1/* YACC parser for C expressions, for GDB. 2 Copyright (C) 1986-2023 Free Software Foundation, Inc. 3 4 This file is part of GDB. 5 6 This program is free software; you can redistribute it and/or modify 7 it under the terms of the GNU General Public License as published by 8 the Free Software Foundation; either version 3 of the License, or 9 (at your option) any later version. 10 11 This program is distributed in the hope that it will be useful, 12 but WITHOUT ANY WARRANTY; without even the implied warranty of 13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 14 GNU General Public License for more details. 15 16 You should have received a copy of the GNU General Public License 17 along with this program. If not, see <http://www.gnu.org/licenses/>. */ 18 19/* Parse a C expression from text in a string, 20 and return the result as a struct expression pointer. 21 That structure contains arithmetic operations in reverse polish, 22 with constants represented by operations that are followed by special data. 23 See expression.h for the details of the format. 24 What is important here is that it can be built up sequentially 25 during the process of parsing; the lower levels of the tree always 26 come first in the result. 27 28 Note that malloc's and realloc's in this file are transformed to 29 xmalloc and xrealloc respectively by the same sed command in the 30 makefile that remaps any other malloc/realloc inserted by the parser 31 generator. Doing this with #defines and trying to control the interaction 32 with include files (<malloc.h> and <stdlib.h> for example) just became 33 too messy, particularly when such includes can be inserted at random 34 times by the parser generator. */ 35 36%{ 37 38#include "defs.h" 39#include <ctype.h> 40#include "expression.h" 41#include "value.h" 42#include "parser-defs.h" 43#include "language.h" 44#include "c-lang.h" 45#include "c-support.h" 46#include "bfd.h" /* Required by objfiles.h. */ 47#include "symfile.h" /* Required by objfiles.h. */ 48#include "objfiles.h" /* For have_full_symbols and have_partial_symbols */ 49#include "charset.h" 50#include "block.h" 51#include "cp-support.h" 52#include "macroscope.h" 53#include "objc-lang.h" 54#include "typeprint.h" 55#include "cp-abi.h" 56#include "type-stack.h" 57#include "target-float.h" 58#include "c-exp.h" 59 60#define parse_type(ps) builtin_type (ps->gdbarch ()) 61 62/* Remap normal yacc parser interface names (yyparse, yylex, yyerror, 63 etc). */ 64#define GDB_YY_REMAP_PREFIX c_ 65#include "yy-remap.h" 66 67/* The state of the parser, used internally when we are parsing the 68 expression. */ 69 70static struct parser_state *pstate = NULL; 71 72/* Data that must be held for the duration of a parse. */ 73 74struct c_parse_state 75{ 76 /* These are used to hold type lists and type stacks that are 77 allocated during the parse. */ 78 std::vector<std::unique_ptr<std::vector<struct type *>>> type_lists; 79 std::vector<std::unique_ptr<struct type_stack>> type_stacks; 80 81 /* Storage for some strings allocated during the parse. */ 82 std::vector<gdb::unique_xmalloc_ptr<char>> strings; 83 84 /* When we find that lexptr (the global var defined in parse.c) is 85 pointing at a macro invocation, we expand the invocation, and call 86 scan_macro_expansion to save the old lexptr here and point lexptr 87 into the expanded text. When we reach the end of that, we call 88 end_macro_expansion to pop back to the value we saved here. The 89 macro expansion code promises to return only fully-expanded text, 90 so we don't need to "push" more than one level. 91 92 This is disgusting, of course. It would be cleaner to do all macro 93 expansion beforehand, and then hand that to lexptr. But we don't 94 really know where the expression ends. Remember, in a command like 95 96 (gdb) break *ADDRESS if CONDITION 97 98 we evaluate ADDRESS in the scope of the current frame, but we 99 evaluate CONDITION in the scope of the breakpoint's location. So 100 it's simply wrong to try to macro-expand the whole thing at once. */ 101 const char *macro_original_text = nullptr; 102 103 /* We save all intermediate macro expansions on this obstack for the 104 duration of a single parse. The expansion text may sometimes have 105 to live past the end of the expansion, due to yacc lookahead. 106 Rather than try to be clever about saving the data for a single 107 token, we simply keep it all and delete it after parsing has 108 completed. */ 109 auto_obstack expansion_obstack; 110 111 /* The type stack. */ 112 struct type_stack type_stack; 113}; 114 115/* This is set and cleared in c_parse. */ 116 117static struct c_parse_state *cpstate; 118 119int yyparse (void); 120 121static int yylex (void); 122 123static void yyerror (const char *); 124 125static int type_aggregate_p (struct type *); 126 127using namespace expr; 128%} 129 130/* Although the yacc "value" of an expression is not used, 131 since the result is stored in the structure being created, 132 other node types do have values. */ 133 134%union 135 { 136 LONGEST lval; 137 struct { 138 LONGEST val; 139 struct type *type; 140 } typed_val_int; 141 struct { 142 gdb_byte val[16]; 143 struct type *type; 144 } typed_val_float; 145 struct type *tval; 146 struct stoken sval; 147 struct typed_stoken tsval; 148 struct ttype tsym; 149 struct symtoken ssym; 150 int voidval; 151 const struct block *bval; 152 enum exp_opcode opcode; 153 154 struct stoken_vector svec; 155 std::vector<struct type *> *tvec; 156 157 struct type_stack *type_stack; 158 159 struct objc_class_str theclass; 160 } 161 162%{ 163/* YYSTYPE gets defined by %union */ 164static int parse_number (struct parser_state *par_state, 165 const char *, int, int, YYSTYPE *); 166static struct stoken operator_stoken (const char *); 167static struct stoken typename_stoken (const char *); 168static void check_parameter_typelist (std::vector<struct type *> *); 169 170#if defined(YYBISON) && YYBISON < 30800 171static void c_print_token (FILE *file, int type, YYSTYPE value); 172#define YYPRINT(FILE, TYPE, VALUE) c_print_token (FILE, TYPE, VALUE) 173#endif 174%} 175 176%type <voidval> exp exp1 type_exp start variable qualified_name lcurly function_method 177%type <lval> rcurly 178%type <tval> type typebase scalar_type 179%type <tvec> nonempty_typelist func_mod parameter_typelist 180/* %type <bval> block */ 181 182/* Fancy type parsing. */ 183%type <tval> ptype 184%type <lval> array_mod 185%type <tval> conversion_type_id 186 187%type <type_stack> ptr_operator_ts abs_decl direct_abs_decl 188 189%token <typed_val_int> INT COMPLEX_INT 190%token <typed_val_float> FLOAT COMPLEX_FLOAT 191 192/* Both NAME and TYPENAME tokens represent symbols in the input, 193 and both convey their data as strings. 194 But a TYPENAME is a string that happens to be defined as a typedef 195 or builtin type name (such as int or char) 196 and a NAME is any other symbol. 197 Contexts where this distinction is not important can use the 198 nonterminal "name", which matches either NAME or TYPENAME. */ 199 200%token <tsval> STRING 201%token <sval> NSSTRING /* ObjC Foundation "NSString" literal */ 202%token SELECTOR /* ObjC "@selector" pseudo-operator */ 203%token <tsval> CHAR 204%token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */ 205%token <ssym> UNKNOWN_CPP_NAME 206%token <voidval> COMPLETE 207%token <tsym> TYPENAME 208%token <theclass> CLASSNAME /* ObjC Class name */ 209%type <sval> name field_name 210%type <svec> string_exp 211%type <ssym> name_not_typename 212%type <tsym> type_name 213 214 /* This is like a '[' token, but is only generated when parsing 215 Objective C. This lets us reuse the same parser without 216 erroneously parsing ObjC-specific expressions in C. */ 217%token OBJC_LBRAC 218 219/* A NAME_OR_INT is a symbol which is not known in the symbol table, 220 but which would parse as a valid number in the current input radix. 221 E.g. "c" when input_radix==16. Depending on the parse, it will be 222 turned into a name or into a number. */ 223 224%token <ssym> NAME_OR_INT 225 226%token OPERATOR 227%token STRUCT CLASS UNION ENUM SIZEOF ALIGNOF UNSIGNED COLONCOLON 228%token TEMPLATE 229%token ERROR 230%token NEW DELETE 231%type <sval> oper 232%token REINTERPRET_CAST DYNAMIC_CAST STATIC_CAST CONST_CAST 233%token ENTRY 234%token TYPEOF 235%token DECLTYPE 236%token TYPEID 237 238/* Special type cases, put in to allow the parser to distinguish different 239 legal basetypes. */ 240%token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD 241%token RESTRICT ATOMIC 242%token FLOAT_KEYWORD COMPLEX 243 244%token <sval> DOLLAR_VARIABLE 245 246%token <opcode> ASSIGN_MODIFY 247 248/* C++ */ 249%token TRUEKEYWORD 250%token FALSEKEYWORD 251 252 253%left ',' 254%left ABOVE_COMMA 255%right '=' ASSIGN_MODIFY 256%right '?' 257%left OROR 258%left ANDAND 259%left '|' 260%left '^' 261%left '&' 262%left EQUAL NOTEQUAL 263%left '<' '>' LEQ GEQ 264%left LSH RSH 265%left '@' 266%left '+' '-' 267%left '*' '/' '%' 268%right UNARY INCREMENT DECREMENT 269%right ARROW ARROW_STAR '.' DOT_STAR '[' OBJC_LBRAC '(' 270%token <ssym> BLOCKNAME 271%token <bval> FILENAME 272%type <bval> block 273%left COLONCOLON 274 275%token DOTDOTDOT 276 277 278%% 279 280start : exp1 281 | type_exp 282 ; 283 284type_exp: type 285 { 286 pstate->push_new<type_operation> ($1); 287 } 288 | TYPEOF '(' exp ')' 289 { 290 pstate->wrap<typeof_operation> (); 291 } 292 | TYPEOF '(' type ')' 293 { 294 pstate->push_new<type_operation> ($3); 295 } 296 | DECLTYPE '(' exp ')' 297 { 298 pstate->wrap<decltype_operation> (); 299 } 300 ; 301 302/* Expressions, including the comma operator. */ 303exp1 : exp 304 | exp1 ',' exp 305 { pstate->wrap2<comma_operation> (); } 306 ; 307 308/* Expressions, not including the comma operator. */ 309exp : '*' exp %prec UNARY 310 { pstate->wrap<unop_ind_operation> (); } 311 ; 312 313exp : '&' exp %prec UNARY 314 { pstate->wrap<unop_addr_operation> (); } 315 ; 316 317exp : '-' exp %prec UNARY 318 { pstate->wrap<unary_neg_operation> (); } 319 ; 320 321exp : '+' exp %prec UNARY 322 { pstate->wrap<unary_plus_operation> (); } 323 ; 324 325exp : '!' exp %prec UNARY 326 { 327 if (pstate->language ()->la_language 328 == language_opencl) 329 pstate->wrap<opencl_not_operation> (); 330 else 331 pstate->wrap<unary_logical_not_operation> (); 332 } 333 ; 334 335exp : '~' exp %prec UNARY 336 { pstate->wrap<unary_complement_operation> (); } 337 ; 338 339exp : INCREMENT exp %prec UNARY 340 { pstate->wrap<preinc_operation> (); } 341 ; 342 343exp : DECREMENT exp %prec UNARY 344 { pstate->wrap<predec_operation> (); } 345 ; 346 347exp : exp INCREMENT %prec UNARY 348 { pstate->wrap<postinc_operation> (); } 349 ; 350 351exp : exp DECREMENT %prec UNARY 352 { pstate->wrap<postdec_operation> (); } 353 ; 354 355exp : TYPEID '(' exp ')' %prec UNARY 356 { pstate->wrap<typeid_operation> (); } 357 ; 358 359exp : TYPEID '(' type_exp ')' %prec UNARY 360 { pstate->wrap<typeid_operation> (); } 361 ; 362 363exp : SIZEOF exp %prec UNARY 364 { pstate->wrap<unop_sizeof_operation> (); } 365 ; 366 367exp : ALIGNOF '(' type_exp ')' %prec UNARY 368 { pstate->wrap<unop_alignof_operation> (); } 369 ; 370 371exp : exp ARROW field_name 372 { 373 pstate->push_new<structop_ptr_operation> 374 (pstate->pop (), copy_name ($3)); 375 } 376 ; 377 378exp : exp ARROW field_name COMPLETE 379 { 380 structop_base_operation *op 381 = new structop_ptr_operation (pstate->pop (), 382 copy_name ($3)); 383 pstate->mark_struct_expression (op); 384 pstate->push (operation_up (op)); 385 } 386 ; 387 388exp : exp ARROW COMPLETE 389 { 390 structop_base_operation *op 391 = new structop_ptr_operation (pstate->pop (), ""); 392 pstate->mark_struct_expression (op); 393 pstate->push (operation_up (op)); 394 } 395 ; 396 397exp : exp ARROW '~' name 398 { 399 pstate->push_new<structop_ptr_operation> 400 (pstate->pop (), "~" + copy_name ($4)); 401 } 402 ; 403 404exp : exp ARROW '~' name COMPLETE 405 { 406 structop_base_operation *op 407 = new structop_ptr_operation (pstate->pop (), 408 "~" + copy_name ($4)); 409 pstate->mark_struct_expression (op); 410 pstate->push (operation_up (op)); 411 } 412 ; 413 414exp : exp ARROW qualified_name 415 { /* exp->type::name becomes exp->*(&type::name) */ 416 /* Note: this doesn't work if name is a 417 static member! FIXME */ 418 pstate->wrap<unop_addr_operation> (); 419 pstate->wrap2<structop_mptr_operation> (); } 420 ; 421 422exp : exp ARROW_STAR exp 423 { pstate->wrap2<structop_mptr_operation> (); } 424 ; 425 426exp : exp '.' field_name 427 { 428 if (pstate->language ()->la_language 429 == language_opencl) 430 pstate->push_new<opencl_structop_operation> 431 (pstate->pop (), copy_name ($3)); 432 else 433 pstate->push_new<structop_operation> 434 (pstate->pop (), copy_name ($3)); 435 } 436 ; 437 438exp : exp '.' field_name COMPLETE 439 { 440 structop_base_operation *op 441 = new structop_operation (pstate->pop (), 442 copy_name ($3)); 443 pstate->mark_struct_expression (op); 444 pstate->push (operation_up (op)); 445 } 446 ; 447 448exp : exp '.' COMPLETE 449 { 450 structop_base_operation *op 451 = new structop_operation (pstate->pop (), ""); 452 pstate->mark_struct_expression (op); 453 pstate->push (operation_up (op)); 454 } 455 ; 456 457exp : exp '.' '~' name 458 { 459 pstate->push_new<structop_operation> 460 (pstate->pop (), "~" + copy_name ($4)); 461 } 462 ; 463 464exp : exp '.' '~' name COMPLETE 465 { 466 structop_base_operation *op 467 = new structop_operation (pstate->pop (), 468 "~" + copy_name ($4)); 469 pstate->mark_struct_expression (op); 470 pstate->push (operation_up (op)); 471 } 472 ; 473 474exp : exp '.' qualified_name 475 { /* exp.type::name becomes exp.*(&type::name) */ 476 /* Note: this doesn't work if name is a 477 static member! FIXME */ 478 pstate->wrap<unop_addr_operation> (); 479 pstate->wrap2<structop_member_operation> (); } 480 ; 481 482exp : exp DOT_STAR exp 483 { pstate->wrap2<structop_member_operation> (); } 484 ; 485 486exp : exp '[' exp1 ']' 487 { pstate->wrap2<subscript_operation> (); } 488 ; 489 490exp : exp OBJC_LBRAC exp1 ']' 491 { pstate->wrap2<subscript_operation> (); } 492 ; 493 494/* 495 * The rules below parse ObjC message calls of the form: 496 * '[' target selector {':' argument}* ']' 497 */ 498 499exp : OBJC_LBRAC TYPENAME 500 { 501 CORE_ADDR theclass; 502 503 std::string copy = copy_name ($2.stoken); 504 theclass = lookup_objc_class (pstate->gdbarch (), 505 copy.c_str ()); 506 if (theclass == 0) 507 error (_("%s is not an ObjC Class"), 508 copy.c_str ()); 509 pstate->push_new<long_const_operation> 510 (parse_type (pstate)->builtin_int, 511 (LONGEST) theclass); 512 start_msglist(); 513 } 514 msglist ']' 515 { end_msglist (pstate); } 516 ; 517 518exp : OBJC_LBRAC CLASSNAME 519 { 520 pstate->push_new<long_const_operation> 521 (parse_type (pstate)->builtin_int, 522 (LONGEST) $2.theclass); 523 start_msglist(); 524 } 525 msglist ']' 526 { end_msglist (pstate); } 527 ; 528 529exp : OBJC_LBRAC exp 530 { start_msglist(); } 531 msglist ']' 532 { end_msglist (pstate); } 533 ; 534 535msglist : name 536 { add_msglist(&$1, 0); } 537 | msgarglist 538 ; 539 540msgarglist : msgarg 541 | msgarglist msgarg 542 ; 543 544msgarg : name ':' exp 545 { add_msglist(&$1, 1); } 546 | ':' exp /* Unnamed arg. */ 547 { add_msglist(0, 1); } 548 | ',' exp /* Variable number of args. */ 549 { add_msglist(0, 0); } 550 ; 551 552exp : exp '(' 553 /* This is to save the value of arglist_len 554 being accumulated by an outer function call. */ 555 { pstate->start_arglist (); } 556 arglist ')' %prec ARROW 557 { 558 std::vector<operation_up> args 559 = pstate->pop_vector (pstate->end_arglist ()); 560 pstate->push_new<funcall_operation> 561 (pstate->pop (), std::move (args)); 562 } 563 ; 564 565/* This is here to disambiguate with the production for 566 "func()::static_var" further below, which uses 567 function_method_void. */ 568exp : exp '(' ')' %prec ARROW 569 { 570 pstate->push_new<funcall_operation> 571 (pstate->pop (), std::vector<operation_up> ()); 572 } 573 ; 574 575 576exp : UNKNOWN_CPP_NAME '(' 577 { 578 /* This could potentially be a an argument defined 579 lookup function (Koenig). */ 580 /* This is to save the value of arglist_len 581 being accumulated by an outer function call. */ 582 pstate->start_arglist (); 583 } 584 arglist ')' %prec ARROW 585 { 586 std::vector<operation_up> args 587 = pstate->pop_vector (pstate->end_arglist ()); 588 pstate->push_new<adl_func_operation> 589 (copy_name ($1.stoken), 590 pstate->expression_context_block, 591 std::move (args)); 592 } 593 ; 594 595lcurly : '{' 596 { pstate->start_arglist (); } 597 ; 598 599arglist : 600 ; 601 602arglist : exp 603 { pstate->arglist_len = 1; } 604 ; 605 606arglist : arglist ',' exp %prec ABOVE_COMMA 607 { pstate->arglist_len++; } 608 ; 609 610function_method: exp '(' parameter_typelist ')' const_or_volatile 611 { 612 std::vector<struct type *> *type_list = $3; 613 /* Save the const/volatile qualifiers as 614 recorded by the const_or_volatile 615 production's actions. */ 616 type_instance_flags flags 617 = (cpstate->type_stack 618 .follow_type_instance_flags ()); 619 pstate->push_new<type_instance_operation> 620 (flags, std::move (*type_list), 621 pstate->pop ()); 622 } 623 ; 624 625function_method_void: exp '(' ')' const_or_volatile 626 { 627 type_instance_flags flags 628 = (cpstate->type_stack 629 .follow_type_instance_flags ()); 630 pstate->push_new<type_instance_operation> 631 (flags, std::vector<type *> (), pstate->pop ()); 632 } 633 ; 634 635exp : function_method 636 ; 637 638/* Normally we must interpret "func()" as a function call, instead of 639 a type. The user needs to write func(void) to disambiguate. 640 However, in the "func()::static_var" case, there's no 641 ambiguity. */ 642function_method_void_or_typelist: function_method 643 | function_method_void 644 ; 645 646exp : function_method_void_or_typelist COLONCOLON name 647 { 648 pstate->push_new<func_static_var_operation> 649 (pstate->pop (), copy_name ($3)); 650 } 651 ; 652 653rcurly : '}' 654 { $$ = pstate->end_arglist () - 1; } 655 ; 656exp : lcurly arglist rcurly %prec ARROW 657 { 658 std::vector<operation_up> args 659 = pstate->pop_vector ($3 + 1); 660 pstate->push_new<array_operation> (0, $3, 661 std::move (args)); 662 } 663 ; 664 665exp : lcurly type_exp rcurly exp %prec UNARY 666 { pstate->wrap2<unop_memval_type_operation> (); } 667 ; 668 669exp : '(' type_exp ')' exp %prec UNARY 670 { 671 if (pstate->language ()->la_language 672 == language_opencl) 673 pstate->wrap2<opencl_cast_type_operation> (); 674 else 675 pstate->wrap2<unop_cast_type_operation> (); 676 } 677 ; 678 679exp : '(' exp1 ')' 680 { } 681 ; 682 683/* Binary operators in order of decreasing precedence. */ 684 685exp : exp '@' exp 686 { pstate->wrap2<repeat_operation> (); } 687 ; 688 689exp : exp '*' exp 690 { pstate->wrap2<mul_operation> (); } 691 ; 692 693exp : exp '/' exp 694 { pstate->wrap2<div_operation> (); } 695 ; 696 697exp : exp '%' exp 698 { pstate->wrap2<rem_operation> (); } 699 ; 700 701exp : exp '+' exp 702 { pstate->wrap2<add_operation> (); } 703 ; 704 705exp : exp '-' exp 706 { pstate->wrap2<sub_operation> (); } 707 ; 708 709exp : exp LSH exp 710 { pstate->wrap2<lsh_operation> (); } 711 ; 712 713exp : exp RSH exp 714 { pstate->wrap2<rsh_operation> (); } 715 ; 716 717exp : exp EQUAL exp 718 { 719 if (pstate->language ()->la_language 720 == language_opencl) 721 pstate->wrap2<opencl_equal_operation> (); 722 else 723 pstate->wrap2<equal_operation> (); 724 } 725 ; 726 727exp : exp NOTEQUAL exp 728 { 729 if (pstate->language ()->la_language 730 == language_opencl) 731 pstate->wrap2<opencl_notequal_operation> (); 732 else 733 pstate->wrap2<notequal_operation> (); 734 } 735 ; 736 737exp : exp LEQ exp 738 { 739 if (pstate->language ()->la_language 740 == language_opencl) 741 pstate->wrap2<opencl_leq_operation> (); 742 else 743 pstate->wrap2<leq_operation> (); 744 } 745 ; 746 747exp : exp GEQ exp 748 { 749 if (pstate->language ()->la_language 750 == language_opencl) 751 pstate->wrap2<opencl_geq_operation> (); 752 else 753 pstate->wrap2<geq_operation> (); 754 } 755 ; 756 757exp : exp '<' exp 758 { 759 if (pstate->language ()->la_language 760 == language_opencl) 761 pstate->wrap2<opencl_less_operation> (); 762 else 763 pstate->wrap2<less_operation> (); 764 } 765 ; 766 767exp : exp '>' exp 768 { 769 if (pstate->language ()->la_language 770 == language_opencl) 771 pstate->wrap2<opencl_gtr_operation> (); 772 else 773 pstate->wrap2<gtr_operation> (); 774 } 775 ; 776 777exp : exp '&' exp 778 { pstate->wrap2<bitwise_and_operation> (); } 779 ; 780 781exp : exp '^' exp 782 { pstate->wrap2<bitwise_xor_operation> (); } 783 ; 784 785exp : exp '|' exp 786 { pstate->wrap2<bitwise_ior_operation> (); } 787 ; 788 789exp : exp ANDAND exp 790 { 791 if (pstate->language ()->la_language 792 == language_opencl) 793 { 794 operation_up rhs = pstate->pop (); 795 operation_up lhs = pstate->pop (); 796 pstate->push_new<opencl_logical_binop_operation> 797 (BINOP_LOGICAL_AND, std::move (lhs), 798 std::move (rhs)); 799 } 800 else 801 pstate->wrap2<logical_and_operation> (); 802 } 803 ; 804 805exp : exp OROR exp 806 { 807 if (pstate->language ()->la_language 808 == language_opencl) 809 { 810 operation_up rhs = pstate->pop (); 811 operation_up lhs = pstate->pop (); 812 pstate->push_new<opencl_logical_binop_operation> 813 (BINOP_LOGICAL_OR, std::move (lhs), 814 std::move (rhs)); 815 } 816 else 817 pstate->wrap2<logical_or_operation> (); 818 } 819 ; 820 821exp : exp '?' exp ':' exp %prec '?' 822 { 823 operation_up last = pstate->pop (); 824 operation_up mid = pstate->pop (); 825 operation_up first = pstate->pop (); 826 if (pstate->language ()->la_language 827 == language_opencl) 828 pstate->push_new<opencl_ternop_cond_operation> 829 (std::move (first), std::move (mid), 830 std::move (last)); 831 else 832 pstate->push_new<ternop_cond_operation> 833 (std::move (first), std::move (mid), 834 std::move (last)); 835 } 836 ; 837 838exp : exp '=' exp 839 { 840 if (pstate->language ()->la_language 841 == language_opencl) 842 pstate->wrap2<opencl_assign_operation> (); 843 else 844 pstate->wrap2<assign_operation> (); 845 } 846 ; 847 848exp : exp ASSIGN_MODIFY exp 849 { 850 operation_up rhs = pstate->pop (); 851 operation_up lhs = pstate->pop (); 852 pstate->push_new<assign_modify_operation> 853 ($2, std::move (lhs), std::move (rhs)); 854 } 855 ; 856 857exp : INT 858 { 859 pstate->push_new<long_const_operation> 860 ($1.type, $1.val); 861 } 862 ; 863 864exp : COMPLEX_INT 865 { 866 operation_up real 867 = (make_operation<long_const_operation> 868 ($1.type->target_type (), 0)); 869 operation_up imag 870 = (make_operation<long_const_operation> 871 ($1.type->target_type (), $1.val)); 872 pstate->push_new<complex_operation> 873 (std::move (real), std::move (imag), $1.type); 874 } 875 ; 876 877exp : CHAR 878 { 879 struct stoken_vector vec; 880 vec.len = 1; 881 vec.tokens = &$1; 882 pstate->push_c_string ($1.type, &vec); 883 } 884 ; 885 886exp : NAME_OR_INT 887 { YYSTYPE val; 888 parse_number (pstate, $1.stoken.ptr, 889 $1.stoken.length, 0, &val); 890 pstate->push_new<long_const_operation> 891 (val.typed_val_int.type, 892 val.typed_val_int.val); 893 } 894 ; 895 896 897exp : FLOAT 898 { 899 float_data data; 900 std::copy (std::begin ($1.val), std::end ($1.val), 901 std::begin (data)); 902 pstate->push_new<float_const_operation> ($1.type, data); 903 } 904 ; 905 906exp : COMPLEX_FLOAT 907 { 908 struct type *underlying = $1.type->target_type (); 909 910 float_data val; 911 target_float_from_host_double (val.data (), 912 underlying, 0); 913 operation_up real 914 = (make_operation<float_const_operation> 915 (underlying, val)); 916 917 std::copy (std::begin ($1.val), std::end ($1.val), 918 std::begin (val)); 919 operation_up imag 920 = (make_operation<float_const_operation> 921 (underlying, val)); 922 923 pstate->push_new<complex_operation> 924 (std::move (real), std::move (imag), 925 $1.type); 926 } 927 ; 928 929exp : variable 930 ; 931 932exp : DOLLAR_VARIABLE 933 { 934 pstate->push_dollar ($1); 935 } 936 ; 937 938exp : SELECTOR '(' name ')' 939 { 940 pstate->push_new<objc_selector_operation> 941 (copy_name ($3)); 942 } 943 ; 944 945exp : SIZEOF '(' type ')' %prec UNARY 946 { struct type *type = $3; 947 struct type *int_type 948 = lookup_signed_typename (pstate->language (), 949 "int"); 950 type = check_typedef (type); 951 952 /* $5.3.3/2 of the C++ Standard (n3290 draft) 953 says of sizeof: "When applied to a reference 954 or a reference type, the result is the size of 955 the referenced type." */ 956 if (TYPE_IS_REFERENCE (type)) 957 type = check_typedef (type->target_type ()); 958 959 pstate->push_new<long_const_operation> 960 (int_type, type->length ()); 961 } 962 ; 963 964exp : REINTERPRET_CAST '<' type_exp '>' '(' exp ')' %prec UNARY 965 { pstate->wrap2<reinterpret_cast_operation> (); } 966 ; 967 968exp : STATIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY 969 { pstate->wrap2<unop_cast_type_operation> (); } 970 ; 971 972exp : DYNAMIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY 973 { pstate->wrap2<dynamic_cast_operation> (); } 974 ; 975 976exp : CONST_CAST '<' type_exp '>' '(' exp ')' %prec UNARY 977 { /* We could do more error checking here, but 978 it doesn't seem worthwhile. */ 979 pstate->wrap2<unop_cast_type_operation> (); } 980 ; 981 982string_exp: 983 STRING 984 { 985 /* We copy the string here, and not in the 986 lexer, to guarantee that we do not leak a 987 string. Note that we follow the 988 NUL-termination convention of the 989 lexer. */ 990 struct typed_stoken *vec = XNEW (struct typed_stoken); 991 $$.len = 1; 992 $$.tokens = vec; 993 994 vec->type = $1.type; 995 vec->length = $1.length; 996 vec->ptr = (char *) malloc ($1.length + 1); 997 memcpy (vec->ptr, $1.ptr, $1.length + 1); 998 } 999 1000 | string_exp STRING 1001 { 1002 /* Note that we NUL-terminate here, but just 1003 for convenience. */ 1004 char *p; 1005 ++$$.len; 1006 $$.tokens = XRESIZEVEC (struct typed_stoken, 1007 $$.tokens, $$.len); 1008 1009 p = (char *) malloc ($2.length + 1); 1010 memcpy (p, $2.ptr, $2.length + 1); 1011 1012 $$.tokens[$$.len - 1].type = $2.type; 1013 $$.tokens[$$.len - 1].length = $2.length; 1014 $$.tokens[$$.len - 1].ptr = p; 1015 } 1016 ; 1017 1018exp : string_exp 1019 { 1020 int i; 1021 c_string_type type = C_STRING; 1022 1023 for (i = 0; i < $1.len; ++i) 1024 { 1025 switch ($1.tokens[i].type) 1026 { 1027 case C_STRING: 1028 break; 1029 case C_WIDE_STRING: 1030 case C_STRING_16: 1031 case C_STRING_32: 1032 if (type != C_STRING 1033 && type != $1.tokens[i].type) 1034 error (_("Undefined string concatenation.")); 1035 type = (enum c_string_type_values) $1.tokens[i].type; 1036 break; 1037 default: 1038 /* internal error */ 1039 internal_error ("unrecognized type in string concatenation"); 1040 } 1041 } 1042 1043 pstate->push_c_string (type, &$1); 1044 for (i = 0; i < $1.len; ++i) 1045 free ($1.tokens[i].ptr); 1046 free ($1.tokens); 1047 } 1048 ; 1049 1050exp : NSSTRING /* ObjC NextStep NSString constant 1051 * of the form '@' '"' string '"'. 1052 */ 1053 { 1054 pstate->push_new<objc_nsstring_operation> 1055 (copy_name ($1)); 1056 } 1057 ; 1058 1059/* C++. */ 1060exp : TRUEKEYWORD 1061 { pstate->push_new<long_const_operation> 1062 (parse_type (pstate)->builtin_bool, 1); 1063 } 1064 ; 1065 1066exp : FALSEKEYWORD 1067 { pstate->push_new<long_const_operation> 1068 (parse_type (pstate)->builtin_bool, 0); 1069 } 1070 ; 1071 1072/* end of C++. */ 1073 1074block : BLOCKNAME 1075 { 1076 if ($1.sym.symbol) 1077 $$ = $1.sym.symbol->value_block (); 1078 else 1079 error (_("No file or function \"%s\"."), 1080 copy_name ($1.stoken).c_str ()); 1081 } 1082 | FILENAME 1083 { 1084 $$ = $1; 1085 } 1086 ; 1087 1088block : block COLONCOLON name 1089 { 1090 std::string copy = copy_name ($3); 1091 struct symbol *tem 1092 = lookup_symbol (copy.c_str (), $1, 1093 VAR_DOMAIN, NULL).symbol; 1094 1095 if (!tem || tem->aclass () != LOC_BLOCK) 1096 error (_("No function \"%s\" in specified context."), 1097 copy.c_str ()); 1098 $$ = tem->value_block (); } 1099 ; 1100 1101variable: name_not_typename ENTRY 1102 { struct symbol *sym = $1.sym.symbol; 1103 1104 if (sym == NULL || !sym->is_argument () 1105 || !symbol_read_needs_frame (sym)) 1106 error (_("@entry can be used only for function " 1107 "parameters, not for \"%s\""), 1108 copy_name ($1.stoken).c_str ()); 1109 1110 pstate->push_new<var_entry_value_operation> (sym); 1111 } 1112 ; 1113 1114variable: block COLONCOLON name 1115 { 1116 std::string copy = copy_name ($3); 1117 struct block_symbol sym 1118 = lookup_symbol (copy.c_str (), $1, 1119 VAR_DOMAIN, NULL); 1120 1121 if (sym.symbol == 0) 1122 error (_("No symbol \"%s\" in specified context."), 1123 copy.c_str ()); 1124 if (symbol_read_needs_frame (sym.symbol)) 1125 pstate->block_tracker->update (sym); 1126 1127 pstate->push_new<var_value_operation> (sym); 1128 } 1129 ; 1130 1131qualified_name: TYPENAME COLONCOLON name 1132 { 1133 struct type *type = $1.type; 1134 type = check_typedef (type); 1135 if (!type_aggregate_p (type)) 1136 error (_("`%s' is not defined as an aggregate type."), 1137 TYPE_SAFE_NAME (type)); 1138 1139 pstate->push_new<scope_operation> (type, 1140 copy_name ($3)); 1141 } 1142 | TYPENAME COLONCOLON '~' name 1143 { 1144 struct type *type = $1.type; 1145 1146 type = check_typedef (type); 1147 if (!type_aggregate_p (type)) 1148 error (_("`%s' is not defined as an aggregate type."), 1149 TYPE_SAFE_NAME (type)); 1150 std::string name = "~" + std::string ($4.ptr, 1151 $4.length); 1152 1153 /* Check for valid destructor name. */ 1154 destructor_name_p (name.c_str (), $1.type); 1155 pstate->push_new<scope_operation> (type, 1156 std::move (name)); 1157 } 1158 | TYPENAME COLONCOLON name COLONCOLON name 1159 { 1160 std::string copy = copy_name ($3); 1161 error (_("No type \"%s\" within class " 1162 "or namespace \"%s\"."), 1163 copy.c_str (), TYPE_SAFE_NAME ($1.type)); 1164 } 1165 ; 1166 1167variable: qualified_name 1168 | COLONCOLON name_not_typename 1169 { 1170 std::string name = copy_name ($2.stoken); 1171 struct block_symbol sym 1172 = lookup_symbol (name.c_str (), 1173 (const struct block *) NULL, 1174 VAR_DOMAIN, NULL); 1175 pstate->push_symbol (name.c_str (), sym); 1176 } 1177 ; 1178 1179variable: name_not_typename 1180 { struct block_symbol sym = $1.sym; 1181 1182 if (sym.symbol) 1183 { 1184 if (symbol_read_needs_frame (sym.symbol)) 1185 pstate->block_tracker->update (sym); 1186 1187 /* If we found a function, see if it's 1188 an ifunc resolver that has the same 1189 address as the ifunc symbol itself. 1190 If so, prefer the ifunc symbol. */ 1191 1192 bound_minimal_symbol resolver 1193 = find_gnu_ifunc (sym.symbol); 1194 if (resolver.minsym != NULL) 1195 pstate->push_new<var_msym_value_operation> 1196 (resolver); 1197 else 1198 pstate->push_new<var_value_operation> (sym); 1199 } 1200 else if ($1.is_a_field_of_this) 1201 { 1202 /* C++: it hangs off of `this'. Must 1203 not inadvertently convert from a method call 1204 to data ref. */ 1205 pstate->block_tracker->update (sym); 1206 operation_up thisop 1207 = make_operation<op_this_operation> (); 1208 pstate->push_new<structop_ptr_operation> 1209 (std::move (thisop), copy_name ($1.stoken)); 1210 } 1211 else 1212 { 1213 std::string arg = copy_name ($1.stoken); 1214 1215 bound_minimal_symbol msymbol 1216 = lookup_bound_minimal_symbol (arg.c_str ()); 1217 if (msymbol.minsym == NULL) 1218 { 1219 if (!have_full_symbols () && !have_partial_symbols ()) 1220 error (_("No symbol table is loaded. Use the \"file\" command.")); 1221 else 1222 error (_("No symbol \"%s\" in current context."), 1223 arg.c_str ()); 1224 } 1225 1226 /* This minsym might be an alias for 1227 another function. See if we can find 1228 the debug symbol for the target, and 1229 if so, use it instead, since it has 1230 return type / prototype info. This 1231 is important for example for "p 1232 *__errno_location()". */ 1233 symbol *alias_target 1234 = ((msymbol.minsym->type () != mst_text_gnu_ifunc 1235 && msymbol.minsym->type () != mst_data_gnu_ifunc) 1236 ? find_function_alias_target (msymbol) 1237 : NULL); 1238 if (alias_target != NULL) 1239 { 1240 block_symbol bsym { alias_target, 1241 alias_target->value_block () }; 1242 pstate->push_new<var_value_operation> (bsym); 1243 } 1244 else 1245 pstate->push_new<var_msym_value_operation> 1246 (msymbol); 1247 } 1248 } 1249 ; 1250 1251const_or_volatile: const_or_volatile_noopt 1252 | 1253 ; 1254 1255single_qualifier: 1256 CONST_KEYWORD 1257 { cpstate->type_stack.insert (tp_const); } 1258 | VOLATILE_KEYWORD 1259 { cpstate->type_stack.insert (tp_volatile); } 1260 | ATOMIC 1261 { cpstate->type_stack.insert (tp_atomic); } 1262 | RESTRICT 1263 { cpstate->type_stack.insert (tp_restrict); } 1264 | '@' NAME 1265 { 1266 cpstate->type_stack.insert (pstate, 1267 copy_name ($2.stoken).c_str ()); 1268 } 1269 | '@' UNKNOWN_CPP_NAME 1270 { 1271 cpstate->type_stack.insert (pstate, 1272 copy_name ($2.stoken).c_str ()); 1273 } 1274 ; 1275 1276qualifier_seq_noopt: 1277 single_qualifier 1278 | qualifier_seq_noopt single_qualifier 1279 ; 1280 1281qualifier_seq: 1282 qualifier_seq_noopt 1283 | 1284 ; 1285 1286ptr_operator: 1287 ptr_operator '*' 1288 { cpstate->type_stack.insert (tp_pointer); } 1289 qualifier_seq 1290 | '*' 1291 { cpstate->type_stack.insert (tp_pointer); } 1292 qualifier_seq 1293 | '&' 1294 { cpstate->type_stack.insert (tp_reference); } 1295 | '&' ptr_operator 1296 { cpstate->type_stack.insert (tp_reference); } 1297 | ANDAND 1298 { cpstate->type_stack.insert (tp_rvalue_reference); } 1299 | ANDAND ptr_operator 1300 { cpstate->type_stack.insert (tp_rvalue_reference); } 1301 ; 1302 1303ptr_operator_ts: ptr_operator 1304 { 1305 $$ = cpstate->type_stack.create (); 1306 cpstate->type_stacks.emplace_back ($$); 1307 } 1308 ; 1309 1310abs_decl: ptr_operator_ts direct_abs_decl 1311 { $$ = $2->append ($1); } 1312 | ptr_operator_ts 1313 | direct_abs_decl 1314 ; 1315 1316direct_abs_decl: '(' abs_decl ')' 1317 { $$ = $2; } 1318 | direct_abs_decl array_mod 1319 { 1320 cpstate->type_stack.push ($1); 1321 cpstate->type_stack.push ($2); 1322 cpstate->type_stack.push (tp_array); 1323 $$ = cpstate->type_stack.create (); 1324 cpstate->type_stacks.emplace_back ($$); 1325 } 1326 | array_mod 1327 { 1328 cpstate->type_stack.push ($1); 1329 cpstate->type_stack.push (tp_array); 1330 $$ = cpstate->type_stack.create (); 1331 cpstate->type_stacks.emplace_back ($$); 1332 } 1333 1334 | direct_abs_decl func_mod 1335 { 1336 cpstate->type_stack.push ($1); 1337 cpstate->type_stack.push ($2); 1338 $$ = cpstate->type_stack.create (); 1339 cpstate->type_stacks.emplace_back ($$); 1340 } 1341 | func_mod 1342 { 1343 cpstate->type_stack.push ($1); 1344 $$ = cpstate->type_stack.create (); 1345 cpstate->type_stacks.emplace_back ($$); 1346 } 1347 ; 1348 1349array_mod: '[' ']' 1350 { $$ = -1; } 1351 | OBJC_LBRAC ']' 1352 { $$ = -1; } 1353 | '[' INT ']' 1354 { $$ = $2.val; } 1355 | OBJC_LBRAC INT ']' 1356 { $$ = $2.val; } 1357 ; 1358 1359func_mod: '(' ')' 1360 { 1361 $$ = new std::vector<struct type *>; 1362 cpstate->type_lists.emplace_back ($$); 1363 } 1364 | '(' parameter_typelist ')' 1365 { $$ = $2; } 1366 ; 1367 1368/* We used to try to recognize pointer to member types here, but 1369 that didn't work (shift/reduce conflicts meant that these rules never 1370 got executed). The problem is that 1371 int (foo::bar::baz::bizzle) 1372 is a function type but 1373 int (foo::bar::baz::bizzle::*) 1374 is a pointer to member type. Stroustrup loses again! */ 1375 1376type : ptype 1377 ; 1378 1379/* A helper production that recognizes scalar types that can validly 1380 be used with _Complex. */ 1381 1382scalar_type: 1383 INT_KEYWORD 1384 { $$ = lookup_signed_typename (pstate->language (), 1385 "int"); } 1386 | LONG 1387 { $$ = lookup_signed_typename (pstate->language (), 1388 "long"); } 1389 | SHORT 1390 { $$ = lookup_signed_typename (pstate->language (), 1391 "short"); } 1392 | LONG INT_KEYWORD 1393 { $$ = lookup_signed_typename (pstate->language (), 1394 "long"); } 1395 | LONG SIGNED_KEYWORD INT_KEYWORD 1396 { $$ = lookup_signed_typename (pstate->language (), 1397 "long"); } 1398 | LONG SIGNED_KEYWORD 1399 { $$ = lookup_signed_typename (pstate->language (), 1400 "long"); } 1401 | SIGNED_KEYWORD LONG INT_KEYWORD 1402 { $$ = lookup_signed_typename (pstate->language (), 1403 "long"); } 1404 | UNSIGNED LONG INT_KEYWORD 1405 { $$ = lookup_unsigned_typename (pstate->language (), 1406 "long"); } 1407 | LONG UNSIGNED INT_KEYWORD 1408 { $$ = lookup_unsigned_typename (pstate->language (), 1409 "long"); } 1410 | LONG UNSIGNED 1411 { $$ = lookup_unsigned_typename (pstate->language (), 1412 "long"); } 1413 | LONG LONG 1414 { $$ = lookup_signed_typename (pstate->language (), 1415 "long long"); } 1416 | LONG LONG INT_KEYWORD 1417 { $$ = lookup_signed_typename (pstate->language (), 1418 "long long"); } 1419 | LONG LONG SIGNED_KEYWORD INT_KEYWORD 1420 { $$ = lookup_signed_typename (pstate->language (), 1421 "long long"); } 1422 | LONG LONG SIGNED_KEYWORD 1423 { $$ = lookup_signed_typename (pstate->language (), 1424 "long long"); } 1425 | SIGNED_KEYWORD LONG LONG 1426 { $$ = lookup_signed_typename (pstate->language (), 1427 "long long"); } 1428 | SIGNED_KEYWORD LONG LONG INT_KEYWORD 1429 { $$ = lookup_signed_typename (pstate->language (), 1430 "long long"); } 1431 | UNSIGNED LONG LONG 1432 { $$ = lookup_unsigned_typename (pstate->language (), 1433 "long long"); } 1434 | UNSIGNED LONG LONG INT_KEYWORD 1435 { $$ = lookup_unsigned_typename (pstate->language (), 1436 "long long"); } 1437 | LONG LONG UNSIGNED 1438 { $$ = lookup_unsigned_typename (pstate->language (), 1439 "long long"); } 1440 | LONG LONG UNSIGNED INT_KEYWORD 1441 { $$ = lookup_unsigned_typename (pstate->language (), 1442 "long long"); } 1443 | SHORT INT_KEYWORD 1444 { $$ = lookup_signed_typename (pstate->language (), 1445 "short"); } 1446 | SHORT SIGNED_KEYWORD INT_KEYWORD 1447 { $$ = lookup_signed_typename (pstate->language (), 1448 "short"); } 1449 | SHORT SIGNED_KEYWORD 1450 { $$ = lookup_signed_typename (pstate->language (), 1451 "short"); } 1452 | UNSIGNED SHORT INT_KEYWORD 1453 { $$ = lookup_unsigned_typename (pstate->language (), 1454 "short"); } 1455 | SHORT UNSIGNED 1456 { $$ = lookup_unsigned_typename (pstate->language (), 1457 "short"); } 1458 | SHORT UNSIGNED INT_KEYWORD 1459 { $$ = lookup_unsigned_typename (pstate->language (), 1460 "short"); } 1461 | DOUBLE_KEYWORD 1462 { $$ = lookup_typename (pstate->language (), 1463 "double", 1464 NULL, 1465 0); } 1466 | FLOAT_KEYWORD 1467 { $$ = lookup_typename (pstate->language (), 1468 "float", 1469 NULL, 1470 0); } 1471 | LONG DOUBLE_KEYWORD 1472 { $$ = lookup_typename (pstate->language (), 1473 "long double", 1474 NULL, 1475 0); } 1476 | UNSIGNED type_name 1477 { $$ = lookup_unsigned_typename (pstate->language (), 1478 $2.type->name ()); } 1479 | UNSIGNED 1480 { $$ = lookup_unsigned_typename (pstate->language (), 1481 "int"); } 1482 | SIGNED_KEYWORD type_name 1483 { $$ = lookup_signed_typename (pstate->language (), 1484 $2.type->name ()); } 1485 | SIGNED_KEYWORD 1486 { $$ = lookup_signed_typename (pstate->language (), 1487 "int"); } 1488 ; 1489 1490/* Implements (approximately): (type-qualifier)* type-specifier. 1491 1492 When type-specifier is only ever a single word, like 'float' then these 1493 arrive as pre-built TYPENAME tokens thanks to the classify_name 1494 function. However, when a type-specifier can contain multiple words, 1495 for example 'double' can appear as just 'double' or 'long double', and 1496 similarly 'long' can appear as just 'long' or in 'long double', then 1497 these type-specifiers are parsed into their own tokens in the function 1498 lex_one_token and the ident_tokens array. These separate tokens are all 1499 recognised here. */ 1500typebase 1501 : TYPENAME 1502 { $$ = $1.type; } 1503 | scalar_type 1504 { $$ = $1; } 1505 | COMPLEX scalar_type 1506 { 1507 $$ = init_complex_type (nullptr, $2); 1508 } 1509 | STRUCT name 1510 { $$ 1511 = lookup_struct (copy_name ($2).c_str (), 1512 pstate->expression_context_block); 1513 } 1514 | STRUCT COMPLETE 1515 { 1516 pstate->mark_completion_tag (TYPE_CODE_STRUCT, 1517 "", 0); 1518 $$ = NULL; 1519 } 1520 | STRUCT name COMPLETE 1521 { 1522 pstate->mark_completion_tag (TYPE_CODE_STRUCT, 1523 $2.ptr, $2.length); 1524 $$ = NULL; 1525 } 1526 | CLASS name 1527 { $$ = lookup_struct 1528 (copy_name ($2).c_str (), 1529 pstate->expression_context_block); 1530 } 1531 | CLASS COMPLETE 1532 { 1533 pstate->mark_completion_tag (TYPE_CODE_STRUCT, 1534 "", 0); 1535 $$ = NULL; 1536 } 1537 | CLASS name COMPLETE 1538 { 1539 pstate->mark_completion_tag (TYPE_CODE_STRUCT, 1540 $2.ptr, $2.length); 1541 $$ = NULL; 1542 } 1543 | UNION name 1544 { $$ 1545 = lookup_union (copy_name ($2).c_str (), 1546 pstate->expression_context_block); 1547 } 1548 | UNION COMPLETE 1549 { 1550 pstate->mark_completion_tag (TYPE_CODE_UNION, 1551 "", 0); 1552 $$ = NULL; 1553 } 1554 | UNION name COMPLETE 1555 { 1556 pstate->mark_completion_tag (TYPE_CODE_UNION, 1557 $2.ptr, $2.length); 1558 $$ = NULL; 1559 } 1560 | ENUM name 1561 { $$ = lookup_enum (copy_name ($2).c_str (), 1562 pstate->expression_context_block); 1563 } 1564 | ENUM COMPLETE 1565 { 1566 pstate->mark_completion_tag (TYPE_CODE_ENUM, "", 0); 1567 $$ = NULL; 1568 } 1569 | ENUM name COMPLETE 1570 { 1571 pstate->mark_completion_tag (TYPE_CODE_ENUM, $2.ptr, 1572 $2.length); 1573 $$ = NULL; 1574 } 1575 /* It appears that this rule for templates is never 1576 reduced; template recognition happens by lookahead 1577 in the token processing code in yylex. */ 1578 | TEMPLATE name '<' type '>' 1579 { $$ = lookup_template_type 1580 (copy_name($2).c_str (), $4, 1581 pstate->expression_context_block); 1582 } 1583 | qualifier_seq_noopt typebase 1584 { $$ = cpstate->type_stack.follow_types ($2); } 1585 | typebase qualifier_seq_noopt 1586 { $$ = cpstate->type_stack.follow_types ($1); } 1587 ; 1588 1589type_name: TYPENAME 1590 | INT_KEYWORD 1591 { 1592 $$.stoken.ptr = "int"; 1593 $$.stoken.length = 3; 1594 $$.type = lookup_signed_typename (pstate->language (), 1595 "int"); 1596 } 1597 | LONG 1598 { 1599 $$.stoken.ptr = "long"; 1600 $$.stoken.length = 4; 1601 $$.type = lookup_signed_typename (pstate->language (), 1602 "long"); 1603 } 1604 | SHORT 1605 { 1606 $$.stoken.ptr = "short"; 1607 $$.stoken.length = 5; 1608 $$.type = lookup_signed_typename (pstate->language (), 1609 "short"); 1610 } 1611 ; 1612 1613parameter_typelist: 1614 nonempty_typelist 1615 { check_parameter_typelist ($1); } 1616 | nonempty_typelist ',' DOTDOTDOT 1617 { 1618 $1->push_back (NULL); 1619 check_parameter_typelist ($1); 1620 $$ = $1; 1621 } 1622 ; 1623 1624nonempty_typelist 1625 : type 1626 { 1627 std::vector<struct type *> *typelist 1628 = new std::vector<struct type *>; 1629 cpstate->type_lists.emplace_back (typelist); 1630 1631 typelist->push_back ($1); 1632 $$ = typelist; 1633 } 1634 | nonempty_typelist ',' type 1635 { 1636 $1->push_back ($3); 1637 $$ = $1; 1638 } 1639 ; 1640 1641ptype : typebase 1642 | ptype abs_decl 1643 { 1644 cpstate->type_stack.push ($2); 1645 $$ = cpstate->type_stack.follow_types ($1); 1646 } 1647 ; 1648 1649conversion_type_id: typebase conversion_declarator 1650 { $$ = cpstate->type_stack.follow_types ($1); } 1651 ; 1652 1653conversion_declarator: /* Nothing. */ 1654 | ptr_operator conversion_declarator 1655 ; 1656 1657const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD 1658 | VOLATILE_KEYWORD CONST_KEYWORD 1659 ; 1660 1661const_or_volatile_noopt: const_and_volatile 1662 { cpstate->type_stack.insert (tp_const); 1663 cpstate->type_stack.insert (tp_volatile); 1664 } 1665 | CONST_KEYWORD 1666 { cpstate->type_stack.insert (tp_const); } 1667 | VOLATILE_KEYWORD 1668 { cpstate->type_stack.insert (tp_volatile); } 1669 ; 1670 1671oper: OPERATOR NEW 1672 { $$ = operator_stoken (" new"); } 1673 | OPERATOR DELETE 1674 { $$ = operator_stoken (" delete"); } 1675 | OPERATOR NEW '[' ']' 1676 { $$ = operator_stoken (" new[]"); } 1677 | OPERATOR DELETE '[' ']' 1678 { $$ = operator_stoken (" delete[]"); } 1679 | OPERATOR NEW OBJC_LBRAC ']' 1680 { $$ = operator_stoken (" new[]"); } 1681 | OPERATOR DELETE OBJC_LBRAC ']' 1682 { $$ = operator_stoken (" delete[]"); } 1683 | OPERATOR '+' 1684 { $$ = operator_stoken ("+"); } 1685 | OPERATOR '-' 1686 { $$ = operator_stoken ("-"); } 1687 | OPERATOR '*' 1688 { $$ = operator_stoken ("*"); } 1689 | OPERATOR '/' 1690 { $$ = operator_stoken ("/"); } 1691 | OPERATOR '%' 1692 { $$ = operator_stoken ("%"); } 1693 | OPERATOR '^' 1694 { $$ = operator_stoken ("^"); } 1695 | OPERATOR '&' 1696 { $$ = operator_stoken ("&"); } 1697 | OPERATOR '|' 1698 { $$ = operator_stoken ("|"); } 1699 | OPERATOR '~' 1700 { $$ = operator_stoken ("~"); } 1701 | OPERATOR '!' 1702 { $$ = operator_stoken ("!"); } 1703 | OPERATOR '=' 1704 { $$ = operator_stoken ("="); } 1705 | OPERATOR '<' 1706 { $$ = operator_stoken ("<"); } 1707 | OPERATOR '>' 1708 { $$ = operator_stoken (">"); } 1709 | OPERATOR ASSIGN_MODIFY 1710 { const char *op = " unknown"; 1711 switch ($2) 1712 { 1713 case BINOP_RSH: 1714 op = ">>="; 1715 break; 1716 case BINOP_LSH: 1717 op = "<<="; 1718 break; 1719 case BINOP_ADD: 1720 op = "+="; 1721 break; 1722 case BINOP_SUB: 1723 op = "-="; 1724 break; 1725 case BINOP_MUL: 1726 op = "*="; 1727 break; 1728 case BINOP_DIV: 1729 op = "/="; 1730 break; 1731 case BINOP_REM: 1732 op = "%="; 1733 break; 1734 case BINOP_BITWISE_IOR: 1735 op = "|="; 1736 break; 1737 case BINOP_BITWISE_AND: 1738 op = "&="; 1739 break; 1740 case BINOP_BITWISE_XOR: 1741 op = "^="; 1742 break; 1743 default: 1744 break; 1745 } 1746 1747 $$ = operator_stoken (op); 1748 } 1749 | OPERATOR LSH 1750 { $$ = operator_stoken ("<<"); } 1751 | OPERATOR RSH 1752 { $$ = operator_stoken (">>"); } 1753 | OPERATOR EQUAL 1754 { $$ = operator_stoken ("=="); } 1755 | OPERATOR NOTEQUAL 1756 { $$ = operator_stoken ("!="); } 1757 | OPERATOR LEQ 1758 { $$ = operator_stoken ("<="); } 1759 | OPERATOR GEQ 1760 { $$ = operator_stoken (">="); } 1761 | OPERATOR ANDAND 1762 { $$ = operator_stoken ("&&"); } 1763 | OPERATOR OROR 1764 { $$ = operator_stoken ("||"); } 1765 | OPERATOR INCREMENT 1766 { $$ = operator_stoken ("++"); } 1767 | OPERATOR DECREMENT 1768 { $$ = operator_stoken ("--"); } 1769 | OPERATOR ',' 1770 { $$ = operator_stoken (","); } 1771 | OPERATOR ARROW_STAR 1772 { $$ = operator_stoken ("->*"); } 1773 | OPERATOR ARROW 1774 { $$ = operator_stoken ("->"); } 1775 | OPERATOR '(' ')' 1776 { $$ = operator_stoken ("()"); } 1777 | OPERATOR '[' ']' 1778 { $$ = operator_stoken ("[]"); } 1779 | OPERATOR OBJC_LBRAC ']' 1780 { $$ = operator_stoken ("[]"); } 1781 | OPERATOR conversion_type_id 1782 { 1783 string_file buf; 1784 c_print_type ($2, NULL, &buf, -1, 0, 1785 pstate->language ()->la_language, 1786 &type_print_raw_options); 1787 std::string name = buf.release (); 1788 1789 /* This also needs canonicalization. */ 1790 gdb::unique_xmalloc_ptr<char> canon 1791 = cp_canonicalize_string (name.c_str ()); 1792 if (canon != nullptr) 1793 name = canon.get (); 1794 $$ = operator_stoken ((" " + name).c_str ()); 1795 } 1796 ; 1797 1798/* This rule exists in order to allow some tokens that would not normally 1799 match the 'name' rule to appear as fields within a struct. The example 1800 that initially motivated this was the RISC-V target which models the 1801 floating point registers as a union with fields called 'float' and 1802 'double'. */ 1803field_name 1804 : name 1805 | DOUBLE_KEYWORD { $$ = typename_stoken ("double"); } 1806 | FLOAT_KEYWORD { $$ = typename_stoken ("float"); } 1807 | INT_KEYWORD { $$ = typename_stoken ("int"); } 1808 | LONG { $$ = typename_stoken ("long"); } 1809 | SHORT { $$ = typename_stoken ("short"); } 1810 | SIGNED_KEYWORD { $$ = typename_stoken ("signed"); } 1811 | UNSIGNED { $$ = typename_stoken ("unsigned"); } 1812 ; 1813 1814name : NAME { $$ = $1.stoken; } 1815 | BLOCKNAME { $$ = $1.stoken; } 1816 | TYPENAME { $$ = $1.stoken; } 1817 | NAME_OR_INT { $$ = $1.stoken; } 1818 | UNKNOWN_CPP_NAME { $$ = $1.stoken; } 1819 | oper { $$ = $1; } 1820 ; 1821 1822name_not_typename : NAME 1823 | BLOCKNAME 1824/* These would be useful if name_not_typename was useful, but it is just 1825 a fake for "variable", so these cause reduce/reduce conflicts because 1826 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable, 1827 =exp) or just an exp. If name_not_typename was ever used in an lvalue 1828 context where only a name could occur, this might be useful. 1829 | NAME_OR_INT 1830 */ 1831 | oper 1832 { 1833 struct field_of_this_result is_a_field_of_this; 1834 1835 $$.stoken = $1; 1836 $$.sym 1837 = lookup_symbol ($1.ptr, 1838 pstate->expression_context_block, 1839 VAR_DOMAIN, 1840 &is_a_field_of_this); 1841 $$.is_a_field_of_this 1842 = is_a_field_of_this.type != NULL; 1843 } 1844 | UNKNOWN_CPP_NAME 1845 ; 1846 1847%% 1848 1849/* Returns a stoken of the operator name given by OP (which does not 1850 include the string "operator"). */ 1851 1852static struct stoken 1853operator_stoken (const char *op) 1854{ 1855 struct stoken st = { NULL, 0 }; 1856 char *buf; 1857 1858 st.length = CP_OPERATOR_LEN + strlen (op); 1859 buf = (char *) malloc (st.length + 1); 1860 strcpy (buf, CP_OPERATOR_STR); 1861 strcat (buf, op); 1862 st.ptr = buf; 1863 1864 /* The toplevel (c_parse) will free the memory allocated here. */ 1865 cpstate->strings.emplace_back (buf); 1866 return st; 1867}; 1868 1869/* Returns a stoken of the type named TYPE. */ 1870 1871static struct stoken 1872typename_stoken (const char *type) 1873{ 1874 struct stoken st = { type, 0 }; 1875 st.length = strlen (type); 1876 return st; 1877}; 1878 1879/* Return true if the type is aggregate-like. */ 1880 1881static int 1882type_aggregate_p (struct type *type) 1883{ 1884 return (type->code () == TYPE_CODE_STRUCT 1885 || type->code () == TYPE_CODE_UNION 1886 || type->code () == TYPE_CODE_NAMESPACE 1887 || (type->code () == TYPE_CODE_ENUM 1888 && type->is_declared_class ())); 1889} 1890 1891/* Validate a parameter typelist. */ 1892 1893static void 1894check_parameter_typelist (std::vector<struct type *> *params) 1895{ 1896 struct type *type; 1897 int ix; 1898 1899 for (ix = 0; ix < params->size (); ++ix) 1900 { 1901 type = (*params)[ix]; 1902 if (type != NULL && check_typedef (type)->code () == TYPE_CODE_VOID) 1903 { 1904 if (ix == 0) 1905 { 1906 if (params->size () == 1) 1907 { 1908 /* Ok. */ 1909 break; 1910 } 1911 error (_("parameter types following 'void'")); 1912 } 1913 else 1914 error (_("'void' invalid as parameter type")); 1915 } 1916 } 1917} 1918 1919/* Take care of parsing a number (anything that starts with a digit). 1920 Set yylval and return the token type; update lexptr. 1921 LEN is the number of characters in it. */ 1922 1923/*** Needs some error checking for the float case ***/ 1924 1925static int 1926parse_number (struct parser_state *par_state, 1927 const char *buf, int len, int parsed_float, YYSTYPE *putithere) 1928{ 1929 ULONGEST n = 0; 1930 ULONGEST prevn = 0; 1931 1932 int i = 0; 1933 int c; 1934 int base = input_radix; 1935 int unsigned_p = 0; 1936 1937 /* Number of "L" suffixes encountered. */ 1938 int long_p = 0; 1939 1940 /* Imaginary number. */ 1941 bool imaginary_p = false; 1942 1943 /* We have found a "L" or "U" (or "i") suffix. */ 1944 int found_suffix = 0; 1945 1946 char *p; 1947 1948 p = (char *) alloca (len); 1949 memcpy (p, buf, len); 1950 1951 if (parsed_float) 1952 { 1953 if (len >= 1 && p[len - 1] == 'i') 1954 { 1955 imaginary_p = true; 1956 --len; 1957 } 1958 1959 /* Handle suffixes for decimal floating-point: "df", "dd" or "dl". */ 1960 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f') 1961 { 1962 putithere->typed_val_float.type 1963 = parse_type (par_state)->builtin_decfloat; 1964 len -= 2; 1965 } 1966 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd') 1967 { 1968 putithere->typed_val_float.type 1969 = parse_type (par_state)->builtin_decdouble; 1970 len -= 2; 1971 } 1972 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l') 1973 { 1974 putithere->typed_val_float.type 1975 = parse_type (par_state)->builtin_declong; 1976 len -= 2; 1977 } 1978 /* Handle suffixes: 'f' for float, 'l' for long double. */ 1979 else if (len >= 1 && TOLOWER (p[len - 1]) == 'f') 1980 { 1981 putithere->typed_val_float.type 1982 = parse_type (par_state)->builtin_float; 1983 len -= 1; 1984 } 1985 else if (len >= 1 && TOLOWER (p[len - 1]) == 'l') 1986 { 1987 putithere->typed_val_float.type 1988 = parse_type (par_state)->builtin_long_double; 1989 len -= 1; 1990 } 1991 /* Default type for floating-point literals is double. */ 1992 else 1993 { 1994 putithere->typed_val_float.type 1995 = parse_type (par_state)->builtin_double; 1996 } 1997 1998 if (!parse_float (p, len, 1999 putithere->typed_val_float.type, 2000 putithere->typed_val_float.val)) 2001 return ERROR; 2002 2003 if (imaginary_p) 2004 putithere->typed_val_float.type 2005 = init_complex_type (nullptr, putithere->typed_val_float.type); 2006 2007 return imaginary_p ? COMPLEX_FLOAT : FLOAT; 2008 } 2009 2010 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */ 2011 if (p[0] == '0' && len > 1) 2012 switch (p[1]) 2013 { 2014 case 'x': 2015 case 'X': 2016 if (len >= 3) 2017 { 2018 p += 2; 2019 base = 16; 2020 len -= 2; 2021 } 2022 break; 2023 2024 case 'b': 2025 case 'B': 2026 if (len >= 3) 2027 { 2028 p += 2; 2029 base = 2; 2030 len -= 2; 2031 } 2032 break; 2033 2034 case 't': 2035 case 'T': 2036 case 'd': 2037 case 'D': 2038 if (len >= 3) 2039 { 2040 p += 2; 2041 base = 10; 2042 len -= 2; 2043 } 2044 break; 2045 2046 default: 2047 base = 8; 2048 break; 2049 } 2050 2051 while (len-- > 0) 2052 { 2053 c = *p++; 2054 if (c >= 'A' && c <= 'Z') 2055 c += 'a' - 'A'; 2056 if (c != 'l' && c != 'u' && c != 'i') 2057 n *= base; 2058 if (c >= '0' && c <= '9') 2059 { 2060 if (found_suffix) 2061 return ERROR; 2062 n += i = c - '0'; 2063 } 2064 else 2065 { 2066 if (base > 10 && c >= 'a' && c <= 'f') 2067 { 2068 if (found_suffix) 2069 return ERROR; 2070 n += i = c - 'a' + 10; 2071 } 2072 else if (c == 'l') 2073 { 2074 ++long_p; 2075 found_suffix = 1; 2076 } 2077 else if (c == 'u') 2078 { 2079 unsigned_p = 1; 2080 found_suffix = 1; 2081 } 2082 else if (c == 'i') 2083 { 2084 imaginary_p = true; 2085 found_suffix = 1; 2086 } 2087 else 2088 return ERROR; /* Char not a digit */ 2089 } 2090 if (i >= base) 2091 return ERROR; /* Invalid digit in this base */ 2092 2093 if (c != 'l' && c != 'u' && c != 'i') 2094 { 2095 /* Test for overflow. */ 2096 if (prevn == 0 && n == 0) 2097 ; 2098 else if (prevn >= n) 2099 error (_("Numeric constant too large.")); 2100 } 2101 prevn = n; 2102 } 2103 2104 /* An integer constant is an int, a long, or a long long. An L 2105 suffix forces it to be long; an LL suffix forces it to be long 2106 long. If not forced to a larger size, it gets the first type of 2107 the above that it fits in. To figure out whether it fits, we 2108 shift it right and see whether anything remains. Note that we 2109 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one 2110 operation, because many compilers will warn about such a shift 2111 (which always produces a zero result). Sometimes gdbarch_int_bit 2112 or gdbarch_long_bit will be that big, sometimes not. To deal with 2113 the case where it is we just always shift the value more than 2114 once, with fewer bits each time. */ 2115 int int_bits = gdbarch_int_bit (par_state->gdbarch ()); 2116 int long_bits = gdbarch_long_bit (par_state->gdbarch ()); 2117 int long_long_bits = gdbarch_long_long_bit (par_state->gdbarch ()); 2118 bool have_signed 2119 /* No 'u' suffix. */ 2120 = !unsigned_p; 2121 bool have_unsigned 2122 = ((/* 'u' suffix. */ 2123 unsigned_p) 2124 || (/* Not a decimal. */ 2125 base != 10) 2126 || (/* Allowed as a convenience, in case decimal doesn't fit in largest 2127 signed type. */ 2128 !fits_in_type (1, n, long_long_bits, true))); 2129 bool have_int 2130 /* No 'l' or 'll' suffix. */ 2131 = long_p == 0; 2132 bool have_long 2133 /* No 'll' suffix. */ 2134 = long_p <= 1; 2135 if (have_int && have_signed && fits_in_type (1, n, int_bits, true)) 2136 putithere->typed_val_int.type = parse_type (par_state)->builtin_int; 2137 else if (have_int && have_unsigned && fits_in_type (1, n, int_bits, false)) 2138 putithere->typed_val_int.type 2139 = parse_type (par_state)->builtin_unsigned_int; 2140 else if (have_long && have_signed && fits_in_type (1, n, long_bits, true)) 2141 putithere->typed_val_int.type = parse_type (par_state)->builtin_long; 2142 else if (have_long && have_unsigned && fits_in_type (1, n, long_bits, false)) 2143 putithere->typed_val_int.type 2144 = parse_type (par_state)->builtin_unsigned_long; 2145 else if (have_signed && fits_in_type (1, n, long_long_bits, true)) 2146 putithere->typed_val_int.type 2147 = parse_type (par_state)->builtin_long_long; 2148 else if (have_unsigned && fits_in_type (1, n, long_long_bits, false)) 2149 putithere->typed_val_int.type 2150 = parse_type (par_state)->builtin_unsigned_long_long; 2151 else 2152 error (_("Numeric constant too large.")); 2153 putithere->typed_val_int.val = n; 2154 2155 if (imaginary_p) 2156 putithere->typed_val_int.type 2157 = init_complex_type (nullptr, putithere->typed_val_int.type); 2158 2159 return imaginary_p ? COMPLEX_INT : INT; 2160} 2161 2162/* Temporary obstack used for holding strings. */ 2163static struct obstack tempbuf; 2164static int tempbuf_init; 2165 2166/* Parse a C escape sequence. The initial backslash of the sequence 2167 is at (*PTR)[-1]. *PTR will be updated to point to just after the 2168 last character of the sequence. If OUTPUT is not NULL, the 2169 translated form of the escape sequence will be written there. If 2170 OUTPUT is NULL, no output is written and the call will only affect 2171 *PTR. If an escape sequence is expressed in target bytes, then the 2172 entire sequence will simply be copied to OUTPUT. Return 1 if any 2173 character was emitted, 0 otherwise. */ 2174 2175int 2176c_parse_escape (const char **ptr, struct obstack *output) 2177{ 2178 const char *tokptr = *ptr; 2179 int result = 1; 2180 2181 /* Some escape sequences undergo character set conversion. Those we 2182 translate here. */ 2183 switch (*tokptr) 2184 { 2185 /* Hex escapes do not undergo character set conversion, so keep 2186 the escape sequence for later. */ 2187 case 'x': 2188 if (output) 2189 obstack_grow_str (output, "\\x"); 2190 ++tokptr; 2191 if (!ISXDIGIT (*tokptr)) 2192 error (_("\\x escape without a following hex digit")); 2193 while (ISXDIGIT (*tokptr)) 2194 { 2195 if (output) 2196 obstack_1grow (output, *tokptr); 2197 ++tokptr; 2198 } 2199 break; 2200 2201 /* Octal escapes do not undergo character set conversion, so 2202 keep the escape sequence for later. */ 2203 case '0': 2204 case '1': 2205 case '2': 2206 case '3': 2207 case '4': 2208 case '5': 2209 case '6': 2210 case '7': 2211 { 2212 int i; 2213 if (output) 2214 obstack_grow_str (output, "\\"); 2215 for (i = 0; 2216 i < 3 && ISDIGIT (*tokptr) && *tokptr != '8' && *tokptr != '9'; 2217 ++i) 2218 { 2219 if (output) 2220 obstack_1grow (output, *tokptr); 2221 ++tokptr; 2222 } 2223 } 2224 break; 2225 2226 /* We handle UCNs later. We could handle them here, but that 2227 would mean a spurious error in the case where the UCN could 2228 be converted to the target charset but not the host 2229 charset. */ 2230 case 'u': 2231 case 'U': 2232 { 2233 char c = *tokptr; 2234 int i, len = c == 'U' ? 8 : 4; 2235 if (output) 2236 { 2237 obstack_1grow (output, '\\'); 2238 obstack_1grow (output, *tokptr); 2239 } 2240 ++tokptr; 2241 if (!ISXDIGIT (*tokptr)) 2242 error (_("\\%c escape without a following hex digit"), c); 2243 for (i = 0; i < len && ISXDIGIT (*tokptr); ++i) 2244 { 2245 if (output) 2246 obstack_1grow (output, *tokptr); 2247 ++tokptr; 2248 } 2249 } 2250 break; 2251 2252 /* We must pass backslash through so that it does not 2253 cause quoting during the second expansion. */ 2254 case '\\': 2255 if (output) 2256 obstack_grow_str (output, "\\\\"); 2257 ++tokptr; 2258 break; 2259 2260 /* Escapes which undergo conversion. */ 2261 case 'a': 2262 if (output) 2263 obstack_1grow (output, '\a'); 2264 ++tokptr; 2265 break; 2266 case 'b': 2267 if (output) 2268 obstack_1grow (output, '\b'); 2269 ++tokptr; 2270 break; 2271 case 'f': 2272 if (output) 2273 obstack_1grow (output, '\f'); 2274 ++tokptr; 2275 break; 2276 case 'n': 2277 if (output) 2278 obstack_1grow (output, '\n'); 2279 ++tokptr; 2280 break; 2281 case 'r': 2282 if (output) 2283 obstack_1grow (output, '\r'); 2284 ++tokptr; 2285 break; 2286 case 't': 2287 if (output) 2288 obstack_1grow (output, '\t'); 2289 ++tokptr; 2290 break; 2291 case 'v': 2292 if (output) 2293 obstack_1grow (output, '\v'); 2294 ++tokptr; 2295 break; 2296 2297 /* GCC extension. */ 2298 case 'e': 2299 if (output) 2300 obstack_1grow (output, HOST_ESCAPE_CHAR); 2301 ++tokptr; 2302 break; 2303 2304 /* Backslash-newline expands to nothing at all. */ 2305 case '\n': 2306 ++tokptr; 2307 result = 0; 2308 break; 2309 2310 /* A few escapes just expand to the character itself. */ 2311 case '\'': 2312 case '\"': 2313 case '?': 2314 /* GCC extensions. */ 2315 case '(': 2316 case '{': 2317 case '[': 2318 case '%': 2319 /* Unrecognized escapes turn into the character itself. */ 2320 default: 2321 if (output) 2322 obstack_1grow (output, *tokptr); 2323 ++tokptr; 2324 break; 2325 } 2326 *ptr = tokptr; 2327 return result; 2328} 2329 2330/* Parse a string or character literal from TOKPTR. The string or 2331 character may be wide or unicode. *OUTPTR is set to just after the 2332 end of the literal in the input string. The resulting token is 2333 stored in VALUE. This returns a token value, either STRING or 2334 CHAR, depending on what was parsed. *HOST_CHARS is set to the 2335 number of host characters in the literal. */ 2336 2337static int 2338parse_string_or_char (const char *tokptr, const char **outptr, 2339 struct typed_stoken *value, int *host_chars) 2340{ 2341 int quote; 2342 c_string_type type; 2343 int is_objc = 0; 2344 2345 /* Build the gdb internal form of the input string in tempbuf. Note 2346 that the buffer is null byte terminated *only* for the 2347 convenience of debugging gdb itself and printing the buffer 2348 contents when the buffer contains no embedded nulls. Gdb does 2349 not depend upon the buffer being null byte terminated, it uses 2350 the length string instead. This allows gdb to handle C strings 2351 (as well as strings in other languages) with embedded null 2352 bytes */ 2353 2354 if (!tempbuf_init) 2355 tempbuf_init = 1; 2356 else 2357 obstack_free (&tempbuf, NULL); 2358 obstack_init (&tempbuf); 2359 2360 /* Record the string type. */ 2361 if (*tokptr == 'L') 2362 { 2363 type = C_WIDE_STRING; 2364 ++tokptr; 2365 } 2366 else if (*tokptr == 'u') 2367 { 2368 type = C_STRING_16; 2369 ++tokptr; 2370 } 2371 else if (*tokptr == 'U') 2372 { 2373 type = C_STRING_32; 2374 ++tokptr; 2375 } 2376 else if (*tokptr == '@') 2377 { 2378 /* An Objective C string. */ 2379 is_objc = 1; 2380 type = C_STRING; 2381 ++tokptr; 2382 } 2383 else 2384 type = C_STRING; 2385 2386 /* Skip the quote. */ 2387 quote = *tokptr; 2388 if (quote == '\'') 2389 type |= C_CHAR; 2390 ++tokptr; 2391 2392 *host_chars = 0; 2393 2394 while (*tokptr) 2395 { 2396 char c = *tokptr; 2397 if (c == '\\') 2398 { 2399 ++tokptr; 2400 *host_chars += c_parse_escape (&tokptr, &tempbuf); 2401 } 2402 else if (c == quote) 2403 break; 2404 else 2405 { 2406 obstack_1grow (&tempbuf, c); 2407 ++tokptr; 2408 /* FIXME: this does the wrong thing with multi-byte host 2409 characters. We could use mbrlen here, but that would 2410 make "set host-charset" a bit less useful. */ 2411 ++*host_chars; 2412 } 2413 } 2414 2415 if (*tokptr != quote) 2416 { 2417 if (quote == '"') 2418 error (_("Unterminated string in expression.")); 2419 else 2420 error (_("Unmatched single quote.")); 2421 } 2422 ++tokptr; 2423 2424 value->type = type; 2425 value->ptr = (char *) obstack_base (&tempbuf); 2426 value->length = obstack_object_size (&tempbuf); 2427 2428 *outptr = tokptr; 2429 2430 return quote == '"' ? (is_objc ? NSSTRING : STRING) : CHAR; 2431} 2432 2433/* This is used to associate some attributes with a token. */ 2434 2435enum token_flag 2436{ 2437 /* If this bit is set, the token is C++-only. */ 2438 2439 FLAG_CXX = 1, 2440 2441 /* If this bit is set, the token is C-only. */ 2442 2443 FLAG_C = 2, 2444 2445 /* If this bit is set, the token is conditional: if there is a 2446 symbol of the same name, then the token is a symbol; otherwise, 2447 the token is a keyword. */ 2448 2449 FLAG_SHADOW = 4 2450}; 2451DEF_ENUM_FLAGS_TYPE (enum token_flag, token_flags); 2452 2453struct token 2454{ 2455 const char *oper; 2456 int token; 2457 enum exp_opcode opcode; 2458 token_flags flags; 2459}; 2460 2461static const struct token tokentab3[] = 2462 { 2463 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0}, 2464 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0}, 2465 {"->*", ARROW_STAR, OP_NULL, FLAG_CXX}, 2466 {"...", DOTDOTDOT, OP_NULL, 0} 2467 }; 2468 2469static const struct token tokentab2[] = 2470 { 2471 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0}, 2472 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0}, 2473 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0}, 2474 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0}, 2475 {"%=", ASSIGN_MODIFY, BINOP_REM, 0}, 2476 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0}, 2477 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0}, 2478 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0}, 2479 {"++", INCREMENT, OP_NULL, 0}, 2480 {"--", DECREMENT, OP_NULL, 0}, 2481 {"->", ARROW, OP_NULL, 0}, 2482 {"&&", ANDAND, OP_NULL, 0}, 2483 {"||", OROR, OP_NULL, 0}, 2484 /* "::" is *not* only C++: gdb overrides its meaning in several 2485 different ways, e.g., 'filename'::func, function::variable. */ 2486 {"::", COLONCOLON, OP_NULL, 0}, 2487 {"<<", LSH, OP_NULL, 0}, 2488 {">>", RSH, OP_NULL, 0}, 2489 {"==", EQUAL, OP_NULL, 0}, 2490 {"!=", NOTEQUAL, OP_NULL, 0}, 2491 {"<=", LEQ, OP_NULL, 0}, 2492 {">=", GEQ, OP_NULL, 0}, 2493 {".*", DOT_STAR, OP_NULL, FLAG_CXX} 2494 }; 2495 2496/* Identifier-like tokens. Only type-specifiers than can appear in 2497 multi-word type names (for example 'double' can appear in 'long 2498 double') need to be listed here. type-specifiers that are only ever 2499 single word (like 'char') are handled by the classify_name function. */ 2500static const struct token ident_tokens[] = 2501 { 2502 {"unsigned", UNSIGNED, OP_NULL, 0}, 2503 {"template", TEMPLATE, OP_NULL, FLAG_CXX}, 2504 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0}, 2505 {"struct", STRUCT, OP_NULL, 0}, 2506 {"signed", SIGNED_KEYWORD, OP_NULL, 0}, 2507 {"sizeof", SIZEOF, OP_NULL, 0}, 2508 {"_Alignof", ALIGNOF, OP_NULL, 0}, 2509 {"alignof", ALIGNOF, OP_NULL, FLAG_CXX}, 2510 {"double", DOUBLE_KEYWORD, OP_NULL, 0}, 2511 {"float", FLOAT_KEYWORD, OP_NULL, 0}, 2512 {"false", FALSEKEYWORD, OP_NULL, FLAG_CXX}, 2513 {"class", CLASS, OP_NULL, FLAG_CXX}, 2514 {"union", UNION, OP_NULL, 0}, 2515 {"short", SHORT, OP_NULL, 0}, 2516 {"const", CONST_KEYWORD, OP_NULL, 0}, 2517 {"restrict", RESTRICT, OP_NULL, FLAG_C | FLAG_SHADOW}, 2518 {"__restrict__", RESTRICT, OP_NULL, 0}, 2519 {"__restrict", RESTRICT, OP_NULL, 0}, 2520 {"_Atomic", ATOMIC, OP_NULL, 0}, 2521 {"enum", ENUM, OP_NULL, 0}, 2522 {"long", LONG, OP_NULL, 0}, 2523 {"_Complex", COMPLEX, OP_NULL, 0}, 2524 {"__complex__", COMPLEX, OP_NULL, 0}, 2525 2526 {"true", TRUEKEYWORD, OP_NULL, FLAG_CXX}, 2527 {"int", INT_KEYWORD, OP_NULL, 0}, 2528 {"new", NEW, OP_NULL, FLAG_CXX}, 2529 {"delete", DELETE, OP_NULL, FLAG_CXX}, 2530 {"operator", OPERATOR, OP_NULL, FLAG_CXX}, 2531 2532 {"and", ANDAND, OP_NULL, FLAG_CXX}, 2533 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, FLAG_CXX}, 2534 {"bitand", '&', OP_NULL, FLAG_CXX}, 2535 {"bitor", '|', OP_NULL, FLAG_CXX}, 2536 {"compl", '~', OP_NULL, FLAG_CXX}, 2537 {"not", '!', OP_NULL, FLAG_CXX}, 2538 {"not_eq", NOTEQUAL, OP_NULL, FLAG_CXX}, 2539 {"or", OROR, OP_NULL, FLAG_CXX}, 2540 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, FLAG_CXX}, 2541 {"xor", '^', OP_NULL, FLAG_CXX}, 2542 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, FLAG_CXX}, 2543 2544 {"const_cast", CONST_CAST, OP_NULL, FLAG_CXX }, 2545 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, FLAG_CXX }, 2546 {"static_cast", STATIC_CAST, OP_NULL, FLAG_CXX }, 2547 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, FLAG_CXX }, 2548 2549 {"__typeof__", TYPEOF, OP_TYPEOF, 0 }, 2550 {"__typeof", TYPEOF, OP_TYPEOF, 0 }, 2551 {"typeof", TYPEOF, OP_TYPEOF, FLAG_SHADOW }, 2552 {"__decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX }, 2553 {"decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX | FLAG_SHADOW }, 2554 2555 {"typeid", TYPEID, OP_TYPEID, FLAG_CXX} 2556 }; 2557 2558 2559static void 2560scan_macro_expansion (const char *expansion) 2561{ 2562 /* We'd better not be trying to push the stack twice. */ 2563 gdb_assert (! cpstate->macro_original_text); 2564 2565 /* Copy to the obstack. */ 2566 const char *copy = obstack_strdup (&cpstate->expansion_obstack, expansion); 2567 2568 /* Save the old lexptr value, so we can return to it when we're done 2569 parsing the expanded text. */ 2570 cpstate->macro_original_text = pstate->lexptr; 2571 pstate->lexptr = copy; 2572} 2573 2574static int 2575scanning_macro_expansion (void) 2576{ 2577 return cpstate->macro_original_text != 0; 2578} 2579 2580static void 2581finished_macro_expansion (void) 2582{ 2583 /* There'd better be something to pop back to. */ 2584 gdb_assert (cpstate->macro_original_text); 2585 2586 /* Pop back to the original text. */ 2587 pstate->lexptr = cpstate->macro_original_text; 2588 cpstate->macro_original_text = 0; 2589} 2590 2591/* Return true iff the token represents a C++ cast operator. */ 2592 2593static int 2594is_cast_operator (const char *token, int len) 2595{ 2596 return (! strncmp (token, "dynamic_cast", len) 2597 || ! strncmp (token, "static_cast", len) 2598 || ! strncmp (token, "reinterpret_cast", len) 2599 || ! strncmp (token, "const_cast", len)); 2600} 2601 2602/* The scope used for macro expansion. */ 2603static struct macro_scope *expression_macro_scope; 2604 2605/* This is set if a NAME token appeared at the very end of the input 2606 string, with no whitespace separating the name from the EOF. This 2607 is used only when parsing to do field name completion. */ 2608static int saw_name_at_eof; 2609 2610/* This is set if the previously-returned token was a structure 2611 operator -- either '.' or ARROW. */ 2612static bool last_was_structop; 2613 2614/* Depth of parentheses. */ 2615static int paren_depth; 2616 2617static int 2618get_namelen (const char *tokstart, bool dot) 2619{ 2620 int c; 2621 int namelen; 2622 2623 for (namelen = 0, c = tokstart[namelen]; 2624 (c == '_' || c == '$' || (dot && c == '.') || c_ident_is_alnum (c) || c == '<');) 2625 { 2626 /* Template parameter lists are part of the name. 2627 FIXME: This mishandles `print $a<4&&$a>3'. */ 2628 2629 if (c == '<') 2630 { 2631 if (! is_cast_operator (tokstart, namelen)) 2632 { 2633 /* Scan ahead to get rest of the template specification. Note 2634 that we look ahead only when the '<' adjoins non-whitespace 2635 characters; for comparison expressions, e.g. "a < b > c", 2636 there must be spaces before the '<', etc. */ 2637 const char *p = find_template_name_end (tokstart + namelen); 2638 2639 if (p) 2640 namelen = p - tokstart; 2641 } 2642 break; 2643 } 2644 c = tokstart[++namelen]; 2645 } 2646 return namelen; 2647} 2648 2649static bool is_generated_symbol (const char *symbol) 2650{ 2651 /* generated symbol are of the form: 2652 2653 <symbol>.<number> 2654 <symbol>.isra.<number> 2655 <symbol>.part.<number> 2656 2657 So we see if the symbol ends with .<number> 2658 */ 2659 2660 int len = get_namelen (symbol, true); 2661 int ndigits; 2662 2663 if (len-- == 0) 2664 return false; 2665 2666 for (ndigits = 0; ndigits <= len && ISDIGIT(symbol[len - ndigits]); ndigits++) 2667 continue; 2668 2669 if (ndigits == 0) 2670 return false; 2671 2672 return symbol[len - ndigits] == '.'; 2673} 2674 2675/* Read one token, getting characters through lexptr. */ 2676 2677static int 2678lex_one_token (struct parser_state *par_state, bool *is_quoted_name) 2679{ 2680 int c; 2681 int namelen; 2682 const char *tokstart; 2683 bool saw_structop = last_was_structop; 2684 2685 last_was_structop = false; 2686 *is_quoted_name = false; 2687 2688 retry: 2689 2690 /* Check if this is a macro invocation that we need to expand. */ 2691 if (! scanning_macro_expansion ()) 2692 { 2693 gdb::unique_xmalloc_ptr<char> expanded 2694 = macro_expand_next (&pstate->lexptr, *expression_macro_scope); 2695 2696 if (expanded != nullptr) 2697 scan_macro_expansion (expanded.get ()); 2698 } 2699 2700 pstate->prev_lexptr = pstate->lexptr; 2701 2702 tokstart = pstate->lexptr; 2703 /* See if it is a special token of length 3. */ 2704 for (const auto &token : tokentab3) 2705 if (strncmp (tokstart, token.oper, 3) == 0) 2706 { 2707 if ((token.flags & FLAG_CXX) != 0 2708 && par_state->language ()->la_language != language_cplus) 2709 break; 2710 gdb_assert ((token.flags & FLAG_C) == 0); 2711 2712 pstate->lexptr += 3; 2713 yylval.opcode = token.opcode; 2714 return token.token; 2715 } 2716 2717 /* See if it is a special token of length 2. */ 2718 for (const auto &token : tokentab2) 2719 if (strncmp (tokstart, token.oper, 2) == 0) 2720 { 2721 if ((token.flags & FLAG_CXX) != 0 2722 && par_state->language ()->la_language != language_cplus) 2723 break; 2724 gdb_assert ((token.flags & FLAG_C) == 0); 2725 2726 pstate->lexptr += 2; 2727 yylval.opcode = token.opcode; 2728 if (token.token == ARROW) 2729 last_was_structop = 1; 2730 return token.token; 2731 } 2732 2733 switch (c = *tokstart) 2734 { 2735 case 0: 2736 /* If we were just scanning the result of a macro expansion, 2737 then we need to resume scanning the original text. 2738 If we're parsing for field name completion, and the previous 2739 token allows such completion, return a COMPLETE token. 2740 Otherwise, we were already scanning the original text, and 2741 we're really done. */ 2742 if (scanning_macro_expansion ()) 2743 { 2744 finished_macro_expansion (); 2745 goto retry; 2746 } 2747 else if (saw_name_at_eof) 2748 { 2749 saw_name_at_eof = 0; 2750 return COMPLETE; 2751 } 2752 else if (par_state->parse_completion && saw_structop) 2753 return COMPLETE; 2754 else 2755 return 0; 2756 2757 case ' ': 2758 case '\t': 2759 case '\n': 2760 pstate->lexptr++; 2761 goto retry; 2762 2763 case '[': 2764 case '(': 2765 paren_depth++; 2766 pstate->lexptr++; 2767 if (par_state->language ()->la_language == language_objc 2768 && c == '[') 2769 return OBJC_LBRAC; 2770 return c; 2771 2772 case ']': 2773 case ')': 2774 if (paren_depth == 0) 2775 return 0; 2776 paren_depth--; 2777 pstate->lexptr++; 2778 return c; 2779 2780 case ',': 2781 if (pstate->comma_terminates 2782 && paren_depth == 0 2783 && ! scanning_macro_expansion ()) 2784 return 0; 2785 pstate->lexptr++; 2786 return c; 2787 2788 case '.': 2789 /* Might be a floating point number. */ 2790 if (pstate->lexptr[1] < '0' || pstate->lexptr[1] > '9') 2791 { 2792 last_was_structop = true; 2793 goto symbol; /* Nope, must be a symbol. */ 2794 } 2795 /* FALL THRU. */ 2796 2797 case '0': 2798 case '1': 2799 case '2': 2800 case '3': 2801 case '4': 2802 case '5': 2803 case '6': 2804 case '7': 2805 case '8': 2806 case '9': 2807 { 2808 /* It's a number. */ 2809 int got_dot = 0, got_e = 0, got_p = 0, toktype; 2810 const char *p = tokstart; 2811 int hex = input_radix > 10; 2812 2813 if (c == '0' && (p[1] == 'x' || p[1] == 'X')) 2814 { 2815 p += 2; 2816 hex = 1; 2817 } 2818 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D')) 2819 { 2820 p += 2; 2821 hex = 0; 2822 } 2823 2824 for (;; ++p) 2825 { 2826 /* This test includes !hex because 'e' is a valid hex digit 2827 and thus does not indicate a floating point number when 2828 the radix is hex. */ 2829 if (!hex && !got_e && !got_p && (*p == 'e' || *p == 'E')) 2830 got_dot = got_e = 1; 2831 else if (!got_e && !got_p && (*p == 'p' || *p == 'P')) 2832 got_dot = got_p = 1; 2833 /* This test does not include !hex, because a '.' always indicates 2834 a decimal floating point number regardless of the radix. */ 2835 else if (!got_dot && *p == '.') 2836 got_dot = 1; 2837 else if (((got_e && (p[-1] == 'e' || p[-1] == 'E')) 2838 || (got_p && (p[-1] == 'p' || p[-1] == 'P'))) 2839 && (*p == '-' || *p == '+')) 2840 /* This is the sign of the exponent, not the end of the 2841 number. */ 2842 continue; 2843 /* We will take any letters or digits. parse_number will 2844 complain if past the radix, or if L or U are not final. */ 2845 else if ((*p < '0' || *p > '9') 2846 && ((*p < 'a' || *p > 'z') 2847 && (*p < 'A' || *p > 'Z'))) 2848 break; 2849 } 2850 toktype = parse_number (par_state, tokstart, p - tokstart, 2851 got_dot | got_e | got_p, &yylval); 2852 if (toktype == ERROR) 2853 { 2854 char *err_copy = (char *) alloca (p - tokstart + 1); 2855 2856 memcpy (err_copy, tokstart, p - tokstart); 2857 err_copy[p - tokstart] = 0; 2858 error (_("Invalid number \"%s\"."), err_copy); 2859 } 2860 pstate->lexptr = p; 2861 return toktype; 2862 } 2863 2864 case '@': 2865 { 2866 const char *p = &tokstart[1]; 2867 2868 if (par_state->language ()->la_language == language_objc) 2869 { 2870 size_t len = strlen ("selector"); 2871 2872 if (strncmp (p, "selector", len) == 0 2873 && (p[len] == '\0' || ISSPACE (p[len]))) 2874 { 2875 pstate->lexptr = p + len; 2876 return SELECTOR; 2877 } 2878 else if (*p == '"') 2879 goto parse_string; 2880 } 2881 2882 while (ISSPACE (*p)) 2883 p++; 2884 size_t len = strlen ("entry"); 2885 if (strncmp (p, "entry", len) == 0 && !c_ident_is_alnum (p[len]) 2886 && p[len] != '_') 2887 { 2888 pstate->lexptr = &p[len]; 2889 return ENTRY; 2890 } 2891 } 2892 /* FALLTHRU */ 2893 case '+': 2894 case '-': 2895 case '*': 2896 case '/': 2897 case '%': 2898 case '|': 2899 case '&': 2900 case '^': 2901 case '~': 2902 case '!': 2903 case '<': 2904 case '>': 2905 case '?': 2906 case ':': 2907 case '=': 2908 case '{': 2909 case '}': 2910 symbol: 2911 pstate->lexptr++; 2912 return c; 2913 2914 case 'L': 2915 case 'u': 2916 case 'U': 2917 if (tokstart[1] != '"' && tokstart[1] != '\'') 2918 break; 2919 /* Fall through. */ 2920 case '\'': 2921 case '"': 2922 2923 parse_string: 2924 { 2925 int host_len; 2926 int result = parse_string_or_char (tokstart, &pstate->lexptr, 2927 &yylval.tsval, &host_len); 2928 if (result == CHAR) 2929 { 2930 if (host_len == 0) 2931 error (_("Empty character constant.")); 2932 else if (host_len > 2 && c == '\'') 2933 { 2934 ++tokstart; 2935 namelen = pstate->lexptr - tokstart - 1; 2936 *is_quoted_name = true; 2937 2938 goto tryname; 2939 } 2940 else if (host_len > 1) 2941 error (_("Invalid character constant.")); 2942 } 2943 return result; 2944 } 2945 } 2946 2947 if (!(c == '_' || c == '$' || c_ident_is_alpha (c))) 2948 /* We must have come across a bad character (e.g. ';'). */ 2949 error (_("Invalid character '%c' in expression."), c); 2950 2951 /* It's a name. See how long it is. */ 2952 namelen = get_namelen (tokstart, is_generated_symbol (tokstart)); 2953 2954 /* The token "if" terminates the expression and is NOT removed from 2955 the input stream. It doesn't count if it appears in the 2956 expansion of a macro. */ 2957 if (namelen == 2 2958 && tokstart[0] == 'i' 2959 && tokstart[1] == 'f' 2960 && ! scanning_macro_expansion ()) 2961 { 2962 return 0; 2963 } 2964 2965 /* For the same reason (breakpoint conditions), "thread N" 2966 terminates the expression. "thread" could be an identifier, but 2967 an identifier is never followed by a number without intervening 2968 punctuation. "task" is similar. Handle abbreviations of these, 2969 similarly to breakpoint.c:find_condition_and_thread. */ 2970 if (namelen >= 1 2971 && (strncmp (tokstart, "thread", namelen) == 0 2972 || strncmp (tokstart, "task", namelen) == 0) 2973 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t') 2974 && ! scanning_macro_expansion ()) 2975 { 2976 const char *p = tokstart + namelen + 1; 2977 2978 while (*p == ' ' || *p == '\t') 2979 p++; 2980 if (*p >= '0' && *p <= '9') 2981 return 0; 2982 } 2983 2984 pstate->lexptr += namelen; 2985 2986 tryname: 2987 2988 yylval.sval.ptr = tokstart; 2989 yylval.sval.length = namelen; 2990 2991 /* Catch specific keywords. */ 2992 std::string copy = copy_name (yylval.sval); 2993 for (const auto &token : ident_tokens) 2994 if (copy == token.oper) 2995 { 2996 if ((token.flags & FLAG_CXX) != 0 2997 && par_state->language ()->la_language != language_cplus) 2998 break; 2999 if ((token.flags & FLAG_C) != 0 3000 && par_state->language ()->la_language != language_c 3001 && par_state->language ()->la_language != language_objc) 3002 break; 3003 3004 if ((token.flags & FLAG_SHADOW) != 0) 3005 { 3006 struct field_of_this_result is_a_field_of_this; 3007 3008 if (lookup_symbol (copy.c_str (), 3009 pstate->expression_context_block, 3010 VAR_DOMAIN, 3011 (par_state->language ()->la_language 3012 == language_cplus ? &is_a_field_of_this 3013 : NULL)).symbol 3014 != NULL) 3015 { 3016 /* The keyword is shadowed. */ 3017 break; 3018 } 3019 } 3020 3021 /* It is ok to always set this, even though we don't always 3022 strictly need to. */ 3023 yylval.opcode = token.opcode; 3024 return token.token; 3025 } 3026 3027 if (*tokstart == '$') 3028 return DOLLAR_VARIABLE; 3029 3030 if (pstate->parse_completion && *pstate->lexptr == '\0') 3031 saw_name_at_eof = 1; 3032 3033 yylval.ssym.stoken = yylval.sval; 3034 yylval.ssym.sym.symbol = NULL; 3035 yylval.ssym.sym.block = NULL; 3036 yylval.ssym.is_a_field_of_this = 0; 3037 return NAME; 3038} 3039 3040/* An object of this type is pushed on a FIFO by the "outer" lexer. */ 3041struct token_and_value 3042{ 3043 int token; 3044 YYSTYPE value; 3045}; 3046 3047/* A FIFO of tokens that have been read but not yet returned to the 3048 parser. */ 3049static std::vector<token_and_value> token_fifo; 3050 3051/* Non-zero if the lexer should return tokens from the FIFO. */ 3052static int popping; 3053 3054/* Temporary storage for c_lex; this holds symbol names as they are 3055 built up. */ 3056static auto_obstack name_obstack; 3057 3058/* Classify a NAME token. The contents of the token are in `yylval'. 3059 Updates yylval and returns the new token type. BLOCK is the block 3060 in which lookups start; this can be NULL to mean the global scope. 3061 IS_QUOTED_NAME is non-zero if the name token was originally quoted 3062 in single quotes. IS_AFTER_STRUCTOP is true if this name follows 3063 a structure operator -- either '.' or ARROW */ 3064 3065static int 3066classify_name (struct parser_state *par_state, const struct block *block, 3067 bool is_quoted_name, bool is_after_structop) 3068{ 3069 struct block_symbol bsym; 3070 struct field_of_this_result is_a_field_of_this; 3071 3072 std::string copy = copy_name (yylval.sval); 3073 3074 /* Initialize this in case we *don't* use it in this call; that way 3075 we can refer to it unconditionally below. */ 3076 memset (&is_a_field_of_this, 0, sizeof (is_a_field_of_this)); 3077 3078 bsym = lookup_symbol (copy.c_str (), block, VAR_DOMAIN, 3079 par_state->language ()->name_of_this () 3080 ? &is_a_field_of_this : NULL); 3081 3082 if (bsym.symbol && bsym.symbol->aclass () == LOC_BLOCK) 3083 { 3084 yylval.ssym.sym = bsym; 3085 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL; 3086 return BLOCKNAME; 3087 } 3088 else if (!bsym.symbol) 3089 { 3090 /* If we found a field of 'this', we might have erroneously 3091 found a constructor where we wanted a type name. Handle this 3092 case by noticing that we found a constructor and then look up 3093 the type tag instead. */ 3094 if (is_a_field_of_this.type != NULL 3095 && is_a_field_of_this.fn_field != NULL 3096 && TYPE_FN_FIELD_CONSTRUCTOR (is_a_field_of_this.fn_field->fn_fields, 3097 0)) 3098 { 3099 struct field_of_this_result inner_is_a_field_of_this; 3100 3101 bsym = lookup_symbol (copy.c_str (), block, STRUCT_DOMAIN, 3102 &inner_is_a_field_of_this); 3103 if (bsym.symbol != NULL) 3104 { 3105 yylval.tsym.type = bsym.symbol->type (); 3106 return TYPENAME; 3107 } 3108 } 3109 3110 /* If we found a field on the "this" object, or we are looking 3111 up a field on a struct, then we want to prefer it over a 3112 filename. However, if the name was quoted, then it is better 3113 to check for a filename or a block, since this is the only 3114 way the user has of requiring the extension to be used. */ 3115 if ((is_a_field_of_this.type == NULL && !is_after_structop) 3116 || is_quoted_name) 3117 { 3118 /* See if it's a file name. */ 3119 struct symtab *symtab; 3120 3121 symtab = lookup_symtab (copy.c_str ()); 3122 if (symtab) 3123 { 3124 yylval.bval 3125 = symtab->compunit ()->blockvector ()->static_block (); 3126 3127 return FILENAME; 3128 } 3129 } 3130 } 3131 3132 if (bsym.symbol && bsym.symbol->aclass () == LOC_TYPEDEF) 3133 { 3134 yylval.tsym.type = bsym.symbol->type (); 3135 return TYPENAME; 3136 } 3137 3138 /* See if it's an ObjC classname. */ 3139 if (par_state->language ()->la_language == language_objc && !bsym.symbol) 3140 { 3141 CORE_ADDR Class = lookup_objc_class (par_state->gdbarch (), 3142 copy.c_str ()); 3143 if (Class) 3144 { 3145 struct symbol *sym; 3146 3147 yylval.theclass.theclass = Class; 3148 sym = lookup_struct_typedef (copy.c_str (), 3149 par_state->expression_context_block, 1); 3150 if (sym) 3151 yylval.theclass.type = sym->type (); 3152 return CLASSNAME; 3153 } 3154 } 3155 3156 /* Input names that aren't symbols but ARE valid hex numbers, when 3157 the input radix permits them, can be names or numbers depending 3158 on the parse. Note we support radixes > 16 here. */ 3159 if (!bsym.symbol 3160 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10) 3161 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10))) 3162 { 3163 YYSTYPE newlval; /* Its value is ignored. */ 3164 int hextype = parse_number (par_state, copy.c_str (), yylval.sval.length, 3165 0, &newlval); 3166 3167 if (hextype == INT) 3168 { 3169 yylval.ssym.sym = bsym; 3170 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL; 3171 return NAME_OR_INT; 3172 } 3173 } 3174 3175 /* Any other kind of symbol */ 3176 yylval.ssym.sym = bsym; 3177 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL; 3178 3179 if (bsym.symbol == NULL 3180 && par_state->language ()->la_language == language_cplus 3181 && is_a_field_of_this.type == NULL 3182 && lookup_minimal_symbol (copy.c_str (), NULL, NULL).minsym == NULL) 3183 return UNKNOWN_CPP_NAME; 3184 3185 return NAME; 3186} 3187 3188/* Like classify_name, but used by the inner loop of the lexer, when a 3189 name might have already been seen. CONTEXT is the context type, or 3190 NULL if this is the first component of a name. */ 3191 3192static int 3193classify_inner_name (struct parser_state *par_state, 3194 const struct block *block, struct type *context) 3195{ 3196 struct type *type; 3197 3198 if (context == NULL) 3199 return classify_name (par_state, block, false, false); 3200 3201 type = check_typedef (context); 3202 if (!type_aggregate_p (type)) 3203 return ERROR; 3204 3205 std::string copy = copy_name (yylval.ssym.stoken); 3206 /* N.B. We assume the symbol can only be in VAR_DOMAIN. */ 3207 yylval.ssym.sym = cp_lookup_nested_symbol (type, copy.c_str (), block, 3208 VAR_DOMAIN); 3209 3210 /* If no symbol was found, search for a matching base class named 3211 COPY. This will allow users to enter qualified names of class members 3212 relative to the `this' pointer. */ 3213 if (yylval.ssym.sym.symbol == NULL) 3214 { 3215 struct type *base_type = cp_find_type_baseclass_by_name (type, 3216 copy.c_str ()); 3217 3218 if (base_type != NULL) 3219 { 3220 yylval.tsym.type = base_type; 3221 return TYPENAME; 3222 } 3223 3224 return ERROR; 3225 } 3226 3227 switch (yylval.ssym.sym.symbol->aclass ()) 3228 { 3229 case LOC_BLOCK: 3230 case LOC_LABEL: 3231 /* cp_lookup_nested_symbol might have accidentally found a constructor 3232 named COPY when we really wanted a base class of the same name. 3233 Double-check this case by looking for a base class. */ 3234 { 3235 struct type *base_type 3236 = cp_find_type_baseclass_by_name (type, copy.c_str ()); 3237 3238 if (base_type != NULL) 3239 { 3240 yylval.tsym.type = base_type; 3241 return TYPENAME; 3242 } 3243 } 3244 return ERROR; 3245 3246 case LOC_TYPEDEF: 3247 yylval.tsym.type = yylval.ssym.sym.symbol->type (); 3248 return TYPENAME; 3249 3250 default: 3251 return NAME; 3252 } 3253 internal_error (_("not reached")); 3254} 3255 3256/* The outer level of a two-level lexer. This calls the inner lexer 3257 to return tokens. It then either returns these tokens, or 3258 aggregates them into a larger token. This lets us work around a 3259 problem in our parsing approach, where the parser could not 3260 distinguish between qualified names and qualified types at the 3261 right point. 3262 3263 This approach is still not ideal, because it mishandles template 3264 types. See the comment in lex_one_token for an example. However, 3265 this is still an improvement over the earlier approach, and will 3266 suffice until we move to better parsing technology. */ 3267 3268static int 3269yylex (void) 3270{ 3271 token_and_value current; 3272 int first_was_coloncolon, last_was_coloncolon; 3273 struct type *context_type = NULL; 3274 int last_to_examine, next_to_examine, checkpoint; 3275 const struct block *search_block; 3276 bool is_quoted_name, last_lex_was_structop; 3277 3278 if (popping && !token_fifo.empty ()) 3279 goto do_pop; 3280 popping = 0; 3281 3282 last_lex_was_structop = last_was_structop; 3283 3284 /* Read the first token and decide what to do. Most of the 3285 subsequent code is C++-only; but also depends on seeing a "::" or 3286 name-like token. */ 3287 current.token = lex_one_token (pstate, &is_quoted_name); 3288 if (current.token == NAME) 3289 current.token = classify_name (pstate, pstate->expression_context_block, 3290 is_quoted_name, last_lex_was_structop); 3291 if (pstate->language ()->la_language != language_cplus 3292 || (current.token != TYPENAME && current.token != COLONCOLON 3293 && current.token != FILENAME)) 3294 return current.token; 3295 3296 /* Read any sequence of alternating "::" and name-like tokens into 3297 the token FIFO. */ 3298 current.value = yylval; 3299 token_fifo.push_back (current); 3300 last_was_coloncolon = current.token == COLONCOLON; 3301 while (1) 3302 { 3303 bool ignore; 3304 3305 /* We ignore quoted names other than the very first one. 3306 Subsequent ones do not have any special meaning. */ 3307 current.token = lex_one_token (pstate, &ignore); 3308 current.value = yylval; 3309 token_fifo.push_back (current); 3310 3311 if ((last_was_coloncolon && current.token != NAME) 3312 || (!last_was_coloncolon && current.token != COLONCOLON)) 3313 break; 3314 last_was_coloncolon = !last_was_coloncolon; 3315 } 3316 popping = 1; 3317 3318 /* We always read one extra token, so compute the number of tokens 3319 to examine accordingly. */ 3320 last_to_examine = token_fifo.size () - 2; 3321 next_to_examine = 0; 3322 3323 current = token_fifo[next_to_examine]; 3324 ++next_to_examine; 3325 3326 name_obstack.clear (); 3327 checkpoint = 0; 3328 if (current.token == FILENAME) 3329 search_block = current.value.bval; 3330 else if (current.token == COLONCOLON) 3331 search_block = NULL; 3332 else 3333 { 3334 gdb_assert (current.token == TYPENAME); 3335 search_block = pstate->expression_context_block; 3336 obstack_grow (&name_obstack, current.value.sval.ptr, 3337 current.value.sval.length); 3338 context_type = current.value.tsym.type; 3339 checkpoint = 1; 3340 } 3341 3342 first_was_coloncolon = current.token == COLONCOLON; 3343 last_was_coloncolon = first_was_coloncolon; 3344 3345 while (next_to_examine <= last_to_examine) 3346 { 3347 token_and_value next; 3348 3349 next = token_fifo[next_to_examine]; 3350 ++next_to_examine; 3351 3352 if (next.token == NAME && last_was_coloncolon) 3353 { 3354 int classification; 3355 3356 yylval = next.value; 3357 classification = classify_inner_name (pstate, search_block, 3358 context_type); 3359 /* We keep going until we either run out of names, or until 3360 we have a qualified name which is not a type. */ 3361 if (classification != TYPENAME && classification != NAME) 3362 break; 3363 3364 /* Accept up to this token. */ 3365 checkpoint = next_to_examine; 3366 3367 /* Update the partial name we are constructing. */ 3368 if (context_type != NULL) 3369 { 3370 /* We don't want to put a leading "::" into the name. */ 3371 obstack_grow_str (&name_obstack, "::"); 3372 } 3373 obstack_grow (&name_obstack, next.value.sval.ptr, 3374 next.value.sval.length); 3375 3376 yylval.sval.ptr = (const char *) obstack_base (&name_obstack); 3377 yylval.sval.length = obstack_object_size (&name_obstack); 3378 current.value = yylval; 3379 current.token = classification; 3380 3381 last_was_coloncolon = 0; 3382 3383 if (classification == NAME) 3384 break; 3385 3386 context_type = yylval.tsym.type; 3387 } 3388 else if (next.token == COLONCOLON && !last_was_coloncolon) 3389 last_was_coloncolon = 1; 3390 else 3391 { 3392 /* We've reached the end of the name. */ 3393 break; 3394 } 3395 } 3396 3397 /* If we have a replacement token, install it as the first token in 3398 the FIFO, and delete the other constituent tokens. */ 3399 if (checkpoint > 0) 3400 { 3401 current.value.sval.ptr 3402 = obstack_strndup (&cpstate->expansion_obstack, 3403 current.value.sval.ptr, 3404 current.value.sval.length); 3405 3406 token_fifo[0] = current; 3407 if (checkpoint > 1) 3408 token_fifo.erase (token_fifo.begin () + 1, 3409 token_fifo.begin () + checkpoint); 3410 } 3411 3412 do_pop: 3413 current = token_fifo[0]; 3414 token_fifo.erase (token_fifo.begin ()); 3415 yylval = current.value; 3416 return current.token; 3417} 3418 3419int 3420c_parse (struct parser_state *par_state) 3421{ 3422 /* Setting up the parser state. */ 3423 scoped_restore pstate_restore = make_scoped_restore (&pstate); 3424 gdb_assert (par_state != NULL); 3425 pstate = par_state; 3426 3427 c_parse_state cstate; 3428 scoped_restore cstate_restore = make_scoped_restore (&cpstate, &cstate); 3429 3430 gdb::unique_xmalloc_ptr<struct macro_scope> macro_scope; 3431 3432 if (par_state->expression_context_block) 3433 macro_scope 3434 = sal_macro_scope (find_pc_line (par_state->expression_context_pc, 0)); 3435 else 3436 macro_scope = default_macro_scope (); 3437 if (! macro_scope) 3438 macro_scope = user_macro_scope (); 3439 3440 scoped_restore restore_macro_scope 3441 = make_scoped_restore (&expression_macro_scope, macro_scope.get ()); 3442 3443 scoped_restore restore_yydebug = make_scoped_restore (&yydebug, 3444 parser_debug); 3445 3446 /* Initialize some state used by the lexer. */ 3447 last_was_structop = false; 3448 saw_name_at_eof = 0; 3449 paren_depth = 0; 3450 3451 token_fifo.clear (); 3452 popping = 0; 3453 name_obstack.clear (); 3454 3455 int result = yyparse (); 3456 if (!result) 3457 pstate->set_operation (pstate->pop ()); 3458 return result; 3459} 3460 3461#if defined(YYBISON) && YYBISON < 30800 3462 3463 3464/* This is called via the YYPRINT macro when parser debugging is 3465 enabled. It prints a token's value. */ 3466 3467static void 3468c_print_token (FILE *file, int type, YYSTYPE value) 3469{ 3470 switch (type) 3471 { 3472 case INT: 3473 parser_fprintf (file, "typed_val_int<%s, %s>", 3474 TYPE_SAFE_NAME (value.typed_val_int.type), 3475 pulongest (value.typed_val_int.val)); 3476 break; 3477 3478 case CHAR: 3479 case STRING: 3480 { 3481 char *copy = (char *) alloca (value.tsval.length + 1); 3482 3483 memcpy (copy, value.tsval.ptr, value.tsval.length); 3484 copy[value.tsval.length] = '\0'; 3485 3486 parser_fprintf (file, "tsval<type=%d, %s>", value.tsval.type, copy); 3487 } 3488 break; 3489 3490 case NSSTRING: 3491 case DOLLAR_VARIABLE: 3492 parser_fprintf (file, "sval<%s>", copy_name (value.sval).c_str ()); 3493 break; 3494 3495 case TYPENAME: 3496 parser_fprintf (file, "tsym<type=%s, name=%s>", 3497 TYPE_SAFE_NAME (value.tsym.type), 3498 copy_name (value.tsym.stoken).c_str ()); 3499 break; 3500 3501 case NAME: 3502 case UNKNOWN_CPP_NAME: 3503 case NAME_OR_INT: 3504 case BLOCKNAME: 3505 parser_fprintf (file, "ssym<name=%s, sym=%s, field_of_this=%d>", 3506 copy_name (value.ssym.stoken).c_str (), 3507 (value.ssym.sym.symbol == NULL 3508 ? "(null)" : value.ssym.sym.symbol->print_name ()), 3509 value.ssym.is_a_field_of_this); 3510 break; 3511 3512 case FILENAME: 3513 parser_fprintf (file, "bval<%s>", host_address_to_string (value.bval)); 3514 break; 3515 } 3516} 3517 3518#endif 3519 3520static void 3521yyerror (const char *msg) 3522{ 3523 if (pstate->prev_lexptr) 3524 pstate->lexptr = pstate->prev_lexptr; 3525 3526 error (_("A %s in expression, near `%s'."), msg, pstate->lexptr); 3527} 3528