Fixed distinction between [LOCAL]SYMBOL/IMMEDIATE for RIP-relative addressing.
[nasm.git] / eval.c
blob35baf27cfc53113da2c6195dcbf07ccdcf533680
1 /* eval.c expression evaluator for the Netwide Assembler
3 * The Netwide Assembler is copyright (C) 1996 Simon Tatham and
4 * Julian Hall. All rights reserved. The software is
5 * redistributable under the licence given in the file "Licence"
6 * distributed in the NASM archive.
8 * initial version 27/iii/95 by Simon Tatham
9 */
11 #include <stdio.h>
12 #include <stdlib.h>
13 #include <stddef.h>
14 #include <string.h>
15 #include <ctype.h>
16 #include <inttypes.h>
18 #include "nasm.h"
19 #include "nasmlib.h"
20 #include "eval.h"
21 #include "labels.h"
23 #define TEMPEXPRS_DELTA 128
24 #define TEMPEXPR_DELTA 8
26 static scanner scan; /* Address of scanner routine */
27 static efunc error; /* Address of error reporting routine */
28 static lfunc labelfunc; /* Address of label routine */
30 static struct ofmt *outfmt; /* Structure of addresses of output routines */
32 static expr **tempexprs = NULL;
33 static int ntempexprs;
34 static int tempexprs_size = 0;
36 static expr *tempexpr;
37 static int ntempexpr;
38 static int tempexpr_size;
40 static struct tokenval *tokval; /* The current token */
41 static int i; /* The t_type of tokval */
43 static void *scpriv;
44 static loc_t *location; /* Pointer to current line's segment,offset */
45 static int *opflags;
47 static struct eval_hints *hint;
49 extern int in_abs_seg; /* ABSOLUTE segment flag */
50 extern int32_t abs_seg; /* ABSOLUTE segment */
51 extern int32_t abs_offset; /* ABSOLUTE segment offset */
54 * Unimportant cleanup is done to avoid confusing people who are trying
55 * to debug real memory leaks
57 void eval_cleanup(void)
59 while (ntempexprs)
60 nasm_free(tempexprs[--ntempexprs]);
61 nasm_free(tempexprs);
65 * Construct a temporary expression.
67 static void begintemp(void)
69 tempexpr = NULL;
70 tempexpr_size = ntempexpr = 0;
73 static void addtotemp(int32_t type, int64_t value)
75 while (ntempexpr >= tempexpr_size) {
76 tempexpr_size += TEMPEXPR_DELTA;
77 tempexpr = nasm_realloc(tempexpr,
78 tempexpr_size * sizeof(*tempexpr));
80 tempexpr[ntempexpr].type = type;
81 tempexpr[ntempexpr++].value = value;
84 static expr *finishtemp(void)
86 addtotemp(0L, 0L); /* terminate */
87 while (ntempexprs >= tempexprs_size) {
88 tempexprs_size += TEMPEXPRS_DELTA;
89 tempexprs = nasm_realloc(tempexprs,
90 tempexprs_size * sizeof(*tempexprs));
92 return tempexprs[ntempexprs++] = tempexpr;
96 * Add two vector datatypes. We have some bizarre behaviour on far-
97 * absolute segment types: we preserve them during addition _only_
98 * if one of the segments is a truly pure scalar.
100 static expr *add_vectors(expr * p, expr * q)
102 int preserve;
104 preserve = is_really_simple(p) || is_really_simple(q);
106 begintemp();
108 while (p->type && q->type &&
109 p->type < EXPR_SEGBASE + SEG_ABS &&
110 q->type < EXPR_SEGBASE + SEG_ABS) {
111 int lasttype;
113 if (p->type > q->type) {
114 addtotemp(q->type, q->value);
115 lasttype = q++->type;
116 } else if (p->type < q->type) {
117 addtotemp(p->type, p->value);
118 lasttype = p++->type;
119 } else { /* *p and *q have same type */
120 int32_t sum = p->value + q->value;
121 if (sum)
122 addtotemp(p->type, sum);
123 lasttype = p->type;
124 p++, q++;
126 if (lasttype == EXPR_UNKNOWN) {
127 return finishtemp();
130 while (p->type && (preserve || p->type < EXPR_SEGBASE + SEG_ABS)) {
131 addtotemp(p->type, p->value);
132 p++;
134 while (q->type && (preserve || q->type < EXPR_SEGBASE + SEG_ABS)) {
135 addtotemp(q->type, q->value);
136 q++;
139 return finishtemp();
143 * Multiply a vector by a scalar. Strip far-absolute segment part
144 * if present.
146 * Explicit treatment of UNKNOWN is not required in this routine,
147 * since it will silently do the Right Thing anyway.
149 * If `affect_hints' is set, we also change the hint type to
150 * NOTBASE if a MAKEBASE hint points at a register being
151 * multiplied. This allows [eax*1+ebx] to hint EBX rather than EAX
152 * as the base register.
154 static expr *scalar_mult(expr * vect, int32_t scalar, int affect_hints)
156 expr *p = vect;
158 while (p->type && p->type < EXPR_SEGBASE + SEG_ABS) {
159 p->value = scalar * (p->value);
160 if (hint && hint->type == EAH_MAKEBASE &&
161 p->type == hint->base && affect_hints)
162 hint->type = EAH_NOTBASE;
163 p++;
165 p->type = 0;
167 return vect;
170 static expr *scalarvect(int32_t scalar)
172 begintemp();
173 addtotemp(EXPR_SIMPLE, scalar);
174 return finishtemp();
177 static expr *unknown_expr(void)
179 begintemp();
180 addtotemp(EXPR_UNKNOWN, 1L);
181 return finishtemp();
185 * The SEG operator: calculate the segment part of a relocatable
186 * value. Return NULL, as usual, if an error occurs. Report the
187 * error too.
189 static expr *segment_part(expr * e)
191 int32_t seg;
193 if (is_unknown(e))
194 return unknown_expr();
196 if (!is_reloc(e)) {
197 error(ERR_NONFATAL, "cannot apply SEG to a non-relocatable value");
198 return NULL;
201 seg = reloc_seg(e);
202 if (seg == NO_SEG) {
203 error(ERR_NONFATAL, "cannot apply SEG to a non-relocatable value");
204 return NULL;
205 } else if (seg & SEG_ABS) {
206 return scalarvect(seg & ~SEG_ABS);
207 } else if (seg & 1) {
208 error(ERR_NONFATAL, "SEG applied to something which"
209 " is already a segment base");
210 return NULL;
211 } else {
212 int32_t base = outfmt->segbase(seg + 1);
214 begintemp();
215 addtotemp((base == NO_SEG ? EXPR_UNKNOWN : EXPR_SEGBASE + base),
216 1L);
217 return finishtemp();
222 * Recursive-descent parser. Called with a single boolean operand,
223 * which is TRUE if the evaluation is critical (i.e. unresolved
224 * symbols are an error condition). Must update the global `i' to
225 * reflect the token after the parsed string. May return NULL.
227 * evaluate() should report its own errors: on return it is assumed
228 * that if NULL has been returned, the error has already been
229 * reported.
233 * Grammar parsed is:
235 * expr : bexpr [ WRT expr6 ]
236 * bexpr : rexp0 or expr0 depending on relative-mode setting
237 * rexp0 : rexp1 [ {||} rexp1...]
238 * rexp1 : rexp2 [ {^^} rexp2...]
239 * rexp2 : rexp3 [ {&&} rexp3...]
240 * rexp3 : expr0 [ {=,==,<>,!=,<,>,<=,>=} expr0 ]
241 * expr0 : expr1 [ {|} expr1...]
242 * expr1 : expr2 [ {^} expr2...]
243 * expr2 : expr3 [ {&} expr3...]
244 * expr3 : expr4 [ {<<,>>} expr4...]
245 * expr4 : expr5 [ {+,-} expr5...]
246 * expr5 : expr6 [ {*,/,%,//,%%} expr6...]
247 * expr6 : { ~,+,-,SEG } expr6
248 * | (bexpr)
249 * | symbol
250 * | $
251 * | number
254 static expr *rexp0(int), *rexp1(int), *rexp2(int), *rexp3(int);
256 static expr *expr0(int), *expr1(int), *expr2(int), *expr3(int);
257 static expr *expr4(int), *expr5(int), *expr6(int);
259 static expr *(*bexpr) (int);
261 static expr *rexp0(int critical)
263 expr *e, *f;
265 e = rexp1(critical);
266 if (!e)
267 return NULL;
269 while (i == TOKEN_DBL_OR) {
270 i = scan(scpriv, tokval);
271 f = rexp1(critical);
272 if (!f)
273 return NULL;
274 if (!(is_simple(e) || is_just_unknown(e)) ||
275 !(is_simple(f) || is_just_unknown(f))) {
276 error(ERR_NONFATAL, "`|' operator may only be applied to"
277 " scalar values");
280 if (is_just_unknown(e) || is_just_unknown(f))
281 e = unknown_expr();
282 else
283 e = scalarvect((int32_t)(reloc_value(e) || reloc_value(f)));
285 return e;
288 static expr *rexp1(int critical)
290 expr *e, *f;
292 e = rexp2(critical);
293 if (!e)
294 return NULL;
296 while (i == TOKEN_DBL_XOR) {
297 i = scan(scpriv, tokval);
298 f = rexp2(critical);
299 if (!f)
300 return NULL;
301 if (!(is_simple(e) || is_just_unknown(e)) ||
302 !(is_simple(f) || is_just_unknown(f))) {
303 error(ERR_NONFATAL, "`^' operator may only be applied to"
304 " scalar values");
307 if (is_just_unknown(e) || is_just_unknown(f))
308 e = unknown_expr();
309 else
310 e = scalarvect((int32_t)(!reloc_value(e) ^ !reloc_value(f)));
312 return e;
315 static expr *rexp2(int critical)
317 expr *e, *f;
319 e = rexp3(critical);
320 if (!e)
321 return NULL;
322 while (i == TOKEN_DBL_AND) {
323 i = scan(scpriv, tokval);
324 f = rexp3(critical);
325 if (!f)
326 return NULL;
327 if (!(is_simple(e) || is_just_unknown(e)) ||
328 !(is_simple(f) || is_just_unknown(f))) {
329 error(ERR_NONFATAL, "`&' operator may only be applied to"
330 " scalar values");
332 if (is_just_unknown(e) || is_just_unknown(f))
333 e = unknown_expr();
334 else
335 e = scalarvect((int32_t)(reloc_value(e) && reloc_value(f)));
337 return e;
340 static expr *rexp3(int critical)
342 expr *e, *f;
343 int32_t v;
345 e = expr0(critical);
346 if (!e)
347 return NULL;
349 while (i == TOKEN_EQ || i == TOKEN_LT || i == TOKEN_GT ||
350 i == TOKEN_NE || i == TOKEN_LE || i == TOKEN_GE) {
351 int j = i;
352 i = scan(scpriv, tokval);
353 f = expr0(critical);
354 if (!f)
355 return NULL;
357 e = add_vectors(e, scalar_mult(f, -1L, FALSE));
359 switch (j) {
360 case TOKEN_EQ:
361 case TOKEN_NE:
362 if (is_unknown(e))
363 v = -1; /* means unknown */
364 else if (!is_really_simple(e) || reloc_value(e) != 0)
365 v = (j == TOKEN_NE); /* unequal, so return TRUE if NE */
366 else
367 v = (j == TOKEN_EQ); /* equal, so return TRUE if EQ */
368 break;
369 default:
370 if (is_unknown(e))
371 v = -1; /* means unknown */
372 else if (!is_really_simple(e)) {
373 error(ERR_NONFATAL,
374 "`%s': operands differ by a non-scalar",
375 (j == TOKEN_LE ? "<=" : j == TOKEN_LT ? "<" : j ==
376 TOKEN_GE ? ">=" : ">"));
377 v = 0; /* must set it to _something_ */
378 } else {
379 int vv = reloc_value(e);
380 if (vv == 0)
381 v = (j == TOKEN_LE || j == TOKEN_GE);
382 else if (vv > 0)
383 v = (j == TOKEN_GE || j == TOKEN_GT);
384 else /* vv < 0 */
385 v = (j == TOKEN_LE || j == TOKEN_LT);
387 break;
390 if (v == -1)
391 e = unknown_expr();
392 else
393 e = scalarvect(v);
395 return e;
398 static expr *expr0(int critical)
400 expr *e, *f;
402 e = expr1(critical);
403 if (!e)
404 return NULL;
406 while (i == '|') {
407 i = scan(scpriv, tokval);
408 f = expr1(critical);
409 if (!f)
410 return NULL;
411 if (!(is_simple(e) || is_just_unknown(e)) ||
412 !(is_simple(f) || is_just_unknown(f))) {
413 error(ERR_NONFATAL, "`|' operator may only be applied to"
414 " scalar values");
416 if (is_just_unknown(e) || is_just_unknown(f))
417 e = unknown_expr();
418 else
419 e = scalarvect(reloc_value(e) | reloc_value(f));
421 return e;
424 static expr *expr1(int critical)
426 expr *e, *f;
428 e = expr2(critical);
429 if (!e)
430 return NULL;
432 while (i == '^') {
433 i = scan(scpriv, tokval);
434 f = expr2(critical);
435 if (!f)
436 return NULL;
437 if (!(is_simple(e) || is_just_unknown(e)) ||
438 !(is_simple(f) || is_just_unknown(f))) {
439 error(ERR_NONFATAL, "`^' operator may only be applied to"
440 " scalar values");
442 if (is_just_unknown(e) || is_just_unknown(f))
443 e = unknown_expr();
444 else
445 e = scalarvect(reloc_value(e) ^ reloc_value(f));
447 return e;
450 static expr *expr2(int critical)
452 expr *e, *f;
454 e = expr3(critical);
455 if (!e)
456 return NULL;
458 while (i == '&') {
459 i = scan(scpriv, tokval);
460 f = expr3(critical);
461 if (!f)
462 return NULL;
463 if (!(is_simple(e) || is_just_unknown(e)) ||
464 !(is_simple(f) || is_just_unknown(f))) {
465 error(ERR_NONFATAL, "`&' operator may only be applied to"
466 " scalar values");
468 if (is_just_unknown(e) || is_just_unknown(f))
469 e = unknown_expr();
470 else
471 e = scalarvect(reloc_value(e) & reloc_value(f));
473 return e;
476 static expr *expr3(int critical)
478 expr *e, *f;
480 e = expr4(critical);
481 if (!e)
482 return NULL;
484 while (i == TOKEN_SHL || i == TOKEN_SHR) {
485 int j = i;
486 i = scan(scpriv, tokval);
487 f = expr4(critical);
488 if (!f)
489 return NULL;
490 if (!(is_simple(e) || is_just_unknown(e)) ||
491 !(is_simple(f) || is_just_unknown(f))) {
492 error(ERR_NONFATAL, "shift operator may only be applied to"
493 " scalar values");
494 } else if (is_just_unknown(e) || is_just_unknown(f)) {
495 e = unknown_expr();
496 } else
497 switch (j) {
498 case TOKEN_SHL:
499 e = scalarvect(reloc_value(e) << reloc_value(f));
500 break;
501 case TOKEN_SHR:
502 e = scalarvect(((uint32_t)reloc_value(e)) >>
503 reloc_value(f));
504 break;
507 return e;
510 static expr *expr4(int critical)
512 expr *e, *f;
514 e = expr5(critical);
515 if (!e)
516 return NULL;
517 while (i == '+' || i == '-') {
518 int j = i;
519 i = scan(scpriv, tokval);
520 f = expr5(critical);
521 if (!f)
522 return NULL;
523 switch (j) {
524 case '+':
525 e = add_vectors(e, f);
526 break;
527 case '-':
528 e = add_vectors(e, scalar_mult(f, -1L, FALSE));
529 break;
532 return e;
535 static expr *expr5(int critical)
537 expr *e, *f;
539 e = expr6(critical);
540 if (!e)
541 return NULL;
542 while (i == '*' || i == '/' || i == '%' ||
543 i == TOKEN_SDIV || i == TOKEN_SMOD) {
544 int j = i;
545 i = scan(scpriv, tokval);
546 f = expr6(critical);
547 if (!f)
548 return NULL;
549 if (j != '*' && (!(is_simple(e) || is_just_unknown(e)) ||
550 !(is_simple(f) || is_just_unknown(f)))) {
551 error(ERR_NONFATAL, "division operator may only be applied to"
552 " scalar values");
553 return NULL;
555 if (j != '*' && !is_unknown(f) && reloc_value(f) == 0) {
556 error(ERR_NONFATAL, "division by zero");
557 return NULL;
559 switch (j) {
560 case '*':
561 if (is_simple(e))
562 e = scalar_mult(f, reloc_value(e), TRUE);
563 else if (is_simple(f))
564 e = scalar_mult(e, reloc_value(f), TRUE);
565 else if (is_just_unknown(e) && is_just_unknown(f))
566 e = unknown_expr();
567 else {
568 error(ERR_NONFATAL, "unable to multiply two "
569 "non-scalar objects");
570 return NULL;
572 break;
573 case '/':
574 if (is_just_unknown(e) || is_just_unknown(f))
575 e = unknown_expr();
576 else
577 e = scalarvect(((uint32_t)reloc_value(e)) /
578 ((uint32_t)reloc_value(f)));
579 break;
580 case '%':
581 if (is_just_unknown(e) || is_just_unknown(f))
582 e = unknown_expr();
583 else
584 e = scalarvect(((uint32_t)reloc_value(e)) %
585 ((uint32_t)reloc_value(f)));
586 break;
587 case TOKEN_SDIV:
588 if (is_just_unknown(e) || is_just_unknown(f))
589 e = unknown_expr();
590 else
591 e = scalarvect(((int32_t)reloc_value(e)) /
592 ((int32_t)reloc_value(f)));
593 break;
594 case TOKEN_SMOD:
595 if (is_just_unknown(e) || is_just_unknown(f))
596 e = unknown_expr();
597 else
598 e = scalarvect(((int32_t)reloc_value(e)) %
599 ((int32_t)reloc_value(f)));
600 break;
603 return e;
606 static expr *expr6(int critical)
608 int32_t type;
609 expr *e;
610 int32_t label_seg, label_ofs;
612 if (i == '-') {
613 i = scan(scpriv, tokval);
614 e = expr6(critical);
615 if (!e)
616 return NULL;
617 return scalar_mult(e, -1L, FALSE);
618 } else if (i == '+') {
619 i = scan(scpriv, tokval);
620 return expr6(critical);
621 } else if (i == '~') {
622 i = scan(scpriv, tokval);
623 e = expr6(critical);
624 if (!e)
625 return NULL;
626 if (is_just_unknown(e))
627 return unknown_expr();
628 else if (!is_simple(e)) {
629 error(ERR_NONFATAL, "`~' operator may only be applied to"
630 " scalar values");
631 return NULL;
633 return scalarvect(~reloc_value(e));
634 } else if (i == TOKEN_SEG) {
635 i = scan(scpriv, tokval);
636 e = expr6(critical);
637 if (!e)
638 return NULL;
639 e = segment_part(e);
640 if (!e)
641 return NULL;
642 if (is_unknown(e) && critical) {
643 error(ERR_NONFATAL, "unable to determine segment base");
644 return NULL;
646 return e;
647 } else if (i == '(') {
648 i = scan(scpriv, tokval);
649 e = bexpr(critical);
650 if (!e)
651 return NULL;
652 if (i != ')') {
653 error(ERR_NONFATAL, "expecting `)'");
654 return NULL;
656 i = scan(scpriv, tokval);
657 return e;
658 } else if (i == TOKEN_NUM || i == TOKEN_REG || i == TOKEN_ID ||
659 i == TOKEN_HERE || i == TOKEN_BASE) {
660 begintemp();
661 switch (i) {
662 case TOKEN_NUM:
663 addtotemp(EXPR_SIMPLE, tokval->t_integer);
664 break;
665 case TOKEN_REG:
666 addtotemp(tokval->t_integer, 1L);
667 if (hint && hint->type == EAH_NOHINT)
668 hint->base = tokval->t_integer, hint->type = EAH_MAKEBASE;
669 break;
670 case TOKEN_ID:
671 case TOKEN_HERE:
672 case TOKEN_BASE:
674 * If !location->known, this indicates that no
675 * symbol, Here or Base references are valid because we
676 * are in preprocess-only mode.
678 if (!location->known) {
679 error(ERR_NONFATAL,
680 "%s not supported in preprocess-only mode",
681 (i == TOKEN_ID ? "symbol references" :
682 i == TOKEN_HERE ? "`$'" : "`$$'"));
683 addtotemp(EXPR_UNKNOWN, 1L);
684 break;
687 type = EXPR_SIMPLE; /* might get overridden by UNKNOWN */
688 if (i == TOKEN_BASE) {
689 label_seg = in_abs_seg ? abs_seg : location->segment;
690 label_ofs = 0;
691 } else if (i == TOKEN_HERE) {
692 label_seg = in_abs_seg ? abs_seg : location->segment;
693 label_ofs = in_abs_seg ? abs_offset : location->offset;
694 } else {
695 if (!labelfunc(tokval->t_charptr, &label_seg, &label_ofs)) {
696 if (critical == 2) {
697 error(ERR_NONFATAL, "symbol `%s' undefined",
698 tokval->t_charptr);
699 return NULL;
700 } else if (critical == 1) {
701 error(ERR_NONFATAL,
702 "symbol `%s' not defined before use",
703 tokval->t_charptr);
704 return NULL;
705 } else {
706 if (opflags)
707 *opflags |= 1;
708 type = EXPR_UNKNOWN;
709 label_seg = NO_SEG;
710 label_ofs = 1;
713 if (opflags && is_extern(tokval->t_charptr))
714 *opflags |= OPFLAG_EXTERN;
716 addtotemp(type, label_ofs);
717 if (label_seg != NO_SEG)
718 addtotemp(EXPR_SEGBASE + label_seg, 1L);
719 break;
721 i = scan(scpriv, tokval);
722 return finishtemp();
723 } else {
724 error(ERR_NONFATAL, "expression syntax error");
725 return NULL;
729 void eval_global_info(struct ofmt *output, lfunc lookup_label,
730 loc_t * locp)
732 outfmt = output;
733 labelfunc = lookup_label;
734 location = locp;
737 expr *evaluate(scanner sc, void *scprivate, struct tokenval *tv,
738 int *fwref, int critical, efunc report_error,
739 struct eval_hints *hints)
741 expr *e;
742 expr *f = NULL;
744 hint = hints;
745 if (hint)
746 hint->type = EAH_NOHINT;
748 if (critical & CRITICAL) {
749 critical &= ~CRITICAL;
750 bexpr = rexp0;
751 } else
752 bexpr = expr0;
754 scan = sc;
755 scpriv = scprivate;
756 tokval = tv;
757 error = report_error;
758 opflags = fwref;
760 if (tokval->t_type == TOKEN_INVALID)
761 i = scan(scpriv, tokval);
762 else
763 i = tokval->t_type;
765 while (ntempexprs) /* initialize temporary storage */
766 nasm_free(tempexprs[--ntempexprs]);
768 e = bexpr(critical);
769 if (!e)
770 return NULL;
772 if (i == TOKEN_WRT) {
773 i = scan(scpriv, tokval); /* eat the WRT */
774 f = expr6(critical);
775 if (!f)
776 return NULL;
778 e = scalar_mult(e, 1L, FALSE); /* strip far-absolute segment part */
779 if (f) {
780 expr *g;
781 if (is_just_unknown(f))
782 g = unknown_expr();
783 else {
784 int64_t value;
785 begintemp();
786 if (!is_reloc(f)) {
787 error(ERR_NONFATAL, "invalid right-hand operand to WRT");
788 return NULL;
790 value = reloc_seg(f);
791 if (value == NO_SEG)
792 value = reloc_value(f) | SEG_ABS;
793 else if (!(value & SEG_ABS) && !(value % 2) && critical) {
794 error(ERR_NONFATAL, "invalid right-hand operand to WRT");
795 return NULL;
797 addtotemp(EXPR_WRT, value);
798 g = finishtemp();
800 e = add_vectors(e, g);
802 return e;