dt_lex.l revision 178529
1%{ 2/* 3 * CDDL HEADER START 4 * 5 * The contents of this file are subject to the terms of the 6 * Common Development and Distribution License (the "License"). 7 * You may not use this file except in compliance with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 * 22 * Copyright 2007 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26#pragma ident "%Z%%M% %I% %E% SMI" 27 28#include <string.h> 29#include <stdlib.h> 30#include <stdio.h> 31#include <assert.h> 32#include <ctype.h> 33#include <errno.h> 34 35#include <dt_impl.h> 36#include <dt_grammar.h> 37#include <dt_parser.h> 38#include <dt_string.h> 39 40/* 41 * We need to undefine lex's input and unput macros so that references to these 42 * call the functions provided at the end of this source file. 43 */ 44#undef input 45#undef unput 46 47static int id_or_type(const char *); 48static int input(void); 49static void unput(int); 50 51/* 52 * We first define a set of labeled states for use in the D lexer and then a 53 * set of regular expressions to simplify things below. The lexer states are: 54 * 55 * S0 - D program clause and expression lexing 56 * S1 - D comments (i.e. skip everything until end of comment) 57 * S2 - D program outer scope (probe specifiers and declarations) 58 * S3 - D control line parsing (i.e. after ^# is seen but before \n) 59 * S4 - D control line scan (locate control directives only and invoke S3) 60 */ 61%} 62 63%e 1500 /* maximum nodes */ 64%p 3700 /* maximum positions */ 65%n 600 /* maximum states */ 66 67%s S0 S1 S2 S3 S4 68 69RGX_AGG "@"[a-zA-Z_][0-9a-zA-Z_]* 70RGX_PSPEC [-$:a-zA-Z_.?*\\\[\]!][-$:0-9a-zA-Z_.`?*\\\[\]!]* 71RGX_IDENT [a-zA-Z_`][0-9a-zA-Z_`]* 72RGX_INT ([0-9]+|0[xX][0-9A-Fa-f]+)[uU]?[lL]?[lL]? 73RGX_FP ([0-9]+("."?)[0-9]*|"."[0-9]+)((e|E)("+"|-)?[0-9]+)?[fFlL]? 74RGX_WS [\f\n\r\t\v ] 75RGX_STR ([^"\\\n]|\\[^"\n]|\\\")* 76RGX_CHR ([^'\\\n]|\\[^'\n]|\\')* 77RGX_INTERP ^[\f\t\v ]*#!.* 78RGX_CTL ^[\f\t\v ]*# 79 80%% 81 82%{ 83 84/* 85 * We insert a special prologue into yylex() itself: if the pcb contains a 86 * context token, we return that prior to running the normal lexer. This 87 * allows libdtrace to force yacc into one of our three parsing contexts: D 88 * expression (DT_CTX_DEXPR), D program (DT_CTX_DPROG) or D type (DT_CTX_DTYPE). 89 * Once the token is returned, we clear it so this only happens once. 90 */ 91if (yypcb->pcb_token != 0) { 92 int tok = yypcb->pcb_token; 93 yypcb->pcb_token = 0; 94 return (tok); 95} 96 97%} 98 99<S0>auto return (DT_KEY_AUTO); 100<S0>break return (DT_KEY_BREAK); 101<S0>case return (DT_KEY_CASE); 102<S0>char return (DT_KEY_CHAR); 103<S0>const return (DT_KEY_CONST); 104<S0>continue return (DT_KEY_CONTINUE); 105<S0>counter return (DT_KEY_COUNTER); 106<S0>default return (DT_KEY_DEFAULT); 107<S0>do return (DT_KEY_DO); 108<S0>double return (DT_KEY_DOUBLE); 109<S0>else return (DT_KEY_ELSE); 110<S0>enum return (DT_KEY_ENUM); 111<S0>extern return (DT_KEY_EXTERN); 112<S0>float return (DT_KEY_FLOAT); 113<S0>for return (DT_KEY_FOR); 114<S0>goto return (DT_KEY_GOTO); 115<S0>if return (DT_KEY_IF); 116<S0>import return (DT_KEY_IMPORT); 117<S0>inline return (DT_KEY_INLINE); 118<S0>int return (DT_KEY_INT); 119<S0>long return (DT_KEY_LONG); 120<S0>offsetof return (DT_TOK_OFFSETOF); 121<S0>probe return (DT_KEY_PROBE); 122<S0>provider return (DT_KEY_PROVIDER); 123<S0>register return (DT_KEY_REGISTER); 124<S0>restrict return (DT_KEY_RESTRICT); 125<S0>return return (DT_KEY_RETURN); 126<S0>self return (DT_KEY_SELF); 127<S0>short return (DT_KEY_SHORT); 128<S0>signed return (DT_KEY_SIGNED); 129<S0>sizeof return (DT_TOK_SIZEOF); 130<S0>static return (DT_KEY_STATIC); 131<S0>string return (DT_KEY_STRING); 132<S0>stringof return (DT_TOK_STRINGOF); 133<S0>struct return (DT_KEY_STRUCT); 134<S0>switch return (DT_KEY_SWITCH); 135<S0>this return (DT_KEY_THIS); 136<S0>translator return (DT_KEY_XLATOR); 137<S0>typedef return (DT_KEY_TYPEDEF); 138<S0>union return (DT_KEY_UNION); 139<S0>unsigned return (DT_KEY_UNSIGNED); 140<S0>void return (DT_KEY_VOID); 141<S0>volatile return (DT_KEY_VOLATILE); 142<S0>while return (DT_KEY_WHILE); 143<S0>xlate return (DT_TOK_XLATE); 144 145<S2>auto { yybegin(YYS_EXPR); return (DT_KEY_AUTO); } 146<S2>char { yybegin(YYS_EXPR); return (DT_KEY_CHAR); } 147<S2>const { yybegin(YYS_EXPR); return (DT_KEY_CONST); } 148<S2>counter { yybegin(YYS_DEFINE); return (DT_KEY_COUNTER); } 149<S2>double { yybegin(YYS_EXPR); return (DT_KEY_DOUBLE); } 150<S2>enum { yybegin(YYS_EXPR); return (DT_KEY_ENUM); } 151<S2>extern { yybegin(YYS_EXPR); return (DT_KEY_EXTERN); } 152<S2>float { yybegin(YYS_EXPR); return (DT_KEY_FLOAT); } 153<S2>import { yybegin(YYS_EXPR); return (DT_KEY_IMPORT); } 154<S2>inline { yybegin(YYS_DEFINE); return (DT_KEY_INLINE); } 155<S2>int { yybegin(YYS_EXPR); return (DT_KEY_INT); } 156<S2>long { yybegin(YYS_EXPR); return (DT_KEY_LONG); } 157<S2>provider { yybegin(YYS_DEFINE); return (DT_KEY_PROVIDER); } 158<S2>register { yybegin(YYS_EXPR); return (DT_KEY_REGISTER); } 159<S2>restrict { yybegin(YYS_EXPR); return (DT_KEY_RESTRICT); } 160<S2>self { yybegin(YYS_EXPR); return (DT_KEY_SELF); } 161<S2>short { yybegin(YYS_EXPR); return (DT_KEY_SHORT); } 162<S2>signed { yybegin(YYS_EXPR); return (DT_KEY_SIGNED); } 163<S2>static { yybegin(YYS_EXPR); return (DT_KEY_STATIC); } 164<S2>string { yybegin(YYS_EXPR); return (DT_KEY_STRING); } 165<S2>struct { yybegin(YYS_EXPR); return (DT_KEY_STRUCT); } 166<S2>this { yybegin(YYS_EXPR); return (DT_KEY_THIS); } 167<S2>translator { yybegin(YYS_DEFINE); return (DT_KEY_XLATOR); } 168<S2>typedef { yybegin(YYS_EXPR); return (DT_KEY_TYPEDEF); } 169<S2>union { yybegin(YYS_EXPR); return (DT_KEY_UNION); } 170<S2>unsigned { yybegin(YYS_EXPR); return (DT_KEY_UNSIGNED); } 171<S2>void { yybegin(YYS_EXPR); return (DT_KEY_VOID); } 172<S2>volatile { yybegin(YYS_EXPR); return (DT_KEY_VOLATILE); } 173 174<S0>"$$"[0-9]+ { 175 int i = atoi(yytext + 2); 176 char *v = ""; 177 178 /* 179 * A macro argument reference substitutes the text of 180 * an argument in place of the current token. When we 181 * see $$<d> we fetch the saved string from pcb_sargv 182 * (or use the default argument if the option has been 183 * set and the argument hasn't been specified) and 184 * return a token corresponding to this string. 185 */ 186 if (i < 0 || (i >= yypcb->pcb_sargc && 187 !(yypcb->pcb_cflags & DTRACE_C_DEFARG))) { 188 xyerror(D_MACRO_UNDEF, "macro argument %s is " 189 "not defined\n", yytext); 190 } 191 192 if (i < yypcb->pcb_sargc) { 193 v = yypcb->pcb_sargv[i]; /* get val from pcb */ 194 yypcb->pcb_sflagv[i] |= DT_IDFLG_REF; 195 } 196 197 if ((yylval.l_str = strdup(v)) == NULL) 198 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 199 200 (void) stresc2chr(yylval.l_str); 201 return (DT_TOK_STRING); 202 } 203 204<S0>"$"[0-9]+ { 205 int i = atoi(yytext + 1); 206 char *p, *v = "0"; 207 208 /* 209 * A macro argument reference substitutes the text of 210 * one identifier or integer pattern for another. When 211 * we see $<d> we fetch the saved string from pcb_sargv 212 * (or use the default argument if the option has been 213 * set and the argument hasn't been specified) and 214 * return a token corresponding to this string. 215 */ 216 if (i < 0 || (i >= yypcb->pcb_sargc && 217 !(yypcb->pcb_cflags & DTRACE_C_DEFARG))) { 218 xyerror(D_MACRO_UNDEF, "macro argument %s is " 219 "not defined\n", yytext); 220 } 221 222 if (i < yypcb->pcb_sargc) { 223 v = yypcb->pcb_sargv[i]; /* get val from pcb */ 224 yypcb->pcb_sflagv[i] |= DT_IDFLG_REF; 225 } 226 227 /* 228 * If the macro text is not a valid integer or ident, 229 * then we treat it as a string. The string may be 230 * optionally enclosed in quotes, which we strip. 231 */ 232 if (strbadidnum(v)) { 233 size_t len = strlen(v); 234 235 if (len != 1 && *v == '"' && v[len - 1] == '"') 236 yylval.l_str = strndup(v + 1, len - 2); 237 else 238 yylval.l_str = strndup(v, len); 239 240 if (yylval.l_str == NULL) 241 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 242 243 (void) stresc2chr(yylval.l_str); 244 return (DT_TOK_STRING); 245 } 246 247 /* 248 * If the macro text is not a string an begins with a 249 * digit or a +/- sign, process it as an integer token. 250 */ 251 if (isdigit(v[0]) || v[0] == '-' || v[0] == '+') { 252 if (isdigit(v[0])) 253 yyintprefix = 0; 254 else 255 yyintprefix = *v++; 256 257 errno = 0; 258 yylval.l_int = strtoull(v, &p, 0); 259 (void) strncpy(yyintsuffix, p, 260 sizeof (yyintsuffix)); 261 yyintdecimal = *v != '0'; 262 263 if (errno == ERANGE) { 264 xyerror(D_MACRO_OFLOW, "macro argument" 265 " %s constant %s results in integer" 266 " overflow\n", yytext, v); 267 } 268 269 return (DT_TOK_INT); 270 } 271 272 return (id_or_type(v)); 273 } 274 275<S0>"$$"{RGX_IDENT} { 276 dt_ident_t *idp = dt_idhash_lookup( 277 yypcb->pcb_hdl->dt_macros, yytext + 2); 278 279 char s[16]; /* enough for UINT_MAX + \0 */ 280 281 if (idp == NULL) { 282 xyerror(D_MACRO_UNDEF, "macro variable %s " 283 "is not defined\n", yytext); 284 } 285 286 /* 287 * For the moment, all current macro variables are of 288 * type id_t (refer to dtrace_update() for details). 289 */ 290 (void) snprintf(s, sizeof (s), "%u", idp->di_id); 291 if ((yylval.l_str = strdup(s)) == NULL) 292 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 293 294 return (DT_TOK_STRING); 295 } 296 297<S0>"$"{RGX_IDENT} { 298 dt_ident_t *idp = dt_idhash_lookup( 299 yypcb->pcb_hdl->dt_macros, yytext + 1); 300 301 if (idp == NULL) { 302 xyerror(D_MACRO_UNDEF, "macro variable %s " 303 "is not defined\n", yytext); 304 } 305 306 /* 307 * For the moment, all current macro variables are of 308 * type id_t (refer to dtrace_update() for details). 309 */ 310 yylval.l_int = (intmax_t)(int)idp->di_id; 311 yyintprefix = 0; 312 yyintsuffix[0] = '\0'; 313 yyintdecimal = 1; 314 315 return (DT_TOK_INT); 316 } 317 318<S0>{RGX_IDENT} { 319 return (id_or_type(yytext)); 320 } 321 322<S0>{RGX_AGG} { 323 if ((yylval.l_str = strdup(yytext)) == NULL) 324 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 325 return (DT_TOK_AGG); 326 } 327 328<S0>"@" { 329 if ((yylval.l_str = strdup("@_")) == NULL) 330 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 331 return (DT_TOK_AGG); 332 } 333 334<S0>{RGX_INT} | 335<S2>{RGX_INT} | 336<S3>{RGX_INT} { 337 char *p; 338 339 errno = 0; 340 yylval.l_int = strtoull(yytext, &p, 0); 341 yyintprefix = 0; 342 (void) strncpy(yyintsuffix, p, sizeof (yyintsuffix)); 343 yyintdecimal = yytext[0] != '0'; 344 345 if (errno == ERANGE) { 346 xyerror(D_INT_OFLOW, "constant %s results in " 347 "integer overflow\n", yytext); 348 } 349 350 if (*p != '\0' && strchr("uUlL", *p) == NULL) { 351 xyerror(D_INT_DIGIT, "constant %s contains " 352 "invalid digit %c\n", yytext, *p); 353 } 354 355 if ((YYSTATE) != S3) 356 return (DT_TOK_INT); 357 358 yypragma = dt_node_link(yypragma, 359 dt_node_int(yylval.l_int)); 360 } 361 362<S0>{RGX_FP} yyerror("floating-point constants are not permitted\n"); 363 364<S0>\"{RGX_STR}$ | 365<S3>\"{RGX_STR}$ xyerror(D_STR_NL, "newline encountered in string literal"); 366 367<S0>\"{RGX_STR}\" | 368<S3>\"{RGX_STR}\" { 369 /* 370 * Quoted string -- convert C escape sequences and 371 * return the string as a token. 372 */ 373 yylval.l_str = strndup(yytext + 1, yyleng - 2); 374 375 if (yylval.l_str == NULL) 376 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 377 378 (void) stresc2chr(yylval.l_str); 379 if ((YYSTATE) != S3) 380 return (DT_TOK_STRING); 381 382 yypragma = dt_node_link(yypragma, 383 dt_node_string(yylval.l_str)); 384 } 385 386<S0>'{RGX_CHR}$ xyerror(D_CHR_NL, "newline encountered in character constant"); 387 388<S0>'{RGX_CHR}' { 389 char *s, *p, *q; 390 size_t nbytes; 391 392 /* 393 * Character constant -- convert C escape sequences and 394 * return the character as an integer immediate value. 395 */ 396 if (yyleng == 2) 397 xyerror(D_CHR_NULL, "empty character constant"); 398 399 s = yytext + 1; 400 yytext[yyleng - 1] = '\0'; 401 nbytes = stresc2chr(s); 402 yylval.l_int = 0; 403 yyintprefix = 0; 404 yyintsuffix[0] = '\0'; 405 yyintdecimal = 1; 406 407 if (nbytes > sizeof (yylval.l_int)) { 408 xyerror(D_CHR_OFLOW, "character constant is " 409 "too long"); 410 } 411#ifdef _LITTLE_ENDIAN 412 p = ((char *)&yylval.l_int) + nbytes - 1; 413 for (q = s; nbytes != 0; nbytes--) 414 *p-- = *q++; 415#else 416 bcopy(s, ((char *)&yylval.l_int) + 417 sizeof (yylval.l_int) - nbytes, nbytes); 418#endif 419 return (DT_TOK_INT); 420 } 421 422<S0>"/*" | 423<S2>"/*" { 424 yypcb->pcb_cstate = (YYSTATE); 425 BEGIN(S1); 426 } 427 428<S0>{RGX_INTERP} | 429<S2>{RGX_INTERP} ; /* discard any #! lines */ 430 431<S0>{RGX_CTL} | 432<S2>{RGX_CTL} | 433<S4>{RGX_CTL} { 434 assert(yypragma == NULL); 435 yypcb->pcb_cstate = (YYSTATE); 436 BEGIN(S3); 437 } 438 439<S4>. ; /* discard */ 440<S4>"\n" ; /* discard */ 441 442<S0>"/" { 443 int c, tok; 444 445 /* 446 * The use of "/" as the predicate delimiter and as the 447 * integer division symbol requires special lookahead 448 * to avoid a shift/reduce conflict in the D grammar. 449 * We look ahead to the next non-whitespace character. 450 * If we encounter EOF, ";", "{", or "/", then this "/" 451 * closes the predicate and we return DT_TOK_EPRED. 452 * If we encounter anything else, it's DT_TOK_DIV. 453 */ 454 while ((c = input()) != 0) { 455 if (strchr("\f\n\r\t\v ", c) == NULL) 456 break; 457 } 458 459 if (c == 0 || c == ';' || c == '{' || c == '/') { 460 if (yypcb->pcb_parens != 0) { 461 yyerror("closing ) expected in " 462 "predicate before /\n"); 463 } 464 if (yypcb->pcb_brackets != 0) { 465 yyerror("closing ] expected in " 466 "predicate before /\n"); 467 } 468 tok = DT_TOK_EPRED; 469 } else 470 tok = DT_TOK_DIV; 471 472 unput(c); 473 return (tok); 474 } 475 476<S0>"(" { 477 yypcb->pcb_parens++; 478 return (DT_TOK_LPAR); 479 } 480 481<S0>")" { 482 if (--yypcb->pcb_parens < 0) 483 yyerror("extra ) in input stream\n"); 484 return (DT_TOK_RPAR); 485 } 486 487<S0>"[" { 488 yypcb->pcb_brackets++; 489 return (DT_TOK_LBRAC); 490 } 491 492<S0>"]" { 493 if (--yypcb->pcb_brackets < 0) 494 yyerror("extra ] in input stream\n"); 495 return (DT_TOK_RBRAC); 496 } 497 498<S0>"{" | 499<S2>"{" { 500 yypcb->pcb_braces++; 501 return ('{'); 502 } 503 504<S0>"}" { 505 if (--yypcb->pcb_braces < 0) 506 yyerror("extra } in input stream\n"); 507 return ('}'); 508 } 509 510<S0>"|" return (DT_TOK_BOR); 511<S0>"^" return (DT_TOK_XOR); 512<S0>"&" return (DT_TOK_BAND); 513<S0>"&&" return (DT_TOK_LAND); 514<S0>"^^" return (DT_TOK_LXOR); 515<S0>"||" return (DT_TOK_LOR); 516<S0>"==" return (DT_TOK_EQU); 517<S0>"!=" return (DT_TOK_NEQ); 518<S0>"<" return (DT_TOK_LT); 519<S0>"<=" return (DT_TOK_LE); 520<S0>">" return (DT_TOK_GT); 521<S0>">=" return (DT_TOK_GE); 522<S0>"<<" return (DT_TOK_LSH); 523<S0>">>" return (DT_TOK_RSH); 524<S0>"+" return (DT_TOK_ADD); 525<S0>"-" return (DT_TOK_SUB); 526<S0>"*" return (DT_TOK_MUL); 527<S0>"%" return (DT_TOK_MOD); 528<S0>"~" return (DT_TOK_BNEG); 529<S0>"!" return (DT_TOK_LNEG); 530<S0>"?" return (DT_TOK_QUESTION); 531<S0>":" return (DT_TOK_COLON); 532<S0>"." return (DT_TOK_DOT); 533<S0>"->" return (DT_TOK_PTR); 534<S0>"=" return (DT_TOK_ASGN); 535<S0>"+=" return (DT_TOK_ADD_EQ); 536<S0>"-=" return (DT_TOK_SUB_EQ); 537<S0>"*=" return (DT_TOK_MUL_EQ); 538<S0>"/=" return (DT_TOK_DIV_EQ); 539<S0>"%=" return (DT_TOK_MOD_EQ); 540<S0>"&=" return (DT_TOK_AND_EQ); 541<S0>"^=" return (DT_TOK_XOR_EQ); 542<S0>"|=" return (DT_TOK_OR_EQ); 543<S0>"<<=" return (DT_TOK_LSH_EQ); 544<S0>">>=" return (DT_TOK_RSH_EQ); 545<S0>"++" return (DT_TOK_ADDADD); 546<S0>"--" return (DT_TOK_SUBSUB); 547<S0>"..." return (DT_TOK_ELLIPSIS); 548<S0>"," return (DT_TOK_COMMA); 549<S0>";" return (';'); 550<S0>{RGX_WS} ; /* discard */ 551<S0>"\\"\n ; /* discard */ 552<S0>. yyerror("syntax error near \"%c\"\n", yytext[0]); 553 554<S1>"/*" yyerror("/* encountered inside a comment\n"); 555<S1>"*/" BEGIN(yypcb->pcb_cstate); 556<S1>.|\n ; /* discard */ 557 558<S2>{RGX_PSPEC} { 559 /* 560 * S2 has an ambiguity because RGX_PSPEC includes '*' 561 * as a glob character and '*' also can be DT_TOK_STAR. 562 * Since lex always matches the longest token, this 563 * rule can be matched by an input string like "int*", 564 * which could begin a global variable declaration such 565 * as "int*x;" or could begin a RGX_PSPEC with globbing 566 * such as "int* { trace(timestamp); }". If C_PSPEC is 567 * not set, we must resolve the ambiguity in favor of 568 * the type and perform lexer pushback if the fragment 569 * before '*' or entire fragment matches a type name. 570 * If C_PSPEC is set, we always return a PSPEC token. 571 * If C_PSPEC is off, the user can avoid ambiguity by 572 * including a ':' delimiter in the specifier, which 573 * they should be doing anyway to specify the provider. 574 */ 575 if (!(yypcb->pcb_cflags & DTRACE_C_PSPEC) && 576 strchr(yytext, ':') == NULL) { 577 578 char *p = strchr(yytext, '*'); 579 char *q = yytext + yyleng - 1; 580 581 if (p != NULL && p > yytext) 582 *p = '\0'; /* prune yytext */ 583 584 if (dt_type_lookup(yytext, NULL) == 0) { 585 yylval.l_str = strdup(yytext); 586 587 if (yylval.l_str == NULL) { 588 longjmp(yypcb->pcb_jmpbuf, 589 EDT_NOMEM); 590 } 591 592 if (p != NULL && p > yytext) { 593 for (*p = '*'; q >= p; q--) 594 unput(*q); 595 } 596 597 yybegin(YYS_EXPR); 598 return (DT_TOK_TNAME); 599 } 600 601 if (p != NULL && p > yytext) 602 *p = '*'; /* restore yytext */ 603 } 604 605 if ((yylval.l_str = strdup(yytext)) == NULL) 606 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 607 608 return (DT_TOK_PSPEC); 609 } 610 611<S2>"/" return (DT_TOK_DIV); 612<S2>"," return (DT_TOK_COMMA); 613 614<S2>{RGX_WS} ; /* discard */ 615<S2>. yyerror("syntax error near \"%c\"\n", yytext[0]); 616 617<S3>\n { 618 dt_pragma(yypragma); 619 yypragma = NULL; 620 BEGIN(yypcb->pcb_cstate); 621 } 622 623<S3>[\f\t\v ]+ ; /* discard */ 624 625<S3>[^\f\n\t\v "]+ { 626 dt_node_t *dnp; 627 628 if ((yylval.l_str = strdup(yytext)) == NULL) 629 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 630 631 /* 632 * We want to call dt_node_ident() here, but we can't 633 * because it will expand inlined identifiers, which we 634 * don't want to do from #pragma context in order to 635 * support pragmas that apply to the ident itself. We 636 * call dt_node_string() and then reset dn_op instead. 637 */ 638 dnp = dt_node_string(yylval.l_str); 639 dnp->dn_kind = DT_NODE_IDENT; 640 dnp->dn_op = DT_TOK_IDENT; 641 yypragma = dt_node_link(yypragma, dnp); 642 } 643 644<S3>. yyerror("syntax error near \"%c\"\n", yytext[0]); 645 646%% 647 648/* 649 * yybegin provides a wrapper for use from C code around the lex BEGIN() macro. 650 * We use two main states for lexing because probe descriptions use a syntax 651 * that is incompatible with the normal D tokens (e.g. names can contain "-"). 652 * yybegin also handles the job of switching between two lists of dt_nodes 653 * as we allocate persistent definitions, like inlines, and transient nodes 654 * that will be freed once we are done parsing the current program file. 655 */ 656void 657yybegin(yystate_t state) 658{ 659#ifdef YYDEBUG 660 yydebug = _dtrace_debug; 661#endif 662 if (yypcb->pcb_yystate == state) 663 return; /* nothing to do if we're in the state already */ 664 665 if (yypcb->pcb_yystate == YYS_DEFINE) { 666 yypcb->pcb_list = yypcb->pcb_hold; 667 yypcb->pcb_hold = NULL; 668 } 669 670 switch (state) { 671 case YYS_CLAUSE: 672 BEGIN(S2); 673 break; 674 case YYS_DEFINE: 675 assert(yypcb->pcb_hold == NULL); 676 yypcb->pcb_hold = yypcb->pcb_list; 677 yypcb->pcb_list = NULL; 678 /*FALLTHRU*/ 679 case YYS_EXPR: 680 BEGIN(S0); 681 break; 682 case YYS_DONE: 683 break; 684 case YYS_CONTROL: 685 BEGIN(S4); 686 break; 687 default: 688 xyerror(D_UNKNOWN, "internal error -- bad yystate %d\n", state); 689 } 690 691 yypcb->pcb_yystate = state; 692} 693 694void 695yyinit(dt_pcb_t *pcb) 696{ 697 yypcb = pcb; 698 yylineno = 1; 699 yypragma = NULL; 700 yysptr = yysbuf; 701} 702 703/* 704 * Given a lexeme 's' (typically yytext), set yylval and return an appropriate 705 * token to the parser indicating either an identifier or a typedef name. 706 * User-defined global variables always take precedence over types, but we do 707 * use some heuristics because D programs can look at an ever-changing set of 708 * kernel types and also can implicitly instantiate variables by assignment, 709 * unlike in C. The code here is ordered carefully as lookups are not cheap. 710 */ 711static int 712id_or_type(const char *s) 713{ 714 dtrace_hdl_t *dtp = yypcb->pcb_hdl; 715 dt_decl_t *ddp = yypcb->pcb_dstack.ds_decl; 716 int c0, c1, ttok = DT_TOK_TNAME; 717 dt_ident_t *idp; 718 719 if ((s = yylval.l_str = strdup(s)) == NULL) 720 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 721 722 /* 723 * If the lexeme is a global variable or likely identifier or *not* a 724 * type_name, then it is an identifier token. 725 */ 726 if (dt_idstack_lookup(&yypcb->pcb_globals, s) != NULL || 727 dt_idhash_lookup(yypcb->pcb_idents, s) != NULL || 728 dt_type_lookup(s, NULL) != 0) 729 return (DT_TOK_IDENT); 730 731 /* 732 * If we're in the midst of parsing a declaration and a type_specifier 733 * has already been shifted, then return DT_TOK_IDENT instead of TNAME. 734 * This semantic is necessary to permit valid ISO C code such as: 735 * 736 * typedef int foo; 737 * struct s { foo foo; }; 738 * 739 * without causing shift/reduce conflicts in the direct_declarator part 740 * of the grammar. The result is that we must check for conflicting 741 * redeclarations of the same identifier as part of dt_node_decl(). 742 */ 743 if (ddp != NULL && ddp->dd_name != NULL) 744 return (DT_TOK_IDENT); 745 746 /* 747 * If the lexeme is a type name and we are not in a program clause, 748 * then always interpret it as a type and return DT_TOK_TNAME. 749 */ 750 if ((YYSTATE) != S0) 751 return (DT_TOK_TNAME); 752 753 /* 754 * If the lexeme matches a type name but is in a program clause, then 755 * it could be a type or it could be an undefined variable. Peek at 756 * the next token to decide. If we see ++, --, [, or =, we know there 757 * might be an assignment that is trying to create a global variable, 758 * so we optimistically return DT_TOK_IDENT. There is no harm in being 759 * wrong: a type_name followed by ++, --, [, or = is a syntax error. 760 */ 761 while ((c0 = input()) != 0) { 762 if (strchr("\f\n\r\t\v ", c0) == NULL) 763 break; 764 } 765 766 switch (c0) { 767 case '+': 768 case '-': 769 if ((c1 = input()) == c0) 770 ttok = DT_TOK_IDENT; 771 unput(c1); 772 break; 773 774 case '=': 775 if ((c1 = input()) != c0) 776 ttok = DT_TOK_IDENT; 777 unput(c1); 778 break; 779 case '[': 780 ttok = DT_TOK_IDENT; 781 break; 782 } 783 784 if (ttok == DT_TOK_IDENT) { 785 idp = dt_idhash_insert(yypcb->pcb_idents, s, DT_IDENT_SCALAR, 0, 786 0, _dtrace_defattr, 0, &dt_idops_thaw, NULL, dtp->dt_gen); 787 788 if (idp == NULL) 789 longjmp(yypcb->pcb_jmpbuf, EDT_NOMEM); 790 } 791 792 unput(c0); 793 return (ttok); 794} 795 796static int 797input(void) 798{ 799 int c; 800 801 if (yysptr > yysbuf) 802 c = *--yysptr; 803 else if (yypcb->pcb_fileptr != NULL) 804 c = fgetc(yypcb->pcb_fileptr); 805 else if (yypcb->pcb_strptr < yypcb->pcb_string + yypcb->pcb_strlen) 806 c = *yypcb->pcb_strptr++; 807 else 808 c = EOF; 809 810 if (c == '\n') 811 yylineno++; 812 813 if (c != EOF) 814 return (c); 815 816 if ((YYSTATE) == S1) 817 yyerror("end-of-file encountered before matching */\n"); 818 819 if ((YYSTATE) == S3) 820 yyerror("end-of-file encountered before end of control line\n"); 821 822 if (yypcb->pcb_fileptr != NULL && ferror(yypcb->pcb_fileptr)) 823 longjmp(yypcb->pcb_jmpbuf, EDT_FIO); 824 825 return (0); /* EOF */ 826} 827 828static void 829unput(int c) 830{ 831 if (c == '\n') 832 yylineno--; 833 834 *yysptr++ = c; 835 yytchar = c; 836} 837