parser.c

开源的nasm编译器源码,研究编译器原理很有帮且

/* parser.c   source line parser for the Netwide Assembler
 *
 * The Netwide Assembler is copyright (C) 1996 Simon Tatham and
 * Julian Hall. All rights reserved. The software is
 * redistributable under the licence given in the file "Licence"
 * distributed in the NASM archive.
 *
 * initial version 27/iii/95 by Simon Tatham
 */

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

#include "nasm.h"
#include "insns.h"
#include "nasmlib.h"
#include "parser.h"
#include "float.h"

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

#include "regflags.c"		/* List of register flags */

enum {				       /* special tokens */
    S_BYTE, S_DWORD, S_FAR, S_LONG, S_NEAR, S_NOSPLIT, S_QWORD,
    S_SHORT, S_STRICT, S_TO, S_TWORD, S_WORD
};

static int is_comma_next (void);

static int i;
static struct tokenval tokval;
static efunc error;
static struct ofmt *outfmt;  /* Structure of addresses of output routines */
static loc_t *location;	     /* Pointer to current line's segment,offset */

void parser_global_info (struct ofmt *output, loc_t *locp) 
{
    outfmt = output;
    location = locp;
}

insn *parse_line (int pass, char *buffer, insn *result,
		  efunc errfunc, evalfunc evaluate, ldfunc ldef) 
{
    int operand;
    int critical;
    struct eval_hints hints;

    result->forw_ref = FALSE;
    error = errfunc;

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

    result->label = NULL;	       /* Assume no label */
    result->eops = NULL;	       /* must do this, whatever happens */
    result->operands = 0;	       /* must initialise this */

    if (i==0) {			       /* blank line - ignore */
	result->opcode = -1;	       /* and no instruction either */
	return result;
    }
    if (i != TOKEN_ID && i != TOKEN_INSN && i != TOKEN_PREFIX &&
	(i!=TOKEN_REG || (REG_SREG & ~reg_flags[tokval.t_integer]))) {
	error (ERR_NONFATAL, "label or instruction expected"
	       " at start of line");
	result->opcode = -1;
	return result;
    }

    if (i == TOKEN_ID) {	       /* there's a label here */
	result->label = tokval.t_charptr;
	i = stdscan(NULL, &tokval);
	if (i == ':') {		       /* skip over the optional colon */
	    i = stdscan(NULL, &tokval);
	} else if (i == 0) {
	    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, outfmt, errfunc);
	}
    }

    if (i==0) {
	result->opcode = -1;	       /* this line contains just a label */
	return result;
    }

    result->nprefix = 0;
    result->times = 1L;

    while (i == TOKEN_PREFIX ||
	   (i==TOKEN_REG && !(REG_SREG & ~reg_flags[tokval.t_integer]))) 
    {
	/*
	 * 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, error, NULL);
	    i = tokval.t_type;
	    if (!value) {	       /* but, error in evaluator */
		result->opcode = -1;   /* unrecoverable parse error: */
		return result;	       /* ignore this instruction */
	    }
	    if (!is_simple (value)) {
		error (ERR_NONFATAL,
		       "non-constant argument supplied to TIMES");
		result->times = 1L;
	    } else {
		result->times = value->value;
		if (value->value < 0) {
		    error(ERR_NONFATAL, "TIMES value %d is negative",
			  value->value);
		    result->times = 0;
		}
	    }
	} else {
	    if (result->nprefix == MAXPREFIX)
		error (ERR_NONFATAL,
		       "instruction has more than %d prefixes", MAXPREFIX);
	    else
		result->prefixes[result->nprefix++] = tokval.t_integer;
	    i = stdscan(NULL, &tokval);
	}
    }

    if (i != TOKEN_INSN) {
	if (result->nprefix > 0 && i == 0) {
	    /*
	     * 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 {
	    error (ERR_NONFATAL, "parser: instruction expected");
	    result->opcode = -1;
	    return result;
	}
    }

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

    /*
     * RESB, RESW and RESD 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. Nasty, but there's nothing we can
     * do about it.
     *
     * For the moment, EQU has the same difficulty, so we'll
     * include that.
     */
    if (result->opcode == I_RESB ||
	result->opcode == I_RESW ||
	result->opcode == I_RESD ||
	result->opcode == I_RESQ ||
	result->opcode == I_REST ||
	result->opcode == I_EQU ||
        result->opcode == I_INCBIN)  /* fbk */
    {
	critical = pass0;
    }
    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_INCBIN) 
    {
	extop *eop, **tail = &result->eops, **fixptr;
	int oper_num = 0;

	result->eops_float = FALSE;

	/*
	 * Begin to read the DB/DW/DD/DQ/DT/INCBIN operands.
	 */
	while (1) {
	    i = stdscan(NULL, &tokval);
	    if (i == 0)
		break;
	    fixptr = tail;
	    eop = *tail = nasm_malloc(sizeof(extop));
	    tail = &eop->next;
	    eop->next = NULL;
	    eop->type = EOT_NOTHING;
	    oper_num++;

	    if (i == TOKEN_NUM && tokval.t_charptr && 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 */
		continue;
	    }

	    if ((i == TOKEN_FLOAT && is_comma_next()) || i == '-') {
		long sign = +1L;

		if (i == '-') {
		    char *save = stdscan_bufptr;
		    i = stdscan(NULL, &tokval);
		    sign = -1L;
		    if (i != TOKEN_FLOAT || !is_comma_next()) {
			stdscan_bufptr = save;
			i = tokval.t_type = '-';
		    }
		}

		if (i == TOKEN_FLOAT) {
		    eop->type = EOT_DB_STRING;
		    result->eops_float = TRUE;
		    if (result->opcode == I_DD)
			eop->stringlen = 4;
		    else if (result->opcode == I_DQ)
			eop->stringlen = 8;
		    else if (result->opcode == I_DT)
			eop->stringlen = 10;
		    else {
			error(ERR_NONFATAL, "floating-point constant"
			      " encountered in `D%c' instruction",
			      result->opcode == I_DW ? 'W' : 'B');
			/*
			 * 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 < 4 ||
			!float_const (tokval.t_charptr, sign,
				      (unsigned char *)eop->stringval,
				      eop->stringlen, error))
			eop->type = EOT_NOTHING;
		    i = stdscan(NULL, &tokval);       /* eat the comma */
		    continue;
		}
	    }

	    /* anything else */ 
	    {
		expr *value;
		value = evaluate (stdscan, NULL, &tokval, NULL,
				  critical, error, NULL);
		i = tokval.t_type;
		if (!value) {	       /* error in evaluator */
		    result->opcode = -1;/* 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 {
		    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 == 0)		       /* also could be EOL */
		break;
	    if (i != ',') {
		error (ERR_NONFATAL, "comma expected after operand %d",
		       oper_num);
		result->opcode = -1;/* 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)
		error (ERR_NONFATAL, "`incbin' expects a file name");
	    else if (result->eops->next &&
		     result->eops->next->type != EOT_DB_NUMBER)
		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)
		error (ERR_NONFATAL, "`incbin': third parameter is",
		       " non-numeric");
	    else if (result->eops->next && result->eops->next->next &&
		     result->eops->next->next->next)
		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 = -1;
	    return result;
	} else /* DB ... */
	    if (oper_num == 0)
		error (ERR_WARNING|ERR_PASS1,
		       "no operand for data declaration");
            else
                result->operands = oper_num;

	return result;
    }

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

    for (operand = 0; operand < 3; 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;

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

	i = stdscan(NULL, &tokval);
	if (i == 0) break;	       /* end of operands: get out of here */
	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;