1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 88, 89, 92, 94, 95, 1996 Free Software Foundation, Inc.
4 This file is part of GNU CC.
6 GNU CC is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2, or (at your option)
11 GNU CC is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with GNU CC; see the file COPYING. If not, write to
18 the Free Software Foundation, 59 Temple Place - Suite 330,
19 Boston, MA 02111-1307, USA. */
38 #ifdef MULTIBYTE_CHARS
47 /* The elements of `ridpointers' are identifier nodes
48 for the reserved type names and storage classes.
49 It is indexed by a RID_... value. */
50 tree ridpointers[(int) RID_MAX];
52 /* Cause the `yydebug' variable to be defined. */
55 /* the declaration found for the last IDENTIFIER token read in.
56 yylex must look this up to detect typedefs, which get token type TYPENAME,
57 so it is left around in case the identifier is not a typedef but is
58 used in a context which makes it a reference to a variable. */
61 /* Nonzero enables objc features. */
65 extern tree is_class_name ();
69 /* File used for outputting assembler code. */
70 extern FILE *asm_out_file;
72 #ifndef WCHAR_TYPE_SIZE
74 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
76 #define WCHAR_TYPE_SIZE BITS_PER_WORD
80 /* Number of bytes in a wide character. */
81 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
83 static int maxtoken; /* Current nominal length of token buffer. */
84 char *token_buffer; /* Pointer to token buffer.
85 Actual allocated length is maxtoken + 2.
86 This is not static because objc-parse.y uses it. */
88 /* Nonzero if end-of-file has been seen on input. */
89 static int end_of_file;
91 /* Buffered-back input character; faster than using ungetc. */
92 static int nextchar = -1;
96 /* Do not insert generated code into the source, instead, include it.
97 This allows us to build gcc automatically even for targets that
98 need to add or modify the reserved keyword lists. */
101 /* Return something to represent absolute declarators containing a *.
102 TARGET is the absolute declarator that the * contains.
103 TYPE_QUALS is a list of modifiers such as const or volatile
104 to apply to the pointer type, represented as identifiers.
106 We return an INDIRECT_REF whose "contents" are TARGET
107 and whose type is the modifier list. */
110 make_pointer_declarator (type_quals, target)
111 tree type_quals, target;
113 return build1 (INDIRECT_REF, type_quals, target);
117 forget_protocol_qualifiers ()
119 int i, n = sizeof wordlist / sizeof (struct resword);
121 for (i = 0; i < n; i++)
122 if ((int) wordlist[i].rid >= (int) RID_IN
123 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
124 wordlist[i].name = "";
128 remember_protocol_qualifiers ()
130 int i, n = sizeof wordlist / sizeof (struct resword);
132 for (i = 0; i < n; i++)
133 if (wordlist[i].rid == RID_IN)
134 wordlist[i].name = "in";
135 else if (wordlist[i].rid == RID_OUT)
136 wordlist[i].name = "out";
137 else if (wordlist[i].rid == RID_INOUT)
138 wordlist[i].name = "inout";
139 else if (wordlist[i].rid == RID_BYCOPY)
140 wordlist[i].name = "bycopy";
141 else if (wordlist[i].rid == RID_ONEWAY)
142 wordlist[i].name = "oneway";
148 /* Make identifier nodes long enough for the language-specific slots. */
149 set_identifier_size (sizeof (struct lang_identifier));
151 /* Start it at 0, because check_newline is called at the very beginning
152 and will increment it to 1. */
155 #ifdef MULTIBYTE_CHARS
156 /* Change to the native locale for multibyte conversions. */
157 setlocale (LC_CTYPE, "");
161 token_buffer = (char *) xmalloc (maxtoken + 2);
163 ridpointers[(int) RID_INT] = get_identifier ("int");
164 ridpointers[(int) RID_CHAR] = get_identifier ("char");
165 ridpointers[(int) RID_VOID] = get_identifier ("void");
166 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
167 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
168 ridpointers[(int) RID_SHORT] = get_identifier ("short");
169 ridpointers[(int) RID_LONG] = get_identifier ("long");
170 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
171 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
172 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
173 ridpointers[(int) RID_CONST] = get_identifier ("const");
174 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
175 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
176 ridpointers[(int) RID_STATIC] = get_identifier ("static");
177 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
178 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
179 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
180 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
181 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
182 ridpointers[(int) RID_ID] = get_identifier ("id");
183 ridpointers[(int) RID_IN] = get_identifier ("in");
184 ridpointers[(int) RID_OUT] = get_identifier ("out");
185 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
186 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
187 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
188 forget_protocol_qualifiers();
190 /* Some options inhibit certain reserved words.
191 Clear those words out of the hash table so they won't be recognized. */
192 #define UNSET_RESERVED_WORD(STRING) \
193 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
194 if (s) s->name = ""; } while (0)
196 if (! doing_objc_thang)
197 UNSET_RESERVED_WORD ("id");
199 if (flag_traditional)
201 UNSET_RESERVED_WORD ("const");
202 UNSET_RESERVED_WORD ("volatile");
203 UNSET_RESERVED_WORD ("typeof");
204 UNSET_RESERVED_WORD ("signed");
205 UNSET_RESERVED_WORD ("inline");
206 UNSET_RESERVED_WORD ("iterator");
207 UNSET_RESERVED_WORD ("complex");
211 UNSET_RESERVED_WORD ("asm");
212 UNSET_RESERVED_WORD ("typeof");
213 UNSET_RESERVED_WORD ("inline");
214 UNSET_RESERVED_WORD ("iterator");
215 UNSET_RESERVED_WORD ("complex");
220 reinit_parse_for_function ()
224 /* Function used when yydebug is set, to print a token in more detail. */
227 yyprint (file, yychar, yylval)
239 if (IDENTIFIER_POINTER (t))
240 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
245 if (TREE_CODE (t) == INTEGER_CST)
247 #if HOST_BITS_PER_WIDE_INT == 64
248 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
254 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
260 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
266 /* If C is not whitespace, return C.
267 Otherwise skip whitespace and return first nonwhite char read. */
273 static int newline_warning = 0;
279 /* We don't recognize comments here, because
280 cpp output can include / and * consecutively as operators.
281 Also, there's no need, since cpp removes all comments. */
284 c = check_newline ();
296 /* ANSI C says the effects of a carriage return in a source file
298 if (pedantic && !newline_warning)
300 warning ("carriage return in source file");
301 warning ("(we only warn about the first carriage return)");
312 error ("stray '\\' in program");
322 /* Skips all of the white space at the current location in the input file.
323 Must use and reset nextchar if it has the next character. */
326 position_after_white_space ()
331 c = nextchar, nextchar = -1;
335 ungetc (skip_white_space (c), finput);
338 /* Make the token buffer longer, preserving the data in it.
339 P should point to just beyond the last valid character in the old buffer.
340 The value we return is a pointer to the new buffer
341 at a place corresponding to P. */
344 extend_token_buffer (p)
347 int offset = p - token_buffer;
349 maxtoken = maxtoken * 2 + 10;
350 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
352 return token_buffer + offset;
355 /* At the beginning of a line, increment the line number
356 and process any #-directive on this line.
357 If the line is a #-directive, read the entire line and return a newline.
358 Otherwise, return the line's first non-whitespace character. */
368 /* Read first nonwhite char on the line. */
371 while (c == ' ' || c == '\t')
376 /* If not #, return it so caller will use it. */
380 /* Read first nonwhite char after the `#'. */
383 while (c == ' ' || c == '\t')
386 /* If a letter follows, then if the word here is `line', skip
387 it and ignore it; otherwise, ignore the line, with an error
388 if the word isn't `pragma', `ident', `define', or `undef'. */
390 if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
394 if (getc (finput) == 'r'
395 && getc (finput) == 'a'
396 && getc (finput) == 'g'
397 && getc (finput) == 'm'
398 && getc (finput) == 'a'
399 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
401 while (c == ' ' || c == '\t')
405 #ifdef HANDLE_SYSV_PRAGMA
408 if (token != IDENTIFIER)
410 return handle_sysv_pragma (finput, token);
411 #else /* !HANDLE_SYSV_PRAGMA */
415 if (token != IDENTIFIER)
417 if (HANDLE_PRAGMA (finput, yylval.ttype))
422 #endif /* HANDLE_PRAGMA */
423 #endif /* !HANDLE_SYSV_PRAGMA */
430 if (getc (finput) == 'e'
431 && getc (finput) == 'f'
432 && getc (finput) == 'i'
433 && getc (finput) == 'n'
434 && getc (finput) == 'e'
435 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
437 #ifdef DWARF_DEBUGGING_INFO
439 && (debug_info_level == DINFO_LEVEL_VERBOSE)
440 && (write_symbols == DWARF_DEBUG))
441 dwarfout_define (lineno, get_directive_line (finput));
442 #endif /* DWARF_DEBUGGING_INFO */
448 if (getc (finput) == 'n'
449 && getc (finput) == 'd'
450 && getc (finput) == 'e'
451 && getc (finput) == 'f'
452 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
454 #ifdef DWARF_DEBUGGING_INFO
456 && (debug_info_level == DINFO_LEVEL_VERBOSE)
457 && (write_symbols == DWARF_DEBUG))
458 dwarfout_undef (lineno, get_directive_line (finput));
459 #endif /* DWARF_DEBUGGING_INFO */
465 if (getc (finput) == 'i'
466 && getc (finput) == 'n'
467 && getc (finput) == 'e'
468 && ((c = getc (finput)) == ' ' || c == '\t'))
473 if (getc (finput) == 'd'
474 && getc (finput) == 'e'
475 && getc (finput) == 'n'
476 && getc (finput) == 't'
477 && ((c = getc (finput)) == ' ' || c == '\t'))
479 /* #ident. The pedantic warning is now in cccp.c. */
481 /* Here we have just seen `#ident '.
482 A string constant should follow. */
484 while (c == ' ' || c == '\t')
487 /* If no argument, ignore the line. */
494 || TREE_CODE (yylval.ttype) != STRING_CST)
496 error ("invalid #ident");
502 #ifdef ASM_OUTPUT_IDENT
503 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
507 /* Skip the rest of this line. */
512 error ("undefined or invalid # directive");
517 /* Here we have either `#line' or `# <nonletter>'.
518 In either case, it should be a line number; a digit should follow. */
520 while (c == ' ' || c == '\t')
523 /* If the # is the only nonwhite char on the line,
524 just ignore it. Check the new newline. */
528 /* Something follows the #; read a token. */
533 if (token == CONSTANT
534 && TREE_CODE (yylval.ttype) == INTEGER_CST)
536 int old_lineno = lineno;
538 /* subtract one, because it is the following line that
539 gets the specified number */
541 int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
543 /* Is this the last nonwhite stuff on the line? */
545 while (c == ' ' || c == '\t')
549 /* No more: store the line number and check following line. */
555 /* More follows: it must be a string constant (filename). */
557 /* Read the string constant. */
560 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
562 error ("invalid #line");
567 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
568 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
571 /* Each change of file name
572 reinitializes whether we are now in a system header. */
573 in_system_header = 0;
575 if (main_input_filename == 0)
576 main_input_filename = input_filename;
578 /* Is this the last nonwhite stuff on the line? */
580 while (c == ' ' || c == '\t')
584 /* Update the name in the top element of input_file_stack. */
585 if (input_file_stack)
586 input_file_stack->name = input_filename;
595 /* `1' after file name means entering new file.
596 `2' after file name means just left a file. */
598 if (token == CONSTANT
599 && TREE_CODE (yylval.ttype) == INTEGER_CST)
601 if (TREE_INT_CST_LOW (yylval.ttype) == 1)
603 /* Pushing to a new file. */
605 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
606 input_file_stack->line = old_lineno;
607 p->next = input_file_stack;
608 p->name = input_filename;
609 input_file_stack = p;
610 input_file_stack_tick++;
611 #ifdef DBX_DEBUGGING_INFO
612 if (write_symbols == DBX_DEBUG)
613 dbxout_start_new_source_file (input_filename);
615 #ifdef DWARF_DEBUGGING_INFO
616 if (debug_info_level == DINFO_LEVEL_VERBOSE
617 && write_symbols == DWARF_DEBUG)
618 dwarfout_start_new_source_file (input_filename);
619 #endif /* DWARF_DEBUGGING_INFO */
623 else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
625 /* Popping out of a file. */
626 if (input_file_stack->next)
628 struct file_stack *p = input_file_stack;
629 input_file_stack = p->next;
631 input_file_stack_tick++;
632 #ifdef DBX_DEBUGGING_INFO
633 if (write_symbols == DBX_DEBUG)
634 dbxout_resume_previous_source_file ();
636 #ifdef DWARF_DEBUGGING_INFO
637 if (debug_info_level == DINFO_LEVEL_VERBOSE
638 && write_symbols == DWARF_DEBUG)
639 dwarfout_resume_previous_source_file (input_file_stack->line);
640 #endif /* DWARF_DEBUGGING_INFO */
643 error ("#-lines for entering and leaving files don't match");
649 /* Now that we've pushed or popped the input stack,
650 update the name in the top element. */
651 if (input_file_stack)
652 input_file_stack->name = input_filename;
654 /* If we have handled a `1' or a `2',
655 see if there is another number to read. */
658 /* Is this the last nonwhite stuff on the line? */
660 while (c == ' ' || c == '\t')
670 /* `3' after file name means this is a system header file. */
672 if (token == CONSTANT
673 && TREE_CODE (yylval.ttype) == INTEGER_CST
674 && TREE_INT_CST_LOW (yylval.ttype) == 3)
675 in_system_header = 1, used_up = 1;
679 /* Is this the last nonwhite stuff on the line? */
681 while (c == ' ' || c == '\t')
688 warning ("unrecognized text at end of #line");
691 error ("invalid #-line");
693 /* skip the rest of this line. */
695 while (c != '\n' && c != EOF)
700 #ifdef HANDLE_SYSV_PRAGMA
702 /* Handle a #pragma directive. INPUT is the current input stream,
703 and TOKEN is the token we read after `#pragma'. Processes the entire input
704 line and returns a character for the caller to reread: either \n or EOF. */
706 /* This function has to be in this file, in order to get at
710 handle_sysv_pragma (input, token)
724 handle_pragma_token (token_buffer, yylval.ttype);
727 handle_pragma_token (token_buffer, 0);
731 c = nextchar, nextchar = -1;
735 while (c == ' ' || c == '\t')
737 if (c == '\n' || c == EOF)
739 handle_pragma_token (0, 0);
747 #endif /* HANDLE_SYSV_PRAGMA */
749 #define ENDFILE -1 /* token that represents end-of-file */
751 /* Read an escape sequence, returning its equivalent as a character,
752 or store 1 in *ignore_ptr if it is backslash-newline. */
755 readescape (ignore_ptr)
758 register int c = getc (finput);
760 register unsigned count;
761 unsigned firstdig = 0;
767 if (warn_traditional)
768 warning ("the meaning of `\\x' varies with -traditional");
770 if (flag_traditional)
779 if (!(c >= 'a' && c <= 'f')
780 && !(c >= 'A' && c <= 'F')
781 && !(c >= '0' && c <= '9'))
787 if (c >= 'a' && c <= 'f')
788 code += c - 'a' + 10;
789 if (c >= 'A' && c <= 'F')
790 code += c - 'A' + 10;
791 if (c >= '0' && c <= '9')
793 if (code != 0 || count != 0)
802 error ("\\x used with no following hex digits");
804 /* Digits are all 0's. Ok. */
806 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
808 && ((1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
810 pedwarn ("hex escape out of range");
813 case '0': case '1': case '2': case '3': case '4':
814 case '5': case '6': case '7':
817 while ((c <= '7') && (c >= '0') && (count++ < 3))
819 code = (code * 8) + (c - '0');
825 case '\\': case '\'': case '"':
834 return TARGET_NEWLINE;
849 if (warn_traditional)
850 warning ("the meaning of `\\a' varies with -traditional");
852 if (flag_traditional)
857 #if 0 /* Vertical tab is present in common usage compilers. */
858 if (flag_traditional)
866 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
872 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
876 /* `\%' is used to prevent SCCS from getting confused. */
879 pedwarn ("non-ANSI escape sequence `\\%c'", c);
882 if (c >= 040 && c < 0177)
883 pedwarn ("unknown escape sequence `\\%c'", c);
885 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
895 strcpy (buf, string);
897 /* We can't print string and character constants well
898 because the token_buffer contains the result of processing escapes. */
900 strcat (buf, " at end of input");
901 else if (token_buffer[0] == 0)
902 strcat (buf, " at null character");
903 else if (token_buffer[0] == '"')
904 strcat (buf, " before string constant");
905 else if (token_buffer[0] == '\'')
906 strcat (buf, " before character constant");
907 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
908 sprintf (buf + strlen (buf), " before character 0%o",
909 (unsigned char) token_buffer[0]);
911 strcat (buf, " before `%s'");
913 error (buf, token_buffer);
926 struct try_type type_sequence[] =
928 { &integer_type_node, 0, 0, 0},
929 { &unsigned_type_node, 1, 0, 0},
930 { &long_integer_type_node, 0, 1, 0},
931 { &long_unsigned_type_node, 1, 1, 0},
932 { &long_long_integer_type_node, 0, 1, 1},
933 { &long_long_unsigned_type_node, 1, 1, 1}
947 c = nextchar, nextchar = -1;
951 /* Effectively do c = skip_white_space (c)
952 but do it faster in the usual cases. */
965 /* Call skip_white_space so we can warn if appropriate. */
970 c = skip_white_space (c);
979 /* yylloc.first_line = lineno; */
990 /* Capital L may start a wide-string or wide-character constant. */
992 register int c = getc (finput);
1001 goto string_constant;
1008 if (!doing_objc_thang)
1015 /* '@' may start a constant string object. */
1016 register int c = getc(finput);
1020 goto string_constant;
1023 /* Fall through to treat '@' as the start of an identifier. */
1026 case 'A': case 'B': case 'C': case 'D': case 'E':
1027 case 'F': case 'G': case 'H': case 'I': case 'J':
1028 case 'K': case 'M': case 'N': case 'O':
1029 case 'P': case 'Q': case 'R': case 'S': case 'T':
1030 case 'U': case 'V': case 'W': case 'X': case 'Y':
1032 case 'a': case 'b': case 'c': case 'd': case 'e':
1033 case 'f': case 'g': case 'h': case 'i': case 'j':
1034 case 'k': case 'l': case 'm': case 'n': case 'o':
1035 case 'p': case 'q': case 'r': case 's': case 't':
1036 case 'u': case 'v': case 'w': case 'x': case 'y':
1042 while (isalnum (c) || c == '_' || c == '$' || c == '@')
1044 /* Make sure this char really belongs in an identifier. */
1045 if (c == '@' && ! doing_objc_thang)
1049 if (! dollars_in_ident)
1050 error ("`$' in identifier");
1052 pedwarn ("`$' in identifier");
1055 if (p >= token_buffer + maxtoken)
1056 p = extend_token_buffer (p);
1068 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1071 register struct resword *ptr;
1073 if (ptr = is_reserved_word (token_buffer, p - token_buffer))
1076 yylval.ttype = ridpointers[(int) ptr->rid];
1077 value = (int) ptr->token;
1079 /* Only return OBJECTNAME if it is a typedef. */
1080 if (doing_objc_thang && value == OBJECTNAME)
1082 lastiddecl = lookup_name(yylval.ttype);
1084 if (lastiddecl == NULL_TREE
1085 || TREE_CODE (lastiddecl) != TYPE_DECL)
1089 /* Even if we decided to recognize asm, still perhaps warn. */
1091 && (value == ASM_KEYWORD || value == TYPEOF
1092 || ptr->rid == RID_INLINE)
1093 && token_buffer[0] != '_')
1094 pedwarn ("ANSI does not permit the keyword `%s'",
1099 /* If we did not find a keyword, look for an identifier
1102 if (value == IDENTIFIER)
1104 if (token_buffer[0] == '@')
1105 error("invalid identifier `%s'", token_buffer);
1107 yylval.ttype = get_identifier (token_buffer);
1108 lastiddecl = lookup_name (yylval.ttype);
1110 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1112 /* A user-invisible read-only initialized variable
1113 should be replaced by its value.
1114 We handle only strings since that's the only case used in C. */
1115 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1116 && DECL_IGNORED_P (lastiddecl)
1117 && TREE_READONLY (lastiddecl)
1118 && DECL_INITIAL (lastiddecl) != 0
1119 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1121 tree stringval = DECL_INITIAL (lastiddecl);
1123 /* Copy the string value so that we won't clobber anything
1124 if we put something in the TREE_CHAIN of this one. */
1125 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1126 TREE_STRING_POINTER (stringval));
1129 else if (doing_objc_thang)
1131 tree objc_interface_decl = is_class_name (yylval.ttype);
1133 if (objc_interface_decl)
1136 yylval.ttype = objc_interface_decl;
1146 /* Check first for common special case: single-digit 0 or 1. */
1148 next_c = getc (finput);
1149 ungetc (next_c, finput); /* Always undo this lookahead. */
1150 if (!isalnum (next_c) && next_c != '.')
1152 token_buffer[0] = (char)c, token_buffer[1] = '\0';
1153 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1159 case '2': case '3': case '4':
1160 case '5': case '6': case '7': case '8': case '9':
1165 int largest_digit = 0;
1167 /* for multi-precision arithmetic,
1168 we actually store only HOST_BITS_PER_CHAR bits in each part.
1169 The number of parts is chosen so as to be sufficient to hold
1170 the enough bits to fit into the two HOST_WIDE_INTs that contain
1171 the integer value (this is always at least as many bits as are
1172 in a target `long long' value, but may be wider). */
1173 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1174 int parts[TOTAL_PARTS];
1177 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag
1180 for (count = 0; count < TOTAL_PARTS; count++)
1188 *p++ = (c = getc (finput));
1189 if ((c == 'x') || (c == 'X'))
1192 *p++ = (c = getc (finput));
1194 /* Leading 0 forces octal unless the 0 is the only digit. */
1195 else if (c >= '0' && c <= '9')
1204 /* Read all the digits-and-decimal-points. */
1207 || (isalnum (c) && c != 'l' && c != 'L'
1208 && c != 'u' && c != 'U'
1209 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1210 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1215 error ("floating constant may not be in radix 16");
1216 if (floatflag == TOO_MANY_POINTS)
1217 /* We have already emitted an error. Don't need another. */
1219 else if (floatflag == AFTER_POINT)
1221 error ("malformed floating constant");
1222 floatflag = TOO_MANY_POINTS;
1223 /* Avoid another error from atof by forcing all characters
1224 from here on to be ignored. */
1228 floatflag = AFTER_POINT;
1231 *p++ = c = getc (finput);
1232 /* Accept '.' as the start of a floating-point number
1233 only when it is followed by a digit.
1234 Otherwise, unread the following non-digit
1235 and use the '.' as a structural token. */
1236 if (p == token_buffer + 2 && !isdigit (c))
1247 error ("parse error at `..'");
1250 token_buffer[1] = 0;
1257 /* It is not a decimal point.
1258 It should be a digit (perhaps a hex digit). */
1264 else if (base <= 10)
1266 if (c == 'e' || c == 'E')
1269 floatflag = AFTER_POINT;
1270 break; /* start of exponent */
1272 error ("nondigits in number and not hexadecimal");
1283 if (c >= largest_digit)
1287 for (count = 0; count < TOTAL_PARTS; count++)
1289 parts[count] *= base;
1293 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1295 &= (1 << HOST_BITS_PER_CHAR) - 1;
1301 /* If the extra highest-order part ever gets anything in it,
1302 the number is certainly too big. */
1303 if (parts[TOTAL_PARTS - 1] != 0)
1306 if (p >= token_buffer + maxtoken - 3)
1307 p = extend_token_buffer (p);
1308 *p++ = (c = getc (finput));
1313 error ("numeric constant with no digits");
1315 if (largest_digit >= base)
1316 error ("numeric constant contains digits beyond the radix");
1318 /* Remove terminating char from the token buffer and delimit the string */
1321 if (floatflag != NOT_FLOAT)
1323 tree type = double_type_node;
1324 int exceeds_double = 0;
1326 REAL_VALUE_TYPE value;
1329 /* Read explicit exponent if any, and put it in tokenbuf. */
1331 if ((c == 'e') || (c == 'E'))
1333 if (p >= token_buffer + maxtoken - 3)
1334 p = extend_token_buffer (p);
1337 if ((c == '+') || (c == '-'))
1343 error ("floating constant exponent has no digits");
1346 if (p >= token_buffer + maxtoken - 3)
1347 p = extend_token_buffer (p);
1356 /* Convert string to a double, checking for overflow. */
1357 if (setjmp (handler))
1359 error ("floating constant out of range");
1364 int fflag = 0, lflag = 0;
1365 /* Copy token_buffer now, while it has just the number
1366 and not the suffixes; once we add `f' or `i',
1367 REAL_VALUE_ATOF may not work any more. */
1368 char *copy = (char *) alloca (p - token_buffer + 1);
1369 bcopy (token_buffer, copy, p - token_buffer + 1);
1371 set_float_handler (handler);
1377 /* Read the suffixes to choose a data type. */
1382 error ("more than one `f' in numeric constant");
1388 error ("more than one `l' in numeric constant");
1394 error ("more than one `i' or `j' in numeric constant");
1396 pedwarn ("ANSI C forbids imaginary numeric constants");
1407 if (p >= token_buffer + maxtoken - 3)
1408 p = extend_token_buffer (p);
1414 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1415 tells the desired precision of the binary result
1416 of decimal-to-binary conversion. */
1421 error ("both `f' and `l' in floating constant");
1423 type = float_type_node;
1424 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1425 /* A diagnostic is required here by some ANSI C testsuites.
1426 This is not pedwarn, become some people don't want
1427 an error for this. */
1428 if (REAL_VALUE_ISINF (value) && pedantic)
1429 warning ("floating point number exceeds range of `float'");
1433 type = long_double_type_node;
1434 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1435 if (REAL_VALUE_ISINF (value) && pedantic)
1436 warning ("floating point number exceeds range of `long double'");
1440 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1441 if (REAL_VALUE_ISINF (value) && pedantic)
1442 warning ("floating point number exceeds range of `double'");
1445 set_float_handler (NULL_PTR);
1448 if (errno == ERANGE && !flag_traditional && pedantic)
1450 /* ERANGE is also reported for underflow,
1451 so test the value to distinguish overflow from that. */
1452 if (REAL_VALUES_LESS (dconst1, value)
1453 || REAL_VALUES_LESS (value, dconstm1))
1455 warning ("floating point number exceeds range of `double'");
1461 /* If the result is not a number, assume it must have been
1462 due to some error message above, so silently convert
1464 if (REAL_VALUE_ISNAN (value))
1467 /* Create a node with determined type and value. */
1469 yylval.ttype = build_complex (NULL_TREE,
1470 convert (type, integer_zero_node),
1471 build_real (type, value));
1473 yylval.ttype = build_real (type, value);
1477 tree traditional_type, ansi_type, type;
1478 HOST_WIDE_INT high, low;
1479 int spec_unsigned = 0;
1481 int spec_long_long = 0;
1487 if (c == 'u' || c == 'U')
1490 error ("two `u's in integer constant");
1493 else if (c == 'l' || c == 'L')
1498 error ("three `l's in integer constant");
1500 pedwarn ("ANSI C forbids long long integer constants");
1505 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1508 error ("more than one `i' or `j' in numeric constant");
1510 pedwarn ("ANSI C forbids imaginary numeric constants");
1515 if (p >= token_buffer + maxtoken - 3)
1516 p = extend_token_buffer (p);
1521 /* If the constant is not long long and it won't fit in an
1522 unsigned long, or if the constant is long long and won't fit
1523 in an unsigned long long, then warn that the constant is out
1526 /* ??? This assumes that long long and long integer types are
1527 a multiple of 8 bits. This better than the original code
1528 though which assumed that long was exactly 32 bits and long
1529 long was exactly 64 bits. */
1532 bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1534 bytes = TYPE_PRECISION (long_integer_type_node) / 8;
1537 for (i = bytes; i < TOTAL_PARTS; i++)
1541 pedwarn ("integer constant out of range");
1543 /* This is simplified by the fact that our constant
1544 is always positive. */
1548 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1550 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1551 / HOST_BITS_PER_CHAR)]
1552 << (i * HOST_BITS_PER_CHAR));
1553 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1556 yylval.ttype = build_int_2 (low, high);
1557 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1559 /* If warn_traditional, calculate both the ANSI type and the
1560 traditional type, then see if they disagree.
1561 Otherwise, calculate only the type for the dialect in use. */
1562 if (warn_traditional || flag_traditional)
1564 /* Calculate the traditional type. */
1565 /* Traditionally, any constant is signed;
1566 but if unsigned is specified explicitly, obey that.
1567 Use the smallest size with the right number of bits,
1568 except for one special case with decimal constants. */
1569 if (! spec_long && base != 10
1570 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1571 traditional_type = (spec_unsigned ? unsigned_type_node
1572 : integer_type_node);
1573 /* A decimal constant must be long
1574 if it does not fit in type int.
1575 I think this is independent of whether
1576 the constant is signed. */
1577 else if (! spec_long && base == 10
1578 && int_fits_type_p (yylval.ttype, integer_type_node))
1579 traditional_type = (spec_unsigned ? unsigned_type_node
1580 : integer_type_node);
1581 else if (! spec_long_long)
1582 traditional_type = (spec_unsigned ? long_unsigned_type_node
1583 : long_integer_type_node);
1585 traditional_type = (spec_unsigned
1586 ? long_long_unsigned_type_node
1587 : long_long_integer_type_node);
1589 if (warn_traditional || ! flag_traditional)
1591 /* Calculate the ANSI type. */
1592 if (! spec_long && ! spec_unsigned
1593 && int_fits_type_p (yylval.ttype, integer_type_node))
1594 ansi_type = integer_type_node;
1595 else if (! spec_long && (base != 10 || spec_unsigned)
1596 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1597 ansi_type = unsigned_type_node;
1598 else if (! spec_unsigned && !spec_long_long
1599 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1600 ansi_type = long_integer_type_node;
1601 else if (! spec_long_long)
1602 ansi_type = long_unsigned_type_node;
1603 else if (! spec_unsigned
1604 /* Verify value does not overflow into sign bit. */
1605 && TREE_INT_CST_HIGH (yylval.ttype) >= 0
1606 && int_fits_type_p (yylval.ttype,
1607 long_long_integer_type_node))
1608 ansi_type = long_long_integer_type_node;
1610 ansi_type = long_long_unsigned_type_node;
1613 type = flag_traditional ? traditional_type : ansi_type;
1615 if (warn_traditional && traditional_type != ansi_type)
1617 if (TYPE_PRECISION (traditional_type)
1618 != TYPE_PRECISION (ansi_type))
1619 warning ("width of integer constant changes with -traditional");
1620 else if (TREE_UNSIGNED (traditional_type)
1621 != TREE_UNSIGNED (ansi_type))
1622 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1624 warning ("width of integer constant may change on other systems with -traditional");
1627 if (!flag_traditional && !int_fits_type_p (yylval.ttype, type)
1629 pedwarn ("integer constant out of range");
1631 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1632 warning ("decimal constant is so large that it is unsigned");
1636 if (TYPE_PRECISION (type)
1637 <= TYPE_PRECISION (integer_type_node))
1639 = build_complex (NULL_TREE, integer_zero_node,
1640 convert (integer_type_node,
1643 error ("complex integer constant is too wide for `complex int'");
1645 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1646 /* The traditional constant 0x80000000 is signed
1647 but doesn't fit in the range of int.
1648 This will change it to -0x80000000, which does fit. */
1650 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1651 yylval.ttype = convert (type, yylval.ttype);
1652 TREE_OVERFLOW (yylval.ttype)
1653 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1656 TREE_TYPE (yylval.ttype) = type;
1662 if (isalnum (c) || c == '.' || c == '_' || c == '$'
1663 || (!flag_traditional && (c == '-' || c == '+')
1664 && (p[-1] == 'e' || p[-1] == 'E')))
1665 error ("missing white space after number `%s'", token_buffer);
1667 value = CONSTANT; break;
1673 register int result = 0;
1674 register int num_chars = 0;
1675 unsigned width = TYPE_PRECISION (char_type_node);
1680 width = WCHAR_TYPE_SIZE;
1681 #ifdef MULTIBYTE_CHARS
1682 max_chars = MB_CUR_MAX;
1688 max_chars = TYPE_PRECISION (integer_type_node) / width;
1696 if (c == '\'' || c == EOF)
1702 c = readescape (&ignore);
1705 if (width < HOST_BITS_PER_INT
1706 && (unsigned) c >= (1 << width))
1707 pedwarn ("escape sequence out of range for character");
1708 #ifdef MAP_CHARACTER
1710 c = MAP_CHARACTER (c);
1716 pedwarn ("ANSI C forbids newline in character constant");
1719 #ifdef MAP_CHARACTER
1721 c = MAP_CHARACTER (c);
1725 if (num_chars > maxtoken - 4)
1726 extend_token_buffer (token_buffer);
1728 token_buffer[num_chars] = c;
1730 /* Merge character into result; ignore excess chars. */
1731 if (num_chars < max_chars + 1)
1733 if (width < HOST_BITS_PER_INT)
1734 result = (result << width) | (c & ((1 << width) - 1));
1740 token_buffer[num_chars + 1] = '\'';
1741 token_buffer[num_chars + 2] = 0;
1744 error ("malformatted character constant");
1745 else if (num_chars == 0)
1746 error ("empty character constant");
1747 else if (num_chars > max_chars)
1749 num_chars = max_chars;
1750 error ("character constant too long");
1752 else if (num_chars != 1 && ! flag_traditional)
1753 warning ("multi-character character constant");
1755 /* If char type is signed, sign-extend the constant. */
1758 int num_bits = num_chars * width;
1760 /* We already got an error; avoid invalid shift. */
1761 yylval.ttype = build_int_2 (0, 0);
1762 else if (TREE_UNSIGNED (char_type_node)
1763 || ((result >> (num_bits - 1)) & 1) == 0)
1765 = build_int_2 (result & ((unsigned HOST_WIDE_INT) ~0
1766 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1770 = build_int_2 (result | ~((unsigned HOST_WIDE_INT) ~0
1771 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1773 TREE_TYPE (yylval.ttype) = integer_type_node;
1777 #ifdef MULTIBYTE_CHARS
1778 /* Set the initial shift state and convert the next sequence. */
1780 /* In all locales L'\0' is zero and mbtowc will return zero,
1783 || (num_chars == 1 && token_buffer[1] != '\0'))
1786 (void) mbtowc (NULL_PTR, NULL_PTR, 0);
1787 if (mbtowc (& wc, token_buffer + 1, num_chars) == num_chars)
1790 warning ("Ignoring invalid multibyte character");
1793 yylval.ttype = build_int_2 (result, 0);
1794 TREE_TYPE (yylval.ttype) = wchar_type_node;
1805 p = token_buffer + 1;
1807 while (c != '"' && c >= 0)
1812 c = readescape (&ignore);
1816 && TYPE_PRECISION (char_type_node) < HOST_BITS_PER_INT
1817 && c >= (1 << TYPE_PRECISION (char_type_node)))
1818 pedwarn ("escape sequence out of range for character");
1823 pedwarn ("ANSI C forbids newline in string constant");
1827 if (p == token_buffer + maxtoken)
1828 p = extend_token_buffer (p);
1837 error ("Unterminated string constant");
1839 /* We have read the entire constant.
1840 Construct a STRING_CST for the result. */
1844 /* If this is a L"..." wide-string, convert the multibyte string
1845 to a wide character string. */
1846 char *widep = (char *) alloca ((p - token_buffer) * WCHAR_BYTES);
1849 #ifdef MULTIBYTE_CHARS
1850 len = mbstowcs ((wchar_t *) widep, token_buffer + 1, p - token_buffer);
1851 if (len < 0 || len >= (p - token_buffer))
1853 warning ("Ignoring invalid multibyte string");
1856 bzero (widep + (len * WCHAR_BYTES), WCHAR_BYTES);
1859 union { long l; char c[sizeof (long)]; } u;
1863 /* Determine whether host is little or big endian. */
1865 big_endian = u.c[sizeof (long) - 1];
1866 wp = widep + (big_endian ? WCHAR_BYTES - 1 : 0);
1868 bzero (widep, (p - token_buffer) * WCHAR_BYTES);
1869 for (cp = token_buffer + 1; cp < p; cp++)
1870 *wp = *cp, wp += WCHAR_BYTES;
1871 len = p - token_buffer - 1;
1874 yylval.ttype = build_string ((len + 1) * WCHAR_BYTES, widep);
1875 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
1880 extern tree build_objc_string();
1881 /* Return an Objective-C @"..." constant string object. */
1882 yylval.ttype = build_objc_string (p - token_buffer,
1884 TREE_TYPE (yylval.ttype) = char_array_type_node;
1885 value = OBJC_STRING;
1889 yylval.ttype = build_string (p - token_buffer, token_buffer + 1);
1890 TREE_TYPE (yylval.ttype) = char_array_type_node;
1921 yylval.code = PLUS_EXPR; break;
1923 yylval.code = MINUS_EXPR; break;
1925 yylval.code = BIT_AND_EXPR; break;
1927 yylval.code = BIT_IOR_EXPR; break;
1929 yylval.code = MULT_EXPR; break;
1931 yylval.code = TRUNC_DIV_EXPR; break;
1933 yylval.code = TRUNC_MOD_EXPR; break;
1935 yylval.code = BIT_XOR_EXPR; break;
1937 yylval.code = LSHIFT_EXPR; break;
1939 yylval.code = RSHIFT_EXPR; break;
1941 yylval.code = LT_EXPR; break;
1943 yylval.code = GT_EXPR; break;
1946 token_buffer[1] = c1 = getc (finput);
1947 token_buffer[2] = 0;
1954 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
1956 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
1958 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
1960 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
1962 value = ASSIGN; goto done;
1968 value = PLUSPLUS; goto done;
1970 value = MINUSMINUS; goto done;
1972 value = ANDAND; goto done;
1974 value = OROR; goto done;
1987 { value = POINTSAT; goto done; }
1991 { value = ']'; goto done; }
1995 { value = '{'; goto done; }
1997 { value = '['; goto done; }
2001 { value = '}'; goto done; }
2004 ungetc (c1, finput);
2005 token_buffer[1] = 0;
2007 if ((c == '<') || (c == '>'))
2008 value = ARITHCOMPARE;
2014 /* Don't make yyparse think this is eof. */
2023 /* yylloc.last_line = lineno; */
2028 /* Sets the value of the 'yydebug' variable to VALUE.
2029 This is a function so we don't have to have YYDEBUG defined
2030 in order to build the compiler. */
2039 warning ("YYDEBUG not defined.");