1 /* YACC parser for C expressions, for GDB. 2 Copyright (C) 1986-2019 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 57 #define parse_type(ps) builtin_type (parse_gdbarch (ps)) 58 59 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, 60 etc). */ 61 #define GDB_YY_REMAP_PREFIX c_ 62 #include "yy-remap.h" 63 64 /* The state of the parser, used internally when we are parsing the 65 expression. */ 66 67 static struct parser_state *pstate = NULL; 68 69 /* Data that must be held for the duration of a parse. */ 70 71 struct c_parse_state 72 { 73 /* These are used to hold type lists and type stacks that are 74 allocated during the parse. */ 75 std::vector<std::unique_ptr<std::vector<struct type *>>> type_lists; 76 std::vector<std::unique_ptr<struct type_stack>> type_stacks; 77 78 /* Storage for some strings allocated during the parse. */ 79 std::vector<gdb::unique_xmalloc_ptr<char>> strings; 80 81 /* When we find that lexptr (the global var defined in parse.c) is 82 pointing at a macro invocation, we expand the invocation, and call 83 scan_macro_expansion to save the old lexptr here and point lexptr 84 into the expanded text. When we reach the end of that, we call 85 end_macro_expansion to pop back to the value we saved here. The 86 macro expansion code promises to return only fully-expanded text, 87 so we don't need to "push" more than one level. 88 89 This is disgusting, of course. It would be cleaner to do all macro 90 expansion beforehand, and then hand that to lexptr. But we don't 91 really know where the expression ends. Remember, in a command like 92 93 (gdb) break *ADDRESS if CONDITION 94 95 we evaluate ADDRESS in the scope of the current frame, but we 96 evaluate CONDITION in the scope of the breakpoint's location. So 97 it's simply wrong to try to macro-expand the whole thing at once. */ 98 const char *macro_original_text = nullptr; 99 100 /* We save all intermediate macro expansions on this obstack for the 101 duration of a single parse. The expansion text may sometimes have 102 to live past the end of the expansion, due to yacc lookahead. 103 Rather than try to be clever about saving the data for a single 104 token, we simply keep it all and delete it after parsing has 105 completed. */ 106 auto_obstack expansion_obstack; 107 }; 108 109 /* This is set and cleared in c_parse. */ 110 111 static struct c_parse_state *cpstate; 112 113 int yyparse (void); 114 115 static int yylex (void); 116 117 static void yyerror (const char *); 118 119 static int type_aggregate_p (struct type *); 120 121 %} 122 123 /* Although the yacc "value" of an expression is not used, 124 since the result is stored in the structure being created, 125 other node types do have values. */ 126 127 %union 128 { 129 LONGEST lval; 130 struct { 131 LONGEST val; 132 struct type *type; 133 } typed_val_int; 134 struct { 135 gdb_byte val[16]; 136 struct type *type; 137 } typed_val_float; 138 struct type *tval; 139 struct stoken sval; 140 struct typed_stoken tsval; 141 struct ttype tsym; 142 struct symtoken ssym; 143 int voidval; 144 const struct block *bval; 145 enum exp_opcode opcode; 146 147 struct stoken_vector svec; 148 std::vector<struct type *> *tvec; 149 150 struct type_stack *type_stack; 151 152 struct objc_class_str theclass; 153 } 154 155 %{ 156 /* YYSTYPE gets defined by %union */ 157 static int parse_number (struct parser_state *par_state, 158 const char *, int, int, YYSTYPE *); 159 static struct stoken operator_stoken (const char *); 160 static struct stoken typename_stoken (const char *); 161 static void check_parameter_typelist (std::vector<struct type *> *); 162 static void write_destructor_name (struct parser_state *par_state, 163 struct stoken); 164 165 #ifdef YYBISON 166 static void c_print_token (FILE *file, int type, YYSTYPE value); 167 #define YYPRINT(FILE, TYPE, VALUE) c_print_token (FILE, TYPE, VALUE) 168 #endif 169 %} 170 171 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly function_method 172 %type <lval> rcurly 173 %type <tval> type typebase 174 %type <tvec> nonempty_typelist func_mod parameter_typelist 175 /* %type <bval> block */ 176 177 /* Fancy type parsing. */ 178 %type <tval> ptype 179 %type <lval> array_mod 180 %type <tval> conversion_type_id 181 182 %type <type_stack> ptr_operator_ts abs_decl direct_abs_decl 183 184 %token <typed_val_int> INT 185 %token <typed_val_float> FLOAT 186 187 /* Both NAME and TYPENAME tokens represent symbols in the input, 188 and both convey their data as strings. 189 But a TYPENAME is a string that happens to be defined as a typedef 190 or builtin type name (such as int or char) 191 and a NAME is any other symbol. 192 Contexts where this distinction is not important can use the 193 nonterminal "name", which matches either NAME or TYPENAME. */ 194 195 %token <tsval> STRING 196 %token <sval> NSSTRING /* ObjC Foundation "NSString" literal */ 197 %token SELECTOR /* ObjC "@selector" pseudo-operator */ 198 %token <tsval> CHAR 199 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */ 200 %token <ssym> UNKNOWN_CPP_NAME 201 %token <voidval> COMPLETE 202 %token <tsym> TYPENAME 203 %token <theclass> CLASSNAME /* ObjC Class name */ 204 %type <sval> name field_name 205 %type <svec> string_exp 206 %type <ssym> name_not_typename 207 %type <tsym> type_name 208 209 /* This is like a '[' token, but is only generated when parsing 210 Objective C. This lets us reuse the same parser without 211 erroneously parsing ObjC-specific expressions in C. */ 212 %token OBJC_LBRAC 213 214 /* A NAME_OR_INT is a symbol which is not known in the symbol table, 215 but which would parse as a valid number in the current input radix. 216 E.g. "c" when input_radix==16. Depending on the parse, it will be 217 turned into a name or into a number. */ 218 219 %token <ssym> NAME_OR_INT 220 221 %token OPERATOR 222 %token STRUCT CLASS UNION ENUM SIZEOF ALIGNOF UNSIGNED COLONCOLON 223 %token TEMPLATE 224 %token ERROR 225 %token NEW DELETE 226 %type <sval> oper 227 %token REINTERPRET_CAST DYNAMIC_CAST STATIC_CAST CONST_CAST 228 %token ENTRY 229 %token TYPEOF 230 %token DECLTYPE 231 %token TYPEID 232 233 /* Special type cases, put in to allow the parser to distinguish different 234 legal basetypes. */ 235 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD 236 237 %token <sval> DOLLAR_VARIABLE 238 239 %token <opcode> ASSIGN_MODIFY 240 241 /* C++ */ 242 %token TRUEKEYWORD 243 %token FALSEKEYWORD 244 245 246 %left ',' 247 %left ABOVE_COMMA 248 %right '=' ASSIGN_MODIFY 249 %right '?' 250 %left OROR 251 %left ANDAND 252 %left '|' 253 %left '^' 254 %left '&' 255 %left EQUAL NOTEQUAL 256 %left '<' '>' LEQ GEQ 257 %left LSH RSH 258 %left '@' 259 %left '+' '-' 260 %left '*' '/' '%' 261 %right UNARY INCREMENT DECREMENT 262 %right ARROW ARROW_STAR '.' DOT_STAR '[' OBJC_LBRAC '(' 263 %token <ssym> BLOCKNAME 264 %token <bval> FILENAME 265 %type <bval> block 266 %left COLONCOLON 267 268 %token DOTDOTDOT 269 270 271 %% 272 273 start : exp1 274 | type_exp 275 ; 276 277 type_exp: type 278 { write_exp_elt_opcode(pstate, OP_TYPE); 279 write_exp_elt_type(pstate, $1); 280 write_exp_elt_opcode(pstate, OP_TYPE);} 281 | TYPEOF '(' exp ')' 282 { 283 write_exp_elt_opcode (pstate, OP_TYPEOF); 284 } 285 | TYPEOF '(' type ')' 286 { 287 write_exp_elt_opcode (pstate, OP_TYPE); 288 write_exp_elt_type (pstate, $3); 289 write_exp_elt_opcode (pstate, OP_TYPE); 290 } 291 | DECLTYPE '(' exp ')' 292 { 293 write_exp_elt_opcode (pstate, OP_DECLTYPE); 294 } 295 ; 296 297 /* Expressions, including the comma operator. */ 298 exp1 : exp 299 | exp1 ',' exp 300 { write_exp_elt_opcode (pstate, BINOP_COMMA); } 301 ; 302 303 /* Expressions, not including the comma operator. */ 304 exp : '*' exp %prec UNARY 305 { write_exp_elt_opcode (pstate, UNOP_IND); } 306 ; 307 308 exp : '&' exp %prec UNARY 309 { write_exp_elt_opcode (pstate, UNOP_ADDR); } 310 ; 311 312 exp : '-' exp %prec UNARY 313 { write_exp_elt_opcode (pstate, UNOP_NEG); } 314 ; 315 316 exp : '+' exp %prec UNARY 317 { write_exp_elt_opcode (pstate, UNOP_PLUS); } 318 ; 319 320 exp : '!' exp %prec UNARY 321 { write_exp_elt_opcode (pstate, UNOP_LOGICAL_NOT); } 322 ; 323 324 exp : '~' exp %prec UNARY 325 { write_exp_elt_opcode (pstate, UNOP_COMPLEMENT); } 326 ; 327 328 exp : INCREMENT exp %prec UNARY 329 { write_exp_elt_opcode (pstate, UNOP_PREINCREMENT); } 330 ; 331 332 exp : DECREMENT exp %prec UNARY 333 { write_exp_elt_opcode (pstate, UNOP_PREDECREMENT); } 334 ; 335 336 exp : exp INCREMENT %prec UNARY 337 { write_exp_elt_opcode (pstate, UNOP_POSTINCREMENT); } 338 ; 339 340 exp : exp DECREMENT %prec UNARY 341 { write_exp_elt_opcode (pstate, UNOP_POSTDECREMENT); } 342 ; 343 344 exp : TYPEID '(' exp ')' %prec UNARY 345 { write_exp_elt_opcode (pstate, OP_TYPEID); } 346 ; 347 348 exp : TYPEID '(' type_exp ')' %prec UNARY 349 { write_exp_elt_opcode (pstate, OP_TYPEID); } 350 ; 351 352 exp : SIZEOF exp %prec UNARY 353 { write_exp_elt_opcode (pstate, UNOP_SIZEOF); } 354 ; 355 356 exp : ALIGNOF '(' type_exp ')' %prec UNARY 357 { write_exp_elt_opcode (pstate, UNOP_ALIGNOF); } 358 ; 359 360 exp : exp ARROW field_name 361 { write_exp_elt_opcode (pstate, STRUCTOP_PTR); 362 write_exp_string (pstate, $3); 363 write_exp_elt_opcode (pstate, STRUCTOP_PTR); } 364 ; 365 366 exp : exp ARROW field_name COMPLETE 367 { mark_struct_expression (pstate); 368 write_exp_elt_opcode (pstate, STRUCTOP_PTR); 369 write_exp_string (pstate, $3); 370 write_exp_elt_opcode (pstate, STRUCTOP_PTR); } 371 ; 372 373 exp : exp ARROW COMPLETE 374 { struct stoken s; 375 mark_struct_expression (pstate); 376 write_exp_elt_opcode (pstate, STRUCTOP_PTR); 377 s.ptr = ""; 378 s.length = 0; 379 write_exp_string (pstate, s); 380 write_exp_elt_opcode (pstate, STRUCTOP_PTR); } 381 ; 382 383 exp : exp ARROW '~' name 384 { write_exp_elt_opcode (pstate, STRUCTOP_PTR); 385 write_destructor_name (pstate, $4); 386 write_exp_elt_opcode (pstate, STRUCTOP_PTR); } 387 ; 388 389 exp : exp ARROW '~' name COMPLETE 390 { mark_struct_expression (pstate); 391 write_exp_elt_opcode (pstate, STRUCTOP_PTR); 392 write_destructor_name (pstate, $4); 393 write_exp_elt_opcode (pstate, STRUCTOP_PTR); } 394 ; 395 396 exp : exp ARROW qualified_name 397 { /* exp->type::name becomes exp->*(&type::name) */ 398 /* Note: this doesn't work if name is a 399 static member! FIXME */ 400 write_exp_elt_opcode (pstate, UNOP_ADDR); 401 write_exp_elt_opcode (pstate, STRUCTOP_MPTR); } 402 ; 403 404 exp : exp ARROW_STAR exp 405 { write_exp_elt_opcode (pstate, STRUCTOP_MPTR); } 406 ; 407 408 exp : exp '.' field_name 409 { write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); 410 write_exp_string (pstate, $3); 411 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); } 412 ; 413 414 exp : exp '.' field_name COMPLETE 415 { mark_struct_expression (pstate); 416 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); 417 write_exp_string (pstate, $3); 418 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); } 419 ; 420 421 exp : exp '.' COMPLETE 422 { struct stoken s; 423 mark_struct_expression (pstate); 424 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); 425 s.ptr = ""; 426 s.length = 0; 427 write_exp_string (pstate, s); 428 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); } 429 ; 430 431 exp : exp '.' '~' name 432 { write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); 433 write_destructor_name (pstate, $4); 434 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); } 435 ; 436 437 exp : exp '.' '~' name COMPLETE 438 { mark_struct_expression (pstate); 439 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); 440 write_destructor_name (pstate, $4); 441 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); } 442 ; 443 444 exp : exp '.' qualified_name 445 { /* exp.type::name becomes exp.*(&type::name) */ 446 /* Note: this doesn't work if name is a 447 static member! FIXME */ 448 write_exp_elt_opcode (pstate, UNOP_ADDR); 449 write_exp_elt_opcode (pstate, STRUCTOP_MEMBER); } 450 ; 451 452 exp : exp DOT_STAR exp 453 { write_exp_elt_opcode (pstate, STRUCTOP_MEMBER); } 454 ; 455 456 exp : exp '[' exp1 ']' 457 { write_exp_elt_opcode (pstate, BINOP_SUBSCRIPT); } 458 ; 459 460 exp : exp OBJC_LBRAC exp1 ']' 461 { write_exp_elt_opcode (pstate, BINOP_SUBSCRIPT); } 462 ; 463 464 /* 465 * The rules below parse ObjC message calls of the form: 466 * '[' target selector {':' argument}* ']' 467 */ 468 469 exp : OBJC_LBRAC TYPENAME 470 { 471 CORE_ADDR theclass; 472 473 theclass = lookup_objc_class (parse_gdbarch (pstate), 474 copy_name ($2.stoken)); 475 if (theclass == 0) 476 error (_("%s is not an ObjC Class"), 477 copy_name ($2.stoken)); 478 write_exp_elt_opcode (pstate, OP_LONG); 479 write_exp_elt_type (pstate, 480 parse_type (pstate)->builtin_int); 481 write_exp_elt_longcst (pstate, (LONGEST) theclass); 482 write_exp_elt_opcode (pstate, OP_LONG); 483 start_msglist(); 484 } 485 msglist ']' 486 { write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL); 487 end_msglist (pstate); 488 write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL); 489 } 490 ; 491 492 exp : OBJC_LBRAC CLASSNAME 493 { 494 write_exp_elt_opcode (pstate, OP_LONG); 495 write_exp_elt_type (pstate, 496 parse_type (pstate)->builtin_int); 497 write_exp_elt_longcst (pstate, (LONGEST) $2.theclass); 498 write_exp_elt_opcode (pstate, OP_LONG); 499 start_msglist(); 500 } 501 msglist ']' 502 { write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL); 503 end_msglist (pstate); 504 write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL); 505 } 506 ; 507 508 exp : OBJC_LBRAC exp 509 { start_msglist(); } 510 msglist ']' 511 { write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL); 512 end_msglist (pstate); 513 write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL); 514 } 515 ; 516 517 msglist : name 518 { add_msglist(&$1, 0); } 519 | msgarglist 520 ; 521 522 msgarglist : msgarg 523 | msgarglist msgarg 524 ; 525 526 msgarg : name ':' exp 527 { add_msglist(&$1, 1); } 528 | ':' exp /* Unnamed arg. */ 529 { add_msglist(0, 1); } 530 | ',' exp /* Variable number of args. */ 531 { add_msglist(0, 0); } 532 ; 533 534 exp : exp '(' 535 /* This is to save the value of arglist_len 536 being accumulated by an outer function call. */ 537 { start_arglist (); } 538 arglist ')' %prec ARROW 539 { write_exp_elt_opcode (pstate, OP_FUNCALL); 540 write_exp_elt_longcst (pstate, 541 (LONGEST) end_arglist ()); 542 write_exp_elt_opcode (pstate, OP_FUNCALL); } 543 ; 544 545 /* This is here to disambiguate with the production for 546 "func()::static_var" further below, which uses 547 function_method_void. */ 548 exp : exp '(' ')' %prec ARROW 549 { start_arglist (); 550 write_exp_elt_opcode (pstate, OP_FUNCALL); 551 write_exp_elt_longcst (pstate, 552 (LONGEST) end_arglist ()); 553 write_exp_elt_opcode (pstate, OP_FUNCALL); } 554 ; 555 556 557 exp : UNKNOWN_CPP_NAME '(' 558 { 559 /* This could potentially be a an argument defined 560 lookup function (Koenig). */ 561 write_exp_elt_opcode (pstate, OP_ADL_FUNC); 562 write_exp_elt_block (pstate, 563 expression_context_block); 564 write_exp_elt_sym (pstate, 565 NULL); /* Placeholder. */ 566 write_exp_string (pstate, $1.stoken); 567 write_exp_elt_opcode (pstate, OP_ADL_FUNC); 568 569 /* This is to save the value of arglist_len 570 being accumulated by an outer function call. */ 571 572 start_arglist (); 573 } 574 arglist ')' %prec ARROW 575 { 576 write_exp_elt_opcode (pstate, OP_FUNCALL); 577 write_exp_elt_longcst (pstate, 578 (LONGEST) end_arglist ()); 579 write_exp_elt_opcode (pstate, OP_FUNCALL); 580 } 581 ; 582 583 lcurly : '{' 584 { start_arglist (); } 585 ; 586 587 arglist : 588 ; 589 590 arglist : exp 591 { arglist_len = 1; } 592 ; 593 594 arglist : arglist ',' exp %prec ABOVE_COMMA 595 { arglist_len++; } 596 ; 597 598 function_method: exp '(' parameter_typelist ')' const_or_volatile 599 { 600 std::vector<struct type *> *type_list = $3; 601 LONGEST len = type_list->size (); 602 603 write_exp_elt_opcode (pstate, TYPE_INSTANCE); 604 /* Save the const/volatile qualifiers as 605 recorded by the const_or_volatile 606 production's actions. */ 607 write_exp_elt_longcst (pstate, 608 follow_type_instance_flags ()); 609 write_exp_elt_longcst (pstate, len); 610 for (type *type_elt : *type_list) 611 write_exp_elt_type (pstate, type_elt); 612 write_exp_elt_longcst(pstate, len); 613 write_exp_elt_opcode (pstate, TYPE_INSTANCE); 614 } 615 ; 616 617 function_method_void: exp '(' ')' const_or_volatile 618 { write_exp_elt_opcode (pstate, TYPE_INSTANCE); 619 /* See above. */ 620 write_exp_elt_longcst (pstate, 621 follow_type_instance_flags ()); 622 write_exp_elt_longcst (pstate, 0); 623 write_exp_elt_longcst (pstate, 0); 624 write_exp_elt_opcode (pstate, TYPE_INSTANCE); 625 } 626 ; 627 628 exp : function_method 629 ; 630 631 /* Normally we must interpret "func()" as a function call, instead of 632 a type. The user needs to write func(void) to disambiguate. 633 However, in the "func()::static_var" case, there's no 634 ambiguity. */ 635 function_method_void_or_typelist: function_method 636 | function_method_void 637 ; 638 639 exp : function_method_void_or_typelist COLONCOLON name 640 { 641 write_exp_elt_opcode (pstate, OP_FUNC_STATIC_VAR); 642 write_exp_string (pstate, $3); 643 write_exp_elt_opcode (pstate, OP_FUNC_STATIC_VAR); 644 } 645 ; 646 647 rcurly : '}' 648 { $$ = end_arglist () - 1; } 649 ; 650 exp : lcurly arglist rcurly %prec ARROW 651 { write_exp_elt_opcode (pstate, OP_ARRAY); 652 write_exp_elt_longcst (pstate, (LONGEST) 0); 653 write_exp_elt_longcst (pstate, (LONGEST) $3); 654 write_exp_elt_opcode (pstate, OP_ARRAY); } 655 ; 656 657 exp : lcurly type_exp rcurly exp %prec UNARY 658 { write_exp_elt_opcode (pstate, UNOP_MEMVAL_TYPE); } 659 ; 660 661 exp : '(' type_exp ')' exp %prec UNARY 662 { write_exp_elt_opcode (pstate, UNOP_CAST_TYPE); } 663 ; 664 665 exp : '(' exp1 ')' 666 { } 667 ; 668 669 /* Binary operators in order of decreasing precedence. */ 670 671 exp : exp '@' exp 672 { write_exp_elt_opcode (pstate, BINOP_REPEAT); } 673 ; 674 675 exp : exp '*' exp 676 { write_exp_elt_opcode (pstate, BINOP_MUL); } 677 ; 678 679 exp : exp '/' exp 680 { write_exp_elt_opcode (pstate, BINOP_DIV); } 681 ; 682 683 exp : exp '%' exp 684 { write_exp_elt_opcode (pstate, BINOP_REM); } 685 ; 686 687 exp : exp '+' exp 688 { write_exp_elt_opcode (pstate, BINOP_ADD); } 689 ; 690 691 exp : exp '-' exp 692 { write_exp_elt_opcode (pstate, BINOP_SUB); } 693 ; 694 695 exp : exp LSH exp 696 { write_exp_elt_opcode (pstate, BINOP_LSH); } 697 ; 698 699 exp : exp RSH exp 700 { write_exp_elt_opcode (pstate, BINOP_RSH); } 701 ; 702 703 exp : exp EQUAL exp 704 { write_exp_elt_opcode (pstate, BINOP_EQUAL); } 705 ; 706 707 exp : exp NOTEQUAL exp 708 { write_exp_elt_opcode (pstate, BINOP_NOTEQUAL); } 709 ; 710 711 exp : exp LEQ exp 712 { write_exp_elt_opcode (pstate, BINOP_LEQ); } 713 ; 714 715 exp : exp GEQ exp 716 { write_exp_elt_opcode (pstate, BINOP_GEQ); } 717 ; 718 719 exp : exp '<' exp 720 { write_exp_elt_opcode (pstate, BINOP_LESS); } 721 ; 722 723 exp : exp '>' exp 724 { write_exp_elt_opcode (pstate, BINOP_GTR); } 725 ; 726 727 exp : exp '&' exp 728 { write_exp_elt_opcode (pstate, BINOP_BITWISE_AND); } 729 ; 730 731 exp : exp '^' exp 732 { write_exp_elt_opcode (pstate, BINOP_BITWISE_XOR); } 733 ; 734 735 exp : exp '|' exp 736 { write_exp_elt_opcode (pstate, BINOP_BITWISE_IOR); } 737 ; 738 739 exp : exp ANDAND exp 740 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_AND); } 741 ; 742 743 exp : exp OROR exp 744 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_OR); } 745 ; 746 747 exp : exp '?' exp ':' exp %prec '?' 748 { write_exp_elt_opcode (pstate, TERNOP_COND); } 749 ; 750 751 exp : exp '=' exp 752 { write_exp_elt_opcode (pstate, BINOP_ASSIGN); } 753 ; 754 755 exp : exp ASSIGN_MODIFY exp 756 { write_exp_elt_opcode (pstate, BINOP_ASSIGN_MODIFY); 757 write_exp_elt_opcode (pstate, $2); 758 write_exp_elt_opcode (pstate, 759 BINOP_ASSIGN_MODIFY); } 760 ; 761 762 exp : INT 763 { write_exp_elt_opcode (pstate, OP_LONG); 764 write_exp_elt_type (pstate, $1.type); 765 write_exp_elt_longcst (pstate, (LONGEST) ($1.val)); 766 write_exp_elt_opcode (pstate, OP_LONG); } 767 ; 768 769 exp : CHAR 770 { 771 struct stoken_vector vec; 772 vec.len = 1; 773 vec.tokens = &$1; 774 write_exp_string_vector (pstate, $1.type, &vec); 775 } 776 ; 777 778 exp : NAME_OR_INT 779 { YYSTYPE val; 780 parse_number (pstate, $1.stoken.ptr, 781 $1.stoken.length, 0, &val); 782 write_exp_elt_opcode (pstate, OP_LONG); 783 write_exp_elt_type (pstate, val.typed_val_int.type); 784 write_exp_elt_longcst (pstate, 785 (LONGEST) val.typed_val_int.val); 786 write_exp_elt_opcode (pstate, OP_LONG); 787 } 788 ; 789 790 791 exp : FLOAT 792 { write_exp_elt_opcode (pstate, OP_FLOAT); 793 write_exp_elt_type (pstate, $1.type); 794 write_exp_elt_floatcst (pstate, $1.val); 795 write_exp_elt_opcode (pstate, OP_FLOAT); } 796 ; 797 798 exp : variable 799 ; 800 801 exp : DOLLAR_VARIABLE 802 { 803 write_dollar_variable (pstate, $1); 804 } 805 ; 806 807 exp : SELECTOR '(' name ')' 808 { 809 write_exp_elt_opcode (pstate, OP_OBJC_SELECTOR); 810 write_exp_string (pstate, $3); 811 write_exp_elt_opcode (pstate, OP_OBJC_SELECTOR); } 812 ; 813 814 exp : SIZEOF '(' type ')' %prec UNARY 815 { struct type *type = $3; 816 write_exp_elt_opcode (pstate, OP_LONG); 817 write_exp_elt_type (pstate, lookup_signed_typename 818 (parse_language (pstate), 819 parse_gdbarch (pstate), 820 "int")); 821 type = check_typedef (type); 822 823 /* $5.3.3/2 of the C++ Standard (n3290 draft) 824 says of sizeof: "When applied to a reference 825 or a reference type, the result is the size of 826 the referenced type." */ 827 if (TYPE_IS_REFERENCE (type)) 828 type = check_typedef (TYPE_TARGET_TYPE (type)); 829 write_exp_elt_longcst (pstate, 830 (LONGEST) TYPE_LENGTH (type)); 831 write_exp_elt_opcode (pstate, OP_LONG); } 832 ; 833 834 exp : REINTERPRET_CAST '<' type_exp '>' '(' exp ')' %prec UNARY 835 { write_exp_elt_opcode (pstate, 836 UNOP_REINTERPRET_CAST); } 837 ; 838 839 exp : STATIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY 840 { write_exp_elt_opcode (pstate, UNOP_CAST_TYPE); } 841 ; 842 843 exp : DYNAMIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY 844 { write_exp_elt_opcode (pstate, UNOP_DYNAMIC_CAST); } 845 ; 846 847 exp : CONST_CAST '<' type_exp '>' '(' exp ')' %prec UNARY 848 { /* We could do more error checking here, but 849 it doesn't seem worthwhile. */ 850 write_exp_elt_opcode (pstate, UNOP_CAST_TYPE); } 851 ; 852 853 string_exp: 854 STRING 855 { 856 /* We copy the string here, and not in the 857 lexer, to guarantee that we do not leak a 858 string. Note that we follow the 859 NUL-termination convention of the 860 lexer. */ 861 struct typed_stoken *vec = XNEW (struct typed_stoken); 862 $$.len = 1; 863 $$.tokens = vec; 864 865 vec->type = $1.type; 866 vec->length = $1.length; 867 vec->ptr = (char *) malloc ($1.length + 1); 868 memcpy (vec->ptr, $1.ptr, $1.length + 1); 869 } 870 871 | string_exp STRING 872 { 873 /* Note that we NUL-terminate here, but just 874 for convenience. */ 875 char *p; 876 ++$$.len; 877 $$.tokens = XRESIZEVEC (struct typed_stoken, 878 $$.tokens, $$.len); 879 880 p = (char *) malloc ($2.length + 1); 881 memcpy (p, $2.ptr, $2.length + 1); 882 883 $$.tokens[$$.len - 1].type = $2.type; 884 $$.tokens[$$.len - 1].length = $2.length; 885 $$.tokens[$$.len - 1].ptr = p; 886 } 887 ; 888 889 exp : string_exp 890 { 891 int i; 892 c_string_type type = C_STRING; 893 894 for (i = 0; i < $1.len; ++i) 895 { 896 switch ($1.tokens[i].type) 897 { 898 case C_STRING: 899 break; 900 case C_WIDE_STRING: 901 case C_STRING_16: 902 case C_STRING_32: 903 if (type != C_STRING 904 && type != $1.tokens[i].type) 905 error (_("Undefined string concatenation.")); 906 type = (enum c_string_type_values) $1.tokens[i].type; 907 break; 908 default: 909 /* internal error */ 910 internal_error (__FILE__, __LINE__, 911 "unrecognized type in string concatenation"); 912 } 913 } 914 915 write_exp_string_vector (pstate, type, &$1); 916 for (i = 0; i < $1.len; ++i) 917 free ($1.tokens[i].ptr); 918 free ($1.tokens); 919 } 920 ; 921 922 exp : NSSTRING /* ObjC NextStep NSString constant 923 * of the form '@' '"' string '"'. 924 */ 925 { write_exp_elt_opcode (pstate, OP_OBJC_NSSTRING); 926 write_exp_string (pstate, $1); 927 write_exp_elt_opcode (pstate, OP_OBJC_NSSTRING); } 928 ; 929 930 /* C++. */ 931 exp : TRUEKEYWORD 932 { write_exp_elt_opcode (pstate, OP_LONG); 933 write_exp_elt_type (pstate, 934 parse_type (pstate)->builtin_bool); 935 write_exp_elt_longcst (pstate, (LONGEST) 1); 936 write_exp_elt_opcode (pstate, OP_LONG); } 937 ; 938 939 exp : FALSEKEYWORD 940 { write_exp_elt_opcode (pstate, OP_LONG); 941 write_exp_elt_type (pstate, 942 parse_type (pstate)->builtin_bool); 943 write_exp_elt_longcst (pstate, (LONGEST) 0); 944 write_exp_elt_opcode (pstate, OP_LONG); } 945 ; 946 947 /* end of C++. */ 948 949 block : BLOCKNAME 950 { 951 if ($1.sym.symbol) 952 $$ = SYMBOL_BLOCK_VALUE ($1.sym.symbol); 953 else 954 error (_("No file or function \"%s\"."), 955 copy_name ($1.stoken)); 956 } 957 | FILENAME 958 { 959 $$ = $1; 960 } 961 ; 962 963 block : block COLONCOLON name 964 { struct symbol *tem 965 = lookup_symbol (copy_name ($3), $1, 966 VAR_DOMAIN, NULL).symbol; 967 968 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK) 969 error (_("No function \"%s\" in specified context."), 970 copy_name ($3)); 971 $$ = SYMBOL_BLOCK_VALUE (tem); } 972 ; 973 974 variable: name_not_typename ENTRY 975 { struct symbol *sym = $1.sym.symbol; 976 977 if (sym == NULL || !SYMBOL_IS_ARGUMENT (sym) 978 || !symbol_read_needs_frame (sym)) 979 error (_("@entry can be used only for function " 980 "parameters, not for \"%s\""), 981 copy_name ($1.stoken)); 982 983 write_exp_elt_opcode (pstate, OP_VAR_ENTRY_VALUE); 984 write_exp_elt_sym (pstate, sym); 985 write_exp_elt_opcode (pstate, OP_VAR_ENTRY_VALUE); 986 } 987 ; 988 989 variable: block COLONCOLON name 990 { struct block_symbol sym 991 = lookup_symbol (copy_name ($3), $1, 992 VAR_DOMAIN, NULL); 993 994 if (sym.symbol == 0) 995 error (_("No symbol \"%s\" in specified context."), 996 copy_name ($3)); 997 if (symbol_read_needs_frame (sym.symbol)) 998 999 innermost_block.update (sym); 1000 1001 write_exp_elt_opcode (pstate, OP_VAR_VALUE); 1002 write_exp_elt_block (pstate, sym.block); 1003 write_exp_elt_sym (pstate, sym.symbol); 1004 write_exp_elt_opcode (pstate, OP_VAR_VALUE); } 1005 ; 1006 1007 qualified_name: TYPENAME COLONCOLON name 1008 { 1009 struct type *type = $1.type; 1010 type = check_typedef (type); 1011 if (!type_aggregate_p (type)) 1012 error (_("`%s' is not defined as an aggregate type."), 1013 TYPE_SAFE_NAME (type)); 1014 1015 write_exp_elt_opcode (pstate, OP_SCOPE); 1016 write_exp_elt_type (pstate, type); 1017 write_exp_string (pstate, $3); 1018 write_exp_elt_opcode (pstate, OP_SCOPE); 1019 } 1020 | TYPENAME COLONCOLON '~' name 1021 { 1022 struct type *type = $1.type; 1023 struct stoken tmp_token; 1024 char *buf; 1025 1026 type = check_typedef (type); 1027 if (!type_aggregate_p (type)) 1028 error (_("`%s' is not defined as an aggregate type."), 1029 TYPE_SAFE_NAME (type)); 1030 buf = (char *) alloca ($4.length + 2); 1031 tmp_token.ptr = buf; 1032 tmp_token.length = $4.length + 1; 1033 buf[0] = '~'; 1034 memcpy (buf+1, $4.ptr, $4.length); 1035 buf[tmp_token.length] = 0; 1036 1037 /* Check for valid destructor name. */ 1038 destructor_name_p (tmp_token.ptr, $1.type); 1039 write_exp_elt_opcode (pstate, OP_SCOPE); 1040 write_exp_elt_type (pstate, type); 1041 write_exp_string (pstate, tmp_token); 1042 write_exp_elt_opcode (pstate, OP_SCOPE); 1043 } 1044 | TYPENAME COLONCOLON name COLONCOLON name 1045 { 1046 char *copy = copy_name ($3); 1047 error (_("No type \"%s\" within class " 1048 "or namespace \"%s\"."), 1049 copy, TYPE_SAFE_NAME ($1.type)); 1050 } 1051 ; 1052 1053 variable: qualified_name 1054 | COLONCOLON name_not_typename 1055 { 1056 char *name = copy_name ($2.stoken); 1057 struct symbol *sym; 1058 struct bound_minimal_symbol msymbol; 1059 1060 sym 1061 = lookup_symbol (name, (const struct block *) NULL, 1062 VAR_DOMAIN, NULL).symbol; 1063 if (sym) 1064 { 1065 write_exp_elt_opcode (pstate, OP_VAR_VALUE); 1066 write_exp_elt_block (pstate, NULL); 1067 write_exp_elt_sym (pstate, sym); 1068 write_exp_elt_opcode (pstate, OP_VAR_VALUE); 1069 break; 1070 } 1071 1072 msymbol = lookup_bound_minimal_symbol (name); 1073 if (msymbol.minsym != NULL) 1074 write_exp_msymbol (pstate, msymbol); 1075 else if (!have_full_symbols () && !have_partial_symbols ()) 1076 error (_("No symbol table is loaded. Use the \"file\" command.")); 1077 else 1078 error (_("No symbol \"%s\" in current context."), name); 1079 } 1080 ; 1081 1082 variable: name_not_typename 1083 { struct block_symbol sym = $1.sym; 1084 1085 if (sym.symbol) 1086 { 1087 if (symbol_read_needs_frame (sym.symbol)) 1088 innermost_block.update (sym); 1089 1090 /* If we found a function, see if it's 1091 an ifunc resolver that has the same 1092 address as the ifunc symbol itself. 1093 If so, prefer the ifunc symbol. */ 1094 1095 bound_minimal_symbol resolver 1096 = find_gnu_ifunc (sym.symbol); 1097 if (resolver.minsym != NULL) 1098 write_exp_msymbol (pstate, resolver); 1099 else 1100 { 1101 write_exp_elt_opcode (pstate, OP_VAR_VALUE); 1102 write_exp_elt_block (pstate, sym.block); 1103 write_exp_elt_sym (pstate, sym.symbol); 1104 write_exp_elt_opcode (pstate, OP_VAR_VALUE); 1105 } 1106 } 1107 else if ($1.is_a_field_of_this) 1108 { 1109 /* C++: it hangs off of `this'. Must 1110 not inadvertently convert from a method call 1111 to data ref. */ 1112 innermost_block.update (sym); 1113 write_exp_elt_opcode (pstate, OP_THIS); 1114 write_exp_elt_opcode (pstate, OP_THIS); 1115 write_exp_elt_opcode (pstate, STRUCTOP_PTR); 1116 write_exp_string (pstate, $1.stoken); 1117 write_exp_elt_opcode (pstate, STRUCTOP_PTR); 1118 } 1119 else 1120 { 1121 char *arg = copy_name ($1.stoken); 1122 1123 bound_minimal_symbol msymbol 1124 = lookup_bound_minimal_symbol (arg); 1125 if (msymbol.minsym == NULL) 1126 { 1127 if (!have_full_symbols () && !have_partial_symbols ()) 1128 error (_("No symbol table is loaded. Use the \"file\" command.")); 1129 else 1130 error (_("No symbol \"%s\" in current context."), 1131 copy_name ($1.stoken)); 1132 } 1133 1134 /* This minsym might be an alias for 1135 another function. See if we can find 1136 the debug symbol for the target, and 1137 if so, use it instead, since it has 1138 return type / prototype info. This 1139 is important for example for "p 1140 *__errno_location()". */ 1141 symbol *alias_target 1142 = ((msymbol.minsym->type != mst_text_gnu_ifunc 1143 && msymbol.minsym->type != mst_data_gnu_ifunc) 1144 ? find_function_alias_target (msymbol) 1145 : NULL); 1146 if (alias_target != NULL) 1147 { 1148 write_exp_elt_opcode (pstate, OP_VAR_VALUE); 1149 write_exp_elt_block 1150 (pstate, SYMBOL_BLOCK_VALUE (alias_target)); 1151 write_exp_elt_sym (pstate, alias_target); 1152 write_exp_elt_opcode (pstate, OP_VAR_VALUE); 1153 } 1154 else 1155 write_exp_msymbol (pstate, msymbol); 1156 } 1157 } 1158 ; 1159 1160 space_identifier : '@' NAME 1161 { insert_type_address_space (pstate, copy_name ($2.stoken)); } 1162 ; 1163 1164 const_or_volatile: const_or_volatile_noopt 1165 | 1166 ; 1167 1168 cv_with_space_id : const_or_volatile space_identifier const_or_volatile 1169 ; 1170 1171 const_or_volatile_or_space_identifier_noopt: cv_with_space_id 1172 | const_or_volatile_noopt 1173 ; 1174 1175 const_or_volatile_or_space_identifier: 1176 const_or_volatile_or_space_identifier_noopt 1177 | 1178 ; 1179 1180 ptr_operator: 1181 ptr_operator '*' 1182 { insert_type (tp_pointer); } 1183 const_or_volatile_or_space_identifier 1184 | '*' 1185 { insert_type (tp_pointer); } 1186 const_or_volatile_or_space_identifier 1187 | '&' 1188 { insert_type (tp_reference); } 1189 | '&' ptr_operator 1190 { insert_type (tp_reference); } 1191 | ANDAND 1192 { insert_type (tp_rvalue_reference); } 1193 | ANDAND ptr_operator 1194 { insert_type (tp_rvalue_reference); } 1195 ; 1196 1197 ptr_operator_ts: ptr_operator 1198 { 1199 $$ = get_type_stack (); 1200 cpstate->type_stacks.emplace_back ($$); 1201 } 1202 ; 1203 1204 abs_decl: ptr_operator_ts direct_abs_decl 1205 { $$ = append_type_stack ($2, $1); } 1206 | ptr_operator_ts 1207 | direct_abs_decl 1208 ; 1209 1210 direct_abs_decl: '(' abs_decl ')' 1211 { $$ = $2; } 1212 | direct_abs_decl array_mod 1213 { 1214 push_type_stack ($1); 1215 push_type_int ($2); 1216 push_type (tp_array); 1217 $$ = get_type_stack (); 1218 cpstate->type_stacks.emplace_back ($$); 1219 } 1220 | array_mod 1221 { 1222 push_type_int ($1); 1223 push_type (tp_array); 1224 $$ = get_type_stack (); 1225 cpstate->type_stacks.emplace_back ($$); 1226 } 1227 1228 | direct_abs_decl func_mod 1229 { 1230 push_type_stack ($1); 1231 push_typelist ($2); 1232 $$ = get_type_stack (); 1233 cpstate->type_stacks.emplace_back ($$); 1234 } 1235 | func_mod 1236 { 1237 push_typelist ($1); 1238 $$ = get_type_stack (); 1239 cpstate->type_stacks.emplace_back ($$); 1240 } 1241 ; 1242 1243 array_mod: '[' ']' 1244 { $$ = -1; } 1245 | OBJC_LBRAC ']' 1246 { $$ = -1; } 1247 | '[' INT ']' 1248 { $$ = $2.val; } 1249 | OBJC_LBRAC INT ']' 1250 { $$ = $2.val; } 1251 ; 1252 1253 func_mod: '(' ')' 1254 { 1255 $$ = new std::vector<struct type *>; 1256 cpstate->type_lists.emplace_back ($$); 1257 } 1258 | '(' parameter_typelist ')' 1259 { $$ = $2; } 1260 ; 1261 1262 /* We used to try to recognize pointer to member types here, but 1263 that didn't work (shift/reduce conflicts meant that these rules never 1264 got executed). The problem is that 1265 int (foo::bar::baz::bizzle) 1266 is a function type but 1267 int (foo::bar::baz::bizzle::*) 1268 is a pointer to member type. Stroustrup loses again! */ 1269 1270 type : ptype 1271 ; 1272 1273 /* Implements (approximately): (type-qualifier)* type-specifier. 1274 1275 When type-specifier is only ever a single word, like 'float' then these 1276 arrive as pre-built TYPENAME tokens thanks to the classify_name 1277 function. However, when a type-specifier can contain multiple words, 1278 for example 'double' can appear as just 'double' or 'long double', and 1279 similarly 'long' can appear as just 'long' or in 'long double', then 1280 these type-specifiers are parsed into their own tokens in the function 1281 lex_one_token and the ident_tokens array. These separate tokens are all 1282 recognised here. */ 1283 typebase 1284 : TYPENAME 1285 { $$ = $1.type; } 1286 | INT_KEYWORD 1287 { $$ = lookup_signed_typename (parse_language (pstate), 1288 parse_gdbarch (pstate), 1289 "int"); } 1290 | LONG 1291 { $$ = lookup_signed_typename (parse_language (pstate), 1292 parse_gdbarch (pstate), 1293 "long"); } 1294 | SHORT 1295 { $$ = lookup_signed_typename (parse_language (pstate), 1296 parse_gdbarch (pstate), 1297 "short"); } 1298 | LONG INT_KEYWORD 1299 { $$ = lookup_signed_typename (parse_language (pstate), 1300 parse_gdbarch (pstate), 1301 "long"); } 1302 | LONG SIGNED_KEYWORD INT_KEYWORD 1303 { $$ = lookup_signed_typename (parse_language (pstate), 1304 parse_gdbarch (pstate), 1305 "long"); } 1306 | LONG SIGNED_KEYWORD 1307 { $$ = lookup_signed_typename (parse_language (pstate), 1308 parse_gdbarch (pstate), 1309 "long"); } 1310 | SIGNED_KEYWORD LONG INT_KEYWORD 1311 { $$ = lookup_signed_typename (parse_language (pstate), 1312 parse_gdbarch (pstate), 1313 "long"); } 1314 | UNSIGNED LONG INT_KEYWORD 1315 { $$ = lookup_unsigned_typename (parse_language (pstate), 1316 parse_gdbarch (pstate), 1317 "long"); } 1318 | LONG UNSIGNED INT_KEYWORD 1319 { $$ = lookup_unsigned_typename (parse_language (pstate), 1320 parse_gdbarch (pstate), 1321 "long"); } 1322 | LONG UNSIGNED 1323 { $$ = lookup_unsigned_typename (parse_language (pstate), 1324 parse_gdbarch (pstate), 1325 "long"); } 1326 | LONG LONG 1327 { $$ = lookup_signed_typename (parse_language (pstate), 1328 parse_gdbarch (pstate), 1329 "long long"); } 1330 | LONG LONG INT_KEYWORD 1331 { $$ = lookup_signed_typename (parse_language (pstate), 1332 parse_gdbarch (pstate), 1333 "long long"); } 1334 | LONG LONG SIGNED_KEYWORD INT_KEYWORD 1335 { $$ = lookup_signed_typename (parse_language (pstate), 1336 parse_gdbarch (pstate), 1337 "long long"); } 1338 | LONG LONG SIGNED_KEYWORD 1339 { $$ = lookup_signed_typename (parse_language (pstate), 1340 parse_gdbarch (pstate), 1341 "long long"); } 1342 | SIGNED_KEYWORD LONG LONG 1343 { $$ = lookup_signed_typename (parse_language (pstate), 1344 parse_gdbarch (pstate), 1345 "long long"); } 1346 | SIGNED_KEYWORD LONG LONG INT_KEYWORD 1347 { $$ = lookup_signed_typename (parse_language (pstate), 1348 parse_gdbarch (pstate), 1349 "long long"); } 1350 | UNSIGNED LONG LONG 1351 { $$ = lookup_unsigned_typename (parse_language (pstate), 1352 parse_gdbarch (pstate), 1353 "long long"); } 1354 | UNSIGNED LONG LONG INT_KEYWORD 1355 { $$ = lookup_unsigned_typename (parse_language (pstate), 1356 parse_gdbarch (pstate), 1357 "long long"); } 1358 | LONG LONG UNSIGNED 1359 { $$ = lookup_unsigned_typename (parse_language (pstate), 1360 parse_gdbarch (pstate), 1361 "long long"); } 1362 | LONG LONG UNSIGNED INT_KEYWORD 1363 { $$ = lookup_unsigned_typename (parse_language (pstate), 1364 parse_gdbarch (pstate), 1365 "long long"); } 1366 | SHORT INT_KEYWORD 1367 { $$ = lookup_signed_typename (parse_language (pstate), 1368 parse_gdbarch (pstate), 1369 "short"); } 1370 | SHORT SIGNED_KEYWORD INT_KEYWORD 1371 { $$ = lookup_signed_typename (parse_language (pstate), 1372 parse_gdbarch (pstate), 1373 "short"); } 1374 | SHORT SIGNED_KEYWORD 1375 { $$ = lookup_signed_typename (parse_language (pstate), 1376 parse_gdbarch (pstate), 1377 "short"); } 1378 | UNSIGNED SHORT INT_KEYWORD 1379 { $$ = lookup_unsigned_typename (parse_language (pstate), 1380 parse_gdbarch (pstate), 1381 "short"); } 1382 | SHORT UNSIGNED 1383 { $$ = lookup_unsigned_typename (parse_language (pstate), 1384 parse_gdbarch (pstate), 1385 "short"); } 1386 | SHORT UNSIGNED INT_KEYWORD 1387 { $$ = lookup_unsigned_typename (parse_language (pstate), 1388 parse_gdbarch (pstate), 1389 "short"); } 1390 | DOUBLE_KEYWORD 1391 { $$ = lookup_typename (parse_language (pstate), 1392 parse_gdbarch (pstate), 1393 "double", 1394 (struct block *) NULL, 1395 0); } 1396 | LONG DOUBLE_KEYWORD 1397 { $$ = lookup_typename (parse_language (pstate), 1398 parse_gdbarch (pstate), 1399 "long double", 1400 (struct block *) NULL, 1401 0); } 1402 | STRUCT name 1403 { $$ = lookup_struct (copy_name ($2), 1404 expression_context_block); } 1405 | STRUCT COMPLETE 1406 { 1407 mark_completion_tag (TYPE_CODE_STRUCT, "", 0); 1408 $$ = NULL; 1409 } 1410 | STRUCT name COMPLETE 1411 { 1412 mark_completion_tag (TYPE_CODE_STRUCT, $2.ptr, 1413 $2.length); 1414 $$ = NULL; 1415 } 1416 | CLASS name 1417 { $$ = lookup_struct (copy_name ($2), 1418 expression_context_block); } 1419 | CLASS COMPLETE 1420 { 1421 mark_completion_tag (TYPE_CODE_STRUCT, "", 0); 1422 $$ = NULL; 1423 } 1424 | CLASS name COMPLETE 1425 { 1426 mark_completion_tag (TYPE_CODE_STRUCT, $2.ptr, 1427 $2.length); 1428 $$ = NULL; 1429 } 1430 | UNION name 1431 { $$ = lookup_union (copy_name ($2), 1432 expression_context_block); } 1433 | UNION COMPLETE 1434 { 1435 mark_completion_tag (TYPE_CODE_UNION, "", 0); 1436 $$ = NULL; 1437 } 1438 | UNION name COMPLETE 1439 { 1440 mark_completion_tag (TYPE_CODE_UNION, $2.ptr, 1441 $2.length); 1442 $$ = NULL; 1443 } 1444 | ENUM name 1445 { $$ = lookup_enum (copy_name ($2), 1446 expression_context_block); } 1447 | ENUM COMPLETE 1448 { 1449 mark_completion_tag (TYPE_CODE_ENUM, "", 0); 1450 $$ = NULL; 1451 } 1452 | ENUM name COMPLETE 1453 { 1454 mark_completion_tag (TYPE_CODE_ENUM, $2.ptr, 1455 $2.length); 1456 $$ = NULL; 1457 } 1458 | UNSIGNED type_name 1459 { $$ = lookup_unsigned_typename (parse_language (pstate), 1460 parse_gdbarch (pstate), 1461 TYPE_NAME($2.type)); } 1462 | UNSIGNED 1463 { $$ = lookup_unsigned_typename (parse_language (pstate), 1464 parse_gdbarch (pstate), 1465 "int"); } 1466 | SIGNED_KEYWORD type_name 1467 { $$ = lookup_signed_typename (parse_language (pstate), 1468 parse_gdbarch (pstate), 1469 TYPE_NAME($2.type)); } 1470 | SIGNED_KEYWORD 1471 { $$ = lookup_signed_typename (parse_language (pstate), 1472 parse_gdbarch (pstate), 1473 "int"); } 1474 /* It appears that this rule for templates is never 1475 reduced; template recognition happens by lookahead 1476 in the token processing code in yylex. */ 1477 | TEMPLATE name '<' type '>' 1478 { $$ = lookup_template_type(copy_name($2), $4, 1479 expression_context_block); 1480 } 1481 | const_or_volatile_or_space_identifier_noopt typebase 1482 { $$ = follow_types ($2); } 1483 | typebase const_or_volatile_or_space_identifier_noopt 1484 { $$ = follow_types ($1); } 1485 ; 1486 1487 type_name: TYPENAME 1488 | INT_KEYWORD 1489 { 1490 $$.stoken.ptr = "int"; 1491 $$.stoken.length = 3; 1492 $$.type = lookup_signed_typename (parse_language (pstate), 1493 parse_gdbarch (pstate), 1494 "int"); 1495 } 1496 | LONG 1497 { 1498 $$.stoken.ptr = "long"; 1499 $$.stoken.length = 4; 1500 $$.type = lookup_signed_typename (parse_language (pstate), 1501 parse_gdbarch (pstate), 1502 "long"); 1503 } 1504 | SHORT 1505 { 1506 $$.stoken.ptr = "short"; 1507 $$.stoken.length = 5; 1508 $$.type = lookup_signed_typename (parse_language (pstate), 1509 parse_gdbarch (pstate), 1510 "short"); 1511 } 1512 ; 1513 1514 parameter_typelist: 1515 nonempty_typelist 1516 { check_parameter_typelist ($1); } 1517 | nonempty_typelist ',' DOTDOTDOT 1518 { 1519 $1->push_back (NULL); 1520 check_parameter_typelist ($1); 1521 $$ = $1; 1522 } 1523 ; 1524 1525 nonempty_typelist 1526 : type 1527 { 1528 std::vector<struct type *> *typelist 1529 = new std::vector<struct type *>; 1530 cpstate->type_lists.emplace_back (typelist); 1531 1532 typelist->push_back ($1); 1533 $$ = typelist; 1534 } 1535 | nonempty_typelist ',' type 1536 { 1537 $1->push_back ($3); 1538 $$ = $1; 1539 } 1540 ; 1541 1542 ptype : typebase 1543 | ptype abs_decl 1544 { 1545 push_type_stack ($2); 1546 $$ = follow_types ($1); 1547 } 1548 ; 1549 1550 conversion_type_id: typebase conversion_declarator 1551 { $$ = follow_types ($1); } 1552 ; 1553 1554 conversion_declarator: /* Nothing. */ 1555 | ptr_operator conversion_declarator 1556 ; 1557 1558 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD 1559 | VOLATILE_KEYWORD CONST_KEYWORD 1560 ; 1561 1562 const_or_volatile_noopt: const_and_volatile 1563 { insert_type (tp_const); 1564 insert_type (tp_volatile); 1565 } 1566 | CONST_KEYWORD 1567 { insert_type (tp_const); } 1568 | VOLATILE_KEYWORD 1569 { insert_type (tp_volatile); } 1570 ; 1571 1572 oper: OPERATOR NEW 1573 { $$ = operator_stoken (" new"); } 1574 | OPERATOR DELETE 1575 { $$ = operator_stoken (" delete"); } 1576 | OPERATOR NEW '[' ']' 1577 { $$ = operator_stoken (" new[]"); } 1578 | OPERATOR DELETE '[' ']' 1579 { $$ = operator_stoken (" delete[]"); } 1580 | OPERATOR NEW OBJC_LBRAC ']' 1581 { $$ = operator_stoken (" new[]"); } 1582 | OPERATOR DELETE OBJC_LBRAC ']' 1583 { $$ = operator_stoken (" delete[]"); } 1584 | OPERATOR '+' 1585 { $$ = operator_stoken ("+"); } 1586 | OPERATOR '-' 1587 { $$ = operator_stoken ("-"); } 1588 | OPERATOR '*' 1589 { $$ = operator_stoken ("*"); } 1590 | OPERATOR '/' 1591 { $$ = operator_stoken ("/"); } 1592 | OPERATOR '%' 1593 { $$ = operator_stoken ("%"); } 1594 | OPERATOR '^' 1595 { $$ = operator_stoken ("^"); } 1596 | OPERATOR '&' 1597 { $$ = operator_stoken ("&"); } 1598 | OPERATOR '|' 1599 { $$ = operator_stoken ("|"); } 1600 | OPERATOR '~' 1601 { $$ = operator_stoken ("~"); } 1602 | OPERATOR '!' 1603 { $$ = operator_stoken ("!"); } 1604 | OPERATOR '=' 1605 { $$ = operator_stoken ("="); } 1606 | OPERATOR '<' 1607 { $$ = operator_stoken ("<"); } 1608 | OPERATOR '>' 1609 { $$ = operator_stoken (">"); } 1610 | OPERATOR ASSIGN_MODIFY 1611 { const char *op = " unknown"; 1612 switch ($2) 1613 { 1614 case BINOP_RSH: 1615 op = ">>="; 1616 break; 1617 case BINOP_LSH: 1618 op = "<<="; 1619 break; 1620 case BINOP_ADD: 1621 op = "+="; 1622 break; 1623 case BINOP_SUB: 1624 op = "-="; 1625 break; 1626 case BINOP_MUL: 1627 op = "*="; 1628 break; 1629 case BINOP_DIV: 1630 op = "/="; 1631 break; 1632 case BINOP_REM: 1633 op = "%="; 1634 break; 1635 case BINOP_BITWISE_IOR: 1636 op = "|="; 1637 break; 1638 case BINOP_BITWISE_AND: 1639 op = "&="; 1640 break; 1641 case BINOP_BITWISE_XOR: 1642 op = "^="; 1643 break; 1644 default: 1645 break; 1646 } 1647 1648 $$ = operator_stoken (op); 1649 } 1650 | OPERATOR LSH 1651 { $$ = operator_stoken ("<<"); } 1652 | OPERATOR RSH 1653 { $$ = operator_stoken (">>"); } 1654 | OPERATOR EQUAL 1655 { $$ = operator_stoken ("=="); } 1656 | OPERATOR NOTEQUAL 1657 { $$ = operator_stoken ("!="); } 1658 | OPERATOR LEQ 1659 { $$ = operator_stoken ("<="); } 1660 | OPERATOR GEQ 1661 { $$ = operator_stoken (">="); } 1662 | OPERATOR ANDAND 1663 { $$ = operator_stoken ("&&"); } 1664 | OPERATOR OROR 1665 { $$ = operator_stoken ("||"); } 1666 | OPERATOR INCREMENT 1667 { $$ = operator_stoken ("++"); } 1668 | OPERATOR DECREMENT 1669 { $$ = operator_stoken ("--"); } 1670 | OPERATOR ',' 1671 { $$ = operator_stoken (","); } 1672 | OPERATOR ARROW_STAR 1673 { $$ = operator_stoken ("->*"); } 1674 | OPERATOR ARROW 1675 { $$ = operator_stoken ("->"); } 1676 | OPERATOR '(' ')' 1677 { $$ = operator_stoken ("()"); } 1678 | OPERATOR '[' ']' 1679 { $$ = operator_stoken ("[]"); } 1680 | OPERATOR OBJC_LBRAC ']' 1681 { $$ = operator_stoken ("[]"); } 1682 | OPERATOR conversion_type_id 1683 { string_file buf; 1684 1685 c_print_type ($2, NULL, &buf, -1, 0, 1686 &type_print_raw_options); 1687 1688 /* This also needs canonicalization. */ 1689 std::string canon 1690 = cp_canonicalize_string (buf.c_str ()); 1691 if (canon.empty ()) 1692 canon = std::move (buf.string ()); 1693 $$ = operator_stoken ((" " + canon).c_str ()); 1694 } 1695 ; 1696 1697 /* This rule exists in order to allow some tokens that would not normally 1698 match the 'name' rule to appear as fields within a struct. The example 1699 that initially motivated this was the RISC-V target which models the 1700 floating point registers as a union with fields called 'float' and 1701 'double'. The 'float' string becomes a TYPENAME token and can appear 1702 anywhere a 'name' can, however 'double' is its own token, 1703 DOUBLE_KEYWORD, and doesn't match the 'name' rule.*/ 1704 field_name 1705 : name 1706 | DOUBLE_KEYWORD { $$ = typename_stoken ("double"); } 1707 | INT_KEYWORD { $$ = typename_stoken ("int"); } 1708 | LONG { $$ = typename_stoken ("long"); } 1709 | SHORT { $$ = typename_stoken ("short"); } 1710 | SIGNED_KEYWORD { $$ = typename_stoken ("signed"); } 1711 | UNSIGNED { $$ = typename_stoken ("unsigned"); } 1712 ; 1713 1714 name : NAME { $$ = $1.stoken; } 1715 | BLOCKNAME { $$ = $1.stoken; } 1716 | TYPENAME { $$ = $1.stoken; } 1717 | NAME_OR_INT { $$ = $1.stoken; } 1718 | UNKNOWN_CPP_NAME { $$ = $1.stoken; } 1719 | oper { $$ = $1; } 1720 ; 1721 1722 name_not_typename : NAME 1723 | BLOCKNAME 1724 /* These would be useful if name_not_typename was useful, but it is just 1725 a fake for "variable", so these cause reduce/reduce conflicts because 1726 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable, 1727 =exp) or just an exp. If name_not_typename was ever used in an lvalue 1728 context where only a name could occur, this might be useful. 1729 | NAME_OR_INT 1730 */ 1731 | oper 1732 { 1733 struct field_of_this_result is_a_field_of_this; 1734 1735 $$.stoken = $1; 1736 $$.sym = lookup_symbol ($1.ptr, 1737 expression_context_block, 1738 VAR_DOMAIN, 1739 &is_a_field_of_this); 1740 $$.is_a_field_of_this 1741 = is_a_field_of_this.type != NULL; 1742 } 1743 | UNKNOWN_CPP_NAME 1744 ; 1745 1746 %% 1747 1748 /* Like write_exp_string, but prepends a '~'. */ 1749 1750 static void 1751 write_destructor_name (struct parser_state *par_state, struct stoken token) 1752 { 1753 char *copy = (char *) alloca (token.length + 1); 1754 1755 copy[0] = '~'; 1756 memcpy (©[1], token.ptr, token.length); 1757 1758 token.ptr = copy; 1759 ++token.length; 1760 1761 write_exp_string (par_state, token); 1762 } 1763 1764 /* Returns a stoken of the operator name given by OP (which does not 1765 include the string "operator"). */ 1766 1767 static struct stoken 1768 operator_stoken (const char *op) 1769 { 1770 struct stoken st = { NULL, 0 }; 1771 char *buf; 1772 1773 st.length = CP_OPERATOR_LEN + strlen (op); 1774 buf = (char *) malloc (st.length + 1); 1775 strcpy (buf, CP_OPERATOR_STR); 1776 strcat (buf, op); 1777 st.ptr = buf; 1778 1779 /* The toplevel (c_parse) will free the memory allocated here. */ 1780 cpstate->strings.emplace_back (buf); 1781 return st; 1782 }; 1783 1784 /* Returns a stoken of the type named TYPE. */ 1785 1786 static struct stoken 1787 typename_stoken (const char *type) 1788 { 1789 struct stoken st = { type, 0 }; 1790 st.length = strlen (type); 1791 return st; 1792 }; 1793 1794 /* Return true if the type is aggregate-like. */ 1795 1796 static int 1797 type_aggregate_p (struct type *type) 1798 { 1799 return (TYPE_CODE (type) == TYPE_CODE_STRUCT 1800 || TYPE_CODE (type) == TYPE_CODE_UNION 1801 || TYPE_CODE (type) == TYPE_CODE_NAMESPACE 1802 || (TYPE_CODE (type) == TYPE_CODE_ENUM 1803 && TYPE_DECLARED_CLASS (type))); 1804 } 1805 1806 /* Validate a parameter typelist. */ 1807 1808 static void 1809 check_parameter_typelist (std::vector<struct type *> *params) 1810 { 1811 struct type *type; 1812 int ix; 1813 1814 for (ix = 0; ix < params->size (); ++ix) 1815 { 1816 type = (*params)[ix]; 1817 if (type != NULL && TYPE_CODE (check_typedef (type)) == TYPE_CODE_VOID) 1818 { 1819 if (ix == 0) 1820 { 1821 if (params->size () == 1) 1822 { 1823 /* Ok. */ 1824 break; 1825 } 1826 error (_("parameter types following 'void'")); 1827 } 1828 else 1829 error (_("'void' invalid as parameter type")); 1830 } 1831 } 1832 } 1833 1834 /* Take care of parsing a number (anything that starts with a digit). 1835 Set yylval and return the token type; update lexptr. 1836 LEN is the number of characters in it. */ 1837 1838 /*** Needs some error checking for the float case ***/ 1839 1840 static int 1841 parse_number (struct parser_state *par_state, 1842 const char *buf, int len, int parsed_float, YYSTYPE *putithere) 1843 { 1844 ULONGEST n = 0; 1845 ULONGEST prevn = 0; 1846 ULONGEST un; 1847 1848 int i = 0; 1849 int c; 1850 int base = input_radix; 1851 int unsigned_p = 0; 1852 1853 /* Number of "L" suffixes encountered. */ 1854 int long_p = 0; 1855 1856 /* We have found a "L" or "U" suffix. */ 1857 int found_suffix = 0; 1858 1859 ULONGEST high_bit; 1860 struct type *signed_type; 1861 struct type *unsigned_type; 1862 char *p; 1863 1864 p = (char *) alloca (len); 1865 memcpy (p, buf, len); 1866 1867 if (parsed_float) 1868 { 1869 /* Handle suffixes for decimal floating-point: "df", "dd" or "dl". */ 1870 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f') 1871 { 1872 putithere->typed_val_float.type 1873 = parse_type (par_state)->builtin_decfloat; 1874 len -= 2; 1875 } 1876 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd') 1877 { 1878 putithere->typed_val_float.type 1879 = parse_type (par_state)->builtin_decdouble; 1880 len -= 2; 1881 } 1882 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l') 1883 { 1884 putithere->typed_val_float.type 1885 = parse_type (par_state)->builtin_declong; 1886 len -= 2; 1887 } 1888 /* Handle suffixes: 'f' for float, 'l' for long double. */ 1889 else if (len >= 1 && TOLOWER (p[len - 1]) == 'f') 1890 { 1891 putithere->typed_val_float.type 1892 = parse_type (par_state)->builtin_float; 1893 len -= 1; 1894 } 1895 else if (len >= 1 && TOLOWER (p[len - 1]) == 'l') 1896 { 1897 putithere->typed_val_float.type 1898 = parse_type (par_state)->builtin_long_double; 1899 len -= 1; 1900 } 1901 /* Default type for floating-point literals is double. */ 1902 else 1903 { 1904 putithere->typed_val_float.type 1905 = parse_type (par_state)->builtin_double; 1906 } 1907 1908 if (!parse_float (p, len, 1909 putithere->typed_val_float.type, 1910 putithere->typed_val_float.val)) 1911 return ERROR; 1912 return FLOAT; 1913 } 1914 1915 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */ 1916 if (p[0] == '0' && len > 1) 1917 switch (p[1]) 1918 { 1919 case 'x': 1920 case 'X': 1921 if (len >= 3) 1922 { 1923 p += 2; 1924 base = 16; 1925 len -= 2; 1926 } 1927 break; 1928 1929 case 'b': 1930 case 'B': 1931 if (len >= 3) 1932 { 1933 p += 2; 1934 base = 2; 1935 len -= 2; 1936 } 1937 break; 1938 1939 case 't': 1940 case 'T': 1941 case 'd': 1942 case 'D': 1943 if (len >= 3) 1944 { 1945 p += 2; 1946 base = 10; 1947 len -= 2; 1948 } 1949 break; 1950 1951 default: 1952 base = 8; 1953 break; 1954 } 1955 1956 while (len-- > 0) 1957 { 1958 c = *p++; 1959 if (c >= 'A' && c <= 'Z') 1960 c += 'a' - 'A'; 1961 if (c != 'l' && c != 'u') 1962 n *= base; 1963 if (c >= '0' && c <= '9') 1964 { 1965 if (found_suffix) 1966 return ERROR; 1967 n += i = c - '0'; 1968 } 1969 else 1970 { 1971 if (base > 10 && c >= 'a' && c <= 'f') 1972 { 1973 if (found_suffix) 1974 return ERROR; 1975 n += i = c - 'a' + 10; 1976 } 1977 else if (c == 'l') 1978 { 1979 ++long_p; 1980 found_suffix = 1; 1981 } 1982 else if (c == 'u') 1983 { 1984 unsigned_p = 1; 1985 found_suffix = 1; 1986 } 1987 else 1988 return ERROR; /* Char not a digit */ 1989 } 1990 if (i >= base) 1991 return ERROR; /* Invalid digit in this base */ 1992 1993 /* Portably test for overflow (only works for nonzero values, so make 1994 a second check for zero). FIXME: Can't we just make n and prevn 1995 unsigned and avoid this? */ 1996 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0) 1997 unsigned_p = 1; /* Try something unsigned */ 1998 1999 /* Portably test for unsigned overflow. 2000 FIXME: This check is wrong; for example it doesn't find overflow 2001 on 0x123456789 when LONGEST is 32 bits. */ 2002 if (c != 'l' && c != 'u' && n != 0) 2003 { 2004 if (unsigned_p && prevn >= n) 2005 error (_("Numeric constant too large.")); 2006 } 2007 prevn = n; 2008 } 2009 2010 /* An integer constant is an int, a long, or a long long. An L 2011 suffix forces it to be long; an LL suffix forces it to be long 2012 long. If not forced to a larger size, it gets the first type of 2013 the above that it fits in. To figure out whether it fits, we 2014 shift it right and see whether anything remains. Note that we 2015 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one 2016 operation, because many compilers will warn about such a shift 2017 (which always produces a zero result). Sometimes gdbarch_int_bit 2018 or gdbarch_long_bit will be that big, sometimes not. To deal with 2019 the case where it is we just always shift the value more than 2020 once, with fewer bits each time. */ 2021 2022 un = n >> 2; 2023 if (long_p == 0 2024 && (un >> (gdbarch_int_bit (parse_gdbarch (par_state)) - 2)) == 0) 2025 { 2026 high_bit 2027 = ((ULONGEST)1) << (gdbarch_int_bit (parse_gdbarch (par_state)) - 1); 2028 2029 /* A large decimal (not hex or octal) constant (between INT_MAX 2030 and UINT_MAX) is a long or unsigned long, according to ANSI, 2031 never an unsigned int, but this code treats it as unsigned 2032 int. This probably should be fixed. GCC gives a warning on 2033 such constants. */ 2034 2035 unsigned_type = parse_type (par_state)->builtin_unsigned_int; 2036 signed_type = parse_type (par_state)->builtin_int; 2037 } 2038 else if (long_p <= 1 2039 && (un >> (gdbarch_long_bit (parse_gdbarch (par_state)) - 2)) == 0) 2040 { 2041 high_bit 2042 = ((ULONGEST)1) << (gdbarch_long_bit (parse_gdbarch (par_state)) - 1); 2043 unsigned_type = parse_type (par_state)->builtin_unsigned_long; 2044 signed_type = parse_type (par_state)->builtin_long; 2045 } 2046 else 2047 { 2048 int shift; 2049 if (sizeof (ULONGEST) * HOST_CHAR_BIT 2050 < gdbarch_long_long_bit (parse_gdbarch (par_state))) 2051 /* A long long does not fit in a LONGEST. */ 2052 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1); 2053 else 2054 shift = (gdbarch_long_long_bit (parse_gdbarch (par_state)) - 1); 2055 high_bit = (ULONGEST) 1 << shift; 2056 unsigned_type = parse_type (par_state)->builtin_unsigned_long_long; 2057 signed_type = parse_type (par_state)->builtin_long_long; 2058 } 2059 2060 putithere->typed_val_int.val = n; 2061 2062 /* If the high bit of the worked out type is set then this number 2063 has to be unsigned. */ 2064 2065 if (unsigned_p || (n & high_bit)) 2066 { 2067 putithere->typed_val_int.type = unsigned_type; 2068 } 2069 else 2070 { 2071 putithere->typed_val_int.type = signed_type; 2072 } 2073 2074 return INT; 2075 } 2076 2077 /* Temporary obstack used for holding strings. */ 2078 static struct obstack tempbuf; 2079 static int tempbuf_init; 2080 2081 /* Parse a C escape sequence. The initial backslash of the sequence 2082 is at (*PTR)[-1]. *PTR will be updated to point to just after the 2083 last character of the sequence. If OUTPUT is not NULL, the 2084 translated form of the escape sequence will be written there. If 2085 OUTPUT is NULL, no output is written and the call will only affect 2086 *PTR. If an escape sequence is expressed in target bytes, then the 2087 entire sequence will simply be copied to OUTPUT. Return 1 if any 2088 character was emitted, 0 otherwise. */ 2089 2090 int 2091 c_parse_escape (const char **ptr, struct obstack *output) 2092 { 2093 const char *tokptr = *ptr; 2094 int result = 1; 2095 2096 /* Some escape sequences undergo character set conversion. Those we 2097 translate here. */ 2098 switch (*tokptr) 2099 { 2100 /* Hex escapes do not undergo character set conversion, so keep 2101 the escape sequence for later. */ 2102 case 'x': 2103 if (output) 2104 obstack_grow_str (output, "\\x"); 2105 ++tokptr; 2106 if (!ISXDIGIT (*tokptr)) 2107 error (_("\\x escape without a following hex digit")); 2108 while (ISXDIGIT (*tokptr)) 2109 { 2110 if (output) 2111 obstack_1grow (output, *tokptr); 2112 ++tokptr; 2113 } 2114 break; 2115 2116 /* Octal escapes do not undergo character set conversion, so 2117 keep the escape sequence for later. */ 2118 case '0': 2119 case '1': 2120 case '2': 2121 case '3': 2122 case '4': 2123 case '5': 2124 case '6': 2125 case '7': 2126 { 2127 int i; 2128 if (output) 2129 obstack_grow_str (output, "\\"); 2130 for (i = 0; 2131 i < 3 && ISDIGIT (*tokptr) && *tokptr != '8' && *tokptr != '9'; 2132 ++i) 2133 { 2134 if (output) 2135 obstack_1grow (output, *tokptr); 2136 ++tokptr; 2137 } 2138 } 2139 break; 2140 2141 /* We handle UCNs later. We could handle them here, but that 2142 would mean a spurious error in the case where the UCN could 2143 be converted to the target charset but not the host 2144 charset. */ 2145 case 'u': 2146 case 'U': 2147 { 2148 char c = *tokptr; 2149 int i, len = c == 'U' ? 8 : 4; 2150 if (output) 2151 { 2152 obstack_1grow (output, '\\'); 2153 obstack_1grow (output, *tokptr); 2154 } 2155 ++tokptr; 2156 if (!ISXDIGIT (*tokptr)) 2157 error (_("\\%c escape without a following hex digit"), c); 2158 for (i = 0; i < len && ISXDIGIT (*tokptr); ++i) 2159 { 2160 if (output) 2161 obstack_1grow (output, *tokptr); 2162 ++tokptr; 2163 } 2164 } 2165 break; 2166 2167 /* We must pass backslash through so that it does not 2168 cause quoting during the second expansion. */ 2169 case '\\': 2170 if (output) 2171 obstack_grow_str (output, "\\\\"); 2172 ++tokptr; 2173 break; 2174 2175 /* Escapes which undergo conversion. */ 2176 case 'a': 2177 if (output) 2178 obstack_1grow (output, '\a'); 2179 ++tokptr; 2180 break; 2181 case 'b': 2182 if (output) 2183 obstack_1grow (output, '\b'); 2184 ++tokptr; 2185 break; 2186 case 'f': 2187 if (output) 2188 obstack_1grow (output, '\f'); 2189 ++tokptr; 2190 break; 2191 case 'n': 2192 if (output) 2193 obstack_1grow (output, '\n'); 2194 ++tokptr; 2195 break; 2196 case 'r': 2197 if (output) 2198 obstack_1grow (output, '\r'); 2199 ++tokptr; 2200 break; 2201 case 't': 2202 if (output) 2203 obstack_1grow (output, '\t'); 2204 ++tokptr; 2205 break; 2206 case 'v': 2207 if (output) 2208 obstack_1grow (output, '\v'); 2209 ++tokptr; 2210 break; 2211 2212 /* GCC extension. */ 2213 case 'e': 2214 if (output) 2215 obstack_1grow (output, HOST_ESCAPE_CHAR); 2216 ++tokptr; 2217 break; 2218 2219 /* Backslash-newline expands to nothing at all. */ 2220 case '\n': 2221 ++tokptr; 2222 result = 0; 2223 break; 2224 2225 /* A few escapes just expand to the character itself. */ 2226 case '\'': 2227 case '\"': 2228 case '?': 2229 /* GCC extensions. */ 2230 case '(': 2231 case '{': 2232 case '[': 2233 case '%': 2234 /* Unrecognized escapes turn into the character itself. */ 2235 default: 2236 if (output) 2237 obstack_1grow (output, *tokptr); 2238 ++tokptr; 2239 break; 2240 } 2241 *ptr = tokptr; 2242 return result; 2243 } 2244 2245 /* Parse a string or character literal from TOKPTR. The string or 2246 character may be wide or unicode. *OUTPTR is set to just after the 2247 end of the literal in the input string. The resulting token is 2248 stored in VALUE. This returns a token value, either STRING or 2249 CHAR, depending on what was parsed. *HOST_CHARS is set to the 2250 number of host characters in the literal. */ 2251 2252 static int 2253 parse_string_or_char (const char *tokptr, const char **outptr, 2254 struct typed_stoken *value, int *host_chars) 2255 { 2256 int quote; 2257 c_string_type type; 2258 int is_objc = 0; 2259 2260 /* Build the gdb internal form of the input string in tempbuf. Note 2261 that the buffer is null byte terminated *only* for the 2262 convenience of debugging gdb itself and printing the buffer 2263 contents when the buffer contains no embedded nulls. Gdb does 2264 not depend upon the buffer being null byte terminated, it uses 2265 the length string instead. This allows gdb to handle C strings 2266 (as well as strings in other languages) with embedded null 2267 bytes */ 2268 2269 if (!tempbuf_init) 2270 tempbuf_init = 1; 2271 else 2272 obstack_free (&tempbuf, NULL); 2273 obstack_init (&tempbuf); 2274 2275 /* Record the string type. */ 2276 if (*tokptr == 'L') 2277 { 2278 type = C_WIDE_STRING; 2279 ++tokptr; 2280 } 2281 else if (*tokptr == 'u') 2282 { 2283 type = C_STRING_16; 2284 ++tokptr; 2285 } 2286 else if (*tokptr == 'U') 2287 { 2288 type = C_STRING_32; 2289 ++tokptr; 2290 } 2291 else if (*tokptr == '@') 2292 { 2293 /* An Objective C string. */ 2294 is_objc = 1; 2295 type = C_STRING; 2296 ++tokptr; 2297 } 2298 else 2299 type = C_STRING; 2300 2301 /* Skip the quote. */ 2302 quote = *tokptr; 2303 if (quote == '\'') 2304 type |= C_CHAR; 2305 ++tokptr; 2306 2307 *host_chars = 0; 2308 2309 while (*tokptr) 2310 { 2311 char c = *tokptr; 2312 if (c == '\\') 2313 { 2314 ++tokptr; 2315 *host_chars += c_parse_escape (&tokptr, &tempbuf); 2316 } 2317 else if (c == quote) 2318 break; 2319 else 2320 { 2321 obstack_1grow (&tempbuf, c); 2322 ++tokptr; 2323 /* FIXME: this does the wrong thing with multi-byte host 2324 characters. We could use mbrlen here, but that would 2325 make "set host-charset" a bit less useful. */ 2326 ++*host_chars; 2327 } 2328 } 2329 2330 if (*tokptr != quote) 2331 { 2332 if (quote == '"') 2333 error (_("Unterminated string in expression.")); 2334 else 2335 error (_("Unmatched single quote.")); 2336 } 2337 ++tokptr; 2338 2339 value->type = type; 2340 value->ptr = (char *) obstack_base (&tempbuf); 2341 value->length = obstack_object_size (&tempbuf); 2342 2343 *outptr = tokptr; 2344 2345 return quote == '"' ? (is_objc ? NSSTRING : STRING) : CHAR; 2346 } 2347 2348 /* This is used to associate some attributes with a token. */ 2349 2350 enum token_flag 2351 { 2352 /* If this bit is set, the token is C++-only. */ 2353 2354 FLAG_CXX = 1, 2355 2356 /* If this bit is set, the token is conditional: if there is a 2357 symbol of the same name, then the token is a symbol; otherwise, 2358 the token is a keyword. */ 2359 2360 FLAG_SHADOW = 2 2361 }; 2362 DEF_ENUM_FLAGS_TYPE (enum token_flag, token_flags); 2363 2364 struct token 2365 { 2366 const char *oper; 2367 int token; 2368 enum exp_opcode opcode; 2369 token_flags flags; 2370 }; 2371 2372 static const struct token tokentab3[] = 2373 { 2374 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0}, 2375 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0}, 2376 {"->*", ARROW_STAR, BINOP_END, FLAG_CXX}, 2377 {"...", DOTDOTDOT, BINOP_END, 0} 2378 }; 2379 2380 static const struct token tokentab2[] = 2381 { 2382 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0}, 2383 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0}, 2384 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0}, 2385 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0}, 2386 {"%=", ASSIGN_MODIFY, BINOP_REM, 0}, 2387 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0}, 2388 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0}, 2389 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0}, 2390 {"++", INCREMENT, BINOP_END, 0}, 2391 {"--", DECREMENT, BINOP_END, 0}, 2392 {"->", ARROW, BINOP_END, 0}, 2393 {"&&", ANDAND, BINOP_END, 0}, 2394 {"||", OROR, BINOP_END, 0}, 2395 /* "::" is *not* only C++: gdb overrides its meaning in several 2396 different ways, e.g., 'filename'::func, function::variable. */ 2397 {"::", COLONCOLON, BINOP_END, 0}, 2398 {"<<", LSH, BINOP_END, 0}, 2399 {">>", RSH, BINOP_END, 0}, 2400 {"==", EQUAL, BINOP_END, 0}, 2401 {"!=", NOTEQUAL, BINOP_END, 0}, 2402 {"<=", LEQ, BINOP_END, 0}, 2403 {">=", GEQ, BINOP_END, 0}, 2404 {".*", DOT_STAR, BINOP_END, FLAG_CXX} 2405 }; 2406 2407 /* Identifier-like tokens. Only type-specifiers than can appear in 2408 multi-word type names (for example 'double' can appear in 'long 2409 double') need to be listed here. type-specifiers that are only ever 2410 single word (like 'float') are handled by the classify_name function. */ 2411 static const struct token ident_tokens[] = 2412 { 2413 {"unsigned", UNSIGNED, OP_NULL, 0}, 2414 {"template", TEMPLATE, OP_NULL, FLAG_CXX}, 2415 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0}, 2416 {"struct", STRUCT, OP_NULL, 0}, 2417 {"signed", SIGNED_KEYWORD, OP_NULL, 0}, 2418 {"sizeof", SIZEOF, OP_NULL, 0}, 2419 {"_Alignof", ALIGNOF, OP_NULL, 0}, 2420 {"alignof", ALIGNOF, OP_NULL, FLAG_CXX}, 2421 {"double", DOUBLE_KEYWORD, OP_NULL, 0}, 2422 {"false", FALSEKEYWORD, OP_NULL, FLAG_CXX}, 2423 {"class", CLASS, OP_NULL, FLAG_CXX}, 2424 {"union", UNION, OP_NULL, 0}, 2425 {"short", SHORT, OP_NULL, 0}, 2426 {"const", CONST_KEYWORD, OP_NULL, 0}, 2427 {"enum", ENUM, OP_NULL, 0}, 2428 {"long", LONG, OP_NULL, 0}, 2429 {"true", TRUEKEYWORD, OP_NULL, FLAG_CXX}, 2430 {"int", INT_KEYWORD, OP_NULL, 0}, 2431 {"new", NEW, OP_NULL, FLAG_CXX}, 2432 {"delete", DELETE, OP_NULL, FLAG_CXX}, 2433 {"operator", OPERATOR, OP_NULL, FLAG_CXX}, 2434 2435 {"and", ANDAND, BINOP_END, FLAG_CXX}, 2436 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, FLAG_CXX}, 2437 {"bitand", '&', OP_NULL, FLAG_CXX}, 2438 {"bitor", '|', OP_NULL, FLAG_CXX}, 2439 {"compl", '~', OP_NULL, FLAG_CXX}, 2440 {"not", '!', OP_NULL, FLAG_CXX}, 2441 {"not_eq", NOTEQUAL, BINOP_END, FLAG_CXX}, 2442 {"or", OROR, BINOP_END, FLAG_CXX}, 2443 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, FLAG_CXX}, 2444 {"xor", '^', OP_NULL, FLAG_CXX}, 2445 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, FLAG_CXX}, 2446 2447 {"const_cast", CONST_CAST, OP_NULL, FLAG_CXX }, 2448 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, FLAG_CXX }, 2449 {"static_cast", STATIC_CAST, OP_NULL, FLAG_CXX }, 2450 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, FLAG_CXX }, 2451 2452 {"__typeof__", TYPEOF, OP_TYPEOF, 0 }, 2453 {"__typeof", TYPEOF, OP_TYPEOF, 0 }, 2454 {"typeof", TYPEOF, OP_TYPEOF, FLAG_SHADOW }, 2455 {"__decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX }, 2456 {"decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX | FLAG_SHADOW }, 2457 2458 {"typeid", TYPEID, OP_TYPEID, FLAG_CXX} 2459 }; 2460 2461 2462 static void 2463 scan_macro_expansion (char *expansion) 2464 { 2465 char *copy; 2466 2467 /* We'd better not be trying to push the stack twice. */ 2468 gdb_assert (! cpstate->macro_original_text); 2469 2470 /* Copy to the obstack, and then free the intermediate 2471 expansion. */ 2472 copy = (char *) obstack_copy0 (&cpstate->expansion_obstack, expansion, 2473 strlen (expansion)); 2474 xfree (expansion); 2475 2476 /* Save the old lexptr value, so we can return to it when we're done 2477 parsing the expanded text. */ 2478 cpstate->macro_original_text = lexptr; 2479 lexptr = copy; 2480 } 2481 2482 static int 2483 scanning_macro_expansion (void) 2484 { 2485 return cpstate->macro_original_text != 0; 2486 } 2487 2488 static void 2489 finished_macro_expansion (void) 2490 { 2491 /* There'd better be something to pop back to. */ 2492 gdb_assert (cpstate->macro_original_text); 2493 2494 /* Pop back to the original text. */ 2495 lexptr = cpstate->macro_original_text; 2496 cpstate->macro_original_text = 0; 2497 } 2498 2499 /* Return true iff the token represents a C++ cast operator. */ 2500 2501 static int 2502 is_cast_operator (const char *token, int len) 2503 { 2504 return (! strncmp (token, "dynamic_cast", len) 2505 || ! strncmp (token, "static_cast", len) 2506 || ! strncmp (token, "reinterpret_cast", len) 2507 || ! strncmp (token, "const_cast", len)); 2508 } 2509 2510 /* The scope used for macro expansion. */ 2511 static struct macro_scope *expression_macro_scope; 2512 2513 /* This is set if a NAME token appeared at the very end of the input 2514 string, with no whitespace separating the name from the EOF. This 2515 is used only when parsing to do field name completion. */ 2516 static int saw_name_at_eof; 2517 2518 /* This is set if the previously-returned token was a structure 2519 operator -- either '.' or ARROW. */ 2520 static bool last_was_structop; 2521 2522 /* Read one token, getting characters through lexptr. */ 2523 2524 static int 2525 lex_one_token (struct parser_state *par_state, bool *is_quoted_name) 2526 { 2527 int c; 2528 int namelen; 2529 unsigned int i; 2530 const char *tokstart; 2531 bool saw_structop = last_was_structop; 2532 char *copy; 2533 2534 last_was_structop = false; 2535 *is_quoted_name = false; 2536 2537 retry: 2538 2539 /* Check if this is a macro invocation that we need to expand. */ 2540 if (! scanning_macro_expansion ()) 2541 { 2542 char *expanded = macro_expand_next (&lexptr, 2543 standard_macro_lookup, 2544 expression_macro_scope); 2545 2546 if (expanded) 2547 scan_macro_expansion (expanded); 2548 } 2549 2550 prev_lexptr = lexptr; 2551 2552 tokstart = lexptr; 2553 /* See if it is a special token of length 3. */ 2554 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++) 2555 if (strncmp (tokstart, tokentab3[i].oper, 3) == 0) 2556 { 2557 if ((tokentab3[i].flags & FLAG_CXX) != 0 2558 && parse_language (par_state)->la_language != language_cplus) 2559 break; 2560 2561 lexptr += 3; 2562 yylval.opcode = tokentab3[i].opcode; 2563 return tokentab3[i].token; 2564 } 2565 2566 /* See if it is a special token of length 2. */ 2567 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++) 2568 if (strncmp (tokstart, tokentab2[i].oper, 2) == 0) 2569 { 2570 if ((tokentab2[i].flags & FLAG_CXX) != 0 2571 && parse_language (par_state)->la_language != language_cplus) 2572 break; 2573 2574 lexptr += 2; 2575 yylval.opcode = tokentab2[i].opcode; 2576 if (tokentab2[i].token == ARROW) 2577 last_was_structop = 1; 2578 return tokentab2[i].token; 2579 } 2580 2581 switch (c = *tokstart) 2582 { 2583 case 0: 2584 /* If we were just scanning the result of a macro expansion, 2585 then we need to resume scanning the original text. 2586 If we're parsing for field name completion, and the previous 2587 token allows such completion, return a COMPLETE token. 2588 Otherwise, we were already scanning the original text, and 2589 we're really done. */ 2590 if (scanning_macro_expansion ()) 2591 { 2592 finished_macro_expansion (); 2593 goto retry; 2594 } 2595 else if (saw_name_at_eof) 2596 { 2597 saw_name_at_eof = 0; 2598 return COMPLETE; 2599 } 2600 else if (parse_completion && saw_structop) 2601 return COMPLETE; 2602 else 2603 return 0; 2604 2605 case ' ': 2606 case '\t': 2607 case '\n': 2608 lexptr++; 2609 goto retry; 2610 2611 case '[': 2612 case '(': 2613 paren_depth++; 2614 lexptr++; 2615 if (parse_language (par_state)->la_language == language_objc 2616 && c == '[') 2617 return OBJC_LBRAC; 2618 return c; 2619 2620 case ']': 2621 case ')': 2622 if (paren_depth == 0) 2623 return 0; 2624 paren_depth--; 2625 lexptr++; 2626 return c; 2627 2628 case ',': 2629 if (comma_terminates 2630 && paren_depth == 0 2631 && ! scanning_macro_expansion ()) 2632 return 0; 2633 lexptr++; 2634 return c; 2635 2636 case '.': 2637 /* Might be a floating point number. */ 2638 if (lexptr[1] < '0' || lexptr[1] > '9') 2639 { 2640 last_was_structop = true; 2641 goto symbol; /* Nope, must be a symbol. */ 2642 } 2643 /* FALL THRU. */ 2644 2645 case '0': 2646 case '1': 2647 case '2': 2648 case '3': 2649 case '4': 2650 case '5': 2651 case '6': 2652 case '7': 2653 case '8': 2654 case '9': 2655 { 2656 /* It's a number. */ 2657 int got_dot = 0, got_e = 0, toktype; 2658 const char *p = tokstart; 2659 int hex = input_radix > 10; 2660 2661 if (c == '0' && (p[1] == 'x' || p[1] == 'X')) 2662 { 2663 p += 2; 2664 hex = 1; 2665 } 2666 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D')) 2667 { 2668 p += 2; 2669 hex = 0; 2670 } 2671 2672 for (;; ++p) 2673 { 2674 /* This test includes !hex because 'e' is a valid hex digit 2675 and thus does not indicate a floating point number when 2676 the radix is hex. */ 2677 if (!hex && !got_e && (*p == 'e' || *p == 'E')) 2678 got_dot = got_e = 1; 2679 /* This test does not include !hex, because a '.' always indicates 2680 a decimal floating point number regardless of the radix. */ 2681 else if (!got_dot && *p == '.') 2682 got_dot = 1; 2683 else if (got_e && (p[-1] == 'e' || p[-1] == 'E') 2684 && (*p == '-' || *p == '+')) 2685 /* This is the sign of the exponent, not the end of the 2686 number. */ 2687 continue; 2688 /* We will take any letters or digits. parse_number will 2689 complain if past the radix, or if L or U are not final. */ 2690 else if ((*p < '0' || *p > '9') 2691 && ((*p < 'a' || *p > 'z') 2692 && (*p < 'A' || *p > 'Z'))) 2693 break; 2694 } 2695 toktype = parse_number (par_state, tokstart, p - tokstart, 2696 got_dot|got_e, &yylval); 2697 if (toktype == ERROR) 2698 { 2699 char *err_copy = (char *) alloca (p - tokstart + 1); 2700 2701 memcpy (err_copy, tokstart, p - tokstart); 2702 err_copy[p - tokstart] = 0; 2703 error (_("Invalid number \"%s\"."), err_copy); 2704 } 2705 lexptr = p; 2706 return toktype; 2707 } 2708 2709 case '@': 2710 { 2711 const char *p = &tokstart[1]; 2712 2713 if (parse_language (par_state)->la_language == language_objc) 2714 { 2715 size_t len = strlen ("selector"); 2716 2717 if (strncmp (p, "selector", len) == 0 2718 && (p[len] == '\0' || ISSPACE (p[len]))) 2719 { 2720 lexptr = p + len; 2721 return SELECTOR; 2722 } 2723 else if (*p == '"') 2724 goto parse_string; 2725 } 2726 2727 while (ISSPACE (*p)) 2728 p++; 2729 size_t len = strlen ("entry"); 2730 if (strncmp (p, "entry", len) == 0 && !c_ident_is_alnum (p[len]) 2731 && p[len] != '_') 2732 { 2733 lexptr = &p[len]; 2734 return ENTRY; 2735 } 2736 } 2737 /* FALLTHRU */ 2738 case '+': 2739 case '-': 2740 case '*': 2741 case '/': 2742 case '%': 2743 case '|': 2744 case '&': 2745 case '^': 2746 case '~': 2747 case '!': 2748 case '<': 2749 case '>': 2750 case '?': 2751 case ':': 2752 case '=': 2753 case '{': 2754 case '}': 2755 symbol: 2756 lexptr++; 2757 return c; 2758 2759 case 'L': 2760 case 'u': 2761 case 'U': 2762 if (tokstart[1] != '"' && tokstart[1] != '\'') 2763 break; 2764 /* Fall through. */ 2765 case '\'': 2766 case '"': 2767 2768 parse_string: 2769 { 2770 int host_len; 2771 int result = parse_string_or_char (tokstart, &lexptr, &yylval.tsval, 2772 &host_len); 2773 if (result == CHAR) 2774 { 2775 if (host_len == 0) 2776 error (_("Empty character constant.")); 2777 else if (host_len > 2 && c == '\'') 2778 { 2779 ++tokstart; 2780 namelen = lexptr - tokstart - 1; 2781 *is_quoted_name = true; 2782 2783 goto tryname; 2784 } 2785 else if (host_len > 1) 2786 error (_("Invalid character constant.")); 2787 } 2788 return result; 2789 } 2790 } 2791 2792 if (!(c == '_' || c == '$' || c_ident_is_alpha (c))) 2793 /* We must have come across a bad character (e.g. ';'). */ 2794 error (_("Invalid character '%c' in expression."), c); 2795 2796 /* It's a name. See how long it is. */ 2797 namelen = 0; 2798 for (c = tokstart[namelen]; 2799 (c == '_' || c == '$' || c_ident_is_alnum (c) || c == '<');) 2800 { 2801 /* Template parameter lists are part of the name. 2802 FIXME: This mishandles `print $a<4&&$a>3'. */ 2803 2804 if (c == '<') 2805 { 2806 if (! is_cast_operator (tokstart, namelen)) 2807 { 2808 /* Scan ahead to get rest of the template specification. Note 2809 that we look ahead only when the '<' adjoins non-whitespace 2810 characters; for comparison expressions, e.g. "a < b > c", 2811 there must be spaces before the '<', etc. */ 2812 const char *p = find_template_name_end (tokstart + namelen); 2813 2814 if (p) 2815 namelen = p - tokstart; 2816 } 2817 break; 2818 } 2819 c = tokstart[++namelen]; 2820 } 2821 2822 /* The token "if" terminates the expression and is NOT removed from 2823 the input stream. It doesn't count if it appears in the 2824 expansion of a macro. */ 2825 if (namelen == 2 2826 && tokstart[0] == 'i' 2827 && tokstart[1] == 'f' 2828 && ! scanning_macro_expansion ()) 2829 { 2830 return 0; 2831 } 2832 2833 /* For the same reason (breakpoint conditions), "thread N" 2834 terminates the expression. "thread" could be an identifier, but 2835 an identifier is never followed by a number without intervening 2836 punctuation. "task" is similar. Handle abbreviations of these, 2837 similarly to breakpoint.c:find_condition_and_thread. */ 2838 if (namelen >= 1 2839 && (strncmp (tokstart, "thread", namelen) == 0 2840 || strncmp (tokstart, "task", namelen) == 0) 2841 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t') 2842 && ! scanning_macro_expansion ()) 2843 { 2844 const char *p = tokstart + namelen + 1; 2845 2846 while (*p == ' ' || *p == '\t') 2847 p++; 2848 if (*p >= '0' && *p <= '9') 2849 return 0; 2850 } 2851 2852 lexptr += namelen; 2853 2854 tryname: 2855 2856 yylval.sval.ptr = tokstart; 2857 yylval.sval.length = namelen; 2858 2859 /* Catch specific keywords. */ 2860 copy = copy_name (yylval.sval); 2861 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++) 2862 if (strcmp (copy, ident_tokens[i].oper) == 0) 2863 { 2864 if ((ident_tokens[i].flags & FLAG_CXX) != 0 2865 && parse_language (par_state)->la_language != language_cplus) 2866 break; 2867 2868 if ((ident_tokens[i].flags & FLAG_SHADOW) != 0) 2869 { 2870 struct field_of_this_result is_a_field_of_this; 2871 2872 if (lookup_symbol (copy, expression_context_block, 2873 VAR_DOMAIN, 2874 (parse_language (par_state)->la_language 2875 == language_cplus ? &is_a_field_of_this 2876 : NULL)).symbol 2877 != NULL) 2878 { 2879 /* The keyword is shadowed. */ 2880 break; 2881 } 2882 } 2883 2884 /* It is ok to always set this, even though we don't always 2885 strictly need to. */ 2886 yylval.opcode = ident_tokens[i].opcode; 2887 return ident_tokens[i].token; 2888 } 2889 2890 if (*tokstart == '$') 2891 return DOLLAR_VARIABLE; 2892 2893 if (parse_completion && *lexptr == '\0') 2894 saw_name_at_eof = 1; 2895 2896 yylval.ssym.stoken = yylval.sval; 2897 yylval.ssym.sym.symbol = NULL; 2898 yylval.ssym.sym.block = NULL; 2899 yylval.ssym.is_a_field_of_this = 0; 2900 return NAME; 2901 } 2902 2903 /* An object of this type is pushed on a FIFO by the "outer" lexer. */ 2904 struct token_and_value 2905 { 2906 int token; 2907 YYSTYPE value; 2908 }; 2909 2910 /* A FIFO of tokens that have been read but not yet returned to the 2911 parser. */ 2912 static std::vector<token_and_value> token_fifo; 2913 2914 /* Non-zero if the lexer should return tokens from the FIFO. */ 2915 static int popping; 2916 2917 /* Temporary storage for c_lex; this holds symbol names as they are 2918 built up. */ 2919 auto_obstack name_obstack; 2920 2921 /* Classify a NAME token. The contents of the token are in `yylval'. 2922 Updates yylval and returns the new token type. BLOCK is the block 2923 in which lookups start; this can be NULL to mean the global scope. 2924 IS_QUOTED_NAME is non-zero if the name token was originally quoted 2925 in single quotes. IS_AFTER_STRUCTOP is true if this name follows 2926 a structure operator -- either '.' or ARROW */ 2927 2928 static int 2929 classify_name (struct parser_state *par_state, const struct block *block, 2930 bool is_quoted_name, bool is_after_structop) 2931 { 2932 struct block_symbol bsym; 2933 char *copy; 2934 struct field_of_this_result is_a_field_of_this; 2935 2936 copy = copy_name (yylval.sval); 2937 2938 /* Initialize this in case we *don't* use it in this call; that way 2939 we can refer to it unconditionally below. */ 2940 memset (&is_a_field_of_this, 0, sizeof (is_a_field_of_this)); 2941 2942 bsym = lookup_symbol (copy, block, VAR_DOMAIN, 2943 parse_language (par_state)->la_name_of_this 2944 ? &is_a_field_of_this : NULL); 2945 2946 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_BLOCK) 2947 { 2948 yylval.ssym.sym = bsym; 2949 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL; 2950 return BLOCKNAME; 2951 } 2952 else if (!bsym.symbol) 2953 { 2954 /* If we found a field of 'this', we might have erroneously 2955 found a constructor where we wanted a type name. Handle this 2956 case by noticing that we found a constructor and then look up 2957 the type tag instead. */ 2958 if (is_a_field_of_this.type != NULL 2959 && is_a_field_of_this.fn_field != NULL 2960 && TYPE_FN_FIELD_CONSTRUCTOR (is_a_field_of_this.fn_field->fn_fields, 2961 0)) 2962 { 2963 struct field_of_this_result inner_is_a_field_of_this; 2964 2965 bsym = lookup_symbol (copy, block, STRUCT_DOMAIN, 2966 &inner_is_a_field_of_this); 2967 if (bsym.symbol != NULL) 2968 { 2969 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol); 2970 return TYPENAME; 2971 } 2972 } 2973 2974 /* If we found a field on the "this" object, or we are looking 2975 up a field on a struct, then we want to prefer it over a 2976 filename. However, if the name was quoted, then it is better 2977 to check for a filename or a block, since this is the only 2978 way the user has of requiring the extension to be used. */ 2979 if ((is_a_field_of_this.type == NULL && !is_after_structop) 2980 || is_quoted_name) 2981 { 2982 /* See if it's a file name. */ 2983 struct symtab *symtab; 2984 2985 symtab = lookup_symtab (copy); 2986 if (symtab) 2987 { 2988 yylval.bval = BLOCKVECTOR_BLOCK (SYMTAB_BLOCKVECTOR (symtab), 2989 STATIC_BLOCK); 2990 return FILENAME; 2991 } 2992 } 2993 } 2994 2995 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_TYPEDEF) 2996 { 2997 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol); 2998 return TYPENAME; 2999 } 3000 3001 /* See if it's an ObjC classname. */ 3002 if (parse_language (par_state)->la_language == language_objc && !bsym.symbol) 3003 { 3004 CORE_ADDR Class = lookup_objc_class (parse_gdbarch (par_state), copy); 3005 if (Class) 3006 { 3007 struct symbol *sym; 3008 3009 yylval.theclass.theclass = Class; 3010 sym = lookup_struct_typedef (copy, expression_context_block, 1); 3011 if (sym) 3012 yylval.theclass.type = SYMBOL_TYPE (sym); 3013 return CLASSNAME; 3014 } 3015 } 3016 3017 /* Input names that aren't symbols but ARE valid hex numbers, when 3018 the input radix permits them, can be names or numbers depending 3019 on the parse. Note we support radixes > 16 here. */ 3020 if (!bsym.symbol 3021 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10) 3022 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10))) 3023 { 3024 YYSTYPE newlval; /* Its value is ignored. */ 3025 int hextype = parse_number (par_state, copy, yylval.sval.length, 3026 0, &newlval); 3027 3028 if (hextype == INT) 3029 { 3030 yylval.ssym.sym = bsym; 3031 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL; 3032 return NAME_OR_INT; 3033 } 3034 } 3035 3036 /* Any other kind of symbol */ 3037 yylval.ssym.sym = bsym; 3038 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL; 3039 3040 if (bsym.symbol == NULL 3041 && parse_language (par_state)->la_language == language_cplus 3042 && is_a_field_of_this.type == NULL 3043 && lookup_minimal_symbol (copy, NULL, NULL).minsym == NULL) 3044 return UNKNOWN_CPP_NAME; 3045 3046 return NAME; 3047 } 3048 3049 /* Like classify_name, but used by the inner loop of the lexer, when a 3050 name might have already been seen. CONTEXT is the context type, or 3051 NULL if this is the first component of a name. */ 3052 3053 static int 3054 classify_inner_name (struct parser_state *par_state, 3055 const struct block *block, struct type *context) 3056 { 3057 struct type *type; 3058 char *copy; 3059 3060 if (context == NULL) 3061 return classify_name (par_state, block, false, false); 3062 3063 type = check_typedef (context); 3064 if (!type_aggregate_p (type)) 3065 return ERROR; 3066 3067 copy = copy_name (yylval.ssym.stoken); 3068 /* N.B. We assume the symbol can only be in VAR_DOMAIN. */ 3069 yylval.ssym.sym = cp_lookup_nested_symbol (type, copy, block, VAR_DOMAIN); 3070 3071 /* If no symbol was found, search for a matching base class named 3072 COPY. This will allow users to enter qualified names of class members 3073 relative to the `this' pointer. */ 3074 if (yylval.ssym.sym.symbol == NULL) 3075 { 3076 struct type *base_type = cp_find_type_baseclass_by_name (type, copy); 3077 3078 if (base_type != NULL) 3079 { 3080 yylval.tsym.type = base_type; 3081 return TYPENAME; 3082 } 3083 3084 return ERROR; 3085 } 3086 3087 switch (SYMBOL_CLASS (yylval.ssym.sym.symbol)) 3088 { 3089 case LOC_BLOCK: 3090 case LOC_LABEL: 3091 /* cp_lookup_nested_symbol might have accidentally found a constructor 3092 named COPY when we really wanted a base class of the same name. 3093 Double-check this case by looking for a base class. */ 3094 { 3095 struct type *base_type = cp_find_type_baseclass_by_name (type, copy); 3096 3097 if (base_type != NULL) 3098 { 3099 yylval.tsym.type = base_type; 3100 return TYPENAME; 3101 } 3102 } 3103 return ERROR; 3104 3105 case LOC_TYPEDEF: 3106 yylval.tsym.type = SYMBOL_TYPE (yylval.ssym.sym.symbol); 3107 return TYPENAME; 3108 3109 default: 3110 return NAME; 3111 } 3112 internal_error (__FILE__, __LINE__, _("not reached")); 3113 } 3114 3115 /* The outer level of a two-level lexer. This calls the inner lexer 3116 to return tokens. It then either returns these tokens, or 3117 aggregates them into a larger token. This lets us work around a 3118 problem in our parsing approach, where the parser could not 3119 distinguish between qualified names and qualified types at the 3120 right point. 3121 3122 This approach is still not ideal, because it mishandles template 3123 types. See the comment in lex_one_token for an example. However, 3124 this is still an improvement over the earlier approach, and will 3125 suffice until we move to better parsing technology. */ 3126 3127 static int 3128 yylex (void) 3129 { 3130 token_and_value current; 3131 int first_was_coloncolon, last_was_coloncolon; 3132 struct type *context_type = NULL; 3133 int last_to_examine, next_to_examine, checkpoint; 3134 const struct block *search_block; 3135 bool is_quoted_name, last_lex_was_structop; 3136 3137 if (popping && !token_fifo.empty ()) 3138 goto do_pop; 3139 popping = 0; 3140 3141 last_lex_was_structop = last_was_structop; 3142 3143 /* Read the first token and decide what to do. Most of the 3144 subsequent code is C++-only; but also depends on seeing a "::" or 3145 name-like token. */ 3146 current.token = lex_one_token (pstate, &is_quoted_name); 3147 if (current.token == NAME) 3148 current.token = classify_name (pstate, expression_context_block, 3149 is_quoted_name, last_lex_was_structop); 3150 if (parse_language (pstate)->la_language != language_cplus 3151 || (current.token != TYPENAME && current.token != COLONCOLON 3152 && current.token != FILENAME)) 3153 return current.token; 3154 3155 /* Read any sequence of alternating "::" and name-like tokens into 3156 the token FIFO. */ 3157 current.value = yylval; 3158 token_fifo.push_back (current); 3159 last_was_coloncolon = current.token == COLONCOLON; 3160 while (1) 3161 { 3162 bool ignore; 3163 3164 /* We ignore quoted names other than the very first one. 3165 Subsequent ones do not have any special meaning. */ 3166 current.token = lex_one_token (pstate, &ignore); 3167 current.value = yylval; 3168 token_fifo.push_back (current); 3169 3170 if ((last_was_coloncolon && current.token != NAME) 3171 || (!last_was_coloncolon && current.token != COLONCOLON)) 3172 break; 3173 last_was_coloncolon = !last_was_coloncolon; 3174 } 3175 popping = 1; 3176 3177 /* We always read one extra token, so compute the number of tokens 3178 to examine accordingly. */ 3179 last_to_examine = token_fifo.size () - 2; 3180 next_to_examine = 0; 3181 3182 current = token_fifo[next_to_examine]; 3183 ++next_to_examine; 3184 3185 name_obstack.clear (); 3186 checkpoint = 0; 3187 if (current.token == FILENAME) 3188 search_block = current.value.bval; 3189 else if (current.token == COLONCOLON) 3190 search_block = NULL; 3191 else 3192 { 3193 gdb_assert (current.token == TYPENAME); 3194 search_block = expression_context_block; 3195 obstack_grow (&name_obstack, current.value.sval.ptr, 3196 current.value.sval.length); 3197 context_type = current.value.tsym.type; 3198 checkpoint = 1; 3199 } 3200 3201 first_was_coloncolon = current.token == COLONCOLON; 3202 last_was_coloncolon = first_was_coloncolon; 3203 3204 while (next_to_examine <= last_to_examine) 3205 { 3206 token_and_value next; 3207 3208 next = token_fifo[next_to_examine]; 3209 ++next_to_examine; 3210 3211 if (next.token == NAME && last_was_coloncolon) 3212 { 3213 int classification; 3214 3215 yylval = next.value; 3216 classification = classify_inner_name (pstate, search_block, 3217 context_type); 3218 /* We keep going until we either run out of names, or until 3219 we have a qualified name which is not a type. */ 3220 if (classification != TYPENAME && classification != NAME) 3221 break; 3222 3223 /* Accept up to this token. */ 3224 checkpoint = next_to_examine; 3225 3226 /* Update the partial name we are constructing. */ 3227 if (context_type != NULL) 3228 { 3229 /* We don't want to put a leading "::" into the name. */ 3230 obstack_grow_str (&name_obstack, "::"); 3231 } 3232 obstack_grow (&name_obstack, next.value.sval.ptr, 3233 next.value.sval.length); 3234 3235 yylval.sval.ptr = (const char *) obstack_base (&name_obstack); 3236 yylval.sval.length = obstack_object_size (&name_obstack); 3237 current.value = yylval; 3238 current.token = classification; 3239 3240 last_was_coloncolon = 0; 3241 3242 if (classification == NAME) 3243 break; 3244 3245 context_type = yylval.tsym.type; 3246 } 3247 else if (next.token == COLONCOLON && !last_was_coloncolon) 3248 last_was_coloncolon = 1; 3249 else 3250 { 3251 /* We've reached the end of the name. */ 3252 break; 3253 } 3254 } 3255 3256 /* If we have a replacement token, install it as the first token in 3257 the FIFO, and delete the other constituent tokens. */ 3258 if (checkpoint > 0) 3259 { 3260 current.value.sval.ptr 3261 = (const char *) obstack_copy0 (&cpstate->expansion_obstack, 3262 current.value.sval.ptr, 3263 current.value.sval.length); 3264 3265 token_fifo[0] = current; 3266 if (checkpoint > 1) 3267 token_fifo.erase (token_fifo.begin () + 1, 3268 token_fifo.begin () + checkpoint); 3269 } 3270 3271 do_pop: 3272 current = token_fifo[0]; 3273 token_fifo.erase (token_fifo.begin ()); 3274 yylval = current.value; 3275 return current.token; 3276 } 3277 3278 int 3279 c_parse (struct parser_state *par_state) 3280 { 3281 /* Setting up the parser state. */ 3282 scoped_restore pstate_restore = make_scoped_restore (&pstate); 3283 gdb_assert (par_state != NULL); 3284 pstate = par_state; 3285 3286 c_parse_state cstate; 3287 scoped_restore cstate_restore = make_scoped_restore (&cpstate, &cstate); 3288 3289 gdb::unique_xmalloc_ptr<struct macro_scope> macro_scope; 3290 3291 if (expression_context_block) 3292 macro_scope = sal_macro_scope (find_pc_line (expression_context_pc, 0)); 3293 else 3294 macro_scope = default_macro_scope (); 3295 if (! macro_scope) 3296 macro_scope = user_macro_scope (); 3297 3298 scoped_restore restore_macro_scope 3299 = make_scoped_restore (&expression_macro_scope, macro_scope.get ()); 3300 3301 scoped_restore restore_yydebug = make_scoped_restore (&yydebug, 3302 parser_debug); 3303 3304 /* Initialize some state used by the lexer. */ 3305 last_was_structop = false; 3306 saw_name_at_eof = 0; 3307 3308 token_fifo.clear (); 3309 popping = 0; 3310 name_obstack.clear (); 3311 3312 return yyparse (); 3313 } 3314 3315 #ifdef YYBISON 3316 3317 /* This is called via the YYPRINT macro when parser debugging is 3318 enabled. It prints a token's value. */ 3319 3320 static void 3321 c_print_token (FILE *file, int type, YYSTYPE value) 3322 { 3323 switch (type) 3324 { 3325 case INT: 3326 parser_fprintf (file, "typed_val_int<%s, %s>", 3327 TYPE_SAFE_NAME (value.typed_val_int.type), 3328 pulongest (value.typed_val_int.val)); 3329 break; 3330 3331 case CHAR: 3332 case STRING: 3333 { 3334 char *copy = (char *) alloca (value.tsval.length + 1); 3335 3336 memcpy (copy, value.tsval.ptr, value.tsval.length); 3337 copy[value.tsval.length] = '\0'; 3338 3339 parser_fprintf (file, "tsval<type=%d, %s>", value.tsval.type, copy); 3340 } 3341 break; 3342 3343 case NSSTRING: 3344 case DOLLAR_VARIABLE: 3345 parser_fprintf (file, "sval<%s>", copy_name (value.sval)); 3346 break; 3347 3348 case TYPENAME: 3349 parser_fprintf (file, "tsym<type=%s, name=%s>", 3350 TYPE_SAFE_NAME (value.tsym.type), 3351 copy_name (value.tsym.stoken)); 3352 break; 3353 3354 case NAME: 3355 case UNKNOWN_CPP_NAME: 3356 case NAME_OR_INT: 3357 case BLOCKNAME: 3358 parser_fprintf (file, "ssym<name=%s, sym=%s, field_of_this=%d>", 3359 copy_name (value.ssym.stoken), 3360 (value.ssym.sym.symbol == NULL 3361 ? "(null)" : SYMBOL_PRINT_NAME (value.ssym.sym.symbol)), 3362 value.ssym.is_a_field_of_this); 3363 break; 3364 3365 case FILENAME: 3366 parser_fprintf (file, "bval<%s>", host_address_to_string (value.bval)); 3367 break; 3368 } 3369 } 3370 3371 #endif 3372 3373 static void 3374 yyerror (const char *msg) 3375 { 3376 if (prev_lexptr) 3377 lexptr = prev_lexptr; 3378 3379 error (_("A %s in expression, near `%s'."), msg, lexptr); 3380 } 3381