Fix infinite loop in free_preprocessor_line()
[smatch.git] / expression.c
blobcf076a9e5bfd4fddcf92aa5bc276218e28ade89e
1 /*
2 * sparse/expression.c
4 * Copyright (C) 2003 Transmeta Corp.
5 * 2003-2004 Linus Torvalds
7 * Licensed under the Open Software License version 1.1
9 * This is the expression parsing part of parsing C.
11 #include <stdarg.h>
12 #include <stdlib.h>
13 #include <stdio.h>
14 #include <string.h>
15 #include <ctype.h>
16 #include <unistd.h>
17 #include <fcntl.h>
18 #include <errno.h>
19 #include <limits.h>
21 #include "lib.h"
22 #include "allocate.h"
23 #include "token.h"
24 #include "parse.h"
25 #include "symbol.h"
26 #include "scope.h"
27 #include "expression.h"
28 #include "target.h"
30 static int match_oplist(int op, ...)
32 va_list args;
34 va_start(args, op);
35 for (;;) {
36 int nextop = va_arg(args, int);
37 if (!nextop)
38 return 0;
39 if (op == nextop)
40 return 1;
44 static struct token *comma_expression(struct token *, struct expression **);
46 struct token *parens_expression(struct token *token, struct expression **expr, const char *where)
48 token = expect(token, '(', where);
49 if (match_op(token, '{')) {
50 struct expression *e = alloc_expression(token->pos, EXPR_STATEMENT);
51 struct statement *stmt = alloc_statement(token->pos, STMT_COMPOUND);
52 *expr = e;
53 e->statement = stmt;
54 start_symbol_scope();
55 token = compound_statement(token->next, stmt);
56 end_symbol_scope();
57 token = expect(token, '}', "at end of statement expression");
58 } else
59 token = parse_expression(token, expr);
60 return expect(token, ')', where);
64 * Handle __func__, __FUNCTION__ and __PRETTY_FUNCTION__ token
65 * conversion
67 static int convert_one_fn_token(struct token *token)
69 struct symbol *sym = current_fn;
71 if (sym) {
72 struct ident *ident = sym->ident;
73 if (ident) {
74 int len = ident->len;
75 struct string *string;
77 string = __alloc_string(len+1);
78 memcpy(string->data, ident->name, len);
79 string->data[len] = 0;
80 string->length = len+1;
81 token_type(token) = TOKEN_STRING;
82 token->string = string;
83 return 1;
86 return 0;
89 static int convert_function(struct token *next)
91 int retval = 0;
92 for (;;) {
93 struct token *token = next;
94 next = next->next;
95 switch (token_type(token)) {
96 case TOKEN_STRING:
97 continue;
98 case TOKEN_IDENT:
99 if (token->ident == &__func___ident ||
100 token->ident == &__FUNCTION___ident ||
101 token->ident == &__PRETTY_FUNCTION___ident) {
102 if (!convert_one_fn_token(token))
103 break;
104 retval = 1;
105 continue;
107 /* Fall through */
108 default:
109 break;
111 break;
113 return retval;
116 static struct token *parse_type(struct token *token, struct expression **tree)
118 struct symbol *sym;
119 *tree = alloc_expression(token->pos, EXPR_TYPE);
120 token = typename(token, &sym);
121 if (sym->ident)
122 sparse_error(token->pos,
123 "type expression should not include identifier "
124 "\"%s\"", sym->ident->name);
125 (*tree)->symbol = sym;
126 return token;
129 static struct token *builtin_types_compatible_p_expr(struct token *token,
130 struct expression **tree)
132 struct expression *expr = alloc_expression(
133 token->pos, EXPR_COMPARE);
134 expr->op = SPECIAL_EQUAL;
135 token = token->next;
136 if (!match_op(token, '('))
137 return expect(token, '(',
138 "after __builtin_types_compatible_p");
139 token = token->next;
140 token = parse_type(token, &expr->left);
141 if (!match_op(token, ','))
142 return expect(token, ',',
143 "in __builtin_types_compatible_p");
144 token = token->next;
145 token = parse_type(token, &expr->right);
146 if (!match_op(token, ')'))
147 return expect(token, ')',
148 "at end of __builtin_types_compatible_p");
149 token = token->next;
151 *tree = expr;
152 return token;
155 static struct token *string_expression(struct token *token, struct expression *expr)
157 struct string *string = token->string;
158 struct token *next = token->next;
160 convert_function(token);
162 if (token_type(next) == TOKEN_STRING) {
163 int totlen = string->length-1;
164 char *data;
166 do {
167 totlen += next->string->length-1;
168 next = next->next;
169 } while (token_type(next) == TOKEN_STRING);
171 if (totlen > MAX_STRING) {
172 warning(token->pos, "trying to concatenate %d-character string (%d bytes max)", totlen, MAX_STRING);
173 totlen = MAX_STRING;
176 string = __alloc_string(totlen+1);
177 string->length = totlen+1;
178 data = string->data;
179 next = token;
180 do {
181 struct string *s = next->string;
182 int len = s->length-1;
184 if (len > totlen)
185 len = totlen;
186 totlen -= len;
188 next = next->next;
189 memcpy(data, s->data, len);
190 data += len;
191 } while (token_type(next) == TOKEN_STRING);
192 *data = '\0';
194 expr->string = string;
195 return next;
198 #ifndef ULLONG_MAX
199 #define ULLONG_MAX (~0ULL)
200 #endif
202 static void get_number_value(struct expression *expr, struct token *token)
204 const char *str = token->number;
205 unsigned long long value;
206 char *end;
207 unsigned long modifiers = 0;
208 int overflow = 0, do_warn = 0;
209 int try_unsigned = 1;
210 int bits;
212 errno = 0;
213 value = strtoull(str, &end, 0);
214 if (end == str)
215 goto Float;
216 if (value == ULLONG_MAX && errno == ERANGE)
217 overflow = 1;
218 while (1) {
219 unsigned long added;
220 char c = *end++;
221 if (!c) {
222 break;
223 } else if (c == 'u' || c == 'U') {
224 added = MOD_UNSIGNED;
225 } else if (c == 'l' || c == 'L') {
226 added = MOD_LONG;
227 if (*end == c) {
228 added |= MOD_LONGLONG;
229 end++;
231 } else
232 goto Float;
233 if (modifiers & added)
234 goto Enoint;
235 modifiers |= added;
237 if (overflow)
238 goto Eoverflow;
239 /* OK, it's a valid integer */
240 /* decimals can be unsigned only if directly specified as such */
241 if (str[0] != '0' && !(modifiers & MOD_UNSIGNED))
242 try_unsigned = 0;
243 if (!(modifiers & MOD_LONG)) {
244 bits = bits_in_int - 1;
245 if (!(value & (~1ULL << bits))) {
246 if (!(value & (1ULL << bits))) {
247 goto got_it;
248 } else if (try_unsigned) {
249 modifiers |= MOD_UNSIGNED;
250 goto got_it;
253 modifiers |= MOD_LONG;
254 do_warn = 1;
256 if (!(modifiers & MOD_LONGLONG)) {
257 bits = bits_in_long - 1;
258 if (!(value & (~1ULL << bits))) {
259 if (!(value & (1ULL << bits))) {
260 goto got_it;
261 } else if (try_unsigned) {
262 modifiers |= MOD_UNSIGNED;
263 goto got_it;
265 do_warn |= 2;
267 modifiers |= MOD_LONGLONG;
268 do_warn |= 1;
270 bits = bits_in_longlong - 1;
271 if (value & (~1ULL << bits))
272 goto Eoverflow;
273 if (!(value & (1ULL << bits)))
274 goto got_it;
275 if (!try_unsigned)
276 warning(expr->pos, "decimal constant %s is too big for long long",
277 show_token(token));
278 modifiers |= MOD_UNSIGNED;
279 got_it:
280 if (do_warn)
281 warning(expr->pos, "constant %s is so big it is%s%s%s",
282 show_token(token),
283 (modifiers & MOD_UNSIGNED) ? " unsigned":"",
284 (modifiers & MOD_LONG) ? " long":"",
285 (modifiers & MOD_LONGLONG) ? " long":"");
286 if (do_warn & 2)
287 warning(expr->pos,
288 "decimal constant %s is between LONG_MAX and ULONG_MAX."
289 " For C99 that means long long, C90 compilers are very "
290 "likely to produce unsigned long (and a warning) here",
291 show_token(token));
292 expr->type = EXPR_VALUE;
293 expr->ctype = ctype_integer(modifiers);
294 expr->value = value;
295 return;
296 Eoverflow:
297 error_die(expr->pos, "constant %s is too big even for unsigned long long",
298 show_token(token));
299 return;
300 Float:
301 expr->fvalue = string_to_ld(str, &end);
302 if (str == end)
303 goto Enoint;
305 if (*end && end[1])
306 goto Enoint;
308 if (*end == 'f' || *end == 'F')
309 expr->ctype = &float_ctype;
310 else if (*end == 'l' || *end == 'L')
311 expr->ctype = &ldouble_ctype;
312 else if (!*end)
313 expr->ctype = &double_ctype;
314 else
315 goto Enoint;
317 expr->type = EXPR_FVALUE;
318 return;
320 Enoint:
321 error_die(expr->pos, "constant %s is not a valid number", show_token(token));
324 struct token *primary_expression(struct token *token, struct expression **tree)
326 struct expression *expr = NULL;
328 switch (token_type(token)) {
329 case TOKEN_CHAR:
330 expr = alloc_expression(token->pos, EXPR_VALUE);
331 expr->ctype = &int_ctype;
332 expr->value = (unsigned char) token->character;
333 token = token->next;
334 break;
336 case TOKEN_NUMBER:
337 expr = alloc_expression(token->pos, EXPR_VALUE);
338 get_number_value(expr, token);
339 token = token->next;
340 break;
342 case TOKEN_ZERO_IDENT: {
343 expr = alloc_expression(token->pos, EXPR_SYMBOL);
344 expr->ctype = &int_ctype;
345 expr->symbol = &zero_int;
346 expr->symbol_name = token->ident;
347 token = token->next;
348 break;
351 case TOKEN_IDENT: {
352 struct symbol *sym = lookup_symbol(token->ident, NS_SYMBOL | NS_TYPEDEF);
353 struct token *next = token->next;
355 if (!sym) {
356 if (convert_function(token))
357 goto handle_string;
358 if (token->ident == &__builtin_types_compatible_p_ident) {
359 token = builtin_types_compatible_p_expr(token, &expr);
360 break;
364 expr = alloc_expression(token->pos, EXPR_SYMBOL);
367 * We support types as real first-class citizens, with type
368 * comparisons etc:
370 * if (typeof(a) == int) ..
372 if (sym && sym->namespace == NS_TYPEDEF) {
373 sparse_error(token->pos, "typename in expression");
374 sym = NULL;
376 expr->symbol_name = token->ident;
377 expr->symbol = sym;
378 token = next;
379 break;
382 case TOKEN_STRING: {
383 handle_string:
384 expr = alloc_expression(token->pos, EXPR_STRING);
385 token = string_expression(token, expr);
386 break;
389 case TOKEN_SPECIAL:
390 if (token->special == '(') {
391 expr = alloc_expression(token->pos, EXPR_PREOP);
392 expr->op = '(';
393 token = parens_expression(token, &expr->unop, "in expression");
394 break;
396 if (token->special == '[' && lookup_type(token->next)) {
397 expr = alloc_expression(token->pos, EXPR_TYPE);
398 token = typename(token->next, &expr->symbol);
399 token = expect(token, ']', "in type expression");
400 break;
403 default:
406 *tree = expr;
407 return token;
410 static struct token *expression_list(struct token *token, struct expression_list **list)
412 while (!match_op(token, ')')) {
413 struct expression *expr = NULL;
414 token = assignment_expression(token, &expr);
415 if (!expr)
416 break;
417 add_expression(list, expr);
418 if (!match_op(token, ','))
419 break;
420 token = token->next;
422 return token;
426 * extend to deal with the ambiguous C grammar for parsing
427 * a cast expressions followed by an initializer.
429 static struct token *postfix_expression(struct token *token, struct expression **tree, struct expression *cast_init_expr)
431 struct expression *expr = cast_init_expr;
433 if (!expr)
434 token = primary_expression(token, &expr);
436 while (expr && token_type(token) == TOKEN_SPECIAL) {
437 switch (token->special) {
438 case '[': { /* Array dereference */
439 struct expression *deref = alloc_expression(token->pos, EXPR_PREOP);
440 struct expression *add = alloc_expression(token->pos, EXPR_BINOP);
442 deref->op = '*';
443 deref->unop = add;
445 add->op = '+';
446 add->left = expr;
447 token = parse_expression(token->next, &add->right);
448 token = expect(token, ']', "at end of array dereference");
449 expr = deref;
450 continue;
452 case SPECIAL_INCREMENT: /* Post-increment */
453 case SPECIAL_DECREMENT: { /* Post-decrement */
454 struct expression *post = alloc_expression(token->pos, EXPR_POSTOP);
455 post->op = token->special;
456 post->unop = expr;
457 expr = post;
458 token = token->next;
459 continue;
461 case SPECIAL_DEREFERENCE: { /* Structure pointer member dereference */
462 /* "x->y" is just shorthand for "(*x).y" */
463 struct expression *inner = alloc_expression(token->pos, EXPR_PREOP);
464 inner->op = '*';
465 inner->unop = expr;
466 expr = inner;
468 /* Fall through!! */
469 case '.': { /* Structure member dereference */
470 struct expression *deref = alloc_expression(token->pos, EXPR_DEREF);
471 deref->op = '.';
472 deref->deref = expr;
473 token = token->next;
474 if (token_type(token) != TOKEN_IDENT) {
475 sparse_error(token->pos, "Expected member name");
476 break;
478 deref->member = token->ident;
479 token = token->next;
480 expr = deref;
481 continue;
484 case '(': { /* Function call */
485 struct expression *call = alloc_expression(token->pos, EXPR_CALL);
486 call->op = '(';
487 call->fn = expr;
488 token = expression_list(token->next, &call->args);
489 token = expect(token, ')', "in function call");
490 expr = call;
491 continue;
494 default:
495 break;
497 break;
499 *tree = expr;
500 return token;
503 static struct token *cast_expression(struct token *token, struct expression **tree);
504 static struct token *unary_expression(struct token *token, struct expression **tree);
506 static struct token *type_info_expression(struct token *token,
507 struct expression **tree, int type)
509 struct expression *expr = alloc_expression(token->pos, type);
511 *tree = expr;
512 token = token->next;
513 if (!match_op(token, '(') || !lookup_type(token->next))
514 return unary_expression(token, &expr->cast_expression);
515 token = typename(token->next, &expr->cast_type);
517 if (!match_op(token, ')')) {
518 static const char * error[] = {
519 [EXPR_SIZEOF] = "at end of sizeof",
520 [EXPR_ALIGNOF] = "at end of __alignof__",
521 [EXPR_PTRSIZEOF] = "at end of __sizeof_ptr__"
523 return expect(token, ')', error[type]);
526 token = token->next;
528 * C99 ambiguity: the typename might have been the beginning
529 * of a typed initializer expression..
531 if (match_op(token, '{'))
532 token = initializer(&expr->cast_expression, token);
533 return token;
536 static struct token *unary_expression(struct token *token, struct expression **tree)
538 if (token_type(token) == TOKEN_IDENT) {
539 struct ident *ident = token->ident;
540 if (ident->reserved) {
541 static const struct {
542 struct ident *id;
543 int type;
544 } type_information[] = {
545 { &sizeof_ident, EXPR_SIZEOF },
546 { &__alignof___ident, EXPR_ALIGNOF },
547 { &__alignof_ident, EXPR_ALIGNOF },
548 { &__sizeof_ptr___ident, EXPR_PTRSIZEOF },
550 int i;
551 for (i = 0; i < 3; i++) {
552 if (ident == type_information[i].id)
553 return type_info_expression(token, tree, type_information[i].type);
558 if (token_type(token) == TOKEN_SPECIAL) {
559 if (match_oplist(token->special,
560 SPECIAL_INCREMENT, SPECIAL_DECREMENT,
561 '&', '*', '+', '-', '~', '!', 0)) {
562 struct expression *unop;
563 struct expression *unary;
564 struct token *next;
566 next = cast_expression(token->next, &unop);
567 if (!unop) {
568 sparse_error(token->pos, "Syntax error in unary expression");
569 return next;
571 unary = alloc_expression(token->pos, EXPR_PREOP);
572 unary->op = token->special;
573 unary->unop = unop;
574 *tree = unary;
575 return next;
578 /* Gcc extension: &&label gives the address of a label */
579 if (match_op(token, SPECIAL_LOGICAL_AND) &&
580 token_type(token->next) == TOKEN_IDENT) {
581 struct expression *label = alloc_expression(token->pos, EXPR_LABEL);
582 struct symbol *sym = label_symbol(token->next);
583 if (!(sym->ctype.modifiers & MOD_ADDRESSABLE)) {
584 sym->ctype.modifiers |= MOD_ADDRESSABLE;
585 add_symbol(&function_computed_target_list, sym);
587 label->label_symbol = sym;
588 *tree = label;
589 return token->next->next;
594 return postfix_expression(token, tree, NULL);
598 * Ambiguity: a '(' can be either a cast-expression or
599 * a primary-expression depending on whether it is followed
600 * by a type or not.
602 * additional ambiguity: a "cast expression" followed by
603 * an initializer is really a postfix-expression.
605 static struct token *cast_expression(struct token *token, struct expression **tree)
607 if (match_op(token, '(')) {
608 struct token *next = token->next;
609 if (lookup_type(next)) {
610 struct expression *cast = alloc_expression(next->pos, EXPR_CAST);
611 struct symbol *sym;
613 token = typename(next, &sym);
614 cast->cast_type = sym;
615 token = expect(token, ')', "at end of cast operator");
616 if (match_op(token, '{')) {
617 token = initializer(&cast->cast_expression, token);
618 return postfix_expression(token, tree, cast);
620 *tree = cast;
621 token = cast_expression(token, &cast->cast_expression);
622 return token;
625 return unary_expression(token, tree);
629 * Generic left-to-right binop parsing
631 * This _really_ needs to be inlined, because that makes the inner
632 * function call statically deterministic rather than a totally
633 * unpredictable indirect call. But gcc-3 is so "clever" that it
634 * doesn't do so by default even when you tell it to inline it.
636 * Making it a macro avoids the inlining problem, and also means
637 * that we can pass in the op-comparison as an expression rather
638 * than create a data structure for it.
641 #define LR_BINOP_EXPRESSION(token, tree, type, inner, compare) \
642 struct expression *left = NULL; \
643 struct token * next = inner(token, &left); \
645 if (left) { \
646 while (token_type(next) == TOKEN_SPECIAL) { \
647 struct expression *top, *right = NULL; \
648 int op = next->special; \
650 if (!(compare)) \
651 goto out; \
652 top = alloc_expression(next->pos, type); \
653 next = inner(next->next, &right); \
654 if (!right) { \
655 sparse_error(next->pos, "No right hand side of '%s'-expression", show_special(op)); \
656 break; \
658 top->op = op; \
659 top->left = left; \
660 top->right = right; \
661 left = top; \
664 out: \
665 *tree = left; \
666 return next; \
669 static struct token *multiplicative_expression(struct token *token, struct expression **tree)
671 LR_BINOP_EXPRESSION(
672 token, tree, EXPR_BINOP, cast_expression,
673 (op == '*') || (op == '/') || (op == '%')
677 static struct token *additive_expression(struct token *token, struct expression **tree)
679 LR_BINOP_EXPRESSION(
680 token, tree, EXPR_BINOP, multiplicative_expression,
681 (op == '+') || (op == '-')
685 static struct token *shift_expression(struct token *token, struct expression **tree)
687 LR_BINOP_EXPRESSION(
688 token, tree, EXPR_BINOP, additive_expression,
689 (op == SPECIAL_LEFTSHIFT) || (op == SPECIAL_RIGHTSHIFT)
693 static struct token *relational_expression(struct token *token, struct expression **tree)
695 LR_BINOP_EXPRESSION(
696 token, tree, EXPR_COMPARE, shift_expression,
697 (op == '<') || (op == '>') ||
698 (op == SPECIAL_LTE) || (op == SPECIAL_GTE)
702 static struct token *equality_expression(struct token *token, struct expression **tree)
704 LR_BINOP_EXPRESSION(
705 token, tree, EXPR_COMPARE, relational_expression,
706 (op == SPECIAL_EQUAL) || (op == SPECIAL_NOTEQUAL)
710 static struct token *bitwise_and_expression(struct token *token, struct expression **tree)
712 LR_BINOP_EXPRESSION(
713 token, tree, EXPR_BINOP, equality_expression,
714 (op == '&')
718 static struct token *bitwise_xor_expression(struct token *token, struct expression **tree)
720 LR_BINOP_EXPRESSION(
721 token, tree, EXPR_BINOP, bitwise_and_expression,
722 (op == '^')
726 static struct token *bitwise_or_expression(struct token *token, struct expression **tree)
728 LR_BINOP_EXPRESSION(
729 token, tree, EXPR_BINOP, bitwise_xor_expression,
730 (op == '|')
734 static struct token *logical_and_expression(struct token *token, struct expression **tree)
736 LR_BINOP_EXPRESSION(
737 token, tree, EXPR_LOGICAL, bitwise_or_expression,
738 (op == SPECIAL_LOGICAL_AND)
742 static struct token *logical_or_expression(struct token *token, struct expression **tree)
744 LR_BINOP_EXPRESSION(
745 token, tree, EXPR_LOGICAL, logical_and_expression,
746 (op == SPECIAL_LOGICAL_OR)
750 struct token *conditional_expression(struct token *token, struct expression **tree)
752 token = logical_or_expression(token, tree);
753 if (*tree && match_op(token, '?')) {
754 struct expression *expr = alloc_expression(token->pos, EXPR_CONDITIONAL);
755 expr->op = token->special;
756 expr->left = *tree;
757 *tree = expr;
758 token = parse_expression(token->next, &expr->cond_true);
759 token = expect(token, ':', "in conditional expression");
760 token = conditional_expression(token, &expr->cond_false);
762 return token;
765 struct token *assignment_expression(struct token *token, struct expression **tree)
767 token = conditional_expression(token, tree);
768 if (*tree && token_type(token) == TOKEN_SPECIAL) {
769 static const int assignments[] = {
770 '=',
771 SPECIAL_ADD_ASSIGN, SPECIAL_SUB_ASSIGN,
772 SPECIAL_MUL_ASSIGN, SPECIAL_DIV_ASSIGN,
773 SPECIAL_MOD_ASSIGN, SPECIAL_SHL_ASSIGN,
774 SPECIAL_SHR_ASSIGN, SPECIAL_AND_ASSIGN,
775 SPECIAL_OR_ASSIGN, SPECIAL_XOR_ASSIGN };
776 int i, op = token->special;
777 for (i = 0; i < sizeof(assignments)/sizeof(int); i++)
778 if (assignments[i] == op) {
779 struct expression * expr = alloc_expression(token->pos, EXPR_ASSIGNMENT);
780 expr->left = *tree;
781 expr->op = op;
782 *tree = expr;
783 return assignment_expression(token->next, &expr->right);
786 return token;
789 static struct token *comma_expression(struct token *token, struct expression **tree)
791 LR_BINOP_EXPRESSION(
792 token, tree, EXPR_COMMA, assignment_expression,
793 (op == ',')
797 struct token *parse_expression(struct token *token, struct expression **tree)
799 return comma_expression(token,tree);