1 /* YACC parser for C expressions, for GDB. 2 Copyright (C) 1986, 1989, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 3 1998, 1999, 2000, 2003, 2004, 2006, 2007, 2008, 2009, 2010, 2011 4 Free Software Foundation, Inc. 5 6 This file is part of GDB. 7 8 This program is free software; you can redistribute it and/or modify 9 it under the terms of the GNU General Public License as published by 10 the Free Software Foundation; either version 3 of the License, or 11 (at your option) any later version. 12 13 This program is distributed in the hope that it will be useful, 14 but WITHOUT ANY WARRANTY; without even the implied warranty of 15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 16 GNU General Public License for more details. 17 18 You should have received a copy of the GNU General Public License 19 along with this program. If not, see <http://www.gnu.org/licenses/>. */ 20 21 /* Parse a C expression from text in a string, 22 and return the result as a struct expression pointer. 23 That structure contains arithmetic operations in reverse polish, 24 with constants represented by operations that are followed by special data. 25 See expression.h for the details of the format. 26 What is important here is that it can be built up sequentially 27 during the process of parsing; the lower levels of the tree always 28 come first in the result. 29 30 Note that malloc's and realloc's in this file are transformed to 31 xmalloc and xrealloc respectively by the same sed command in the 32 makefile that remaps any other malloc/realloc inserted by the parser 33 generator. Doing this with #defines and trying to control the interaction 34 with include files (<malloc.h> and <stdlib.h> for example) just became 35 too messy, particularly when such includes can be inserted at random 36 times by the parser generator. */ 37 38 %{ 39 40 #include "defs.h" 41 #include "gdb_string.h" 42 #include <ctype.h> 43 #include "expression.h" 44 #include "value.h" 45 #include "parser-defs.h" 46 #include "language.h" 47 #include "c-lang.h" 48 #include "bfd.h" /* Required by objfiles.h. */ 49 #include "symfile.h" /* Required by objfiles.h. */ 50 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */ 51 #include "charset.h" 52 #include "block.h" 53 #include "cp-support.h" 54 #include "dfp.h" 55 #include "gdb_assert.h" 56 #include "macroscope.h" 57 58 #define parse_type builtin_type (parse_gdbarch) 59 60 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc), 61 as well as gratuitiously global symbol names, so we can have multiple 62 yacc generated parsers in gdb. Note that these are only the variables 63 produced by yacc. If other parser generators (bison, byacc, etc) produce 64 additional global names that conflict at link time, then those parser 65 generators need to be fixed instead of adding those names to this list. */ 66 67 #define yymaxdepth c_maxdepth 68 #define yyparse c_parse_internal 69 #define yylex c_lex 70 #define yyerror c_error 71 #define yylval c_lval 72 #define yychar c_char 73 #define yydebug c_debug 74 #define yypact c_pact 75 #define yyr1 c_r1 76 #define yyr2 c_r2 77 #define yydef c_def 78 #define yychk c_chk 79 #define yypgo c_pgo 80 #define yyact c_act 81 #define yyexca c_exca 82 #define yyerrflag c_errflag 83 #define yynerrs c_nerrs 84 #define yyps c_ps 85 #define yypv c_pv 86 #define yys c_s 87 #define yy_yys c_yys 88 #define yystate c_state 89 #define yytmp c_tmp 90 #define yyv c_v 91 #define yy_yyv c_yyv 92 #define yyval c_val 93 #define yylloc c_lloc 94 #define yyreds c_reds /* With YYDEBUG defined */ 95 #define yytoks c_toks /* With YYDEBUG defined */ 96 #define yyname c_name /* With YYDEBUG defined */ 97 #define yyrule c_rule /* With YYDEBUG defined */ 98 #define yylhs c_yylhs 99 #define yylen c_yylen 100 #define yydefred c_yydefred 101 #define yydgoto c_yydgoto 102 #define yysindex c_yysindex 103 #define yyrindex c_yyrindex 104 #define yygindex c_yygindex 105 #define yytable c_yytable 106 #define yycheck c_yycheck 107 108 #ifndef YYDEBUG 109 #define YYDEBUG 1 /* Default to yydebug support */ 110 #endif 111 112 #define YYFPRINTF parser_fprintf 113 114 int yyparse (void); 115 116 static int yylex (void); 117 118 void yyerror (char *); 119 120 %} 121 122 /* Although the yacc "value" of an expression is not used, 123 since the result is stored in the structure being created, 124 other node types do have values. */ 125 126 %union 127 { 128 LONGEST lval; 129 struct { 130 LONGEST val; 131 struct type *type; 132 } typed_val_int; 133 struct { 134 DOUBLEST dval; 135 struct type *type; 136 } typed_val_float; 137 struct { 138 gdb_byte val[16]; 139 struct type *type; 140 } typed_val_decfloat; 141 struct symbol *sym; 142 struct type *tval; 143 struct stoken sval; 144 struct typed_stoken tsval; 145 struct ttype tsym; 146 struct symtoken ssym; 147 int voidval; 148 struct block *bval; 149 enum exp_opcode opcode; 150 struct internalvar *ivar; 151 152 struct stoken_vector svec; 153 struct type **tvec; 154 int *ivec; 155 } 156 157 %{ 158 /* YYSTYPE gets defined by %union */ 159 static int parse_number (char *, int, int, YYSTYPE *); 160 static struct stoken operator_stoken (const char *); 161 %} 162 163 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly 164 %type <lval> rcurly 165 %type <tval> type typebase 166 %type <tvec> nonempty_typelist 167 /* %type <bval> block */ 168 169 /* Fancy type parsing. */ 170 %type <voidval> func_mod direct_abs_decl abs_decl 171 %type <tval> ptype 172 %type <lval> array_mod 173 174 %token <typed_val_int> INT 175 %token <typed_val_float> FLOAT 176 %token <typed_val_decfloat> DECFLOAT 177 178 /* Both NAME and TYPENAME tokens represent symbols in the input, 179 and both convey their data as strings. 180 But a TYPENAME is a string that happens to be defined as a typedef 181 or builtin type name (such as int or char) 182 and a NAME is any other symbol. 183 Contexts where this distinction is not important can use the 184 nonterminal "name", which matches either NAME or TYPENAME. */ 185 186 %token <tsval> STRING 187 %token <tsval> CHAR 188 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */ 189 %token <ssym> UNKNOWN_CPP_NAME 190 %token <voidval> COMPLETE 191 %token <tsym> TYPENAME 192 %type <sval> name 193 %type <svec> string_exp 194 %type <ssym> name_not_typename 195 %type <tsym> typename 196 197 /* A NAME_OR_INT is a symbol which is not known in the symbol table, 198 but which would parse as a valid number in the current input radix. 199 E.g. "c" when input_radix==16. Depending on the parse, it will be 200 turned into a name or into a number. */ 201 202 %token <ssym> NAME_OR_INT 203 204 %token OPERATOR 205 %token STRUCT CLASS UNION ENUM SIZEOF UNSIGNED COLONCOLON 206 %token TEMPLATE 207 %token ERROR 208 %token NEW DELETE 209 %type <sval> operator 210 %token REINTERPRET_CAST DYNAMIC_CAST STATIC_CAST CONST_CAST 211 212 /* Special type cases, put in to allow the parser to distinguish different 213 legal basetypes. */ 214 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD 215 216 %token <sval> VARIABLE 217 218 %token <opcode> ASSIGN_MODIFY 219 220 /* C++ */ 221 %token TRUEKEYWORD 222 %token FALSEKEYWORD 223 224 225 %left ',' 226 %left ABOVE_COMMA 227 %right '=' ASSIGN_MODIFY 228 %right '?' 229 %left OROR 230 %left ANDAND 231 %left '|' 232 %left '^' 233 %left '&' 234 %left EQUAL NOTEQUAL 235 %left '<' '>' LEQ GEQ 236 %left LSH RSH 237 %left '@' 238 %left '+' '-' 239 %left '*' '/' '%' 240 %right UNARY INCREMENT DECREMENT 241 %right ARROW ARROW_STAR '.' DOT_STAR '[' '(' 242 %token <ssym> BLOCKNAME 243 %token <bval> FILENAME 244 %type <bval> block 245 %left COLONCOLON 246 247 248 %% 249 250 start : exp1 251 | type_exp 252 ; 253 254 type_exp: type 255 { write_exp_elt_opcode(OP_TYPE); 256 write_exp_elt_type($1); 257 write_exp_elt_opcode(OP_TYPE);} 258 ; 259 260 /* Expressions, including the comma operator. */ 261 exp1 : exp 262 | exp1 ',' exp 263 { write_exp_elt_opcode (BINOP_COMMA); } 264 ; 265 266 /* Expressions, not including the comma operator. */ 267 exp : '*' exp %prec UNARY 268 { write_exp_elt_opcode (UNOP_IND); } 269 ; 270 271 exp : '&' exp %prec UNARY 272 { write_exp_elt_opcode (UNOP_ADDR); } 273 ; 274 275 exp : '-' exp %prec UNARY 276 { write_exp_elt_opcode (UNOP_NEG); } 277 ; 278 279 exp : '+' exp %prec UNARY 280 { write_exp_elt_opcode (UNOP_PLUS); } 281 ; 282 283 exp : '!' exp %prec UNARY 284 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); } 285 ; 286 287 exp : '~' exp %prec UNARY 288 { write_exp_elt_opcode (UNOP_COMPLEMENT); } 289 ; 290 291 exp : INCREMENT exp %prec UNARY 292 { write_exp_elt_opcode (UNOP_PREINCREMENT); } 293 ; 294 295 exp : DECREMENT exp %prec UNARY 296 { write_exp_elt_opcode (UNOP_PREDECREMENT); } 297 ; 298 299 exp : exp INCREMENT %prec UNARY 300 { write_exp_elt_opcode (UNOP_POSTINCREMENT); } 301 ; 302 303 exp : exp DECREMENT %prec UNARY 304 { write_exp_elt_opcode (UNOP_POSTDECREMENT); } 305 ; 306 307 exp : SIZEOF exp %prec UNARY 308 { write_exp_elt_opcode (UNOP_SIZEOF); } 309 ; 310 311 exp : exp ARROW name 312 { write_exp_elt_opcode (STRUCTOP_PTR); 313 write_exp_string ($3); 314 write_exp_elt_opcode (STRUCTOP_PTR); } 315 ; 316 317 exp : exp ARROW name COMPLETE 318 { mark_struct_expression (); 319 write_exp_elt_opcode (STRUCTOP_PTR); 320 write_exp_string ($3); 321 write_exp_elt_opcode (STRUCTOP_PTR); } 322 ; 323 324 exp : exp ARROW COMPLETE 325 { struct stoken s; 326 mark_struct_expression (); 327 write_exp_elt_opcode (STRUCTOP_PTR); 328 s.ptr = ""; 329 s.length = 0; 330 write_exp_string (s); 331 write_exp_elt_opcode (STRUCTOP_PTR); } 332 ; 333 334 exp : exp ARROW qualified_name 335 { /* exp->type::name becomes exp->*(&type::name) */ 336 /* Note: this doesn't work if name is a 337 static member! FIXME */ 338 write_exp_elt_opcode (UNOP_ADDR); 339 write_exp_elt_opcode (STRUCTOP_MPTR); } 340 ; 341 342 exp : exp ARROW_STAR exp 343 { write_exp_elt_opcode (STRUCTOP_MPTR); } 344 ; 345 346 exp : exp '.' name 347 { write_exp_elt_opcode (STRUCTOP_STRUCT); 348 write_exp_string ($3); 349 write_exp_elt_opcode (STRUCTOP_STRUCT); } 350 ; 351 352 exp : exp '.' name COMPLETE 353 { mark_struct_expression (); 354 write_exp_elt_opcode (STRUCTOP_STRUCT); 355 write_exp_string ($3); 356 write_exp_elt_opcode (STRUCTOP_STRUCT); } 357 ; 358 359 exp : exp '.' COMPLETE 360 { struct stoken s; 361 mark_struct_expression (); 362 write_exp_elt_opcode (STRUCTOP_STRUCT); 363 s.ptr = ""; 364 s.length = 0; 365 write_exp_string (s); 366 write_exp_elt_opcode (STRUCTOP_STRUCT); } 367 ; 368 369 exp : exp '.' qualified_name 370 { /* exp.type::name becomes exp.*(&type::name) */ 371 /* Note: this doesn't work if name is a 372 static member! FIXME */ 373 write_exp_elt_opcode (UNOP_ADDR); 374 write_exp_elt_opcode (STRUCTOP_MEMBER); } 375 ; 376 377 exp : exp DOT_STAR exp 378 { write_exp_elt_opcode (STRUCTOP_MEMBER); } 379 ; 380 381 exp : exp '[' exp1 ']' 382 { write_exp_elt_opcode (BINOP_SUBSCRIPT); } 383 ; 384 385 exp : exp '(' 386 /* This is to save the value of arglist_len 387 being accumulated by an outer function call. */ 388 { start_arglist (); } 389 arglist ')' %prec ARROW 390 { write_exp_elt_opcode (OP_FUNCALL); 391 write_exp_elt_longcst ((LONGEST) end_arglist ()); 392 write_exp_elt_opcode (OP_FUNCALL); } 393 ; 394 395 exp : UNKNOWN_CPP_NAME '(' 396 { 397 /* This could potentially be a an argument defined 398 lookup function (Koenig). */ 399 write_exp_elt_opcode (OP_ADL_FUNC); 400 write_exp_elt_block (expression_context_block); 401 write_exp_elt_sym (NULL); /* Placeholder. */ 402 write_exp_string ($1.stoken); 403 write_exp_elt_opcode (OP_ADL_FUNC); 404 405 /* This is to save the value of arglist_len 406 being accumulated by an outer function call. */ 407 408 start_arglist (); 409 } 410 arglist ')' %prec ARROW 411 { 412 write_exp_elt_opcode (OP_FUNCALL); 413 write_exp_elt_longcst ((LONGEST) end_arglist ()); 414 write_exp_elt_opcode (OP_FUNCALL); 415 } 416 ; 417 418 lcurly : '{' 419 { start_arglist (); } 420 ; 421 422 arglist : 423 ; 424 425 arglist : exp 426 { arglist_len = 1; } 427 ; 428 429 arglist : arglist ',' exp %prec ABOVE_COMMA 430 { arglist_len++; } 431 ; 432 433 exp : exp '(' nonempty_typelist ')' const_or_volatile 434 { int i; 435 write_exp_elt_opcode (TYPE_INSTANCE); 436 write_exp_elt_longcst ((LONGEST) $<ivec>3[0]); 437 for (i = 0; i < $<ivec>3[0]; ++i) 438 write_exp_elt_type ($<tvec>3[i + 1]); 439 write_exp_elt_longcst((LONGEST) $<ivec>3[0]); 440 write_exp_elt_opcode (TYPE_INSTANCE); 441 free ($3); 442 } 443 ; 444 445 rcurly : '}' 446 { $$ = end_arglist () - 1; } 447 ; 448 exp : lcurly arglist rcurly %prec ARROW 449 { write_exp_elt_opcode (OP_ARRAY); 450 write_exp_elt_longcst ((LONGEST) 0); 451 write_exp_elt_longcst ((LONGEST) $3); 452 write_exp_elt_opcode (OP_ARRAY); } 453 ; 454 455 exp : lcurly type rcurly exp %prec UNARY 456 { write_exp_elt_opcode (UNOP_MEMVAL); 457 write_exp_elt_type ($2); 458 write_exp_elt_opcode (UNOP_MEMVAL); } 459 ; 460 461 exp : '(' type ')' exp %prec UNARY 462 { write_exp_elt_opcode (UNOP_CAST); 463 write_exp_elt_type ($2); 464 write_exp_elt_opcode (UNOP_CAST); } 465 ; 466 467 exp : '(' exp1 ')' 468 { } 469 ; 470 471 /* Binary operators in order of decreasing precedence. */ 472 473 exp : exp '@' exp 474 { write_exp_elt_opcode (BINOP_REPEAT); } 475 ; 476 477 exp : exp '*' exp 478 { write_exp_elt_opcode (BINOP_MUL); } 479 ; 480 481 exp : exp '/' exp 482 { write_exp_elt_opcode (BINOP_DIV); } 483 ; 484 485 exp : exp '%' exp 486 { write_exp_elt_opcode (BINOP_REM); } 487 ; 488 489 exp : exp '+' exp 490 { write_exp_elt_opcode (BINOP_ADD); } 491 ; 492 493 exp : exp '-' exp 494 { write_exp_elt_opcode (BINOP_SUB); } 495 ; 496 497 exp : exp LSH exp 498 { write_exp_elt_opcode (BINOP_LSH); } 499 ; 500 501 exp : exp RSH exp 502 { write_exp_elt_opcode (BINOP_RSH); } 503 ; 504 505 exp : exp EQUAL exp 506 { write_exp_elt_opcode (BINOP_EQUAL); } 507 ; 508 509 exp : exp NOTEQUAL exp 510 { write_exp_elt_opcode (BINOP_NOTEQUAL); } 511 ; 512 513 exp : exp LEQ exp 514 { write_exp_elt_opcode (BINOP_LEQ); } 515 ; 516 517 exp : exp GEQ exp 518 { write_exp_elt_opcode (BINOP_GEQ); } 519 ; 520 521 exp : exp '<' exp 522 { write_exp_elt_opcode (BINOP_LESS); } 523 ; 524 525 exp : exp '>' exp 526 { write_exp_elt_opcode (BINOP_GTR); } 527 ; 528 529 exp : exp '&' exp 530 { write_exp_elt_opcode (BINOP_BITWISE_AND); } 531 ; 532 533 exp : exp '^' exp 534 { write_exp_elt_opcode (BINOP_BITWISE_XOR); } 535 ; 536 537 exp : exp '|' exp 538 { write_exp_elt_opcode (BINOP_BITWISE_IOR); } 539 ; 540 541 exp : exp ANDAND exp 542 { write_exp_elt_opcode (BINOP_LOGICAL_AND); } 543 ; 544 545 exp : exp OROR exp 546 { write_exp_elt_opcode (BINOP_LOGICAL_OR); } 547 ; 548 549 exp : exp '?' exp ':' exp %prec '?' 550 { write_exp_elt_opcode (TERNOP_COND); } 551 ; 552 553 exp : exp '=' exp 554 { write_exp_elt_opcode (BINOP_ASSIGN); } 555 ; 556 557 exp : exp ASSIGN_MODIFY exp 558 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); 559 write_exp_elt_opcode ($2); 560 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); } 561 ; 562 563 exp : INT 564 { write_exp_elt_opcode (OP_LONG); 565 write_exp_elt_type ($1.type); 566 write_exp_elt_longcst ((LONGEST)($1.val)); 567 write_exp_elt_opcode (OP_LONG); } 568 ; 569 570 exp : CHAR 571 { 572 struct stoken_vector vec; 573 vec.len = 1; 574 vec.tokens = &$1; 575 write_exp_string_vector ($1.type, &vec); 576 } 577 ; 578 579 exp : NAME_OR_INT 580 { YYSTYPE val; 581 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val); 582 write_exp_elt_opcode (OP_LONG); 583 write_exp_elt_type (val.typed_val_int.type); 584 write_exp_elt_longcst ((LONGEST)val.typed_val_int.val); 585 write_exp_elt_opcode (OP_LONG); 586 } 587 ; 588 589 590 exp : FLOAT 591 { write_exp_elt_opcode (OP_DOUBLE); 592 write_exp_elt_type ($1.type); 593 write_exp_elt_dblcst ($1.dval); 594 write_exp_elt_opcode (OP_DOUBLE); } 595 ; 596 597 exp : DECFLOAT 598 { write_exp_elt_opcode (OP_DECFLOAT); 599 write_exp_elt_type ($1.type); 600 write_exp_elt_decfloatcst ($1.val); 601 write_exp_elt_opcode (OP_DECFLOAT); } 602 ; 603 604 exp : variable 605 ; 606 607 exp : VARIABLE 608 { 609 write_dollar_variable ($1); 610 } 611 ; 612 613 exp : SIZEOF '(' type ')' %prec UNARY 614 { write_exp_elt_opcode (OP_LONG); 615 write_exp_elt_type (lookup_signed_typename 616 (parse_language, parse_gdbarch, 617 "int")); 618 CHECK_TYPEDEF ($3); 619 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3)); 620 write_exp_elt_opcode (OP_LONG); } 621 ; 622 623 exp : REINTERPRET_CAST '<' type '>' '(' exp ')' %prec UNARY 624 { write_exp_elt_opcode (UNOP_REINTERPRET_CAST); 625 write_exp_elt_type ($3); 626 write_exp_elt_opcode (UNOP_REINTERPRET_CAST); } 627 ; 628 629 exp : STATIC_CAST '<' type '>' '(' exp ')' %prec UNARY 630 { write_exp_elt_opcode (UNOP_CAST); 631 write_exp_elt_type ($3); 632 write_exp_elt_opcode (UNOP_CAST); } 633 ; 634 635 exp : DYNAMIC_CAST '<' type '>' '(' exp ')' %prec UNARY 636 { write_exp_elt_opcode (UNOP_DYNAMIC_CAST); 637 write_exp_elt_type ($3); 638 write_exp_elt_opcode (UNOP_DYNAMIC_CAST); } 639 ; 640 641 exp : CONST_CAST '<' type '>' '(' exp ')' %prec UNARY 642 { /* We could do more error checking here, but 643 it doesn't seem worthwhile. */ 644 write_exp_elt_opcode (UNOP_CAST); 645 write_exp_elt_type ($3); 646 write_exp_elt_opcode (UNOP_CAST); } 647 ; 648 649 string_exp: 650 STRING 651 { 652 /* We copy the string here, and not in the 653 lexer, to guarantee that we do not leak a 654 string. Note that we follow the 655 NUL-termination convention of the 656 lexer. */ 657 struct typed_stoken *vec = XNEW (struct typed_stoken); 658 $$.len = 1; 659 $$.tokens = vec; 660 661 vec->type = $1.type; 662 vec->length = $1.length; 663 vec->ptr = malloc ($1.length + 1); 664 memcpy (vec->ptr, $1.ptr, $1.length + 1); 665 } 666 667 | string_exp STRING 668 { 669 /* Note that we NUL-terminate here, but just 670 for convenience. */ 671 char *p; 672 ++$$.len; 673 $$.tokens = realloc ($$.tokens, 674 $$.len * sizeof (struct typed_stoken)); 675 676 p = malloc ($2.length + 1); 677 memcpy (p, $2.ptr, $2.length + 1); 678 679 $$.tokens[$$.len - 1].type = $2.type; 680 $$.tokens[$$.len - 1].length = $2.length; 681 $$.tokens[$$.len - 1].ptr = p; 682 } 683 ; 684 685 exp : string_exp 686 { 687 int i; 688 enum c_string_type type = C_STRING; 689 690 for (i = 0; i < $1.len; ++i) 691 { 692 switch ($1.tokens[i].type) 693 { 694 case C_STRING: 695 break; 696 case C_WIDE_STRING: 697 case C_STRING_16: 698 case C_STRING_32: 699 if (type != C_STRING 700 && type != $1.tokens[i].type) 701 error (_("Undefined string concatenation.")); 702 type = $1.tokens[i].type; 703 break; 704 default: 705 /* internal error */ 706 internal_error (__FILE__, __LINE__, 707 "unrecognized type in string concatenation"); 708 } 709 } 710 711 write_exp_string_vector (type, &$1); 712 for (i = 0; i < $1.len; ++i) 713 free ($1.tokens[i].ptr); 714 free ($1.tokens); 715 } 716 ; 717 718 /* C++. */ 719 exp : TRUEKEYWORD 720 { write_exp_elt_opcode (OP_LONG); 721 write_exp_elt_type (parse_type->builtin_bool); 722 write_exp_elt_longcst ((LONGEST) 1); 723 write_exp_elt_opcode (OP_LONG); } 724 ; 725 726 exp : FALSEKEYWORD 727 { write_exp_elt_opcode (OP_LONG); 728 write_exp_elt_type (parse_type->builtin_bool); 729 write_exp_elt_longcst ((LONGEST) 0); 730 write_exp_elt_opcode (OP_LONG); } 731 ; 732 733 /* end of C++. */ 734 735 block : BLOCKNAME 736 { 737 if ($1.sym) 738 $$ = SYMBOL_BLOCK_VALUE ($1.sym); 739 else 740 error (_("No file or function \"%s\"."), 741 copy_name ($1.stoken)); 742 } 743 | FILENAME 744 { 745 $$ = $1; 746 } 747 ; 748 749 block : block COLONCOLON name 750 { struct symbol *tem 751 = lookup_symbol (copy_name ($3), $1, 752 VAR_DOMAIN, (int *) NULL); 753 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK) 754 error (_("No function \"%s\" in specified context."), 755 copy_name ($3)); 756 $$ = SYMBOL_BLOCK_VALUE (tem); } 757 ; 758 759 variable: block COLONCOLON name 760 { struct symbol *sym; 761 sym = lookup_symbol (copy_name ($3), $1, 762 VAR_DOMAIN, (int *) NULL); 763 if (sym == 0) 764 error (_("No symbol \"%s\" in specified context."), 765 copy_name ($3)); 766 767 write_exp_elt_opcode (OP_VAR_VALUE); 768 /* block_found is set by lookup_symbol. */ 769 write_exp_elt_block (block_found); 770 write_exp_elt_sym (sym); 771 write_exp_elt_opcode (OP_VAR_VALUE); } 772 ; 773 774 qualified_name: TYPENAME COLONCOLON name 775 { 776 struct type *type = $1.type; 777 CHECK_TYPEDEF (type); 778 if (TYPE_CODE (type) != TYPE_CODE_STRUCT 779 && TYPE_CODE (type) != TYPE_CODE_UNION 780 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE) 781 error (_("`%s' is not defined as an aggregate type."), 782 TYPE_NAME (type)); 783 784 write_exp_elt_opcode (OP_SCOPE); 785 write_exp_elt_type (type); 786 write_exp_string ($3); 787 write_exp_elt_opcode (OP_SCOPE); 788 } 789 | TYPENAME COLONCOLON '~' name 790 { 791 struct type *type = $1.type; 792 struct stoken tmp_token; 793 CHECK_TYPEDEF (type); 794 if (TYPE_CODE (type) != TYPE_CODE_STRUCT 795 && TYPE_CODE (type) != TYPE_CODE_UNION 796 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE) 797 error (_("`%s' is not defined as an aggregate type."), 798 TYPE_NAME (type)); 799 800 tmp_token.ptr = (char*) alloca ($4.length + 2); 801 tmp_token.length = $4.length + 1; 802 tmp_token.ptr[0] = '~'; 803 memcpy (tmp_token.ptr+1, $4.ptr, $4.length); 804 tmp_token.ptr[tmp_token.length] = 0; 805 806 /* Check for valid destructor name. */ 807 destructor_name_p (tmp_token.ptr, type); 808 write_exp_elt_opcode (OP_SCOPE); 809 write_exp_elt_type (type); 810 write_exp_string (tmp_token); 811 write_exp_elt_opcode (OP_SCOPE); 812 } 813 | TYPENAME COLONCOLON name COLONCOLON name 814 { 815 char *copy = copy_name ($3); 816 error (_("No type \"%s\" within class " 817 "or namespace \"%s\"."), 818 copy, TYPE_NAME ($1.type)); 819 } 820 ; 821 822 variable: qualified_name 823 | COLONCOLON name_not_typename 824 { 825 char *name = copy_name ($2.stoken); 826 struct symbol *sym; 827 struct minimal_symbol *msymbol; 828 829 sym = 830 lookup_symbol (name, (const struct block *) NULL, 831 VAR_DOMAIN, (int *) NULL); 832 if (sym) 833 { 834 write_exp_elt_opcode (OP_VAR_VALUE); 835 write_exp_elt_block (NULL); 836 write_exp_elt_sym (sym); 837 write_exp_elt_opcode (OP_VAR_VALUE); 838 break; 839 } 840 841 msymbol = lookup_minimal_symbol (name, NULL, NULL); 842 if (msymbol != NULL) 843 write_exp_msymbol (msymbol); 844 else if (!have_full_symbols () && !have_partial_symbols ()) 845 error (_("No symbol table is loaded. Use the \"file\" command.")); 846 else 847 error (_("No symbol \"%s\" in current context."), name); 848 } 849 ; 850 851 variable: name_not_typename 852 { struct symbol *sym = $1.sym; 853 854 if (sym) 855 { 856 if (symbol_read_needs_frame (sym)) 857 { 858 if (innermost_block == 0 859 || contained_in (block_found, 860 innermost_block)) 861 innermost_block = block_found; 862 } 863 864 write_exp_elt_opcode (OP_VAR_VALUE); 865 /* We want to use the selected frame, not 866 another more inner frame which happens to 867 be in the same block. */ 868 write_exp_elt_block (NULL); 869 write_exp_elt_sym (sym); 870 write_exp_elt_opcode (OP_VAR_VALUE); 871 } 872 else if ($1.is_a_field_of_this) 873 { 874 /* C++: it hangs off of `this'. Must 875 not inadvertently convert from a method call 876 to data ref. */ 877 if (innermost_block == 0 878 || contained_in (block_found, 879 innermost_block)) 880 innermost_block = block_found; 881 write_exp_elt_opcode (OP_THIS); 882 write_exp_elt_opcode (OP_THIS); 883 write_exp_elt_opcode (STRUCTOP_PTR); 884 write_exp_string ($1.stoken); 885 write_exp_elt_opcode (STRUCTOP_PTR); 886 } 887 else 888 { 889 struct minimal_symbol *msymbol; 890 char *arg = copy_name ($1.stoken); 891 892 msymbol = 893 lookup_minimal_symbol (arg, NULL, NULL); 894 if (msymbol != NULL) 895 write_exp_msymbol (msymbol); 896 else if (!have_full_symbols () && !have_partial_symbols ()) 897 error (_("No symbol table is loaded. Use the \"file\" command.")); 898 else 899 error (_("No symbol \"%s\" in current context."), 900 copy_name ($1.stoken)); 901 } 902 } 903 ; 904 905 space_identifier : '@' NAME 906 { push_type_address_space (copy_name ($2.stoken)); 907 push_type (tp_space_identifier); 908 } 909 ; 910 911 const_or_volatile: const_or_volatile_noopt 912 | 913 ; 914 915 cv_with_space_id : const_or_volatile space_identifier const_or_volatile 916 ; 917 918 const_or_volatile_or_space_identifier_noopt: cv_with_space_id 919 | const_or_volatile_noopt 920 ; 921 922 const_or_volatile_or_space_identifier: 923 const_or_volatile_or_space_identifier_noopt 924 | 925 ; 926 927 abs_decl: '*' 928 { push_type (tp_pointer); $$ = 0; } 929 | '*' abs_decl 930 { push_type (tp_pointer); $$ = $2; } 931 | '&' 932 { push_type (tp_reference); $$ = 0; } 933 | '&' abs_decl 934 { push_type (tp_reference); $$ = $2; } 935 | direct_abs_decl 936 ; 937 938 direct_abs_decl: '(' abs_decl ')' 939 { $$ = $2; } 940 | direct_abs_decl array_mod 941 { 942 push_type_int ($2); 943 push_type (tp_array); 944 } 945 | array_mod 946 { 947 push_type_int ($1); 948 push_type (tp_array); 949 $$ = 0; 950 } 951 952 | direct_abs_decl func_mod 953 { push_type (tp_function); } 954 | func_mod 955 { push_type (tp_function); } 956 ; 957 958 array_mod: '[' ']' 959 { $$ = -1; } 960 | '[' INT ']' 961 { $$ = $2.val; } 962 ; 963 964 func_mod: '(' ')' 965 { $$ = 0; } 966 | '(' nonempty_typelist ')' 967 { free ($2); $$ = 0; } 968 ; 969 970 /* We used to try to recognize pointer to member types here, but 971 that didn't work (shift/reduce conflicts meant that these rules never 972 got executed). The problem is that 973 int (foo::bar::baz::bizzle) 974 is a function type but 975 int (foo::bar::baz::bizzle::*) 976 is a pointer to member type. Stroustrup loses again! */ 977 978 type : ptype 979 ; 980 981 typebase /* Implements (approximately): (type-qualifier)* type-specifier */ 982 : TYPENAME 983 { $$ = $1.type; } 984 | INT_KEYWORD 985 { $$ = lookup_signed_typename (parse_language, 986 parse_gdbarch, 987 "int"); } 988 | LONG 989 { $$ = lookup_signed_typename (parse_language, 990 parse_gdbarch, 991 "long"); } 992 | SHORT 993 { $$ = lookup_signed_typename (parse_language, 994 parse_gdbarch, 995 "short"); } 996 | LONG INT_KEYWORD 997 { $$ = lookup_signed_typename (parse_language, 998 parse_gdbarch, 999 "long"); } 1000 | LONG SIGNED_KEYWORD INT_KEYWORD 1001 { $$ = lookup_signed_typename (parse_language, 1002 parse_gdbarch, 1003 "long"); } 1004 | LONG SIGNED_KEYWORD 1005 { $$ = lookup_signed_typename (parse_language, 1006 parse_gdbarch, 1007 "long"); } 1008 | SIGNED_KEYWORD LONG INT_KEYWORD 1009 { $$ = lookup_signed_typename (parse_language, 1010 parse_gdbarch, 1011 "long"); } 1012 | UNSIGNED LONG INT_KEYWORD 1013 { $$ = lookup_unsigned_typename (parse_language, 1014 parse_gdbarch, 1015 "long"); } 1016 | LONG UNSIGNED INT_KEYWORD 1017 { $$ = lookup_unsigned_typename (parse_language, 1018 parse_gdbarch, 1019 "long"); } 1020 | LONG UNSIGNED 1021 { $$ = lookup_unsigned_typename (parse_language, 1022 parse_gdbarch, 1023 "long"); } 1024 | LONG LONG 1025 { $$ = lookup_signed_typename (parse_language, 1026 parse_gdbarch, 1027 "long long"); } 1028 | LONG LONG INT_KEYWORD 1029 { $$ = lookup_signed_typename (parse_language, 1030 parse_gdbarch, 1031 "long long"); } 1032 | LONG LONG SIGNED_KEYWORD INT_KEYWORD 1033 { $$ = lookup_signed_typename (parse_language, 1034 parse_gdbarch, 1035 "long long"); } 1036 | LONG LONG SIGNED_KEYWORD 1037 { $$ = lookup_signed_typename (parse_language, 1038 parse_gdbarch, 1039 "long long"); } 1040 | SIGNED_KEYWORD LONG LONG 1041 { $$ = lookup_signed_typename (parse_language, 1042 parse_gdbarch, 1043 "long long"); } 1044 | SIGNED_KEYWORD LONG LONG INT_KEYWORD 1045 { $$ = lookup_signed_typename (parse_language, 1046 parse_gdbarch, 1047 "long long"); } 1048 | UNSIGNED LONG LONG 1049 { $$ = lookup_unsigned_typename (parse_language, 1050 parse_gdbarch, 1051 "long long"); } 1052 | UNSIGNED LONG LONG INT_KEYWORD 1053 { $$ = lookup_unsigned_typename (parse_language, 1054 parse_gdbarch, 1055 "long long"); } 1056 | LONG LONG UNSIGNED 1057 { $$ = lookup_unsigned_typename (parse_language, 1058 parse_gdbarch, 1059 "long long"); } 1060 | LONG LONG UNSIGNED INT_KEYWORD 1061 { $$ = lookup_unsigned_typename (parse_language, 1062 parse_gdbarch, 1063 "long long"); } 1064 | SHORT INT_KEYWORD 1065 { $$ = lookup_signed_typename (parse_language, 1066 parse_gdbarch, 1067 "short"); } 1068 | SHORT SIGNED_KEYWORD INT_KEYWORD 1069 { $$ = lookup_signed_typename (parse_language, 1070 parse_gdbarch, 1071 "short"); } 1072 | SHORT SIGNED_KEYWORD 1073 { $$ = lookup_signed_typename (parse_language, 1074 parse_gdbarch, 1075 "short"); } 1076 | UNSIGNED SHORT INT_KEYWORD 1077 { $$ = lookup_unsigned_typename (parse_language, 1078 parse_gdbarch, 1079 "short"); } 1080 | SHORT UNSIGNED 1081 { $$ = lookup_unsigned_typename (parse_language, 1082 parse_gdbarch, 1083 "short"); } 1084 | SHORT UNSIGNED INT_KEYWORD 1085 { $$ = lookup_unsigned_typename (parse_language, 1086 parse_gdbarch, 1087 "short"); } 1088 | DOUBLE_KEYWORD 1089 { $$ = lookup_typename (parse_language, parse_gdbarch, 1090 "double", (struct block *) NULL, 1091 0); } 1092 | LONG DOUBLE_KEYWORD 1093 { $$ = lookup_typename (parse_language, parse_gdbarch, 1094 "long double", 1095 (struct block *) NULL, 0); } 1096 | STRUCT name 1097 { $$ = lookup_struct (copy_name ($2), 1098 expression_context_block); } 1099 | CLASS name 1100 { $$ = lookup_struct (copy_name ($2), 1101 expression_context_block); } 1102 | UNION name 1103 { $$ = lookup_union (copy_name ($2), 1104 expression_context_block); } 1105 | ENUM name 1106 { $$ = lookup_enum (copy_name ($2), 1107 expression_context_block); } 1108 | UNSIGNED typename 1109 { $$ = lookup_unsigned_typename (parse_language, 1110 parse_gdbarch, 1111 TYPE_NAME($2.type)); } 1112 | UNSIGNED 1113 { $$ = lookup_unsigned_typename (parse_language, 1114 parse_gdbarch, 1115 "int"); } 1116 | SIGNED_KEYWORD typename 1117 { $$ = lookup_signed_typename (parse_language, 1118 parse_gdbarch, 1119 TYPE_NAME($2.type)); } 1120 | SIGNED_KEYWORD 1121 { $$ = lookup_signed_typename (parse_language, 1122 parse_gdbarch, 1123 "int"); } 1124 /* It appears that this rule for templates is never 1125 reduced; template recognition happens by lookahead 1126 in the token processing code in yylex. */ 1127 | TEMPLATE name '<' type '>' 1128 { $$ = lookup_template_type(copy_name($2), $4, 1129 expression_context_block); 1130 } 1131 | const_or_volatile_or_space_identifier_noopt typebase 1132 { $$ = follow_types ($2); } 1133 | typebase const_or_volatile_or_space_identifier_noopt 1134 { $$ = follow_types ($1); } 1135 ; 1136 1137 typename: TYPENAME 1138 | INT_KEYWORD 1139 { 1140 $$.stoken.ptr = "int"; 1141 $$.stoken.length = 3; 1142 $$.type = lookup_signed_typename (parse_language, 1143 parse_gdbarch, 1144 "int"); 1145 } 1146 | LONG 1147 { 1148 $$.stoken.ptr = "long"; 1149 $$.stoken.length = 4; 1150 $$.type = lookup_signed_typename (parse_language, 1151 parse_gdbarch, 1152 "long"); 1153 } 1154 | SHORT 1155 { 1156 $$.stoken.ptr = "short"; 1157 $$.stoken.length = 5; 1158 $$.type = lookup_signed_typename (parse_language, 1159 parse_gdbarch, 1160 "short"); 1161 } 1162 ; 1163 1164 nonempty_typelist 1165 : type 1166 { $$ = (struct type **) malloc (sizeof (struct type *) * 2); 1167 $<ivec>$[0] = 1; /* Number of types in vector */ 1168 $$[1] = $1; 1169 } 1170 | nonempty_typelist ',' type 1171 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1); 1172 $$ = (struct type **) realloc ((char *) $1, len); 1173 $$[$<ivec>$[0]] = $3; 1174 } 1175 ; 1176 1177 ptype : typebase 1178 | ptype const_or_volatile_or_space_identifier abs_decl const_or_volatile_or_space_identifier 1179 { $$ = follow_types ($1); } 1180 ; 1181 1182 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD 1183 | VOLATILE_KEYWORD CONST_KEYWORD 1184 ; 1185 1186 const_or_volatile_noopt: const_and_volatile 1187 { push_type (tp_const); 1188 push_type (tp_volatile); 1189 } 1190 | CONST_KEYWORD 1191 { push_type (tp_const); } 1192 | VOLATILE_KEYWORD 1193 { push_type (tp_volatile); } 1194 ; 1195 1196 operator: OPERATOR NEW 1197 { $$ = operator_stoken (" new"); } 1198 | OPERATOR DELETE 1199 { $$ = operator_stoken (" delete"); } 1200 | OPERATOR NEW '[' ']' 1201 { $$ = operator_stoken (" new[]"); } 1202 | OPERATOR DELETE '[' ']' 1203 { $$ = operator_stoken (" delete[]"); } 1204 | OPERATOR '+' 1205 { $$ = operator_stoken ("+"); } 1206 | OPERATOR '-' 1207 { $$ = operator_stoken ("-"); } 1208 | OPERATOR '*' 1209 { $$ = operator_stoken ("*"); } 1210 | OPERATOR '/' 1211 { $$ = operator_stoken ("/"); } 1212 | OPERATOR '%' 1213 { $$ = operator_stoken ("%"); } 1214 | OPERATOR '^' 1215 { $$ = operator_stoken ("^"); } 1216 | OPERATOR '&' 1217 { $$ = operator_stoken ("&"); } 1218 | OPERATOR '|' 1219 { $$ = operator_stoken ("|"); } 1220 | OPERATOR '~' 1221 { $$ = operator_stoken ("~"); } 1222 | OPERATOR '!' 1223 { $$ = operator_stoken ("!"); } 1224 | OPERATOR '=' 1225 { $$ = operator_stoken ("="); } 1226 | OPERATOR '<' 1227 { $$ = operator_stoken ("<"); } 1228 | OPERATOR '>' 1229 { $$ = operator_stoken (">"); } 1230 | OPERATOR ASSIGN_MODIFY 1231 { const char *op = "unknown"; 1232 switch ($2) 1233 { 1234 case BINOP_RSH: 1235 op = ">>="; 1236 break; 1237 case BINOP_LSH: 1238 op = "<<="; 1239 break; 1240 case BINOP_ADD: 1241 op = "+="; 1242 break; 1243 case BINOP_SUB: 1244 op = "-="; 1245 break; 1246 case BINOP_MUL: 1247 op = "*="; 1248 break; 1249 case BINOP_DIV: 1250 op = "/="; 1251 break; 1252 case BINOP_REM: 1253 op = "%="; 1254 break; 1255 case BINOP_BITWISE_IOR: 1256 op = "|="; 1257 break; 1258 case BINOP_BITWISE_AND: 1259 op = "&="; 1260 break; 1261 case BINOP_BITWISE_XOR: 1262 op = "^="; 1263 break; 1264 default: 1265 break; 1266 } 1267 1268 $$ = operator_stoken (op); 1269 } 1270 | OPERATOR LSH 1271 { $$ = operator_stoken ("<<"); } 1272 | OPERATOR RSH 1273 { $$ = operator_stoken (">>"); } 1274 | OPERATOR EQUAL 1275 { $$ = operator_stoken ("=="); } 1276 | OPERATOR NOTEQUAL 1277 { $$ = operator_stoken ("!="); } 1278 | OPERATOR LEQ 1279 { $$ = operator_stoken ("<="); } 1280 | OPERATOR GEQ 1281 { $$ = operator_stoken (">="); } 1282 | OPERATOR ANDAND 1283 { $$ = operator_stoken ("&&"); } 1284 | OPERATOR OROR 1285 { $$ = operator_stoken ("||"); } 1286 | OPERATOR INCREMENT 1287 { $$ = operator_stoken ("++"); } 1288 | OPERATOR DECREMENT 1289 { $$ = operator_stoken ("--"); } 1290 | OPERATOR ',' 1291 { $$ = operator_stoken (","); } 1292 | OPERATOR ARROW_STAR 1293 { $$ = operator_stoken ("->*"); } 1294 | OPERATOR ARROW 1295 { $$ = operator_stoken ("->"); } 1296 | OPERATOR '(' ')' 1297 { $$ = operator_stoken ("()"); } 1298 | OPERATOR '[' ']' 1299 { $$ = operator_stoken ("[]"); } 1300 | OPERATOR ptype 1301 { char *name; 1302 long length; 1303 struct ui_file *buf = mem_fileopen (); 1304 1305 c_print_type ($2, NULL, buf, -1, 0); 1306 name = ui_file_xstrdup (buf, &length); 1307 ui_file_delete (buf); 1308 $$ = operator_stoken (name); 1309 free (name); 1310 } 1311 ; 1312 1313 1314 1315 name : NAME { $$ = $1.stoken; } 1316 | BLOCKNAME { $$ = $1.stoken; } 1317 | TYPENAME { $$ = $1.stoken; } 1318 | NAME_OR_INT { $$ = $1.stoken; } 1319 | UNKNOWN_CPP_NAME { $$ = $1.stoken; } 1320 | operator { $$ = $1; } 1321 ; 1322 1323 name_not_typename : NAME 1324 | BLOCKNAME 1325 /* These would be useful if name_not_typename was useful, but it is just 1326 a fake for "variable", so these cause reduce/reduce conflicts because 1327 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable, 1328 =exp) or just an exp. If name_not_typename was ever used in an lvalue 1329 context where only a name could occur, this might be useful. 1330 | NAME_OR_INT 1331 */ 1332 | operator 1333 { 1334 $$.stoken = $1; 1335 $$.sym = lookup_symbol ($1.ptr, 1336 expression_context_block, 1337 VAR_DOMAIN, 1338 &$$.is_a_field_of_this); 1339 } 1340 | UNKNOWN_CPP_NAME 1341 ; 1342 1343 %% 1344 1345 /* Returns a stoken of the operator name given by OP (which does not 1346 include the string "operator"). */ 1347 static struct stoken 1348 operator_stoken (const char *op) 1349 { 1350 static const char *operator_string = "operator"; 1351 struct stoken st = { NULL, 0 }; 1352 st.length = strlen (operator_string) + strlen (op); 1353 st.ptr = malloc (st.length + 1); 1354 strcpy (st.ptr, operator_string); 1355 strcat (st.ptr, op); 1356 1357 /* The toplevel (c_parse) will free the memory allocated here. */ 1358 make_cleanup (free, st.ptr); 1359 return st; 1360 }; 1361 1362 /* Take care of parsing a number (anything that starts with a digit). 1363 Set yylval and return the token type; update lexptr. 1364 LEN is the number of characters in it. */ 1365 1366 /*** Needs some error checking for the float case ***/ 1367 1368 static int 1369 parse_number (char *p, int len, int parsed_float, YYSTYPE *putithere) 1370 { 1371 /* FIXME: Shouldn't these be unsigned? We don't deal with negative values 1372 here, and we do kind of silly things like cast to unsigned. */ 1373 LONGEST n = 0; 1374 LONGEST prevn = 0; 1375 ULONGEST un; 1376 1377 int i = 0; 1378 int c; 1379 int base = input_radix; 1380 int unsigned_p = 0; 1381 1382 /* Number of "L" suffixes encountered. */ 1383 int long_p = 0; 1384 1385 /* We have found a "L" or "U" suffix. */ 1386 int found_suffix = 0; 1387 1388 ULONGEST high_bit; 1389 struct type *signed_type; 1390 struct type *unsigned_type; 1391 1392 if (parsed_float) 1393 { 1394 const char *suffix; 1395 int suffix_len; 1396 1397 /* If it ends at "df", "dd" or "dl", take it as type of decimal floating 1398 point. Return DECFLOAT. */ 1399 1400 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f') 1401 { 1402 p[len - 2] = '\0'; 1403 putithere->typed_val_decfloat.type 1404 = parse_type->builtin_decfloat; 1405 decimal_from_string (putithere->typed_val_decfloat.val, 4, 1406 gdbarch_byte_order (parse_gdbarch), p); 1407 p[len - 2] = 'd'; 1408 return DECFLOAT; 1409 } 1410 1411 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd') 1412 { 1413 p[len - 2] = '\0'; 1414 putithere->typed_val_decfloat.type 1415 = parse_type->builtin_decdouble; 1416 decimal_from_string (putithere->typed_val_decfloat.val, 8, 1417 gdbarch_byte_order (parse_gdbarch), p); 1418 p[len - 2] = 'd'; 1419 return DECFLOAT; 1420 } 1421 1422 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l') 1423 { 1424 p[len - 2] = '\0'; 1425 putithere->typed_val_decfloat.type 1426 = parse_type->builtin_declong; 1427 decimal_from_string (putithere->typed_val_decfloat.val, 16, 1428 gdbarch_byte_order (parse_gdbarch), p); 1429 p[len - 2] = 'd'; 1430 return DECFLOAT; 1431 } 1432 1433 if (! parse_c_float (parse_gdbarch, p, len, 1434 &putithere->typed_val_float.dval, 1435 &putithere->typed_val_float.type)) 1436 return ERROR; 1437 return FLOAT; 1438 } 1439 1440 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */ 1441 if (p[0] == '0') 1442 switch (p[1]) 1443 { 1444 case 'x': 1445 case 'X': 1446 if (len >= 3) 1447 { 1448 p += 2; 1449 base = 16; 1450 len -= 2; 1451 } 1452 break; 1453 1454 case 'b': 1455 case 'B': 1456 if (len >= 3) 1457 { 1458 p += 2; 1459 base = 2; 1460 len -= 2; 1461 } 1462 break; 1463 1464 case 't': 1465 case 'T': 1466 case 'd': 1467 case 'D': 1468 if (len >= 3) 1469 { 1470 p += 2; 1471 base = 10; 1472 len -= 2; 1473 } 1474 break; 1475 1476 default: 1477 base = 8; 1478 break; 1479 } 1480 1481 while (len-- > 0) 1482 { 1483 c = *p++; 1484 if (c >= 'A' && c <= 'Z') 1485 c += 'a' - 'A'; 1486 if (c != 'l' && c != 'u') 1487 n *= base; 1488 if (c >= '0' && c <= '9') 1489 { 1490 if (found_suffix) 1491 return ERROR; 1492 n += i = c - '0'; 1493 } 1494 else 1495 { 1496 if (base > 10 && c >= 'a' && c <= 'f') 1497 { 1498 if (found_suffix) 1499 return ERROR; 1500 n += i = c - 'a' + 10; 1501 } 1502 else if (c == 'l') 1503 { 1504 ++long_p; 1505 found_suffix = 1; 1506 } 1507 else if (c == 'u') 1508 { 1509 unsigned_p = 1; 1510 found_suffix = 1; 1511 } 1512 else 1513 return ERROR; /* Char not a digit */ 1514 } 1515 if (i >= base) 1516 return ERROR; /* Invalid digit in this base */ 1517 1518 /* Portably test for overflow (only works for nonzero values, so make 1519 a second check for zero). FIXME: Can't we just make n and prevn 1520 unsigned and avoid this? */ 1521 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0) 1522 unsigned_p = 1; /* Try something unsigned */ 1523 1524 /* Portably test for unsigned overflow. 1525 FIXME: This check is wrong; for example it doesn't find overflow 1526 on 0x123456789 when LONGEST is 32 bits. */ 1527 if (c != 'l' && c != 'u' && n != 0) 1528 { 1529 if ((unsigned_p && (ULONGEST) prevn >= (ULONGEST) n)) 1530 error (_("Numeric constant too large.")); 1531 } 1532 prevn = n; 1533 } 1534 1535 /* An integer constant is an int, a long, or a long long. An L 1536 suffix forces it to be long; an LL suffix forces it to be long 1537 long. If not forced to a larger size, it gets the first type of 1538 the above that it fits in. To figure out whether it fits, we 1539 shift it right and see whether anything remains. Note that we 1540 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one 1541 operation, because many compilers will warn about such a shift 1542 (which always produces a zero result). Sometimes gdbarch_int_bit 1543 or gdbarch_long_bit will be that big, sometimes not. To deal with 1544 the case where it is we just always shift the value more than 1545 once, with fewer bits each time. */ 1546 1547 un = (ULONGEST)n >> 2; 1548 if (long_p == 0 1549 && (un >> (gdbarch_int_bit (parse_gdbarch) - 2)) == 0) 1550 { 1551 high_bit = ((ULONGEST)1) << (gdbarch_int_bit (parse_gdbarch) - 1); 1552 1553 /* A large decimal (not hex or octal) constant (between INT_MAX 1554 and UINT_MAX) is a long or unsigned long, according to ANSI, 1555 never an unsigned int, but this code treats it as unsigned 1556 int. This probably should be fixed. GCC gives a warning on 1557 such constants. */ 1558 1559 unsigned_type = parse_type->builtin_unsigned_int; 1560 signed_type = parse_type->builtin_int; 1561 } 1562 else if (long_p <= 1 1563 && (un >> (gdbarch_long_bit (parse_gdbarch) - 2)) == 0) 1564 { 1565 high_bit = ((ULONGEST)1) << (gdbarch_long_bit (parse_gdbarch) - 1); 1566 unsigned_type = parse_type->builtin_unsigned_long; 1567 signed_type = parse_type->builtin_long; 1568 } 1569 else 1570 { 1571 int shift; 1572 if (sizeof (ULONGEST) * HOST_CHAR_BIT 1573 < gdbarch_long_long_bit (parse_gdbarch)) 1574 /* A long long does not fit in a LONGEST. */ 1575 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1); 1576 else 1577 shift = (gdbarch_long_long_bit (parse_gdbarch) - 1); 1578 high_bit = (ULONGEST) 1 << shift; 1579 unsigned_type = parse_type->builtin_unsigned_long_long; 1580 signed_type = parse_type->builtin_long_long; 1581 } 1582 1583 putithere->typed_val_int.val = n; 1584 1585 /* If the high bit of the worked out type is set then this number 1586 has to be unsigned. */ 1587 1588 if (unsigned_p || (n & high_bit)) 1589 { 1590 putithere->typed_val_int.type = unsigned_type; 1591 } 1592 else 1593 { 1594 putithere->typed_val_int.type = signed_type; 1595 } 1596 1597 return INT; 1598 } 1599 1600 /* Temporary obstack used for holding strings. */ 1601 static struct obstack tempbuf; 1602 static int tempbuf_init; 1603 1604 /* Parse a C escape sequence. The initial backslash of the sequence 1605 is at (*PTR)[-1]. *PTR will be updated to point to just after the 1606 last character of the sequence. If OUTPUT is not NULL, the 1607 translated form of the escape sequence will be written there. If 1608 OUTPUT is NULL, no output is written and the call will only affect 1609 *PTR. If an escape sequence is expressed in target bytes, then the 1610 entire sequence will simply be copied to OUTPUT. Return 1 if any 1611 character was emitted, 0 otherwise. */ 1612 1613 int 1614 c_parse_escape (char **ptr, struct obstack *output) 1615 { 1616 char *tokptr = *ptr; 1617 int result = 1; 1618 1619 /* Some escape sequences undergo character set conversion. Those we 1620 translate here. */ 1621 switch (*tokptr) 1622 { 1623 /* Hex escapes do not undergo character set conversion, so keep 1624 the escape sequence for later. */ 1625 case 'x': 1626 if (output) 1627 obstack_grow_str (output, "\\x"); 1628 ++tokptr; 1629 if (!isxdigit (*tokptr)) 1630 error (_("\\x escape without a following hex digit")); 1631 while (isxdigit (*tokptr)) 1632 { 1633 if (output) 1634 obstack_1grow (output, *tokptr); 1635 ++tokptr; 1636 } 1637 break; 1638 1639 /* Octal escapes do not undergo character set conversion, so 1640 keep the escape sequence for later. */ 1641 case '0': 1642 case '1': 1643 case '2': 1644 case '3': 1645 case '4': 1646 case '5': 1647 case '6': 1648 case '7': 1649 { 1650 int i; 1651 if (output) 1652 obstack_grow_str (output, "\\"); 1653 for (i = 0; 1654 i < 3 && isdigit (*tokptr) && *tokptr != '8' && *tokptr != '9'; 1655 ++i) 1656 { 1657 if (output) 1658 obstack_1grow (output, *tokptr); 1659 ++tokptr; 1660 } 1661 } 1662 break; 1663 1664 /* We handle UCNs later. We could handle them here, but that 1665 would mean a spurious error in the case where the UCN could 1666 be converted to the target charset but not the host 1667 charset. */ 1668 case 'u': 1669 case 'U': 1670 { 1671 char c = *tokptr; 1672 int i, len = c == 'U' ? 8 : 4; 1673 if (output) 1674 { 1675 obstack_1grow (output, '\\'); 1676 obstack_1grow (output, *tokptr); 1677 } 1678 ++tokptr; 1679 if (!isxdigit (*tokptr)) 1680 error (_("\\%c escape without a following hex digit"), c); 1681 for (i = 0; i < len && isxdigit (*tokptr); ++i) 1682 { 1683 if (output) 1684 obstack_1grow (output, *tokptr); 1685 ++tokptr; 1686 } 1687 } 1688 break; 1689 1690 /* We must pass backslash through so that it does not 1691 cause quoting during the second expansion. */ 1692 case '\\': 1693 if (output) 1694 obstack_grow_str (output, "\\\\"); 1695 ++tokptr; 1696 break; 1697 1698 /* Escapes which undergo conversion. */ 1699 case 'a': 1700 if (output) 1701 obstack_1grow (output, '\a'); 1702 ++tokptr; 1703 break; 1704 case 'b': 1705 if (output) 1706 obstack_1grow (output, '\b'); 1707 ++tokptr; 1708 break; 1709 case 'f': 1710 if (output) 1711 obstack_1grow (output, '\f'); 1712 ++tokptr; 1713 break; 1714 case 'n': 1715 if (output) 1716 obstack_1grow (output, '\n'); 1717 ++tokptr; 1718 break; 1719 case 'r': 1720 if (output) 1721 obstack_1grow (output, '\r'); 1722 ++tokptr; 1723 break; 1724 case 't': 1725 if (output) 1726 obstack_1grow (output, '\t'); 1727 ++tokptr; 1728 break; 1729 case 'v': 1730 if (output) 1731 obstack_1grow (output, '\v'); 1732 ++tokptr; 1733 break; 1734 1735 /* GCC extension. */ 1736 case 'e': 1737 if (output) 1738 obstack_1grow (output, HOST_ESCAPE_CHAR); 1739 ++tokptr; 1740 break; 1741 1742 /* Backslash-newline expands to nothing at all. */ 1743 case '\n': 1744 ++tokptr; 1745 result = 0; 1746 break; 1747 1748 /* A few escapes just expand to the character itself. */ 1749 case '\'': 1750 case '\"': 1751 case '?': 1752 /* GCC extensions. */ 1753 case '(': 1754 case '{': 1755 case '[': 1756 case '%': 1757 /* Unrecognized escapes turn into the character itself. */ 1758 default: 1759 if (output) 1760 obstack_1grow (output, *tokptr); 1761 ++tokptr; 1762 break; 1763 } 1764 *ptr = tokptr; 1765 return result; 1766 } 1767 1768 /* Parse a string or character literal from TOKPTR. The string or 1769 character may be wide or unicode. *OUTPTR is set to just after the 1770 end of the literal in the input string. The resulting token is 1771 stored in VALUE. This returns a token value, either STRING or 1772 CHAR, depending on what was parsed. *HOST_CHARS is set to the 1773 number of host characters in the literal. */ 1774 static int 1775 parse_string_or_char (char *tokptr, char **outptr, struct typed_stoken *value, 1776 int *host_chars) 1777 { 1778 int quote; 1779 enum c_string_type type; 1780 1781 /* Build the gdb internal form of the input string in tempbuf. Note 1782 that the buffer is null byte terminated *only* for the 1783 convenience of debugging gdb itself and printing the buffer 1784 contents when the buffer contains no embedded nulls. Gdb does 1785 not depend upon the buffer being null byte terminated, it uses 1786 the length string instead. This allows gdb to handle C strings 1787 (as well as strings in other languages) with embedded null 1788 bytes */ 1789 1790 if (!tempbuf_init) 1791 tempbuf_init = 1; 1792 else 1793 obstack_free (&tempbuf, NULL); 1794 obstack_init (&tempbuf); 1795 1796 /* Record the string type. */ 1797 if (*tokptr == 'L') 1798 { 1799 type = C_WIDE_STRING; 1800 ++tokptr; 1801 } 1802 else if (*tokptr == 'u') 1803 { 1804 type = C_STRING_16; 1805 ++tokptr; 1806 } 1807 else if (*tokptr == 'U') 1808 { 1809 type = C_STRING_32; 1810 ++tokptr; 1811 } 1812 else 1813 type = C_STRING; 1814 1815 /* Skip the quote. */ 1816 quote = *tokptr; 1817 if (quote == '\'') 1818 type |= C_CHAR; 1819 ++tokptr; 1820 1821 *host_chars = 0; 1822 1823 while (*tokptr) 1824 { 1825 char c = *tokptr; 1826 if (c == '\\') 1827 { 1828 ++tokptr; 1829 *host_chars += c_parse_escape (&tokptr, &tempbuf); 1830 } 1831 else if (c == quote) 1832 break; 1833 else 1834 { 1835 obstack_1grow (&tempbuf, c); 1836 ++tokptr; 1837 /* FIXME: this does the wrong thing with multi-byte host 1838 characters. We could use mbrlen here, but that would 1839 make "set host-charset" a bit less useful. */ 1840 ++*host_chars; 1841 } 1842 } 1843 1844 if (*tokptr != quote) 1845 { 1846 if (quote == '"') 1847 error (_("Unterminated string in expression.")); 1848 else 1849 error (_("Unmatched single quote.")); 1850 } 1851 ++tokptr; 1852 1853 value->type = type; 1854 value->ptr = obstack_base (&tempbuf); 1855 value->length = obstack_object_size (&tempbuf); 1856 1857 *outptr = tokptr; 1858 1859 return quote == '"' ? STRING : CHAR; 1860 } 1861 1862 struct token 1863 { 1864 char *operator; 1865 int token; 1866 enum exp_opcode opcode; 1867 int cxx_only; 1868 }; 1869 1870 static const struct token tokentab3[] = 1871 { 1872 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0}, 1873 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0}, 1874 {"->*", ARROW_STAR, BINOP_END, 1} 1875 }; 1876 1877 static const struct token tokentab2[] = 1878 { 1879 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0}, 1880 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0}, 1881 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0}, 1882 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0}, 1883 {"%=", ASSIGN_MODIFY, BINOP_REM, 0}, 1884 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0}, 1885 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0}, 1886 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0}, 1887 {"++", INCREMENT, BINOP_END, 0}, 1888 {"--", DECREMENT, BINOP_END, 0}, 1889 {"->", ARROW, BINOP_END, 0}, 1890 {"&&", ANDAND, BINOP_END, 0}, 1891 {"||", OROR, BINOP_END, 0}, 1892 /* "::" is *not* only C++: gdb overrides its meaning in several 1893 different ways, e.g., 'filename'::func, function::variable. */ 1894 {"::", COLONCOLON, BINOP_END, 0}, 1895 {"<<", LSH, BINOP_END, 0}, 1896 {">>", RSH, BINOP_END, 0}, 1897 {"==", EQUAL, BINOP_END, 0}, 1898 {"!=", NOTEQUAL, BINOP_END, 0}, 1899 {"<=", LEQ, BINOP_END, 0}, 1900 {">=", GEQ, BINOP_END, 0}, 1901 {".*", DOT_STAR, BINOP_END, 1} 1902 }; 1903 1904 /* Identifier-like tokens. */ 1905 static const struct token ident_tokens[] = 1906 { 1907 {"unsigned", UNSIGNED, OP_NULL, 0}, 1908 {"template", TEMPLATE, OP_NULL, 1}, 1909 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0}, 1910 {"struct", STRUCT, OP_NULL, 0}, 1911 {"signed", SIGNED_KEYWORD, OP_NULL, 0}, 1912 {"sizeof", SIZEOF, OP_NULL, 0}, 1913 {"double", DOUBLE_KEYWORD, OP_NULL, 0}, 1914 {"false", FALSEKEYWORD, OP_NULL, 1}, 1915 {"class", CLASS, OP_NULL, 1}, 1916 {"union", UNION, OP_NULL, 0}, 1917 {"short", SHORT, OP_NULL, 0}, 1918 {"const", CONST_KEYWORD, OP_NULL, 0}, 1919 {"enum", ENUM, OP_NULL, 0}, 1920 {"long", LONG, OP_NULL, 0}, 1921 {"true", TRUEKEYWORD, OP_NULL, 1}, 1922 {"int", INT_KEYWORD, OP_NULL, 0}, 1923 {"new", NEW, OP_NULL, 1}, 1924 {"delete", DELETE, OP_NULL, 1}, 1925 {"operator", OPERATOR, OP_NULL, 1}, 1926 1927 {"and", ANDAND, BINOP_END, 1}, 1928 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, 1}, 1929 {"bitand", '&', OP_NULL, 1}, 1930 {"bitor", '|', OP_NULL, 1}, 1931 {"compl", '~', OP_NULL, 1}, 1932 {"not", '!', OP_NULL, 1}, 1933 {"not_eq", NOTEQUAL, BINOP_END, 1}, 1934 {"or", OROR, BINOP_END, 1}, 1935 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 1}, 1936 {"xor", '^', OP_NULL, 1}, 1937 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 1}, 1938 1939 {"const_cast", CONST_CAST, OP_NULL, 1 }, 1940 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, 1 }, 1941 {"static_cast", STATIC_CAST, OP_NULL, 1 }, 1942 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, 1 } 1943 }; 1944 1945 /* When we find that lexptr (the global var defined in parse.c) is 1946 pointing at a macro invocation, we expand the invocation, and call 1947 scan_macro_expansion to save the old lexptr here and point lexptr 1948 into the expanded text. When we reach the end of that, we call 1949 end_macro_expansion to pop back to the value we saved here. The 1950 macro expansion code promises to return only fully-expanded text, 1951 so we don't need to "push" more than one level. 1952 1953 This is disgusting, of course. It would be cleaner to do all macro 1954 expansion beforehand, and then hand that to lexptr. But we don't 1955 really know where the expression ends. Remember, in a command like 1956 1957 (gdb) break *ADDRESS if CONDITION 1958 1959 we evaluate ADDRESS in the scope of the current frame, but we 1960 evaluate CONDITION in the scope of the breakpoint's location. So 1961 it's simply wrong to try to macro-expand the whole thing at once. */ 1962 static char *macro_original_text; 1963 1964 /* We save all intermediate macro expansions on this obstack for the 1965 duration of a single parse. The expansion text may sometimes have 1966 to live past the end of the expansion, due to yacc lookahead. 1967 Rather than try to be clever about saving the data for a single 1968 token, we simply keep it all and delete it after parsing has 1969 completed. */ 1970 static struct obstack expansion_obstack; 1971 1972 static void 1973 scan_macro_expansion (char *expansion) 1974 { 1975 char *copy; 1976 1977 /* We'd better not be trying to push the stack twice. */ 1978 gdb_assert (! macro_original_text); 1979 1980 /* Copy to the obstack, and then free the intermediate 1981 expansion. */ 1982 copy = obstack_copy0 (&expansion_obstack, expansion, strlen (expansion)); 1983 xfree (expansion); 1984 1985 /* Save the old lexptr value, so we can return to it when we're done 1986 parsing the expanded text. */ 1987 macro_original_text = lexptr; 1988 lexptr = copy; 1989 } 1990 1991 1992 static int 1993 scanning_macro_expansion (void) 1994 { 1995 return macro_original_text != 0; 1996 } 1997 1998 1999 static void 2000 finished_macro_expansion (void) 2001 { 2002 /* There'd better be something to pop back to. */ 2003 gdb_assert (macro_original_text); 2004 2005 /* Pop back to the original text. */ 2006 lexptr = macro_original_text; 2007 macro_original_text = 0; 2008 } 2009 2010 2011 static void 2012 scan_macro_cleanup (void *dummy) 2013 { 2014 if (macro_original_text) 2015 finished_macro_expansion (); 2016 2017 obstack_free (&expansion_obstack, NULL); 2018 } 2019 2020 /* Return true iff the token represents a C++ cast operator. */ 2021 2022 static int 2023 is_cast_operator (const char *token, int len) 2024 { 2025 return (! strncmp (token, "dynamic_cast", len) 2026 || ! strncmp (token, "static_cast", len) 2027 || ! strncmp (token, "reinterpret_cast", len) 2028 || ! strncmp (token, "const_cast", len)); 2029 } 2030 2031 /* The scope used for macro expansion. */ 2032 static struct macro_scope *expression_macro_scope; 2033 2034 /* This is set if a NAME token appeared at the very end of the input 2035 string, with no whitespace separating the name from the EOF. This 2036 is used only when parsing to do field name completion. */ 2037 static int saw_name_at_eof; 2038 2039 /* This is set if the previously-returned token was a structure 2040 operator -- either '.' or ARROW. This is used only when parsing to 2041 do field name completion. */ 2042 static int last_was_structop; 2043 2044 /* Read one token, getting characters through lexptr. */ 2045 2046 static int 2047 lex_one_token (void) 2048 { 2049 int c; 2050 int namelen; 2051 unsigned int i; 2052 char *tokstart; 2053 int saw_structop = last_was_structop; 2054 char *copy; 2055 2056 last_was_structop = 0; 2057 2058 retry: 2059 2060 /* Check if this is a macro invocation that we need to expand. */ 2061 if (! scanning_macro_expansion ()) 2062 { 2063 char *expanded = macro_expand_next (&lexptr, 2064 standard_macro_lookup, 2065 expression_macro_scope); 2066 2067 if (expanded) 2068 scan_macro_expansion (expanded); 2069 } 2070 2071 prev_lexptr = lexptr; 2072 2073 tokstart = lexptr; 2074 /* See if it is a special token of length 3. */ 2075 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++) 2076 if (strncmp (tokstart, tokentab3[i].operator, 3) == 0) 2077 { 2078 if (tokentab3[i].cxx_only 2079 && parse_language->la_language != language_cplus) 2080 break; 2081 2082 lexptr += 3; 2083 yylval.opcode = tokentab3[i].opcode; 2084 return tokentab3[i].token; 2085 } 2086 2087 /* See if it is a special token of length 2. */ 2088 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++) 2089 if (strncmp (tokstart, tokentab2[i].operator, 2) == 0) 2090 { 2091 if (tokentab2[i].cxx_only 2092 && parse_language->la_language != language_cplus) 2093 break; 2094 2095 lexptr += 2; 2096 yylval.opcode = tokentab2[i].opcode; 2097 if (in_parse_field && tokentab2[i].token == ARROW) 2098 last_was_structop = 1; 2099 return tokentab2[i].token; 2100 } 2101 2102 switch (c = *tokstart) 2103 { 2104 case 0: 2105 /* If we were just scanning the result of a macro expansion, 2106 then we need to resume scanning the original text. 2107 If we're parsing for field name completion, and the previous 2108 token allows such completion, return a COMPLETE token. 2109 Otherwise, we were already scanning the original text, and 2110 we're really done. */ 2111 if (scanning_macro_expansion ()) 2112 { 2113 finished_macro_expansion (); 2114 goto retry; 2115 } 2116 else if (saw_name_at_eof) 2117 { 2118 saw_name_at_eof = 0; 2119 return COMPLETE; 2120 } 2121 else if (saw_structop) 2122 return COMPLETE; 2123 else 2124 return 0; 2125 2126 case ' ': 2127 case '\t': 2128 case '\n': 2129 lexptr++; 2130 goto retry; 2131 2132 case '[': 2133 case '(': 2134 paren_depth++; 2135 lexptr++; 2136 return c; 2137 2138 case ']': 2139 case ')': 2140 if (paren_depth == 0) 2141 return 0; 2142 paren_depth--; 2143 lexptr++; 2144 return c; 2145 2146 case ',': 2147 if (comma_terminates 2148 && paren_depth == 0 2149 && ! scanning_macro_expansion ()) 2150 return 0; 2151 lexptr++; 2152 return c; 2153 2154 case '.': 2155 /* Might be a floating point number. */ 2156 if (lexptr[1] < '0' || lexptr[1] > '9') 2157 { 2158 if (in_parse_field) 2159 last_was_structop = 1; 2160 goto symbol; /* Nope, must be a symbol. */ 2161 } 2162 /* FALL THRU into number case. */ 2163 2164 case '0': 2165 case '1': 2166 case '2': 2167 case '3': 2168 case '4': 2169 case '5': 2170 case '6': 2171 case '7': 2172 case '8': 2173 case '9': 2174 { 2175 /* It's a number. */ 2176 int got_dot = 0, got_e = 0, toktype; 2177 char *p = tokstart; 2178 int hex = input_radix > 10; 2179 2180 if (c == '0' && (p[1] == 'x' || p[1] == 'X')) 2181 { 2182 p += 2; 2183 hex = 1; 2184 } 2185 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D')) 2186 { 2187 p += 2; 2188 hex = 0; 2189 } 2190 2191 for (;; ++p) 2192 { 2193 /* This test includes !hex because 'e' is a valid hex digit 2194 and thus does not indicate a floating point number when 2195 the radix is hex. */ 2196 if (!hex && !got_e && (*p == 'e' || *p == 'E')) 2197 got_dot = got_e = 1; 2198 /* This test does not include !hex, because a '.' always indicates 2199 a decimal floating point number regardless of the radix. */ 2200 else if (!got_dot && *p == '.') 2201 got_dot = 1; 2202 else if (got_e && (p[-1] == 'e' || p[-1] == 'E') 2203 && (*p == '-' || *p == '+')) 2204 /* This is the sign of the exponent, not the end of the 2205 number. */ 2206 continue; 2207 /* We will take any letters or digits. parse_number will 2208 complain if past the radix, or if L or U are not final. */ 2209 else if ((*p < '0' || *p > '9') 2210 && ((*p < 'a' || *p > 'z') 2211 && (*p < 'A' || *p > 'Z'))) 2212 break; 2213 } 2214 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e, &yylval); 2215 if (toktype == ERROR) 2216 { 2217 char *err_copy = (char *) alloca (p - tokstart + 1); 2218 2219 memcpy (err_copy, tokstart, p - tokstart); 2220 err_copy[p - tokstart] = 0; 2221 error (_("Invalid number \"%s\"."), err_copy); 2222 } 2223 lexptr = p; 2224 return toktype; 2225 } 2226 2227 case '+': 2228 case '-': 2229 case '*': 2230 case '/': 2231 case '%': 2232 case '|': 2233 case '&': 2234 case '^': 2235 case '~': 2236 case '!': 2237 case '@': 2238 case '<': 2239 case '>': 2240 case '?': 2241 case ':': 2242 case '=': 2243 case '{': 2244 case '}': 2245 symbol: 2246 lexptr++; 2247 return c; 2248 2249 case 'L': 2250 case 'u': 2251 case 'U': 2252 if (tokstart[1] != '"' && tokstart[1] != '\'') 2253 break; 2254 /* Fall through. */ 2255 case '\'': 2256 case '"': 2257 { 2258 int host_len; 2259 int result = parse_string_or_char (tokstart, &lexptr, &yylval.tsval, 2260 &host_len); 2261 if (result == CHAR) 2262 { 2263 if (host_len == 0) 2264 error (_("Empty character constant.")); 2265 else if (host_len > 2 && c == '\'') 2266 { 2267 ++tokstart; 2268 namelen = lexptr - tokstart - 1; 2269 goto tryname; 2270 } 2271 else if (host_len > 1) 2272 error (_("Invalid character constant.")); 2273 } 2274 return result; 2275 } 2276 } 2277 2278 if (!(c == '_' || c == '$' 2279 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))) 2280 /* We must have come across a bad character (e.g. ';'). */ 2281 error (_("Invalid character '%c' in expression."), c); 2282 2283 /* It's a name. See how long it is. */ 2284 namelen = 0; 2285 for (c = tokstart[namelen]; 2286 (c == '_' || c == '$' || (c >= '0' && c <= '9') 2287 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '<');) 2288 { 2289 /* Template parameter lists are part of the name. 2290 FIXME: This mishandles `print $a<4&&$a>3'. */ 2291 2292 if (c == '<') 2293 { 2294 if (! is_cast_operator (tokstart, namelen)) 2295 { 2296 /* Scan ahead to get rest of the template specification. Note 2297 that we look ahead only when the '<' adjoins non-whitespace 2298 characters; for comparison expressions, e.g. "a < b > c", 2299 there must be spaces before the '<', etc. */ 2300 2301 char * p = find_template_name_end (tokstart + namelen); 2302 if (p) 2303 namelen = p - tokstart; 2304 } 2305 break; 2306 } 2307 c = tokstart[++namelen]; 2308 } 2309 2310 /* The token "if" terminates the expression and is NOT removed from 2311 the input stream. It doesn't count if it appears in the 2312 expansion of a macro. */ 2313 if (namelen == 2 2314 && tokstart[0] == 'i' 2315 && tokstart[1] == 'f' 2316 && ! scanning_macro_expansion ()) 2317 { 2318 return 0; 2319 } 2320 2321 /* For the same reason (breakpoint conditions), "thread N" 2322 terminates the expression. "thread" could be an identifier, but 2323 an identifier is never followed by a number without intervening 2324 punctuation. "task" is similar. Handle abbreviations of these, 2325 similarly to breakpoint.c:find_condition_and_thread. */ 2326 if (namelen >= 1 2327 && (strncmp (tokstart, "thread", namelen) == 0 2328 || strncmp (tokstart, "task", namelen) == 0) 2329 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t') 2330 && ! scanning_macro_expansion ()) 2331 { 2332 char *p = tokstart + namelen + 1; 2333 while (*p == ' ' || *p == '\t') 2334 p++; 2335 if (*p >= '0' && *p <= '9') 2336 return 0; 2337 } 2338 2339 lexptr += namelen; 2340 2341 tryname: 2342 2343 yylval.sval.ptr = tokstart; 2344 yylval.sval.length = namelen; 2345 2346 /* Catch specific keywords. */ 2347 copy = copy_name (yylval.sval); 2348 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++) 2349 if (strcmp (copy, ident_tokens[i].operator) == 0) 2350 { 2351 if (ident_tokens[i].cxx_only 2352 && parse_language->la_language != language_cplus) 2353 break; 2354 2355 /* It is ok to always set this, even though we don't always 2356 strictly need to. */ 2357 yylval.opcode = ident_tokens[i].opcode; 2358 return ident_tokens[i].token; 2359 } 2360 2361 if (*tokstart == '$') 2362 return VARIABLE; 2363 2364 if (in_parse_field && *lexptr == '\0') 2365 saw_name_at_eof = 1; 2366 return NAME; 2367 } 2368 2369 /* An object of this type is pushed on a FIFO by the "outer" lexer. */ 2370 typedef struct 2371 { 2372 int token; 2373 YYSTYPE value; 2374 } token_and_value; 2375 2376 DEF_VEC_O (token_and_value); 2377 2378 /* A FIFO of tokens that have been read but not yet returned to the 2379 parser. */ 2380 static VEC (token_and_value) *token_fifo; 2381 2382 /* Non-zero if the lexer should return tokens from the FIFO. */ 2383 static int popping; 2384 2385 /* Temporary storage for c_lex; this holds symbol names as they are 2386 built up. */ 2387 static struct obstack name_obstack; 2388 2389 /* Classify a NAME token. The contents of the token are in `yylval'. 2390 Updates yylval and returns the new token type. BLOCK is the block 2391 in which lookups start; this can be NULL to mean the global 2392 scope. */ 2393 static int 2394 classify_name (struct block *block) 2395 { 2396 struct symbol *sym; 2397 char *copy; 2398 int is_a_field_of_this = 0; 2399 2400 copy = copy_name (yylval.sval); 2401 2402 sym = lookup_symbol (copy, block, VAR_DOMAIN, 2403 parse_language->la_language == language_cplus 2404 ? &is_a_field_of_this : (int *) NULL); 2405 2406 if (sym && SYMBOL_CLASS (sym) == LOC_BLOCK) 2407 { 2408 yylval.ssym.sym = sym; 2409 yylval.ssym.is_a_field_of_this = is_a_field_of_this; 2410 return BLOCKNAME; 2411 } 2412 else if (!sym) 2413 { 2414 /* See if it's a file name. */ 2415 struct symtab *symtab; 2416 2417 symtab = lookup_symtab (copy); 2418 if (symtab) 2419 { 2420 yylval.bval = BLOCKVECTOR_BLOCK (BLOCKVECTOR (symtab), STATIC_BLOCK); 2421 return FILENAME; 2422 } 2423 } 2424 2425 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF) 2426 { 2427 yylval.tsym.type = SYMBOL_TYPE (sym); 2428 return TYPENAME; 2429 } 2430 2431 yylval.tsym.type 2432 = language_lookup_primitive_type_by_name (parse_language, 2433 parse_gdbarch, copy); 2434 if (yylval.tsym.type != NULL) 2435 return TYPENAME; 2436 2437 /* Input names that aren't symbols but ARE valid hex numbers, when 2438 the input radix permits them, can be names or numbers depending 2439 on the parse. Note we support radixes > 16 here. */ 2440 if (!sym 2441 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10) 2442 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10))) 2443 { 2444 YYSTYPE newlval; /* Its value is ignored. */ 2445 int hextype = parse_number (copy, yylval.sval.length, 0, &newlval); 2446 if (hextype == INT) 2447 { 2448 yylval.ssym.sym = sym; 2449 yylval.ssym.is_a_field_of_this = is_a_field_of_this; 2450 return NAME_OR_INT; 2451 } 2452 } 2453 2454 /* Any other kind of symbol */ 2455 yylval.ssym.sym = sym; 2456 yylval.ssym.is_a_field_of_this = is_a_field_of_this; 2457 2458 if (sym == NULL 2459 && parse_language->la_language == language_cplus 2460 && !is_a_field_of_this 2461 && !lookup_minimal_symbol (copy, NULL, NULL)) 2462 return UNKNOWN_CPP_NAME; 2463 2464 return NAME; 2465 } 2466 2467 /* Like classify_name, but used by the inner loop of the lexer, when a 2468 name might have already been seen. FIRST_NAME is true if the token 2469 in `yylval' is the first component of a name, false otherwise. If 2470 this function returns NAME, it might not have updated `yylval'. 2471 This is ok because the caller only cares about TYPENAME. */ 2472 static int 2473 classify_inner_name (struct block *block, int first_name) 2474 { 2475 struct type *type, *new_type; 2476 char *copy; 2477 2478 if (first_name) 2479 return classify_name (block); 2480 2481 type = check_typedef (yylval.tsym.type); 2482 if (TYPE_CODE (type) != TYPE_CODE_STRUCT 2483 && TYPE_CODE (type) != TYPE_CODE_UNION 2484 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE) 2485 /* We know the caller won't expect us to update yylval. */ 2486 return NAME; 2487 2488 copy = copy_name (yylval.tsym.stoken); 2489 new_type = cp_lookup_nested_type (type, copy, block); 2490 2491 if (new_type == NULL) 2492 /* We know the caller won't expect us to update yylval. */ 2493 return NAME; 2494 2495 yylval.tsym.type = new_type; 2496 return TYPENAME; 2497 } 2498 2499 /* The outer level of a two-level lexer. This calls the inner lexer 2500 to return tokens. It then either returns these tokens, or 2501 aggregates them into a larger token. This lets us work around a 2502 problem in our parsing approach, where the parser could not 2503 distinguish between qualified names and qualified types at the 2504 right point. 2505 2506 This approach is still not ideal, because it mishandles template 2507 types. See the comment in lex_one_token for an example. However, 2508 this is still an improvement over the earlier approach, and will 2509 suffice until we move to better parsing technology. */ 2510 static int 2511 yylex (void) 2512 { 2513 token_and_value current; 2514 int first_was_coloncolon, last_was_coloncolon, first_iter; 2515 2516 if (popping && !VEC_empty (token_and_value, token_fifo)) 2517 { 2518 token_and_value tv = *VEC_index (token_and_value, token_fifo, 0); 2519 VEC_ordered_remove (token_and_value, token_fifo, 0); 2520 yylval = tv.value; 2521 return tv.token; 2522 } 2523 popping = 0; 2524 2525 current.token = lex_one_token (); 2526 if (current.token == NAME) 2527 current.token = classify_name (expression_context_block); 2528 if (parse_language->la_language != language_cplus 2529 || (current.token != TYPENAME && current.token != COLONCOLON)) 2530 return current.token; 2531 2532 first_was_coloncolon = current.token == COLONCOLON; 2533 last_was_coloncolon = first_was_coloncolon; 2534 obstack_free (&name_obstack, obstack_base (&name_obstack)); 2535 if (!last_was_coloncolon) 2536 obstack_grow (&name_obstack, yylval.sval.ptr, yylval.sval.length); 2537 current.value = yylval; 2538 first_iter = 1; 2539 while (1) 2540 { 2541 token_and_value next; 2542 2543 next.token = lex_one_token (); 2544 next.value = yylval; 2545 2546 if (next.token == NAME && last_was_coloncolon) 2547 { 2548 int classification; 2549 2550 classification = classify_inner_name (first_was_coloncolon 2551 ? NULL 2552 : expression_context_block, 2553 first_iter); 2554 /* We keep going until we either run out of names, or until 2555 we have a qualified name which is not a type. */ 2556 if (classification != TYPENAME) 2557 { 2558 /* Push the final component and leave the loop. */ 2559 VEC_safe_push (token_and_value, token_fifo, &next); 2560 break; 2561 } 2562 2563 /* Update the partial name we are constructing. */ 2564 if (!first_iter) 2565 { 2566 /* We don't want to put a leading "::" into the name. */ 2567 obstack_grow_str (&name_obstack, "::"); 2568 } 2569 obstack_grow (&name_obstack, next.value.sval.ptr, 2570 next.value.sval.length); 2571 2572 yylval.sval.ptr = obstack_base (&name_obstack); 2573 yylval.sval.length = obstack_object_size (&name_obstack); 2574 current.value = yylval; 2575 current.token = classification; 2576 2577 last_was_coloncolon = 0; 2578 } 2579 else if (next.token == COLONCOLON && !last_was_coloncolon) 2580 last_was_coloncolon = 1; 2581 else 2582 { 2583 /* We've reached the end of the name. */ 2584 VEC_safe_push (token_and_value, token_fifo, &next); 2585 break; 2586 } 2587 2588 first_iter = 0; 2589 } 2590 2591 popping = 1; 2592 2593 /* If we ended with a "::", insert it too. */ 2594 if (last_was_coloncolon) 2595 { 2596 token_and_value cc; 2597 memset (&cc, 0, sizeof (token_and_value)); 2598 if (first_was_coloncolon && first_iter) 2599 { 2600 yylval = cc.value; 2601 return COLONCOLON; 2602 } 2603 cc.token = COLONCOLON; 2604 VEC_safe_insert (token_and_value, token_fifo, 0, &cc); 2605 } 2606 2607 yylval = current.value; 2608 yylval.sval.ptr = obstack_copy0 (&expansion_obstack, 2609 yylval.sval.ptr, 2610 yylval.sval.length); 2611 return current.token; 2612 } 2613 2614 int 2615 c_parse (void) 2616 { 2617 int result; 2618 struct cleanup *back_to = make_cleanup (free_current_contents, 2619 &expression_macro_scope); 2620 2621 /* Set up the scope for macro expansion. */ 2622 expression_macro_scope = NULL; 2623 2624 if (expression_context_block) 2625 expression_macro_scope 2626 = sal_macro_scope (find_pc_line (expression_context_pc, 0)); 2627 else 2628 expression_macro_scope = default_macro_scope (); 2629 if (! expression_macro_scope) 2630 expression_macro_scope = user_macro_scope (); 2631 2632 /* Initialize macro expansion code. */ 2633 obstack_init (&expansion_obstack); 2634 gdb_assert (! macro_original_text); 2635 make_cleanup (scan_macro_cleanup, 0); 2636 2637 make_cleanup_restore_integer (&yydebug); 2638 yydebug = parser_debug; 2639 2640 /* Initialize some state used by the lexer. */ 2641 last_was_structop = 0; 2642 saw_name_at_eof = 0; 2643 2644 VEC_free (token_and_value, token_fifo); 2645 popping = 0; 2646 obstack_init (&name_obstack); 2647 make_cleanup_obstack_free (&name_obstack); 2648 2649 result = yyparse (); 2650 do_cleanups (back_to); 2651 return result; 2652 } 2653 2654 2655 void 2656 yyerror (char *msg) 2657 { 2658 if (prev_lexptr) 2659 lexptr = prev_lexptr; 2660 2661 error (_("A %s in expression, near `%s'."), (msg ? msg : "error"), lexptr); 2662 } 2663