined in an object file due to a .symver
2056
// directive, it looks up that symbol name in that version tag.  If it
2057
// finds it, it matches the symbol name against the patterns for that
2058
// version.  If there is no match with a global pattern, but there is
2059
// a match with a local pattern, then the GNU linker marks the symbol
2060
// as local.
2061
 
2062
// We want gold to be generally compatible, but we also want gold to
2063
// be fast.  These are the rules that gold implements:
2064
//   * If there is an exact match for the mangled name, we use it.
2065
//     + If there is more than one exact match, we give a warning, and
2066
//       we use the first tag in the script which matches.
2067
//     + If a symbol has an exact match as both global and local for
2068
//       the same version tag, we give an error.
2069
//   * Otherwise, we look for an extern C++ or an extern Java exact
2070
//     match.  If we find an exact match, we use it.
2071
//     + If there is more than one exact match, we give a warning, and
2072
//       we use the first tag in the script which matches.
2073
//     + If a symbol has an exact match as both global and local for
2074
//       the same version tag, we give an error.
2075
//   * Otherwise, we look through the wildcard patterns, ignoring "*"
2076
//     patterns.  We look through the version tags in reverse order.
2077
//     For each version tag, we look through the global patterns and
2078
//     then the local patterns.  We use the first match we find (i.e.,
2079
//     the last matching version tag in the file).
2080
//   * Otherwise, we use the "*" pattern if there is one.  We give an
2081
//     error if there are multiple "*" patterns.
2082
 
2083
// At least for now, gold does not look up the version tag for a
2084
// symbol version found in an object file to see if it should be
2085
// forced local.  There are other ways to force a symbol to be local,
2086
// and I don't understand why this one is useful.
2087
 
2088
// Build a set of fast lookup tables for a version script.
2089
 
2090
void
2091
Version_script_info::build_lookup_tables()
2092
{
2093
  size_t size = this->version_trees_.size();
2094
  for (size_t j = 0; j < size; ++j)
2095
    {
2096
      const Version_tree* v = this->version_trees_[j];
2097
      this->build_expression_list_lookup(v->local, v, false);
2098
      this->build_expression_list_lookup(v->global, v, true);
2099
    }
2100
}
2101
 
2102
// If a pattern has backlashes but no unquoted wildcard characters,
2103
// then we apply backslash unquoting and look for an exact match.
2104
// Otherwise we treat it as a wildcard pattern.  This function returns
2105
// true for a wildcard pattern.  Otherwise, it does backslash
2106
// unquoting on *PATTERN and returns false.  If this returns true,
2107
// *PATTERN may have been partially unquoted.
2108
 
2109
bool
2110
Version_script_info::unquote(std::string* pattern) const
2111
{
2112
  bool saw_backslash = false;
2113
  size_t len = pattern->length();
2114
  size_t j = 0;
2115
  for (size_t i = 0; i < len; ++i)
2116
    {
2117
      if (saw_backslash)
2118
        saw_backslash = false;
2119
      else
2120
        {
2121
          switch ((*pattern)[i])
2122
            {
2123
            case '?': case '[': case '*':
2124
              return true;
2125
            case '\\':
2126
              saw_backslash = true;
2127
              continue;
2128
            default:
2129
              break;
2130
            }
2131
        }
2132
 
2133
      if (i != j)
2134
        (*pattern)[j] = (*pattern)[i];
2135
      ++j;
2136
    }
2137
  return false;
2138
}
2139
 
2140
// Add an exact match for MATCH to *PE.  The result of the match is
2141
// V/IS_GLOBAL.
2142
 
2143
void
2144
Version_script_info::add_exact_match(const std::string& match,
2145
                                     const Version_tree* v, bool is_global,
2146
                                     const Version_expression* ve,
2147
                                     Exact* pe)
2148
{
2149
  std::pair<Exact::iterator, bool> ins =
2150
    pe->insert(std::make_pair(match, Version_tree_match(v, is_global, ve)));
2151
  if (ins.second)
2152
    {
2153
      // This is the first time we have seen this match.
2154
      return;
2155
    }
2156
 
2157
  Version_tree_match& vtm(ins.first->second);
2158
  if (vtm.real->tag != v->tag)
2159
    {
2160
      // This is an ambiguous match.  We still return the
2161
      // first version that we found in the script, but we
2162
      // record the new version to issue a warning if we
2163
      // wind up looking up this symbol.
2164
      if (vtm.ambiguous == NULL)
2165
        vtm.ambiguous = v;
2166
    }
2167
  else if (is_global != vtm.is_global)
2168
    {
2169
      // We have a match for both the global and local entries for a
2170
      // version tag.  That's got to be wrong.
2171
      gold_error(_("'%s' appears as both a global and a local symbol "
2172
                   "for version '%s' in script"),
2173
                 match.c_str(), v->tag.c_str());
2174
    }
2175
}
2176
 
2177
// Build fast lookup information for EXPLIST and store it in LOOKUP.
2178
// All matches go to V, and IS_GLOBAL is true if they are global
2179
// matches.
2180
 
2181
void
2182
Version_script_info::build_expression_list_lookup(
2183
    const Version_expression_list* explist,
2184
    const Version_tree* v,
2185
    bool is_global)
2186
{
2187
  if (explist == NULL)
2188
    return;
2189
  size_t size = explist->expressions.size();
2190
  for (size_t i = 0; i < size; ++i)
2191
    {
2192
      const Version_expression& exp(explist->expressions[i]);
2193
 
2194
      if (exp.pattern.length() == 1 && exp.pattern[0] == '*')
2195
        {
2196
          if (this->default_version_ != NULL
2197
              && this->default_version_->tag != v->tag)
2198
            gold_warning(_("wildcard match appears in both version '%s' "
2199
                           "and '%s' in script"),
2200
                         this->default_version_->tag.c_str(), v->tag.c_str());
2201
          else if (this->default_version_ != NULL
2202
                   && this->default_is_global_ != is_global)
2203
            gold_error(_("wildcard match appears as both global and local "
2204
                         "in version '%s' in script"),
2205
                       v->tag.c_str());
2206
          this->default_version_ = v;
2207
          this->default_is_global_ = is_global;
2208
          continue;
2209
        }
2210
 
2211
      std::string pattern = exp.pattern;
2212
      if (!exp.exact_match)
2213
        {
2214
          if (this->unquote(&pattern))
2215
            {
2216
              this->globs_.push_back(Glob(&exp, v, is_global));
2217
              continue;
2218
            }
2219
        }
2220
 
2221
      if (this->exact_[exp.language] == NULL)
2222
        this->exact_[exp.language] = new Exact();
2223
      this->add_exact_match(pattern, v, is_global, &exp,
2224
                            this->exact_[exp.language]);
2225
    }
2226
}
2227
 
2228
// Return the name to match given a name, a language code, and two
2229
// lazy demanglers.
2230
 
2231
const char*
2232
Version_script_info::get_name_to_match(const char* name,
2233
                                       int language,
2234
                                       Lazy_demangler* cpp_demangler,
2235
                                       Lazy_demangler* java_demangler) const
2236
{
2237
  switch (language)
2238
    {
2239
    case LANGUAGE_C:
2240
      return name;
2241
    case LANGUAGE_CXX:
2242
      return cpp_demangler->get();
2243
    case LANGUAGE_JAVA:
2244
      return java_demangler->get();
2245
    default:
2246
      gold_unreachable();
2247
    }
2248
}
2249
 
2250
// Look up SYMBOL_NAME in the list of versions.  Return true if the
2251
// symbol is found, false if not.  If the symbol is found, then if
2252
// PVERSION is not NULL, set *PVERSION to the version tag, and if
2253
// P_IS_GLOBAL is not NULL, set *P_IS_GLOBAL according to whether the
2254
// symbol is global or not.
2255
 
2256
bool
2257
Version_script_info::get_symbol_version(const char* symbol_name,
2258
                                        std::string* pversion,
2259
                                        bool* p_is_global) const
2260
{
2261
  Lazy_demangler cpp_demangled_name(symbol_name, DMGL_ANSI | DMGL_PARAMS);
2262
  Lazy_demangler java_demangled_name(symbol_name,
2263
                                     DMGL_ANSI | DMGL_PARAMS | DMGL_JAVA);
2264
 
2265
  gold_assert(this->is_finalized_);
2266
  for (int i = 0; i < LANGUAGE_COUNT; ++i)
2267
    {
2268
      Exact* exact = this->exact_[i];
2269
      if (exact == NULL)
2270
        continue;
2271
 
2272
      const char* name_to_match = this->get_name_to_match(symbol_name, i,
2273
                                                          &cpp_demangled_name,
2274
                                                          &java_demangled_name);
2275
      if (name_to_match == NULL)
2276
        {
2277
          // If the name can not be demangled, the GNU linker goes
2278
          // ahead and tries to match it anyhow.  That does not
2279
          // make sense to me and I have not implemented it.
2280
          continue;
2281
        }
2282
 
2283
      Exact::const_iterator pe = exact->find(name_to_match);
2284
      if (pe != exact->end())
2285
        {
2286
          const Version_tree_match& vtm(pe->second);
2287
          if (vtm.ambiguous != NULL)
2288
            gold_warning(_("using '%s' as version for '%s' which is also "
2289
                           "named in version '%s' in script"),
2290
                         vtm.real->tag.c_str(), name_to_match,
2291
                         vtm.ambiguous->tag.c_str());
2292
 
2293
          if (pversion != NULL)
2294
            *pversion = vtm.real->tag;
2295
          if (p_is_global != NULL)
2296
            *p_is_global = vtm.is_global;
2297
 
2298
          // If we are using --no-undefined-version, and this is a
2299
          // global symbol, we have to record that we have found this
2300
          // symbol, so that we don't warn about it.  We have to do
2301
          // this now, because otherwise we have no way to get from a
2302
          // non-C language back to the demangled name that we
2303
          // matched.
2304
          if (p_is_global != NULL && vtm.is_global)
2305
            vtm.expression->was_matched_by_symbol = true;
2306
 
2307
          return true;
2308
        }
2309
    }
2310
 
2311
  // Look through the glob patterns in reverse order.
2312
 
2313
  for (Globs::const_reverse_iterator p = this->globs_.rbegin();
2314
       p != this->globs_.rend();
2315
       ++p)
2316
    {
2317
      int language = p->expression->language;
2318
      const char* name_to_match = this->get_name_to_match(symbol_name,
2319
                                                          language,
2320
                                                          &cpp_demangled_name,
2321
                                                          &java_demangled_name);
2322
      if (name_to_match == NULL)
2323
        continue;
2324
 
2325
      if (fnmatch(p->expression->pattern.c_str(), name_to_match,
2326
                  FNM_NOESCAPE) == 0)
2327
        {
2328
          if (pversion != NULL)
2329
            *pversion = p->version->tag;
2330
          if (p_is_global != NULL)
2331
            *p_is_global = p->is_global;
2332
          return true;
2333
        }
2334
    }
2335
 
2336
  // Finally, there may be a wildcard.
2337
  if (this->default_version_ != NULL)
2338
    {
2339
      if (pversion != NULL)
2340
        *pversion = this->default_version_->tag;
2341
      if (p_is_global != NULL)
2342
        *p_is_global = this->default_is_global_;
2343
      return true;
2344
    }
2345
 
2346
  return false;
2347
}
2348
 
2349
// Give an error if any exact symbol names (not wildcards) appear in a
2350
// version script, but there is no such symbol.
2351
 
2352
void
2353
Version_script_info::check_unmatched_names(const Symbol_table* symtab) const
2354
{
2355
  for (size_t i = 0; i < this->version_trees_.size(); ++i)
2356
    {
2357
      const Version_tree* vt = this->version_trees_[i];
2358
      if (vt->global == NULL)
2359
        continue;
2360
      for (size_t j = 0; j < vt->global->expressions.size(); ++j)
2361
        {
2362
          const Version_expression& expression(vt->global->expressions[j]);
2363
 
2364
          // Ignore cases where we used the version because we saw a
2365
          // symbol that we looked up.  Note that
2366
          // WAS_MATCHED_BY_SYMBOL will be true even if the symbol was
2367
          // not a definition.  That's OK as in that case we most
2368
          // likely gave an undefined symbol error anyhow.
2369
          if (expression.was_matched_by_symbol)
2370
            continue;
2371
 
2372
          // Just ignore names which are in languages other than C.
2373
          // We have no way to look them up in the symbol table.
2374
          if (expression.language != LANGUAGE_C)
2375
            continue;
2376
 
2377
          // Remove backslash quoting, and ignore wildcard patterns.
2378
          std::string pattern = expression.pattern;
2379
          if (!expression.exact_match)
2380
            {
2381
              if (this->unquote(&pattern))
2382
                continue;
2383
            }
2384
 
2385
          if (symtab->lookup(pattern.c_str(), vt->tag.c_str()) == NULL)
2386
            gold_error(_("version script assignment of %s to symbol %s "
2387
                         "failed: symbol not defined"),
2388
                       vt->tag.c_str(), pattern.c_str());
2389
        }
2390
    }
2391
}
2392
 
2393
struct Version_dependency_list*
2394
Version_script_info::allocate_dependency_list()
2395
{
2396
  dependency_lists_.push_back(new Version_dependency_list);
2397
  return dependency_lists_.back();
2398
}
2399
 
2400
struct Version_expression_list*
2401
Version_script_info::allocate_expression_list()
2402
{
2403
  expression_lists_.push_back(new Version_expression_list);
2404
  return expression_lists_.back();
2405
}
2406
 
2407
struct Version_tree*
2408
Version_script_info::allocate_version_tree()
2409
{
2410
  version_trees_.push_back(new Version_tree);
2411
  return version_trees_.back();
2412
}
2413
 
2414
// Print for debugging.
2415
 
2416
void
2417
Version_script_info::print(FILE* f) const
2418
{
2419
  if (this->empty())
2420
    return;
2421
 
2422
  fprintf(f, "VERSION {");
2423
 
2424
  for (size_t i = 0; i < this->version_trees_.size(); ++i)
2425
    {
2426
      const Version_tree* vt = this->version_trees_[i];
2427
 
2428
      if (vt->tag.empty())
2429
        fprintf(f, "  {\n");
2430
      else
2431
        fprintf(f, "  %s {\n", vt->tag.c_str());
2432
 
2433
      if (vt->global != NULL)
2434
        {
2435
          fprintf(f, "    global :\n");
2436
          this->print_expression_list(f, vt->global);
2437
        }
2438
 
2439
      if (vt->local != NULL)
2440
        {
2441
          fprintf(f, "    local :\n");
2442
          this->print_expression_list(f, vt->local);
2443
        }
2444
 
2445
      fprintf(f, "  }");
2446
      if (vt->dependencies != NULL)
2447
        {
2448
          const Version_dependency_list* deps = vt->dependencies;
2449
          for (size_t j = 0; j < deps->dependencies.size(); ++j)
2450
            {
2451
              if (j < deps->dependencies.size() - 1)
2452
                fprintf(f, "\n");
2453
              fprintf(f, "    %s", deps->dependencies[j].c_str());
2454
            }
2455
        }
2456
      fprintf(f, ";\n");
2457
    }
2458
 
2459
  fprintf(f, "}\n");
2460
}
2461
 
2462
void
2463
Version_script_info::print_expression_list(
2464
    FILE* f,
2465
    const Version_expression_list* vel) const
2466
{
2467
  Version_script_info::Language current_language = LANGUAGE_C;
2468
  for (size_t i = 0; i < vel->expressions.size(); ++i)
2469
    {
2470
      const Version_expression& ve(vel->expressions[i]);
2471
 
2472
      if (ve.language != current_language)
2473
        {
2474
          if (current_language != LANGUAGE_C)
2475
            fprintf(f, "      }\n");
2476
          switch (ve.language)
2477
            {
2478
            case LANGUAGE_C:
2479
              break;
2480
            case LANGUAGE_CXX:
2481
              fprintf(f, "      extern \"C++\" {\n");
2482
              break;
2483
            case LANGUAGE_JAVA:
2484
              fprintf(f, "      extern \"Java\" {\n");
2485
              break;
2486
            default:
2487
              gold_unreachable();
2488
            }
2489
          current_language = ve.language;
2490
        }
2491
 
2492
      fprintf(f, "      ");
2493
      if (current_language != LANGUAGE_C)
2494
        fprintf(f, "  ");
2495
 
2496
      if (ve.exact_match)
2497
        fprintf(f, "\"");
2498
      fprintf(f, "%s", ve.pattern.c_str());
2499
      if (ve.exact_match)
2500
        fprintf(f, "\"");
2501
 
2502
      fprintf(f, "\n");
2503
    }
2504
 
2505
  if (current_language != LANGUAGE_C)
2506
    fprintf(f, "      }\n");
2507
}
2508
 
2509
} // End namespace gold.
2510
 
2511
// The remaining functions are extern "C", so it's clearer to not put
2512
// them in namespace gold.
2513
 
2514
using namespace gold;
2515
 
2516
// This function is called by the bison parser to return the next
2517
// token.
2518
 
2519
extern "C" int
2520
yylex(YYSTYPE* lvalp, void* closurev)
2521
{
2522
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2523
  const Token* token = closure->next_token();
2524
  switch (token->classification())
2525
    {
2526
    default:
2527
      gold_unreachable();
2528
 
2529
    case Token::TOKEN_INVALID:
2530
      yyerror(closurev, "invalid character");
2531
      return 0;
2532
 
2533
    case Token::TOKEN_EOF:
2534
      return 0;
2535
 
2536
    case Token::TOKEN_STRING:
2537
      {
2538
        // This is either a keyword or a STRING.
2539
        size_t len;
2540
        const char* str = token->string_value(&len);
2541
        int parsecode = 0;
2542
        switch (closure->lex_mode())
2543
          {
2544
          case Lex::LINKER_SCRIPT:
2545
            parsecode = script_keywords.keyword_to_parsecode(str, len);
2546
            break;
2547
          case Lex::VERSION_SCRIPT:
2548
            parsecode = version_script_keywords.keyword_to_parsecode(str, len);
2549
            break;
2550
          case Lex::DYNAMIC_LIST:
2551
            parsecode = dynamic_list_keywords.keyword_to_parsecode(str, len);
2552
            break;
2553
          default:
2554
            break;
2555
          }
2556
        if (parsecode != 0)
2557
          return parsecode;
2558
        lvalp->string.value = str;
2559
        lvalp->string.length = len;
2560
        return STRING;
2561
      }
2562
 
2563
    case Token::TOKEN_QUOTED_STRING:
2564
      lvalp->string.value = token->string_value(&lvalp->string.length);
2565
      return QUOTED_STRING;
2566
 
2567
    case Token::TOKEN_OPERATOR:
2568
      return token->operator_value();
2569
 
2570
    case Token::TOKEN_INTEGER:
2571
      lvalp->integer = token->integer_value();
2572
      return INTEGER;
2573
    }
2574
}
2575
 
2576
// This function is called by the bison parser to report an error.
2577
 
2578
extern "C" void
2579
yyerror(void* closurev, const char* message)
2580
{
2581
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2582
  gold_error(_("%s:%d:%d: %s"), closure->filename(), closure->lineno(),
2583
             closure->charpos(), message);
2584
}
2585
 
2586
// Called by the bison parser to add an external symbol to the link.
2587
 
2588
extern "C" void
2589
script_add_extern(void* closurev, const char* name, size_t length)
2590
{
2591
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2592
  closure->script_options()->add_symbol_reference(name, length);
2593
}
2594
 
2595
// Called by the bison parser to add a file to the link.
2596
 
2597
extern "C" void
2598
script_add_file(void* closurev, const char* name, size_t length)
2599
{
2600
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2601
 
2602
  // If this is an absolute path, and we found the script in the
2603
  // sysroot, then we want to prepend the sysroot to the file name.
2604
  // For example, this is how we handle a cross link to the x86_64
2605
  // libc.so, which refers to /lib/libc.so.6.
2606
  std::string name_string(name, length);
2607
  const char* extra_search_path = ".";
2608
  std::string script_directory;
2609
  if (IS_ABSOLUTE_PATH(name_string.c_str()))
2610
    {
2611
      if (closure->is_in_sysroot())
2612
        {
2613
          const std::string& sysroot(parameters->options().sysroot());
2614
          gold_assert(!sysroot.empty());
2615
          name_string = sysroot + name_string;
2616
        }
2617
    }
2618
  else
2619
    {
2620
      // In addition to checking the normal library search path, we
2621
      // also want to check in the script-directory.
2622
      const char* slash = strrchr(closure->filename(), '/');
2623
      if (slash != NULL)
2624
        {
2625
          script_directory.assign(closure->filename(),
2626
                                  slash - closure->filename() + 1);
2627
          extra_search_path = script_directory.c_str();
2628
        }
2629
    }
2630
 
2631
  Input_file_argument file(name_string.c_str(),
2632
                           Input_file_argument::INPUT_FILE_TYPE_FILE,
2633
                           extra_search_path, false,
2634
                           closure->position_dependent_options());
2635
  Input_argument& arg = closure->inputs()->add_file(file);
2636
  arg.set_script_info(closure->script_info());
2637
}
2638
 
2639
// Called by the bison parser to add a library to the link.
2640
 
2641
extern "C" void
2642
script_add_library(void* closurev, const char* name, size_t length)
2643
{
2644
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2645
  std::string name_string(name, length);
2646
 
2647
  if (name_string[0] != 'l')
2648
    gold_error(_("library name must be prefixed with -l"));
2649
 
2650
  Input_file_argument file(name_string.c_str() + 1,
2651
                           Input_file_argument::INPUT_FILE_TYPE_LIBRARY,
2652
                           "", false,
2653
                           closure->position_dependent_options());
2654
  Input_argument& arg = closure->inputs()->add_file(file);
2655
  arg.set_script_info(closure->script_info());
2656
}
2657
 
2658
// Called by the bison parser to start a group.  If we are already in
2659
// a group, that means that this script was invoked within a
2660
// --start-group --end-group sequence on the command line, or that
2661
// this script was found in a GROUP of another script.  In that case,
2662
// we simply continue the existing group, rather than starting a new
2663
// one.  It is possible to construct a case in which this will do
2664
// something other than what would happen if we did a recursive group,
2665
// but it's hard to imagine why the different behaviour would be
2666
// useful for a real program.  Avoiding recursive groups is simpler
2667
// and more efficient.
2668
 
2669
extern "C" void
2670
script_start_group(void* closurev)
2671
{
2672
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2673
  if (!closure->in_group())
2674
    closure->inputs()->start_group();
2675
}
2676
 
2677
// Called by the bison parser at the end of a group.
2678
 
2679
extern "C" void
2680
script_end_group(void* closurev)
2681
{
2682
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2683
  if (!closure->in_group())
2684
    closure->inputs()->end_group();
2685
}
2686
 
2687
// Called by the bison parser to start an AS_NEEDED list.
2688
 
2689
extern "C" void
2690
script_start_as_needed(void* closurev)
2691
{
2692
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2693
  closure->position_dependent_options().set_as_needed(true);
2694
}
2695
 
2696
// Called by the bison parser at the end of an AS_NEEDED list.
2697
 
2698
extern "C" void
2699
script_end_as_needed(void* closurev)
2700
{
2701
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2702
  closure->position_dependent_options().set_as_needed(false);
2703
}
2704
 
2705
// Called by the bison parser to set the entry symbol.
2706
 
2707
extern "C" void
2708
script_set_entry(void* closurev, const char* entry, size_t length)
2709
{
2710
  // We'll parse this exactly the same as --entry=ENTRY on the commandline
2711
  // TODO(csilvers): FIXME -- call set_entry directly.
2712
  std::string arg("--entry=");
2713
  arg.append(entry, length);
2714
  script_parse_option(closurev, arg.c_str(), arg.size());
2715
}
2716
 
2717
// Called by the bison parser to set whether to define common symbols.
2718
 
2719
extern "C" void
2720
script_set_common_allocation(void* closurev, int set)
2721
{
2722
  const char* arg = set != 0 ? "--define-common" : "--no-define-common";
2723
  script_parse_option(closurev, arg, strlen(arg));
2724
}
2725
 
2726
// Called by the bison parser to refer to a symbol.
2727
 
2728
extern "C" Expression*
2729
script_symbol(void* closurev, const char* name, size_t length)
2730
{
2731
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2732
  if (length != 1 || name[0] != '.')
2733
    closure->script_options()->add_symbol_reference(name, length);
2734
  return script_exp_string(name, length);
2735
}
2736
 
2737
// Called by the bison parser to define a symbol.
2738
 
2739
extern "C" void
2740
script_set_symbol(void* closurev, const char* name, size_t length,
2741
                  Expression* value, int providei, int hiddeni)
2742
{
2743
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2744
  const bool provide = providei != 0;
2745
  const bool hidden = hiddeni != 0;
2746
  closure->script_options()->add_symbol_assignment(name, length,
2747
                                                   closure->parsing_defsym(),
2748
                                                   value, provide, hidden);
2749
  closure->clear_skip_on_incompatible_target();
2750
}
2751
 
2752
// Called by the bison parser to add an assertion.
2753
 
2754
extern "C" void
2755
script_add_assertion(void* closurev, Expression* check, const char* message,
2756
                     size_t messagelen)
2757
{
2758
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2759
  closure->script_options()->add_assertion(check, message, messagelen);
2760
  closure->clear_skip_on_incompatible_target();
2761
}
2762
 
2763
// Called by the bison parser to parse an OPTION.
2764
 
2765
extern "C" void
2766
script_parse_option(void* closurev, const char* option, size_t length)
2767
{
2768
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2769
  // We treat the option as a single command-line option, even if
2770
  // it has internal whitespace.
2771
  if (closure->command_line() == NULL)
2772
    {
2773
      // There are some options that we could handle here--e.g.,
2774
      // -lLIBRARY.  Should we bother?
2775
      gold_warning(_("%s:%d:%d: ignoring command OPTION; OPTION is only valid"
2776
                     " for scripts specified via -T/--script"),
2777
                   closure->filename(), closure->lineno(), closure->charpos());
2778
    }
2779
  else
2780
    {
2781
      bool past_a_double_dash_option = false;
2782
      const char* mutable_option = strndup(option, length);
2783
      gold_assert(mutable_option != NULL);
2784
      closure->command_line()->process_one_option(1, &mutable_option, 0,
2785
                                                  &past_a_double_dash_option);
2786
      // The General_options class will quite possibly store a pointer
2787
      // into mutable_option, so we can't free it.  In cases the class
2788
      // does not store such a pointer, this is a memory leak.  Alas. :(
2789
    }
2790
  closure->clear_skip_on_incompatible_target();
2791
}
2792
 
2793
// Called by the bison parser to handle OUTPUT_FORMAT.  OUTPUT_FORMAT
2794
// takes either one or three arguments.  In the three argument case,
2795
// the format depends on the endianness option, which we don't
2796
// currently support (FIXME).  If we see an OUTPUT_FORMAT for the
2797
// wrong format, then we want to search for a new file.  Returning 0
2798
// here will cause the parser to immediately abort.
2799
 
2800
extern "C" int
2801
script_check_output_format(void* closurev,
2802
                           const char* default_name, size_t default_length,
2803
                           const char*, size_t, const char*, size_t)
2804
{
2805
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2806
  std::string name(default_name, default_length);
2807
  Target* target = select_target_by_name(name.c_str());
2808
  if (target == NULL || !parameters->is_compatible_target(target))
2809
    {
2810
      if (closure->skip_on_incompatible_target())
2811
        {
2812
          closure->set_found_incompatible_target();
2813
          return 0;
2814
        }
2815
      // FIXME: Should we warn about the unknown target?
2816
    }
2817
  return 1;
2818
}
2819
 
2820
// Called by the bison parser to handle TARGET.
2821
 
2822
extern "C" void
2823
script_set_target(void* closurev, const char* target, size_t len)
2824
{
2825
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2826
  std::string s(target, len);
2827
  General_options::Object_format format_enum;
2828
  format_enum = General_options::string_to_object_format(s.c_str());
2829
  closure->position_dependent_options().set_format_enum(format_enum);
2830
}
2831
 
2832
// Called by the bison parser to handle SEARCH_DIR.  This is handled
2833
// exactly like a -L option.
2834
 
2835
extern "C" void
2836
script_add_search_dir(void* closurev, const char* option, size_t length)
2837
{
2838
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2839
  if (closure->command_line() == NULL)
2840
    gold_warning(_("%s:%d:%d: ignoring SEARCH_DIR; SEARCH_DIR is only valid"
2841
                   " for scripts specified via -T/--script"),
2842
                 closure->filename(), closure->lineno(), closure->charpos());
2843
  else if (!closure->command_line()->options().nostdlib())
2844
    {
2845
      std::string s = "-L" + std::string(option, length);
2846
      script_parse_option(closurev, s.c_str(), s.size());
2847
    }
2848
}
2849
 
2850
/* Called by the bison parser to push the lexer into expression
2851
   mode.  */
2852
 
2853
extern "C" void
2854
script_push_lex_into_expression_mode(void* closurev)
2855
{
2856
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2857
  closure->push_lex_mode(Lex::EXPRESSION);
2858
}
2859
 
2860
/* Called by the bison parser to push the lexer into version
2861
   mode.  */
2862
 
2863
extern "C" void
2864
script_push_lex_into_version_mode(void* closurev)
2865
{
2866
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2867
  if (closure->version_script()->is_finalized())
2868
    gold_error(_("%s:%d:%d: invalid use of VERSION in input file"),
2869
               closure->filename(), closure->lineno(), closure->charpos());
2870
  closure->push_lex_mode(Lex::VERSION_SCRIPT);
2871
}
2872
 
2873
/* Called by the bison parser to pop the lexer mode.  */
2874
 
2875
extern "C" void
2876
script_pop_lex_mode(void* closurev)
2877
{
2878
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2879
  closure->pop_lex_mode();
2880
}
2881
 
2882
// Register an entire version node. For example:
2883
//
2884
// GLIBC_2.1 {
2885
//   global: foo;
2886
// } GLIBC_2.0;
2887
//
2888
// - tag is "GLIBC_2.1"
2889
// - tree contains the information "global: foo"
2890
// - deps contains "GLIBC_2.0"
2891
 
2892
extern "C" void
2893
script_register_vers_node(void*,
2894
                          const char* tag,
2895
                          int taglen,
2896
                          struct Version_tree* tree,
2897
                          struct Version_dependency_list* deps)
2898
{
2899
  gold_assert(tree != NULL);
2900
  tree->dependencies = deps;
2901
  if (tag != NULL)
2902
    tree->tag = std::string(tag, taglen);
2903
}
2904
 
2905
// Add a dependencies to the list of existing dependencies, if any,
2906
// and return the expanded list.
2907
 
2908
extern "C" struct Version_dependency_list*
2909
script_add_vers_depend(void* closurev,
2910
                       struct Version_dependency_list* all_deps,
2911
                       const char* depend_to_add, int deplen)
2912
{
2913
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2914
  if (all_deps == NULL)
2915
    all_deps = closure->version_script()->allocate_dependency_list();
2916
  all_deps->dependencies.push_back(std::string(depend_to_add, deplen));
2917
  return all_deps;
2918
}
2919
 
2920
// Add a pattern expression to an existing list of expressions, if any.
2921
 
2922
extern "C" struct Version_expression_list*
2923
script_new_vers_pattern(void* closurev,
2924
                        struct Version_expression_list* expressions,
2925
                        const char* pattern, int patlen, int exact_match)
2926
{
2927
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2928
  if (expressions == NULL)
2929
    expressions = closure->version_script()->allocate_expression_list();
2930
  expressions->expressions.push_back(
2931
      Version_expression(std::string(pattern, patlen),
2932
                         closure->get_current_language(),
2933
                         static_cast<bool>(exact_match)));
2934
  return expressions;
2935
}
2936
 
2937
// Attaches b to the end of a, and clears b.  So a = a + b and b = {}.
2938
 
2939
extern "C" struct Version_expression_list*
2940
script_merge_expressions(struct Version_expression_list* a,
2941
                         struct Version_expression_list* b)
2942
{
2943
  a->expressions.insert(a->expressions.end(),
2944
                        b->expressions.begin(), b->expressions.end());
2945
  // We could delete b and remove it from expressions_lists_, but
2946
  // that's a lot of work.  This works just as well.
2947
  b->expressions.clear();
2948
  return a;
2949
}
2950
 
2951
// Combine the global and local expressions into a a Version_tree.
2952
 
2953
extern "C" struct Version_tree*
2954
script_new_vers_node(void* closurev,
2955
                     struct Version_expression_list* global,
2956
                     struct Version_expression_list* local)
2957
{
2958
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2959
  Version_tree* tree = closure->version_script()->allocate_version_tree();
2960
  tree->global = global;
2961
  tree->local = local;
2962
  return tree;
2963
}
2964
 
2965
// Handle a transition in language, such as at the
2966
// start or end of 'extern "C++"'
2967
 
2968
extern "C" void
2969
version_script_push_lang(void* closurev, const char* lang, int langlen)
2970
{
2971
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2972
  std::string language(lang, langlen);
2973
  Version_script_info::Language code;
2974
  if (language.empty() || language == "C")
2975
    code = Version_script_info::LANGUAGE_C;
2976
  else if (language == "C++")
2977
    code = Version_script_info::LANGUAGE_CXX;
2978
  else if (language == "Java")
2979
    code = Version_script_info::LANGUAGE_JAVA;
2980
  else
2981
    {
2982
      char* buf = new char[langlen + 100];
2983
      snprintf(buf, langlen + 100,
2984
               _("unrecognized version script language '%s'"),
2985
               language.c_str());
2986
      yyerror(closurev, buf);
2987
      delete[] buf;
2988
      code = Version_script_info::LANGUAGE_C;
2989
    }
2990
  closure->push_language(code);
2991
}
2992
 
2993
extern "C" void
2994
version_script_pop_lang(void* closurev)
2995
{
2996
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2997
  closure->pop_language();
2998
}
2999
 
3000
// Called by the bison parser to start a SECTIONS clause.
3001
 
3002
extern "C" void
3003
script_start_sections(void* closurev)
3004
{
3005
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3006
  closure->script_options()->script_sections()->start_sections();
3007
  closure->clear_skip_on_incompatible_target();
3008
}
3009
 
3010
// Called by the bison parser to finish a SECTIONS clause.
3011
 
3012
extern "C" void
3013
script_finish_sections(void* closurev)
3014
{
3015
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3016
  closure->script_options()->script_sections()->finish_sections();
3017
}
3018
 
3019
// Start processing entries for an output section.
3020
 
3021
extern "C" void
3022
script_start_output_section(void* closurev, const char* name, size_t namelen,
3023
                            const struct Parser_output_section_header* header)
3024
{
3025
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3026
  closure->script_options()->script_sections()->start_output_section(name,
3027
                                                                     namelen,
3028
                                                                     header);
3029
}
3030
 
3031
// Finish processing entries for an output section.
3032
 
3033
extern "C" void
3034
script_finish_output_section(void* closurev,
3035
                             const struct Parser_output_section_trailer* trail)
3036
{
3037
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3038
  closure->script_options()->script_sections()->finish_output_section(trail);
3039
}
3040
 
3041
// Add a data item (e.g., "WORD (0)") to the current output section.
3042
 
3043
extern "C" void
3044
script_add_data(void* closurev, int data_token, Expression* val)
3045
{
3046
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3047
  int size;
3048
  bool is_signed = true;
3049
  switch (data_token)
3050
    {
3051
    case QUAD:
3052
      size = 8;
3053
      is_signed = false;
3054
      break;
3055
    case SQUAD:
3056
      size = 8;
3057
      break;
3058
    case LONG:
3059
      size = 4;
3060
      break;
3061
    case SHORT:
3062
      size = 2;
3063
      break;
3064
    case BYTE:
3065
      size = 1;
3066
      break;
3067
    default:
3068
      gold_unreachable();
3069
    }
3070
  closure->script_options()->script_sections()->add_data(size, is_signed, val);
3071
}
3072
 
3073
// Add a clause setting the fill value to the current output section.
3074
 
3075
extern "C" void
3076
script_add_fill(void* closurev, Expression* val)
3077
{
3078
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3079
  closure->script_options()->script_sections()->add_fill(val);
3080
}
3081
 
3082
// Add a new input section specification to the current output
3083
// section.
3084
 
3085
extern "C" void
3086
script_add_input_section(void* closurev,
3087
                         const struct Input_section_spec* spec,
3088
                         int keepi)
3089
{
3090
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3091
  bool keep = keepi != 0;
3092
  closure->script_options()->script_sections()->add_input_section(spec, keep);
3093
}
3094
 
3095
// When we see DATA_SEGMENT_ALIGN we record that following output
3096
// sections may be relro.
3097
 
3098
extern "C" void
3099
script_data_segment_align(void* closurev)
3100
{
3101
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3102
  if (!closure->script_options()->saw_sections_clause())
3103
    gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
3104
               closure->filename(), closure->lineno(), closure->charpos());
3105
  else
3106
    closure->script_options()->script_sections()->data_segment_align();
3107
}
3108
 
3109
// When we see DATA_SEGMENT_RELRO_END we know that all output sections
3110
// since DATA_SEGMENT_ALIGN should be relro.
3111
 
3112
extern "C" void
3113
script_data_segment_relro_end(void* closurev)
3114
{
3115
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3116
  if (!closure->script_options()->saw_sections_clause())
3117
    gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
3118
               closure->filename(), closure->lineno(), closure->charpos());
3119
  else
3120
    closure->script_options()->script_sections()->data_segment_relro_end();
3121
}
3122
 
3123
// Create a new list of string/sort pairs.
3124
 
3125
extern "C" String_sort_list_ptr
3126
script_new_string_sort_list(const struct Wildcard_section* string_sort)
3127
{
3128
  return new String_sort_list(1, *string_sort);
3129
}
3130
 
3131
// Add an entry to a list of string/sort pairs.  The way the parser
3132
// works permits us to simply modify the first parameter, rather than
3133
// copy the vector.
3134
 
3135
extern "C" String_sort_list_ptr
3136
script_string_sort_list_add(String_sort_list_ptr pv,
3137
                            const struct Wildcard_section* string_sort)
3138
{
3139
  if (pv == NULL)
3140
    return script_new_string_sort_list(string_sort);
3141
  else
3142
    {
3143
      pv->push_back(*string_sort);
3144
      return pv;
3145
    }
3146
}
3147
 
3148
// Create a new list of strings.
3149
 
3150
extern "C" String_list_ptr
3151
script_new_string_list(const char* str, size_t len)
3152
{
3153
  return new String_list(1, std::string(str, len));
3154
}
3155
 
3156
// Add an element to a list of strings.  The way the parser works
3157
// permits us to simply modify the first parameter, rather than copy
3158
// the vector.
3159
 
3160
extern "C" String_list_ptr
3161
script_string_list_push_back(String_list_ptr pv, const char* str, size_t len)
3162
{
3163
  if (pv == NULL)
3164
    return script_new_string_list(str, len);
3165
  else
3166
    {
3167
      pv->push_back(std::string(str, len));
3168
      return pv;
3169
    }
3170
}
3171
 
3172
// Concatenate two string lists.  Either or both may be NULL.  The way
3173
// the parser works permits us to modify the parameters, rather than
3174
// copy the vector.
3175
 
3176
extern "C" String_list_ptr
3177
script_string_list_append(String_list_ptr pv1, String_list_ptr pv2)
3178
{
3179
  if (pv1 == NULL)
3180
    return pv2;
3181
  if (pv2 == NULL)
3182
    return pv1;
3183
  pv1->insert(pv1->end(), pv2->begin(), pv2->end());
3184
  return pv1;
3185
}
3186
 
3187
// Add a new program header.
3188
 
3189
extern "C" void
3190
script_add_phdr(void* closurev, const char* name, size_t namelen,
3191
                unsigned int type, const Phdr_info* info)
3192
{
3193
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3194
  bool includes_filehdr = info->includes_filehdr != 0;
3195
  bool includes_phdrs = info->includes_phdrs != 0;
3196
  bool is_flags_valid = info->is_flags_valid != 0;
3197
  Script_sections* ss = closure->script_options()->script_sections();
3198
  ss->add_phdr(name, namelen, type, includes_filehdr, includes_phdrs,
3199
               is_flags_valid, info->flags, info->load_address);
3200
  closure->clear_skip_on_incompatible_target();
3201
}
3202
 
3203
// Convert a program header string to a type.
3204
 
3205
#define PHDR_TYPE(NAME) { #NAME, sizeof(#NAME) - 1, elfcpp::NAME }
3206
 
3207
static struct
3208
{
3209
  const char* name;
3210
  size_t namelen;
3211
  unsigned int val;
3212
} phdr_type_names[] =
3213
{
3214
  PHDR_TYPE(PT_NULL),
3215
  PHDR_TYPE(PT_LOAD),
3216
  PHDR_TYPE(PT_DYNAMIC),
3217
  PHDR_TYPE(PT_INTERP),
3218
  PHDR_TYPE(PT_NOTE),
3219
  PHDR_TYPE(PT_SHLIB),
3220
  PHDR_TYPE(PT_PHDR),
3221
  PHDR_TYPE(PT_TLS),
3222
  PHDR_TYPE(PT_GNU_EH_FRAME),
3223
  PHDR_TYPE(PT_GNU_STACK),
3224
  PHDR_TYPE(PT_GNU_RELRO)
3225
};
3226
 
3227
extern "C" unsigned int
3228
script_phdr_string_to_type(void* closurev, const char* name, size_t namelen)
3229
{
3230
  for (unsigned int i = 0;
3231
       i < sizeof(phdr_type_names) / sizeof(phdr_type_names[0]);
3232
       ++i)
3233
    if (namelen == phdr_type_names[i].namelen
3234
        && strncmp(name, phdr_type_names[i].name, namelen) == 0)
3235
      return phdr_type_names[i].val;
3236
  yyerror(closurev, _("unknown PHDR type (try integer)"));
3237
  return elfcpp::PT_NULL;
3238
}
3239
 
3240
extern "C" void
3241
script_saw_segment_start_expression(void* closurev)
3242
{
3243
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3244
  Script_sections* ss = closure->script_options()->script_sections();
3245
  ss->set_saw_segment_start_expression(true);
3246
}
3247
 
3248
extern "C" void
3249
script_set_section_region(void* closurev, const char* name, size_t namelen,
3250
                          int set_vma)
3251
{
3252
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3253
  if (!closure->script_options()->saw_sections_clause())
3254
    {
3255
      gold_error(_("%s:%d:%d: MEMORY region '%.*s' referred to outside of "
3256
                   "SECTIONS clause"),
3257
                 closure->filename(), closure->lineno(), closure->charpos(),
3258
                 static_cast<int>(namelen), name);
3259
      return;
3260
    }
3261
 
3262
  Script_sections* ss = closure->script_options()->script_sections();
3263
  Memory_region* mr = ss->find_memory_region(name, namelen);
3264
  if (mr == NULL)
3265
    {
3266
      gold_error(_("%s:%d:%d: MEMORY region '%.*s' not declared"),
3267
                 closure->filename(), closure->lineno(), closure->charpos(),
3268
                 static_cast<int>(namelen), name);
3269
      return;
3270
    }
3271
 
3272
  ss->set_memory_region(mr, set_vma);
3273
}
3274
 
3275
extern "C" void
3276
script_add_memory(void* closurev, const char* name, size_t namelen,
3277
                  unsigned int attrs, Expression* origin, Expression* length)
3278
{
3279
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3280
  Script_sections* ss = closure->script_options()->script_sections();
3281
  ss->add_memory_region(name, namelen, attrs, origin, length);
3282
}
3283
 
3284
extern "C" unsigned int
3285
script_parse_memory_attr(void* closurev, const char* attrs, size_t attrlen,
3286
                         int invert)
3287
{
3288
  int attributes = 0;
3289
 
3290
  while (attrlen--)
3291
    switch (*attrs++)
3292
      {
3293
      case 'R':
3294
      case 'r':
3295
        attributes |= MEM_READABLE; break;
3296
      case 'W':
3297
      case 'w':
3298
        attributes |= MEM_READABLE | MEM_WRITEABLE; break;
3299
      case 'X':
3300
      case 'x':
3301
        attributes |= MEM_EXECUTABLE; break;
3302
      case 'A':
3303
      case 'a':
3304
        attributes |= MEM_ALLOCATABLE; break;
3305
      case 'I':
3306
      case 'i':
3307
      case 'L':
3308
      case 'l':
3309
        attributes |= MEM_INITIALIZED; break;
3310
      default:
3311
        yyerror(closurev, _("unknown MEMORY attribute"));
3312
      }
3313
 
3314
  if (invert)
3315
    attributes = (~ attributes) & MEM_ATTR_MASK;
3316
 
3317
  return attributes;
3318
}
3319
 
3320
extern "C" void
3321
script_include_directive(void* closurev, const char*, size_t)
3322
{
3323
  // FIXME: Implement ?
3324
  yyerror (closurev, _("GOLD does not currently support INCLUDE directives"));
3325
}
3326
 
3327
// Functions for memory regions.
3328
 
3329
extern "C" Expression*
3330
script_exp_function_origin(void* closurev, const char* name, size_t namelen)
3331
{
3332
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3333
  Script_sections* ss = closure->script_options()->script_sections();
3334
  Expression* origin = ss->find_memory_region_origin(name, namelen);
3335
 
3336
  if (origin == NULL)
3337
    {
3338
      gold_error(_("undefined memory region '%s' referenced "
3339
                   "in ORIGIN expression"),
3340
                 name);
3341
      // Create a dummy expression to prevent crashes later on.
3342
      origin = script_exp_integer(0);
3343
    }
3344
 
3345
  return origin;
3346
}
3347
 
3348
extern "C" Expression*
3349
script_exp_function_length(void* closurev, const char* name, size_t namelen)
3350
{
3351
  Parser_closure* closure = static_cast<Parser_closure*>(closurev);
3352
  Script_sections* ss = closure->script_options()->script_sections();
3353
  Expression* length = ss->find_memory_region_length(name, namelen);
3354
 
3355
  if (length == NULL)
3356
    {
3357
      gold_error(_("undefined memory region '%s' referenced "
3358
                   "in LENGTH expression"),
3359
                 name);
3360
      // Create a dummy expression to prevent crashes later on.
3361
      length = script_exp_integer(0);
3362
    }
3363
 
3364
  return length;
3365
}

powered by: WebSVN 2.1.0

© copyright 1999-2024 OpenCores.org, equivalent to Oliscience, all rights reserved. OpenCores®, registered trademark.