parser.c

/* ----------------------------------------------------------------------- *
 *
 *   Copyright 1996-2013 The NASM Authors - All Rights Reserved
 *   See the file AUTHORS included with the NASM distribution for
 *   the specific copyright holders.
 *
 *   Redistribution and use in source and binary forms, with or without
 *   modification, are permitted provided that the following
 *   conditions are met:
 *
 *   * Redistributions of source code must retain the above copyright
 *     notice, this list of conditions and the following disclaimer.
 *   * Redistributions in binary form must reproduce the above
 *     copyright notice, this list of conditions and the following
 *     disclaimer in the documentation and/or other materials provided
 *     with the distribution.
 *
 *     THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
 *     CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
 *     INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 *     MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 *     DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
 *     CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 *     SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 *     NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 *     LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 *     HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 *     CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
 *     OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
 *     EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * ----------------------------------------------------------------------- */

/*
 * parser.c   source line parser for the Netwide Assembler
 */

#include "compiler.h"

#include <stdio.h>
#include <stdlib.h>
#include <stddef.h>
#include <string.h>
#include <ctype.h>
#include <inttypes.h>

#include "nasm.h"
#include "insns.h"
#include "nasmlib.h"
#include "stdscan.h"
#include "eval.h"
#include "parser.h"
#include "float.h"
#include "tables.h"

extern int in_abs_seg;          /* ABSOLUTE segment flag */
extern int32_t abs_seg;         /* ABSOLUTE segment */
extern int32_t abs_offset;      /* ABSOLUTE segment offset */

static int is_comma_next(void);

static int i;
static struct tokenval tokval;
static struct location *location;       /* Pointer to current line's segment,offset */

void parser_global_info(struct location * locp)
{
    location = locp;
}

static int prefix_slot(int prefix)
{
    switch (prefix) {
    case P_WAIT:
        return PPS_WAIT;
    case R_CS:
    case R_DS:
    case R_SS:
    case R_ES:
    case R_FS:
    case R_GS:
        return PPS_SEG;
    case P_LOCK:
        return PPS_LOCK;
    case P_REP:
    case P_REPE:
    case P_REPZ:
    case P_REPNE:
    case P_REPNZ:
    case P_XACQUIRE:
    case P_XRELEASE:
        return PPS_REP;
    case P_O16:
    case P_O32:
    case P_O64:
    case P_OSP:
        return PPS_OSIZE;
    case P_A16:
    case P_A32:
    case P_A64:
    case P_ASP:
        return PPS_ASIZE;
    default:
        nasm_error(ERR_PANIC, "Invalid value %d passed to prefix_slot()", prefix);
        return -1;
    }
}

static void process_size_override(insn *result, int operand)
{
    if (tasm_compatible_mode) {
        switch ((int)tokval.t_integer) {
            /* For TASM compatibility a size override inside the
             * brackets changes the size of the operand, not the
             * address type of the operand as it does in standard
             * NASM syntax. Hence:
             *
             *  mov     eax,[DWORD val]
             *
             * is valid syntax in TASM compatibility mode. Note that
             * you lose the ability to override the default address
             * type for the instruction, but we never use anything
             * but 32-bit flat model addressing in our code.
             */
        case S_BYTE:
            result->oprs[operand].type |= BITS8;
            break;
        case S_WORD:
            result->oprs[operand].type |= BITS16;
            break;
        case S_DWORD:
        case S_LONG:
            result->oprs[operand].type |= BITS32;
            break;
        case S_QWORD:
            result->oprs[operand].type |= BITS64;
            break;
        case S_TWORD:
            result->oprs[operand].type |= BITS80;
            break;
        case S_OWORD:
            result->oprs[operand].type |= BITS128;
            break;
        default:
            nasm_error(ERR_NONFATAL,
                       "invalid operand size specification");
            break;
        }
    } else {
        /* Standard NASM compatible syntax */
        switch ((int)tokval.t_integer) {
        case S_NOSPLIT:
            result->oprs[operand].eaflags |= EAF_TIMESTWO;
            break;
        case S_REL:
            result->oprs[operand].eaflags |= EAF_REL;
            break;
        case S_ABS:
            result->oprs[operand].eaflags |= EAF_ABS;
            break;
        case S_BYTE:
            result->oprs[operand].disp_size = 8;
            result->oprs[operand].eaflags |= EAF_BYTEOFFS;
            break;
        case P_A16:
        case P_A32:
        case P_A64:
            if (result->prefixes[PPS_ASIZE] &&
                result->prefixes[PPS_ASIZE] != tokval.t_integer)
                nasm_error(ERR_NONFATAL,
                           "conflicting address size specifications");
            else
                result->prefixes[PPS_ASIZE] = tokval.t_integer;
            break;
        case S_WORD:
            result->oprs[operand].disp_size = 16;
            result->oprs[operand].eaflags |= EAF_WORDOFFS;
            break;
        case S_DWORD:
        case S_LONG:
            result->oprs[operand].disp_size = 32;
            result->oprs[operand].eaflags |= EAF_WORDOFFS;
            break;
        case S_QWORD:
            result->oprs[operand].disp_size = 64;
            result->oprs[operand].eaflags |= EAF_WORDOFFS;
            break;
        default:
            nasm_error(ERR_NONFATAL, "invalid size specification in"
                       " effective address");
            break;
        }
    }
}

/*
 * when two or more decorators follow a register operand,
 * consecutive decorators are parsed here.
 * opmask and zeroing decorators can be placed in any order.
 * e.g. zmm1 {k2}{z} or zmm2 {z,k3}
 * decorator(s) are placed at the end of an operand.
 */
static bool parse_braces(decoflags_t *decoflags)
{
    int i;
    bool recover = false;

    i = tokval.t_type;
    do {
        if (i == TOKEN_OPMASK) {
            if (*decoflags & OPMASK_MASK) {
                nasm_error(ERR_NONFATAL, "opmask k%lu is already set",
                           *decoflags & OPMASK_MASK);
                *decoflags &= ~OPMASK_MASK;
            }
            *decoflags |= VAL_OPMASK(nasm_regvals[tokval.t_integer]);
        } else if (i == TOKEN_DECORATOR) {
            switch (tokval.t_integer) {
            case BRC_Z:
                /*
                 * according to AVX512 spec, only zeroing/merging decorator
                 * is supported with opmask
                 */
                *decoflags |= GEN_Z(0);
                break;
            default:
                nasm_error(ERR_NONFATAL, "{%s} is not an expected decorator",
                                         tokval.t_charptr);
                break;
            }
        } else if (i == ',' || i == TOKEN_EOS){
            break;
        } else {
            nasm_error(ERR_NONFATAL, "only a series of valid decorators"
                                     " expected");
            recover = true;
            break;
        }
        i = stdscan(NULL, &tokval);
    } while(1);

    return recover;
}

insn *parse_line(int pass, char *buffer, insn *result, ldfunc ldef)
{
    bool insn_is_label = false;
    struct eval_hints hints;
    int operand;
    int critical;
    bool first;
    bool recover;

restart_parse:
    first               = true;
    result->forw_ref    = false;

    stdscan_reset();
    stdscan_set(buffer);
    i = stdscan(NULL, &tokval);

    result->label       = NULL; /* Assume no label */
    result->eops        = NULL; /* must do this, whatever happens */
    result->operands    = 0;    /* must initialize this */
    result->evex_rm     = 0;    /* Ensure EVEX rounding mode is reset */
    result->evex_brerop = -1;   /* Reset EVEX broadcasting/ER op position */

    /* Ignore blank lines */
    if (i == TOKEN_EOS) {
        result->opcode = I_none;
        return result;
    }

    if (i != TOKEN_ID       &&
        i != TOKEN_INSN     &&
        i != TOKEN_PREFIX   &&
        (i != TOKEN_REG || !IS_SREG(tokval.t_integer))) {
        nasm_error(ERR_NONFATAL,
                   "label or instruction expected at start of line");
        result->opcode = I_none;
        return result;
    }

    if (i == TOKEN_ID || (insn_is_label && i == TOKEN_INSN)) {
        /* there's a label here */
        first = false;
        result->label = tokval.t_charptr;
        i = stdscan(NULL, &tokval);
        if (i == ':') {         /* skip over the optional colon */
            i = stdscan(NULL, &tokval);
        } else if (i == 0) {
            nasm_error(ERR_WARNING | ERR_WARN_OL | ERR_PASS1,
                  "label alone on a line without a colon might be in error");
        }
        if (i != TOKEN_INSN || tokval.t_integer != I_EQU) {
            /*
             * FIXME: location->segment could be NO_SEG, in which case
             * it is possible we should be passing 'abs_seg'. Look into this.
             * Work out whether that is *really* what we should be doing.
             * Generally fix things. I think this is right as it is, but
             * am still not certain.
             */
            ldef(result->label, in_abs_seg ? abs_seg : location->segment,
                 location->offset, NULL, true, false);
        }
    }

    /* Just a label here */
    if (i == TOKEN_EOS) {
        result->opcode = I_none;
        return result;
    }

    nasm_build_assert(P_none != 0);
    memset(result->prefixes, P_none, sizeof(result->prefixes));
    result->times = 1L;

    while (i == TOKEN_PREFIX ||
           (i == TOKEN_REG && IS_SREG(tokval.t_integer))) {
        first = false;

        /*
         * Handle special case: the TIMES prefix.
         */
        if (i == TOKEN_PREFIX && tokval.t_integer == P_TIMES) {
            expr *value;

            i = stdscan(NULL, &tokval);
            value = evaluate(stdscan, NULL, &tokval, NULL, pass0, nasm_error, NULL);
            i = tokval.t_type;
            if (!value) {       /* but, error in evaluator */
                result->opcode = I_none;    /* unrecoverable parse error: */
                return result;  /* ignore this instruction */
            }
            if (!is_simple(value)) {
                nasm_error(ERR_NONFATAL,
                      "non-constant argument supplied to TIMES");
                result->times = 1L;
            } else {
                result->times = value->value;
                if (value->value < 0 && pass0 == 2) {
                    nasm_error(ERR_NONFATAL, "TIMES value %"PRId64" is negative",
                          value->value);
                    result->times = 0;
                }
            }
        } else {
            int slot = prefix_slot(tokval.t_integer);
            if (result->prefixes[slot]) {
               if (result->prefixes[slot] == tokval.t_integer)
                    nasm_error(ERR_WARNING | ERR_PASS1,
                               "instruction has redundant prefixes");
               else
                    nasm_error(ERR_NONFATAL,
                               "instruction has conflicting prefixes");
            }
            result->prefixes[slot] = tokval.t_integer;
            i = stdscan(NULL, &tokval);
        }
    }

    if (i != TOKEN_INSN) {
        int j;
        enum prefixes pfx;

        for (j = 0; j < MAXPREFIX; j++) {
            if ((pfx = result->prefixes[j]) != P_none)
                break;
        }

        if (i == 0 && pfx != P_none) {
            /*
             * Instruction prefixes are present, but no actual
             * instruction. This is allowed: at this point we
             * invent a notional instruction of RESB 0.
             */
            result->opcode          = I_RESB;
            result->operands        = 1;
            result->oprs[0].type    = IMMEDIATE;
            result->oprs[0].offset  = 0L;
            result->oprs[0].segment = result->oprs[0].wrt = NO_SEG;
            return result;
        } else {
            nasm_error(ERR_NONFATAL, "parser: instruction expected");
            result->opcode = I_none;
            return result;
        }
    }

    result->opcode = tokval.t_integer;
    result->condition = tokval.t_inttwo;

    /*
     * INCBIN cannot be satisfied with incorrectly
     * evaluated operands, since the correct values _must_ be known
     * on the first pass. Hence, even in pass one, we set the
     * `critical' flag on calling evaluate(), so that it will bomb
     * out on undefined symbols.
     */
    if (result->opcode == I_INCBIN) {
        critical = (pass0 < 2 ? 1 : 2);

    } else
        critical = (pass == 2 ? 2 : 0);

    if (result->opcode == I_DB || result->opcode == I_DW ||
        result->opcode == I_DD || result->opcode == I_DQ ||
        result->opcode == I_DT || result->opcode == I_DO ||
        result->opcode == I_DY || result->opcode == I_INCBIN) {
        extop *eop, **tail = &result->eops, **fixptr;
        int oper_num = 0;
        int32_t sign;

        result->eops_float = false;

        /*
         * Begin to read the DB/DW/DD/DQ/DT/DO/INCBIN operands.
         */
        while (1) {
            i = stdscan(NULL, &tokval);
            if (i == TOKEN_EOS)
                break;
            else if (first && i == ':') {
                insn_is_label = true;
                goto restart_parse;
            }
            first = false;
            fixptr = tail;
            eop = *tail = nasm_malloc(sizeof(extop));
            tail = &eop->next;
            eop->next = NULL;
            eop->type = EOT_NOTHING;
            oper_num++;
            sign = +1;

            /*
             * is_comma_next() here is to distinguish this from
             * a string used as part of an expression...
             */
            if (i == TOKEN_STR && is_comma_next()) {
                eop->type       = EOT_DB_STRING;
                eop->stringval  = tokval.t_charptr;
                eop->stringlen  = tokval.t_inttwo;
                i = stdscan(NULL, &tokval);     /* eat the comma */
            } else if (i == TOKEN_STRFUNC) {
                bool parens = false;
                const char *funcname = tokval.t_charptr;
                enum strfunc func = tokval.t_integer;
                i = stdscan(NULL, &tokval);
                if (i == '(') {
                    parens = true;
                    i = stdscan(NULL, &tokval);
                }
                if (i != TOKEN_STR) {
                    nasm_error(ERR_NONFATAL,
                               "%s must be followed by a string constant",
                               funcname);
                        eop->type = EOT_NOTHING;
                } else {
                    eop->type = EOT_DB_STRING_FREE;
                    eop->stringlen =
                        string_transform(tokval.t_charptr, tokval.t_inttwo,
                                         &eop->stringval, func);
                    if (eop->stringlen == (size_t)-1) {
                        nasm_error(ERR_NONFATAL, "invalid string for transform");
                        eop->type = EOT_NOTHING;
                    }
                }
                if (parens && i && i != ')') {
                    i = stdscan(NULL, &tokval);
                    if (i != ')') {
                        nasm_error(ERR_NONFATAL, "unterminated %s function",
                                   funcname);
                    }
                }
                if (i && i != ',')
                    i = stdscan(NULL, &tokval);
            } else if (i == '-' || i == '+') {
                char *save = stdscan_get();
                int token = i;
                sign = (i == '-') ? -1 : 1;
                i = stdscan(NULL, &tokval);
                if (i != TOKEN_FLOAT) {
                    stdscan_set(save);
                    i = tokval.t_type = token;
                    goto is_expression;
                } else {
                    goto is_float;
                }
            } else if (i == TOKEN_FLOAT) {
is_float:
                eop->type = EOT_DB_STRING;
                result->eops_float = true;

                eop->stringlen = idata_bytes(result->opcode);
                if (eop->stringlen > 16) {
                    nasm_error(ERR_NONFATAL, "floating-point constant"
                               " encountered in DY instruction");
                    eop->stringlen = 0;
                } else if (eop->stringlen < 1) {
                    nasm_error(ERR_NONFATAL, "floating-point constant"
                               " encountered in unknown instruction");
                    /*
                     * fix suggested by Pedro Gimeno... original line was:
                     * eop->type = EOT_NOTHING;
                     */
                    eop->stringlen = 0;
                }

                eop = nasm_realloc(eop, sizeof(extop) + eop->stringlen);
                tail = &eop->next;
                *fixptr = eop;
                eop->stringval = (char *)eop + sizeof(extop);
                if (!eop->stringlen ||
                    !float_const(tokval.t_charptr, sign,
                                 (uint8_t *)eop->stringval,
                                 eop->stringlen, nasm_error))
                    eop->type = EOT_NOTHING;
                i = stdscan(NULL, &tokval); /* eat the comma */
            } else {
                /* anything else, assume it is an expression */
                expr *value;

is_expression:
                value = evaluate(stdscan, NULL, &tokval, NULL,
                                 critical, nasm_error, NULL);
                i = tokval.t_type;
                if (!value) {   /* error in evaluator */
                    result->opcode = I_none;        /* unrecoverable parse error: */
                    return result;      /* ignore this instruction */
                }
                if (is_unknown(value)) {
                    eop->type = EOT_DB_NUMBER;
                    eop->offset = 0;    /* doesn't matter what we put */
                    eop->segment = eop->wrt = NO_SEG;   /* likewise */
                } else if (is_reloc(value)) {
                    eop->type = EOT_DB_NUMBER;
                    eop->offset = reloc_value(value);
                    eop->segment = reloc_seg(value);
                    eop->wrt = reloc_wrt(value);
                } else {
                    nasm_error(ERR_NONFATAL,
                          "operand %d: expression is not simple"
                          " or relocatable", oper_num);
                }
            }

            /*
             * We're about to call stdscan(), which will eat the
             * comma that we're currently sitting on between
             * arguments. However, we'd better check first that it
             * _is_ a comma.
             */
            if (i == TOKEN_EOS) /* also could be EOL */
                break;
            if (i != ',') {
                nasm_error(ERR_NONFATAL, "comma expected after operand %d",
                           oper_num);
                result->opcode = I_none;/* unrecoverable parse error: */
                return result;          /* ignore this instruction */
            }
        }

        if (result->opcode == I_INCBIN) {
            /*
             * Correct syntax for INCBIN is that there should be
             * one string operand, followed by one or two numeric
             * operands.
             */
            if (!result->eops || result->eops->type != EOT_DB_STRING)
                nasm_error(ERR_NONFATAL, "`incbin' expects a file name");
            else if (result->eops->next &&
                     result->eops->next->type != EOT_DB_NUMBER)
                nasm_error(ERR_NONFATAL, "`incbin': second parameter is"
                           " non-numeric");
            else if (result->eops->next && result->eops->next->next &&
                     result->eops->next->next->type != EOT_DB_NUMBER)
                nasm_error(ERR_NONFATAL, "`incbin': third parameter is"
                           " non-numeric");
            else if (result->eops->next && result->eops->next->next &&
                     result->eops->next->next->next)
                nasm_error(ERR_NONFATAL,
                           "`incbin': more than three parameters");
            else
                return result;
            /*
             * If we reach here, one of the above errors happened.
             * Throw the instruction away.
             */
            result->opcode = I_none;
            return result;
        } else /* DB ... */ if (oper_num == 0)
            nasm_error(ERR_WARNING | ERR_PASS1,
                  "no operand for data declaration");
        else
            result->operands = oper_num;

        return result;
    }

    /*
     * Now we begin to parse the operands. There may be up to four
     * of these, separated by commas, and terminated by a zero token.
     */

    for (operand = 0; operand < MAX_OPERANDS; operand++) {
        expr *value;            /* used most of the time */
        int mref;               /* is this going to be a memory ref? */
        int bracket;            /* is it a [] mref, or a & mref? */
        int setsize = 0;
        decoflags_t brace_flags = 0;    /* flags for decorators in braces */

        result->oprs[operand].disp_size = 0;    /* have to zero this whatever */
        result->oprs[operand].eaflags   = 0;    /* and this */
        result->oprs[operand].opflags   = 0;
        result->oprs[operand].decoflags = 0;

        i = stdscan(NULL, &tokval);
        if (i == TOKEN_EOS)
            break;              /* end of operands: get out of here */
        else if (first && i == ':') {
            insn_is_label = true;
            goto restart_parse;
        }
        first = false;
        result->oprs[operand].type = 0; /* so far, no override */
        while (i == TOKEN_SPECIAL) {    /* size specifiers */
            switch ((int)tokval.t_integer) {
            case S_BYTE:
                if (!setsize)   /* we want to use only the first */
                    result->oprs[operand].type |= BITS8;
                setsize = 1;
                break;
            case S_WORD:
                if (!setsize)
                    result->oprs[operand].type |= BITS16;
                setsize = 1;
                break;
            case S_DWORD:
            case S_LONG:
                if (!setsize)
                    result->oprs[operand].type |= BITS32;
                setsize = 1;
                break;
            case S_QWORD:
                if (!setsize)
                    result->oprs[operand].type |= BITS64;
                setsize = 1;
                break;
            case S_TWORD:
                if (!setsize)
                    result->oprs[operand].type |= BITS80;
                setsize = 1;
                break;
            case S_OWORD:
                if (!setsize)
                    result->oprs[operand].type |= BITS128;
                setsize = 1;
                break;
            case S_YWORD:
                if (!setsize)
                    result->oprs[operand].type |= BITS256;
                setsize = 1;
                break;
            case S_ZWORD:
                if (!setsize)
                    result->oprs[operand].type |= BITS512;
                setsize = 1;
                break;
            case S_TO:
                result->oprs[operand].type |= TO;
                break;
            case S_STRICT:
                result->oprs[operand].type |= STRICT;
                break;
            case S_FAR:
                result->oprs[operand].type |= FAR;
                break;
            case S_NEAR:
                result->oprs[operand].type |= NEAR;
                break;
            case S_SHORT:
                result->oprs[operand].type |= SHORT;
                break;
            default:
                nasm_error(ERR_NONFATAL, "invalid operand size specification");
            }
            i = stdscan(NULL, &tokval);
        }

        if (i == '[' || i == '&') {     /* memory reference */
            mref = true;
            bracket = (i == '[');
            i = stdscan(NULL, &tokval); /* then skip the colon */
            while (i == TOKEN_SPECIAL || i == TOKEN_PREFIX) {
                process_size_override(result, operand);
                i = stdscan(NULL, &tokval);
            }
        } else {                /* immediate operand, or register */
            mref = false;
            bracket = false;    /* placate optimisers */
        }

        if ((result->oprs[operand].type & FAR) && !mref &&
            result->opcode != I_JMP && result->opcode != I_CALL) {
            nasm_error(ERR_NONFATAL, "invalid use of FAR operand specifier");
        }

        value = evaluate(stdscan, NULL, &tokval,
                         &result->oprs[operand].opflags,
                         critical, nasm_error, &hints);
        i = tokval.t_type;
        if (result->oprs[operand].opflags & OPFLAG_FORWARD) {
            result->forw_ref = true;
        }
        if (!value) {           /* nasm_error in evaluator */
            result->opcode = I_none;        /* unrecoverable parse error: */
            return result;      /* ignore this instruction */
        }
        if (i == ':' && mref) { /* it was seg:offset */
            /*
             * Process the segment override.
             */
            if (value[1].type   != 0    ||
                value->value    != 1    ||
                !IS_SREG(value->type))
                nasm_error(ERR_NONFATAL, "invalid segment override");
            else if (result->prefixes[PPS_SEG])
                nasm_error(ERR_NONFATAL,
                      "instruction has conflicting segment overrides");
            else {
                result->prefixes[PPS_SEG] = value->type;
                if (IS_FSGS(value->type))
                    result->oprs[operand].eaflags |= EAF_FSGS;
            }

            i = stdscan(NULL, &tokval); /* then skip the colon */
            while (i == TOKEN_SPECIAL || i == TOKEN_PREFIX) {
                process_size_override(result, operand);
                i = stdscan(NULL, &tokval);
            }
            value = evaluate(stdscan, NULL, &tokval,
                             &result->oprs[operand].opflags,
                             critical, nasm_error, &hints);
            i = tokval.t_type;
            if (result->oprs[operand].opflags & OPFLAG_FORWARD) {
                result->forw_ref = true;
            }
            /* and get the offset */
            if (!value) {       /* but, error in evaluator */
                result->opcode = I_none;    /* unrecoverable parse error: */
                return result;  /* ignore this instruction */
            }
        }

        recover = false;
        if (mref && bracket) {  /* find ] at the end */
            if (i != ']') {
                nasm_error(ERR_NONFATAL, "parser: expecting ]");
                recover = true;
            } else {            /* we got the required ] */
                i = stdscan(NULL, &tokval);
                if ((i == TOKEN_DECORATOR) || (i == TOKEN_OPMASK)) {
                    /*
                     * according to AVX512 spec, broacast or opmask decorator
                     * is expected for memory reference operands
                     */
                    if (tokval.t_flag & TFLAG_BRDCAST) {
                        brace_flags |= GEN_BRDCAST(0);
                        i = stdscan(NULL, &tokval);
                    } else if (i == TOKEN_OPMASK) {
                        brace_flags |= VAL_OPMASK(nasm_regvals[tokval.t_integer]);
                        i = stdscan(NULL, &tokval);
                    } else {
                        nasm_error(ERR_NONFATAL, "broadcast or opmask "
                                   "decorator expected inside braces");
                        recover = true;
                    }
                }

                if (i != 0 && i != ',') {
                    nasm_error(ERR_NONFATAL, "comma or end of line expected");
                    recover = true;
                }
            }
        } else {                /* immediate operand */
            if (i != 0 && i != ',' && i != ':' &&
                i != TOKEN_DECORATOR && i != TOKEN_OPMASK) {
                nasm_error(ERR_NONFATAL, "comma, colon, decorator or end of "
                                         "line expected after operand");
                recover = true;
            } else if (i == ':') {
                result->oprs[operand].type |= COLON;
            } else if (i == TOKEN_DECORATOR || i == TOKEN_OPMASK) {
                /* parse opmask (and zeroing) after an operand */
                recover = parse_braces(&brace_flags);
            }
        }
        if (recover) {
            do {                /* error recovery */
                i = stdscan(NULL, &tokval);
            } while (i != 0 && i != ',');
        }

        /*
         * now convert the exprs returned from evaluate()
         * into operand descriptions...
         */

        if (mref) {             /* it's a memory reference */
            expr *e = value;
            int b, i, s;        /* basereg, indexreg, scale */
            int64_t o;          /* offset */

            b = i = -1, o = s = 0;
            result->oprs[operand].hintbase = hints.base;
            result->oprs[operand].hinttype = hints.type;

            if (e->type && e->type <= EXPR_REG_END) {   /* this bit's a register */
                bool is_gpr = is_class(REG_GPR,nasm_reg_flags[e->type]);
                
                if (is_gpr && e->value == 1)
                    b = e->type;	/* It can be basereg */
                else            	/* No, it has to be indexreg */
                    i = e->type, s = e->value;
                e++;
            }
            if (e->type && e->type <= EXPR_REG_END) {   /* it's a 2nd register */
                bool is_gpr = is_class(REG_GPR,nasm_reg_flags[e->type]);

                if (b != -1)    /* If the first was the base, ... */
                    i = e->type, s = e->value;  /* second has to be indexreg */

                else if (!is_gpr || e->value != 1) {
                    /* If both want to be index */
                    nasm_error(ERR_NONFATAL,
                               "invalid effective address: two index registers");
                    result->opcode = I_none;
                    return result;
                } else
                    b = e->type;
                e++;
            }
            if (e->type != 0) { /* is there an offset? */
                if (e->type <= EXPR_REG_END) {  /* in fact, is there an error? */
                    nasm_error(ERR_NONFATAL,
                          "beroset-p-603-invalid effective address");
                    result->opcode = I_none;
                    return result;
                } else {
                    if (e->type == EXPR_UNKNOWN) {
                        result->oprs[operand].opflags |= OPFLAG_UNKNOWN;
                        o = 0;  /* doesn't matter what */
                        result->oprs[operand].wrt = NO_SEG;     /* nor this */
                        result->oprs[operand].segment = NO_SEG; /* or this */
                        while (e->type)
                            e++;        /* go to the end of the line */
                    } else {
                        if (e->type == EXPR_SIMPLE) {
                            o = e->value;
                            e++;
                        }
                        if (e->type == EXPR_WRT) {
                            result->oprs[operand].wrt = e->value;
                            e++;
                        } else
                            result->oprs[operand].wrt = NO_SEG;
                        /*
                         * Look for a segment base type.
                         */
                        if (e->type && e->type < EXPR_SEGBASE) {
                            nasm_error(ERR_NONFATAL,
                                  "beroset-p-630-invalid effective address");
                            result->opcode = I_none;
                            return result;
                        }
                        while (e->type && e->value == 0)
                            e++;
                        if (e->type && e->value != 1) {
                            nasm_error(ERR_NONFATAL,
                                  "beroset-p-637-invalid effective address");
                            result->opcode = I_none;
                            return result;
                        }
                        if (e->type) {
                            result->oprs[operand].segment =
                                e->type - EXPR_SEGBASE;
                            e++;
                        } else
                            result->oprs[operand].segment = NO_SEG;
                        while (e->type && e->value == 0)
                            e++;
                        if (e->type) {
                            nasm_error(ERR_NONFATAL,
                                  "beroset-p-650-invalid effective address");
                            result->opcode = I_none;
                            return result;
                        }
                    }
                }
            } else {
                o = 0;
                result->oprs[operand].wrt = NO_SEG;
                result->oprs[operand].segment = NO_SEG;
            }

            if (e->type != 0) { /* there'd better be nothing left! */
                nasm_error(ERR_NONFATAL,
                      "beroset-p-663-invalid effective address");
                result->opcode = I_none;
                return result;
            }

            /* It is memory, but it can match any r/m operand */
            result->oprs[operand].type |= MEMORY_ANY;

            if (b == -1 && (i == -1 || s == 0)) {
                int is_rel = globalbits == 64 &&
                    !(result->oprs[operand].eaflags & EAF_ABS) &&
                    ((globalrel &&
                      !(result->oprs[operand].eaflags & EAF_FSGS)) ||
                     (result->oprs[operand].eaflags & EAF_REL));

                result->oprs[operand].type |= is_rel ? IP_REL : MEM_OFFS;
            }

            if (i != -1) {
                opflags_t iclass = nasm_reg_flags[i];

                if (is_class(XMMREG,iclass))
                    result->oprs[operand].type |= XMEM;
                else if (is_class(YMMREG,iclass))
                    result->oprs[operand].type |= YMEM;
                else if (is_class(ZMMREG,iclass))
                    result->oprs[operand].type |= ZMEM;
            }

            result->oprs[operand].basereg = b;
            result->oprs[operand].indexreg = i;
            result->oprs[operand].scale = s;
            result->oprs[operand].offset = o;
            result->oprs[operand].decoflags |= brace_flags;
        } else {                /* it's not a memory reference */
            if (is_just_unknown(value)) {       /* it's immediate but unknown */
                result->oprs[operand].type      |= IMMEDIATE;
                result->oprs[operand].opflags   |= OPFLAG_UNKNOWN;
                result->oprs[operand].offset    = 0;        /* don't care */
                result->oprs[operand].segment   = NO_SEG;   /* don't care again */
                result->oprs[operand].wrt       = NO_SEG;   /* still don't care */

                if(optimizing >= 0 && !(result->oprs[operand].type & STRICT)) {
                    /* Be optimistic */
                    result->oprs[operand].type |=
                        UNITY | SBYTEWORD | SBYTEDWORD | UDWORD | SDWORD;
                }
            } else if (is_reloc(value)) {       /* it's immediate */
                result->oprs[operand].type      |= IMMEDIATE;
                result->oprs[operand].offset    = reloc_value(value);
                result->oprs[operand].segment   = reloc_seg(value);
                result->oprs[operand].wrt       = reloc_wrt(value);

                if (is_simple(value)) {
                    uint64_t n = reloc_value(value);
                    if (n == 1)
                        result->oprs[operand].type |= UNITY;
                    if (optimizing >= 0 &&
                        !(result->oprs[operand].type & STRICT)) {
                        if ((uint32_t) (n + 128) <= 255)
                            result->oprs[operand].type |= SBYTEDWORD;
                        if ((uint16_t) (n + 128) <= 255)
                            result->oprs[operand].type |= SBYTEWORD;
                        if (n <= 0xFFFFFFFF)
                            result->oprs[operand].type |= UDWORD;
                        if (n + 0x80000000 <= 0xFFFFFFFF)
                            result->oprs[operand].type |= SDWORD;
                    }
                }
            } else if(value->type == EXPR_RDSAE) {
                /*
                 * it's not an operand but a rounding or SAE decorator.
                 * put the decorator information in the (opflag_t) type field
                 * of previous operand.
                 */
                operand --;
                switch (value->value) {
                case BRC_RN:
                case BRC_RU:
                case BRC_RD:
                case BRC_RZ:
                case BRC_SAE:
                    result->oprs[operand].decoflags  |=
                                        (value->value == BRC_SAE ? SAE : ER);
                    result->evex_rm = value->value;
                    break;
                default:
                    nasm_error(ERR_NONFATAL, "invalid decorator");
                    break;
                }
            } else {            /* it's a register */
                opflags_t rs;

                if (value->type >= EXPR_SIMPLE || value->value != 1) {
                    nasm_error(ERR_NONFATAL, "invalid operand type");
                    result->opcode = I_none;
                    return result;
                }

                /*
                 * check that its only 1 register, not an expression...
                 */
                for (i = 1; value[i].type; i++)
                    if (value[i].value) {
                        nasm_error(ERR_NONFATAL, "invalid operand type");
                        result->opcode = I_none;
                        return result;
                    }

                /* clear overrides, except TO which applies to FPU regs */
                if (result->oprs[operand].type & ~TO) {
                    /*
                     * we want to produce a warning iff the specified size
                     * is different from the register size
                     */
                    rs = result->oprs[operand].type & SIZE_MASK;
                } else
                    rs = 0;

                result->oprs[operand].type      &= TO;
                result->oprs[operand].type      |= REGISTER;
                result->oprs[operand].type      |= nasm_reg_flags[value->type];
                result->oprs[operand].decoflags |= brace_flags;
                result->oprs[operand].basereg   = value->type;

                if (rs && (result->oprs[operand].type & SIZE_MASK) != rs)
                    nasm_error(ERR_WARNING | ERR_PASS1,
                          "register size specification ignored");
            }
        }

        /* remember the position of operand having broadcasting/ER mode */
        if (result->oprs[operand].decoflags & (BRDCAST_MASK | ER | SAE))
            result->evex_brerop = operand;
    }

    result->operands = operand; /* set operand count */

    /* clear remaining operands */
    while (operand < MAX_OPERANDS)
        result->oprs[operand++].type = 0;

    /*
     * Transform RESW, RESD, RESQ, REST, RESO, RESY into RESB.
     */
    switch (result->opcode) {
    case I_RESW:
        result->opcode = I_RESB;
        result->oprs[0].offset *= 2;
        break;
    case I_RESD:
        result->opcode = I_RESB;
        result->oprs[0].offset *= 4;
        break;
    case I_RESQ:
        result->opcode = I_RESB;
        result->oprs[0].offset *= 8;
        break;
    case I_REST:
        result->opcode = I_RESB;
        result->oprs[0].offset *= 10;
        break;
    case I_RESO:
        result->opcode = I_RESB;
        result->oprs[0].offset *= 16;
        break;
    case I_RESY:
        result->opcode = I_RESB;
        result->oprs[0].offset *= 32;
        break;
    default:
        break;
    }

    return result;
}

static int is_comma_next(void)
{
    struct tokenval tv;
    char *p;
    int i;

    p = stdscan_get();
    i = stdscan(NULL, &tv);
    stdscan_set(p);

    return (i == ',' || i == ';' || !i);
}

void cleanup_insn(insn * i)
{
    extop *e;

    while ((e = i->eops)) {
        i->eops = e->next;
        if (e->type == EOT_DB_STRING_FREE)
            nasm_free(e->stringval);
        nasm_free(e);
    }
}