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.
31 #include "filenames.h"
35 #include "dirsearch.h"
38 #include "workqueue.h"
40 #include "parameters.h"
43 #include "target-select.h"
50 // A token read from a script file. We don't implement keywords here;
51 // all keywords are simply represented as a string.
56 // Token classification.
61 // Token indicates end of input.
63 // Token is a string of characters.
65 // Token is a quoted string of characters.
67 // Token is an operator.
69 // Token is a number (an integer).
73 // We need an empty constructor so that we can put this STL objects.
75 : classification_(TOKEN_INVALID
), value_(NULL
), value_length_(0),
76 opcode_(0), lineno_(0), charpos_(0)
79 // A general token with no value.
80 Token(Classification classification
, int lineno
, int charpos
)
81 : classification_(classification
), value_(NULL
), value_length_(0),
82 opcode_(0), lineno_(lineno
), charpos_(charpos
)
84 gold_assert(classification
== TOKEN_INVALID
85 || classification
== TOKEN_EOF
);
88 // A general token with a value.
89 Token(Classification classification
, const char* value
, size_t length
,
90 int lineno
, int charpos
)
91 : classification_(classification
), value_(value
), value_length_(length
),
92 opcode_(0), lineno_(lineno
), charpos_(charpos
)
94 gold_assert(classification
!= TOKEN_INVALID
95 && classification
!= TOKEN_EOF
);
98 // A token representing an operator.
99 Token(int opcode
, int lineno
, int charpos
)
100 : classification_(TOKEN_OPERATOR
), value_(NULL
), value_length_(0),
101 opcode_(opcode
), lineno_(lineno
), charpos_(charpos
)
104 // Return whether the token is invalid.
107 { return this->classification_
== TOKEN_INVALID
; }
109 // Return whether this is an EOF token.
112 { return this->classification_
== TOKEN_EOF
; }
114 // Return the token classification.
116 classification() const
117 { return this->classification_
; }
119 // Return the line number at which the token starts.
122 { return this->lineno_
; }
124 // Return the character position at this the token starts.
127 { return this->charpos_
; }
129 // Get the value of a token.
132 string_value(size_t* length
) const
134 gold_assert(this->classification_
== TOKEN_STRING
135 || this->classification_
== TOKEN_QUOTED_STRING
);
136 *length
= this->value_length_
;
141 operator_value() const
143 gold_assert(this->classification_
== TOKEN_OPERATOR
);
144 return this->opcode_
;
148 integer_value() const
150 gold_assert(this->classification_
== TOKEN_INTEGER
);
152 std::string
s(this->value_
, this->value_length_
);
153 return strtoull(s
.c_str(), NULL
, 0);
157 // The token classification.
158 Classification classification_
;
159 // The token value, for TOKEN_STRING or TOKEN_QUOTED_STRING or
162 // The length of the token value.
163 size_t value_length_
;
164 // The token value, for TOKEN_OPERATOR.
166 // The line number where this token started (one based).
168 // The character position within the line where this token started
173 // This class handles lexing a file into a sequence of tokens.
178 // We unfortunately have to support different lexing modes, because
179 // when reading different parts of a linker script we need to parse
180 // things differently.
183 // Reading an ordinary linker script.
185 // Reading an expression in a linker script.
187 // Reading a version script.
189 // Reading a --dynamic-list file.
193 Lex(const char* input_string
, size_t input_length
, int parsing_token
)
194 : input_string_(input_string
), input_length_(input_length
),
195 current_(input_string
), mode_(LINKER_SCRIPT
),
196 first_token_(parsing_token
), token_(),
197 lineno_(1), linestart_(input_string
)
200 // Read a file into a string.
202 read_file(Input_file
*, std::string
*);
204 // Return the next token.
208 // Return the current lexing mode.
211 { return this->mode_
; }
213 // Set the lexing mode.
216 { this->mode_
= mode
; }
220 Lex
& operator=(const Lex
&);
222 // Make a general token with no value at the current location.
224 make_token(Token::Classification c
, const char* start
) const
225 { return Token(c
, this->lineno_
, start
- this->linestart_
+ 1); }
227 // Make a general token with a value at the current location.
229 make_token(Token::Classification c
, const char* v
, size_t len
,
232 { return Token(c
, v
, len
, this->lineno_
, start
- this->linestart_
+ 1); }
234 // Make an operator token at the current location.
236 make_token(int opcode
, const char* start
) const
237 { return Token(opcode
, this->lineno_
, start
- this->linestart_
+ 1); }
239 // Make an invalid token at the current location.
241 make_invalid_token(const char* start
)
242 { return this->make_token(Token::TOKEN_INVALID
, start
); }
244 // Make an EOF token at the current location.
246 make_eof_token(const char* start
)
247 { return this->make_token(Token::TOKEN_EOF
, start
); }
249 // Return whether C can be the first character in a name. C2 is the
250 // next character, since we sometimes need that.
252 can_start_name(char c
, char c2
);
254 // If C can appear in a name which has already started, return a
255 // pointer to a character later in the token or just past
256 // it. Otherwise, return NULL.
258 can_continue_name(const char* c
);
260 // Return whether C, C2, C3 can start a hex number.
262 can_start_hex(char c
, char c2
, char c3
);
264 // If C can appear in a hex number which has already started, return
265 // a pointer to a character later in the token or just past
266 // it. Otherwise, return NULL.
268 can_continue_hex(const char* c
);
270 // Return whether C can start a non-hex number.
272 can_start_number(char c
);
274 // If C can appear in a decimal number which has already started,
275 // return a pointer to a character later in the token or just past
276 // it. Otherwise, return NULL.
278 can_continue_number(const char* c
)
279 { return Lex::can_start_number(*c
) ? c
+ 1 : NULL
; }
281 // If C1 C2 C3 form a valid three character operator, return the
282 // opcode. Otherwise return 0.
284 three_char_operator(char c1
, char c2
, char c3
);
286 // If C1 C2 form a valid two character operator, return the opcode.
287 // Otherwise return 0.
289 two_char_operator(char c1
, char c2
);
291 // If C1 is a valid one character operator, return the opcode.
292 // Otherwise return 0.
294 one_char_operator(char c1
);
296 // Read the next token.
298 get_token(const char**);
300 // Skip a C style /* */ comment. Return false if the comment did
303 skip_c_comment(const char**);
305 // Skip a line # comment. Return false if there was no newline.
307 skip_line_comment(const char**);
309 // Build a token CLASSIFICATION from all characters that match
310 // CAN_CONTINUE_FN. The token starts at START. Start matching from
311 // MATCH. Set *PP to the character following the token.
313 gather_token(Token::Classification
,
314 const char* (Lex::*can_continue_fn
)(const char*),
315 const char* start
, const char* match
, const char** pp
);
317 // Build a token from a quoted string.
319 gather_quoted_string(const char** pp
);
321 // The string we are tokenizing.
322 const char* input_string_
;
323 // The length of the string.
324 size_t input_length_
;
325 // The current offset into the string.
326 const char* current_
;
327 // The current lexing mode.
329 // The code to use for the first token. This is set to 0 after it
332 // The current token.
334 // The current line number.
336 // The start of the current line in the string.
337 const char* linestart_
;
340 // Read the whole file into memory. We don't expect linker scripts to
341 // be large, so we just use a std::string as a buffer. We ignore the
342 // data we've already read, so that we read aligned buffers.
345 Lex::read_file(Input_file
* input_file
, std::string
* contents
)
347 off_t filesize
= input_file
->file().filesize();
349 contents
->reserve(filesize
);
352 unsigned char buf
[BUFSIZ
];
353 while (off
< filesize
)
356 if (get
> filesize
- off
)
357 get
= filesize
- off
;
358 input_file
->file().read(off
, get
, buf
);
359 contents
->append(reinterpret_cast<char*>(&buf
[0]), get
);
364 // Return whether C can be the start of a name, if the next character
365 // is C2. A name can being with a letter, underscore, period, or
366 // dollar sign. Because a name can be a file name, we also permit
367 // forward slash, backslash, and tilde. Tilde is the tricky case
368 // here; GNU ld also uses it as a bitwise not operator. It is only
369 // recognized as the operator if it is not immediately followed by
370 // some character which can appear in a symbol. That is, when we
371 // don't know that we are looking at an expression, "~0" is a file
372 // name, and "~ 0" is an expression using bitwise not. We are
376 Lex::can_start_name(char c
, char c2
)
380 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
381 case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
382 case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R':
383 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
385 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
386 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
387 case 'm': case 'n': case 'o': case 'q': case 'p': case 'r':
388 case 's': case 't': case 'u': case 'v': case 'w': case 'x':
390 case '_': case '.': case '$':
394 return this->mode_
== LINKER_SCRIPT
;
397 return this->mode_
== LINKER_SCRIPT
&& can_continue_name(&c2
);
400 return (this->mode_
== VERSION_SCRIPT
401 || this->mode_
== DYNAMIC_LIST
402 || (this->mode_
== LINKER_SCRIPT
403 && can_continue_name(&c2
)));
410 // Return whether C can continue a name which has already started.
411 // Subsequent characters in a name are the same as the leading
412 // characters, plus digits and "=+-:[],?*". So in general the linker
413 // script language requires spaces around operators, unless we know
414 // that we are parsing an expression.
417 Lex::can_continue_name(const char* c
)
421 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
422 case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
423 case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R':
424 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
426 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
427 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
428 case 'm': case 'n': case 'o': case 'q': case 'p': case 'r':
429 case 's': case 't': case 'u': case 'v': case 'w': case 'x':
431 case '_': case '.': case '$':
432 case '0': case '1': case '2': case '3': case '4':
433 case '5': case '6': case '7': case '8': case '9':
436 // TODO(csilvers): why not allow ~ in names for version-scripts?
437 case '/': case '\\': case '~':
440 if (this->mode_
== LINKER_SCRIPT
)
444 case '[': case ']': case '*': case '?': case '-':
445 if (this->mode_
== LINKER_SCRIPT
|| this->mode_
== VERSION_SCRIPT
446 || this->mode_
== DYNAMIC_LIST
)
450 // TODO(csilvers): why allow this? ^ is meaningless in version scripts.
452 if (this->mode_
== VERSION_SCRIPT
|| this->mode_
== DYNAMIC_LIST
)
457 if (this->mode_
== LINKER_SCRIPT
)
459 else if ((this->mode_
== VERSION_SCRIPT
|| this->mode_
== DYNAMIC_LIST
)
462 // A name can have '::' in it, as that's a c++ namespace
463 // separator. But a single colon is not part of a name.
473 // For a number we accept 0x followed by hex digits, or any sequence
474 // of digits. The old linker accepts leading '$' for hex, and
475 // trailing HXBOD. Those are for MRI compatibility and we don't
476 // accept them. The old linker also accepts trailing MK for mega or
477 // kilo. FIXME: Those are mentioned in the documentation, and we
478 // should accept them.
480 // Return whether C1 C2 C3 can start a hex number.
483 Lex::can_start_hex(char c1
, char c2
, char c3
)
485 if (c1
== '0' && (c2
== 'x' || c2
== 'X'))
486 return this->can_continue_hex(&c3
);
490 // Return whether C can appear in a hex number.
493 Lex::can_continue_hex(const char* c
)
497 case '0': case '1': case '2': case '3': case '4':
498 case '5': case '6': case '7': case '8': case '9':
499 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
500 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
508 // Return whether C can start a non-hex number.
511 Lex::can_start_number(char c
)
515 case '0': case '1': case '2': case '3': case '4':
516 case '5': case '6': case '7': case '8': case '9':
524 // If C1 C2 C3 form a valid three character operator, return the
525 // opcode (defined in the yyscript.h file generated from yyscript.y).
526 // Otherwise return 0.
529 Lex::three_char_operator(char c1
, char c2
, char c3
)
534 if (c2
== '<' && c3
== '=')
538 if (c2
== '>' && c3
== '=')
547 // If C1 C2 form a valid two character operator, return the opcode
548 // (defined in the yyscript.h file generated from yyscript.y).
549 // Otherwise return 0.
552 Lex::two_char_operator(char c1
, char c2
)
610 // If C1 is a valid operator, return the opcode. Otherwise return 0.
613 Lex::one_char_operator(char c1
)
646 // Skip a C style comment. *PP points to just after the "/*". Return
647 // false if the comment did not end.
650 Lex::skip_c_comment(const char** pp
)
653 while (p
[0] != '*' || p
[1] != '/')
664 this->linestart_
= p
+ 1;
673 // Skip a line # comment. Return false if there was no newline.
676 Lex::skip_line_comment(const char** pp
)
679 size_t skip
= strcspn(p
, "\n");
688 this->linestart_
= p
;
694 // Build a token CLASSIFICATION from all characters that match
695 // CAN_CONTINUE_FN. Update *PP.
698 Lex::gather_token(Token::Classification classification
,
699 const char* (Lex::*can_continue_fn
)(const char*),
704 const char* new_match
= NULL
;
705 while ((new_match
= (this->*can_continue_fn
)(match
)))
708 return this->make_token(classification
, start
, match
- start
, start
);
711 // Build a token from a quoted string.
714 Lex::gather_quoted_string(const char** pp
)
716 const char* start
= *pp
;
717 const char* p
= start
;
719 size_t skip
= strcspn(p
, "\"\n");
721 return this->make_invalid_token(start
);
723 return this->make_token(Token::TOKEN_QUOTED_STRING
, p
, skip
, start
);
726 // Return the next token at *PP. Update *PP. General guideline: we
727 // require linker scripts to be simple ASCII. No unicode linker
728 // scripts. In particular we can assume that any '\0' is the end of
732 Lex::get_token(const char** pp
)
741 return this->make_eof_token(p
);
744 // Skip whitespace quickly.
745 while (*p
== ' ' || *p
== '\t')
752 this->linestart_
= p
;
756 // Skip C style comments.
757 if (p
[0] == '/' && p
[1] == '*')
759 int lineno
= this->lineno_
;
760 int charpos
= p
- this->linestart_
+ 1;
763 if (!this->skip_c_comment(pp
))
764 return Token(Token::TOKEN_INVALID
, lineno
, charpos
);
770 // Skip line comments.
774 if (!this->skip_line_comment(pp
))
775 return this->make_eof_token(p
);
781 if (this->can_start_name(p
[0], p
[1]))
782 return this->gather_token(Token::TOKEN_STRING
,
783 &Lex::can_continue_name
,
786 // We accept any arbitrary name in double quotes, as long as it
787 // does not cross a line boundary.
791 return this->gather_quoted_string(pp
);
794 // Check for a number.
796 if (this->can_start_hex(p
[0], p
[1], p
[2]))
797 return this->gather_token(Token::TOKEN_INTEGER
,
798 &Lex::can_continue_hex
,
801 if (Lex::can_start_number(p
[0]))
802 return this->gather_token(Token::TOKEN_INTEGER
,
803 &Lex::can_continue_number
,
806 // Check for operators.
808 int opcode
= Lex::three_char_operator(p
[0], p
[1], p
[2]);
812 return this->make_token(opcode
, p
);
815 opcode
= Lex::two_char_operator(p
[0], p
[1]);
819 return this->make_token(opcode
, p
);
822 opcode
= Lex::one_char_operator(p
[0]);
826 return this->make_token(opcode
, p
);
829 return this->make_token(Token::TOKEN_INVALID
, p
);
833 // Return the next token.
838 // The first token is special.
839 if (this->first_token_
!= 0)
841 this->token_
= Token(this->first_token_
, 0, 0);
842 this->first_token_
= 0;
843 return &this->token_
;
846 this->token_
= this->get_token(&this->current_
);
848 // Don't let an early null byte fool us into thinking that we've
849 // reached the end of the file.
850 if (this->token_
.is_eof()
851 && (static_cast<size_t>(this->current_
- this->input_string_
)
852 < this->input_length_
))
853 this->token_
= this->make_invalid_token(this->current_
);
855 return &this->token_
;
858 // class Symbol_assignment.
860 // Add the symbol to the symbol table. This makes sure the symbol is
861 // there and defined. The actual value is stored later. We can't
862 // determine the actual value at this point, because we can't
863 // necessarily evaluate the expression until all ordinary symbols have
866 // The GNU linker lets symbol assignments in the linker script
867 // silently override defined symbols in object files. We are
868 // compatible. FIXME: Should we issue a warning?
871 Symbol_assignment::add_to_table(Symbol_table
* symtab
)
873 elfcpp::STV vis
= this->hidden_
? elfcpp::STV_HIDDEN
: elfcpp::STV_DEFAULT
;
874 this->sym_
= symtab
->define_as_constant(this->name_
.c_str(),
883 true); // force_override
886 // Finalize a symbol value.
889 Symbol_assignment::finalize(Symbol_table
* symtab
, const Layout
* layout
)
891 this->finalize_maybe_dot(symtab
, layout
, false, 0, NULL
);
894 // Finalize a symbol value which can refer to the dot symbol.
897 Symbol_assignment::finalize_with_dot(Symbol_table
* symtab
,
898 const Layout
* layout
,
900 Output_section
* dot_section
)
902 this->finalize_maybe_dot(symtab
, layout
, true, dot_value
, dot_section
);
905 // Finalize a symbol value, internal version.
908 Symbol_assignment::finalize_maybe_dot(Symbol_table
* symtab
,
909 const Layout
* layout
,
910 bool is_dot_available
,
912 Output_section
* dot_section
)
914 // If we were only supposed to provide this symbol, the sym_ field
915 // will be NULL if the symbol was not referenced.
916 if (this->sym_
== NULL
)
918 gold_assert(this->provide_
);
922 if (parameters
->target().get_size() == 32)
924 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
925 this->sized_finalize
<32>(symtab
, layout
, is_dot_available
, dot_value
,
931 else if (parameters
->target().get_size() == 64)
933 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
934 this->sized_finalize
<64>(symtab
, layout
, is_dot_available
, dot_value
,
946 Symbol_assignment::sized_finalize(Symbol_table
* symtab
, const Layout
* layout
,
947 bool is_dot_available
, uint64_t dot_value
,
948 Output_section
* dot_section
)
950 Output_section
* section
;
951 uint64_t final_val
= this->val_
->eval_maybe_dot(symtab
, layout
, true,
953 dot_value
, dot_section
,
955 Sized_symbol
<size
>* ssym
= symtab
->get_sized_symbol
<size
>(this->sym_
);
956 ssym
->set_value(final_val
);
958 ssym
->set_output_section(section
);
961 // Set the symbol value if the expression yields an absolute value.
964 Symbol_assignment::set_if_absolute(Symbol_table
* symtab
, const Layout
* layout
,
965 bool is_dot_available
, uint64_t dot_value
)
967 if (this->sym_
== NULL
)
970 Output_section
* val_section
;
971 uint64_t val
= this->val_
->eval_maybe_dot(symtab
, layout
, false,
972 is_dot_available
, dot_value
,
974 if (val_section
!= NULL
)
977 if (parameters
->target().get_size() == 32)
979 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
980 Sized_symbol
<32>* ssym
= symtab
->get_sized_symbol
<32>(this->sym_
);
981 ssym
->set_value(val
);
986 else if (parameters
->target().get_size() == 64)
988 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
989 Sized_symbol
<64>* ssym
= symtab
->get_sized_symbol
<64>(this->sym_
);
990 ssym
->set_value(val
);
999 // Print for debugging.
1002 Symbol_assignment::print(FILE* f
) const
1004 if (this->provide_
&& this->hidden_
)
1005 fprintf(f
, "PROVIDE_HIDDEN(");
1006 else if (this->provide_
)
1007 fprintf(f
, "PROVIDE(");
1008 else if (this->hidden_
)
1011 fprintf(f
, "%s = ", this->name_
.c_str());
1012 this->val_
->print(f
);
1014 if (this->provide_
|| this->hidden_
)
1020 // Class Script_assertion.
1022 // Check the assertion.
1025 Script_assertion::check(const Symbol_table
* symtab
, const Layout
* layout
)
1027 if (!this->check_
->eval(symtab
, layout
, true))
1028 gold_error("%s", this->message_
.c_str());
1031 // Print for debugging.
1034 Script_assertion::print(FILE* f
) const
1036 fprintf(f
, "ASSERT(");
1037 this->check_
->print(f
);
1038 fprintf(f
, ", \"%s\")\n", this->message_
.c_str());
1041 // Class Script_options.
1043 Script_options::Script_options()
1044 : entry_(), symbol_assignments_(), version_script_info_(),
1049 // Add a symbol to be defined.
1052 Script_options::add_symbol_assignment(const char* name
, size_t length
,
1053 Expression
* value
, bool provide
,
1056 if (length
!= 1 || name
[0] != '.')
1058 if (this->script_sections_
.in_sections_clause())
1059 this->script_sections_
.add_symbol_assignment(name
, length
, value
,
1063 Symbol_assignment
* p
= new Symbol_assignment(name
, length
, value
,
1065 this->symbol_assignments_
.push_back(p
);
1070 if (provide
|| hidden
)
1071 gold_error(_("invalid use of PROVIDE for dot symbol"));
1072 if (!this->script_sections_
.in_sections_clause())
1073 gold_error(_("invalid assignment to dot outside of SECTIONS"));
1075 this->script_sections_
.add_dot_assignment(value
);
1079 // Add an assertion.
1082 Script_options::add_assertion(Expression
* check
, const char* message
,
1085 if (this->script_sections_
.in_sections_clause())
1086 this->script_sections_
.add_assertion(check
, message
, messagelen
);
1089 Script_assertion
* p
= new Script_assertion(check
, message
, messagelen
);
1090 this->assertions_
.push_back(p
);
1094 // Create sections required by any linker scripts.
1097 Script_options::create_script_sections(Layout
* layout
)
1099 if (this->saw_sections_clause())
1100 this->script_sections_
.create_sections(layout
);
1103 // Add any symbols we are defining to the symbol table.
1106 Script_options::add_symbols_to_table(Symbol_table
* symtab
)
1108 for (Symbol_assignments::iterator p
= this->symbol_assignments_
.begin();
1109 p
!= this->symbol_assignments_
.end();
1111 (*p
)->add_to_table(symtab
);
1112 this->script_sections_
.add_symbols_to_table(symtab
);
1115 // Finalize symbol values. Also check assertions.
1118 Script_options::finalize_symbols(Symbol_table
* symtab
, const Layout
* layout
)
1120 // We finalize the symbols defined in SECTIONS first, because they
1121 // are the ones which may have changed. This way if symbol outside
1122 // SECTIONS are defined in terms of symbols inside SECTIONS, they
1123 // will get the right value.
1124 this->script_sections_
.finalize_symbols(symtab
, layout
);
1126 for (Symbol_assignments::iterator p
= this->symbol_assignments_
.begin();
1127 p
!= this->symbol_assignments_
.end();
1129 (*p
)->finalize(symtab
, layout
);
1131 for (Assertions::iterator p
= this->assertions_
.begin();
1132 p
!= this->assertions_
.end();
1134 (*p
)->check(symtab
, layout
);
1137 // Set section addresses. We set all the symbols which have absolute
1138 // values. Then we let the SECTIONS clause do its thing. This
1139 // returns the segment which holds the file header and segment
1143 Script_options::set_section_addresses(Symbol_table
* symtab
, Layout
* layout
)
1145 for (Symbol_assignments::iterator p
= this->symbol_assignments_
.begin();
1146 p
!= this->symbol_assignments_
.end();
1148 (*p
)->set_if_absolute(symtab
, layout
, false, 0);
1150 return this->script_sections_
.set_section_addresses(symtab
, layout
);
1153 // This class holds data passed through the parser to the lexer and to
1154 // the parser support functions. This avoids global variables. We
1155 // can't use global variables because we need not be called by a
1156 // singleton thread.
1158 class Parser_closure
1161 Parser_closure(const char* filename
,
1162 const Position_dependent_options
& posdep_options
,
1163 bool in_group
, bool is_in_sysroot
,
1164 Command_line
* command_line
,
1165 Script_options
* script_options
,
1167 bool skip_on_incompatible_target
)
1168 : filename_(filename
), posdep_options_(posdep_options
),
1169 in_group_(in_group
), is_in_sysroot_(is_in_sysroot
),
1170 skip_on_incompatible_target_(skip_on_incompatible_target
),
1171 found_incompatible_target_(false),
1172 command_line_(command_line
), script_options_(script_options
),
1173 version_script_info_(script_options
->version_script_info()),
1174 lex_(lex
), lineno_(0), charpos_(0), lex_mode_stack_(), inputs_(NULL
)
1176 // We start out processing C symbols in the default lex mode.
1177 language_stack_
.push_back("");
1178 lex_mode_stack_
.push_back(lex
->mode());
1181 // Return the file name.
1184 { return this->filename_
; }
1186 // Return the position dependent options. The caller may modify
1188 Position_dependent_options
&
1189 position_dependent_options()
1190 { return this->posdep_options_
; }
1192 // Return whether this script is being run in a group.
1195 { return this->in_group_
; }
1197 // Return whether this script was found using a directory in the
1200 is_in_sysroot() const
1201 { return this->is_in_sysroot_
; }
1203 // Whether to skip to the next file with the same name if we find an
1204 // incompatible target in an OUTPUT_FORMAT statement.
1206 skip_on_incompatible_target() const
1207 { return this->skip_on_incompatible_target_
; }
1209 // Stop skipping to the next flie on an incompatible target. This
1210 // is called when we make some unrevocable change to the data
1213 clear_skip_on_incompatible_target()
1214 { this->skip_on_incompatible_target_
= false; }
1216 // Whether we found an incompatible target in an OUTPUT_FORMAT
1219 found_incompatible_target() const
1220 { return this->found_incompatible_target_
; }
1222 // Note that we found an incompatible target.
1224 set_found_incompatible_target()
1225 { this->found_incompatible_target_
= true; }
1227 // Returns the Command_line structure passed in at constructor time.
1228 // This value may be NULL. The caller may modify this, which modifies
1229 // the passed-in Command_line object (not a copy).
1232 { return this->command_line_
; }
1234 // Return the options which may be set by a script.
1237 { return this->script_options_
; }
1239 // Return the object in which version script information should be stored.
1240 Version_script_info
*
1242 { return this->version_script_info_
; }
1244 // Return the next token, and advance.
1248 const Token
* token
= this->lex_
->next_token();
1249 this->lineno_
= token
->lineno();
1250 this->charpos_
= token
->charpos();
1254 // Set a new lexer mode, pushing the current one.
1256 push_lex_mode(Lex::Mode mode
)
1258 this->lex_mode_stack_
.push_back(this->lex_
->mode());
1259 this->lex_
->set_mode(mode
);
1262 // Pop the lexer mode.
1266 gold_assert(!this->lex_mode_stack_
.empty());
1267 this->lex_
->set_mode(this->lex_mode_stack_
.back());
1268 this->lex_mode_stack_
.pop_back();
1271 // Return the current lexer mode.
1274 { return this->lex_mode_stack_
.back(); }
1276 // Return the line number of the last token.
1279 { return this->lineno_
; }
1281 // Return the character position in the line of the last token.
1284 { return this->charpos_
; }
1286 // Return the list of input files, creating it if necessary. This
1287 // is a space leak--we never free the INPUTS_ pointer.
1291 if (this->inputs_
== NULL
)
1292 this->inputs_
= new Input_arguments();
1293 return this->inputs_
;
1296 // Return whether we saw any input files.
1299 { return this->inputs_
!= NULL
&& !this->inputs_
->empty(); }
1301 // Return the current language being processed in a version script
1302 // (eg, "C++"). The empty string represents unmangled C names.
1304 get_current_language() const
1305 { return this->language_stack_
.back(); }
1307 // Push a language onto the stack when entering an extern block.
1308 void push_language(const std::string
& lang
)
1309 { this->language_stack_
.push_back(lang
); }
1311 // Pop a language off of the stack when exiting an extern block.
1314 gold_assert(!this->language_stack_
.empty());
1315 this->language_stack_
.pop_back();
1319 // The name of the file we are reading.
1320 const char* filename_
;
1321 // The position dependent options.
1322 Position_dependent_options posdep_options_
;
1323 // Whether we are currently in a --start-group/--end-group.
1325 // Whether the script was found in a sysrooted directory.
1326 bool is_in_sysroot_
;
1327 // If this is true, then if we find an OUTPUT_FORMAT with an
1328 // incompatible target, then we tell the parser to abort so that we
1329 // can search for the next file with the same name.
1330 bool skip_on_incompatible_target_
;
1331 // True if we found an OUTPUT_FORMAT with an incompatible target.
1332 bool found_incompatible_target_
;
1333 // May be NULL if the user chooses not to pass one in.
1334 Command_line
* command_line_
;
1335 // Options which may be set from any linker script.
1336 Script_options
* script_options_
;
1337 // Information parsed from a version script.
1338 Version_script_info
* version_script_info_
;
1341 // The line number of the last token returned by next_token.
1343 // The column number of the last token returned by next_token.
1345 // A stack of lexer modes.
1346 std::vector
<Lex::Mode
> lex_mode_stack_
;
1347 // A stack of which extern/language block we're inside. Can be C++,
1348 // java, or empty for C.
1349 std::vector
<std::string
> language_stack_
;
1350 // New input files found to add to the link.
1351 Input_arguments
* inputs_
;
1354 // FILE was found as an argument on the command line. Try to read it
1355 // as a script. Return true if the file was handled.
1358 read_input_script(Workqueue
* workqueue
, Symbol_table
* symtab
, Layout
* layout
,
1359 Dirsearch
* dirsearch
, int dirindex
,
1360 Input_objects
* input_objects
, Mapfile
* mapfile
,
1361 Input_group
* input_group
,
1362 const Input_argument
* input_argument
,
1363 Input_file
* input_file
, Task_token
* next_blocker
,
1364 bool* used_next_blocker
)
1366 *used_next_blocker
= false;
1368 std::string input_string
;
1369 Lex::read_file(input_file
, &input_string
);
1371 Lex
lex(input_string
.c_str(), input_string
.length(), PARSING_LINKER_SCRIPT
);
1373 Parser_closure
closure(input_file
->filename().c_str(),
1374 input_argument
->file().options(),
1375 input_group
!= NULL
,
1376 input_file
->is_in_sysroot(),
1378 layout
->script_options(),
1380 input_file
->will_search_for());
1382 if (yyparse(&closure
) != 0)
1384 if (closure
.found_incompatible_target())
1386 Read_symbols::incompatible_warning(input_argument
, input_file
);
1387 Read_symbols::requeue(workqueue
, input_objects
, symtab
, layout
,
1388 dirsearch
, dirindex
, mapfile
, input_argument
,
1389 input_group
, next_blocker
);
1395 if (!closure
.saw_inputs())
1398 Task_token
* this_blocker
= NULL
;
1399 for (Input_arguments::const_iterator p
= closure
.inputs()->begin();
1400 p
!= closure
.inputs()->end();
1404 if (p
+ 1 == closure
.inputs()->end())
1408 nb
= new Task_token(true);
1411 workqueue
->queue_soon(new Read_symbols(input_objects
, symtab
,
1412 layout
, dirsearch
, 0, mapfile
, &*p
,
1413 input_group
, this_blocker
, nb
));
1417 *used_next_blocker
= true;
1422 // Helper function for read_version_script() and
1423 // read_commandline_script(). Processes the given file in the mode
1424 // indicated by first_token and lex_mode.
1427 read_script_file(const char* filename
, Command_line
* cmdline
,
1428 Script_options
* script_options
,
1429 int first_token
, Lex::Mode lex_mode
)
1431 // TODO: if filename is a relative filename, search for it manually
1432 // using "." + cmdline->options()->search_path() -- not dirsearch.
1433 Dirsearch dirsearch
;
1435 // The file locking code wants to record a Task, but we haven't
1436 // started the workqueue yet. This is only for debugging purposes,
1437 // so we invent a fake value.
1438 const Task
* task
= reinterpret_cast<const Task
*>(-1);
1440 // We don't want this file to be opened in binary mode.
1441 Position_dependent_options posdep
= cmdline
->position_dependent_options();
1442 if (posdep
.format_enum() == General_options::OBJECT_FORMAT_BINARY
)
1443 posdep
.set_format_enum(General_options::OBJECT_FORMAT_ELF
);
1444 Input_file_argument
input_argument(filename
, false, "", false, posdep
);
1445 Input_file
input_file(&input_argument
);
1447 if (!input_file
.open(dirsearch
, task
, &dummy
))
1450 std::string input_string
;
1451 Lex::read_file(&input_file
, &input_string
);
1453 Lex
lex(input_string
.c_str(), input_string
.length(), first_token
);
1454 lex
.set_mode(lex_mode
);
1456 Parser_closure
closure(filename
,
1457 cmdline
->position_dependent_options(),
1459 input_file
.is_in_sysroot(),
1464 if (yyparse(&closure
) != 0)
1466 input_file
.file().unlock(task
);
1470 input_file
.file().unlock(task
);
1472 gold_assert(!closure
.saw_inputs());
1477 // FILENAME was found as an argument to --script (-T).
1478 // Read it as a script, and execute its contents immediately.
1481 read_commandline_script(const char* filename
, Command_line
* cmdline
)
1483 return read_script_file(filename
, cmdline
, &cmdline
->script_options(),
1484 PARSING_LINKER_SCRIPT
, Lex::LINKER_SCRIPT
);
1487 // FILENAME was found as an argument to --version-script. Read it as
1488 // a version script, and store its contents in
1489 // cmdline->script_options()->version_script_info().
1492 read_version_script(const char* filename
, Command_line
* cmdline
)
1494 return read_script_file(filename
, cmdline
, &cmdline
->script_options(),
1495 PARSING_VERSION_SCRIPT
, Lex::VERSION_SCRIPT
);
1498 // FILENAME was found as an argument to --dynamic-list. Read it as a
1499 // list of symbols, and store its contents in DYNAMIC_LIST.
1502 read_dynamic_list(const char* filename
, Command_line
* cmdline
,
1503 Script_options
* dynamic_list
)
1505 return read_script_file(filename
, cmdline
, dynamic_list
,
1506 PARSING_DYNAMIC_LIST
, Lex::DYNAMIC_LIST
);
1509 // Implement the --defsym option on the command line. Return true if
1513 Script_options::define_symbol(const char* definition
)
1515 Lex
lex(definition
, strlen(definition
), PARSING_DEFSYM
);
1516 lex
.set_mode(Lex::EXPRESSION
);
1519 Position_dependent_options posdep_options
;
1521 Parser_closure
closure("command line", posdep_options
, false, false, NULL
,
1524 if (yyparse(&closure
) != 0)
1527 gold_assert(!closure
.saw_inputs());
1532 // Print the script to F for debugging.
1535 Script_options::print(FILE* f
) const
1537 fprintf(f
, "%s: Dumping linker script\n", program_name
);
1539 if (!this->entry_
.empty())
1540 fprintf(f
, "ENTRY(%s)\n", this->entry_
.c_str());
1542 for (Symbol_assignments::const_iterator p
=
1543 this->symbol_assignments_
.begin();
1544 p
!= this->symbol_assignments_
.end();
1548 for (Assertions::const_iterator p
= this->assertions_
.begin();
1549 p
!= this->assertions_
.end();
1553 this->script_sections_
.print(f
);
1555 this->version_script_info_
.print(f
);
1558 // Manage mapping from keywords to the codes expected by the bison
1559 // parser. We construct one global object for each lex mode with
1562 class Keyword_to_parsecode
1565 // The structure which maps keywords to parsecodes.
1566 struct Keyword_parsecode
1569 const char* keyword
;
1570 // Corresponding parsecode.
1574 Keyword_to_parsecode(const Keyword_parsecode
* keywords
,
1576 : keyword_parsecodes_(keywords
), keyword_count_(keyword_count
)
1579 // Return the parsecode corresponding KEYWORD, or 0 if it is not a
1582 keyword_to_parsecode(const char* keyword
, size_t len
) const;
1585 const Keyword_parsecode
* keyword_parsecodes_
;
1586 const int keyword_count_
;
1589 // Mapping from keyword string to keyword parsecode. This array must
1590 // be kept in sorted order. Parsecodes are looked up using bsearch.
1591 // This array must correspond to the list of parsecodes in yyscript.y.
1593 static const Keyword_to_parsecode::Keyword_parsecode
1594 script_keyword_parsecodes
[] =
1596 { "ABSOLUTE", ABSOLUTE
},
1598 { "ALIGN", ALIGN_K
},
1599 { "ALIGNOF", ALIGNOF
},
1600 { "ASSERT", ASSERT_K
},
1601 { "AS_NEEDED", AS_NEEDED
},
1606 { "CONSTANT", CONSTANT
},
1607 { "CONSTRUCTORS", CONSTRUCTORS
},
1608 { "CREATE_OBJECT_SYMBOLS", CREATE_OBJECT_SYMBOLS
},
1609 { "DATA_SEGMENT_ALIGN", DATA_SEGMENT_ALIGN
},
1610 { "DATA_SEGMENT_END", DATA_SEGMENT_END
},
1611 { "DATA_SEGMENT_RELRO_END", DATA_SEGMENT_RELRO_END
},
1612 { "DEFINED", DEFINED
},
1614 { "EXCLUDE_FILE", EXCLUDE_FILE
},
1615 { "EXTERN", EXTERN
},
1618 { "FORCE_COMMON_ALLOCATION", FORCE_COMMON_ALLOCATION
},
1621 { "INCLUDE", INCLUDE
},
1622 { "INHIBIT_COMMON_ALLOCATION", INHIBIT_COMMON_ALLOCATION
},
1625 { "LENGTH", LENGTH
},
1626 { "LOADADDR", LOADADDR
},
1630 { "MEMORY", MEMORY
},
1633 { "NOCROSSREFS", NOCROSSREFS
},
1634 { "NOFLOAT", NOFLOAT
},
1635 { "ONLY_IF_RO", ONLY_IF_RO
},
1636 { "ONLY_IF_RW", ONLY_IF_RW
},
1637 { "OPTION", OPTION
},
1638 { "ORIGIN", ORIGIN
},
1639 { "OUTPUT", OUTPUT
},
1640 { "OUTPUT_ARCH", OUTPUT_ARCH
},
1641 { "OUTPUT_FORMAT", OUTPUT_FORMAT
},
1642 { "OVERLAY", OVERLAY
},
1644 { "PROVIDE", PROVIDE
},
1645 { "PROVIDE_HIDDEN", PROVIDE_HIDDEN
},
1647 { "SEARCH_DIR", SEARCH_DIR
},
1648 { "SECTIONS", SECTIONS
},
1649 { "SEGMENT_START", SEGMENT_START
},
1651 { "SIZEOF", SIZEOF
},
1652 { "SIZEOF_HEADERS", SIZEOF_HEADERS
},
1653 { "SORT", SORT_BY_NAME
},
1654 { "SORT_BY_ALIGNMENT", SORT_BY_ALIGNMENT
},
1655 { "SORT_BY_NAME", SORT_BY_NAME
},
1656 { "SPECIAL", SPECIAL
},
1658 { "STARTUP", STARTUP
},
1659 { "SUBALIGN", SUBALIGN
},
1660 { "SYSLIB", SYSLIB
},
1661 { "TARGET", TARGET_K
},
1662 { "TRUNCATE", TRUNCATE
},
1663 { "VERSION", VERSIONK
},
1664 { "global", GLOBAL
},
1670 { "sizeof_headers", SIZEOF_HEADERS
},
1673 static const Keyword_to_parsecode
1674 script_keywords(&script_keyword_parsecodes
[0],
1675 (sizeof(script_keyword_parsecodes
)
1676 / sizeof(script_keyword_parsecodes
[0])));
1678 static const Keyword_to_parsecode::Keyword_parsecode
1679 version_script_keyword_parsecodes
[] =
1681 { "extern", EXTERN
},
1682 { "global", GLOBAL
},
1686 static const Keyword_to_parsecode
1687 version_script_keywords(&version_script_keyword_parsecodes
[0],
1688 (sizeof(version_script_keyword_parsecodes
)
1689 / sizeof(version_script_keyword_parsecodes
[0])));
1691 static const Keyword_to_parsecode::Keyword_parsecode
1692 dynamic_list_keyword_parsecodes
[] =
1694 { "extern", EXTERN
},
1697 static const Keyword_to_parsecode
1698 dynamic_list_keywords(&dynamic_list_keyword_parsecodes
[0],
1699 (sizeof(dynamic_list_keyword_parsecodes
)
1700 / sizeof(dynamic_list_keyword_parsecodes
[0])));
1704 // Comparison function passed to bsearch.
1716 ktt_compare(const void* keyv
, const void* kttv
)
1718 const Ktt_key
* key
= static_cast<const Ktt_key
*>(keyv
);
1719 const Keyword_to_parsecode::Keyword_parsecode
* ktt
=
1720 static_cast<const Keyword_to_parsecode::Keyword_parsecode
*>(kttv
);
1721 int i
= strncmp(key
->str
, ktt
->keyword
, key
->len
);
1724 if (ktt
->keyword
[key
->len
] != '\0')
1729 } // End extern "C".
1732 Keyword_to_parsecode::keyword_to_parsecode(const char* keyword
,
1738 void* kttv
= bsearch(&key
,
1739 this->keyword_parsecodes_
,
1740 this->keyword_count_
,
1741 sizeof(this->keyword_parsecodes_
[0]),
1745 Keyword_parsecode
* ktt
= static_cast<Keyword_parsecode
*>(kttv
);
1746 return ktt
->parsecode
;
1749 // Helper class that calls cplus_demangle when needed and takes care of freeing
1752 class Lazy_demangler
1755 Lazy_demangler(const char* symbol
, int options
)
1756 : symbol_(symbol
), options_(options
), demangled_(NULL
), did_demangle_(false)
1760 { free(this->demangled_
); }
1762 // Return the demangled name. The actual demangling happens on the first call,
1763 // and the result is later cached.
1769 // The symbol to demangle.
1770 const char *symbol_
;
1771 // Option flags to pass to cplus_demagle.
1773 // The cached demangled value, or NULL if demangling didn't happen yet or
1776 // Whether we already called cplus_demangle
1780 // Return the demangled name. The actual demangling happens on the first call,
1781 // and the result is later cached. Returns NULL if the symbol cannot be
1785 Lazy_demangler::get()
1787 if (!this->did_demangle_
)
1789 this->demangled_
= cplus_demangle(this->symbol_
, this->options_
);
1790 this->did_demangle_
= true;
1792 return this->demangled_
;
1795 // The following structs are used within the VersionInfo class as well
1796 // as in the bison helper functions. They store the information
1797 // parsed from the version script.
1799 // A single version expression.
1800 // For example, pattern="std::map*" and language="C++".
1801 // pattern and language should be from the stringpool
1802 struct Version_expression
{
1803 Version_expression(const std::string
& pattern
,
1804 const std::string
& language
,
1806 : pattern(pattern
), language(language
), exact_match(exact_match
) {}
1808 std::string pattern
;
1809 std::string language
;
1810 // If false, we use glob() to match pattern. If true, we use strcmp().
1815 // A list of expressions.
1816 struct Version_expression_list
{
1817 std::vector
<struct Version_expression
> expressions
;
1821 // A list of which versions upon which another version depends.
1822 // Strings should be from the Stringpool.
1823 struct Version_dependency_list
{
1824 std::vector
<std::string
> dependencies
;
1828 // The total definition of a version. It includes the tag for the
1829 // version, its global and local expressions, and any dependencies.
1830 struct Version_tree
{
1832 : tag(), global(NULL
), local(NULL
), dependencies(NULL
) {}
1835 const struct Version_expression_list
* global
;
1836 const struct Version_expression_list
* local
;
1837 const struct Version_dependency_list
* dependencies
;
1840 Version_script_info::~Version_script_info()
1846 Version_script_info::clear()
1848 for (size_t k
= 0; k
< dependency_lists_
.size(); ++k
)
1849 delete dependency_lists_
[k
];
1850 this->dependency_lists_
.clear();
1851 for (size_t k
= 0; k
< version_trees_
.size(); ++k
)
1852 delete version_trees_
[k
];
1853 this->version_trees_
.clear();
1854 for (size_t k
= 0; k
< expression_lists_
.size(); ++k
)
1855 delete expression_lists_
[k
];
1856 this->expression_lists_
.clear();
1859 std::vector
<std::string
>
1860 Version_script_info::get_versions() const
1862 std::vector
<std::string
> ret
;
1863 for (size_t j
= 0; j
< version_trees_
.size(); ++j
)
1864 if (!this->version_trees_
[j
]->tag
.empty())
1865 ret
.push_back(this->version_trees_
[j
]->tag
);
1869 std::vector
<std::string
>
1870 Version_script_info::get_dependencies(const char* version
) const
1872 std::vector
<std::string
> ret
;
1873 for (size_t j
= 0; j
< version_trees_
.size(); ++j
)
1874 if (version_trees_
[j
]->tag
== version
)
1876 const struct Version_dependency_list
* deps
=
1877 version_trees_
[j
]->dependencies
;
1879 for (size_t k
= 0; k
< deps
->dependencies
.size(); ++k
)
1880 ret
.push_back(deps
->dependencies
[k
]);
1886 // Look up SYMBOL_NAME in the list of versions. If CHECK_GLOBAL is
1887 // true look at the globally visible symbols, otherwise look at the
1888 // symbols listed as "local:". Return true if the symbol is found,
1889 // false otherwise. If the symbol is found, then if PVERSION is not
1890 // NULL, set *PVERSION to the version.
1893 Version_script_info::get_symbol_version_helper(const char* symbol_name
,
1895 std::string
* pversion
) const
1897 Lazy_demangler
cpp_demangled_name(symbol_name
, DMGL_ANSI
| DMGL_PARAMS
);
1898 Lazy_demangler
java_demangled_name(symbol_name
,
1899 DMGL_ANSI
| DMGL_PARAMS
| DMGL_JAVA
);
1900 for (size_t j
= 0; j
< version_trees_
.size(); ++j
)
1902 // Is it a global symbol for this version?
1903 const Version_expression_list
* explist
=
1904 check_global
? version_trees_
[j
]->global
: version_trees_
[j
]->local
;
1905 if (explist
!= NULL
)
1906 for (size_t k
= 0; k
< explist
->expressions
.size(); ++k
)
1908 const char* name_to_match
= symbol_name
;
1909 const struct Version_expression
& exp
= explist
->expressions
[k
];
1910 if (exp
.language
== "C++")
1912 name_to_match
= cpp_demangled_name
.get();
1913 // This isn't a C++ symbol.
1914 if (name_to_match
== NULL
)
1917 else if (exp
.language
== "Java")
1919 name_to_match
= java_demangled_name
.get();
1920 // This isn't a Java symbol.
1921 if (name_to_match
== NULL
)
1925 if (exp
.exact_match
)
1926 matched
= strcmp(exp
.pattern
.c_str(), name_to_match
) == 0;
1928 matched
= fnmatch(exp
.pattern
.c_str(), name_to_match
,
1932 if (pversion
!= NULL
)
1933 *pversion
= this->version_trees_
[j
]->tag
;
1941 struct Version_dependency_list
*
1942 Version_script_info::allocate_dependency_list()
1944 dependency_lists_
.push_back(new Version_dependency_list
);
1945 return dependency_lists_
.back();
1948 struct Version_expression_list
*
1949 Version_script_info::allocate_expression_list()
1951 expression_lists_
.push_back(new Version_expression_list
);
1952 return expression_lists_
.back();
1955 struct Version_tree
*
1956 Version_script_info::allocate_version_tree()
1958 version_trees_
.push_back(new Version_tree
);
1959 return version_trees_
.back();
1962 // Print for debugging.
1965 Version_script_info::print(FILE* f
) const
1970 fprintf(f
, "VERSION {");
1972 for (size_t i
= 0; i
< this->version_trees_
.size(); ++i
)
1974 const Version_tree
* vt
= this->version_trees_
[i
];
1976 if (vt
->tag
.empty())
1979 fprintf(f
, " %s {\n", vt
->tag
.c_str());
1981 if (vt
->global
!= NULL
)
1983 fprintf(f
, " global :\n");
1984 this->print_expression_list(f
, vt
->global
);
1987 if (vt
->local
!= NULL
)
1989 fprintf(f
, " local :\n");
1990 this->print_expression_list(f
, vt
->local
);
1994 if (vt
->dependencies
!= NULL
)
1996 const Version_dependency_list
* deps
= vt
->dependencies
;
1997 for (size_t j
= 0; j
< deps
->dependencies
.size(); ++j
)
1999 if (j
< deps
->dependencies
.size() - 1)
2001 fprintf(f
, " %s", deps
->dependencies
[j
].c_str());
2011 Version_script_info::print_expression_list(
2013 const Version_expression_list
* vel
) const
2015 std::string current_language
;
2016 for (size_t i
= 0; i
< vel
->expressions
.size(); ++i
)
2018 const Version_expression
& ve(vel
->expressions
[i
]);
2020 if (ve
.language
!= current_language
)
2022 if (!current_language
.empty())
2024 fprintf(f
, " extern \"%s\" {\n", ve
.language
.c_str());
2025 current_language
= ve
.language
;
2029 if (!current_language
.empty())
2034 fprintf(f
, "%s", ve
.pattern
.c_str());
2041 if (!current_language
.empty())
2045 } // End namespace gold.
2047 // The remaining functions are extern "C", so it's clearer to not put
2048 // them in namespace gold.
2050 using namespace gold
;
2052 // This function is called by the bison parser to return the next
2056 yylex(YYSTYPE
* lvalp
, void* closurev
)
2058 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2059 const Token
* token
= closure
->next_token();
2060 switch (token
->classification())
2065 case Token::TOKEN_INVALID
:
2066 yyerror(closurev
, "invalid character");
2069 case Token::TOKEN_EOF
:
2072 case Token::TOKEN_STRING
:
2074 // This is either a keyword or a STRING.
2076 const char* str
= token
->string_value(&len
);
2078 switch (closure
->lex_mode())
2080 case Lex::LINKER_SCRIPT
:
2081 parsecode
= script_keywords
.keyword_to_parsecode(str
, len
);
2083 case Lex::VERSION_SCRIPT
:
2084 parsecode
= version_script_keywords
.keyword_to_parsecode(str
, len
);
2086 case Lex::DYNAMIC_LIST
:
2087 parsecode
= dynamic_list_keywords
.keyword_to_parsecode(str
, len
);
2094 lvalp
->string
.value
= str
;
2095 lvalp
->string
.length
= len
;
2099 case Token::TOKEN_QUOTED_STRING
:
2100 lvalp
->string
.value
= token
->string_value(&lvalp
->string
.length
);
2101 return QUOTED_STRING
;
2103 case Token::TOKEN_OPERATOR
:
2104 return token
->operator_value();
2106 case Token::TOKEN_INTEGER
:
2107 lvalp
->integer
= token
->integer_value();
2112 // This function is called by the bison parser to report an error.
2115 yyerror(void* closurev
, const char* message
)
2117 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2118 gold_error(_("%s:%d:%d: %s"), closure
->filename(), closure
->lineno(),
2119 closure
->charpos(), message
);
2122 // Called by the bison parser to add an external symbol to the link.
2125 script_add_extern(void* closurev
, const char* name
, size_t length
)
2127 // We treat exactly like -u NAME. FIXME: If it seems useful, we
2128 // could handle this after the command line has been read, by adding
2129 // entries to the symbol table directly.
2130 std::string
arg("--undefined=");
2131 arg
.append(name
, length
);
2132 script_parse_option(closurev
, arg
.c_str(), arg
.size());
2135 // Called by the bison parser to add a file to the link.
2138 script_add_file(void* closurev
, const char* name
, size_t length
)
2140 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2142 // If this is an absolute path, and we found the script in the
2143 // sysroot, then we want to prepend the sysroot to the file name.
2144 // For example, this is how we handle a cross link to the x86_64
2145 // libc.so, which refers to /lib/libc.so.6.
2146 std::string
name_string(name
, length
);
2147 const char* extra_search_path
= ".";
2148 std::string script_directory
;
2149 if (IS_ABSOLUTE_PATH(name_string
.c_str()))
2151 if (closure
->is_in_sysroot())
2153 const std::string
& sysroot(parameters
->options().sysroot());
2154 gold_assert(!sysroot
.empty());
2155 name_string
= sysroot
+ name_string
;
2160 // In addition to checking the normal library search path, we
2161 // also want to check in the script-directory.
2162 const char *slash
= strrchr(closure
->filename(), '/');
2165 script_directory
.assign(closure
->filename(),
2166 slash
- closure
->filename() + 1);
2167 extra_search_path
= script_directory
.c_str();
2171 Input_file_argument
file(name_string
.c_str(), false, extra_search_path
,
2172 false, closure
->position_dependent_options());
2173 closure
->inputs()->add_file(file
);
2176 // Called by the bison parser to start a group. If we are already in
2177 // a group, that means that this script was invoked within a
2178 // --start-group --end-group sequence on the command line, or that
2179 // this script was found in a GROUP of another script. In that case,
2180 // we simply continue the existing group, rather than starting a new
2181 // one. It is possible to construct a case in which this will do
2182 // something other than what would happen if we did a recursive group,
2183 // but it's hard to imagine why the different behaviour would be
2184 // useful for a real program. Avoiding recursive groups is simpler
2185 // and more efficient.
2188 script_start_group(void* closurev
)
2190 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2191 if (!closure
->in_group())
2192 closure
->inputs()->start_group();
2195 // Called by the bison parser at the end of a group.
2198 script_end_group(void* closurev
)
2200 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2201 if (!closure
->in_group())
2202 closure
->inputs()->end_group();
2205 // Called by the bison parser to start an AS_NEEDED list.
2208 script_start_as_needed(void* closurev
)
2210 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2211 closure
->position_dependent_options().set_as_needed(true);
2214 // Called by the bison parser at the end of an AS_NEEDED list.
2217 script_end_as_needed(void* closurev
)
2219 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2220 closure
->position_dependent_options().set_as_needed(false);
2223 // Called by the bison parser to set the entry symbol.
2226 script_set_entry(void* closurev
, const char* entry
, size_t length
)
2228 // We'll parse this exactly the same as --entry=ENTRY on the commandline
2229 // TODO(csilvers): FIXME -- call set_entry directly.
2230 std::string
arg("--entry=");
2231 arg
.append(entry
, length
);
2232 script_parse_option(closurev
, arg
.c_str(), arg
.size());
2235 // Called by the bison parser to set whether to define common symbols.
2238 script_set_common_allocation(void* closurev
, int set
)
2240 const char* arg
= set
!= 0 ? "--define-common" : "--no-define-common";
2241 script_parse_option(closurev
, arg
, strlen(arg
));
2244 // Called by the bison parser to define a symbol.
2247 script_set_symbol(void* closurev
, const char* name
, size_t length
,
2248 Expression
* value
, int providei
, int hiddeni
)
2250 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2251 const bool provide
= providei
!= 0;
2252 const bool hidden
= hiddeni
!= 0;
2253 closure
->script_options()->add_symbol_assignment(name
, length
, value
,
2255 closure
->clear_skip_on_incompatible_target();
2258 // Called by the bison parser to add an assertion.
2261 script_add_assertion(void* closurev
, Expression
* check
, const char* message
,
2264 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2265 closure
->script_options()->add_assertion(check
, message
, messagelen
);
2266 closure
->clear_skip_on_incompatible_target();
2269 // Called by the bison parser to parse an OPTION.
2272 script_parse_option(void* closurev
, const char* option
, size_t length
)
2274 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2275 // We treat the option as a single command-line option, even if
2276 // it has internal whitespace.
2277 if (closure
->command_line() == NULL
)
2279 // There are some options that we could handle here--e.g.,
2280 // -lLIBRARY. Should we bother?
2281 gold_warning(_("%s:%d:%d: ignoring command OPTION; OPTION is only valid"
2282 " for scripts specified via -T/--script"),
2283 closure
->filename(), closure
->lineno(), closure
->charpos());
2287 bool past_a_double_dash_option
= false;
2288 const char* mutable_option
= strndup(option
, length
);
2289 gold_assert(mutable_option
!= NULL
);
2290 closure
->command_line()->process_one_option(1, &mutable_option
, 0,
2291 &past_a_double_dash_option
);
2292 // The General_options class will quite possibly store a pointer
2293 // into mutable_option, so we can't free it. In cases the class
2294 // does not store such a pointer, this is a memory leak. Alas. :(
2296 closure
->clear_skip_on_incompatible_target();
2299 // Called by the bison parser to handle OUTPUT_FORMAT. OUTPUT_FORMAT
2300 // takes either one or three arguments. In the three argument case,
2301 // the format depends on the endianness option, which we don't
2302 // currently support (FIXME). If we see an OUTPUT_FORMAT for the
2303 // wrong format, then we want to search for a new file. Returning 0
2304 // here will cause the parser to immediately abort.
2307 script_check_output_format(void* closurev
,
2308 const char* default_name
, size_t default_length
,
2309 const char*, size_t, const char*, size_t)
2311 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2312 std::string
name(default_name
, default_length
);
2313 Target
* target
= select_target_by_name(name
.c_str());
2314 if (target
== NULL
|| !parameters
->is_compatible_target(target
))
2316 if (closure
->skip_on_incompatible_target())
2318 closure
->set_found_incompatible_target();
2321 // FIXME: Should we warn about the unknown target?
2326 // Called by the bison parser to handle SEARCH_DIR. This is handled
2327 // exactly like a -L option.
2330 script_add_search_dir(void* closurev
, const char* option
, size_t length
)
2332 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2333 if (closure
->command_line() == NULL
)
2334 gold_warning(_("%s:%d:%d: ignoring SEARCH_DIR; SEARCH_DIR is only valid"
2335 " for scripts specified via -T/--script"),
2336 closure
->filename(), closure
->lineno(), closure
->charpos());
2339 std::string s
= "-L" + std::string(option
, length
);
2340 script_parse_option(closurev
, s
.c_str(), s
.size());
2344 /* Called by the bison parser to push the lexer into expression
2348 script_push_lex_into_expression_mode(void* closurev
)
2350 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2351 closure
->push_lex_mode(Lex::EXPRESSION
);
2354 /* Called by the bison parser to push the lexer into version
2358 script_push_lex_into_version_mode(void* closurev
)
2360 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2361 closure
->push_lex_mode(Lex::VERSION_SCRIPT
);
2364 /* Called by the bison parser to pop the lexer mode. */
2367 script_pop_lex_mode(void* closurev
)
2369 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2370 closure
->pop_lex_mode();
2373 // Register an entire version node. For example:
2379 // - tag is "GLIBC_2.1"
2380 // - tree contains the information "global: foo"
2381 // - deps contains "GLIBC_2.0"
2384 script_register_vers_node(void*,
2387 struct Version_tree
*tree
,
2388 struct Version_dependency_list
*deps
)
2390 gold_assert(tree
!= NULL
);
2391 tree
->dependencies
= deps
;
2393 tree
->tag
= std::string(tag
, taglen
);
2396 // Add a dependencies to the list of existing dependencies, if any,
2397 // and return the expanded list.
2399 extern "C" struct Version_dependency_list
*
2400 script_add_vers_depend(void* closurev
,
2401 struct Version_dependency_list
*all_deps
,
2402 const char *depend_to_add
, int deplen
)
2404 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2405 if (all_deps
== NULL
)
2406 all_deps
= closure
->version_script()->allocate_dependency_list();
2407 all_deps
->dependencies
.push_back(std::string(depend_to_add
, deplen
));
2411 // Add a pattern expression to an existing list of expressions, if any.
2412 // TODO: In the old linker, the last argument used to be a bool, but I
2413 // don't know what it meant.
2415 extern "C" struct Version_expression_list
*
2416 script_new_vers_pattern(void* closurev
,
2417 struct Version_expression_list
*expressions
,
2418 const char *pattern
, int patlen
, int exact_match
)
2420 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2421 if (expressions
== NULL
)
2422 expressions
= closure
->version_script()->allocate_expression_list();
2423 expressions
->expressions
.push_back(
2424 Version_expression(std::string(pattern
, patlen
),
2425 closure
->get_current_language(),
2426 static_cast<bool>(exact_match
)));
2430 // Attaches b to the end of a, and clears b. So a = a + b and b = {}.
2432 extern "C" struct Version_expression_list
*
2433 script_merge_expressions(struct Version_expression_list
*a
,
2434 struct Version_expression_list
*b
)
2436 a
->expressions
.insert(a
->expressions
.end(),
2437 b
->expressions
.begin(), b
->expressions
.end());
2438 // We could delete b and remove it from expressions_lists_, but
2439 // that's a lot of work. This works just as well.
2440 b
->expressions
.clear();
2444 // Combine the global and local expressions into a a Version_tree.
2446 extern "C" struct Version_tree
*
2447 script_new_vers_node(void* closurev
,
2448 struct Version_expression_list
*global
,
2449 struct Version_expression_list
*local
)
2451 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2452 Version_tree
* tree
= closure
->version_script()->allocate_version_tree();
2453 tree
->global
= global
;
2454 tree
->local
= local
;
2458 // Handle a transition in language, such as at the
2459 // start or end of 'extern "C++"'
2462 version_script_push_lang(void* closurev
, const char* lang
, int langlen
)
2464 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2465 closure
->push_language(std::string(lang
, langlen
));
2469 version_script_pop_lang(void* closurev
)
2471 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2472 closure
->pop_language();
2475 // Called by the bison parser to start a SECTIONS clause.
2478 script_start_sections(void* closurev
)
2480 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2481 closure
->script_options()->script_sections()->start_sections();
2482 closure
->clear_skip_on_incompatible_target();
2485 // Called by the bison parser to finish a SECTIONS clause.
2488 script_finish_sections(void* closurev
)
2490 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2491 closure
->script_options()->script_sections()->finish_sections();
2494 // Start processing entries for an output section.
2497 script_start_output_section(void* closurev
, const char* name
, size_t namelen
,
2498 const struct Parser_output_section_header
* header
)
2500 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2501 closure
->script_options()->script_sections()->start_output_section(name
,
2506 // Finish processing entries for an output section.
2509 script_finish_output_section(void* closurev
,
2510 const struct Parser_output_section_trailer
* trail
)
2512 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2513 closure
->script_options()->script_sections()->finish_output_section(trail
);
2516 // Add a data item (e.g., "WORD (0)") to the current output section.
2519 script_add_data(void* closurev
, int data_token
, Expression
* val
)
2521 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2523 bool is_signed
= true;
2545 closure
->script_options()->script_sections()->add_data(size
, is_signed
, val
);
2548 // Add a clause setting the fill value to the current output section.
2551 script_add_fill(void* closurev
, Expression
* val
)
2553 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2554 closure
->script_options()->script_sections()->add_fill(val
);
2557 // Add a new input section specification to the current output
2561 script_add_input_section(void* closurev
,
2562 const struct Input_section_spec
* spec
,
2565 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2566 bool keep
= keepi
!= 0;
2567 closure
->script_options()->script_sections()->add_input_section(spec
, keep
);
2570 // When we see DATA_SEGMENT_ALIGN we record that following output
2571 // sections may be relro.
2574 script_data_segment_align(void* closurev
)
2576 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2577 if (!closure
->script_options()->saw_sections_clause())
2578 gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
2579 closure
->filename(), closure
->lineno(), closure
->charpos());
2581 closure
->script_options()->script_sections()->data_segment_align();
2584 // When we see DATA_SEGMENT_RELRO_END we know that all output sections
2585 // since DATA_SEGMENT_ALIGN should be relro.
2588 script_data_segment_relro_end(void* closurev
)
2590 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2591 if (!closure
->script_options()->saw_sections_clause())
2592 gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
2593 closure
->filename(), closure
->lineno(), closure
->charpos());
2595 closure
->script_options()->script_sections()->data_segment_relro_end();
2598 // Create a new list of string/sort pairs.
2600 extern "C" String_sort_list_ptr
2601 script_new_string_sort_list(const struct Wildcard_section
* string_sort
)
2603 return new String_sort_list(1, *string_sort
);
2606 // Add an entry to a list of string/sort pairs. The way the parser
2607 // works permits us to simply modify the first parameter, rather than
2610 extern "C" String_sort_list_ptr
2611 script_string_sort_list_add(String_sort_list_ptr pv
,
2612 const struct Wildcard_section
* string_sort
)
2615 return script_new_string_sort_list(string_sort
);
2618 pv
->push_back(*string_sort
);
2623 // Create a new list of strings.
2625 extern "C" String_list_ptr
2626 script_new_string_list(const char* str
, size_t len
)
2628 return new String_list(1, std::string(str
, len
));
2631 // Add an element to a list of strings. The way the parser works
2632 // permits us to simply modify the first parameter, rather than copy
2635 extern "C" String_list_ptr
2636 script_string_list_push_back(String_list_ptr pv
, const char* str
, size_t len
)
2639 return script_new_string_list(str
, len
);
2642 pv
->push_back(std::string(str
, len
));
2647 // Concatenate two string lists. Either or both may be NULL. The way
2648 // the parser works permits us to modify the parameters, rather than
2651 extern "C" String_list_ptr
2652 script_string_list_append(String_list_ptr pv1
, String_list_ptr pv2
)
2658 pv1
->insert(pv1
->end(), pv2
->begin(), pv2
->end());
2662 // Add a new program header.
2665 script_add_phdr(void* closurev
, const char* name
, size_t namelen
,
2666 unsigned int type
, const Phdr_info
* info
)
2668 Parser_closure
* closure
= static_cast<Parser_closure
*>(closurev
);
2669 bool includes_filehdr
= info
->includes_filehdr
!= 0;
2670 bool includes_phdrs
= info
->includes_phdrs
!= 0;
2671 bool is_flags_valid
= info
->is_flags_valid
!= 0;
2672 Script_sections
* ss
= closure
->script_options()->script_sections();
2673 ss
->add_phdr(name
, namelen
, type
, includes_filehdr
, includes_phdrs
,
2674 is_flags_valid
, info
->flags
, info
->load_address
);
2675 closure
->clear_skip_on_incompatible_target();
2678 // Convert a program header string to a type.
2680 #define PHDR_TYPE(NAME) { #NAME, sizeof(#NAME) - 1, elfcpp::NAME }
2687 } phdr_type_names
[] =
2691 PHDR_TYPE(PT_DYNAMIC
),
2692 PHDR_TYPE(PT_INTERP
),
2694 PHDR_TYPE(PT_SHLIB
),
2697 PHDR_TYPE(PT_GNU_EH_FRAME
),
2698 PHDR_TYPE(PT_GNU_STACK
),
2699 PHDR_TYPE(PT_GNU_RELRO
)
2702 extern "C" unsigned int
2703 script_phdr_string_to_type(void* closurev
, const char* name
, size_t namelen
)
2705 for (unsigned int i
= 0;
2706 i
< sizeof(phdr_type_names
) / sizeof(phdr_type_names
[0]);
2708 if (namelen
== phdr_type_names
[i
].namelen
2709 && strncmp(name
, phdr_type_names
[i
].name
, namelen
) == 0)
2710 return phdr_type_names
[i
].val
;
2711 yyerror(closurev
, _("unknown PHDR type (try integer)"));
2712 return elfcpp::PT_NULL
;