1 /* Parse C expressions for CCCP.
2 Copyright (C) 1987, 1992, 1994, 1995 Free Software Foundation.
4 This program is free software; you can redistribute it and/or modify it
5 under the terms of the GNU General Public License as published by the
6 Free Software Foundation; either version 2, or (at your option) any
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, 59 Temple Place - Suite 330,
17 Boston, MA 02111-1307, USA.
19 In other words, you are welcome to use, share and improve this program.
20 You are forbidden to forbid anyone else to use, share and improve
21 what you give them. Help stamp out software-hoarding!
23 Written by Per Bothner 1994. */
25 /* Parse a C expression from text in a string */
30 extern char *xmalloc PARAMS ((unsigned));
31 extern char *xrealloc PARAMS ((char *, unsigned));
33 #ifdef MULTIBYTE_CHARS
40 /* This is used for communicating lists of keywords with cccp.c. */
48 /* Define a generic NULL if one hasn't already been defined. */
55 #if defined (USE_PROTOTYPES) ? USE_PROTOTYPES : defined (__STDC__)
56 #define GENERIC_PTR void *
58 #define GENERIC_PTR char *
63 #define NULL_PTR ((GENERIC_PTR)0)
66 extern char *xmalloc ();
68 #ifndef CHAR_TYPE_SIZE
69 #define CHAR_TYPE_SIZE BITS_PER_UNIT
73 #define INT_TYPE_SIZE BITS_PER_WORD
76 #ifndef LONG_TYPE_SIZE
77 #define LONG_TYPE_SIZE BITS_PER_WORD
80 #ifndef WCHAR_TYPE_SIZE
81 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
84 #ifndef MAX_CHAR_TYPE_SIZE
85 #define MAX_CHAR_TYPE_SIZE CHAR_TYPE_SIZE
88 #ifndef MAX_INT_TYPE_SIZE
89 #define MAX_INT_TYPE_SIZE INT_TYPE_SIZE
92 #ifndef MAX_LONG_TYPE_SIZE
93 #define MAX_LONG_TYPE_SIZE LONG_TYPE_SIZE
96 #ifndef MAX_WCHAR_TYPE_SIZE
97 #define MAX_WCHAR_TYPE_SIZE WCHAR_TYPE_SIZE
100 /* Yield nonzero if adding two numbers with A's and B's signs can yield a
101 number with SUM's sign, where A, B, and SUM are all C integers. */
102 #define possible_sum_sign(a, b, sum) ((((a) ^ (b)) | ~ ((a) ^ (sum))) < 0)
104 static void integer_overflow ();
105 static long left_shift ();
106 static long right_shift ();
121 #define LEFT_OPERAND_REQUIRED 1
122 #define RIGHT_OPERAND_REQUIRED 2
124 /* SKIP_OPERAND is set for '&&' '||' '?' and ':' when the
125 following operand should be short-circuited instead of evaluated. */
126 #define SKIP_OPERAND 8
127 /*#define UNSIGNEDP 16*/
129 #ifndef HOST_BITS_PER_WIDE_INT
131 #if HOST_BITS_PER_LONG > HOST_BITS_PER_INT
132 #define HOST_BITS_PER_WIDE_INT HOST_BITS_PER_LONG
133 #define HOST_WIDE_INT long
135 #define HOST_BITS_PER_WIDE_INT HOST_BITS_PER_INT
136 #define HOST_WIDE_INT int
143 char rprio; /* Priority of op (relative to it right operand). */
145 char unsignedp; /* true if value should be treated as unsigned */
146 HOST_WIDE_INT value; /* The value logically "right" of op. */
149 /* Take care of parsing a number (anything that starts with a digit).
150 LEN is the number of characters in it. */
152 /* maybe needs to actually deal with floating point numbers */
155 parse_number (pfile, start, olen)
161 register char *p = start;
163 register unsigned long n = 0, nd, ULONG_MAX_over_base;
164 register int base = 10;
165 register int len = olen;
166 register int overflow = 0;
167 register int digit, largest_digit = 0;
172 for (c = 0; c < len; c++)
174 /* It's a float since it contains a point. */
176 "floating point numbers not allowed in #if expressions");
181 if (len >= 3 && (!strncmp (p, "0x", 2) || !strncmp (p, "0X", 2))) {
189 /* Some buggy compilers (e.g. MPW C) seem to need both casts. */
190 ULONG_MAX_over_base = ((unsigned long) -1) / ((unsigned long) base);
192 for (; len > 0; len--) {
195 if (c >= '0' && c <= '9')
197 else if (base == 16 && c >= 'a' && c <= 'f')
198 digit = c - 'a' + 10;
199 else if (base == 16 && c >= 'A' && c <= 'F')
200 digit = c - 'A' + 10;
202 /* `l' means long, and `u' means unsigned. */
204 if (c == 'l' || c == 'L')
207 cpp_error (pfile, "two `l's in integer constant");
210 else if (c == 'u' || c == 'U')
213 cpp_error (pfile, "two `u's in integer constant");
223 /* Don't look for any more digits after the suffixes. */
226 if (largest_digit < digit)
227 largest_digit = digit;
228 nd = n * base + digit;
229 overflow |= ULONG_MAX_over_base < n | nd < n;
235 cpp_error (pfile, "Invalid number in #if expression");
240 if (base <= largest_digit)
241 cpp_warning (pfile, "integer constant contains digits beyond the radix");
244 cpp_warning (pfile, "integer constant out of range");
246 /* If too big to be signed, consider it unsigned. */
247 if ((long) n < 0 && ! op.unsignedp)
250 cpp_warning (pfile, "integer constant is so large that it is unsigned");
264 static struct token tokentab2[] = {
278 /* Read one token. */
285 register int namelen;
286 register struct token *toktab;
287 enum cpp_token token;
289 U_CHAR *tok_start, *tok_end;
294 old_written = CPP_WRITTEN (pfile);
295 cpp_skip_hspace (pfile);
296 c = CPP_BUF_PEEK (CPP_BUFFER (pfile));
298 return parse_number (pfile,
299 cpp_read_check_assertion (pfile) ? "1" : "0", 1);
307 token = cpp_get_token (pfile);
308 tok_start = pfile->token_buffer + old_written;
309 tok_end = CPP_PWRITTEN (pfile);
310 pfile->limit = tok_start;
313 case CPP_EOF: /* Should not happen ... */
318 if (CPP_BUFFER (pfile)->fname != NULL)
323 cpp_pop_buffer (pfile);
325 case CPP_HSPACE: case CPP_COMMENT:
328 return parse_number (pfile, tok_start, tok_end - tok_start);
330 cpp_error (pfile, "string constants not allowed in #if expressions");
334 /* This code for reading a character constant
335 handles multicharacter constants and wide characters.
336 It is mostly copied from c-lex.c. */
338 register int result = 0;
339 register num_chars = 0;
340 unsigned width = MAX_CHAR_TYPE_SIZE;
343 U_CHAR *ptr = tok_start;
344 #ifdef MULTIBYTE_CHARS
345 char token_buffer[MAX_LONG_TYPE_SIZE/MAX_CHAR_TYPE_SIZE + MB_CUR_MAX];
347 char token_buffer[MAX_LONG_TYPE_SIZE/MAX_CHAR_TYPE_SIZE + 1];
354 width = MAX_WCHAR_TYPE_SIZE;
355 #ifdef MULTIBYTE_CHARS
356 max_chars = MB_CUR_MAX;
362 max_chars = MAX_LONG_TYPE_SIZE / width;
365 while (ptr < tok_end && ((c = *ptr++) != '\''))
369 c = cpp_parse_escape (pfile, &ptr);
370 if (width < HOST_BITS_PER_INT
371 && (unsigned) c >= (1 << width))
373 "escape sequence out of range for character");
378 /* Merge character into result; ignore excess chars. */
379 if (num_chars < max_chars + 1)
381 if (width < HOST_BITS_PER_INT)
382 result = (result << width) | (c & ((1 << width) - 1));
385 token_buffer[num_chars - 1] = c;
389 token_buffer[num_chars] = 0;
392 cpp_error (pfile, "malformatted character constant");
393 else if (num_chars == 0)
394 cpp_error (pfile, "empty character constant");
395 else if (num_chars > max_chars)
397 num_chars = max_chars;
398 cpp_error (pfile, "character constant too long");
400 else if (num_chars != 1 && ! CPP_TRADITIONAL (pfile))
401 cpp_warning (pfile, "multi-character character constant");
403 /* If char type is signed, sign-extend the constant. */
406 int num_bits = num_chars * width;
408 if (cpp_lookup (pfile, "__CHAR_UNSIGNED__",
409 sizeof ("__CHAR_UNSIGNED__")-1, -1)
410 || ((result >> (num_bits - 1)) & 1) == 0)
412 = result & ((unsigned long) ~0 >> (HOST_BITS_PER_LONG - num_bits));
415 = result | ~((unsigned long) ~0 >> (HOST_BITS_PER_LONG - num_bits));
419 #ifdef MULTIBYTE_CHARS
420 /* Set the initial shift state and convert the next sequence. */
422 /* In all locales L'\0' is zero and mbtowc will return zero,
425 || (num_chars == 1 && token_buffer[0] != '\0'))
428 (void) mbtowc (NULL_PTR, NULL_PTR, 0);
429 if (mbtowc (& wc, token_buffer, num_chars) == num_chars)
432 cpp_warning (pfile,"Ignoring invalid multibyte character");
439 /* This is always a signed type. */
446 return parse_number (pfile, "0", 0);
449 /* See if it is a special token of length 2. */
450 if (tok_start + 2 == tok_end)
452 for (toktab = tokentab2; toktab->operator != NULL; toktab++)
453 if (tok_start[0] == toktab->operator[0]
454 && tok_start[1] == toktab->operator[1])
456 if (toktab->token == ERROR)
458 char *buf = (char *) alloca (40);
459 sprintf (buf, "`%s' not allowed in operand of `#if'", tok_start);
460 cpp_error (pfile, buf);
462 op.op = toktab->token;
473 /* Parse a C escape sequence. STRING_PTR points to a variable
474 containing a pointer to the string to parse. That pointer
475 is updated past the characters we use. The value of the
476 escape sequence is returned.
478 A negative value means the sequence \ newline was seen,
479 which is supposed to be equivalent to nothing at all.
481 If \ is followed by a null character, we return a negative
482 value and leave the string pointer pointing at the null character.
484 If \ is followed by 000, we return 0 and leave the string pointer
485 after the zeros. A value of 0 does not mean end of string. */
488 cpp_parse_escape (pfile, string_ptr)
492 register int c = *(*string_ptr)++;
501 if (CPP_PEDANTIC (pfile))
502 cpp_pedwarn (pfile, "non-ANSI-standard escape sequence, `\\%c'", c);
507 return TARGET_NEWLINE;
529 register int i = c - '0';
530 register int count = 0;
533 c = *(*string_ptr)++;
534 if (c >= '0' && c <= '7')
535 i = (i << 3) + c - '0';
542 if ((i & ~((1 << MAX_CHAR_TYPE_SIZE) - 1)) != 0)
544 i &= (1 << MAX_CHAR_TYPE_SIZE) - 1;
546 "octal character constant does not fit in a byte");
552 register unsigned i = 0, overflow = 0, digits_found = 0, digit;
555 c = *(*string_ptr)++;
556 if (c >= '0' && c <= '9')
558 else if (c >= 'a' && c <= 'f')
559 digit = c - 'a' + 10;
560 else if (c >= 'A' && c <= 'F')
561 digit = c - 'A' + 10;
567 overflow |= i ^ (i << 4 >> 4);
568 i = (i << 4) + digit;
572 cpp_error (pfile, "\\x used with no following hex digits");
573 if (overflow | (i & ~((1 << BITS_PER_UNIT) - 1)))
575 i &= (1 << BITS_PER_UNIT) - 1;
577 "hex character constant does not fit in a byte");
587 integer_overflow (pfile)
590 if (CPP_PEDANTIC (pfile))
591 cpp_pedwarn (pfile, "integer overflow in preprocessor expression");
595 left_shift (pfile, a, unsignedp, b)
601 if (b >= HOST_BITS_PER_LONG)
603 if (! unsignedp && a != 0)
604 integer_overflow (pfile);
608 return (unsigned long) a << b;
613 integer_overflow (pfile);
619 right_shift (pfile, a, unsignedp, b)
625 if (b >= HOST_BITS_PER_LONG)
626 return unsignedp ? 0 : a >> (HOST_BITS_PER_LONG - 1);
628 return (unsigned long) a >> b;
633 /* These priorities are all even, so we can handle associatively. */
634 #define PAREN_INNER_PRIO 0
636 #define COND_PRIO (COMMA_PRIO+2)
637 #define OROR_PRIO (COND_PRIO+2)
638 #define ANDAND_PRIO (OROR_PRIO+2)
639 #define OR_PRIO (ANDAND_PRIO+2)
640 #define XOR_PRIO (OR_PRIO+2)
641 #define AND_PRIO (XOR_PRIO+2)
642 #define EQUAL_PRIO (AND_PRIO+2)
643 #define LESS_PRIO (EQUAL_PRIO+2)
644 #define SHIFT_PRIO (LESS_PRIO+2)
645 #define PLUS_PRIO (SHIFT_PRIO+2)
646 #define MUL_PRIO (PLUS_PRIO+2)
647 #define UNARY_PRIO (MUL_PRIO+2)
648 #define PAREN_OUTER_PRIO (UNARY_PRIO+2)
650 #define COMPARE(OP) \
652 top->value = (unsigned1 || unsigned2) ? (unsigned long) v1 OP v2 : (v1 OP v2)
654 /* Parse and evaluate a C expression, reading from PFILE.
655 Returns the value of the expression. */
658 cpp_parse_expr (pfile)
661 /* The implementation is an operator precedence parser,
662 i.e. a bottom-up parser, using a stack for not-yet-reduced tokens.
664 The stack base is 'stack', and the current stack pointer is 'top'.
665 There is a stack element for each operator (only),
666 and the most recently pushed operator is 'top->op'.
667 An operand (value) is stored in the 'value' field of the stack
668 element of the operator that precedes it.
669 In that case the 'flags' field has the HAVE_VALUE flag set. */
671 #define INIT_STACK_SIZE 20
672 struct operation init_stack[INIT_STACK_SIZE];
673 struct operation *stack = init_stack;
674 struct operation *limit = stack + INIT_STACK_SIZE;
675 register struct operation *top = stack;
677 int skip_evaluation = 0;
687 op = cpp_lex (pfile);
689 /* See if the token is an operand, in which case go to set_value.
690 If the token is an operator, figure out its left and right
691 priorities, and then goto maybe_reduce. */
696 top->value = 0, top->unsignedp = 0;
699 top->value = op.value;
700 top->unsignedp = op.unsignedp;
703 lprio = 0; goto maybe_reduce;
705 /* Is this correct if unary ? FIXME */
706 flags = RIGHT_OPERAND_REQUIRED;
707 lprio = PLUS_PRIO; rprio = lprio + 1; goto maybe_reduce;
709 flags = RIGHT_OPERAND_REQUIRED;
710 rprio = UNARY_PRIO; lprio = rprio + 1; goto maybe_reduce;
711 case '*': case '/': case '%':
712 lprio = MUL_PRIO; goto binop;
713 case '<': case '>': case LEQ: case GEQ:
714 lprio = LESS_PRIO; goto binop;
715 case EQUAL: case NOTEQUAL:
716 lprio = EQUAL_PRIO; goto binop;
718 lprio = SHIFT_PRIO; goto binop;
719 case '&': lprio = AND_PRIO; goto binop;
720 case '^': lprio = XOR_PRIO; goto binop;
721 case '|': lprio = OR_PRIO; goto binop;
722 case ANDAND: lprio = ANDAND_PRIO; goto binop;
723 case OROR: lprio = OROR_PRIO; goto binop;
725 lprio = COMMA_PRIO; goto binop;
727 lprio = PAREN_OUTER_PRIO; rprio = PAREN_INNER_PRIO;
730 lprio = PAREN_INNER_PRIO; rprio = PAREN_OUTER_PRIO;
733 lprio = COND_PRIO; rprio = COND_PRIO;
736 lprio = COND_PRIO + 1; rprio = COND_PRIO;
739 flags = LEFT_OPERAND_REQUIRED|RIGHT_OPERAND_REQUIRED;
743 cpp_error (pfile, "invalid character in #if");
748 /* Push a value onto the stack. */
749 if (top->flags & HAVE_VALUE)
751 cpp_error (pfile, "syntax error in #if");
754 top->flags |= HAVE_VALUE;
758 /* Push an operator, and check if we can reduce now. */
759 while (top->rprio > lprio)
761 long v1 = top[-1].value, v2 = top[0].value;
762 int unsigned1 = top[-1].unsignedp, unsigned2 = top[0].unsignedp;
764 if ((top[1].flags & LEFT_OPERAND_REQUIRED)
765 && ! (top[0].flags & HAVE_VALUE))
767 cpp_error (pfile, "syntax error - missing left operand");
770 if ((top[1].flags & RIGHT_OPERAND_REQUIRED)
771 && ! (top[1].flags & HAVE_VALUE))
773 cpp_error (pfile, "syntax error - missing right operand");
776 /* top[0].value = (top[1].op)(v1, v2);*/
780 if (!(top->flags & HAVE_VALUE))
783 top->unsignedp = unsigned2;
784 top->flags |= HAVE_VALUE;
788 top->value = v1 + v2;
789 top->unsignedp = unsigned1 || unsigned2;
790 if (! top->unsignedp && ! skip_evaluation
791 && ! possible_sum_sign (v1, v2, top->value))
792 integer_overflow (pfile);
796 if (skip_evaluation) ; /* do nothing */
797 else if (!(top->flags & HAVE_VALUE))
800 if ((top->value & v2) < 0 && ! unsigned2)
801 integer_overflow (pfile);
802 top->unsignedp = unsigned2;
803 top->flags |= HAVE_VALUE;
807 top->value = v1 - v2;
808 top->unsignedp = unsigned1 || unsigned2;
810 && ! possible_sum_sign (top->value, v2, v1))
811 integer_overflow (pfile);
815 top->unsignedp = unsigned1 || unsigned2;
817 top->value = (unsigned long) v1 * v2;
818 else if (!skip_evaluation)
820 top->value = v1 * v2;
822 && (top->value / v1 != v2
823 || (top->value & v1 & v2) < 0))
824 integer_overflow (pfile);
832 cpp_error (pfile, "division by zero in #if");
835 top->unsignedp = unsigned1 || unsigned2;
837 top->value = (unsigned long) v1 / v2;
840 top->value = v1 / v2;
841 if ((top->value & v1 & v2) < 0)
842 integer_overflow (pfile);
850 cpp_error (pfile, "division by zero in #if");
853 top->unsignedp = unsigned1 || unsigned2;
855 top->value = (unsigned long) v1 % v2;
857 top->value = v1 % v2;
860 if (top->flags & HAVE_VALUE)
862 cpp_error (pfile, "syntax error");
867 top->flags |= HAVE_VALUE;
870 if (top->flags & HAVE_VALUE)
872 cpp_error (pfile, "syntax error");
876 top->unsignedp = unsigned2;
877 top->flags |= HAVE_VALUE;
879 case '<': COMPARE(<); break;
880 case '>': COMPARE(>); break;
881 case LEQ: COMPARE(<=); break;
882 case GEQ: COMPARE(>=); break;
884 top->value = (v1 == v2);
888 top->value = (v1 != v2);
894 top->unsignedp = unsigned1;
895 if (v2 < 0 && ! unsigned2)
896 top->value = right_shift (pfile, v1, unsigned1, -v2);
898 top->value = left_shift (pfile, v1, unsigned1, v2);
903 top->unsignedp = unsigned1;
904 if (v2 < 0 && ! unsigned2)
905 top->value = left_shift (pfile, v1, unsigned1, -v2);
907 top->value = right_shift (pfile, v1, unsigned1, v2);
909 #define LOGICAL(OP) \
910 top->value = v1 OP v2;\
911 top->unsignedp = unsigned1 || unsigned2;
912 case '&': LOGICAL(&); break;
913 case '^': LOGICAL(^); break;
914 case '|': LOGICAL(|); break;
916 top->value = v1 && v2; top->unsignedp = 0;
917 if (!v1) skip_evaluation--;
920 top->value = v1 || v2; top->unsignedp = 0;
921 if (v1) skip_evaluation--;
924 if (CPP_PEDANTIC (pfile))
925 cpp_pedwarn (pfile, "comma operator in operand of `#if'");
927 top->unsignedp = unsigned2;
930 cpp_error (pfile, "syntax error in #if");
933 if (top[0].op != '?')
936 "syntax error ':' without preceding '?'");
939 else if (! (top[1].flags & HAVE_VALUE)
940 || !(top[-1].flags & HAVE_VALUE)
941 || !(top[0].flags & HAVE_VALUE))
943 cpp_error (pfile, "bad syntax for ?: operator");
949 if (top->value) skip_evaluation--;
950 top->value = top->value ? v1 : v2;
951 top->unsignedp = unsigned1 || unsigned2;
955 if ((top[1].flags & HAVE_VALUE)
956 || ! (top[0].flags & HAVE_VALUE)
958 || (top[-1].flags & HAVE_VALUE))
960 cpp_error (pfile, "mismatched parentheses in #if");
967 top->unsignedp = unsigned1;
968 top->flags |= HAVE_VALUE;
973 top[1].op >= ' ' && top[1].op <= '~'
974 ? "unimplemented operator '%c'\n"
975 : "unimplemented operator '\\%03o'\n",
982 cpp_error (pfile, "internal error in #if expression");
983 if (stack != init_stack)
989 /* Check for and handle stack overflow. */
992 struct operation *new_stack;
993 int old_size = (char*)limit - (char*)stack;
994 int new_size = 2 * old_size;
995 if (stack != init_stack)
996 new_stack = (struct operation*) xrealloc (stack, new_size);
999 new_stack = (struct operation*) xmalloc (new_size);
1000 bcopy ((char *) stack, (char *) new_stack, old_size);
1003 top = (struct operation*)((char*) new_stack + old_size);
1004 limit = (struct operation*)((char*) new_stack + new_size);
1010 if ((op.op == OROR && top[-1].value)
1011 || (op.op == ANDAND && !top[-1].value)
1012 || (op.op == '?' && !top[-1].value))
1016 else if (op.op == ':')
1018 if (top[-2].value) /* Was condition true? */
1025 if (stack != init_stack)
1027 skip_rest_of_line (pfile);