parser.c

   1 /* ----------------------------------------------------------------------- *
   2  *
   3  *   Copyright 1996-2009 The NASM Authors - All Rights Reserved
   4  *   See the file AUTHORS included with the NASM distribution for
   5  *   the specific copyright holders.
   6  *
   7  *   Redistribution and use in source and binary forms, with or without
   8  *   modification, are permitted provided that the following
   9  *   conditions are met:
  10  *
  11  *   * Redistributions of source code must retain the above copyright
  12  *     notice, this list of conditions and the following disclaimer.
  13  *   * Redistributions in binary form must reproduce the above
  14  *     copyright notice, this list of conditions and the following
  15  *     disclaimer in the documentation and/or other materials provided
  16  *     with the distribution.
  17  *
  18  *     THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
  19  *     CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
  20  *     INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  21  *     MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  22  *     DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
  23  *     CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  24  *     SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  25  *     NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  26  *     LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  *     HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  28  *     CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
  29  *     OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
  30  *     EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  31  *
  32  * ----------------------------------------------------------------------- */
  33
  34 /*
  35  * parser.c   source line parser for the Netwide Assembler
  36  */
  37
  38 #include "compiler.h"
  39
  40 #include <stdio.h>
  41 #include <stdlib.h>
  42 #include <stddef.h>
  43 #include <string.h>
  44 #include <ctype.h>
  45 #include <inttypes.h>
  46
  47 #include "nasm.h"
  48 #include "insns.h"
  49 #include "nasmlib.h"
  50 #include "stdscan.h"
  51 #include "eval.h"
  52 #include "parser.h"
  53 #include "float.h"
  54 #include "tables.h"
  55
  56 extern int in_abs_seg;          /* ABSOLUTE segment flag */
  57 extern int32_t abs_seg;         /* ABSOLUTE segment */
  58 extern int32_t abs_offset;      /* ABSOLUTE segment offset */
  59
  60 static int is_comma_next(void);
  61
  62 static int i;
  63 static struct tokenval tokval;
  64 static struct location *location;       /* Pointer to current line's segment,offset */
  65
  66 void parser_global_info(struct location * locp)
  67 {
  68     location = locp;
  69 }
  70
  71 static int prefix_slot(int prefix)
  72 {
  73     switch (prefix) {
  74     case P_WAIT:
  75         return PPS_WAIT;
  76     case R_CS:
  77     case R_DS:
  78     case R_SS:
  79     case R_ES:
  80     case R_FS:
  81     case R_GS:
  82         return PPS_SEG;
  83     case P_LOCK:
  84         return PPS_LOCK;
  85     case P_REP:
  86     case P_REPE:
  87     case P_REPZ:
  88     case P_REPNE:
  89     case P_REPNZ:
  90     case P_XACQUIRE:
  91     case P_XRELEASE:
  92         return PPS_REP;
  93     case P_O16:
  94     case P_O32:
  95     case P_O64:
  96     case P_OSP:
  97         return PPS_OSIZE;
  98     case P_A16:
  99     case P_A32:
 100     case P_A64:
 101     case P_ASP:
 102         return PPS_ASIZE;
 103     default:
 104         nasm_error(ERR_PANIC, "Invalid value %d passed to prefix_slot()", prefix);
 105         return -1;
 106     }
 107 }
 108
 109 static void process_size_override(insn *result, int operand)
 110 {
 111     if (tasm_compatible_mode) {
 112         switch ((int)tokval.t_integer) {
 113             /* For TASM compatibility a size override inside the
 114              * brackets changes the size of the operand, not the
 115              * address type of the operand as it does in standard
 116              * NASM syntax. Hence:
 117              *
 118              *  mov     eax,[DWORD val]
 119              *
 120              * is valid syntax in TASM compatibility mode. Note that
 121              * you lose the ability to override the default address
 122              * type for the instruction, but we never use anything
 123              * but 32-bit flat model addressing in our code.
 124              */
 125         case S_BYTE:
 126             result->oprs[operand].type |= BITS8;
 127             break;
 128         case S_WORD:
 129             result->oprs[operand].type |= BITS16;
 130             break;
 131         case S_DWORD:
 132         case S_LONG:
 133             result->oprs[operand].type |= BITS32;
 134             break;
 135         case S_QWORD:
 136             result->oprs[operand].type |= BITS64;
 137             break;
 138         case S_TWORD:
 139             result->oprs[operand].type |= BITS80;
 140             break;
 141         case S_OWORD:
 142             result->oprs[operand].type |= BITS128;
 143             break;
 144         default:
 145             nasm_error(ERR_NONFATAL,
 146                        "invalid operand size specification");
 147             break;
 148         }
 149     } else {
 150         /* Standard NASM compatible syntax */
 151         switch ((int)tokval.t_integer) {
 152         case S_NOSPLIT:
 153             result->oprs[operand].eaflags |= EAF_TIMESTWO;
 154             break;
 155         case S_REL:
 156             result->oprs[operand].eaflags |= EAF_REL;
 157             break;
 158         case S_ABS:
 159             result->oprs[operand].eaflags |= EAF_ABS;
 160             break;
 161         case S_BYTE:
 162             result->oprs[operand].disp_size = 8;
 163             result->oprs[operand].eaflags |= EAF_BYTEOFFS;
 164             break;
 165         case P_A16:
 166         case P_A32:
 167         case P_A64:
 168             if (result->prefixes[PPS_ASIZE] &&
 169                 result->prefixes[PPS_ASIZE] != tokval.t_integer)
 170                 nasm_error(ERR_NONFATAL,
 171                            "conflicting address size specifications");
 172             else
 173                 result->prefixes[PPS_ASIZE] = tokval.t_integer;
 174             break;
 175         case S_WORD:
 176             result->oprs[operand].disp_size = 16;
 177             result->oprs[operand].eaflags |= EAF_WORDOFFS;
 178             break;
 179         case S_DWORD:
 180         case S_LONG:
 181             result->oprs[operand].disp_size = 32;
 182             result->oprs[operand].eaflags |= EAF_WORDOFFS;
 183             break;
 184         case S_QWORD:
 185             result->oprs[operand].disp_size = 64;
 186             result->oprs[operand].eaflags |= EAF_WORDOFFS;
 187             break;
 188         default:
 189             nasm_error(ERR_NONFATAL, "invalid size specification in"
 190                        " effective address");
 191             break;
 192         }
 193     }
 194 }
 195
 196 insn *parse_line(int pass, char *buffer, insn *result, ldfunc ldef)
 197 {
 198     bool insn_is_label = false;
 199     struct eval_hints hints;
 200     int operand;
 201     int critical;
 202     bool first;
 203     bool recover;
 204     int j;
 205
 206 restart_parse:
 207     first               = true;
 208     result->forw_ref    = false;
 209
 210     stdscan_reset();
 211     stdscan_set(buffer);
 212     i = stdscan(NULL, &tokval);
 213
 214     result->label       = NULL; /* Assume no label */
 215     result->eops        = NULL; /* must do this, whatever happens */
 216     result->operands    = 0;    /* must initialize this */
 217
 218     /* Ignore blank lines */
 219     if (i == TOKEN_EOS) {
 220         result->opcode = I_none;
 221         return result;
 222     }
 223
 224     if (i != TOKEN_ID       &&
 225         i != TOKEN_INSN     &&
 226         i != TOKEN_PREFIX   &&
 227         (i != TOKEN_REG || !IS_SREG(tokval.t_integer))) {
 228         nasm_error(ERR_NONFATAL,
 229                    "label or instruction expected at start of line");
 230         result->opcode = I_none;
 231         return result;
 232     }
 233
 234     if (i == TOKEN_ID || (insn_is_label && i == TOKEN_INSN)) {
 235         /* there's a label here */
 236         first = false;
 237         result->label = tokval.t_charptr;
 238         i = stdscan(NULL, &tokval);
 239         if (i == ':') {         /* skip over the optional colon */
 240             i = stdscan(NULL, &tokval);
 241         } else if (i == 0) {
 242             nasm_error(ERR_WARNING | ERR_WARN_OL | ERR_PASS1,
 243                   "label alone on a line without a colon might be in error");
 244         }
 245         if (i != TOKEN_INSN || tokval.t_integer != I_EQU) {
 246             /*
 247              * FIXME: location->segment could be NO_SEG, in which case
 248              * it is possible we should be passing 'abs_seg'. Look into this.
 249              * Work out whether that is *really* what we should be doing.
 250              * Generally fix things. I think this is right as it is, but
 251              * am still not certain.
 252              */
 253             ldef(result->label, in_abs_seg ? abs_seg : location->segment,
 254                  location->offset, NULL, true, false);
 255         }
 256     }
 257
 258     /* Just a label here */
 259     if (i == TOKEN_EOS) {
 260         result->opcode = I_none;
 261         return result;
 262     }
 263
 264     for (j = 0; j < MAXPREFIX; j++)
 265         result->prefixes[j] = P_none;
 266     result->times = 1L;
 267
 268     while (i == TOKEN_PREFIX ||
 269            (i == TOKEN_REG && IS_SREG(tokval.t_integer))) {
 270         first = false;
 271
 272         /*
 273          * Handle special case: the TIMES prefix.
 274          */
 275         if (i == TOKEN_PREFIX && tokval.t_integer == P_TIMES) {
 276             expr *value;
 277
 278             i = stdscan(NULL, &tokval);
 279             value = evaluate(stdscan, NULL, &tokval, NULL, pass0, nasm_error, NULL);
 280             i = tokval.t_type;
 281             if (!value) {       /* but, error in evaluator */
 282                 result->opcode = I_none;    /* unrecoverable parse error: */
 283                 return result;  /* ignore this instruction */
 284             }
 285             if (!is_simple(value)) {
 286                 nasm_error(ERR_NONFATAL,
 287                       "non-constant argument supplied to TIMES");
 288                 result->times = 1L;
 289             } else {
 290                 result->times = value->value;
 291                 if (value->value < 0 && pass0 == 2) {
 292                     nasm_error(ERR_NONFATAL, "TIMES value %"PRId64" is negative",
 293                           value->value);
 294                     result->times = 0;
 295                 }
 296             }
 297         } else {
 298             int slot = prefix_slot(tokval.t_integer);
 299             if (result->prefixes[slot]) {
 300                if (result->prefixes[slot] == tokval.t_integer)
 301                     nasm_error(ERR_WARNING | ERR_PASS1,
 302                                "instruction has redundant prefixes");
 303                else
 304                     nasm_error(ERR_NONFATAL,
 305                                "instruction has conflicting prefixes");
 306             }
 307             result->prefixes[slot] = tokval.t_integer;
 308             i = stdscan(NULL, &tokval);
 309         }
 310     }
 311
 312     if (i != TOKEN_INSN) {
 313         int j;
 314         enum prefixes pfx;
 315
 316         for (j = 0; j < MAXPREFIX; j++) {
 317             if ((pfx = result->prefixes[j]) != P_none)
 318                 break;
 319         }
 320
 321         if (i == 0 && pfx != P_none) {
 322             /*
 323              * Instruction prefixes are present, but no actual
 324              * instruction. This is allowed: at this point we
 325              * invent a notional instruction of RESB 0.
 326              */
 327             result->opcode          = I_RESB;
 328             result->operands        = 1;
 329             result->oprs[0].type    = IMMEDIATE;
 330             result->oprs[0].offset  = 0L;
 331             result->oprs[0].segment = result->oprs[0].wrt = NO_SEG;
 332             return result;
 333         } else {
 334             nasm_error(ERR_NONFATAL, "parser: instruction expected");
 335             result->opcode = I_none;
 336             return result;
 337         }
 338     }
 339
 340     result->opcode = tokval.t_integer;
 341     result->condition = tokval.t_inttwo;
 342
 343     /*
 344      * INCBIN cannot be satisfied with incorrectly
 345      * evaluated operands, since the correct values _must_ be known
 346      * on the first pass. Hence, even in pass one, we set the
 347      * `critical' flag on calling evaluate(), so that it will bomb
 348      * out on undefined symbols.
 349      */
 350     if (result->opcode == I_INCBIN) {
 351         critical = (pass0 < 2 ? 1 : 2);
 352
 353     } else
 354         critical = (pass == 2 ? 2 : 0);
 355
 356     if (result->opcode == I_DB || result->opcode == I_DW ||
 357         result->opcode == I_DD || result->opcode == I_DQ ||
 358         result->opcode == I_DT || result->opcode == I_DO ||
 359         result->opcode == I_DY || result->opcode == I_INCBIN) {
 360         extop *eop, **tail = &result->eops, **fixptr;
 361         int oper_num = 0;
 362         int32_t sign;
 363
 364         result->eops_float = false;
 365
 366         /*
 367          * Begin to read the DB/DW/DD/DQ/DT/DO/INCBIN operands.
 368          */
 369         while (1) {
 370             i = stdscan(NULL, &tokval);
 371             if (i == TOKEN_EOS)
 372                 break;
 373             else if (first && i == ':') {
 374                 insn_is_label = true;
 375                 goto restart_parse;
 376             }
 377             first = false;
 378             fixptr = tail;
 379             eop = *tail = nasm_malloc(sizeof(extop));
 380             tail = &eop->next;
 381             eop->next = NULL;
 382             eop->type = EOT_NOTHING;
 383             oper_num++;
 384             sign = +1;
 385
 386             /*
 387              * is_comma_next() here is to distinguish this from
 388              * a string used as part of an expression...
 389              */
 390             if (i == TOKEN_STR && is_comma_next()) {
 391                 eop->type       = EOT_DB_STRING;
 392                 eop->stringval  = tokval.t_charptr;
 393                 eop->stringlen  = tokval.t_inttwo;
 394                 i = stdscan(NULL, &tokval);     /* eat the comma */
 395             } else if (i == TOKEN_STRFUNC) {
 396                 bool parens = false;
 397                 const char *funcname = tokval.t_charptr;
 398                 enum strfunc func = tokval.t_integer;
 399                 i = stdscan(NULL, &tokval);
 400                 if (i == '(') {
 401                     parens = true;
 402                     i = stdscan(NULL, &tokval);
 403                 }
 404                 if (i != TOKEN_STR) {
 405                     nasm_error(ERR_NONFATAL,
 406                                "%s must be followed by a string constant",
 407                                funcname);
 408                         eop->type = EOT_NOTHING;
 409                 } else {
 410                     eop->type = EOT_DB_STRING_FREE;
 411                     eop->stringlen =
 412                         string_transform(tokval.t_charptr, tokval.t_inttwo,
 413                                          &eop->stringval, func);
 414                     if (eop->stringlen == (size_t)-1) {
 415                         nasm_error(ERR_NONFATAL, "invalid string for transform");
 416                         eop->type = EOT_NOTHING;
 417                     }
 418                 }
 419                 if (parens && i && i != ')') {
 420                     i = stdscan(NULL, &tokval);
 421                     if (i != ')') {
 422                         nasm_error(ERR_NONFATAL, "unterminated %s function",
 423                                    funcname);
 424                     }
 425                 }
 426                 if (i && i != ',')
 427                     i = stdscan(NULL, &tokval);
 428             } else if (i == '-' || i == '+') {
 429                 char *save = stdscan_get();
 430                 int token = i;
 431                 sign = (i == '-') ? -1 : 1;
 432                 i = stdscan(NULL, &tokval);
 433                 if (i != TOKEN_FLOAT) {
 434                     stdscan_set(save);
 435                     i = tokval.t_type = token;
 436                     goto is_expression;
 437                 } else {
 438                     goto is_float;
 439                 }
 440             } else if (i == TOKEN_FLOAT) {
 441 is_float:
 442                 eop->type = EOT_DB_STRING;
 443                 result->eops_float = true;
 444
 445                 eop->stringlen = idata_bytes(result->opcode);
 446                 if (eop->stringlen > 16) {
 447                     nasm_error(ERR_NONFATAL, "floating-point constant"
 448                                " encountered in DY instruction");
 449                     eop->stringlen = 0;
 450                 } else if (eop->stringlen < 1) {
 451                     nasm_error(ERR_NONFATAL, "floating-point constant"
 452                                " encountered in unknown instruction");
 453                     /*
 454                      * fix suggested by Pedro Gimeno... original line was:
 455                      * eop->type = EOT_NOTHING;
 456                      */
 457                     eop->stringlen = 0;
 458                 }
 459
 460                 eop = nasm_realloc(eop, sizeof(extop) + eop->stringlen);
 461                 tail = &eop->next;
 462                 *fixptr = eop;
 463                 eop->stringval = (char *)eop + sizeof(extop);
 464                 if (!eop->stringlen ||
 465                     !float_const(tokval.t_charptr, sign,
 466                                  (uint8_t *)eop->stringval,
 467                                  eop->stringlen, nasm_error))
 468                     eop->type = EOT_NOTHING;
 469                 i = stdscan(NULL, &tokval); /* eat the comma */
 470             } else {
 471                 /* anything else, assume it is an expression */
 472                 expr *value;
 473
 474 is_expression:
 475                 value = evaluate(stdscan, NULL, &tokval, NULL,
 476                                  critical, nasm_error, NULL);
 477                 i = tokval.t_type;
 478                 if (!value) {   /* error in evaluator */
 479                     result->opcode = I_none;        /* unrecoverable parse error: */
 480                     return result;      /* ignore this instruction */
 481                 }
 482                 if (is_unknown(value)) {
 483                     eop->type = EOT_DB_NUMBER;
 484                     eop->offset = 0;    /* doesn't matter what we put */
 485                     eop->segment = eop->wrt = NO_SEG;   /* likewise */
 486                 } else if (is_reloc(value)) {
 487                     eop->type = EOT_DB_NUMBER;
 488                     eop->offset = reloc_value(value);
 489                     eop->segment = reloc_seg(value);
 490                     eop->wrt = reloc_wrt(value);
 491                 } else {
 492                     nasm_error(ERR_NONFATAL,
 493                           "operand %d: expression is not simple"
 494                           " or relocatable", oper_num);
 495                 }
 496             }
 497
 498             /*
 499              * We're about to call stdscan(), which will eat the
 500              * comma that we're currently sitting on between
 501              * arguments. However, we'd better check first that it
 502              * _is_ a comma.
 503              */
 504             if (i == TOKEN_EOS) /* also could be EOL */
 505                 break;
 506             if (i != ',') {
 507                 nasm_error(ERR_NONFATAL, "comma expected after operand %d",
 508                            oper_num);
 509                 result->opcode = I_none;/* unrecoverable parse error: */
 510                 return result;          /* ignore this instruction */
 511             }
 512         }
 513
 514         if (result->opcode == I_INCBIN) {
 515             /*
 516              * Correct syntax for INCBIN is that there should be
 517              * one string operand, followed by one or two numeric
 518              * operands.
 519              */
 520             if (!result->eops || result->eops->type != EOT_DB_STRING)
 521                 nasm_error(ERR_NONFATAL, "`incbin' expects a file name");
 522             else if (result->eops->next &&
 523                      result->eops->next->type != EOT_DB_NUMBER)
 524                 nasm_error(ERR_NONFATAL, "`incbin': second parameter is"
 525                            " non-numeric");
 526             else if (result->eops->next && result->eops->next->next &&
 527                      result->eops->next->next->type != EOT_DB_NUMBER)
 528                 nasm_error(ERR_NONFATAL, "`incbin': third parameter is"
 529                            " non-numeric");
 530             else if (result->eops->next && result->eops->next->next &&
 531                      result->eops->next->next->next)
 532                 nasm_error(ERR_NONFATAL,
 533                            "`incbin': more than three parameters");
 534             else
 535                 return result;
 536             /*
 537              * If we reach here, one of the above errors happened.
 538              * Throw the instruction away.
 539              */
 540             result->opcode = I_none;
 541             return result;
 542         } else /* DB ... */ if (oper_num == 0)
 543             nasm_error(ERR_WARNING | ERR_PASS1,
 544                   "no operand for data declaration");
 545         else
 546             result->operands = oper_num;
 547
 548         return result;
 549     }
 550
 551     /*
 552      * Now we begin to parse the operands. There may be up to four
 553      * of these, separated by commas, and terminated by a zero token.
 554      */
 555
 556     for (operand = 0; operand < MAX_OPERANDS; operand++) {
 557         expr *value;            /* used most of the time */
 558         int mref;               /* is this going to be a memory ref? */
 559         int bracket;            /* is it a [] mref, or a & mref? */
 560         int setsize = 0;
 561
 562         result->oprs[operand].disp_size = 0;    /* have to zero this whatever */
 563         result->oprs[operand].eaflags   = 0;    /* and this */
 564         result->oprs[operand].opflags   = 0;
 565
 566         i = stdscan(NULL, &tokval);
 567         if (i == TOKEN_EOS)
 568             break;              /* end of operands: get out of here */
 569         else if (first && i == ':') {
 570             insn_is_label = true;
 571             goto restart_parse;
 572         }
 573         first = false;
 574         result->oprs[operand].type = 0; /* so far, no override */
 575         while (i == TOKEN_SPECIAL) {    /* size specifiers */
 576             switch ((int)tokval.t_integer) {
 577             case S_BYTE:
 578                 if (!setsize)   /* we want to use only the first */
 579                     result->oprs[operand].type |= BITS8;
 580                 setsize = 1;
 581                 break;
 582             case S_WORD:
 583                 if (!setsize)
 584                     result->oprs[operand].type |= BITS16;
 585                 setsize = 1;
 586                 break;
 587             case S_DWORD:
 588             case S_LONG:
 589                 if (!setsize)
 590                     result->oprs[operand].type |= BITS32;
 591                 setsize = 1;
 592                 break;
 593             case S_QWORD:
 594                 if (!setsize)
 595                     result->oprs[operand].type |= BITS64;
 596                 setsize = 1;
 597                 break;
 598             case S_TWORD:
 599                 if (!setsize)
 600                     result->oprs[operand].type |= BITS80;
 601                 setsize = 1;
 602                 break;
 603             case S_OWORD:
 604                 if (!setsize)
 605                     result->oprs[operand].type |= BITS128;
 606                 setsize = 1;
 607                 break;
 608             case S_YWORD:
 609                 if (!setsize)
 610                     result->oprs[operand].type |= BITS256;
 611                 setsize = 1;
 612                 break;
 613             case S_TO:
 614                 result->oprs[operand].type |= TO;
 615                 break;
 616             case S_STRICT:
 617                 result->oprs[operand].type |= STRICT;
 618                 break;
 619             case S_FAR:
 620                 result->oprs[operand].type |= FAR;
 621                 break;
 622             case S_NEAR:
 623                 result->oprs[operand].type |= NEAR;
 624                 break;
 625             case S_SHORT:
 626                 result->oprs[operand].type |= SHORT;
 627                 break;
 628             default:
 629                 nasm_error(ERR_NONFATAL, "invalid operand size specification");
 630             }
 631             i = stdscan(NULL, &tokval);
 632         }
 633
 634         if (i == '[' || i == '&') {     /* memory reference */
 635             mref = true;
 636             bracket = (i == '[');
 637             i = stdscan(NULL, &tokval); /* then skip the colon */
 638             while (i == TOKEN_SPECIAL || i == TOKEN_PREFIX) {
 639                 process_size_override(result, operand);
 640                 i = stdscan(NULL, &tokval);
 641             }
 642         } else {                /* immediate operand, or register */
 643             mref = false;
 644             bracket = false;    /* placate optimisers */
 645         }
 646
 647         if ((result->oprs[operand].type & FAR) && !mref &&
 648             result->opcode != I_JMP && result->opcode != I_CALL) {
 649             nasm_error(ERR_NONFATAL, "invalid use of FAR operand specifier");
 650         }
 651
 652         value = evaluate(stdscan, NULL, &tokval,
 653                          &result->oprs[operand].opflags,
 654                          critical, nasm_error, &hints);
 655         i = tokval.t_type;
 656         if (result->oprs[operand].opflags & OPFLAG_FORWARD) {
 657             result->forw_ref = true;
 658         }
 659         if (!value) {           /* nasm_error in evaluator */
 660             result->opcode = I_none;        /* unrecoverable parse error: */
 661             return result;      /* ignore this instruction */
 662         }
 663         if (i == ':' && mref) { /* it was seg:offset */
 664             /*
 665              * Process the segment override.
 666              */
 667             if (value[1].type   != 0    ||
 668                 value->value    != 1    ||
 669                 !IS_SREG(value->type))
 670                 nasm_error(ERR_NONFATAL, "invalid segment override");
 671             else if (result->prefixes[PPS_SEG])
 672                 nasm_error(ERR_NONFATAL,
 673                       "instruction has conflicting segment overrides");
 674             else {
 675                 result->prefixes[PPS_SEG] = value->type;
 676                 if (IS_FSGS(value->type))
 677                     result->oprs[operand].eaflags |= EAF_FSGS;
 678             }
 679
 680             i = stdscan(NULL, &tokval); /* then skip the colon */
 681             while (i == TOKEN_SPECIAL || i == TOKEN_PREFIX) {
 682                 process_size_override(result, operand);
 683                 i = stdscan(NULL, &tokval);
 684             }
 685             value = evaluate(stdscan, NULL, &tokval,
 686                              &result->oprs[operand].opflags,
 687                              critical, nasm_error, &hints);
 688             i = tokval.t_type;
 689             if (result->oprs[operand].opflags & OPFLAG_FORWARD) {
 690                 result->forw_ref = true;
 691             }
 692             /* and get the offset */
 693             if (!value) {       /* but, error in evaluator */
 694                 result->opcode = I_none;    /* unrecoverable parse error: */
 695                 return result;  /* ignore this instruction */
 696             }
 697         }
 698
 699         recover = false;
 700         if (mref && bracket) {  /* find ] at the end */
 701             if (i != ']') {
 702                 nasm_error(ERR_NONFATAL, "parser: expecting ]");
 703                 recover = true;
 704             } else {            /* we got the required ] */
 705                 i = stdscan(NULL, &tokval);
 706                 if (i != 0 && i != ',') {
 707                     nasm_error(ERR_NONFATAL, "comma or end of line expected");
 708                     recover = true;
 709                 }
 710             }
 711         } else {                /* immediate operand */
 712             if (i != 0 && i != ',' && i != ':') {
 713                 nasm_error(ERR_NONFATAL, "comma, colon or end of line expected");
 714                 recover = true;
 715             } else if (i == ':') {
 716                 result->oprs[operand].type |= COLON;
 717             }
 718         }
 719         if (recover) {
 720             do {                /* error recovery */
 721                 i = stdscan(NULL, &tokval);
 722             } while (i != 0 && i != ',');
 723         }
 724
 725         /*
 726          * now convert the exprs returned from evaluate()
 727          * into operand descriptions...
 728          */
 729
 730         if (mref) {             /* it's a memory reference */
 731             expr *e = value;
 732             int b, i, s;        /* basereg, indexreg, scale */
 733             int64_t o;          /* offset */
 734
 735             b = i = -1, o = s = 0;
 736             result->oprs[operand].hintbase = hints.base;
 737             result->oprs[operand].hinttype = hints.type;
 738
 739             if (e->type && e->type <= EXPR_REG_END) {   /* this bit's a register */
 740                 if (e->value == 1)      /* in fact it can be basereg */
 741                     b = e->type;
 742                 else            /* no, it has to be indexreg */
 743                     i = e->type, s = e->value;
 744                 e++;
 745             }
 746             if (e->type && e->type <= EXPR_REG_END) {   /* it's a 2nd register */
 747                 if (b != -1)    /* If the first was the base, ... */
 748                     i = e->type, s = e->value;  /* second has to be indexreg */
 749
 750                 else if (e->value != 1) {       /* If both want to be index */
 751                     nasm_error(ERR_NONFATAL,
 752                           "beroset-p-592-invalid effective address");
 753                     result->opcode = I_none;
 754                     return result;
 755                 } else
 756                     b = e->type;
 757                 e++;
 758             }
 759             if (e->type != 0) { /* is there an offset? */
 760                 if (e->type <= EXPR_REG_END) {  /* in fact, is there an error? */
 761                     nasm_error(ERR_NONFATAL,
 762                           "beroset-p-603-invalid effective address");
 763                     result->opcode = I_none;
 764                     return result;
 765                 } else {
 766                     if (e->type == EXPR_UNKNOWN) {
 767                         result->oprs[operand].opflags |= OPFLAG_UNKNOWN;
 768                         o = 0;  /* doesn't matter what */
 769                         result->oprs[operand].wrt = NO_SEG;     /* nor this */
 770                         result->oprs[operand].segment = NO_SEG; /* or this */
 771                         while (e->type)
 772                             e++;        /* go to the end of the line */
 773                     } else {
 774                         if (e->type == EXPR_SIMPLE) {
 775                             o = e->value;
 776                             e++;
 777                         }
 778                         if (e->type == EXPR_WRT) {
 779                             result->oprs[operand].wrt = e->value;
 780                             e++;
 781                         } else
 782                             result->oprs[operand].wrt = NO_SEG;
 783                         /*
 784                          * Look for a segment base type.
 785                          */
 786                         if (e->type && e->type < EXPR_SEGBASE) {
 787                             nasm_error(ERR_NONFATAL,
 788                                   "beroset-p-630-invalid effective address");
 789                             result->opcode = I_none;
 790                             return result;
 791                         }
 792                         while (e->type && e->value == 0)
 793                             e++;
 794                         if (e->type && e->value != 1) {
 795                             nasm_error(ERR_NONFATAL,
 796                                   "beroset-p-637-invalid effective address");
 797                             result->opcode = I_none;
 798                             return result;
 799                         }
 800                         if (e->type) {
 801                             result->oprs[operand].segment =
 802                                 e->type - EXPR_SEGBASE;
 803                             e++;
 804                         } else
 805                             result->oprs[operand].segment = NO_SEG;
 806                         while (e->type && e->value == 0)
 807                             e++;
 808                         if (e->type) {
 809                             nasm_error(ERR_NONFATAL,
 810                                   "beroset-p-650-invalid effective address");
 811                             result->opcode = I_none;
 812                             return result;
 813                         }
 814                     }
 815                 }
 816             } else {
 817                 o = 0;
 818                 result->oprs[operand].wrt = NO_SEG;
 819                 result->oprs[operand].segment = NO_SEG;
 820             }
 821
 822             if (e->type != 0) { /* there'd better be nothing left! */
 823                 nasm_error(ERR_NONFATAL,
 824                       "beroset-p-663-invalid effective address");
 825                 result->opcode = I_none;
 826                 return result;
 827             }
 828
 829             /* It is memory, but it can match any r/m operand */
 830             result->oprs[operand].type |= MEMORY_ANY;
 831
 832             if (b == -1 && (i == -1 || s == 0)) {
 833                 int is_rel = globalbits == 64 &&
 834                     !(result->oprs[operand].eaflags & EAF_ABS) &&
 835                     ((globalrel &&
 836                       !(result->oprs[operand].eaflags & EAF_FSGS)) ||
 837                      (result->oprs[operand].eaflags & EAF_REL));
 838
 839                 result->oprs[operand].type |= is_rel ? IP_REL : MEM_OFFS;
 840             }
 841             result->oprs[operand].basereg = b;
 842             result->oprs[operand].indexreg = i;
 843             result->oprs[operand].scale = s;
 844             result->oprs[operand].offset = o;
 845         } else {                /* it's not a memory reference */
 846             if (is_just_unknown(value)) {       /* it's immediate but unknown */
 847                 result->oprs[operand].type      |= IMMEDIATE;
 848                 result->oprs[operand].opflags   |= OPFLAG_UNKNOWN;
 849                 result->oprs[operand].offset    = 0;        /* don't care */
 850                 result->oprs[operand].segment   = NO_SEG;   /* don't care again */
 851                 result->oprs[operand].wrt       = NO_SEG;   /* still don't care */
 852
 853                 if(optimizing >= 0 && !(result->oprs[operand].type & STRICT)) {
 854                     /* Be optimistic */
 855                     result->oprs[operand].type |=
 856                             SBYTE16 | SBYTE32 | SBYTE64 | UDWORD64 | SDWORD64;
 857                 }
 858             } else if (is_reloc(value)) {       /* it's immediate */
 859                 result->oprs[operand].type      |= IMMEDIATE;
 860                 result->oprs[operand].offset    = reloc_value(value);
 861                 result->oprs[operand].segment   = reloc_seg(value);
 862                 result->oprs[operand].wrt       = reloc_wrt(value);
 863
 864                 if (is_simple(value)) {
 865                     if (reloc_value(value) == 1)
 866                         result->oprs[operand].type |= UNITY;
 867                     if (optimizing >= 0 &&
 868                         !(result->oprs[operand].type & STRICT)) {
 869                         int64_t v64 = reloc_value(value);
 870                         int32_t v32 = (int32_t)v64;
 871                         int16_t v16 = (int16_t)v32;
 872
 873                         if (v64 >= -128 && v64 <= 127)
 874                             result->oprs[operand].type |= SBYTE64;
 875                         if (v32 >= -128 && v32 <= 127)
 876                             result->oprs[operand].type |= SBYTE32;
 877                         if (v16 >= -128 && v16 <= 127)
 878                             result->oprs[operand].type |= SBYTE16;
 879                         if ((uint64_t)v64 <= UINT64_C(0xffffffff))
 880                             result->oprs[operand].type |= UDWORD64;
 881                         if (v64 >= -INT64_C(0x80000000) &&
 882                             v64 <=  INT64_C(0x7fffffff))
 883                             result->oprs[operand].type |= SDWORD64;
 884                     }
 885                 }
 886             } else {            /* it's a register */
 887                 opflags_t rs;
 888
 889                 if (value->type >= EXPR_SIMPLE || value->value != 1) {
 890                     nasm_error(ERR_NONFATAL, "invalid operand type");
 891                     result->opcode = I_none;
 892                     return result;
 893                 }
 894
 895                 /*
 896                  * check that its only 1 register, not an expression...
 897                  */
 898                 for (i = 1; value[i].type; i++)
 899                     if (value[i].value) {
 900                         nasm_error(ERR_NONFATAL, "invalid operand type");
 901                         result->opcode = I_none;
 902                         return result;
 903                     }
 904
 905                 /* clear overrides, except TO which applies to FPU regs */
 906                 if (result->oprs[operand].type & ~TO) {
 907                     /*
 908                      * we want to produce a warning iff the specified size
 909                      * is different from the register size
 910                      */
 911                     rs = result->oprs[operand].type & SIZE_MASK;
 912                 } else
 913                     rs = 0;
 914
 915                 result->oprs[operand].type      &= TO;
 916                 result->oprs[operand].type      |= REGISTER;
 917                 result->oprs[operand].type      |= nasm_reg_flags[value->type];
 918                 result->oprs[operand].basereg   = value->type;
 919
 920                 if (rs && (result->oprs[operand].type & SIZE_MASK) != rs)
 921                     nasm_error(ERR_WARNING | ERR_PASS1,
 922                           "register size specification ignored");
 923             }
 924         }
 925     }
 926
 927     result->operands = operand; /* set operand count */
 928
 929     /* clear remaining operands */
 930     while (operand < MAX_OPERANDS)
 931         result->oprs[operand++].type = 0;
 932
 933     /*
 934      * Transform RESW, RESD, RESQ, REST, RESO, RESY into RESB.
 935      */
 936     switch (result->opcode) {
 937     case I_RESW:
 938         result->opcode = I_RESB;
 939         result->oprs[0].offset *= 2;
 940         break;
 941     case I_RESD:
 942         result->opcode = I_RESB;
 943         result->oprs[0].offset *= 4;
 944         break;
 945     case I_RESQ:
 946         result->opcode = I_RESB;
 947         result->oprs[0].offset *= 8;
 948         break;
 949     case I_REST:
 950         result->opcode = I_RESB;
 951         result->oprs[0].offset *= 10;
 952         break;
 953     case I_RESO:
 954         result->opcode = I_RESB;
 955         result->oprs[0].offset *= 16;
 956         break;
 957     case I_RESY:
 958         result->opcode = I_RESB;
 959         result->oprs[0].offset *= 32;
 960         break;
 961     default:
 962         break;
 963     }
 964
 965     return result;
 966 }
 967
 968 static int is_comma_next(void)
 969 {
 970     struct tokenval tv;
 971     char *p;
 972     int i;
 973
 974     p = stdscan_get();
 975     i = stdscan(NULL, &tv);
 976     stdscan_set(p);
 977
 978     return (i == ',' || i == ';' || !i);
 979 }
 980
 981 void cleanup_insn(insn * i)
 982 {
 983     extop *e;
 984
 985     while ((e = i->eops)) {
 986         i->eops = e->next;
 987         if (e->type == EOT_DB_STRING_FREE)
 988             nasm_free(e->stringval);
 989         nasm_free(e);
 990     }
 991 }