1 // script.cc -- handle linker scripts for gold.
3 // Copyright 2006, 2007, 2008 Free Software Foundation, Inc.
4 // Written by Ian Lance Taylor <iant@google.com>.
6 // This file is part of gold.
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.
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.
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.
30 #include "filenames.h"
34 #include "dirsearch.h"
37 #include "workqueue.h"
39 #include "parameters.h"
48 // A token read from a script file. We don't implement keywords here;
49 // all keywords are simply represented as a string.
54 // Token classification.
59 // Token indicates end of input.
61 // Token is a string of characters.
63 // Token is a quoted string of characters.
65 // Token is an operator.
67 // Token is a number (an integer).
71 // We need an empty constructor so that we can put this STL objects.
73 : classification_(TOKEN_INVALID
), value_(NULL
), value_length_(0),
74 opcode_(0), lineno_(0), charpos_(0)
77 // A general token with no value.
78 Token(Classification classification
, int lineno
, int charpos
)
79 : classification_(classification
), value_(NULL
), value_length_(0),
80 opcode_(0), lineno_(lineno
), charpos_(charpos
)
82 gold_assert(classification
== TOKEN_INVALID
83 || classification
== TOKEN_EOF
);
86 // A general token with a value.
87 Token(Classification classification
, const char* value
, size_t length
,
88 int lineno
, int charpos
)
89 : classification_(classification
), value_(value
), value_length_(length
),
90 opcode_(0), lineno_(lineno
), charpos_(charpos
)
92 gold_assert(classification
!= TOKEN_INVALID
93 && classification
!= TOKEN_EOF
);
96 // A token representing an operator.
97 Token(int opcode
, int lineno
, int charpos
)
98 : classification_(TOKEN_OPERATOR
), value_(NULL
), value_length_(0),
99 opcode_(opcode
), lineno_(lineno
), charpos_(charpos
)
102 // Return whether the token is invalid.
105 { return this->classification_
== TOKEN_INVALID
; }
107 // Return whether this is an EOF token.
110 { return this->classification_
== TOKEN_EOF
; }
112 // Return the token classification.
114 classification() const
115 { return this->classification_
; }
117 // Return the line number at which the token starts.
120 { return this->lineno_
; }
122 // Return the character position at this the token starts.
125 { return this->charpos_
; }
127 // Get the value of a token.
130 string_value(size_t* length
) const
132 gold_assert(this->classification_
== TOKEN_STRING
133 || this->classification_
== TOKEN_QUOTED_STRING
);
134 *length
= this->value_length_
;
139 operator_value() const
141 gold_assert(this->classification_
== TOKEN_OPERATOR
);
142 return this->opcode_
;
146 integer_value() const
148 gold_assert(this->classification_
== TOKEN_INTEGER
);
150 std::string
s(this->value_
, this->value_length_
);
151 return strtoull(s
.c_str(), NULL
, 0);
155 // The token classification.
156 Classification classification_
;
157 // The token value, for TOKEN_STRING or TOKEN_QUOTED_STRING or
160 // The length of the token value.
161 size_t value_length_
;
162 // The token value, for TOKEN_OPERATOR.
164 // The line number where this token started (one based).
166 // The character position within the line where this token started
171 // This class handles lexing a file into a sequence of tokens.
176 // We unfortunately have to support different lexing modes, because
177 // when reading different parts of a linker script we need to parse
178 // things differently.
181 // Reading an ordinary linker script.
183 // Reading an expression in a linker script.
185 // Reading a version script.
189 Lex(const char* input_string
, size_t input_length
, int parsing_token
)
190 : input_string_(input_string
), input_length_(input_length
),
191 current_(input_string
), mode_(LINKER_SCRIPT
),
192 first_token_(parsing_token
), token_(),
193 lineno_(1), linestart_(input_string
)
196 // Read a file into a string.
198 read_file(Input_file
*, std::string
*);
200 // Return the next token.
204 // Return the current lexing mode.
207 { return this->mode_
; }
209 // Set the lexing mode.
212 { this->mode_
= mode
; }
216 Lex
& operator=(const Lex
&);
218 // Make a general token with no value at the current location.
220 make_token(Token::Classification c
, const char* start
) const
221 { return Token(c
, this->lineno_
, start
- this->linestart_
+ 1); }
223 // Make a general token with a value at the current location.
225 make_token(Token::Classification c
, const char* v
, size_t len
,
228 { return Token(c
, v
, len
, this->lineno_
, start
- this->linestart_
+ 1); }
230 // Make an operator token at the current location.
232 make_token(int opcode
, const char* start
) const
233 { return Token(opcode
, this->lineno_
, start
- this->linestart_
+ 1); }
235 // Make an invalid token at the current location.
237 make_invalid_token(const char* start
)
238 { return this->make_token(Token::TOKEN_INVALID
, start
); }
240 // Make an EOF token at the current location.
242 make_eof_token(const char* start
)
243 { return this->make_token(Token::TOKEN_EOF
, start
); }
245 // Return whether C can be the first character in a name. C2 is the
246 // next character, since we sometimes need that.
248 can_start_name(char c
, char c2
);
250 // If C can appear in a name which has already started, return a
251 // pointer to a character later in the token or just past
252 // it. Otherwise, return NULL.
254 can_continue_name(const char* c
);
256 // Return whether C, C2, C3 can start a hex number.
258 can_start_hex(char c
, char c2
, char c3
);
260 // If C can appear in a hex number which has already started, return
261 // a pointer to a character later in the token or just past
262 // it. Otherwise, return NULL.
264 can_continue_hex(const char* c
);
266 // Return whether C can start a non-hex number.
268 can_start_number(char c
);
270 // If C can appear in a decimal number which has already started,
271 // return a pointer to a character later in the token or just past
272 // it. Otherwise, return NULL.
274 can_continue_number(const char* c
)
275 { return Lex::can_start_number(*c
) ? c
+ 1 : NULL
; }
277 // If C1 C2 C3 form a valid three character operator, return the
278 // opcode. Otherwise return 0.
280 three_char_operator(char c1
, char c2
, char c3
);
282 // If C1 C2 form a valid two character operator, return the opcode.
283 // Otherwise return 0.
285 two_char_operator(char c1
, char c2
);
287 // If C1 is a valid one character operator, return the opcode.
288 // Otherwise return 0.
290 one_char_operator(char c1
);
292 // Read the next token.
294 get_token(const char**);
296 // Skip a C style /* */ comment. Return false if the comment did
299 skip_c_comment(const char**);
301 // Skip a line # comment. Return false if there was no newline.
303 skip_line_comment(const char**);
305 // Build a token CLASSIFICATION from all characters that match
306 // CAN_CONTINUE_FN. The token starts at START. Start matching from
307 // MATCH. Set *PP to the character following the token.
309 gather_token(Token::Classification
,
310 const char* (Lex::*can_continue_fn
)(const char*),
311 const char* start
, const char* match
, const char** pp
);
313 // Build a token from a quoted string.
315 gather_quoted_string(const char** pp
);
317 // The string we are tokenizing.
318 const char* input_string_
;
319 // The length of the string.
320 size_t input_length_
;
321 // The current offset into the string.
322 const char* current_
;
323 // The current lexing mode.
325 // The code to use for the first token. This is set to 0 after it
328 // The current token.
330 // The current line number.
332 // The start of the current line in the string.
333 const char* linestart_
;
336 // Read the whole file into memory. We don't expect linker scripts to
337 // be large, so we just use a std::string as a buffer. We ignore the
338 // data we've already read, so that we read aligned buffers.
341 Lex::read_file(Input_file
* input_file
, std::string
* contents
)
343 off_t filesize
= input_file
->file().filesize();
345 contents
->reserve(filesize
);
348 unsigned char buf
[BUFSIZ
];
349 while (off
< filesize
)
352 if (get
> filesize
- off
)
353 get
= filesize
- off
;
354 input_file
->file().read(off
, get
, buf
);
355 contents
->append(reinterpret_cast<char*>(&buf
[0]), get
);
360 // Return whether C can be the start of a name, if the next character
361 // is C2. A name can being with a letter, underscore, period, or
362 // dollar sign. Because a name can be a file name, we also permit
363 // forward slash, backslash, and tilde. Tilde is the tricky case
364 // here; GNU ld also uses it as a bitwise not operator. It is only
365 // recognized as the operator if it is not immediately followed by
366 // some character which can appear in a symbol. That is, when we
367 // don't know that we are looking at an expression, "~0" is a file
368 // name, and "~ 0" is an expression using bitwise not. We are
372 Lex::can_start_name(char c
, char c2
)
376 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
377 case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
378 case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R':
379 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
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':
386 case '_': case '.': case '$':
390 return this->mode_
== LINKER_SCRIPT
;
393 return this->mode_
== LINKER_SCRIPT
&& can_continue_name(&c2
);
396 return (this->mode_
== VERSION_SCRIPT
397 || (this->mode_
== LINKER_SCRIPT
398 && can_continue_name(&c2
)));
405 // Return whether C can continue a name which has already started.
406 // Subsequent characters in a name are the same as the leading
407 // characters, plus digits and "=+-:[],?*". So in general the linker
408 // script language requires spaces around operators, unless we know
409 // that we are parsing an expression.
412 Lex::can_continue_name(const char* c
)
416 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
417 case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
418 case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R':
419 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
421 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
422 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
423 case 'm': case 'n': case 'o': case 'q': case 'p': case 'r':
424 case 's': case 't': case 'u': case 'v': case 'w': case 'x':
426 case '_': case '.': case '$':
427 case '0': case '1': case '2': case '3': case '4':
428 case '5': case '6': case '7': case '8': case '9':
431 case '/': case '\\': case '~':
434 if (this->mode_
== LINKER_SCRIPT
)
438 case '[': case ']': case '*': case '-':
439 if (this->mode_
== LINKER_SCRIPT
|| this->mode_
== VERSION_SCRIPT
)
444 if (this->mode_
== VERSION_SCRIPT
)
449 if (this->mode_
== LINKER_SCRIPT
)
451 else if (this->mode_
== VERSION_SCRIPT
&& (c
[1] == ':'))
453 // A name can have '::' in it, as that's a c++ namespace
454 // separator. But a single colon is not part of a name.
464 // For a number we accept 0x followed by hex digits, or any sequence
465 // of digits. The old linker accepts leading '$' for hex, and
466 // trailing HXBOD. Those are for MRI compatibility and we don't
467 // accept them. The old linker also accepts trailing MK for mega or
468 // kilo. FIXME: Those are mentioned in the documentation, and we
469 // should accept them.
471 // Return whether C1 C2 C3 can start a hex number.
474 Lex::can_start_hex(char c1
, char c2
, char c3
)
476 if (c1
== '0' && (c2
== 'x' || c2
== 'X'))
477 return this->can_continue_hex(&c3
);
481 // Return whether C can appear in a hex number.
484 Lex::can_continue_hex(const char* c
)
488 case '0': case '1': case '2': case '3': case '4':
489 case '5': case '6': case '7': case '8': case '9':
490 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
491 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
499 // Return whether C can start a non-hex number.
502 Lex::can_start_number(char c
)
506 case '0': case '1': case '2': case '3': case '4':
507 case '5': case '6': case '7': case '8': case '9':
515 // If C1 C2 C3 form a valid three character operator, return the
516 // opcode (defined in the yyscript.h file generated from yyscript.y).
517 // Otherwise return 0.
520 Lex::three_char_operator(char c1
, char c2
, char c3
)
525 if (c2
== '<' && c3
== '=')
529 if (c2
== '>' && c3
== '=')
538 // If C1 C2 form a valid two character operator, return the opcode
539 // (defined in the yyscript.h file generated from yyscript.y).
540 // Otherwise return 0.
543 Lex::two_char_operator(char c1
, char c2
)
601 // If C1 is a valid operator, return the opcode. Otherwise return 0.
604 Lex::one_char_operator(char c1
)
637 // Skip a C style comment. *PP points to just after the "/*". Return
638 // false if the comment did not end.
641 Lex::skip_c_comment(const char** pp
)
644 while (p
[0] != '*' || p
[1] != '/')
655 this->linestart_
= p
+ 1;
664 // Skip a line # comment. Return false if there was no newline.
667 Lex::skip_line_comment(const char** pp
)
670 size_t skip
= strcspn(p
, "\n");
679 this->linestart_
= p
;
685 // Build a token CLASSIFICATION from all characters that match
686 // CAN_CONTINUE_FN. Update *PP.
689 Lex::gather_token(Token::Classification classification
,
690 const char* (Lex::*can_continue_fn
)(const char*),
695 const char* new_match
= NULL
;
696 while ((new_match
= (this->*can_continue_fn
)(match
)))
699 return this->make_token(classification
, start
, match
- start
, start
);
702 // Build a token from a quoted string.
705 Lex::gather_quoted_string(const char** pp
)
707 const char* start
= *pp
;
708 const char* p
= start
;
710 size_t skip
= strcspn(p
, "\"\n");
712 return this->make_invalid_token(start
);
714 return this->make_token(Token::TOKEN_QUOTED_STRING
, p
, skip
, start
);
717 // Return the next token at *PP. Update *PP. General guideline: we
718 // require linker scripts to be simple ASCII. No unicode linker
719 // scripts. In particular we can assume that any '\0' is the end of
723 Lex::get_token(const char** pp
)
732 return this->make_eof_token(p
);
735 // Skip whitespace quickly.
736 while (*p
== ' ' || *p
== '\t')
743 this->linestart_
= p
;
747 // Skip C style comments.
748 if (p
[0] == '/' && p
[1] == '*')
750 int lineno
= this->lineno_
;
751 int charpos
= p
- this->linestart_
+ 1;
754 if (!this->skip_c_comment(pp
))
755 return Token(Token::TOKEN_INVALID
, lineno
, charpos
);
761 // Skip line comments.
765 if (!this->skip_line_comment(pp
))
766 return this->make_eof_token(p
);
772 if (this->can_start_name(p
[0], p
[1]))
773 return this->gather_token(Token::TOKEN_STRING
,
774 &Lex::can_continue_name
,
777 // We accept any arbitrary name in double quotes, as long as it
778 // does not cross a line boundary.
782 return this->gather_quoted_string(pp
);
785 // Check for a number.
787 if (this->can_start_hex(p
[0], p
[1], p
[2]))
788 return this->gather_token(Token::TOKEN_INTEGER
,
789 &Lex::can_continue_hex
,
792 if (Lex::can_start_number(p
[0]))
793 return this->gather_token(Token::TOKEN_INTEGER
,
794 &Lex::can_continue_number
,
797 // Check for operators.
799 int opcode
= Lex::three_char_operator(p
[0], p
[1], p
[2]);
803 return this->make_token(opcode
, p
);
806 opcode
= Lex::two_char_operator(p
[0], p
[1]);
810 return this->make_token(opcode
, p
);
813 opcode
= Lex::one_char_operator(p
[0]);
817 return this->make_token(opcode
, p
);
820 return this->make_token(Token::TOKEN_INVALID
, p
);
824 // Return the next token.
829 // The first token is special.
830 if (this->first_token_
!= 0)
832 this->token_
= Token(this->first_token_
, 0, 0);
833 this->first_token_
= 0;
834 return &this->token_
;
837 this->token_
= this->get_token(&this->current_
);
839 // Don't let an early null byte fool us into thinking that we've
840 // reached the end of the file.
841 if (this->token_
.is_eof()
842 && (static_cast<size_t>(this->current_
- this->input_string_
)
843 < this->input_length_
))
844 this->token_
= this->make_invalid_token(this->current_
);
846 return &this->token_
;
849 // A trivial task which waits for THIS_BLOCKER to be clear and then
850 // clears NEXT_BLOCKER. THIS_BLOCKER may be NULL.
852 class Script_unblock
: public Task
855 Script_unblock(Task_token
* this_blocker
, Task_token
* next_blocker
)
856 : this_blocker_(this_blocker
), next_blocker_(next_blocker
)
861 if (this->this_blocker_
!= NULL
)
862 delete this->this_blocker_
;
868 if (this->this_blocker_
!= NULL
&& this->this_blocker_
->is_blocked())
869 return this->this_blocker_
;
874 locks(Task_locker
* tl
)
875 { tl
->add(this, this->next_blocker_
); }
883 { return "Script_unblock"; }
886 Task_token
* this_blocker_
;
887 Task_token
* next_blocker_
;
890 // class Symbol_assignment.
892 // Add the symbol to the symbol table. This makes sure the symbol is
893 // there and defined. The actual value is stored later. We can't
894 // determine the actual value at this point, because we can't
895 // necessarily evaluate the expression until all ordinary symbols have
899 Symbol_assignment::add_to_table(Symbol_table
* symtab
)
901 elfcpp::STV vis
= this->hidden_
? elfcpp::STV_HIDDEN
: elfcpp::STV_DEFAULT
;
902 this->sym_
= symtab
->define_as_constant(this->name_
.c_str(),
913 // Finalize a symbol value.
916 Symbol_assignment::finalize(Symbol_table
* symtab
, const Layout
* layout
)
918 this->finalize_maybe_dot(symtab
, layout
, false, 0, NULL
);
921 // Finalize a symbol value which can refer to the dot symbol.
924 Symbol_assignment::finalize_with_dot(Symbol_table
* symtab
,
925 const Layout
* layout
,
927 Output_section
* dot_section
)
929 this->finalize_maybe_dot(symtab
, layout
, true, dot_value
, dot_section
);
932 // Finalize a symbol value, internal version.
935 Symbol_assignment::finalize_maybe_dot(Symbol_table
* symtab
,
936 const Layout
* layout
,
937 bool is_dot_available
,
939 Output_section
* dot_section
)
941 // If we were only supposed to provide this symbol, the sym_ field
942 // will be NULL if the symbol was not referenced.
943 if (this->sym_
== NULL
)
945 gold_assert(this->provide_
);
949 if (parameters
->get_size() == 32)
951 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
952 this->sized_finalize
<32>(symtab
, layout
, is_dot_available
, dot_value
,
958 else if (parameters
->get_size() == 64)
960 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
961 this->sized_finalize
<64>(symtab
, layout
, is_dot_available
, dot_value
,
973 Symbol_assignment::sized_finalize(Symbol_table
* symtab
, const Layout
* layout
,
974 bool is_dot_available
, uint64_t dot_value
,
975 Output_section
* dot_section
)
977 Output_section
* section
;
978 uint64_t final_val
= this->val_
->eval_maybe_dot(symtab
, layout
,
980 dot_value
, dot_section
,
982 Sized_symbol
<size
>* ssym
= symtab
->get_sized_symbol
<size
>(this->sym_
);
983 ssym
->set_value(final_val
);
985 ssym
->set_output_section(section
);
988 // Set the symbol value if the expression yields an absolute value.
991 Symbol_assignment::set_if_absolute(Symbol_table
* symtab
, const Layout
* layout
,
992 bool is_dot_available
, uint64_t dot_value
)
994 if (this->sym_
== NULL
)
997 Output_section
* val_section
;
998 uint64_t val
= this->val_
->eval_maybe_dot(symtab
, layout
, is_dot_available
,
999 dot_value
, NULL
, &val_section
);
1000 if (val_section
!= NULL
)
1003 if (parameters
->get_size() == 32)
1005 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
1006 Sized_symbol
<32>* ssym
= symtab
->get_sized_symbol
<32>(this->sym_
);
1007 ssym
->set_value(val
);
1012 else if (parameters
->get_size() == 64)
1014 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
1015 Sized_symbol
<64>* ssym
= symtab
->get_sized_symbol
<64>(this->sym_
);
1016 ssym
->set_value(val
);
1025 // Print for debugging.
1028 Symbol_assignment::print(FILE* f
) const
1030 if (this->provide_
&& this->hidden_
)
1031 fprintf(f
, "PROVIDE_HIDDEN(");
1032 else if (this->provide_
)
1033 fprintf(f
, "PROVIDE(");
1034 else if (this->hidden_
)
1037 fprintf(f
, "%s = ", this->name_
.c_str());
1038 this->val_
->print(f
);
1040 if (this->provide_
|| this->hidden_
)
1046 // Class Script_assertion.
1048 // Check the assertion.
1051 Script_assertion::check(const Symbol_table
* symtab
, const Layout
* layout
)
1053 if (!this->check_
->eval(symtab
, layout
))
1054 gold_error("%s", this->message_
.c_str());
1057 // Print for debugging.
1060 Script_assertion::print(FILE* f
) const
1062 fprintf(f
, "ASSERT(");
1063 this->check_
->print(f
);
1064 fprintf(f
, ", \"%s\")\n", this->message_
.c_str());
1067 // Class Script_options.
1069 Script_options::Script_options()
1070 : entry_(), symbol_assignments_(), version_script_info_(),
1075 // Add a symbol to be defined.
1078 Script_options::add_symbol_assignment(const char* name
, size_t length
,
1079 Expression
* value
, bool provide
,
1082 if (length
!= 1 || name
[0] != '.')
1084 if (this->script_sections_
.in_sections_clause())
1085 this->script_sections_
.add_symbol_assignment(name
, length
, value
,
1089 Symbol_assignment
* p
= new Symbol_assignment(name
, length
, value
,
1091 this->symbol_assignments_
.push_back(p
);
1096 if (provide
|| hidden
)
1097 gold_error(_("invalid use of PROVIDE for dot symbol"));
1098 if (!this->script_sections_
.in_sections_clause())
1099 gold_error(_("invalid assignment to dot outside of SECTIONS"));
1101 this->script_sections_
.add_dot_assignment(value
);
1105 // Add an assertion.
1108 Script_options::add_assertion(Expression
* check
, const char* message
,
1111 if (this->script_sections_
.in_sections_clause())
1112 this->script_sections_
.add_assertion(check
, message
, messagelen
);
1115 Script_assertion
* p
= new Script_assertion(check
, message
, messagelen
);
1116 this->assertions_
.push_back(p
);
1120 // Add any symbols we are defining to the symbol table.
1123 Script_options::add_symbols_to_table(Symbol_table
* symtab
)
1125 for (Symbol_assignments::iterator p
= this->symbol_assignments_
.begin();
1126 p
!= this->symbol_assignments_
.end();
1128 (*p
)->add_to_table(symtab
);
1129 this->script_sections_
.add_symbols_to_table(symtab
);
1132 // Finalize symbol values. Also check assertions.
1135 Script_options::finalize_symbols(Symbol_table
* symtab
, const Layout
* layout
)
1137 for (Symbol_assignments::iterator p
= this->symbol_assignments_
.begin();
1138 p
!= this->symbol_assignments_
.end();
1140 (*p
)->finalize(symtab
, layout
);
1142 for (Assertions::iterator p
= this->assertions_
.begin();
1143 p
!= this->assertions_
.end();
1145 (*p
)->check(symtab
, layout
);
1147 this->script_sections_
.finalize_symbols(symtab
, layout
);
1150 // Set section addresses. We set all the symbols which have absolute
1151 // values. Then we let the SECTIONS clause do its thing. This
1152 // returns the segment which holds the file header and segment
1156 Script_options::set_section_addresses(Symbol_table
* symtab
, Layout
* layout
)
1158 for (Symbol_assignments::iterator p
= this->symbol_assignments_
.begin();
1159 p
!= this->symbol_assignments_
.end();
1161 (*p
)->set_if_absolute(symtab
, layout
, false, 0);
1163 return this->script_sections_
.set_section_addresses(symtab
, layout
);
1166 // This class holds data passed through the parser to the lexer and to
1167 // the parser support functions. This avoids global variables. We
1168 // can't use global variables because we need not be called by a
1169 // singleton thread.
1171 class Parser_closure
1174 Parser_closure(const char* filename
,
1175 const Position_dependent_options
& posdep_options
,
1176 bool in_group
, bool is_in_sysroot
,
1177 Command_line
* command_line
,
1178 Script_options
* script_options
,
1180 : filename_(filename
), posdep_options_(posdep_options
),
1181 in_group_(in_group
), is_in_sysroot_(is_in_sysroot
),
1182 command_line_(command_line
), script_options_(script_options
),
1183 version_script_info_(script_options
->version_script_info()),
1184 lex_(lex
), lineno_(0), charpos_(0), lex_mode_stack_(), inputs_(NULL
)
1186 // We start out processing C symbols in the default lex mode.
1187 language_stack_
.push_back("");
1188 lex_mode_stack_
.push_back(lex
->mode());
1191 // Return the file name.
1194 { return this->filename_
; }
1196 // Return the position dependent options. The caller may modify
1198 Position_dependent_options
&
1199 position_dependent_options()
1200 { return this->posdep_options_
; }
1202 // Return whether this script is being run in a group.
1205 { return this->in_group_
; }
1207 // Return whether this script was found using a directory in the
1210 is_in_sysroot() const
1211 { return this->is_in_sysroot_
; }
1213 // Returns the Command_line structure passed in at constructor time.
1214 // This value may be NULL. The caller may modify this, which modifies
1215 // the passed-in Command_line object (not a copy).
1218 { return this->command_line_
; }
1220 // Return the options which may be set by a script.
1223 { return this->script_options_
; }
1225 // Return the object in which version script information should be stored.
1226 Version_script_info
*
1228 { return this->version_script_info_
; }
1230 // Return the next token, and advance.
1234 const Token
* token
= this->lex_
->next_token();
1235 this->lineno_
= token
->lineno();
1236 this->charpos_
= token
->charpos();
1240 // Set a new lexer mode, pushing the current one.
1242 push_lex_mode(Lex::Mode mode
)
1244 this->lex_mode_stack_
.push_back(this->lex_
->mode());
1245 this->lex_
->set_mode(mode
);
1248 // Pop the lexer mode.
1252 gold_assert(!this->lex_mode_stack_
.empty());
1253 this->lex_
->set_mode(this->lex_mode_stack_
.back());
1254 this->lex_mode_stack_
.pop_back();
1257 // Return the current lexer mode.
1260 { return this->lex_mode_stack_
.back(); }
1262 // Return the line number of the last token.
1265 { return this->lineno_
; }
1267 // Return the character position in the line of the last token.
1270 { return this->charpos_
; }
1272 // Return the list of input files, creating it if necessary. This
1273 // is a space leak--we never free the INPUTS_ pointer.
1277 if (this->inputs_
== NULL
)
1278 this->inputs_
= new Input_arguments();
1279 return this->inputs_
;
1282 // Return whether we saw any input files.
1285 { return this->inputs_
!= NULL
&& !this->inputs_
->empty(); }
1287 // Return the current language being processed in a version script
1288 // (eg, "C++"). The empty string represents unmangled C names.
1290 get_current_language() const
1291 { return this->language_stack_
.back(); }
1293 // Push a language onto the stack when entering an extern block.
1294 void push_language(const std::string
& lang
)
1295 { this->language_stack_
.push_back(lang
); }
1297 // Pop a language off of the stack when exiting an extern block.
1300 gold_assert(!this->language_stack_
.empty());
1301 this->language_stack_
.pop_back();
1305 // The name of the file we are reading.
1306 const char* filename_
;
1307 // The position dependent options.
1308 Position_dependent_options posdep_options_
;
1309 // Whether we are currently in a --start-group/--end-group.
1311 // Whether the script was found in a sysrooted directory.
1312 bool is_in_sysroot_
;
1313 // May be NULL if the user chooses not to pass one in.
1314 Command_line
* command_line_
;
1315 // Options which may be set from any linker script.
1316 Script_options
* script_options_
;
1317 // Information parsed from a version script.
1318 Version_script_info
* version_script_info_
;
1321 // The line number of the last token returned by next_token.
1323 // The column number of the last token returned by next_token.
1325 // A stack of lexer modes.
1326 std::vector
<Lex::Mode
> lex_mode_stack_
;
1327 // A stack of which extern/language block we're inside. Can be C++,
1328 // java, or empty for C.
1329 std::vector
<std::string
> language_stack_
;
1330 // New input files found to add to the link.
1331 Input_arguments
* inputs_
;
1334 // FILE was found as an argument on the command line. Try to read it
1335 // as a script. We've already read BYTES of data into P, but we
1336 // ignore that. Return true if the file was handled.
1339 read_input_script(Workqueue
* workqueue
, const General_options
& options
,
1340 Symbol_table
* symtab
, Layout
* layout
,
1341 Dirsearch
* dirsearch
, Input_objects
* input_objects
,
1342 Input_group
* input_group
,
1343 const Input_argument
* input_argument
,
1344 Input_file
* input_file
, const unsigned char*, off_t
,
1345 Task_token
* this_blocker
, Task_token
* next_blocker
)
1347 std::string input_string
;
1348 Lex::read_file(input_file
, &input_string
);
1350 Lex
lex(input_string
.c_str(), input_string
.length(), PARSING_LINKER_SCRIPT
);
1352 Parser_closure
closure(input_file
->filename().c_str(),
1353 input_argument
->file().options(),
1354 input_group
!= NULL
,
1355 input_file
->is_in_sysroot(),
1357 layout
->script_options(),
1360 if (yyparse(&closure
) != 0)
1363 // THIS_BLOCKER must be clear before we may add anything to the
1364 // symbol table. We are responsible for unblocking NEXT_BLOCKER
1365 // when we are done. We are responsible for deleting THIS_BLOCKER
1366 // when it is unblocked.
1368 if (!closure
.saw_inputs())
1370 // The script did not add any files to read. Note that we are
1371 // not permitted to call NEXT_BLOCKER->unblock() here even if
1372 // THIS_BLOCKER is NULL, as we do not hold the workqueue lock.
1373 workqueue
->queue(new Script_unblock(this_blocker
, next_blocker
));
1377 for (Input_arguments::const_iterator p
= closure
.inputs()->begin();
1378 p
!= closure
.inputs()->end();
1382 if (p
+ 1 == closure
.inputs()->end())
1386 nb
= new Task_token(true);
1389 workqueue
->queue(new Read_symbols(options
, input_objects
, symtab
,
1390 layout
, dirsearch
, &*p
,
1391 input_group
, this_blocker
, nb
));
1398 // Helper function for read_version_script() and
1399 // read_commandline_script(). Processes the given file in the mode
1400 // indicated by first_token and lex_mode.
1403 read_script_file(const char* filename
, Command_line
* cmdline
,
1404 int first_token
, Lex::Mode lex_mode
)
1406 // TODO: if filename is a relative filename, search for it manually
1407 // using "." + cmdline->options()->search_path() -- not dirsearch.
1408 Dirsearch dirsearch
;
1410 // The file locking code wants to record a Task, but we haven't
1411 // started the workqueue yet. This is only for debugging purposes,
1412 // so we invent a fake value.
1413 const Task
* task
= reinterpret_cast<const Task
*>(-1);
1415 // We don't want this file to be opened in binary mode.
1416 Position_dependent_options posdep
= cmdline
->position_dependent_options();
1417 if (posdep
.input_format() == General_options::OBJECT_FORMAT_BINARY
)
1418 posdep
.set_input_format("elf");
1419 Input_file_argument
input_argument(filename
, false, "", false, posdep
);
1420 Input_file
input_file(&input_argument
);
1421 if (!input_file
.open(cmdline
->options(), dirsearch
, task
))
1424 std::string input_string
;
1425 Lex::read_file(&input_file
, &input_string
);
1427 Lex
lex(input_string
.c_str(), input_string
.length(), first_token
);
1428 lex
.set_mode(lex_mode
);
1430 Parser_closure
closure(filename
,
1431 cmdline
->position_dependent_options(),
1433 input_file
.is_in_sysroot(),
1435 cmdline
->script_options(),
1437 if (yyparse(&closure
) != 0)
1439 input_file
.file().unlock(task
);
1443 input_file
.file().unlock(task
);
1445 gold_assert(!closure
.saw_inputs());
1450 // FILENAME was found as an argument to --script (-T).
1451 // Read it as a script, and execute its contents immediately.
1454 read_commandline_script(const char* filename
, Command_line
* cmdline
)
1456 return read_script_file(filename
, cmdline
,
1457 PARSING_LINKER_SCRIPT
, Lex::LINKER_SCRIPT
);
1460 // FILE was found as an argument to --version-script. Read it as a
1461 // version script, and store its contents in
1462 // cmdline->script_options()->version_script_info().
1465 read_version_script(const char* filename
, Command_line
* cmdline
)
1467 return read_script_file(filename
, cmdline
,
1468 PARSING_VERSION_SCRIPT
, Lex::VERSION_SCRIPT
);
1471 // Implement the --defsym option on the command line. Return true if
1475 Script_options::define_symbol(const char* definition
)
1477 Lex
lex(definition
, strlen(definition
), PARSING_DEFSYM
);
1478 lex
.set_mode(Lex::EXPRESSION
);
1481 Position_dependent_options posdep_options
;
1483 Parser_closure
closure("command line", posdep_options
, false, false, NULL
,
1486 if (yyparse(&closure
) != 0)
1489 gold_assert(!closure
.saw_inputs());
1494 // Print the script to F for debugging.
1497 Script_options::print(FILE* f
) const
1499 fprintf(f
, "%s: Dumping linker script\n", program_name
);
1501 if (!this->entry_
.empty())
1502 fprintf(f
, "ENTRY(%s)\n", this->entry_
.c_str());
1504 for (Symbol_assignments::const_iterator p
=
1505 this->symbol_assignments_
.begin();
1506 p
!= this->symbol_assignments_
.end();
1510 for (Assertions::const_iterator p
= this->assertions_
.begin();
1511 p
!= this->assertions_
.end();
1515 this->script_sections_
.print(f
);
1517 this->version_script_info_
.print(f
);
1520 // Manage mapping from keywords to the codes expected by the bison
1521 // parser. We construct one global object for each lex mode with
1524 class Keyword_to_parsecode
1527 // The structure which maps keywords to parsecodes.
1528 struct Keyword_parsecode
1531 const char* keyword
;
1532 // Corresponding parsecode.
1536 Keyword_to_parsecode(const Keyword_parsecode
* keywords
,
1538 : keyword_parsecodes_(keywords
), keyword_count_(keyword_count
)
1541 // Return the parsecode corresponding KEYWORD, or 0 if it is not a
1544 keyword_to_parsecode(const char* keyword
, size_t len
) const;
1547 const Keyword_parsecode
* keyword_parsecodes_
;
1548 const int keyword_count_
;
1551 // Mapping from keyword string to keyword parsecode. This array must
1552 // be kept in sorted order. Parsecodes are looked up using bsearch.
1553 // This array must correspond to the list of parsecodes in yyscript.y.
1555 static const Keyword_to_parsecode::Keyword_parsecode
1556 script_keyword_parsecodes
[] =
1558 { "ABSOLUTE", ABSOLUTE
},
1560 { "ALIGN", ALIGN_K
},
1561 { "ALIGNOF", ALIGNOF
},
1562 { "ASSERT", ASSERT_K
},
1563 { "AS_NEEDED", AS_NEEDED
},
1568 { "CONSTANT", CONSTANT
},
1569 { "CONSTRUCTORS", CONSTRUCTORS
},
1570 { "CREATE_OBJECT_SYMBOLS", CREATE_OBJECT_SYMBOLS
},
1571 { "DATA_SEGMENT_ALIGN", DATA_SEGMENT_ALIGN
},
1572 { "DATA_SEGMENT_END", DATA_SEGMENT_END
},
1573 { "DATA_SEGMENT_RELRO_END", DATA_SEGMENT_RELRO_END
},
1574 { "DEFINED", DEFINED
},
1576 { "EXCLUDE_FILE", EXCLUDE_FILE
},
1577 { "EXTERN", EXTERN
},
1580 { "FORCE_COMMON_ALLOCATION", FORCE_COMMON_ALLOCATION
},
1583 { "INCLUDE", INCLUDE
},
1584 { "INHIBIT_COMMON_ALLOCATION", INHIBIT_COMMON_ALLOCATION
},
1587 { "LENGTH", LENGTH
},
1588 { "LOADADDR", LOADADDR
},
1592 { "MEMORY", MEMORY
},
1595 { "NOCROSSREFS", NOCROSSREFS
},
1596 { "NOFLOAT", NOFLOAT
},
1597 { "ONLY_IF_RO", ONLY_IF_RO
},
1598 { "ONLY_IF_RW", ONLY_IF_RW
},
1599 { "OPTION", OPTION
},
1600 { "ORIGIN", ORIGIN
},
1601 { "OUTPUT", OUTPUT
},
1602 { "OUTPUT_ARCH", OUTPUT_ARCH
},
1603 { "OUTPUT_FORMAT", OUTPUT_FORMAT
},
1604 { "OVERLAY", OVERLAY
},
1606 { "PROVIDE", PROVIDE
},
1607 { "PROVIDE_HIDDEN", PROVIDE_HIDDEN
},
1609 { "SEARCH_DIR", SEARCH_DIR
},
1610 { "SECTIONS", SECTIONS
},
1611 { "SEGMENT_START", SEGMENT_START
},
1613 { "SIZEOF", SIZEOF
},
1614 { "SIZEOF_HEADERS", SIZEOF_HEADERS
},
1615 { "SORT", SORT_BY_NAME
},
1616 { "SORT_BY_ALIGNMENT", SORT_BY_ALIGNMENT
},
1617 { "SORT_BY_NAME", SORT_BY_NAME
},
1618 { "SPECIAL", SPECIAL
},
1620 { "STARTUP", STARTUP
},
1621 { "SUBALIGN", SUBALIGN
},
1622 { "SYSLIB", SYSLIB
},
1623 { "TARGET", TARGET_K
},
1624 { "TRUNCATE", TRUNCATE
},
1625 { "VERSION", VERSIONK
},
1626 { "global", GLOBAL
},
1632 { "sizeof_headers", SIZEOF_HEADERS
},
1635 static const Keyword_to_parsecode
1636 script_keywords(&script_keyword_parsecodes
[0],
1637 (sizeof(script_keyword_parsecodes
)
1638 / sizeof(script_keyword_parsecodes
[0])));
1640 static const Keyword_to_parsecode::Keyword_parsecode
1641 version_script_keyword_parsecodes
[] =
1643 { "extern", EXTERN
},
1644 { "global", GLOBAL
},
1648 static const Keyword_to_parsecode
1649 version_script_keywords(&version_script_keyword_parsecodes
[0],
1650 (sizeof(version_script_keyword_parsecodes
)
1651 / sizeof(version_script_keyword_parsecodes
[0])));
1653 // Comparison function passed to bsearch.
1665 ktt_compare(const void* keyv
, const void* kttv
)
1667 const Ktt_key
* key
= static_cast<const Ktt_key
*>(keyv
);
1668 const Keyword_to_parsecode::Keyword_parsecode
* ktt
=
1669 static_cast<const Keyword_to_parsecode::Keyword_parsecode
*>(kttv
);
1670 int i
= strncmp(key
->str
, ktt
->keyword
, key
->len
);
1673 if (ktt
->keyword
[key
->len
] != '\0')
1678 } // End extern "C".
1681 Keyword_to_parsecode::keyword_to_parsecode(const char* keyword
,
1687 void* kttv
= bsearch(&key
,
1688 this->keyword_parsecodes_
,
1689 this->keyword_count_
,
1690 sizeof(this->keyword_parsecodes_
[0]),
1694 Keyword_parsecode
* ktt
= static_cast<Keyword_parsecode
*>(kttv
);
1695 return ktt
->parsecode
;
1698 // The following structs are used within the VersionInfo class as well
1699 // as in the bison helper functions. They store the information
1700 // parsed from the version script.
1702 // A single version expression.
1703 // For example, pattern="std::map*" and language="C++".
1704 // pattern and language should be from the stringpool
1705 struct Version_expression
{
1706 Version_expression(const std::string
& pattern
,
1707 const std::string
& language
,
1709 : pattern(pattern
), language(language
), exact_match(exact_match
) {}
1711 std::string pattern
;
1712 std::string language
;
1713 // If false, we use glob() to match pattern. If true, we use strcmp().
1718 // A list of expressions.
1719 struct Version_expression_list
{
1720 std::vector
<struct Version_expression
> expressions
;
1724 // A list of which versions upon which another version depends.
1725 // Strings should be from the Stringpool.
1726 struct Version_dependency_list
{
1727 std::vector
<std::string
> dependencies
;
1731 // The total definition of a version. It includes the tag for the
1732 // version, its global and local expressions, and any dependencies.
1733 struct Version_tree
{
1735 : tag(), global(NULL
), local(NULL
), dependencies(NULL
) {}
1738 const struct Version_expression_list
* global
;
1739 const struct Version_expression_list
* local
;
1740 const struct Version_dependency_list
* dependencies
;
1743 Version_script_info::~Version_script_info()
1749 Version_script_info::clear()
1751 for (size_t k
= 0; k
< dependency_lists_
.size(); ++k
)
1752 delete dependency_lists_
[k
];
1753 this->dependency_lists_
.clear();
1754 for (size_t k
= 0; k
< version_trees_
.size(); ++k
)
1755 delete version_trees_
[k
];
1756 this->version_trees_
.clear();
1757 for (size_t k
= 0; k
< expression_lists_
.size(); ++k
)
1758 delete expression_lists_
[k
];
1759 this->expression_lists_
.clear();
1762 std::vector
<std::string
>
1763 Version_script_info::get_versions() const
1765 std::vector
<std::string
> ret
;
1766 for (size_t j
= 0; j
< version_trees_
.size(); ++j
)
1767 ret
.push_back(version_trees_
[j
]->tag
);
1771 std::vector
<std::string
>
1772 Version_script_info::get_dependencies(const char* version
) const
1774 std::vector
<std::string
> ret
;
1775 for (size_t j
= 0; j
< version_trees_
.size(); ++j
)
1776 if (version_trees_
[j
]->tag
== version
)
1778 const struct Version_dependency_list
* deps
=
1779 version_trees_
[j
]->dependencies
;
1781 for (size_t k
= 0; k
< deps
->dependencies
.size(); ++k
)
1782 ret
.push_back(deps
->dependencies
[k
]);
1789 Version_script_info::get_symbol_version_helper(const char* symbol_name
,
1790 bool check_global
) const
1792 for (size_t j
= 0; j
< version_trees_
.size(); ++j
)
1794 // Is it a global symbol for this version?
1795 const Version_expression_list
* explist
=
1796 check_global
? version_trees_
[j
]->global
: version_trees_
[j
]->local
;
1797 if (explist
!= NULL
)
1798 for (size_t k
= 0; k
< explist
->expressions
.size(); ++k
)
1800 const char* name_to_match
= symbol_name
;
1801 const struct Version_expression
& exp
= explist
->expressions
[k
];
1802 char* demangled_name
= NULL
;
1803 if (exp
.language
== "C++")
1805 demangled_name
= cplus_demangle(symbol_name
,
1806 DMGL_ANSI
| DMGL_PARAMS
);
1807 // This isn't a C++ symbol.
1808 if (demangled_name
== NULL
)
1810 name_to_match
= demangled_name
;
1812 else if (exp
.language
== "Java")
1814 demangled_name
= cplus_demangle(symbol_name
,
1815 (DMGL_ANSI
| DMGL_PARAMS
1817 // This isn't a Java symbol.
1818 if (demangled_name
== NULL
)
1820 name_to_match
= demangled_name
;
1823 if (exp
.exact_match
)
1824 matched
= strcmp(exp
.pattern
.c_str(), name_to_match
) == 0;
1826 matched
= fnmatch(exp
.pattern
.c_str(), name_to_match
,
1828 if (demangled_name
!= NULL
)
1829 free(demangled_name
);
1831 return version_trees_
[j
]->tag
;
1834 static const std::string empty
= "";
1838 struct Version_dependency_list
*
1839 Version_script_info::allocate_dependency_list()
1841 dependency_lists_
.push_back(new Version_dependency_list
);
1842 return dependency_lists_
.back();
1845 struct Version_expression_list
*
1846 Version_script_info::allocate_expression_list()
1848 expression_lists_
.push_back(new Version_expression_list
);
1849 return expression_lists_
.back();
1852 struct Version_tree
*
1853 Version_script_info::allocate_version_tree()
1855 version_trees_
.push_back(new Version_tree
);
1856 return version_trees_
.back();
1859 // Print for debugging.
1862 Version_script_info::print(FILE* f
) const
1867 fprintf(f
, "VERSION {");
1869 for (size_t i
= 0; i
< this->version_trees_
.size(); ++i
)
1871 const Version_tree
* vt
= this->version_trees_
[i
];
1873 if (vt
->tag
.empty())
1876 fprintf(f
, " %s {\n", vt
->tag
.c_str());
1878 if (vt
->global
!= NULL
)
1880 fprintf(f
, " global :\n");
1881 this->print_expression_list(f
, vt
->global
);
1884 if (vt
->local
!= NULL
)
1886 fprintf(f
, " local :\n");
1887 this->print_expression_list(f
, vt
->local
);
1891 if (vt
->dependencies
!= NULL
)
1893 const Version_dependency_list
* deps
= vt
->dependencies
;
1894 for (size_t j
= 0; j
< deps
->dependencies
.size(); ++j
)
1896 if (j
< deps
->dependencies
.size() - 1)
1898 fprintf(f
, " %s", deps
->dependencies
[j
].c_str());
1908 Version_script_info::print_expression_list(
1910 const Version_expression_list
* vel
) const
1912 std::string current_language
;
1913 for (size_t i
= 0; i
< vel
->expressions
.size(); ++i
)
1915 const Version_expression
& ve(vel
->expressions
[i
]);
1917 if (ve
.language
!= current_language
)
1919 if (!current_language
.empty())
1921 fprintf(f
, " extern \"%s\" {\n", ve
.language
.c_str());
1922 current_language
= ve
.language
;
1926 if (!current_language
.empty())
1931 fprintf(f
, "%s", ve
.pattern
.c_str());
1938 if (!current_language
.empty())
1942 } // End namespace gold.
1944 // The remaining functions are extern "C", so it's clearer to not put
1945 // them in namespace gold.
1947 using namespace gold
;
1949 // This function is called by the bison parser to return the next
1953 yylex(YYSTYPE
* lvalp
, void* closurev
)
1955 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
1956 const Token
* token
= closure
->next_token();
1957 switch (token
->classification())
1962 case Token::TOKEN_INVALID
:
1963 yyerror(closurev
, "invalid character");
1966 case Token::TOKEN_EOF
:
1969 case Token::TOKEN_STRING
:
1971 // This is either a keyword or a STRING.
1973 const char* str
= token
->string_value(&len
);
1975 switch (closure
->lex_mode())
1977 case Lex::LINKER_SCRIPT
:
1978 parsecode
= script_keywords
.keyword_to_parsecode(str
, len
);
1980 case Lex::VERSION_SCRIPT
:
1981 parsecode
= version_script_keywords
.keyword_to_parsecode(str
, len
);
1988 lvalp
->string
.value
= str
;
1989 lvalp
->string
.length
= len
;
1993 case Token::TOKEN_QUOTED_STRING
:
1994 lvalp
->string
.value
= token
->string_value(&lvalp
->string
.length
);
1995 return QUOTED_STRING
;
1997 case Token::TOKEN_OPERATOR
:
1998 return token
->operator_value();
2000 case Token::TOKEN_INTEGER
:
2001 lvalp
->integer
= token
->integer_value();
2006 // This function is called by the bison parser to report an error.
2009 yyerror(void* closurev
, const char* message
)
2011 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2012 gold_error(_("%s:%d:%d: %s"), closure
->filename(), closure
->lineno(),
2013 closure
->charpos(), message
);
2016 // Called by the bison parser to add a file to the link.
2019 script_add_file(void* closurev
, const char* name
, size_t length
)
2021 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2023 // If this is an absolute path, and we found the script in the
2024 // sysroot, then we want to prepend the sysroot to the file name.
2025 // For example, this is how we handle a cross link to the x86_64
2026 // libc.so, which refers to /lib/libc.so.6.
2027 std::string
name_string(name
, length
);
2028 const char* extra_search_path
= ".";
2029 std::string script_directory
;
2030 if (IS_ABSOLUTE_PATH(name_string
.c_str()))
2032 if (closure
->is_in_sysroot())
2034 const std::string
& sysroot(parameters
->sysroot());
2035 gold_assert(!sysroot
.empty());
2036 name_string
= sysroot
+ name_string
;
2041 // In addition to checking the normal library search path, we
2042 // also want to check in the script-directory.
2043 const char *slash
= strrchr(closure
->filename(), '/');
2046 script_directory
.assign(closure
->filename(),
2047 slash
- closure
->filename() + 1);
2048 extra_search_path
= script_directory
.c_str();
2052 Input_file_argument
file(name_string
.c_str(), false, extra_search_path
,
2053 false, closure
->position_dependent_options());
2054 closure
->inputs()->add_file(file
);
2057 // Called by the bison parser to start a group. If we are already in
2058 // a group, that means that this script was invoked within a
2059 // --start-group --end-group sequence on the command line, or that
2060 // this script was found in a GROUP of another script. In that case,
2061 // we simply continue the existing group, rather than starting a new
2062 // one. It is possible to construct a case in which this will do
2063 // something other than what would happen if we did a recursive group,
2064 // but it's hard to imagine why the different behaviour would be
2065 // useful for a real program. Avoiding recursive groups is simpler
2066 // and more efficient.
2069 script_start_group(void* closurev
)
2071 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2072 if (!closure
->in_group())
2073 closure
->inputs()->start_group();
2076 // Called by the bison parser at the end of a group.
2079 script_end_group(void* closurev
)
2081 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2082 if (!closure
->in_group())
2083 closure
->inputs()->end_group();
2086 // Called by the bison parser to start an AS_NEEDED list.
2089 script_start_as_needed(void* closurev
)
2091 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2092 closure
->position_dependent_options().set_as_needed();
2095 // Called by the bison parser at the end of an AS_NEEDED list.
2098 script_end_as_needed(void* closurev
)
2100 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2101 closure
->position_dependent_options().clear_as_needed();
2104 // Called by the bison parser to set the entry symbol.
2107 script_set_entry(void* closurev
, const char* entry
, size_t length
)
2109 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2110 closure
->script_options()->set_entry(entry
, length
);
2113 // Called by the bison parser to define a symbol.
2116 script_set_symbol(void* closurev
, const char* name
, size_t length
,
2117 Expression
* value
, int providei
, int hiddeni
)
2119 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2120 const bool provide
= providei
!= 0;
2121 const bool hidden
= hiddeni
!= 0;
2122 closure
->script_options()->add_symbol_assignment(name
, length
, value
,
2126 // Called by the bison parser to add an assertion.
2129 script_add_assertion(void* closurev
, Expression
* check
, const char* message
,
2132 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2133 closure
->script_options()->add_assertion(check
, message
, messagelen
);
2136 // Called by the bison parser to parse an OPTION.
2139 script_parse_option(void* closurev
, const char* option
, size_t length
)
2141 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2142 // We treat the option as a single command-line option, even if
2143 // it has internal whitespace.
2144 if (closure
->command_line() == NULL
)
2146 // There are some options that we could handle here--e.g.,
2147 // -lLIBRARY. Should we bother?
2148 gold_warning(_("%s:%d:%d: ignoring command OPTION; OPTION is only valid"
2149 " for scripts specified via -T/--script"),
2150 closure
->filename(), closure
->lineno(), closure
->charpos());
2154 bool past_a_double_dash_option
= false;
2155 char* mutable_option
= strndup(option
, length
);
2156 gold_assert(mutable_option
!= NULL
);
2157 closure
->command_line()->process_one_option(1, &mutable_option
, 0,
2158 &past_a_double_dash_option
);
2159 free(mutable_option
);
2163 // Called by the bison parser to handle SEARCH_DIR. This is handled
2164 // exactly like a -L option.
2167 script_add_search_dir(void* closurev
, const char* option
, size_t length
)
2169 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2170 if (closure
->command_line() == NULL
)
2171 gold_warning(_("%s:%d:%d: ignoring SEARCH_DIR; SEARCH_DIR is only valid"
2172 " for scripts specified via -T/--script"),
2173 closure
->filename(), closure
->lineno(), closure
->charpos());
2176 std::string s
= "-L" + std::string(option
, length
);
2177 script_parse_option(closurev
, s
.c_str(), s
.size());
2181 /* Called by the bison parser to push the lexer into expression
2185 script_push_lex_into_expression_mode(void* closurev
)
2187 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2188 closure
->push_lex_mode(Lex::EXPRESSION
);
2191 /* Called by the bison parser to push the lexer into version
2195 script_push_lex_into_version_mode(void* closurev
)
2197 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2198 closure
->push_lex_mode(Lex::VERSION_SCRIPT
);
2201 /* Called by the bison parser to pop the lexer mode. */
2204 script_pop_lex_mode(void* closurev
)
2206 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2207 closure
->pop_lex_mode();
2210 // Register an entire version node. For example:
2216 // - tag is "GLIBC_2.1"
2217 // - tree contains the information "global: foo"
2218 // - deps contains "GLIBC_2.0"
2221 script_register_vers_node(void*,
2224 struct Version_tree
*tree
,
2225 struct Version_dependency_list
*deps
)
2227 gold_assert(tree
!= NULL
);
2228 gold_assert(tag
!= NULL
);
2229 tree
->dependencies
= deps
;
2230 tree
->tag
= std::string(tag
, taglen
);
2233 // Add a dependencies to the list of existing dependencies, if any,
2234 // and return the expanded list.
2236 extern "C" struct Version_dependency_list
*
2237 script_add_vers_depend(void* closurev
,
2238 struct Version_dependency_list
*all_deps
,
2239 const char *depend_to_add
, int deplen
)
2241 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2242 if (all_deps
== NULL
)
2243 all_deps
= closure
->version_script()->allocate_dependency_list();
2244 all_deps
->dependencies
.push_back(std::string(depend_to_add
, deplen
));
2248 // Add a pattern expression to an existing list of expressions, if any.
2249 // TODO: In the old linker, the last argument used to be a bool, but I
2250 // don't know what it meant.
2252 extern "C" struct Version_expression_list
*
2253 script_new_vers_pattern(void* closurev
,
2254 struct Version_expression_list
*expressions
,
2255 const char *pattern
, int patlen
, int exact_match
)
2257 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2258 if (expressions
== NULL
)
2259 expressions
= closure
->version_script()->allocate_expression_list();
2260 expressions
->expressions
.push_back(
2261 Version_expression(std::string(pattern
, patlen
),
2262 closure
->get_current_language(),
2263 static_cast<bool>(exact_match
)));
2267 // Attaches b to the end of a, and clears b. So a = a + b and b = {}.
2269 extern "C" struct Version_expression_list
*
2270 script_merge_expressions(struct Version_expression_list
*a
,
2271 struct Version_expression_list
*b
)
2273 a
->expressions
.insert(a
->expressions
.end(),
2274 b
->expressions
.begin(), b
->expressions
.end());
2275 // We could delete b and remove it from expressions_lists_, but
2276 // that's a lot of work. This works just as well.
2277 b
->expressions
.clear();
2281 // Combine the global and local expressions into a a Version_tree.
2283 extern "C" struct Version_tree
*
2284 script_new_vers_node(void* closurev
,
2285 struct Version_expression_list
*global
,
2286 struct Version_expression_list
*local
)
2288 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2289 Version_tree
* tree
= closure
->version_script()->allocate_version_tree();
2290 tree
->global
= global
;
2291 tree
->local
= local
;
2295 // Handle a transition in language, such as at the
2296 // start or end of 'extern "C++"'
2299 version_script_push_lang(void* closurev
, const char* lang
, int langlen
)
2301 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2302 closure
->push_language(std::string(lang
, langlen
));
2306 version_script_pop_lang(void* closurev
)
2308 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2309 closure
->pop_language();
2312 // Called by the bison parser to start a SECTIONS clause.
2315 script_start_sections(void* closurev
)
2317 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2318 closure
->script_options()->script_sections()->start_sections();
2321 // Called by the bison parser to finish a SECTIONS clause.
2324 script_finish_sections(void* closurev
)
2326 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2327 closure
->script_options()->script_sections()->finish_sections();
2330 // Start processing entries for an output section.
2333 script_start_output_section(void* closurev
, const char* name
, size_t namelen
,
2334 const struct Parser_output_section_header
* header
)
2336 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2337 closure
->script_options()->script_sections()->start_output_section(name
,
2342 // Finish processing entries for an output section.
2345 script_finish_output_section(void* closurev
,
2346 const struct Parser_output_section_trailer
* trail
)
2348 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2349 closure
->script_options()->script_sections()->finish_output_section(trail
);
2352 // Add a data item (e.g., "WORD (0)") to the current output section.
2355 script_add_data(void* closurev
, int data_token
, Expression
* val
)
2357 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2359 bool is_signed
= true;
2381 closure
->script_options()->script_sections()->add_data(size
, is_signed
, val
);
2384 // Add a clause setting the fill value to the current output section.
2387 script_add_fill(void* closurev
, Expression
* val
)
2389 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2390 closure
->script_options()->script_sections()->add_fill(val
);
2393 // Add a new input section specification to the current output
2397 script_add_input_section(void* closurev
,
2398 const struct Input_section_spec
* spec
,
2401 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2402 bool keep
= keepi
!= 0;
2403 closure
->script_options()->script_sections()->add_input_section(spec
, keep
);
2406 // Create a new list of string/sort pairs.
2408 extern "C" String_sort_list_ptr
2409 script_new_string_sort_list(const struct Wildcard_section
* string_sort
)
2411 return new String_sort_list(1, *string_sort
);
2414 // Add an entry to a list of string/sort pairs. The way the parser
2415 // works permits us to simply modify the first parameter, rather than
2418 extern "C" String_sort_list_ptr
2419 script_string_sort_list_add(String_sort_list_ptr pv
,
2420 const struct Wildcard_section
* string_sort
)
2423 return script_new_string_sort_list(string_sort
);
2426 pv
->push_back(*string_sort
);
2431 // Create a new list of strings.
2433 extern "C" String_list_ptr
2434 script_new_string_list(const char* str
, size_t len
)
2436 return new String_list(1, std::string(str
, len
));
2439 // Add an element to a list of strings. The way the parser works
2440 // permits us to simply modify the first parameter, rather than copy
2443 extern "C" String_list_ptr
2444 script_string_list_push_back(String_list_ptr pv
, const char* str
, size_t len
)
2447 return script_new_string_list(str
, len
);
2450 pv
->push_back(std::string(str
, len
));
2455 // Concatenate two string lists. Either or both may be NULL. The way
2456 // the parser works permits us to modify the parameters, rather than
2459 extern "C" String_list_ptr
2460 script_string_list_append(String_list_ptr pv1
, String_list_ptr pv2
)
2466 pv1
->insert(pv1
->end(), pv2
->begin(), pv2
->end());
2470 // Add a new program header.
2473 script_add_phdr(void* closurev
, const char* name
, size_t namelen
,
2474 unsigned int type
, const Phdr_info
* info
)
2476 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2477 bool includes_filehdr
= info
->includes_filehdr
!= 0;
2478 bool includes_phdrs
= info
->includes_phdrs
!= 0;
2479 bool is_flags_valid
= info
->is_flags_valid
!= 0;
2480 Script_sections
* ss
= closure
->script_options()->script_sections();
2481 ss
->add_phdr(name
, namelen
, type
, includes_filehdr
, includes_phdrs
,
2482 is_flags_valid
, info
->flags
, info
->load_address
);
2485 // Convert a program header string to a type.
2487 #define PHDR_TYPE(NAME) { #NAME, sizeof(#NAME) - 1, elfcpp::NAME }
2494 } phdr_type_names
[] =
2498 PHDR_TYPE(PT_DYNAMIC
),
2499 PHDR_TYPE(PT_INTERP
),
2501 PHDR_TYPE(PT_SHLIB
),
2504 PHDR_TYPE(PT_GNU_EH_FRAME
),
2505 PHDR_TYPE(PT_GNU_STACK
),
2506 PHDR_TYPE(PT_GNU_RELRO
)
2509 extern "C" unsigned int
2510 script_phdr_string_to_type(void* closurev
, const char* name
, size_t namelen
)
2512 for (unsigned int i
= 0;
2513 i
< sizeof(phdr_type_names
) / sizeof(phdr_type_names
[0]);
2515 if (namelen
== phdr_type_names
[i
].namelen
2516 && strncmp(name
, phdr_type_names
[i
].name
, namelen
) == 0)
2517 return phdr_type_names
[i
].val
;
2518 yyerror(closurev
, _("unknown PHDR type (try integer)"));
2519 return elfcpp::PT_NULL
;