| 1 | /* YACC parser for C expressions, for GDB. |
| 2 | Copyright (C) 1986, 1989, 1990, 1991 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 2 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, write to the Free Software |
| 18 | Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ |
| 19 | |
| 20 | /* Parse a C expression from text in a string, |
| 21 | and return the result as a struct expression pointer. |
| 22 | That structure contains arithmetic operations in reverse polish, |
| 23 | with constants represented by operations that are followed by special data. |
| 24 | See expression.h for the details of the format. |
| 25 | What is important here is that it can be built up sequentially |
| 26 | during the process of parsing; the lower levels of the tree always |
| 27 | come first in the result. |
| 28 | |
| 29 | Note that malloc's and realloc's in this file are transformed to |
| 30 | xmalloc and xrealloc respectively by the same sed command in the |
| 31 | makefile that remaps any other malloc/realloc inserted by the parser |
| 32 | generator. Doing this with #defines and trying to control the interaction |
| 33 | with include files (<malloc.h> and <stdlib.h> for example) just became |
| 34 | too messy, particularly when such includes can be inserted at random |
| 35 | times by the parser generator. */ |
| 36 | |
| 37 | %{ |
| 38 | |
| 39 | #include "defs.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 "bfd.h" /* Required by objfiles.h. */ |
| 46 | #include "symfile.h" /* Required by objfiles.h. */ |
| 47 | #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */ |
| 48 | |
| 49 | /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc), |
| 50 | as well as gratuitiously global symbol names, so we can have multiple |
| 51 | yacc generated parsers in gdb. Note that these are only the variables |
| 52 | produced by yacc. If other parser generators (bison, byacc, etc) produce |
| 53 | additional global names that conflict at link time, then those parser |
| 54 | generators need to be fixed instead of adding those names to this list. */ |
| 55 | |
| 56 | #define yymaxdepth c_maxdepth |
| 57 | #define yyparse c_parse |
| 58 | #define yylex c_lex |
| 59 | #define yyerror c_error |
| 60 | #define yylval c_lval |
| 61 | #define yychar c_char |
| 62 | #define yydebug c_debug |
| 63 | #define yypact c_pact |
| 64 | #define yyr1 c_r1 |
| 65 | #define yyr2 c_r2 |
| 66 | #define yydef c_def |
| 67 | #define yychk c_chk |
| 68 | #define yypgo c_pgo |
| 69 | #define yyact c_act |
| 70 | #define yyexca c_exca |
| 71 | #define yyerrflag c_errflag |
| 72 | #define yynerrs c_nerrs |
| 73 | #define yyps c_ps |
| 74 | #define yypv c_pv |
| 75 | #define yys c_s |
| 76 | #define yy_yys c_yys |
| 77 | #define yystate c_state |
| 78 | #define yytmp c_tmp |
| 79 | #define yyv c_v |
| 80 | #define yy_yyv c_yyv |
| 81 | #define yyval c_val |
| 82 | #define yylloc c_lloc |
| 83 | #define yyreds c_reds /* With YYDEBUG defined */ |
| 84 | #define yytoks c_toks /* With YYDEBUG defined */ |
| 85 | |
| 86 | #ifndef YYDEBUG |
| 87 | #define YYDEBUG 0 /* Default to no yydebug support */ |
| 88 | #endif |
| 89 | |
| 90 | int |
| 91 | yyparse PARAMS ((void)); |
| 92 | |
| 93 | static int |
| 94 | yylex PARAMS ((void)); |
| 95 | |
| 96 | void |
| 97 | yyerror PARAMS ((char *)); |
| 98 | |
| 99 | %} |
| 100 | |
| 101 | /* Although the yacc "value" of an expression is not used, |
| 102 | since the result is stored in the structure being created, |
| 103 | other node types do have values. */ |
| 104 | |
| 105 | %union |
| 106 | { |
| 107 | LONGEST lval; |
| 108 | struct { |
| 109 | LONGEST val; |
| 110 | struct type *type; |
| 111 | } typed_val; |
| 112 | double dval; |
| 113 | struct symbol *sym; |
| 114 | struct type *tval; |
| 115 | struct stoken sval; |
| 116 | struct ttype tsym; |
| 117 | struct symtoken ssym; |
| 118 | int voidval; |
| 119 | struct block *bval; |
| 120 | enum exp_opcode opcode; |
| 121 | struct internalvar *ivar; |
| 122 | |
| 123 | struct type **tvec; |
| 124 | int *ivec; |
| 125 | } |
| 126 | |
| 127 | %{ |
| 128 | /* YYSTYPE gets defined by %union */ |
| 129 | static int |
| 130 | parse_number PARAMS ((char *, int, int, YYSTYPE *)); |
| 131 | %} |
| 132 | |
| 133 | %type <voidval> exp exp1 type_exp start variable qualified_name lcurly |
| 134 | %type <lval> rcurly |
| 135 | %type <tval> type typebase |
| 136 | %type <tvec> nonempty_typelist |
| 137 | /* %type <bval> block */ |
| 138 | |
| 139 | /* Fancy type parsing. */ |
| 140 | %type <voidval> func_mod direct_abs_decl abs_decl |
| 141 | %type <tval> ptype |
| 142 | %type <lval> array_mod |
| 143 | |
| 144 | %token <typed_val> INT |
| 145 | %token <dval> FLOAT |
| 146 | |
| 147 | /* Both NAME and TYPENAME tokens represent symbols in the input, |
| 148 | and both convey their data as strings. |
| 149 | But a TYPENAME is a string that happens to be defined as a typedef |
| 150 | or builtin type name (such as int or char) |
| 151 | and a NAME is any other symbol. |
| 152 | Contexts where this distinction is not important can use the |
| 153 | nonterminal "name", which matches either NAME or TYPENAME. */ |
| 154 | |
| 155 | %token <sval> STRING |
| 156 | %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */ |
| 157 | %token <tsym> TYPENAME |
| 158 | %type <sval> name |
| 159 | %type <ssym> name_not_typename |
| 160 | %type <tsym> typename |
| 161 | |
| 162 | /* A NAME_OR_INT is a symbol which is not known in the symbol table, |
| 163 | but which would parse as a valid number in the current input radix. |
| 164 | E.g. "c" when input_radix==16. Depending on the parse, it will be |
| 165 | turned into a name or into a number. */ |
| 166 | |
| 167 | %token <ssym> NAME_OR_INT |
| 168 | |
| 169 | %token STRUCT CLASS UNION ENUM SIZEOF UNSIGNED COLONCOLON |
| 170 | %token TEMPLATE |
| 171 | %token ERROR |
| 172 | |
| 173 | /* Special type cases, put in to allow the parser to distinguish different |
| 174 | legal basetypes. */ |
| 175 | %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD |
| 176 | %token <lval> LAST REGNAME |
| 177 | |
| 178 | %token <ivar> VARIABLE |
| 179 | |
| 180 | %token <opcode> ASSIGN_MODIFY |
| 181 | |
| 182 | /* C++ */ |
| 183 | %token THIS |
| 184 | |
| 185 | %left ',' |
| 186 | %left ABOVE_COMMA |
| 187 | %right '=' ASSIGN_MODIFY |
| 188 | %right '?' |
| 189 | %left OROR |
| 190 | %left ANDAND |
| 191 | %left '|' |
| 192 | %left '^' |
| 193 | %left '&' |
| 194 | %left EQUAL NOTEQUAL |
| 195 | %left '<' '>' LEQ GEQ |
| 196 | %left LSH RSH |
| 197 | %left '@' |
| 198 | %left '+' '-' |
| 199 | %left '*' '/' '%' |
| 200 | %right UNARY INCREMENT DECREMENT |
| 201 | %right ARROW '.' '[' '(' |
| 202 | %token <ssym> BLOCKNAME |
| 203 | %type <bval> block |
| 204 | %left COLONCOLON |
| 205 | |
| 206 | \f |
| 207 | %% |
| 208 | |
| 209 | start : exp1 |
| 210 | | type_exp |
| 211 | ; |
| 212 | |
| 213 | type_exp: type |
| 214 | { write_exp_elt_opcode(OP_TYPE); |
| 215 | write_exp_elt_type($1); |
| 216 | write_exp_elt_opcode(OP_TYPE);} |
| 217 | ; |
| 218 | |
| 219 | /* Expressions, including the comma operator. */ |
| 220 | exp1 : exp |
| 221 | | exp1 ',' exp |
| 222 | { write_exp_elt_opcode (BINOP_COMMA); } |
| 223 | ; |
| 224 | |
| 225 | /* Expressions, not including the comma operator. */ |
| 226 | exp : '*' exp %prec UNARY |
| 227 | { write_exp_elt_opcode (UNOP_IND); } |
| 228 | |
| 229 | exp : '&' exp %prec UNARY |
| 230 | { write_exp_elt_opcode (UNOP_ADDR); } |
| 231 | |
| 232 | exp : '-' exp %prec UNARY |
| 233 | { write_exp_elt_opcode (UNOP_NEG); } |
| 234 | ; |
| 235 | |
| 236 | exp : '!' exp %prec UNARY |
| 237 | { write_exp_elt_opcode (UNOP_LOGICAL_NOT); } |
| 238 | ; |
| 239 | |
| 240 | exp : '~' exp %prec UNARY |
| 241 | { write_exp_elt_opcode (UNOP_COMPLEMENT); } |
| 242 | ; |
| 243 | |
| 244 | exp : INCREMENT exp %prec UNARY |
| 245 | { write_exp_elt_opcode (UNOP_PREINCREMENT); } |
| 246 | ; |
| 247 | |
| 248 | exp : DECREMENT exp %prec UNARY |
| 249 | { write_exp_elt_opcode (UNOP_PREDECREMENT); } |
| 250 | ; |
| 251 | |
| 252 | exp : exp INCREMENT %prec UNARY |
| 253 | { write_exp_elt_opcode (UNOP_POSTINCREMENT); } |
| 254 | ; |
| 255 | |
| 256 | exp : exp DECREMENT %prec UNARY |
| 257 | { write_exp_elt_opcode (UNOP_POSTDECREMENT); } |
| 258 | ; |
| 259 | |
| 260 | exp : SIZEOF exp %prec UNARY |
| 261 | { write_exp_elt_opcode (UNOP_SIZEOF); } |
| 262 | ; |
| 263 | |
| 264 | exp : exp ARROW name |
| 265 | { write_exp_elt_opcode (STRUCTOP_PTR); |
| 266 | write_exp_string ($3); |
| 267 | write_exp_elt_opcode (STRUCTOP_PTR); } |
| 268 | ; |
| 269 | |
| 270 | exp : exp ARROW qualified_name |
| 271 | { /* exp->type::name becomes exp->*(&type::name) */ |
| 272 | /* Note: this doesn't work if name is a |
| 273 | static member! FIXME */ |
| 274 | write_exp_elt_opcode (UNOP_ADDR); |
| 275 | write_exp_elt_opcode (STRUCTOP_MPTR); } |
| 276 | ; |
| 277 | exp : exp ARROW '*' exp |
| 278 | { write_exp_elt_opcode (STRUCTOP_MPTR); } |
| 279 | ; |
| 280 | |
| 281 | exp : exp '.' name |
| 282 | { write_exp_elt_opcode (STRUCTOP_STRUCT); |
| 283 | write_exp_string ($3); |
| 284 | write_exp_elt_opcode (STRUCTOP_STRUCT); } |
| 285 | ; |
| 286 | |
| 287 | exp : exp '.' qualified_name |
| 288 | { /* exp.type::name becomes exp.*(&type::name) */ |
| 289 | /* Note: this doesn't work if name is a |
| 290 | static member! FIXME */ |
| 291 | write_exp_elt_opcode (UNOP_ADDR); |
| 292 | write_exp_elt_opcode (STRUCTOP_MEMBER); } |
| 293 | ; |
| 294 | |
| 295 | exp : exp '.' '*' exp |
| 296 | { write_exp_elt_opcode (STRUCTOP_MEMBER); } |
| 297 | ; |
| 298 | |
| 299 | exp : exp '[' exp1 ']' |
| 300 | { write_exp_elt_opcode (BINOP_SUBSCRIPT); } |
| 301 | ; |
| 302 | |
| 303 | exp : exp '(' |
| 304 | /* This is to save the value of arglist_len |
| 305 | being accumulated by an outer function call. */ |
| 306 | { start_arglist (); } |
| 307 | arglist ')' %prec ARROW |
| 308 | { write_exp_elt_opcode (OP_FUNCALL); |
| 309 | write_exp_elt_longcst ((LONGEST) end_arglist ()); |
| 310 | write_exp_elt_opcode (OP_FUNCALL); } |
| 311 | ; |
| 312 | |
| 313 | lcurly : '{' |
| 314 | { start_arglist (); } |
| 315 | ; |
| 316 | |
| 317 | arglist : |
| 318 | ; |
| 319 | |
| 320 | arglist : exp |
| 321 | { arglist_len = 1; } |
| 322 | ; |
| 323 | |
| 324 | arglist : arglist ',' exp %prec ABOVE_COMMA |
| 325 | { arglist_len++; } |
| 326 | ; |
| 327 | |
| 328 | rcurly : '}' |
| 329 | { $$ = end_arglist () - 1; } |
| 330 | ; |
| 331 | exp : lcurly arglist rcurly %prec ARROW |
| 332 | { write_exp_elt_opcode (OP_ARRAY); |
| 333 | write_exp_elt_longcst ((LONGEST) 0); |
| 334 | write_exp_elt_longcst ((LONGEST) $3); |
| 335 | write_exp_elt_opcode (OP_ARRAY); } |
| 336 | ; |
| 337 | |
| 338 | exp : lcurly type rcurly exp %prec UNARY |
| 339 | { write_exp_elt_opcode (UNOP_MEMVAL); |
| 340 | write_exp_elt_type ($2); |
| 341 | write_exp_elt_opcode (UNOP_MEMVAL); } |
| 342 | ; |
| 343 | |
| 344 | exp : '(' type ')' exp %prec UNARY |
| 345 | { write_exp_elt_opcode (UNOP_CAST); |
| 346 | write_exp_elt_type ($2); |
| 347 | write_exp_elt_opcode (UNOP_CAST); } |
| 348 | ; |
| 349 | |
| 350 | exp : '(' exp1 ')' |
| 351 | { } |
| 352 | ; |
| 353 | |
| 354 | /* Binary operators in order of decreasing precedence. */ |
| 355 | |
| 356 | exp : exp '@' exp |
| 357 | { write_exp_elt_opcode (BINOP_REPEAT); } |
| 358 | ; |
| 359 | |
| 360 | exp : exp '*' exp |
| 361 | { write_exp_elt_opcode (BINOP_MUL); } |
| 362 | ; |
| 363 | |
| 364 | exp : exp '/' exp |
| 365 | { write_exp_elt_opcode (BINOP_DIV); } |
| 366 | ; |
| 367 | |
| 368 | exp : exp '%' exp |
| 369 | { write_exp_elt_opcode (BINOP_REM); } |
| 370 | ; |
| 371 | |
| 372 | exp : exp '+' exp |
| 373 | { write_exp_elt_opcode (BINOP_ADD); } |
| 374 | ; |
| 375 | |
| 376 | exp : exp '-' exp |
| 377 | { write_exp_elt_opcode (BINOP_SUB); } |
| 378 | ; |
| 379 | |
| 380 | exp : exp LSH exp |
| 381 | { write_exp_elt_opcode (BINOP_LSH); } |
| 382 | ; |
| 383 | |
| 384 | exp : exp RSH exp |
| 385 | { write_exp_elt_opcode (BINOP_RSH); } |
| 386 | ; |
| 387 | |
| 388 | exp : exp EQUAL exp |
| 389 | { write_exp_elt_opcode (BINOP_EQUAL); } |
| 390 | ; |
| 391 | |
| 392 | exp : exp NOTEQUAL exp |
| 393 | { write_exp_elt_opcode (BINOP_NOTEQUAL); } |
| 394 | ; |
| 395 | |
| 396 | exp : exp LEQ exp |
| 397 | { write_exp_elt_opcode (BINOP_LEQ); } |
| 398 | ; |
| 399 | |
| 400 | exp : exp GEQ exp |
| 401 | { write_exp_elt_opcode (BINOP_GEQ); } |
| 402 | ; |
| 403 | |
| 404 | exp : exp '<' exp |
| 405 | { write_exp_elt_opcode (BINOP_LESS); } |
| 406 | ; |
| 407 | |
| 408 | exp : exp '>' exp |
| 409 | { write_exp_elt_opcode (BINOP_GTR); } |
| 410 | ; |
| 411 | |
| 412 | exp : exp '&' exp |
| 413 | { write_exp_elt_opcode (BINOP_BITWISE_AND); } |
| 414 | ; |
| 415 | |
| 416 | exp : exp '^' exp |
| 417 | { write_exp_elt_opcode (BINOP_BITWISE_XOR); } |
| 418 | ; |
| 419 | |
| 420 | exp : exp '|' exp |
| 421 | { write_exp_elt_opcode (BINOP_BITWISE_IOR); } |
| 422 | ; |
| 423 | |
| 424 | exp : exp ANDAND exp |
| 425 | { write_exp_elt_opcode (BINOP_LOGICAL_AND); } |
| 426 | ; |
| 427 | |
| 428 | exp : exp OROR exp |
| 429 | { write_exp_elt_opcode (BINOP_LOGICAL_OR); } |
| 430 | ; |
| 431 | |
| 432 | exp : exp '?' exp ':' exp %prec '?' |
| 433 | { write_exp_elt_opcode (TERNOP_COND); } |
| 434 | ; |
| 435 | |
| 436 | exp : exp '=' exp |
| 437 | { write_exp_elt_opcode (BINOP_ASSIGN); } |
| 438 | ; |
| 439 | |
| 440 | exp : exp ASSIGN_MODIFY exp |
| 441 | { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); |
| 442 | write_exp_elt_opcode ($2); |
| 443 | write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); } |
| 444 | ; |
| 445 | |
| 446 | exp : INT |
| 447 | { write_exp_elt_opcode (OP_LONG); |
| 448 | write_exp_elt_type ($1.type); |
| 449 | write_exp_elt_longcst ((LONGEST)($1.val)); |
| 450 | write_exp_elt_opcode (OP_LONG); } |
| 451 | ; |
| 452 | |
| 453 | exp : NAME_OR_INT |
| 454 | { YYSTYPE val; |
| 455 | parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val); |
| 456 | write_exp_elt_opcode (OP_LONG); |
| 457 | write_exp_elt_type (val.typed_val.type); |
| 458 | write_exp_elt_longcst ((LONGEST)val.typed_val.val); |
| 459 | write_exp_elt_opcode (OP_LONG); |
| 460 | } |
| 461 | ; |
| 462 | |
| 463 | |
| 464 | exp : FLOAT |
| 465 | { write_exp_elt_opcode (OP_DOUBLE); |
| 466 | write_exp_elt_type (builtin_type_double); |
| 467 | write_exp_elt_dblcst ($1); |
| 468 | write_exp_elt_opcode (OP_DOUBLE); } |
| 469 | ; |
| 470 | |
| 471 | exp : variable |
| 472 | ; |
| 473 | |
| 474 | exp : LAST |
| 475 | { write_exp_elt_opcode (OP_LAST); |
| 476 | write_exp_elt_longcst ((LONGEST) $1); |
| 477 | write_exp_elt_opcode (OP_LAST); } |
| 478 | ; |
| 479 | |
| 480 | exp : REGNAME |
| 481 | { write_exp_elt_opcode (OP_REGISTER); |
| 482 | write_exp_elt_longcst ((LONGEST) $1); |
| 483 | write_exp_elt_opcode (OP_REGISTER); } |
| 484 | ; |
| 485 | |
| 486 | exp : VARIABLE |
| 487 | { write_exp_elt_opcode (OP_INTERNALVAR); |
| 488 | write_exp_elt_intern ($1); |
| 489 | write_exp_elt_opcode (OP_INTERNALVAR); } |
| 490 | ; |
| 491 | |
| 492 | exp : SIZEOF '(' type ')' %prec UNARY |
| 493 | { write_exp_elt_opcode (OP_LONG); |
| 494 | write_exp_elt_type (builtin_type_int); |
| 495 | write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3)); |
| 496 | write_exp_elt_opcode (OP_LONG); } |
| 497 | ; |
| 498 | |
| 499 | exp : STRING |
| 500 | { /* C strings are converted into array constants with |
| 501 | an explicit null byte added at the end. Thus |
| 502 | the array upper bound is the string length. |
| 503 | There is no such thing in C as a completely empty |
| 504 | string. */ |
| 505 | char *sp = $1.ptr; int count = $1.length; |
| 506 | while (count-- > 0) |
| 507 | { |
| 508 | write_exp_elt_opcode (OP_LONG); |
| 509 | write_exp_elt_type (builtin_type_char); |
| 510 | write_exp_elt_longcst ((LONGEST)(*sp++)); |
| 511 | write_exp_elt_opcode (OP_LONG); |
| 512 | } |
| 513 | write_exp_elt_opcode (OP_LONG); |
| 514 | write_exp_elt_type (builtin_type_char); |
| 515 | write_exp_elt_longcst ((LONGEST)'\0'); |
| 516 | write_exp_elt_opcode (OP_LONG); |
| 517 | write_exp_elt_opcode (OP_ARRAY); |
| 518 | write_exp_elt_longcst ((LONGEST) 0); |
| 519 | write_exp_elt_longcst ((LONGEST) ($1.length)); |
| 520 | write_exp_elt_opcode (OP_ARRAY); } |
| 521 | ; |
| 522 | |
| 523 | /* C++. */ |
| 524 | exp : THIS |
| 525 | { write_exp_elt_opcode (OP_THIS); |
| 526 | write_exp_elt_opcode (OP_THIS); } |
| 527 | ; |
| 528 | |
| 529 | /* end of C++. */ |
| 530 | |
| 531 | block : BLOCKNAME |
| 532 | { |
| 533 | if ($1.sym != 0) |
| 534 | $$ = SYMBOL_BLOCK_VALUE ($1.sym); |
| 535 | else |
| 536 | { |
| 537 | struct symtab *tem = |
| 538 | lookup_symtab (copy_name ($1.stoken)); |
| 539 | if (tem) |
| 540 | $$ = BLOCKVECTOR_BLOCK |
| 541 | (BLOCKVECTOR (tem), STATIC_BLOCK); |
| 542 | else |
| 543 | error ("No file or function \"%s\".", |
| 544 | copy_name ($1.stoken)); |
| 545 | } |
| 546 | } |
| 547 | ; |
| 548 | |
| 549 | block : block COLONCOLON name |
| 550 | { struct symbol *tem |
| 551 | = lookup_symbol (copy_name ($3), $1, |
| 552 | VAR_NAMESPACE, (int *) NULL, |
| 553 | (struct symtab **) NULL); |
| 554 | if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK) |
| 555 | error ("No function \"%s\" in specified context.", |
| 556 | copy_name ($3)); |
| 557 | $$ = SYMBOL_BLOCK_VALUE (tem); } |
| 558 | ; |
| 559 | |
| 560 | variable: block COLONCOLON name |
| 561 | { struct symbol *sym; |
| 562 | sym = lookup_symbol (copy_name ($3), $1, |
| 563 | VAR_NAMESPACE, (int *) NULL, |
| 564 | (struct symtab **) NULL); |
| 565 | if (sym == 0) |
| 566 | error ("No symbol \"%s\" in specified context.", |
| 567 | copy_name ($3)); |
| 568 | |
| 569 | write_exp_elt_opcode (OP_VAR_VALUE); |
| 570 | /* block_found is set by lookup_symbol. */ |
| 571 | write_exp_elt_block (block_found); |
| 572 | write_exp_elt_sym (sym); |
| 573 | write_exp_elt_opcode (OP_VAR_VALUE); } |
| 574 | ; |
| 575 | |
| 576 | qualified_name: typebase COLONCOLON name |
| 577 | { |
| 578 | struct type *type = $1; |
| 579 | if (TYPE_CODE (type) != TYPE_CODE_STRUCT |
| 580 | && TYPE_CODE (type) != TYPE_CODE_UNION) |
| 581 | error ("`%s' is not defined as an aggregate type.", |
| 582 | TYPE_NAME (type)); |
| 583 | |
| 584 | write_exp_elt_opcode (OP_SCOPE); |
| 585 | write_exp_elt_type (type); |
| 586 | write_exp_string ($3); |
| 587 | write_exp_elt_opcode (OP_SCOPE); |
| 588 | } |
| 589 | | typebase COLONCOLON '~' name |
| 590 | { |
| 591 | struct type *type = $1; |
| 592 | struct stoken tmp_token; |
| 593 | if (TYPE_CODE (type) != TYPE_CODE_STRUCT |
| 594 | && TYPE_CODE (type) != TYPE_CODE_UNION) |
| 595 | error ("`%s' is not defined as an aggregate type.", |
| 596 | TYPE_NAME (type)); |
| 597 | |
| 598 | if (!STREQ (type_name_no_tag (type), $4.ptr)) |
| 599 | error ("invalid destructor `%s::~%s'", |
| 600 | type_name_no_tag (type), $4.ptr); |
| 601 | |
| 602 | tmp_token.ptr = (char*) alloca ($4.length + 2); |
| 603 | tmp_token.length = $4.length + 1; |
| 604 | tmp_token.ptr[0] = '~'; |
| 605 | memcpy (tmp_token.ptr+1, $4.ptr, $4.length); |
| 606 | tmp_token.ptr[tmp_token.length] = 0; |
| 607 | write_exp_elt_opcode (OP_SCOPE); |
| 608 | write_exp_elt_type (type); |
| 609 | write_exp_string (tmp_token); |
| 610 | write_exp_elt_opcode (OP_SCOPE); |
| 611 | } |
| 612 | ; |
| 613 | |
| 614 | variable: qualified_name |
| 615 | | COLONCOLON name |
| 616 | { |
| 617 | char *name = copy_name ($2); |
| 618 | struct symbol *sym; |
| 619 | struct minimal_symbol *msymbol; |
| 620 | |
| 621 | sym = |
| 622 | lookup_symbol (name, (const struct block *) NULL, |
| 623 | VAR_NAMESPACE, (int *) NULL, |
| 624 | (struct symtab **) NULL); |
| 625 | if (sym) |
| 626 | { |
| 627 | write_exp_elt_opcode (OP_VAR_VALUE); |
| 628 | write_exp_elt_block (NULL); |
| 629 | write_exp_elt_sym (sym); |
| 630 | write_exp_elt_opcode (OP_VAR_VALUE); |
| 631 | break; |
| 632 | } |
| 633 | |
| 634 | msymbol = lookup_minimal_symbol (name, |
| 635 | (struct objfile *) NULL); |
| 636 | if (msymbol != NULL) |
| 637 | { |
| 638 | write_exp_msymbol (msymbol, |
| 639 | lookup_function_type (builtin_type_int), |
| 640 | builtin_type_int); |
| 641 | } |
| 642 | else |
| 643 | if (!have_full_symbols () && !have_partial_symbols ()) |
| 644 | error ("No symbol table is loaded. Use the \"file\" command."); |
| 645 | else |
| 646 | error ("No symbol \"%s\" in current context.", name); |
| 647 | } |
| 648 | ; |
| 649 | |
| 650 | variable: name_not_typename |
| 651 | { struct symbol *sym = $1.sym; |
| 652 | |
| 653 | if (sym) |
| 654 | { |
| 655 | if (symbol_read_needs_frame (sym)) |
| 656 | { |
| 657 | if (innermost_block == 0 || |
| 658 | contained_in (block_found, |
| 659 | innermost_block)) |
| 660 | innermost_block = block_found; |
| 661 | } |
| 662 | |
| 663 | write_exp_elt_opcode (OP_VAR_VALUE); |
| 664 | /* We want to use the selected frame, not |
| 665 | another more inner frame which happens to |
| 666 | be in the same block. */ |
| 667 | write_exp_elt_block (NULL); |
| 668 | write_exp_elt_sym (sym); |
| 669 | write_exp_elt_opcode (OP_VAR_VALUE); |
| 670 | } |
| 671 | else if ($1.is_a_field_of_this) |
| 672 | { |
| 673 | /* C++: it hangs off of `this'. Must |
| 674 | not inadvertently convert from a method call |
| 675 | to data ref. */ |
| 676 | if (innermost_block == 0 || |
| 677 | contained_in (block_found, innermost_block)) |
| 678 | innermost_block = block_found; |
| 679 | write_exp_elt_opcode (OP_THIS); |
| 680 | write_exp_elt_opcode (OP_THIS); |
| 681 | write_exp_elt_opcode (STRUCTOP_PTR); |
| 682 | write_exp_string ($1.stoken); |
| 683 | write_exp_elt_opcode (STRUCTOP_PTR); |
| 684 | } |
| 685 | else |
| 686 | { |
| 687 | struct minimal_symbol *msymbol; |
| 688 | register char *arg = copy_name ($1.stoken); |
| 689 | |
| 690 | msymbol = lookup_minimal_symbol (arg, |
| 691 | (struct objfile *) NULL); |
| 692 | if (msymbol != NULL) |
| 693 | { |
| 694 | write_exp_msymbol (msymbol, |
| 695 | lookup_function_type (builtin_type_int), |
| 696 | builtin_type_int); |
| 697 | } |
| 698 | else if (!have_full_symbols () && !have_partial_symbols ()) |
| 699 | error ("No symbol table is loaded. Use the \"file\" command."); |
| 700 | else |
| 701 | error ("No symbol \"%s\" in current context.", |
| 702 | copy_name ($1.stoken)); |
| 703 | } |
| 704 | } |
| 705 | ; |
| 706 | |
| 707 | |
| 708 | ptype : typebase |
| 709 | /* "const" and "volatile" are curently ignored. A type qualifier |
| 710 | before the type is currently handled in the typebase rule. |
| 711 | The reason for recognizing these here (shift/reduce conflicts) |
| 712 | might be obsolete now that some pointer to member rules have |
| 713 | been deleted. */ |
| 714 | | typebase CONST_KEYWORD |
| 715 | | typebase VOLATILE_KEYWORD |
| 716 | | typebase abs_decl |
| 717 | { $$ = follow_types ($1); } |
| 718 | | typebase CONST_KEYWORD abs_decl |
| 719 | { $$ = follow_types ($1); } |
| 720 | | typebase VOLATILE_KEYWORD abs_decl |
| 721 | { $$ = follow_types ($1); } |
| 722 | ; |
| 723 | |
| 724 | abs_decl: '*' |
| 725 | { push_type (tp_pointer); $$ = 0; } |
| 726 | | '*' abs_decl |
| 727 | { push_type (tp_pointer); $$ = $2; } |
| 728 | | '&' |
| 729 | { push_type (tp_reference); $$ = 0; } |
| 730 | | '&' abs_decl |
| 731 | { push_type (tp_reference); $$ = $2; } |
| 732 | | direct_abs_decl |
| 733 | ; |
| 734 | |
| 735 | direct_abs_decl: '(' abs_decl ')' |
| 736 | { $$ = $2; } |
| 737 | | direct_abs_decl array_mod |
| 738 | { |
| 739 | push_type_int ($2); |
| 740 | push_type (tp_array); |
| 741 | } |
| 742 | | array_mod |
| 743 | { |
| 744 | push_type_int ($1); |
| 745 | push_type (tp_array); |
| 746 | $$ = 0; |
| 747 | } |
| 748 | |
| 749 | | direct_abs_decl func_mod |
| 750 | { push_type (tp_function); } |
| 751 | | func_mod |
| 752 | { push_type (tp_function); } |
| 753 | ; |
| 754 | |
| 755 | array_mod: '[' ']' |
| 756 | { $$ = -1; } |
| 757 | | '[' INT ']' |
| 758 | { $$ = $2.val; } |
| 759 | ; |
| 760 | |
| 761 | func_mod: '(' ')' |
| 762 | { $$ = 0; } |
| 763 | | '(' nonempty_typelist ')' |
| 764 | { free ((PTR)$2); $$ = 0; } |
| 765 | ; |
| 766 | |
| 767 | /* We used to try to recognize more pointer to member types here, but |
| 768 | that didn't work (shift/reduce conflicts meant that these rules never |
| 769 | got executed). The problem is that |
| 770 | int (foo::bar::baz::bizzle) |
| 771 | is a function type but |
| 772 | int (foo::bar::baz::bizzle::*) |
| 773 | is a pointer to member type. Stroustrup loses again! */ |
| 774 | |
| 775 | type : ptype |
| 776 | | typebase COLONCOLON '*' |
| 777 | { $$ = lookup_member_type (builtin_type_int, $1); } |
| 778 | ; |
| 779 | |
| 780 | typebase /* Implements (approximately): (type-qualifier)* type-specifier */ |
| 781 | : TYPENAME |
| 782 | { $$ = $1.type; } |
| 783 | | INT_KEYWORD |
| 784 | { $$ = builtin_type_int; } |
| 785 | | LONG |
| 786 | { $$ = builtin_type_long; } |
| 787 | | SHORT |
| 788 | { $$ = builtin_type_short; } |
| 789 | | LONG INT_KEYWORD |
| 790 | { $$ = builtin_type_long; } |
| 791 | | UNSIGNED LONG INT_KEYWORD |
| 792 | { $$ = builtin_type_unsigned_long; } |
| 793 | | LONG LONG |
| 794 | { $$ = builtin_type_long_long; } |
| 795 | | LONG LONG INT_KEYWORD |
| 796 | { $$ = builtin_type_long_long; } |
| 797 | | UNSIGNED LONG LONG |
| 798 | { $$ = builtin_type_unsigned_long_long; } |
| 799 | | UNSIGNED LONG LONG INT_KEYWORD |
| 800 | { $$ = builtin_type_unsigned_long_long; } |
| 801 | | SHORT INT_KEYWORD |
| 802 | { $$ = builtin_type_short; } |
| 803 | | UNSIGNED SHORT INT_KEYWORD |
| 804 | { $$ = builtin_type_unsigned_short; } |
| 805 | | STRUCT name |
| 806 | { $$ = lookup_struct (copy_name ($2), |
| 807 | expression_context_block); } |
| 808 | | CLASS name |
| 809 | { $$ = lookup_struct (copy_name ($2), |
| 810 | expression_context_block); } |
| 811 | | UNION name |
| 812 | { $$ = lookup_union (copy_name ($2), |
| 813 | expression_context_block); } |
| 814 | | ENUM name |
| 815 | { $$ = lookup_enum (copy_name ($2), |
| 816 | expression_context_block); } |
| 817 | | UNSIGNED typename |
| 818 | { $$ = lookup_unsigned_typename (TYPE_NAME($2.type)); } |
| 819 | | UNSIGNED |
| 820 | { $$ = builtin_type_unsigned_int; } |
| 821 | | SIGNED_KEYWORD typename |
| 822 | { $$ = lookup_signed_typename (TYPE_NAME($2.type)); } |
| 823 | | SIGNED_KEYWORD |
| 824 | { $$ = builtin_type_int; } |
| 825 | | TEMPLATE name '<' type '>' |
| 826 | { $$ = lookup_template_type(copy_name($2), $4, |
| 827 | expression_context_block); |
| 828 | } |
| 829 | /* "const" and "volatile" are curently ignored. A type qualifier |
| 830 | after the type is handled in the ptype rule. I think these could |
| 831 | be too. */ |
| 832 | | CONST_KEYWORD typebase { $$ = $2; } |
| 833 | | VOLATILE_KEYWORD typebase { $$ = $2; } |
| 834 | ; |
| 835 | |
| 836 | typename: TYPENAME |
| 837 | | INT_KEYWORD |
| 838 | { |
| 839 | $$.stoken.ptr = "int"; |
| 840 | $$.stoken.length = 3; |
| 841 | $$.type = builtin_type_int; |
| 842 | } |
| 843 | | LONG |
| 844 | { |
| 845 | $$.stoken.ptr = "long"; |
| 846 | $$.stoken.length = 4; |
| 847 | $$.type = builtin_type_long; |
| 848 | } |
| 849 | | SHORT |
| 850 | { |
| 851 | $$.stoken.ptr = "short"; |
| 852 | $$.stoken.length = 5; |
| 853 | $$.type = builtin_type_short; |
| 854 | } |
| 855 | ; |
| 856 | |
| 857 | nonempty_typelist |
| 858 | : type |
| 859 | { $$ = (struct type **) malloc (sizeof (struct type *) * 2); |
| 860 | $<ivec>$[0] = 1; /* Number of types in vector */ |
| 861 | $$[1] = $1; |
| 862 | } |
| 863 | | nonempty_typelist ',' type |
| 864 | { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1); |
| 865 | $$ = (struct type **) realloc ((char *) $1, len); |
| 866 | $$[$<ivec>$[0]] = $3; |
| 867 | } |
| 868 | ; |
| 869 | |
| 870 | name : NAME { $$ = $1.stoken; } |
| 871 | | BLOCKNAME { $$ = $1.stoken; } |
| 872 | | TYPENAME { $$ = $1.stoken; } |
| 873 | | NAME_OR_INT { $$ = $1.stoken; } |
| 874 | ; |
| 875 | |
| 876 | name_not_typename : NAME |
| 877 | | BLOCKNAME |
| 878 | /* These would be useful if name_not_typename was useful, but it is just |
| 879 | a fake for "variable", so these cause reduce/reduce conflicts because |
| 880 | the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable, |
| 881 | =exp) or just an exp. If name_not_typename was ever used in an lvalue |
| 882 | context where only a name could occur, this might be useful. |
| 883 | | NAME_OR_INT |
| 884 | */ |
| 885 | ; |
| 886 | |
| 887 | %% |
| 888 | |
| 889 | /* Take care of parsing a number (anything that starts with a digit). |
| 890 | Set yylval and return the token type; update lexptr. |
| 891 | LEN is the number of characters in it. */ |
| 892 | |
| 893 | /*** Needs some error checking for the float case ***/ |
| 894 | |
| 895 | static int |
| 896 | parse_number (p, len, parsed_float, putithere) |
| 897 | register char *p; |
| 898 | register int len; |
| 899 | int parsed_float; |
| 900 | YYSTYPE *putithere; |
| 901 | { |
| 902 | /* FIXME: Shouldn't these be unsigned? We don't deal with negative values |
| 903 | here, and we do kind of silly things like cast to unsigned. */ |
| 904 | register LONGEST n = 0; |
| 905 | register LONGEST prevn = 0; |
| 906 | unsigned LONGEST un; |
| 907 | |
| 908 | register int i = 0; |
| 909 | register int c; |
| 910 | register int base = input_radix; |
| 911 | int unsigned_p = 0; |
| 912 | |
| 913 | /* Number of "L" suffixes encountered. */ |
| 914 | int long_p = 0; |
| 915 | |
| 916 | /* We have found a "L" or "U" suffix. */ |
| 917 | int found_suffix = 0; |
| 918 | |
| 919 | unsigned LONGEST high_bit; |
| 920 | struct type *signed_type; |
| 921 | struct type *unsigned_type; |
| 922 | |
| 923 | if (parsed_float) |
| 924 | { |
| 925 | /* It's a float since it contains a point or an exponent. */ |
| 926 | putithere->dval = atof (p); |
| 927 | return FLOAT; |
| 928 | } |
| 929 | |
| 930 | /* Handle base-switching prefixes 0x, 0t, 0d, 0 */ |
| 931 | if (p[0] == '0') |
| 932 | switch (p[1]) |
| 933 | { |
| 934 | case 'x': |
| 935 | case 'X': |
| 936 | if (len >= 3) |
| 937 | { |
| 938 | p += 2; |
| 939 | base = 16; |
| 940 | len -= 2; |
| 941 | } |
| 942 | break; |
| 943 | |
| 944 | case 't': |
| 945 | case 'T': |
| 946 | case 'd': |
| 947 | case 'D': |
| 948 | if (len >= 3) |
| 949 | { |
| 950 | p += 2; |
| 951 | base = 10; |
| 952 | len -= 2; |
| 953 | } |
| 954 | break; |
| 955 | |
| 956 | default: |
| 957 | base = 8; |
| 958 | break; |
| 959 | } |
| 960 | |
| 961 | while (len-- > 0) |
| 962 | { |
| 963 | c = *p++; |
| 964 | if (c >= 'A' && c <= 'Z') |
| 965 | c += 'a' - 'A'; |
| 966 | if (c != 'l' && c != 'u') |
| 967 | n *= base; |
| 968 | if (c >= '0' && c <= '9') |
| 969 | { |
| 970 | if (found_suffix) |
| 971 | return ERROR; |
| 972 | n += i = c - '0'; |
| 973 | } |
| 974 | else |
| 975 | { |
| 976 | if (base > 10 && c >= 'a' && c <= 'f') |
| 977 | { |
| 978 | if (found_suffix) |
| 979 | return ERROR; |
| 980 | n += i = c - 'a' + 10; |
| 981 | } |
| 982 | else if (c == 'l') |
| 983 | { |
| 984 | ++long_p; |
| 985 | found_suffix = 1; |
| 986 | } |
| 987 | else if (c == 'u') |
| 988 | { |
| 989 | unsigned_p = 1; |
| 990 | found_suffix = 1; |
| 991 | } |
| 992 | else |
| 993 | return ERROR; /* Char not a digit */ |
| 994 | } |
| 995 | if (i >= base) |
| 996 | return ERROR; /* Invalid digit in this base */ |
| 997 | |
| 998 | /* Portably test for overflow (only works for nonzero values, so make |
| 999 | a second check for zero). FIXME: Can't we just make n and prevn |
| 1000 | unsigned and avoid this? */ |
| 1001 | if (c != 'l' && c != 'u' && (prevn >= n) && n != 0) |
| 1002 | unsigned_p = 1; /* Try something unsigned */ |
| 1003 | |
| 1004 | /* Portably test for unsigned overflow. |
| 1005 | FIXME: This check is wrong; for example it doesn't find overflow |
| 1006 | on 0x123456789 when LONGEST is 32 bits. */ |
| 1007 | if (c != 'l' && c != 'u' && n != 0) |
| 1008 | { |
| 1009 | if ((unsigned_p && (unsigned LONGEST) prevn >= (unsigned LONGEST) n)) |
| 1010 | error ("Numeric constant too large."); |
| 1011 | } |
| 1012 | prevn = n; |
| 1013 | } |
| 1014 | |
| 1015 | /* An integer constant is an int, a long, or a long long. An L |
| 1016 | suffix forces it to be long; an LL suffix forces it to be long |
| 1017 | long. If not forced to a larger size, it gets the first type of |
| 1018 | the above that it fits in. To figure out whether it fits, we |
| 1019 | shift it right and see whether anything remains. Note that we |
| 1020 | can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one |
| 1021 | operation, because many compilers will warn about such a shift |
| 1022 | (which always produces a zero result). Sometimes TARGET_INT_BIT |
| 1023 | or TARGET_LONG_BIT will be that big, sometimes not. To deal with |
| 1024 | the case where it is we just always shift the value more than |
| 1025 | once, with fewer bits each time. */ |
| 1026 | |
| 1027 | un = (unsigned LONGEST)n >> 2; |
| 1028 | if (long_p == 0 |
| 1029 | && (un >> (TARGET_INT_BIT - 2)) == 0) |
| 1030 | { |
| 1031 | high_bit = ((unsigned LONGEST)1) << (TARGET_INT_BIT-1); |
| 1032 | |
| 1033 | /* A large decimal (not hex or octal) constant (between INT_MAX |
| 1034 | and UINT_MAX) is a long or unsigned long, according to ANSI, |
| 1035 | never an unsigned int, but this code treats it as unsigned |
| 1036 | int. This probably should be fixed. GCC gives a warning on |
| 1037 | such constants. */ |
| 1038 | |
| 1039 | unsigned_type = builtin_type_unsigned_int; |
| 1040 | signed_type = builtin_type_int; |
| 1041 | } |
| 1042 | else if (long_p <= 1 |
| 1043 | && (un >> (TARGET_LONG_BIT - 2)) == 0) |
| 1044 | { |
| 1045 | high_bit = ((unsigned LONGEST)1) << (TARGET_LONG_BIT-1); |
| 1046 | unsigned_type = builtin_type_unsigned_long; |
| 1047 | signed_type = builtin_type_long; |
| 1048 | } |
| 1049 | else |
| 1050 | { |
| 1051 | high_bit = (((unsigned LONGEST)1) |
| 1052 | << (TARGET_LONG_LONG_BIT - 32 - 1) |
| 1053 | << 16 |
| 1054 | << 16); |
| 1055 | if (high_bit == 0) |
| 1056 | /* A long long does not fit in a LONGEST. */ |
| 1057 | high_bit = (unsigned LONGEST)1 << sizeof (LONGEST) * HOST_CHAR_BIT - 1; |
| 1058 | unsigned_type = builtin_type_unsigned_long_long; |
| 1059 | signed_type = builtin_type_long_long; |
| 1060 | } |
| 1061 | |
| 1062 | putithere->typed_val.val = n; |
| 1063 | |
| 1064 | /* If the high bit of the worked out type is set then this number |
| 1065 | has to be unsigned. */ |
| 1066 | |
| 1067 | if (unsigned_p || (n & high_bit)) |
| 1068 | { |
| 1069 | putithere->typed_val.type = unsigned_type; |
| 1070 | } |
| 1071 | else |
| 1072 | { |
| 1073 | putithere->typed_val.type = signed_type; |
| 1074 | } |
| 1075 | |
| 1076 | return INT; |
| 1077 | } |
| 1078 | |
| 1079 | struct token |
| 1080 | { |
| 1081 | char *operator; |
| 1082 | int token; |
| 1083 | enum exp_opcode opcode; |
| 1084 | }; |
| 1085 | |
| 1086 | static const struct token tokentab3[] = |
| 1087 | { |
| 1088 | {">>=", ASSIGN_MODIFY, BINOP_RSH}, |
| 1089 | {"<<=", ASSIGN_MODIFY, BINOP_LSH} |
| 1090 | }; |
| 1091 | |
| 1092 | static const struct token tokentab2[] = |
| 1093 | { |
| 1094 | {"+=", ASSIGN_MODIFY, BINOP_ADD}, |
| 1095 | {"-=", ASSIGN_MODIFY, BINOP_SUB}, |
| 1096 | {"*=", ASSIGN_MODIFY, BINOP_MUL}, |
| 1097 | {"/=", ASSIGN_MODIFY, BINOP_DIV}, |
| 1098 | {"%=", ASSIGN_MODIFY, BINOP_REM}, |
| 1099 | {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR}, |
| 1100 | {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND}, |
| 1101 | {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR}, |
| 1102 | {"++", INCREMENT, BINOP_END}, |
| 1103 | {"--", DECREMENT, BINOP_END}, |
| 1104 | {"->", ARROW, BINOP_END}, |
| 1105 | {"&&", ANDAND, BINOP_END}, |
| 1106 | {"||", OROR, BINOP_END}, |
| 1107 | {"::", COLONCOLON, BINOP_END}, |
| 1108 | {"<<", LSH, BINOP_END}, |
| 1109 | {">>", RSH, BINOP_END}, |
| 1110 | {"==", EQUAL, BINOP_END}, |
| 1111 | {"!=", NOTEQUAL, BINOP_END}, |
| 1112 | {"<=", LEQ, BINOP_END}, |
| 1113 | {">=", GEQ, BINOP_END} |
| 1114 | }; |
| 1115 | |
| 1116 | /* Read one token, getting characters through lexptr. */ |
| 1117 | |
| 1118 | static int |
| 1119 | yylex () |
| 1120 | { |
| 1121 | int c; |
| 1122 | int namelen; |
| 1123 | unsigned int i; |
| 1124 | char *tokstart; |
| 1125 | char *tokptr; |
| 1126 | int tempbufindex; |
| 1127 | static char *tempbuf; |
| 1128 | static int tempbufsize; |
| 1129 | |
| 1130 | retry: |
| 1131 | |
| 1132 | tokstart = lexptr; |
| 1133 | /* See if it is a special token of length 3. */ |
| 1134 | for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++) |
| 1135 | if (STREQN (tokstart, tokentab3[i].operator, 3)) |
| 1136 | { |
| 1137 | lexptr += 3; |
| 1138 | yylval.opcode = tokentab3[i].opcode; |
| 1139 | return tokentab3[i].token; |
| 1140 | } |
| 1141 | |
| 1142 | /* See if it is a special token of length 2. */ |
| 1143 | for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++) |
| 1144 | if (STREQN (tokstart, tokentab2[i].operator, 2)) |
| 1145 | { |
| 1146 | lexptr += 2; |
| 1147 | yylval.opcode = tokentab2[i].opcode; |
| 1148 | return tokentab2[i].token; |
| 1149 | } |
| 1150 | |
| 1151 | switch (c = *tokstart) |
| 1152 | { |
| 1153 | case 0: |
| 1154 | return 0; |
| 1155 | |
| 1156 | case ' ': |
| 1157 | case '\t': |
| 1158 | case '\n': |
| 1159 | lexptr++; |
| 1160 | goto retry; |
| 1161 | |
| 1162 | case '\'': |
| 1163 | /* We either have a character constant ('0' or '\177' for example) |
| 1164 | or we have a quoted symbol reference ('foo(int,int)' in C++ |
| 1165 | for example). */ |
| 1166 | lexptr++; |
| 1167 | c = *lexptr++; |
| 1168 | if (c == '\\') |
| 1169 | c = parse_escape (&lexptr); |
| 1170 | |
| 1171 | yylval.typed_val.val = c; |
| 1172 | yylval.typed_val.type = builtin_type_char; |
| 1173 | |
| 1174 | c = *lexptr++; |
| 1175 | if (c != '\'') |
| 1176 | { |
| 1177 | namelen = skip_quoted (tokstart) - tokstart; |
| 1178 | if (namelen > 2) |
| 1179 | { |
| 1180 | lexptr = tokstart + namelen; |
| 1181 | if (lexptr[-1] != '\'') |
| 1182 | error ("Unmatched single quote."); |
| 1183 | namelen -= 2; |
| 1184 | tokstart++; |
| 1185 | goto tryname; |
| 1186 | } |
| 1187 | error ("Invalid character constant."); |
| 1188 | } |
| 1189 | return INT; |
| 1190 | |
| 1191 | case '(': |
| 1192 | paren_depth++; |
| 1193 | lexptr++; |
| 1194 | return c; |
| 1195 | |
| 1196 | case ')': |
| 1197 | if (paren_depth == 0) |
| 1198 | return 0; |
| 1199 | paren_depth--; |
| 1200 | lexptr++; |
| 1201 | return c; |
| 1202 | |
| 1203 | case ',': |
| 1204 | if (comma_terminates && paren_depth == 0) |
| 1205 | return 0; |
| 1206 | lexptr++; |
| 1207 | return c; |
| 1208 | |
| 1209 | case '.': |
| 1210 | /* Might be a floating point number. */ |
| 1211 | if (lexptr[1] < '0' || lexptr[1] > '9') |
| 1212 | goto symbol; /* Nope, must be a symbol. */ |
| 1213 | /* FALL THRU into number case. */ |
| 1214 | |
| 1215 | case '0': |
| 1216 | case '1': |
| 1217 | case '2': |
| 1218 | case '3': |
| 1219 | case '4': |
| 1220 | case '5': |
| 1221 | case '6': |
| 1222 | case '7': |
| 1223 | case '8': |
| 1224 | case '9': |
| 1225 | { |
| 1226 | /* It's a number. */ |
| 1227 | int got_dot = 0, got_e = 0, toktype; |
| 1228 | register char *p = tokstart; |
| 1229 | int hex = input_radix > 10; |
| 1230 | |
| 1231 | if (c == '0' && (p[1] == 'x' || p[1] == 'X')) |
| 1232 | { |
| 1233 | p += 2; |
| 1234 | hex = 1; |
| 1235 | } |
| 1236 | else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D')) |
| 1237 | { |
| 1238 | p += 2; |
| 1239 | hex = 0; |
| 1240 | } |
| 1241 | |
| 1242 | for (;; ++p) |
| 1243 | { |
| 1244 | /* This test includes !hex because 'e' is a valid hex digit |
| 1245 | and thus does not indicate a floating point number when |
| 1246 | the radix is hex. */ |
| 1247 | if (!hex && !got_e && (*p == 'e' || *p == 'E')) |
| 1248 | got_dot = got_e = 1; |
| 1249 | /* This test does not include !hex, because a '.' always indicates |
| 1250 | a decimal floating point number regardless of the radix. */ |
| 1251 | else if (!got_dot && *p == '.') |
| 1252 | got_dot = 1; |
| 1253 | else if (got_e && (p[-1] == 'e' || p[-1] == 'E') |
| 1254 | && (*p == '-' || *p == '+')) |
| 1255 | /* This is the sign of the exponent, not the end of the |
| 1256 | number. */ |
| 1257 | continue; |
| 1258 | /* We will take any letters or digits. parse_number will |
| 1259 | complain if past the radix, or if L or U are not final. */ |
| 1260 | else if ((*p < '0' || *p > '9') |
| 1261 | && ((*p < 'a' || *p > 'z') |
| 1262 | && (*p < 'A' || *p > 'Z'))) |
| 1263 | break; |
| 1264 | } |
| 1265 | toktype = parse_number (tokstart, p - tokstart, got_dot|got_e, &yylval); |
| 1266 | if (toktype == ERROR) |
| 1267 | { |
| 1268 | char *err_copy = (char *) alloca (p - tokstart + 1); |
| 1269 | |
| 1270 | memcpy (err_copy, tokstart, p - tokstart); |
| 1271 | err_copy[p - tokstart] = 0; |
| 1272 | error ("Invalid number \"%s\".", err_copy); |
| 1273 | } |
| 1274 | lexptr = p; |
| 1275 | return toktype; |
| 1276 | } |
| 1277 | |
| 1278 | case '+': |
| 1279 | case '-': |
| 1280 | case '*': |
| 1281 | case '/': |
| 1282 | case '%': |
| 1283 | case '|': |
| 1284 | case '&': |
| 1285 | case '^': |
| 1286 | case '~': |
| 1287 | case '!': |
| 1288 | case '@': |
| 1289 | case '<': |
| 1290 | case '>': |
| 1291 | case '[': |
| 1292 | case ']': |
| 1293 | case '?': |
| 1294 | case ':': |
| 1295 | case '=': |
| 1296 | case '{': |
| 1297 | case '}': |
| 1298 | symbol: |
| 1299 | lexptr++; |
| 1300 | return c; |
| 1301 | |
| 1302 | case '"': |
| 1303 | |
| 1304 | /* Build the gdb internal form of the input string in tempbuf, |
| 1305 | translating any standard C escape forms seen. Note that the |
| 1306 | buffer is null byte terminated *only* for the convenience of |
| 1307 | debugging gdb itself and printing the buffer contents when |
| 1308 | the buffer contains no embedded nulls. Gdb does not depend |
| 1309 | upon the buffer being null byte terminated, it uses the length |
| 1310 | string instead. This allows gdb to handle C strings (as well |
| 1311 | as strings in other languages) with embedded null bytes */ |
| 1312 | |
| 1313 | tokptr = ++tokstart; |
| 1314 | tempbufindex = 0; |
| 1315 | |
| 1316 | do { |
| 1317 | /* Grow the static temp buffer if necessary, including allocating |
| 1318 | the first one on demand. */ |
| 1319 | if (tempbufindex + 1 >= tempbufsize) |
| 1320 | { |
| 1321 | tempbuf = (char *) realloc (tempbuf, tempbufsize += 64); |
| 1322 | } |
| 1323 | switch (*tokptr) |
| 1324 | { |
| 1325 | case '\0': |
| 1326 | case '"': |
| 1327 | /* Do nothing, loop will terminate. */ |
| 1328 | break; |
| 1329 | case '\\': |
| 1330 | tokptr++; |
| 1331 | c = parse_escape (&tokptr); |
| 1332 | if (c == -1) |
| 1333 | { |
| 1334 | continue; |
| 1335 | } |
| 1336 | tempbuf[tempbufindex++] = c; |
| 1337 | break; |
| 1338 | default: |
| 1339 | tempbuf[tempbufindex++] = *tokptr++; |
| 1340 | break; |
| 1341 | } |
| 1342 | } while ((*tokptr != '"') && (*tokptr != '\0')); |
| 1343 | if (*tokptr++ != '"') |
| 1344 | { |
| 1345 | error ("Unterminated string in expression."); |
| 1346 | } |
| 1347 | tempbuf[tempbufindex] = '\0'; /* See note above */ |
| 1348 | yylval.sval.ptr = tempbuf; |
| 1349 | yylval.sval.length = tempbufindex; |
| 1350 | lexptr = tokptr; |
| 1351 | return (STRING); |
| 1352 | } |
| 1353 | |
| 1354 | if (!(c == '_' || c == '$' |
| 1355 | || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))) |
| 1356 | /* We must have come across a bad character (e.g. ';'). */ |
| 1357 | error ("Invalid character '%c' in expression.", c); |
| 1358 | |
| 1359 | /* It's a name. See how long it is. */ |
| 1360 | namelen = 0; |
| 1361 | for (c = tokstart[namelen]; |
| 1362 | (c == '_' || c == '$' || (c >= '0' && c <= '9') |
| 1363 | || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')); |
| 1364 | c = tokstart[++namelen]) |
| 1365 | ; |
| 1366 | |
| 1367 | /* The token "if" terminates the expression and is NOT |
| 1368 | removed from the input stream. */ |
| 1369 | if (namelen == 2 && tokstart[0] == 'i' && tokstart[1] == 'f') |
| 1370 | { |
| 1371 | return 0; |
| 1372 | } |
| 1373 | |
| 1374 | lexptr += namelen; |
| 1375 | |
| 1376 | /* Handle the tokens $digits; also $ (short for $0) and $$ (short for $$1) |
| 1377 | and $$digits (equivalent to $<-digits> if you could type that). |
| 1378 | Make token type LAST, and put the number (the digits) in yylval. */ |
| 1379 | |
| 1380 | tryname: |
| 1381 | if (*tokstart == '$') |
| 1382 | { |
| 1383 | register int negate = 0; |
| 1384 | c = 1; |
| 1385 | /* Double dollar means negate the number and add -1 as well. |
| 1386 | Thus $$ alone means -1. */ |
| 1387 | if (namelen >= 2 && tokstart[1] == '$') |
| 1388 | { |
| 1389 | negate = 1; |
| 1390 | c = 2; |
| 1391 | } |
| 1392 | if (c == namelen) |
| 1393 | { |
| 1394 | /* Just dollars (one or two) */ |
| 1395 | yylval.lval = - negate; |
| 1396 | return LAST; |
| 1397 | } |
| 1398 | /* Is the rest of the token digits? */ |
| 1399 | for (; c < namelen; c++) |
| 1400 | if (!(tokstart[c] >= '0' && tokstart[c] <= '9')) |
| 1401 | break; |
| 1402 | if (c == namelen) |
| 1403 | { |
| 1404 | yylval.lval = atoi (tokstart + 1 + negate); |
| 1405 | if (negate) |
| 1406 | yylval.lval = - yylval.lval; |
| 1407 | return LAST; |
| 1408 | } |
| 1409 | } |
| 1410 | |
| 1411 | /* Handle tokens that refer to machine registers: |
| 1412 | $ followed by a register name. */ |
| 1413 | |
| 1414 | if (*tokstart == '$') { |
| 1415 | for (c = 0; c < NUM_REGS; c++) |
| 1416 | if (namelen - 1 == strlen (reg_names[c]) |
| 1417 | && STREQN (tokstart + 1, reg_names[c], namelen - 1)) |
| 1418 | { |
| 1419 | yylval.lval = c; |
| 1420 | return REGNAME; |
| 1421 | } |
| 1422 | for (c = 0; c < num_std_regs; c++) |
| 1423 | if (namelen - 1 == strlen (std_regs[c].name) |
| 1424 | && STREQN (tokstart + 1, std_regs[c].name, namelen - 1)) |
| 1425 | { |
| 1426 | yylval.lval = std_regs[c].regnum; |
| 1427 | return REGNAME; |
| 1428 | } |
| 1429 | } |
| 1430 | /* Catch specific keywords. Should be done with a data structure. */ |
| 1431 | switch (namelen) |
| 1432 | { |
| 1433 | case 8: |
| 1434 | if (STREQN (tokstart, "unsigned", 8)) |
| 1435 | return UNSIGNED; |
| 1436 | if (current_language->la_language == language_cplus |
| 1437 | && STREQN (tokstart, "template", 8)) |
| 1438 | return TEMPLATE; |
| 1439 | if (STREQN (tokstart, "volatile", 8)) |
| 1440 | return VOLATILE_KEYWORD; |
| 1441 | break; |
| 1442 | case 6: |
| 1443 | if (STREQN (tokstart, "struct", 6)) |
| 1444 | return STRUCT; |
| 1445 | if (STREQN (tokstart, "signed", 6)) |
| 1446 | return SIGNED_KEYWORD; |
| 1447 | if (STREQN (tokstart, "sizeof", 6)) |
| 1448 | return SIZEOF; |
| 1449 | break; |
| 1450 | case 5: |
| 1451 | if (current_language->la_language == language_cplus |
| 1452 | && STREQN (tokstart, "class", 5)) |
| 1453 | return CLASS; |
| 1454 | if (STREQN (tokstart, "union", 5)) |
| 1455 | return UNION; |
| 1456 | if (STREQN (tokstart, "short", 5)) |
| 1457 | return SHORT; |
| 1458 | if (STREQN (tokstart, "const", 5)) |
| 1459 | return CONST_KEYWORD; |
| 1460 | break; |
| 1461 | case 4: |
| 1462 | if (STREQN (tokstart, "enum", 4)) |
| 1463 | return ENUM; |
| 1464 | if (STREQN (tokstart, "long", 4)) |
| 1465 | return LONG; |
| 1466 | if (current_language->la_language == language_cplus |
| 1467 | && STREQN (tokstart, "this", 4)) |
| 1468 | { |
| 1469 | static const char this_name[] = |
| 1470 | { CPLUS_MARKER, 't', 'h', 'i', 's', '\0' }; |
| 1471 | |
| 1472 | if (lookup_symbol (this_name, expression_context_block, |
| 1473 | VAR_NAMESPACE, (int *) NULL, |
| 1474 | (struct symtab **) NULL)) |
| 1475 | return THIS; |
| 1476 | } |
| 1477 | break; |
| 1478 | case 3: |
| 1479 | if (STREQN (tokstart, "int", 3)) |
| 1480 | return INT_KEYWORD; |
| 1481 | break; |
| 1482 | default: |
| 1483 | break; |
| 1484 | } |
| 1485 | |
| 1486 | yylval.sval.ptr = tokstart; |
| 1487 | yylval.sval.length = namelen; |
| 1488 | |
| 1489 | /* Any other names starting in $ are debugger internal variables. */ |
| 1490 | |
| 1491 | if (*tokstart == '$') |
| 1492 | { |
| 1493 | yylval.ivar = lookup_internalvar (copy_name (yylval.sval) + 1); |
| 1494 | return VARIABLE; |
| 1495 | } |
| 1496 | |
| 1497 | /* Use token-type BLOCKNAME for symbols that happen to be defined as |
| 1498 | functions or symtabs. If this is not so, then ... |
| 1499 | Use token-type TYPENAME for symbols that happen to be defined |
| 1500 | currently as names of types; NAME for other symbols. |
| 1501 | The caller is not constrained to care about the distinction. */ |
| 1502 | { |
| 1503 | char *tmp = copy_name (yylval.sval); |
| 1504 | struct symbol *sym; |
| 1505 | int is_a_field_of_this = 0; |
| 1506 | int hextype; |
| 1507 | |
| 1508 | sym = lookup_symbol (tmp, expression_context_block, |
| 1509 | VAR_NAMESPACE, |
| 1510 | current_language->la_language == language_cplus |
| 1511 | ? &is_a_field_of_this : (int *) NULL, |
| 1512 | (struct symtab **) NULL); |
| 1513 | /* Call lookup_symtab, not lookup_partial_symtab, in case there are |
| 1514 | no psymtabs (coff, xcoff, or some future change to blow away the |
| 1515 | psymtabs once once symbols are read). */ |
| 1516 | if ((sym && SYMBOL_CLASS (sym) == LOC_BLOCK) || |
| 1517 | lookup_symtab (tmp)) |
| 1518 | { |
| 1519 | yylval.ssym.sym = sym; |
| 1520 | yylval.ssym.is_a_field_of_this = is_a_field_of_this; |
| 1521 | return BLOCKNAME; |
| 1522 | } |
| 1523 | if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF) |
| 1524 | { |
| 1525 | #if 1 |
| 1526 | /* Despite the following flaw, we need to keep this code enabled. |
| 1527 | Because we can get called from check_stub_method, if we don't |
| 1528 | handle nested types then it screws many operations in any |
| 1529 | program which uses nested types. */ |
| 1530 | /* In "A::x", if x is a member function of A and there happens |
| 1531 | to be a type (nested or not, since the stabs don't make that |
| 1532 | distinction) named x, then this code incorrectly thinks we |
| 1533 | are dealing with nested types rather than a member function. */ |
| 1534 | |
| 1535 | char *p; |
| 1536 | char *namestart; |
| 1537 | struct symbol *best_sym; |
| 1538 | |
| 1539 | /* Look ahead to detect nested types. This probably should be |
| 1540 | done in the grammar, but trying seemed to introduce a lot |
| 1541 | of shift/reduce and reduce/reduce conflicts. It's possible |
| 1542 | that it could be done, though. Or perhaps a non-grammar, but |
| 1543 | less ad hoc, approach would work well. */ |
| 1544 | |
| 1545 | /* Since we do not currently have any way of distinguishing |
| 1546 | a nested type from a non-nested one (the stabs don't tell |
| 1547 | us whether a type is nested), we just ignore the |
| 1548 | containing type. */ |
| 1549 | |
| 1550 | p = lexptr; |
| 1551 | best_sym = sym; |
| 1552 | while (1) |
| 1553 | { |
| 1554 | /* Skip whitespace. */ |
| 1555 | while (*p == ' ' || *p == '\t' || *p == '\n') |
| 1556 | ++p; |
| 1557 | if (*p == ':' && p[1] == ':') |
| 1558 | { |
| 1559 | /* Skip the `::'. */ |
| 1560 | p += 2; |
| 1561 | /* Skip whitespace. */ |
| 1562 | while (*p == ' ' || *p == '\t' || *p == '\n') |
| 1563 | ++p; |
| 1564 | namestart = p; |
| 1565 | while (*p == '_' || *p == '$' || (*p >= '0' && *p <= '9') |
| 1566 | || (*p >= 'a' && *p <= 'z') |
| 1567 | || (*p >= 'A' && *p <= 'Z')) |
| 1568 | ++p; |
| 1569 | if (p != namestart) |
| 1570 | { |
| 1571 | struct symbol *cur_sym; |
| 1572 | /* As big as the whole rest of the expression, which is |
| 1573 | at least big enough. */ |
| 1574 | char *tmp = alloca (strlen (namestart)); |
| 1575 | |
| 1576 | memcpy (tmp, namestart, p - namestart); |
| 1577 | tmp[p - namestart] = '\0'; |
| 1578 | cur_sym = lookup_symbol (tmp, expression_context_block, |
| 1579 | VAR_NAMESPACE, (int *) NULL, |
| 1580 | (struct symtab **) NULL); |
| 1581 | if (cur_sym) |
| 1582 | { |
| 1583 | if (SYMBOL_CLASS (cur_sym) == LOC_TYPEDEF) |
| 1584 | { |
| 1585 | best_sym = cur_sym; |
| 1586 | lexptr = p; |
| 1587 | } |
| 1588 | else |
| 1589 | break; |
| 1590 | } |
| 1591 | else |
| 1592 | break; |
| 1593 | } |
| 1594 | else |
| 1595 | break; |
| 1596 | } |
| 1597 | else |
| 1598 | break; |
| 1599 | } |
| 1600 | |
| 1601 | yylval.tsym.type = SYMBOL_TYPE (best_sym); |
| 1602 | #else /* not 0 */ |
| 1603 | yylval.tsym.type = SYMBOL_TYPE (sym); |
| 1604 | #endif /* not 0 */ |
| 1605 | return TYPENAME; |
| 1606 | } |
| 1607 | if ((yylval.tsym.type = lookup_primitive_typename (tmp)) != 0) |
| 1608 | return TYPENAME; |
| 1609 | |
| 1610 | /* Input names that aren't symbols but ARE valid hex numbers, |
| 1611 | when the input radix permits them, can be names or numbers |
| 1612 | depending on the parse. Note we support radixes > 16 here. */ |
| 1613 | if (!sym && |
| 1614 | ((tokstart[0] >= 'a' && tokstart[0] < 'a' + input_radix - 10) || |
| 1615 | (tokstart[0] >= 'A' && tokstart[0] < 'A' + input_radix - 10))) |
| 1616 | { |
| 1617 | YYSTYPE newlval; /* Its value is ignored. */ |
| 1618 | hextype = parse_number (tokstart, namelen, 0, &newlval); |
| 1619 | if (hextype == INT) |
| 1620 | { |
| 1621 | yylval.ssym.sym = sym; |
| 1622 | yylval.ssym.is_a_field_of_this = is_a_field_of_this; |
| 1623 | return NAME_OR_INT; |
| 1624 | } |
| 1625 | } |
| 1626 | |
| 1627 | /* Any other kind of symbol */ |
| 1628 | yylval.ssym.sym = sym; |
| 1629 | yylval.ssym.is_a_field_of_this = is_a_field_of_this; |
| 1630 | return NAME; |
| 1631 | } |
| 1632 | } |
| 1633 | |
| 1634 | void |
| 1635 | yyerror (msg) |
| 1636 | char *msg; |
| 1637 | { |
| 1638 | error (msg ? msg : "Invalid syntax in expression."); |
| 1639 | } |