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"
49 // A token read from a script file. We don't implement keywords here;
50 // all keywords are simply represented as a string.
55 // Token classification.
60 // Token indicates end of input.
62 // Token is a string of characters.
64 // Token is a quoted string of characters.
66 // Token is an operator.
68 // Token is a number (an integer).
72 // We need an empty constructor so that we can put this STL objects.
74 : classification_(TOKEN_INVALID), value_(NULL), value_length_(0),
75 opcode_(0), lineno_(0), charpos_(0)
78 // A general token with no value.
79 Token(Classification classification, int lineno, int charpos)
80 : classification_(classification), value_(NULL), value_length_(0),
81 opcode_(0), lineno_(lineno), charpos_(charpos)
83 gold_assert(classification == TOKEN_INVALID
84 || classification == TOKEN_EOF);
87 // A general token with a value.
88 Token(Classification classification, const char* value, size_t length,
89 int lineno, int charpos)
90 : classification_(classification), value_(value), value_length_(length),
91 opcode_(0), lineno_(lineno), charpos_(charpos)
93 gold_assert(classification != TOKEN_INVALID
94 && classification != TOKEN_EOF);
97 // A token representing an operator.
98 Token(int opcode, int lineno, int charpos)
99 : classification_(TOKEN_OPERATOR), value_(NULL), value_length_(0),
100 opcode_(opcode), lineno_(lineno), charpos_(charpos)
103 // Return whether the token is invalid.
106 { return this->classification_ == TOKEN_INVALID; }
108 // Return whether this is an EOF token.
111 { return this->classification_ == TOKEN_EOF; }
113 // Return the token classification.
115 classification() const
116 { return this->classification_; }
118 // Return the line number at which the token starts.
121 { return this->lineno_; }
123 // Return the character position at this the token starts.
126 { return this->charpos_; }
128 // Get the value of a token.
131 string_value(size_t* length) const
133 gold_assert(this->classification_ == TOKEN_STRING
134 || this->classification_ == TOKEN_QUOTED_STRING);
135 *length = this->value_length_;
140 operator_value() const
142 gold_assert(this->classification_ == TOKEN_OPERATOR);
143 return this->opcode_;
147 integer_value() const
149 gold_assert(this->classification_ == TOKEN_INTEGER);
151 std::string s(this->value_, this->value_length_);
152 return strtoull(s.c_str(), NULL, 0);
156 // The token classification.
157 Classification classification_;
158 // The token value, for TOKEN_STRING or TOKEN_QUOTED_STRING or
161 // The length of the token value.
162 size_t value_length_;
163 // The token value, for TOKEN_OPERATOR.
165 // The line number where this token started (one based).
167 // The character position within the line where this token started
172 // This class handles lexing a file into a sequence of tokens.
177 // We unfortunately have to support different lexing modes, because
178 // when reading different parts of a linker script we need to parse
179 // things differently.
182 // Reading an ordinary linker script.
184 // Reading an expression in a linker script.
186 // Reading a version script.
188 // Reading a --dynamic-list file.
192 Lex(const char* input_string, size_t input_length, int parsing_token)
193 : input_string_(input_string), input_length_(input_length),
194 current_(input_string), mode_(LINKER_SCRIPT),
195 first_token_(parsing_token), token_(),
196 lineno_(1), linestart_(input_string)
199 // Read a file into a string.
201 read_file(Input_file*, std::string*);
203 // Return the next token.
207 // Return the current lexing mode.
210 { return this->mode_; }
212 // Set the lexing mode.
215 { this->mode_ = mode; }
219 Lex& operator=(const Lex&);
221 // Make a general token with no value at the current location.
223 make_token(Token::Classification c, const char* start) const
224 { return Token(c, this->lineno_, start - this->linestart_ + 1); }
226 // Make a general token with a value at the current location.
228 make_token(Token::Classification c, const char* v, size_t len,
231 { return Token(c, v, len, this->lineno_, start - this->linestart_ + 1); }
233 // Make an operator token at the current location.
235 make_token(int opcode, const char* start) const
236 { return Token(opcode, this->lineno_, start - this->linestart_ + 1); }
238 // Make an invalid token at the current location.
240 make_invalid_token(const char* start)
241 { return this->make_token(Token::TOKEN_INVALID, start); }
243 // Make an EOF token at the current location.
245 make_eof_token(const char* start)
246 { return this->make_token(Token::TOKEN_EOF, start); }
248 // Return whether C can be the first character in a name. C2 is the
249 // next character, since we sometimes need that.
251 can_start_name(char c, char c2);
253 // If C can appear in a name which has already started, return a
254 // pointer to a character later in the token or just past
255 // it. Otherwise, return NULL.
257 can_continue_name(const char* c);
259 // Return whether C, C2, C3 can start a hex number.
261 can_start_hex(char c, char c2, char c3);
263 // If C can appear in a hex number which has already started, return
264 // a pointer to a character later in the token or just past
265 // it. Otherwise, return NULL.
267 can_continue_hex(const char* c);
269 // Return whether C can start a non-hex number.
271 can_start_number(char c);
273 // If C can appear in a decimal number which has already started,
274 // return a pointer to a character later in the token or just past
275 // it. Otherwise, return NULL.
277 can_continue_number(const char* c)
278 { return Lex::can_start_number(*c) ? c + 1 : NULL; }
280 // If C1 C2 C3 form a valid three character operator, return the
281 // opcode. Otherwise return 0.
283 three_char_operator(char c1, char c2, char c3);
285 // If C1 C2 form a valid two character operator, return the opcode.
286 // Otherwise return 0.
288 two_char_operator(char c1, char c2);
290 // If C1 is a valid one character operator, return the opcode.
291 // Otherwise return 0.
293 one_char_operator(char c1);
295 // Read the next token.
297 get_token(const char**);
299 // Skip a C style /* */ comment. Return false if the comment did
302 skip_c_comment(const char**);
304 // Skip a line # comment. Return false if there was no newline.
306 skip_line_comment(const char**);
308 // Build a token CLASSIFICATION from all characters that match
309 // CAN_CONTINUE_FN. The token starts at START. Start matching from
310 // MATCH. Set *PP to the character following the token.
312 gather_token(Token::Classification,
313 const char* (Lex::*can_continue_fn)(const char*),
314 const char* start, const char* match, const char** pp);
316 // Build a token from a quoted string.
318 gather_quoted_string(const char** pp);
320 // The string we are tokenizing.
321 const char* input_string_;
322 // The length of the string.
323 size_t input_length_;
324 // The current offset into the string.
325 const char* current_;
326 // The current lexing mode.
328 // The code to use for the first token. This is set to 0 after it
331 // The current token.
333 // The current line number.
335 // The start of the current line in the string.
336 const char* linestart_;
339 // Read the whole file into memory. We don't expect linker scripts to
340 // be large, so we just use a std::string as a buffer. We ignore the
341 // data we've already read, so that we read aligned buffers.
344 Lex::read_file(Input_file* input_file, std::string* contents)
346 off_t filesize = input_file->file().filesize();
348 contents->reserve(filesize);
351 unsigned char buf[BUFSIZ];
352 while (off < filesize)
355 if (get > filesize - off)
356 get = filesize - off;
357 input_file->file().read(off, get, buf);
358 contents->append(reinterpret_cast<char*>(&buf[0]), get);
363 // Return whether C can be the start of a name, if the next character
364 // is C2. A name can being with a letter, underscore, period, or
365 // dollar sign. Because a name can be a file name, we also permit
366 // forward slash, backslash, and tilde. Tilde is the tricky case
367 // here; GNU ld also uses it as a bitwise not operator. It is only
368 // recognized as the operator if it is not immediately followed by
369 // some character which can appear in a symbol. That is, when we
370 // don't know that we are looking at an expression, "~0" is a file
371 // name, and "~ 0" is an expression using bitwise not. We are
375 Lex::can_start_name(char c, char c2)
379 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
380 case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
381 case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R':
382 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
384 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
385 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
386 case 'm': case 'n': case 'o': case 'q': case 'p': case 'r':
387 case 's': case 't': case 'u': case 'v': case 'w': case 'x':
389 case '_': case '.': case '$':
393 return this->mode_ == LINKER_SCRIPT;
396 return this->mode_ == LINKER_SCRIPT && can_continue_name(&c2);
399 return (this->mode_ == VERSION_SCRIPT
400 || this->mode_ == DYNAMIC_LIST
401 || (this->mode_ == LINKER_SCRIPT
402 && can_continue_name(&c2)));
409 // Return whether C can continue a name which has already started.
410 // Subsequent characters in a name are the same as the leading
411 // characters, plus digits and "=+-:[],?*". So in general the linker
412 // script language requires spaces around operators, unless we know
413 // that we are parsing an expression.
416 Lex::can_continue_name(const char* c)
420 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
421 case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
422 case 'M': case 'N': case 'O': case 'Q': case 'P': case 'R':
423 case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
425 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
426 case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
427 case 'm': case 'n': case 'o': case 'q': case 'p': case 'r':
428 case 's': case 't': case 'u': case 'v': case 'w': case 'x':
430 case '_': case '.': case '$':
431 case '0': case '1': case '2': case '3': case '4':
432 case '5': case '6': case '7': case '8': case '9':
435 // TODO(csilvers): why not allow ~ in names for version-scripts?
436 case '/': case '\\': case '~':
439 if (this->mode_ == LINKER_SCRIPT)
443 case '[': case ']': case '*': case '?': case '-':
444 if (this->mode_ == LINKER_SCRIPT || this->mode_ == VERSION_SCRIPT
445 || this->mode_ == DYNAMIC_LIST)
449 // TODO(csilvers): why allow this? ^ is meaningless in version scripts.
451 if (this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST)
456 if (this->mode_ == LINKER_SCRIPT)
458 else if ((this->mode_ == VERSION_SCRIPT || this->mode_ == DYNAMIC_LIST)
461 // A name can have '::' in it, as that's a c++ namespace
462 // separator. But a single colon is not part of a name.
472 // For a number we accept 0x followed by hex digits, or any sequence
473 // of digits. The old linker accepts leading '$' for hex, and
474 // trailing HXBOD. Those are for MRI compatibility and we don't
475 // accept them. The old linker also accepts trailing MK for mega or
476 // kilo. FIXME: Those are mentioned in the documentation, and we
477 // should accept them.
479 // Return whether C1 C2 C3 can start a hex number.
482 Lex::can_start_hex(char c1, char c2, char c3)
484 if (c1 == '0' && (c2 == 'x' || c2 == 'X'))
485 return this->can_continue_hex(&c3);
489 // Return whether C can appear in a hex number.
492 Lex::can_continue_hex(const char* c)
496 case '0': case '1': case '2': case '3': case '4':
497 case '5': case '6': case '7': case '8': case '9':
498 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
499 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
507 // Return whether C can start a non-hex number.
510 Lex::can_start_number(char c)
514 case '0': case '1': case '2': case '3': case '4':
515 case '5': case '6': case '7': case '8': case '9':
523 // If C1 C2 C3 form a valid three character operator, return the
524 // opcode (defined in the yyscript.h file generated from yyscript.y).
525 // Otherwise return 0.
528 Lex::three_char_operator(char c1, char c2, char c3)
533 if (c2 == '<' && c3 == '=')
537 if (c2 == '>' && c3 == '=')
546 // If C1 C2 form a valid two character operator, return the opcode
547 // (defined in the yyscript.h file generated from yyscript.y).
548 // Otherwise return 0.
551 Lex::two_char_operator(char c1, char c2)
609 // If C1 is a valid operator, return the opcode. Otherwise return 0.
612 Lex::one_char_operator(char c1)
645 // Skip a C style comment. *PP points to just after the "/*". Return
646 // false if the comment did not end.
649 Lex::skip_c_comment(const char** pp)
652 while (p[0] != '*' || p[1] != '/')
663 this->linestart_ = p + 1;
672 // Skip a line # comment. Return false if there was no newline.
675 Lex::skip_line_comment(const char** pp)
678 size_t skip = strcspn(p, "\n");
687 this->linestart_ = p;
693 // Build a token CLASSIFICATION from all characters that match
694 // CAN_CONTINUE_FN. Update *PP.
697 Lex::gather_token(Token::Classification classification,
698 const char* (Lex::*can_continue_fn)(const char*),
703 const char* new_match = NULL;
704 while ((new_match = (this->*can_continue_fn)(match)))
707 return this->make_token(classification, start, match - start, start);
710 // Build a token from a quoted string.
713 Lex::gather_quoted_string(const char** pp)
715 const char* start = *pp;
716 const char* p = start;
718 size_t skip = strcspn(p, "\"\n");
720 return this->make_invalid_token(start);
722 return this->make_token(Token::TOKEN_QUOTED_STRING, p, skip, start);
725 // Return the next token at *PP. Update *PP. General guideline: we
726 // require linker scripts to be simple ASCII. No unicode linker
727 // scripts. In particular we can assume that any '\0' is the end of
731 Lex::get_token(const char** pp)
740 return this->make_eof_token(p);
743 // Skip whitespace quickly.
744 while (*p == ' ' || *p == '\t')
751 this->linestart_ = p;
755 // Skip C style comments.
756 if (p[0] == '/' && p[1] == '*')
758 int lineno = this->lineno_;
759 int charpos = p - this->linestart_ + 1;
762 if (!this->skip_c_comment(pp))
763 return Token(Token::TOKEN_INVALID, lineno, charpos);
769 // Skip line comments.
773 if (!this->skip_line_comment(pp))
774 return this->make_eof_token(p);
780 if (this->can_start_name(p[0], p[1]))
781 return this->gather_token(Token::TOKEN_STRING,
782 &Lex::can_continue_name,
785 // We accept any arbitrary name in double quotes, as long as it
786 // does not cross a line boundary.
790 return this->gather_quoted_string(pp);
793 // Check for a number.
795 if (this->can_start_hex(p[0], p[1], p[2]))
796 return this->gather_token(Token::TOKEN_INTEGER,
797 &Lex::can_continue_hex,
800 if (Lex::can_start_number(p[0]))
801 return this->gather_token(Token::TOKEN_INTEGER,
802 &Lex::can_continue_number,
805 // Check for operators.
807 int opcode = Lex::three_char_operator(p[0], p[1], p[2]);
811 return this->make_token(opcode, p);
814 opcode = Lex::two_char_operator(p[0], p[1]);
818 return this->make_token(opcode, p);
821 opcode = Lex::one_char_operator(p[0]);
825 return this->make_token(opcode, p);
828 return this->make_token(Token::TOKEN_INVALID, p);
832 // Return the next token.
837 // The first token is special.
838 if (this->first_token_ != 0)
840 this->token_ = Token(this->first_token_, 0, 0);
841 this->first_token_ = 0;
842 return &this->token_;
845 this->token_ = this->get_token(&this->current_);
847 // Don't let an early null byte fool us into thinking that we've
848 // reached the end of the file.
849 if (this->token_.is_eof()
850 && (static_cast<size_t>(this->current_ - this->input_string_)
851 < this->input_length_))
852 this->token_ = this->make_invalid_token(this->current_);
854 return &this->token_;
857 // class Symbol_assignment.
859 // Add the symbol to the symbol table. This makes sure the symbol is
860 // there and defined. The actual value is stored later. We can't
861 // determine the actual value at this point, because we can't
862 // necessarily evaluate the expression until all ordinary symbols have
865 // The GNU linker lets symbol assignments in the linker script
866 // silently override defined symbols in object files. We are
867 // compatible. FIXME: Should we issue a warning?
870 Symbol_assignment::add_to_table(Symbol_table* symtab)
872 elfcpp::STV vis = this->hidden_ ? elfcpp::STV_HIDDEN : elfcpp::STV_DEFAULT;
873 this->sym_ = symtab->define_as_constant(this->name_.c_str(),
882 true); // force_override
885 // Finalize a symbol value.
888 Symbol_assignment::finalize(Symbol_table* symtab, const Layout* layout)
890 this->finalize_maybe_dot(symtab, layout, false, 0, NULL);
893 // Finalize a symbol value which can refer to the dot symbol.
896 Symbol_assignment::finalize_with_dot(Symbol_table* symtab,
897 const Layout* layout,
899 Output_section* dot_section)
901 this->finalize_maybe_dot(symtab, layout, true, dot_value, dot_section);
904 // Finalize a symbol value, internal version.
907 Symbol_assignment::finalize_maybe_dot(Symbol_table* symtab,
908 const Layout* layout,
909 bool is_dot_available,
911 Output_section* dot_section)
913 // If we were only supposed to provide this symbol, the sym_ field
914 // will be NULL if the symbol was not referenced.
915 if (this->sym_ == NULL)
917 gold_assert(this->provide_);
921 if (parameters->target().get_size() == 32)
923 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
924 this->sized_finalize<32>(symtab, layout, is_dot_available, dot_value,
930 else if (parameters->target().get_size() == 64)
932 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
933 this->sized_finalize<64>(symtab, layout, is_dot_available, dot_value,
945 Symbol_assignment::sized_finalize(Symbol_table* symtab, const Layout* layout,
946 bool is_dot_available, uint64_t dot_value,
947 Output_section* dot_section)
949 Output_section* section;
950 uint64_t final_val = this->val_->eval_maybe_dot(symtab, layout, true,
952 dot_value, dot_section,
954 Sized_symbol<size>* ssym = symtab->get_sized_symbol<size>(this->sym_);
955 ssym->set_value(final_val);
957 ssym->set_output_section(section);
960 // Set the symbol value if the expression yields an absolute value.
963 Symbol_assignment::set_if_absolute(Symbol_table* symtab, const Layout* layout,
964 bool is_dot_available, uint64_t dot_value)
966 if (this->sym_ == NULL)
969 Output_section* val_section;
970 uint64_t val = this->val_->eval_maybe_dot(symtab, layout, false,
971 is_dot_available, dot_value,
973 if (val_section != NULL)
976 if (parameters->target().get_size() == 32)
978 #if defined(HAVE_TARGET_32_LITTLE) || defined(HAVE_TARGET_32_BIG)
979 Sized_symbol<32>* ssym = symtab->get_sized_symbol<32>(this->sym_);
980 ssym->set_value(val);
985 else if (parameters->target().get_size() == 64)
987 #if defined(HAVE_TARGET_64_LITTLE) || defined(HAVE_TARGET_64_BIG)
988 Sized_symbol<64>* ssym = symtab->get_sized_symbol<64>(this->sym_);
989 ssym->set_value(val);
998 // Print for debugging.
1001 Symbol_assignment::print(FILE* f) const
1003 if (this->provide_ && this->hidden_)
1004 fprintf(f, "PROVIDE_HIDDEN(");
1005 else if (this->provide_)
1006 fprintf(f, "PROVIDE(");
1007 else if (this->hidden_)
1010 fprintf(f, "%s = ", this->name_.c_str());
1011 this->val_->print(f);
1013 if (this->provide_ || this->hidden_)
1019 // Class Script_assertion.
1021 // Check the assertion.
1024 Script_assertion::check(const Symbol_table* symtab, const Layout* layout)
1026 if (!this->check_->eval(symtab, layout, true))
1027 gold_error("%s", this->message_.c_str());
1030 // Print for debugging.
1033 Script_assertion::print(FILE* f) const
1035 fprintf(f, "ASSERT(");
1036 this->check_->print(f);
1037 fprintf(f, ", \"%s\")\n", this->message_.c_str());
1040 // Class Script_options.
1042 Script_options::Script_options()
1043 : entry_(), symbol_assignments_(), version_script_info_(),
1048 // Add a symbol to be defined.
1051 Script_options::add_symbol_assignment(const char* name, size_t length,
1052 Expression* value, bool provide,
1055 if (length != 1 || name[0] != '.')
1057 if (this->script_sections_.in_sections_clause())
1058 this->script_sections_.add_symbol_assignment(name, length, value,
1062 Symbol_assignment* p = new Symbol_assignment(name, length, value,
1064 this->symbol_assignments_.push_back(p);
1069 if (provide || hidden)
1070 gold_error(_("invalid use of PROVIDE for dot symbol"));
1071 if (!this->script_sections_.in_sections_clause())
1072 gold_error(_("invalid assignment to dot outside of SECTIONS"));
1074 this->script_sections_.add_dot_assignment(value);
1078 // Add an assertion.
1081 Script_options::add_assertion(Expression* check, const char* message,
1084 if (this->script_sections_.in_sections_clause())
1085 this->script_sections_.add_assertion(check, message, messagelen);
1088 Script_assertion* p = new Script_assertion(check, message, messagelen);
1089 this->assertions_.push_back(p);
1093 // Create sections required by any linker scripts.
1096 Script_options::create_script_sections(Layout* layout)
1098 if (this->saw_sections_clause())
1099 this->script_sections_.create_sections(layout);
1102 // Add any symbols we are defining to the symbol table.
1105 Script_options::add_symbols_to_table(Symbol_table* symtab)
1107 for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1108 p != this->symbol_assignments_.end();
1110 (*p)->add_to_table(symtab);
1111 this->script_sections_.add_symbols_to_table(symtab);
1114 // Finalize symbol values. Also check assertions.
1117 Script_options::finalize_symbols(Symbol_table* symtab, const Layout* layout)
1119 // We finalize the symbols defined in SECTIONS first, because they
1120 // are the ones which may have changed. This way if symbol outside
1121 // SECTIONS are defined in terms of symbols inside SECTIONS, they
1122 // will get the right value.
1123 this->script_sections_.finalize_symbols(symtab, layout);
1125 for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1126 p != this->symbol_assignments_.end();
1128 (*p)->finalize(symtab, layout);
1130 for (Assertions::iterator p = this->assertions_.begin();
1131 p != this->assertions_.end();
1133 (*p)->check(symtab, layout);
1136 // Set section addresses. We set all the symbols which have absolute
1137 // values. Then we let the SECTIONS clause do its thing. This
1138 // returns the segment which holds the file header and segment
1142 Script_options::set_section_addresses(Symbol_table* symtab, Layout* layout)
1144 for (Symbol_assignments::iterator p = this->symbol_assignments_.begin();
1145 p != this->symbol_assignments_.end();
1147 (*p)->set_if_absolute(symtab, layout, false, 0);
1149 return this->script_sections_.set_section_addresses(symtab, layout);
1152 // This class holds data passed through the parser to the lexer and to
1153 // the parser support functions. This avoids global variables. We
1154 // can't use global variables because we need not be called by a
1155 // singleton thread.
1157 class Parser_closure
1160 Parser_closure(const char* filename,
1161 const Position_dependent_options& posdep_options,
1162 bool in_group, bool is_in_sysroot,
1163 Command_line* command_line,
1164 Script_options* script_options,
1166 : filename_(filename), posdep_options_(posdep_options),
1167 in_group_(in_group), is_in_sysroot_(is_in_sysroot),
1168 command_line_(command_line), script_options_(script_options),
1169 version_script_info_(script_options->version_script_info()),
1170 lex_(lex), lineno_(0), charpos_(0), lex_mode_stack_(), inputs_(NULL)
1172 // We start out processing C symbols in the default lex mode.
1173 language_stack_.push_back("");
1174 lex_mode_stack_.push_back(lex->mode());
1177 // Return the file name.
1180 { return this->filename_; }
1182 // Return the position dependent options. The caller may modify
1184 Position_dependent_options&
1185 position_dependent_options()
1186 { return this->posdep_options_; }
1188 // Return whether this script is being run in a group.
1191 { return this->in_group_; }
1193 // Return whether this script was found using a directory in the
1196 is_in_sysroot() const
1197 { return this->is_in_sysroot_; }
1199 // Returns the Command_line structure passed in at constructor time.
1200 // This value may be NULL. The caller may modify this, which modifies
1201 // the passed-in Command_line object (not a copy).
1204 { return this->command_line_; }
1206 // Return the options which may be set by a script.
1209 { return this->script_options_; }
1211 // Return the object in which version script information should be stored.
1212 Version_script_info*
1214 { return this->version_script_info_; }
1216 // Return the next token, and advance.
1220 const Token* token = this->lex_->next_token();
1221 this->lineno_ = token->lineno();
1222 this->charpos_ = token->charpos();
1226 // Set a new lexer mode, pushing the current one.
1228 push_lex_mode(Lex::Mode mode)
1230 this->lex_mode_stack_.push_back(this->lex_->mode());
1231 this->lex_->set_mode(mode);
1234 // Pop the lexer mode.
1238 gold_assert(!this->lex_mode_stack_.empty());
1239 this->lex_->set_mode(this->lex_mode_stack_.back());
1240 this->lex_mode_stack_.pop_back();
1243 // Return the current lexer mode.
1246 { return this->lex_mode_stack_.back(); }
1248 // Return the line number of the last token.
1251 { return this->lineno_; }
1253 // Return the character position in the line of the last token.
1256 { return this->charpos_; }
1258 // Return the list of input files, creating it if necessary. This
1259 // is a space leak--we never free the INPUTS_ pointer.
1263 if (this->inputs_ == NULL)
1264 this->inputs_ = new Input_arguments();
1265 return this->inputs_;
1268 // Return whether we saw any input files.
1271 { return this->inputs_ != NULL && !this->inputs_->empty(); }
1273 // Return the current language being processed in a version script
1274 // (eg, "C++"). The empty string represents unmangled C names.
1276 get_current_language() const
1277 { return this->language_stack_.back(); }
1279 // Push a language onto the stack when entering an extern block.
1280 void push_language(const std::string& lang)
1281 { this->language_stack_.push_back(lang); }
1283 // Pop a language off of the stack when exiting an extern block.
1286 gold_assert(!this->language_stack_.empty());
1287 this->language_stack_.pop_back();
1291 // The name of the file we are reading.
1292 const char* filename_;
1293 // The position dependent options.
1294 Position_dependent_options posdep_options_;
1295 // Whether we are currently in a --start-group/--end-group.
1297 // Whether the script was found in a sysrooted directory.
1298 bool is_in_sysroot_;
1299 // May be NULL if the user chooses not to pass one in.
1300 Command_line* command_line_;
1301 // Options which may be set from any linker script.
1302 Script_options* script_options_;
1303 // Information parsed from a version script.
1304 Version_script_info* version_script_info_;
1307 // The line number of the last token returned by next_token.
1309 // The column number of the last token returned by next_token.
1311 // A stack of lexer modes.
1312 std::vector<Lex::Mode> lex_mode_stack_;
1313 // A stack of which extern/language block we're inside. Can be C++,
1314 // java, or empty for C.
1315 std::vector<std::string> language_stack_;
1316 // New input files found to add to the link.
1317 Input_arguments* inputs_;
1320 // FILE was found as an argument on the command line. Try to read it
1321 // as a script. Return true if the file was handled.
1324 read_input_script(Workqueue* workqueue, const General_options& options,
1325 Symbol_table* symtab, Layout* layout,
1326 Dirsearch* dirsearch, Input_objects* input_objects,
1327 Mapfile* mapfile, Input_group* input_group,
1328 const Input_argument* input_argument,
1329 Input_file* input_file, Task_token* next_blocker,
1330 bool* used_next_blocker)
1332 *used_next_blocker = false;
1334 std::string input_string;
1335 Lex::read_file(input_file, &input_string);
1337 Lex lex(input_string.c_str(), input_string.length(), PARSING_LINKER_SCRIPT);
1339 Parser_closure closure(input_file->filename().c_str(),
1340 input_argument->file().options(),
1341 input_group != NULL,
1342 input_file->is_in_sysroot(),
1344 layout->script_options(),
1347 if (yyparse(&closure) != 0)
1350 if (!closure.saw_inputs())
1353 Task_token* this_blocker = NULL;
1354 for (Input_arguments::const_iterator p = closure.inputs()->begin();
1355 p != closure.inputs()->end();
1359 if (p + 1 == closure.inputs()->end())
1363 nb = new Task_token(true);
1366 workqueue->queue_soon(new Read_symbols(options, input_objects, symtab,
1367 layout, dirsearch, mapfile, &*p,
1368 input_group, this_blocker, nb));
1372 *used_next_blocker = true;
1377 // Helper function for read_version_script() and
1378 // read_commandline_script(). Processes the given file in the mode
1379 // indicated by first_token and lex_mode.
1382 read_script_file(const char* filename, Command_line* cmdline,
1383 Script_options* script_options,
1384 int first_token, Lex::Mode lex_mode)
1386 // TODO: if filename is a relative filename, search for it manually
1387 // using "." + cmdline->options()->search_path() -- not dirsearch.
1388 Dirsearch dirsearch;
1390 // The file locking code wants to record a Task, but we haven't
1391 // started the workqueue yet. This is only for debugging purposes,
1392 // so we invent a fake value.
1393 const Task* task = reinterpret_cast<const Task*>(-1);
1395 // We don't want this file to be opened in binary mode.
1396 Position_dependent_options posdep = cmdline->position_dependent_options();
1397 if (posdep.format_enum() == General_options::OBJECT_FORMAT_BINARY)
1398 posdep.set_format_enum(General_options::OBJECT_FORMAT_ELF);
1399 Input_file_argument input_argument(filename, false, "", false, posdep);
1400 Input_file input_file(&input_argument);
1401 if (!input_file.open(cmdline->options(), dirsearch, task))
1404 std::string input_string;
1405 Lex::read_file(&input_file, &input_string);
1407 Lex lex(input_string.c_str(), input_string.length(), first_token);
1408 lex.set_mode(lex_mode);
1410 Parser_closure closure(filename,
1411 cmdline->position_dependent_options(),
1413 input_file.is_in_sysroot(),
1417 if (yyparse(&closure) != 0)
1419 input_file.file().unlock(task);
1423 input_file.file().unlock(task);
1425 gold_assert(!closure.saw_inputs());
1430 // FILENAME was found as an argument to --script (-T).
1431 // Read it as a script, and execute its contents immediately.
1434 read_commandline_script(const char* filename, Command_line* cmdline)
1436 return read_script_file(filename, cmdline, &cmdline->script_options(),
1437 PARSING_LINKER_SCRIPT, Lex::LINKER_SCRIPT);
1440 // FILENAME was found as an argument to --version-script. Read it as
1441 // a version script, and store its contents in
1442 // cmdline->script_options()->version_script_info().
1445 read_version_script(const char* filename, Command_line* cmdline)
1447 return read_script_file(filename, cmdline, &cmdline->script_options(),
1448 PARSING_VERSION_SCRIPT, Lex::VERSION_SCRIPT);
1451 // FILENAME was found as an argument to --dynamic-list. Read it as a
1452 // list of symbols, and store its contents in DYNAMIC_LIST.
1455 read_dynamic_list(const char* filename, Command_line* cmdline,
1456 Script_options* dynamic_list)
1458 return read_script_file(filename, cmdline, dynamic_list,
1459 PARSING_DYNAMIC_LIST, Lex::DYNAMIC_LIST);
1462 // Implement the --defsym option on the command line. Return true if
1466 Script_options::define_symbol(const char* definition)
1468 Lex lex(definition, strlen(definition), PARSING_DEFSYM);
1469 lex.set_mode(Lex::EXPRESSION);
1472 Position_dependent_options posdep_options;
1474 Parser_closure closure("command line", posdep_options, false, false, NULL,
1477 if (yyparse(&closure) != 0)
1480 gold_assert(!closure.saw_inputs());
1485 // Print the script to F for debugging.
1488 Script_options::print(FILE* f) const
1490 fprintf(f, "%s: Dumping linker script\n", program_name);
1492 if (!this->entry_.empty())
1493 fprintf(f, "ENTRY(%s)\n", this->entry_.c_str());
1495 for (Symbol_assignments::const_iterator p =
1496 this->symbol_assignments_.begin();
1497 p != this->symbol_assignments_.end();
1501 for (Assertions::const_iterator p = this->assertions_.begin();
1502 p != this->assertions_.end();
1506 this->script_sections_.print(f);
1508 this->version_script_info_.print(f);
1511 // Manage mapping from keywords to the codes expected by the bison
1512 // parser. We construct one global object for each lex mode with
1515 class Keyword_to_parsecode
1518 // The structure which maps keywords to parsecodes.
1519 struct Keyword_parsecode
1522 const char* keyword;
1523 // Corresponding parsecode.
1527 Keyword_to_parsecode(const Keyword_parsecode* keywords,
1529 : keyword_parsecodes_(keywords), keyword_count_(keyword_count)
1532 // Return the parsecode corresponding KEYWORD, or 0 if it is not a
1535 keyword_to_parsecode(const char* keyword, size_t len) const;
1538 const Keyword_parsecode* keyword_parsecodes_;
1539 const int keyword_count_;
1542 // Mapping from keyword string to keyword parsecode. This array must
1543 // be kept in sorted order. Parsecodes are looked up using bsearch.
1544 // This array must correspond to the list of parsecodes in yyscript.y.
1546 static const Keyword_to_parsecode::Keyword_parsecode
1547 script_keyword_parsecodes[] =
1549 { "ABSOLUTE", ABSOLUTE },
1551 { "ALIGN", ALIGN_K },
1552 { "ALIGNOF", ALIGNOF },
1553 { "ASSERT", ASSERT_K },
1554 { "AS_NEEDED", AS_NEEDED },
1559 { "CONSTANT", CONSTANT },
1560 { "CONSTRUCTORS", CONSTRUCTORS },
1561 { "CREATE_OBJECT_SYMBOLS", CREATE_OBJECT_SYMBOLS },
1562 { "DATA_SEGMENT_ALIGN", DATA_SEGMENT_ALIGN },
1563 { "DATA_SEGMENT_END", DATA_SEGMENT_END },
1564 { "DATA_SEGMENT_RELRO_END", DATA_SEGMENT_RELRO_END },
1565 { "DEFINED", DEFINED },
1567 { "EXCLUDE_FILE", EXCLUDE_FILE },
1568 { "EXTERN", EXTERN },
1571 { "FORCE_COMMON_ALLOCATION", FORCE_COMMON_ALLOCATION },
1574 { "INCLUDE", INCLUDE },
1575 { "INHIBIT_COMMON_ALLOCATION", INHIBIT_COMMON_ALLOCATION },
1578 { "LENGTH", LENGTH },
1579 { "LOADADDR", LOADADDR },
1583 { "MEMORY", MEMORY },
1586 { "NOCROSSREFS", NOCROSSREFS },
1587 { "NOFLOAT", NOFLOAT },
1588 { "ONLY_IF_RO", ONLY_IF_RO },
1589 { "ONLY_IF_RW", ONLY_IF_RW },
1590 { "OPTION", OPTION },
1591 { "ORIGIN", ORIGIN },
1592 { "OUTPUT", OUTPUT },
1593 { "OUTPUT_ARCH", OUTPUT_ARCH },
1594 { "OUTPUT_FORMAT", OUTPUT_FORMAT },
1595 { "OVERLAY", OVERLAY },
1597 { "PROVIDE", PROVIDE },
1598 { "PROVIDE_HIDDEN", PROVIDE_HIDDEN },
1600 { "SEARCH_DIR", SEARCH_DIR },
1601 { "SECTIONS", SECTIONS },
1602 { "SEGMENT_START", SEGMENT_START },
1604 { "SIZEOF", SIZEOF },
1605 { "SIZEOF_HEADERS", SIZEOF_HEADERS },
1606 { "SORT", SORT_BY_NAME },
1607 { "SORT_BY_ALIGNMENT", SORT_BY_ALIGNMENT },
1608 { "SORT_BY_NAME", SORT_BY_NAME },
1609 { "SPECIAL", SPECIAL },
1611 { "STARTUP", STARTUP },
1612 { "SUBALIGN", SUBALIGN },
1613 { "SYSLIB", SYSLIB },
1614 { "TARGET", TARGET_K },
1615 { "TRUNCATE", TRUNCATE },
1616 { "VERSION", VERSIONK },
1617 { "global", GLOBAL },
1623 { "sizeof_headers", SIZEOF_HEADERS },
1626 static const Keyword_to_parsecode
1627 script_keywords(&script_keyword_parsecodes[0],
1628 (sizeof(script_keyword_parsecodes)
1629 / sizeof(script_keyword_parsecodes[0])));
1631 static const Keyword_to_parsecode::Keyword_parsecode
1632 version_script_keyword_parsecodes[] =
1634 { "extern", EXTERN },
1635 { "global", GLOBAL },
1639 static const Keyword_to_parsecode
1640 version_script_keywords(&version_script_keyword_parsecodes[0],
1641 (sizeof(version_script_keyword_parsecodes)
1642 / sizeof(version_script_keyword_parsecodes[0])));
1644 static const Keyword_to_parsecode::Keyword_parsecode
1645 dynamic_list_keyword_parsecodes[] =
1647 { "extern", EXTERN },
1650 static const Keyword_to_parsecode
1651 dynamic_list_keywords(&dynamic_list_keyword_parsecodes[0],
1652 (sizeof(dynamic_list_keyword_parsecodes)
1653 / sizeof(dynamic_list_keyword_parsecodes[0])));
1657 // Comparison function passed to bsearch.
1669 ktt_compare(const void* keyv, const void* kttv)
1671 const Ktt_key* key = static_cast<const Ktt_key*>(keyv);
1672 const Keyword_to_parsecode::Keyword_parsecode* ktt =
1673 static_cast<const Keyword_to_parsecode::Keyword_parsecode*>(kttv);
1674 int i = strncmp(key->str, ktt->keyword, key->len);
1677 if (ktt->keyword[key->len] != '\0')
1682 } // End extern "C".
1685 Keyword_to_parsecode::keyword_to_parsecode(const char* keyword,
1691 void* kttv = bsearch(&key,
1692 this->keyword_parsecodes_,
1693 this->keyword_count_,
1694 sizeof(this->keyword_parsecodes_[0]),
1698 Keyword_parsecode* ktt = static_cast<Keyword_parsecode*>(kttv);
1699 return ktt->parsecode;
1702 // Helper class that calls cplus_demangle when needed and takes care of freeing
1705 class Lazy_demangler
1708 Lazy_demangler(const char* symbol, int options)
1709 : symbol_(symbol), options_(options), demangled_(NULL), did_demangle_(false)
1713 { free(this->demangled_); }
1715 // Return the demangled name. The actual demangling happens on the first call,
1716 // and the result is later cached.
1722 // The symbol to demangle.
1723 const char *symbol_;
1724 // Option flags to pass to cplus_demagle.
1726 // The cached demangled value, or NULL if demangling didn't happen yet or
1729 // Whether we already called cplus_demangle
1733 // Return the demangled name. The actual demangling happens on the first call,
1734 // and the result is later cached. Returns NULL if the symbol cannot be
1738 Lazy_demangler::get()
1740 if (!this->did_demangle_)
1742 this->demangled_ = cplus_demangle(this->symbol_, this->options_);
1743 this->did_demangle_ = true;
1745 return this->demangled_;
1748 // The following structs are used within the VersionInfo class as well
1749 // as in the bison helper functions. They store the information
1750 // parsed from the version script.
1752 // A single version expression.
1753 // For example, pattern="std::map*" and language="C++".
1754 // pattern and language should be from the stringpool
1755 struct Version_expression {
1756 Version_expression(const std::string& pattern,
1757 const std::string& language,
1759 : pattern(pattern), language(language), exact_match(exact_match) {}
1761 std::string pattern;
1762 std::string language;
1763 // If false, we use glob() to match pattern. If true, we use strcmp().
1768 // A list of expressions.
1769 struct Version_expression_list {
1770 std::vector<struct Version_expression> expressions;
1774 // A list of which versions upon which another version depends.
1775 // Strings should be from the Stringpool.
1776 struct Version_dependency_list {
1777 std::vector<std::string> dependencies;
1781 // The total definition of a version. It includes the tag for the
1782 // version, its global and local expressions, and any dependencies.
1783 struct Version_tree {
1785 : tag(), global(NULL), local(NULL), dependencies(NULL) {}
1788 const struct Version_expression_list* global;
1789 const struct Version_expression_list* local;
1790 const struct Version_dependency_list* dependencies;
1793 Version_script_info::~Version_script_info()
1799 Version_script_info::clear()
1801 for (size_t k = 0; k < dependency_lists_.size(); ++k)
1802 delete dependency_lists_[k];
1803 this->dependency_lists_.clear();
1804 for (size_t k = 0; k < version_trees_.size(); ++k)
1805 delete version_trees_[k];
1806 this->version_trees_.clear();
1807 for (size_t k = 0; k < expression_lists_.size(); ++k)
1808 delete expression_lists_[k];
1809 this->expression_lists_.clear();
1812 std::vector<std::string>
1813 Version_script_info::get_versions() const
1815 std::vector<std::string> ret;
1816 for (size_t j = 0; j < version_trees_.size(); ++j)
1817 if (!this->version_trees_[j]->tag.empty())
1818 ret.push_back(this->version_trees_[j]->tag);
1822 std::vector<std::string>
1823 Version_script_info::get_dependencies(const char* version) const
1825 std::vector<std::string> ret;
1826 for (size_t j = 0; j < version_trees_.size(); ++j)
1827 if (version_trees_[j]->tag == version)
1829 const struct Version_dependency_list* deps =
1830 version_trees_[j]->dependencies;
1832 for (size_t k = 0; k < deps->dependencies.size(); ++k)
1833 ret.push_back(deps->dependencies[k]);
1839 // Look up SYMBOL_NAME in the list of versions. If CHECK_GLOBAL is
1840 // true look at the globally visible symbols, otherwise look at the
1841 // symbols listed as "local:". Return true if the symbol is found,
1842 // false otherwise. If the symbol is found, then if PVERSION is not
1843 // NULL, set *PVERSION to the version.
1846 Version_script_info::get_symbol_version_helper(const char* symbol_name,
1848 std::string* pversion) const
1850 Lazy_demangler cpp_demangled_name(symbol_name, DMGL_ANSI | DMGL_PARAMS);
1851 Lazy_demangler java_demangled_name(symbol_name,
1852 DMGL_ANSI | DMGL_PARAMS | DMGL_JAVA);
1853 for (size_t j = 0; j < version_trees_.size(); ++j)
1855 // Is it a global symbol for this version?
1856 const Version_expression_list* explist =
1857 check_global ? version_trees_[j]->global : version_trees_[j]->local;
1858 if (explist != NULL)
1859 for (size_t k = 0; k < explist->expressions.size(); ++k)
1861 const char* name_to_match = symbol_name;
1862 const struct Version_expression& exp = explist->expressions[k];
1863 if (exp.language == "C++")
1865 name_to_match = cpp_demangled_name.get();
1866 // This isn't a C++ symbol.
1867 if (name_to_match == NULL)
1870 else if (exp.language == "Java")
1872 name_to_match = java_demangled_name.get();
1873 // This isn't a Java symbol.
1874 if (name_to_match == NULL)
1878 if (exp.exact_match)
1879 matched = strcmp(exp.pattern.c_str(), name_to_match) == 0;
1881 matched = fnmatch(exp.pattern.c_str(), name_to_match,
1885 if (pversion != NULL)
1886 *pversion = this->version_trees_[j]->tag;
1894 struct Version_dependency_list*
1895 Version_script_info::allocate_dependency_list()
1897 dependency_lists_.push_back(new Version_dependency_list);
1898 return dependency_lists_.back();
1901 struct Version_expression_list*
1902 Version_script_info::allocate_expression_list()
1904 expression_lists_.push_back(new Version_expression_list);
1905 return expression_lists_.back();
1908 struct Version_tree*
1909 Version_script_info::allocate_version_tree()
1911 version_trees_.push_back(new Version_tree);
1912 return version_trees_.back();
1915 // Print for debugging.
1918 Version_script_info::print(FILE* f) const
1923 fprintf(f, "VERSION {");
1925 for (size_t i = 0; i < this->version_trees_.size(); ++i)
1927 const Version_tree* vt = this->version_trees_[i];
1929 if (vt->tag.empty())
1932 fprintf(f, " %s {\n", vt->tag.c_str());
1934 if (vt->global != NULL)
1936 fprintf(f, " global :\n");
1937 this->print_expression_list(f, vt->global);
1940 if (vt->local != NULL)
1942 fprintf(f, " local :\n");
1943 this->print_expression_list(f, vt->local);
1947 if (vt->dependencies != NULL)
1949 const Version_dependency_list* deps = vt->dependencies;
1950 for (size_t j = 0; j < deps->dependencies.size(); ++j)
1952 if (j < deps->dependencies.size() - 1)
1954 fprintf(f, " %s", deps->dependencies[j].c_str());
1964 Version_script_info::print_expression_list(
1966 const Version_expression_list* vel) const
1968 std::string current_language;
1969 for (size_t i = 0; i < vel->expressions.size(); ++i)
1971 const Version_expression& ve(vel->expressions[i]);
1973 if (ve.language != current_language)
1975 if (!current_language.empty())
1977 fprintf(f, " extern \"%s\" {\n", ve.language.c_str());
1978 current_language = ve.language;
1982 if (!current_language.empty())
1987 fprintf(f, "%s", ve.pattern.c_str());
1994 if (!current_language.empty())
1998 } // End namespace gold.
2000 // The remaining functions are extern "C", so it's clearer to not put
2001 // them in namespace gold.
2003 using namespace gold;
2005 // This function is called by the bison parser to return the next
2009 yylex(YYSTYPE* lvalp, void* closurev)
2011 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2012 const Token* token = closure->next_token();
2013 switch (token->classification())
2018 case Token::TOKEN_INVALID:
2019 yyerror(closurev, "invalid character");
2022 case Token::TOKEN_EOF:
2025 case Token::TOKEN_STRING:
2027 // This is either a keyword or a STRING.
2029 const char* str = token->string_value(&len);
2031 switch (closure->lex_mode())
2033 case Lex::LINKER_SCRIPT:
2034 parsecode = script_keywords.keyword_to_parsecode(str, len);
2036 case Lex::VERSION_SCRIPT:
2037 parsecode = version_script_keywords.keyword_to_parsecode(str, len);
2039 case Lex::DYNAMIC_LIST:
2040 parsecode = dynamic_list_keywords.keyword_to_parsecode(str, len);
2047 lvalp->string.value = str;
2048 lvalp->string.length = len;
2052 case Token::TOKEN_QUOTED_STRING:
2053 lvalp->string.value = token->string_value(&lvalp->string.length);
2054 return QUOTED_STRING;
2056 case Token::TOKEN_OPERATOR:
2057 return token->operator_value();
2059 case Token::TOKEN_INTEGER:
2060 lvalp->integer = token->integer_value();
2065 // This function is called by the bison parser to report an error.
2068 yyerror(void* closurev, const char* message)
2070 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2071 gold_error(_("%s:%d:%d: %s"), closure->filename(), closure->lineno(),
2072 closure->charpos(), message);
2075 // Called by the bison parser to add a file to the link.
2078 script_add_file(void* closurev, const char* name, size_t length)
2080 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2082 // If this is an absolute path, and we found the script in the
2083 // sysroot, then we want to prepend the sysroot to the file name.
2084 // For example, this is how we handle a cross link to the x86_64
2085 // libc.so, which refers to /lib/libc.so.6.
2086 std::string name_string(name, length);
2087 const char* extra_search_path = ".";
2088 std::string script_directory;
2089 if (IS_ABSOLUTE_PATH(name_string.c_str()))
2091 if (closure->is_in_sysroot())
2093 const std::string& sysroot(parameters->options().sysroot());
2094 gold_assert(!sysroot.empty());
2095 name_string = sysroot + name_string;
2100 // In addition to checking the normal library search path, we
2101 // also want to check in the script-directory.
2102 const char *slash = strrchr(closure->filename(), '/');
2105 script_directory.assign(closure->filename(),
2106 slash - closure->filename() + 1);
2107 extra_search_path = script_directory.c_str();
2111 Input_file_argument file(name_string.c_str(), false, extra_search_path,
2112 false, closure->position_dependent_options());
2113 closure->inputs()->add_file(file);
2116 // Called by the bison parser to start a group. If we are already in
2117 // a group, that means that this script was invoked within a
2118 // --start-group --end-group sequence on the command line, or that
2119 // this script was found in a GROUP of another script. In that case,
2120 // we simply continue the existing group, rather than starting a new
2121 // one. It is possible to construct a case in which this will do
2122 // something other than what would happen if we did a recursive group,
2123 // but it's hard to imagine why the different behaviour would be
2124 // useful for a real program. Avoiding recursive groups is simpler
2125 // and more efficient.
2128 script_start_group(void* closurev)
2130 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2131 if (!closure->in_group())
2132 closure->inputs()->start_group();
2135 // Called by the bison parser at the end of a group.
2138 script_end_group(void* closurev)
2140 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2141 if (!closure->in_group())
2142 closure->inputs()->end_group();
2145 // Called by the bison parser to start an AS_NEEDED list.
2148 script_start_as_needed(void* closurev)
2150 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2151 closure->position_dependent_options().set_as_needed(true);
2154 // Called by the bison parser at the end of an AS_NEEDED list.
2157 script_end_as_needed(void* closurev)
2159 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2160 closure->position_dependent_options().set_as_needed(false);
2163 // Called by the bison parser to set the entry symbol.
2166 script_set_entry(void* closurev, const char* entry, size_t length)
2168 // We'll parse this exactly the same as --entry=ENTRY on the commandline
2169 // TODO(csilvers): FIXME -- call set_entry directly.
2170 std::string arg("--entry=");
2171 arg.append(entry, length);
2172 script_parse_option(closurev, arg.c_str(), arg.size());
2175 // Called by the bison parser to set whether to define common symbols.
2178 script_set_common_allocation(void* closurev, int set)
2180 const char* arg = set != 0 ? "--define-common" : "--no-define-common";
2181 script_parse_option(closurev, arg, strlen(arg));
2184 // Called by the bison parser to define a symbol.
2187 script_set_symbol(void* closurev, const char* name, size_t length,
2188 Expression* value, int providei, int hiddeni)
2190 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2191 const bool provide = providei != 0;
2192 const bool hidden = hiddeni != 0;
2193 closure->script_options()->add_symbol_assignment(name, length, value,
2197 // Called by the bison parser to add an assertion.
2200 script_add_assertion(void* closurev, Expression* check, const char* message,
2203 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2204 closure->script_options()->add_assertion(check, message, messagelen);
2207 // Called by the bison parser to parse an OPTION.
2210 script_parse_option(void* closurev, const char* option, size_t length)
2212 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2213 // We treat the option as a single command-line option, even if
2214 // it has internal whitespace.
2215 if (closure->command_line() == NULL)
2217 // There are some options that we could handle here--e.g.,
2218 // -lLIBRARY. Should we bother?
2219 gold_warning(_("%s:%d:%d: ignoring command OPTION; OPTION is only valid"
2220 " for scripts specified via -T/--script"),
2221 closure->filename(), closure->lineno(), closure->charpos());
2225 bool past_a_double_dash_option = false;
2226 const char* mutable_option = strndup(option, length);
2227 gold_assert(mutable_option != NULL);
2228 closure->command_line()->process_one_option(1, &mutable_option, 0,
2229 &past_a_double_dash_option);
2230 // The General_options class will quite possibly store a pointer
2231 // into mutable_option, so we can't free it. In cases the class
2232 // does not store such a pointer, this is a memory leak. Alas. :(
2236 // Called by the bison parser to handle SEARCH_DIR. This is handled
2237 // exactly like a -L option.
2240 script_add_search_dir(void* closurev, const char* option, size_t length)
2242 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2243 if (closure->command_line() == NULL)
2244 gold_warning(_("%s:%d:%d: ignoring SEARCH_DIR; SEARCH_DIR is only valid"
2245 " for scripts specified via -T/--script"),
2246 closure->filename(), closure->lineno(), closure->charpos());
2249 std::string s = "-L" + std::string(option, length);
2250 script_parse_option(closurev, s.c_str(), s.size());
2254 /* Called by the bison parser to push the lexer into expression
2258 script_push_lex_into_expression_mode(void* closurev)
2260 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2261 closure->push_lex_mode(Lex::EXPRESSION);
2264 /* Called by the bison parser to push the lexer into version
2268 script_push_lex_into_version_mode(void* closurev)
2270 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2271 closure->push_lex_mode(Lex::VERSION_SCRIPT);
2274 /* Called by the bison parser to pop the lexer mode. */
2277 script_pop_lex_mode(void* closurev)
2279 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2280 closure->pop_lex_mode();
2283 // Register an entire version node. For example:
2289 // - tag is "GLIBC_2.1"
2290 // - tree contains the information "global: foo"
2291 // - deps contains "GLIBC_2.0"
2294 script_register_vers_node(void*,
2297 struct Version_tree *tree,
2298 struct Version_dependency_list *deps)
2300 gold_assert(tree != NULL);
2301 tree->dependencies = deps;
2303 tree->tag = std::string(tag, taglen);
2306 // Add a dependencies to the list of existing dependencies, if any,
2307 // and return the expanded list.
2309 extern "C" struct Version_dependency_list *
2310 script_add_vers_depend(void* closurev,
2311 struct Version_dependency_list *all_deps,
2312 const char *depend_to_add, int deplen)
2314 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2315 if (all_deps == NULL)
2316 all_deps = closure->version_script()->allocate_dependency_list();
2317 all_deps->dependencies.push_back(std::string(depend_to_add, deplen));
2321 // Add a pattern expression to an existing list of expressions, if any.
2322 // TODO: In the old linker, the last argument used to be a bool, but I
2323 // don't know what it meant.
2325 extern "C" struct Version_expression_list *
2326 script_new_vers_pattern(void* closurev,
2327 struct Version_expression_list *expressions,
2328 const char *pattern, int patlen, int exact_match)
2330 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2331 if (expressions == NULL)
2332 expressions = closure->version_script()->allocate_expression_list();
2333 expressions->expressions.push_back(
2334 Version_expression(std::string(pattern, patlen),
2335 closure->get_current_language(),
2336 static_cast<bool>(exact_match)));
2340 // Attaches b to the end of a, and clears b. So a = a + b and b = {}.
2342 extern "C" struct Version_expression_list*
2343 script_merge_expressions(struct Version_expression_list *a,
2344 struct Version_expression_list *b)
2346 a->expressions.insert(a->expressions.end(),
2347 b->expressions.begin(), b->expressions.end());
2348 // We could delete b and remove it from expressions_lists_, but
2349 // that's a lot of work. This works just as well.
2350 b->expressions.clear();
2354 // Combine the global and local expressions into a a Version_tree.
2356 extern "C" struct Version_tree *
2357 script_new_vers_node(void* closurev,
2358 struct Version_expression_list *global,
2359 struct Version_expression_list *local)
2361 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2362 Version_tree* tree = closure->version_script()->allocate_version_tree();
2363 tree->global = global;
2364 tree->local = local;
2368 // Handle a transition in language, such as at the
2369 // start or end of 'extern "C++"'
2372 version_script_push_lang(void* closurev, const char* lang, int langlen)
2374 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2375 closure->push_language(std::string(lang, langlen));
2379 version_script_pop_lang(void* closurev)
2381 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2382 closure->pop_language();
2385 // Called by the bison parser to start a SECTIONS clause.
2388 script_start_sections(void* closurev)
2390 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2391 closure->script_options()->script_sections()->start_sections();
2394 // Called by the bison parser to finish a SECTIONS clause.
2397 script_finish_sections(void* closurev)
2399 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2400 closure->script_options()->script_sections()->finish_sections();
2403 // Start processing entries for an output section.
2406 script_start_output_section(void* closurev, const char* name, size_t namelen,
2407 const struct Parser_output_section_header* header)
2409 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2410 closure->script_options()->script_sections()->start_output_section(name,
2415 // Finish processing entries for an output section.
2418 script_finish_output_section(void* closurev,
2419 const struct Parser_output_section_trailer* trail)
2421 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2422 closure->script_options()->script_sections()->finish_output_section(trail);
2425 // Add a data item (e.g., "WORD (0)") to the current output section.
2428 script_add_data(void* closurev, int data_token, Expression* val)
2430 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2432 bool is_signed = true;
2454 closure->script_options()->script_sections()->add_data(size, is_signed, val);
2457 // Add a clause setting the fill value to the current output section.
2460 script_add_fill(void* closurev, Expression* val)
2462 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2463 closure->script_options()->script_sections()->add_fill(val);
2466 // Add a new input section specification to the current output
2470 script_add_input_section(void* closurev,
2471 const struct Input_section_spec* spec,
2474 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2475 bool keep = keepi != 0;
2476 closure->script_options()->script_sections()->add_input_section(spec, keep);
2479 // When we see DATA_SEGMENT_ALIGN we record that following output
2480 // sections may be relro.
2483 script_data_segment_align(void* closurev)
2485 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2486 if (!closure->script_options()->saw_sections_clause())
2487 gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
2488 closure->filename(), closure->lineno(), closure->charpos());
2490 closure->script_options()->script_sections()->data_segment_align();
2493 // When we see DATA_SEGMENT_RELRO_END we know that all output sections
2494 // since DATA_SEGMENT_ALIGN should be relro.
2497 script_data_segment_relro_end(void* closurev)
2499 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2500 if (!closure->script_options()->saw_sections_clause())
2501 gold_error(_("%s:%d:%d: DATA_SEGMENT_ALIGN not in SECTIONS clause"),
2502 closure->filename(), closure->lineno(), closure->charpos());
2504 closure->script_options()->script_sections()->data_segment_relro_end();
2507 // Create a new list of string/sort pairs.
2509 extern "C" String_sort_list_ptr
2510 script_new_string_sort_list(const struct Wildcard_section* string_sort)
2512 return new String_sort_list(1, *string_sort);
2515 // Add an entry to a list of string/sort pairs. The way the parser
2516 // works permits us to simply modify the first parameter, rather than
2519 extern "C" String_sort_list_ptr
2520 script_string_sort_list_add(String_sort_list_ptr pv,
2521 const struct Wildcard_section* string_sort)
2524 return script_new_string_sort_list(string_sort);
2527 pv->push_back(*string_sort);
2532 // Create a new list of strings.
2534 extern "C" String_list_ptr
2535 script_new_string_list(const char* str, size_t len)
2537 return new String_list(1, std::string(str, len));
2540 // Add an element to a list of strings. The way the parser works
2541 // permits us to simply modify the first parameter, rather than copy
2544 extern "C" String_list_ptr
2545 script_string_list_push_back(String_list_ptr pv, const char* str, size_t len)
2548 return script_new_string_list(str, len);
2551 pv->push_back(std::string(str, len));
2556 // Concatenate two string lists. Either or both may be NULL. The way
2557 // the parser works permits us to modify the parameters, rather than
2560 extern "C" String_list_ptr
2561 script_string_list_append(String_list_ptr pv1, String_list_ptr pv2)
2567 pv1->insert(pv1->end(), pv2->begin(), pv2->end());
2571 // Add a new program header.
2574 script_add_phdr(void* closurev, const char* name, size_t namelen,
2575 unsigned int type, const Phdr_info* info)
2577 Parser_closure* closure = static_cast<Parser_closure*>(closurev);
2578 bool includes_filehdr = info->includes_filehdr != 0;
2579 bool includes_phdrs = info->includes_phdrs != 0;
2580 bool is_flags_valid = info->is_flags_valid != 0;
2581 Script_sections* ss = closure->script_options()->script_sections();
2582 ss->add_phdr(name, namelen, type, includes_filehdr, includes_phdrs,
2583 is_flags_valid, info->flags, info->load_address);
2586 // Convert a program header string to a type.
2588 #define PHDR_TYPE(NAME) { #NAME, sizeof(#NAME) - 1, elfcpp::NAME }
2595 } phdr_type_names[] =
2599 PHDR_TYPE(PT_DYNAMIC),
2600 PHDR_TYPE(PT_INTERP),
2602 PHDR_TYPE(PT_SHLIB),
2605 PHDR_TYPE(PT_GNU_EH_FRAME),
2606 PHDR_TYPE(PT_GNU_STACK),
2607 PHDR_TYPE(PT_GNU_RELRO)
2610 extern "C" unsigned int
2611 script_phdr_string_to_type(void* closurev, const char* name, size_t namelen)
2613 for (unsigned int i = 0;
2614 i < sizeof(phdr_type_names) / sizeof(phdr_type_names[0]);
2616 if (namelen == phdr_type_names[i].namelen
2617 && strncmp(name, phdr_type_names[i].name, namelen) == 0)
2618 return phdr_type_names[i].val;
2619 yyerror(closurev, _("unknown PHDR type (try integer)"));
2620 return elfcpp::PT_NULL;