1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 88, 89, 92, 94, 1995 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 #ifdef HANDLE_SYSV_PRAGMA
402 return handle_sysv_pragma (finput, c);
403 #else /* !HANDLE_SYSV_PRAGMA */
405 HANDLE_PRAGMA (finput);
406 #endif /* HANDLE_PRAGMA */
408 #endif /* !HANDLE_SYSV_PRAGMA */
414 if (getc (finput) == 'e'
415 && getc (finput) == 'f'
416 && getc (finput) == 'i'
417 && getc (finput) == 'n'
418 && getc (finput) == 'e'
419 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
421 #ifdef DWARF_DEBUGGING_INFO
422 if ((debug_info_level == DINFO_LEVEL_VERBOSE)
423 && (write_symbols == DWARF_DEBUG))
424 dwarfout_define (lineno, get_directive_line (finput));
425 #endif /* DWARF_DEBUGGING_INFO */
431 if (getc (finput) == 'n'
432 && getc (finput) == 'd'
433 && getc (finput) == 'e'
434 && getc (finput) == 'f'
435 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
437 #ifdef DWARF_DEBUGGING_INFO
438 if ((debug_info_level == DINFO_LEVEL_VERBOSE)
439 && (write_symbols == DWARF_DEBUG))
440 dwarfout_undef (lineno, get_directive_line (finput));
441 #endif /* DWARF_DEBUGGING_INFO */
447 if (getc (finput) == 'i'
448 && getc (finput) == 'n'
449 && getc (finput) == 'e'
450 && ((c = getc (finput)) == ' ' || c == '\t'))
455 if (getc (finput) == 'd'
456 && getc (finput) == 'e'
457 && getc (finput) == 'n'
458 && getc (finput) == 't'
459 && ((c = getc (finput)) == ' ' || c == '\t'))
461 /* #ident. The pedantic warning is now in cccp.c. */
463 /* Here we have just seen `#ident '.
464 A string constant should follow. */
466 while (c == ' ' || c == '\t')
469 /* If no argument, ignore the line. */
476 || TREE_CODE (yylval.ttype) != STRING_CST)
478 error ("invalid #ident");
484 #ifdef ASM_OUTPUT_IDENT
485 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
489 /* Skip the rest of this line. */
494 error ("undefined or invalid # directive");
499 /* Here we have either `#line' or `# <nonletter>'.
500 In either case, it should be a line number; a digit should follow. */
502 while (c == ' ' || c == '\t')
505 /* If the # is the only nonwhite char on the line,
506 just ignore it. Check the new newline. */
510 /* Something follows the #; read a token. */
515 if (token == CONSTANT
516 && TREE_CODE (yylval.ttype) == INTEGER_CST)
518 int old_lineno = lineno;
520 /* subtract one, because it is the following line that
521 gets the specified number */
523 int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
525 /* Is this the last nonwhite stuff on the line? */
527 while (c == ' ' || c == '\t')
531 /* No more: store the line number and check following line. */
537 /* More follows: it must be a string constant (filename). */
539 /* Read the string constant. */
542 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
544 error ("invalid #line");
549 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
550 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
553 /* Each change of file name
554 reinitializes whether we are now in a system header. */
555 in_system_header = 0;
557 if (main_input_filename == 0)
558 main_input_filename = input_filename;
560 /* Is this the last nonwhite stuff on the line? */
562 while (c == ' ' || c == '\t')
566 /* Update the name in the top element of input_file_stack. */
567 if (input_file_stack)
568 input_file_stack->name = input_filename;
577 /* `1' after file name means entering new file.
578 `2' after file name means just left a file. */
580 if (token == CONSTANT
581 && TREE_CODE (yylval.ttype) == INTEGER_CST)
583 if (TREE_INT_CST_LOW (yylval.ttype) == 1)
585 /* Pushing to a new file. */
587 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
588 input_file_stack->line = old_lineno;
589 p->next = input_file_stack;
590 p->name = input_filename;
591 input_file_stack = p;
592 input_file_stack_tick++;
593 #ifdef DBX_DEBUGGING_INFO
594 if (write_symbols == DBX_DEBUG)
595 dbxout_start_new_source_file (input_filename);
597 #ifdef DWARF_DEBUGGING_INFO
598 if (debug_info_level == DINFO_LEVEL_VERBOSE
599 && write_symbols == DWARF_DEBUG)
600 dwarfout_start_new_source_file (input_filename);
601 #endif /* DWARF_DEBUGGING_INFO */
605 else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
607 /* Popping out of a file. */
608 if (input_file_stack->next)
610 struct file_stack *p = input_file_stack;
611 input_file_stack = p->next;
613 input_file_stack_tick++;
614 #ifdef DBX_DEBUGGING_INFO
615 if (write_symbols == DBX_DEBUG)
616 dbxout_resume_previous_source_file ();
618 #ifdef DWARF_DEBUGGING_INFO
619 if (debug_info_level == DINFO_LEVEL_VERBOSE
620 && write_symbols == DWARF_DEBUG)
621 dwarfout_resume_previous_source_file (input_file_stack->line);
622 #endif /* DWARF_DEBUGGING_INFO */
625 error ("#-lines for entering and leaving files don't match");
631 /* Now that we've pushed or popped the input stack,
632 update the name in the top element. */
633 if (input_file_stack)
634 input_file_stack->name = input_filename;
636 /* If we have handled a `1' or a `2',
637 see if there is another number to read. */
640 /* Is this the last nonwhite stuff on the line? */
642 while (c == ' ' || c == '\t')
652 /* `3' after file name means this is a system header file. */
654 if (token == CONSTANT
655 && TREE_CODE (yylval.ttype) == INTEGER_CST
656 && TREE_INT_CST_LOW (yylval.ttype) == 3)
657 in_system_header = 1, used_up = 1;
661 /* Is this the last nonwhite stuff on the line? */
663 while (c == ' ' || c == '\t')
670 warning ("unrecognized text at end of #line");
673 error ("invalid #-line");
675 /* skip the rest of this line. */
679 while ((c = getc (finput)) != EOF && c != '\n');
683 #ifdef HANDLE_SYSV_PRAGMA
685 /* Handle a #pragma directive. INPUT is the current input stream,
686 and C is a character to reread. Processes the entire input line
687 and returns a character for the caller to reread: either \n or EOF. */
689 /* This function has to be in this file, in order to get at
693 handle_sysv_pragma (input, c)
699 while (c == ' ' || c == '\t')
701 if (c == '\n' || c == EOF)
703 handle_pragma_token (0, 0);
713 handle_pragma_token (token_buffer, yylval.ttype);
716 handle_pragma_token (token_buffer, 0);
719 c = nextchar, nextchar = -1;
725 #endif /* HANDLE_SYSV_PRAGMA */
727 #define ENDFILE -1 /* token that represents end-of-file */
729 /* Read an escape sequence, returning its equivalent as a character,
730 or store 1 in *ignore_ptr if it is backslash-newline. */
733 readescape (ignore_ptr)
736 register int c = getc (finput);
738 register unsigned count;
739 unsigned firstdig = 0;
745 if (warn_traditional)
746 warning ("the meaning of `\\x' varies with -traditional");
748 if (flag_traditional)
757 if (!(c >= 'a' && c <= 'f')
758 && !(c >= 'A' && c <= 'F')
759 && !(c >= '0' && c <= '9'))
765 if (c >= 'a' && c <= 'f')
766 code += c - 'a' + 10;
767 if (c >= 'A' && c <= 'F')
768 code += c - 'A' + 10;
769 if (c >= '0' && c <= '9')
771 if (code != 0 || count != 0)
780 error ("\\x used with no following hex digits");
782 /* Digits are all 0's. Ok. */
784 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
786 && ((1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
788 pedwarn ("hex escape out of range");
791 case '0': case '1': case '2': case '3': case '4':
792 case '5': case '6': case '7':
795 while ((c <= '7') && (c >= '0') && (count++ < 3))
797 code = (code * 8) + (c - '0');
803 case '\\': case '\'': case '"':
812 return TARGET_NEWLINE;
827 if (warn_traditional)
828 warning ("the meaning of `\\a' varies with -traditional");
830 if (flag_traditional)
835 #if 0 /* Vertical tab is present in common usage compilers. */
836 if (flag_traditional)
844 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
850 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
854 /* `\%' is used to prevent SCCS from getting confused. */
857 pedwarn ("non-ANSI escape sequence `\\%c'", c);
860 if (c >= 040 && c < 0177)
861 pedwarn ("unknown escape sequence `\\%c'", c);
863 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
873 strcpy (buf, string);
875 /* We can't print string and character constants well
876 because the token_buffer contains the result of processing escapes. */
878 strcat (buf, " at end of input");
879 else if (token_buffer[0] == 0)
880 strcat (buf, " at null character");
881 else if (token_buffer[0] == '"')
882 strcat (buf, " before string constant");
883 else if (token_buffer[0] == '\'')
884 strcat (buf, " before character constant");
885 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
886 sprintf (buf + strlen (buf), " before character 0%o",
887 (unsigned char) token_buffer[0]);
889 strcat (buf, " before `%s'");
891 error (buf, token_buffer);
904 struct try_type type_sequence[] =
906 { &integer_type_node, 0, 0, 0},
907 { &unsigned_type_node, 1, 0, 0},
908 { &long_integer_type_node, 0, 1, 0},
909 { &long_unsigned_type_node, 1, 1, 0},
910 { &long_long_integer_type_node, 0, 1, 1},
911 { &long_long_unsigned_type_node, 1, 1, 1}
925 c = nextchar, nextchar = -1;
929 /* Effectively do c = skip_white_space (c)
930 but do it faster in the usual cases. */
943 /* Call skip_white_space so we can warn if appropriate. */
948 c = skip_white_space (c);
957 /* yylloc.first_line = lineno; */
968 if (dollars_in_ident)
973 /* Capital L may start a wide-string or wide-character constant. */
975 register int c = getc (finput);
984 goto string_constant;
991 if (!doing_objc_thang)
998 /* '@' may start a constant string object. */
999 register int c = getc(finput);
1003 goto string_constant;
1006 /* Fall through to treat '@' as the start of an identifier. */
1009 case 'A': case 'B': case 'C': case 'D': case 'E':
1010 case 'F': case 'G': case 'H': case 'I': case 'J':
1011 case 'K': case 'M': case 'N': case 'O':
1012 case 'P': case 'Q': case 'R': case 'S': case 'T':
1013 case 'U': case 'V': case 'W': case 'X': case 'Y':
1015 case 'a': case 'b': case 'c': case 'd': case 'e':
1016 case 'f': case 'g': case 'h': case 'i': case 'j':
1017 case 'k': case 'l': case 'm': case 'n': case 'o':
1018 case 'p': case 'q': case 'r': case 's': case 't':
1019 case 'u': case 'v': case 'w': case 'x': case 'y':
1024 while (isalnum (c) || c == '_' || c == '$' || c == '@')
1026 /* Make sure this char really belongs in an identifier. */
1027 if (c == '@' && ! doing_objc_thang)
1029 if (c == '$' && ! dollars_in_ident)
1032 if (p >= token_buffer + maxtoken)
1033 p = extend_token_buffer (p);
1045 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1048 register struct resword *ptr;
1050 if (ptr = is_reserved_word (token_buffer, p - token_buffer))
1053 yylval.ttype = ridpointers[(int) ptr->rid];
1054 value = (int) ptr->token;
1056 /* Only return OBJECTNAME if it is a typedef. */
1057 if (doing_objc_thang && value == OBJECTNAME)
1059 lastiddecl = lookup_name(yylval.ttype);
1061 if (lastiddecl == NULL_TREE
1062 || TREE_CODE (lastiddecl) != TYPE_DECL)
1066 /* Even if we decided to recognize asm, still perhaps warn. */
1068 && (value == ASM_KEYWORD || value == TYPEOF
1069 || ptr->rid == RID_INLINE)
1070 && token_buffer[0] != '_')
1071 pedwarn ("ANSI does not permit the keyword `%s'",
1076 /* If we did not find a keyword, look for an identifier
1079 if (value == IDENTIFIER)
1081 if (token_buffer[0] == '@')
1082 error("invalid identifier `%s'", token_buffer);
1084 yylval.ttype = get_identifier (token_buffer);
1085 lastiddecl = lookup_name (yylval.ttype);
1087 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1089 /* A user-invisible read-only initialized variable
1090 should be replaced by its value.
1091 We handle only strings since that's the only case used in C. */
1092 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1093 && DECL_IGNORED_P (lastiddecl)
1094 && TREE_READONLY (lastiddecl)
1095 && DECL_INITIAL (lastiddecl) != 0
1096 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1098 tree stringval = DECL_INITIAL (lastiddecl);
1100 /* Copy the string value so that we won't clobber anything
1101 if we put something in the TREE_CHAIN of this one. */
1102 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1103 TREE_STRING_POINTER (stringval));
1106 else if (doing_objc_thang)
1108 tree objc_interface_decl = is_class_name (yylval.ttype);
1110 if (objc_interface_decl)
1113 yylval.ttype = objc_interface_decl;
1120 case '0': case '1': case '2': case '3': case '4':
1121 case '5': case '6': case '7': case '8': case '9':
1126 int largest_digit = 0;
1128 /* for multi-precision arithmetic,
1129 we actually store only HOST_BITS_PER_CHAR bits in each part.
1130 The number of parts is chosen so as to be sufficient to hold
1131 the enough bits to fit into the two HOST_WIDE_INTs that contain
1132 the integer value (this is always at least as many bits as are
1133 in a target `long long' value, but may be wider). */
1134 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1135 int parts[TOTAL_PARTS];
1138 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag
1141 for (count = 0; count < TOTAL_PARTS; count++)
1149 *p++ = (c = getc (finput));
1150 if ((c == 'x') || (c == 'X'))
1153 *p++ = (c = getc (finput));
1155 /* Leading 0 forces octal unless the 0 is the only digit. */
1156 else if (c >= '0' && c <= '9')
1165 /* Read all the digits-and-decimal-points. */
1168 || (isalnum (c) && c != 'l' && c != 'L'
1169 && c != 'u' && c != 'U'
1170 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1171 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1176 error ("floating constant may not be in radix 16");
1177 if (floatflag == TOO_MANY_POINTS)
1178 /* We have already emitted an error. Don't need another. */
1180 else if (floatflag == AFTER_POINT)
1182 error ("malformed floating constant");
1183 floatflag = TOO_MANY_POINTS;
1184 /* Avoid another error from atof by forcing all characters
1185 from here on to be ignored. */
1189 floatflag = AFTER_POINT;
1192 *p++ = c = getc (finput);
1193 /* Accept '.' as the start of a floating-point number
1194 only when it is followed by a digit.
1195 Otherwise, unread the following non-digit
1196 and use the '.' as a structural token. */
1197 if (p == token_buffer + 2 && !isdigit (c))
1208 error ("parse error at `..'");
1211 token_buffer[1] = 0;
1218 /* It is not a decimal point.
1219 It should be a digit (perhaps a hex digit). */
1225 else if (base <= 10)
1227 if (c == 'e' || c == 'E')
1230 floatflag = AFTER_POINT;
1231 break; /* start of exponent */
1233 error ("nondigits in number and not hexadecimal");
1244 if (c >= largest_digit)
1248 for (count = 0; count < TOTAL_PARTS; count++)
1250 parts[count] *= base;
1254 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1256 &= (1 << HOST_BITS_PER_CHAR) - 1;
1262 /* If the extra highest-order part ever gets anything in it,
1263 the number is certainly too big. */
1264 if (parts[TOTAL_PARTS - 1] != 0)
1267 if (p >= token_buffer + maxtoken - 3)
1268 p = extend_token_buffer (p);
1269 *p++ = (c = getc (finput));
1274 error ("numeric constant with no digits");
1276 if (largest_digit >= base)
1277 error ("numeric constant contains digits beyond the radix");
1279 /* Remove terminating char from the token buffer and delimit the string */
1282 if (floatflag != NOT_FLOAT)
1284 tree type = double_type_node;
1285 int garbage_chars = 0, exceeds_double = 0;
1287 REAL_VALUE_TYPE value;
1290 /* Read explicit exponent if any, and put it in tokenbuf. */
1292 if ((c == 'e') || (c == 'E'))
1294 if (p >= token_buffer + maxtoken - 3)
1295 p = extend_token_buffer (p);
1298 if ((c == '+') || (c == '-'))
1304 error ("floating constant exponent has no digits");
1307 if (p >= token_buffer + maxtoken - 3)
1308 p = extend_token_buffer (p);
1317 /* Convert string to a double, checking for overflow. */
1318 if (setjmp (handler))
1320 error ("floating constant out of range");
1325 int fflag = 0, lflag = 0;
1326 /* Copy token_buffer now, while it has just the number
1327 and not the suffixes; once we add `f' or `i',
1328 REAL_VALUE_ATOF may not work any more. */
1329 char *copy = (char *) alloca (p - token_buffer + 1);
1330 bcopy (token_buffer, copy, p - token_buffer + 1);
1332 set_float_handler (handler);
1338 /* Read the suffixes to choose a data type. */
1343 error ("more than one `f' in numeric constant");
1349 error ("more than one `l' in numeric constant");
1355 error ("more than one `i' or `j' in numeric constant");
1357 pedwarn ("ANSI C forbids imaginary numeric constants");
1368 if (p >= token_buffer + maxtoken - 3)
1369 p = extend_token_buffer (p);
1375 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1376 tells the desired precision of the binary result
1377 of decimal-to-binary conversion. */
1382 error ("both `f' and `l' in floating constant");
1384 type = float_type_node;
1385 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1386 /* A diagnostic is required here by some ANSI C testsuites.
1387 This is not pedwarn, become some people don't want
1388 an error for this. */
1389 if (REAL_VALUE_ISINF (value) && pedantic)
1390 warning ("floating point number exceeds range of `float'");
1394 type = long_double_type_node;
1395 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1396 if (REAL_VALUE_ISINF (value) && pedantic)
1397 warning ("floating point number exceeds range of `long double'");
1401 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1402 if (REAL_VALUE_ISINF (value) && pedantic)
1403 warning ("floating point number exceeds range of `double'");
1406 set_float_handler (NULL_PTR);
1409 if (errno == ERANGE && !flag_traditional && pedantic)
1411 /* ERANGE is also reported for underflow,
1412 so test the value to distinguish overflow from that. */
1413 if (REAL_VALUES_LESS (dconst1, value)
1414 || REAL_VALUES_LESS (value, dconstm1))
1416 warning ("floating point number exceeds range of `double'");
1422 while (isalnum (c) || c == '.' || c == '_'
1423 || (!flag_traditional && (c == '+' || c == '-')
1424 && (p[-1] == 'e' || p[-1] == 'E')))
1426 if (p >= token_buffer + maxtoken - 3)
1427 p = extend_token_buffer (p);
1432 if (garbage_chars > 0)
1433 error ("garbage at end of number");
1435 /* If the result is not a number, assume it must have been
1436 due to some error message above, so silently convert
1438 if (REAL_VALUE_ISNAN (value))
1441 /* Create a node with determined type and value. */
1443 yylval.ttype = build_complex (convert (type, integer_zero_node),
1444 build_real (type, value));
1446 yylval.ttype = build_real (type, value);
1453 tree traditional_type, ansi_type, type;
1454 HOST_WIDE_INT high, low;
1455 int spec_unsigned = 0;
1457 int spec_long_long = 0;
1463 if (c == 'u' || c == 'U')
1466 error ("two `u's in integer constant");
1469 else if (c == 'l' || c == 'L')
1474 error ("three `l's in integer constant");
1476 pedwarn ("ANSI C forbids long long integer constants");
1481 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1484 error ("more than one `i' or `j' in numeric constant");
1486 pedwarn ("ANSI C forbids imaginary numeric constants");
1491 if (isalnum (c) || c == '.' || c == '_'
1492 || (!flag_traditional && (c == '+' || c == '-')
1493 && (p[-1] == 'e' || p[-1] == 'E')))
1495 error ("garbage at end of number");
1496 while (isalnum (c) || c == '.' || c == '_'
1497 || (!flag_traditional && (c == '+' || c == '-')
1498 && (p[-1] == 'e' || p[-1] == 'E')))
1500 if (p >= token_buffer + maxtoken - 3)
1501 p = extend_token_buffer (p);
1508 if (p >= token_buffer + maxtoken - 3)
1509 p = extend_token_buffer (p);
1516 /* If the constant is not long long and it won't fit in an
1517 unsigned long, or if the constant is long long and won't fit
1518 in an unsigned long long, then warn that the constant is out
1521 /* ??? This assumes that long long and long integer types are
1522 a multiple of 8 bits. This better than the original code
1523 though which assumed that long was exactly 32 bits and long
1524 long was exactly 64 bits. */
1527 bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1529 bytes = TYPE_PRECISION (long_integer_type_node) / 8;
1532 for (i = bytes; i < TOTAL_PARTS; i++)
1536 pedwarn ("integer constant out of range");
1538 /* This is simplified by the fact that our constant
1539 is always positive. */
1543 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1545 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1546 / HOST_BITS_PER_CHAR)]
1547 << (i * HOST_BITS_PER_CHAR));
1548 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1551 yylval.ttype = build_int_2 (low, high);
1552 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1554 /* If warn_traditional, calculate both the ANSI type and the
1555 traditional type, then see if they disagree.
1556 Otherwise, calculate only the type for the dialect in use. */
1557 if (warn_traditional || flag_traditional)
1559 /* Calculate the traditional type. */
1560 /* Traditionally, any constant is signed;
1561 but if unsigned is specified explicitly, obey that.
1562 Use the smallest size with the right number of bits,
1563 except for one special case with decimal constants. */
1564 if (! spec_long && base != 10
1565 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1566 traditional_type = (spec_unsigned ? unsigned_type_node
1567 : integer_type_node);
1568 /* A decimal constant must be long
1569 if it does not fit in type int.
1570 I think this is independent of whether
1571 the constant is signed. */
1572 else if (! spec_long && base == 10
1573 && int_fits_type_p (yylval.ttype, integer_type_node))
1574 traditional_type = (spec_unsigned ? unsigned_type_node
1575 : integer_type_node);
1576 else if (! spec_long_long)
1577 traditional_type = (spec_unsigned ? long_unsigned_type_node
1578 : long_integer_type_node);
1580 traditional_type = (spec_unsigned
1581 ? long_long_unsigned_type_node
1582 : long_long_integer_type_node);
1584 if (warn_traditional || ! flag_traditional)
1586 /* Calculate the ANSI type. */
1587 if (! spec_long && ! spec_unsigned
1588 && int_fits_type_p (yylval.ttype, integer_type_node))
1589 ansi_type = integer_type_node;
1590 else if (! spec_long && (base != 10 || spec_unsigned)
1591 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1592 ansi_type = unsigned_type_node;
1593 else if (! spec_unsigned && !spec_long_long
1594 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1595 ansi_type = long_integer_type_node;
1596 else if (! spec_long_long)
1597 ansi_type = long_unsigned_type_node;
1598 else if (! spec_unsigned
1599 /* Verify value does not overflow into sign bit. */
1600 && TREE_INT_CST_HIGH (yylval.ttype) >= 0
1601 && int_fits_type_p (yylval.ttype,
1602 long_long_integer_type_node))
1603 ansi_type = long_long_integer_type_node;
1605 ansi_type = long_long_unsigned_type_node;
1608 type = flag_traditional ? traditional_type : ansi_type;
1610 if (warn_traditional && traditional_type != ansi_type)
1612 if (TYPE_PRECISION (traditional_type)
1613 != TYPE_PRECISION (ansi_type))
1614 warning ("width of integer constant changes with -traditional");
1615 else if (TREE_UNSIGNED (traditional_type)
1616 != TREE_UNSIGNED (ansi_type))
1617 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1619 warning ("width of integer constant may change on other systems with -traditional");
1622 if (!flag_traditional && !int_fits_type_p (yylval.ttype, type)
1624 pedwarn ("integer constant out of range");
1626 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1627 warning ("decimal constant is so large that it is unsigned");
1631 if (TYPE_PRECISION (type)
1632 <= TYPE_PRECISION (integer_type_node))
1634 = build_complex (integer_zero_node,
1635 convert (integer_type_node, yylval.ttype));
1637 error ("complex integer constant is too wide for `complex int'");
1639 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1640 /* The traditional constant 0x80000000 is signed
1641 but doesn't fit in the range of int.
1642 This will change it to -0x80000000, which does fit. */
1644 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1645 yylval.ttype = convert (type, yylval.ttype);
1646 TREE_OVERFLOW (yylval.ttype)
1647 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1650 TREE_TYPE (yylval.ttype) = type;
1655 value = CONSTANT; break;
1661 register int result = 0;
1662 register int num_chars = 0;
1663 unsigned width = TYPE_PRECISION (char_type_node);
1668 width = WCHAR_TYPE_SIZE;
1669 #ifdef MULTIBYTE_CHARS
1670 max_chars = MB_CUR_MAX;
1676 max_chars = TYPE_PRECISION (integer_type_node) / width;
1684 if (c == '\'' || c == EOF)
1690 c = readescape (&ignore);
1693 if (width < HOST_BITS_PER_INT
1694 && (unsigned) c >= (1 << width))
1695 pedwarn ("escape sequence out of range for character");
1696 #ifdef MAP_CHARACTER
1698 c = MAP_CHARACTER (c);
1704 pedwarn ("ANSI C forbids newline in character constant");
1707 #ifdef MAP_CHARACTER
1709 c = MAP_CHARACTER (c);
1713 if (num_chars > maxtoken - 4)
1714 extend_token_buffer (token_buffer);
1716 token_buffer[num_chars] = c;
1718 /* Merge character into result; ignore excess chars. */
1719 if (num_chars < max_chars + 1)
1721 if (width < HOST_BITS_PER_INT)
1722 result = (result << width) | (c & ((1 << width) - 1));
1728 token_buffer[num_chars + 1] = '\'';
1729 token_buffer[num_chars + 2] = 0;
1732 error ("malformatted character constant");
1733 else if (num_chars == 0)
1734 error ("empty character constant");
1735 else if (num_chars > max_chars)
1737 num_chars = max_chars;
1738 error ("character constant too long");
1740 else if (num_chars != 1 && ! flag_traditional)
1741 warning ("multi-character character constant");
1743 /* If char type is signed, sign-extend the constant. */
1746 int num_bits = num_chars * width;
1748 /* We already got an error; avoid invalid shift. */
1749 yylval.ttype = build_int_2 (0, 0);
1750 else if (TREE_UNSIGNED (char_type_node)
1751 || ((result >> (num_bits - 1)) & 1) == 0)
1753 = build_int_2 (result & ((unsigned HOST_WIDE_INT) ~0
1754 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1758 = build_int_2 (result | ~((unsigned HOST_WIDE_INT) ~0
1759 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1761 TREE_TYPE (yylval.ttype) = integer_type_node;
1765 #ifdef MULTIBYTE_CHARS
1766 /* Set the initial shift state and convert the next sequence. */
1768 /* In all locales L'\0' is zero and mbtowc will return zero,
1771 || (num_chars == 1 && token_buffer[1] != '\0'))
1774 (void) mbtowc (NULL_PTR, NULL_PTR, 0);
1775 if (mbtowc (& wc, token_buffer + 1, num_chars) == num_chars)
1778 warning ("Ignoring invalid multibyte character");
1781 yylval.ttype = build_int_2 (result, 0);
1782 TREE_TYPE (yylval.ttype) = wchar_type_node;
1793 p = token_buffer + 1;
1795 while (c != '"' && c >= 0)
1800 c = readescape (&ignore);
1804 && TYPE_PRECISION (char_type_node) < HOST_BITS_PER_INT
1805 && c >= (1 << TYPE_PRECISION (char_type_node)))
1806 pedwarn ("escape sequence out of range for character");
1811 pedwarn ("ANSI C forbids newline in string constant");
1815 if (p == token_buffer + maxtoken)
1816 p = extend_token_buffer (p);
1825 error ("Unterminated string constant");
1827 /* We have read the entire constant.
1828 Construct a STRING_CST for the result. */
1832 /* If this is a L"..." wide-string, convert the multibyte string
1833 to a wide character string. */
1834 char *widep = (char *) alloca ((p - token_buffer) * WCHAR_BYTES);
1837 #ifdef MULTIBYTE_CHARS
1838 len = mbstowcs ((wchar_t *) widep, token_buffer + 1, p - token_buffer);
1839 if (len < 0 || len >= (p - token_buffer))
1841 warning ("Ignoring invalid multibyte string");
1844 bzero (widep + (len * WCHAR_BYTES), WCHAR_BYTES);
1847 union { long l; char c[sizeof (long)]; } u;
1851 /* Determine whether host is little or big endian. */
1853 big_endian = u.c[sizeof (long) - 1];
1854 wp = widep + (big_endian ? WCHAR_BYTES - 1 : 0);
1856 bzero (widep, (p - token_buffer) * WCHAR_BYTES);
1857 for (cp = token_buffer + 1; cp < p; cp++)
1858 *wp = *cp, wp += WCHAR_BYTES;
1859 len = p - token_buffer - 1;
1862 yylval.ttype = build_string ((len + 1) * WCHAR_BYTES, widep);
1863 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
1868 extern tree build_objc_string();
1869 /* Return an Objective-C @"..." constant string object. */
1870 yylval.ttype = build_objc_string (p - token_buffer,
1872 TREE_TYPE (yylval.ttype) = char_array_type_node;
1873 value = OBJC_STRING;
1877 yylval.ttype = build_string (p - token_buffer, token_buffer + 1);
1878 TREE_TYPE (yylval.ttype) = char_array_type_node;
1909 yylval.code = PLUS_EXPR; break;
1911 yylval.code = MINUS_EXPR; break;
1913 yylval.code = BIT_AND_EXPR; break;
1915 yylval.code = BIT_IOR_EXPR; break;
1917 yylval.code = MULT_EXPR; break;
1919 yylval.code = TRUNC_DIV_EXPR; break;
1921 yylval.code = TRUNC_MOD_EXPR; break;
1923 yylval.code = BIT_XOR_EXPR; break;
1925 yylval.code = LSHIFT_EXPR; break;
1927 yylval.code = RSHIFT_EXPR; break;
1929 yylval.code = LT_EXPR; break;
1931 yylval.code = GT_EXPR; break;
1934 token_buffer[1] = c1 = getc (finput);
1935 token_buffer[2] = 0;
1942 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
1944 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
1946 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
1948 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
1950 value = ASSIGN; goto done;
1956 value = PLUSPLUS; goto done;
1958 value = MINUSMINUS; goto done;
1960 value = ANDAND; goto done;
1962 value = OROR; goto done;
1975 { value = POINTSAT; goto done; }
1979 { value = ']'; goto done; }
1983 { value = '{'; goto done; }
1985 { value = '['; goto done; }
1989 { value = '}'; goto done; }
1992 ungetc (c1, finput);
1993 token_buffer[1] = 0;
1995 if ((c == '<') || (c == '>'))
1996 value = ARITHCOMPARE;
2002 /* Don't make yyparse think this is eof. */
2011 /* yylloc.last_line = lineno; */
2016 /* Sets the value of the 'yydebug' variable to VALUE.
2017 This is a function so we don't have to have YYDEBUG defined
2018 in order to build the compiler. */
2027 warning ("YYDEBUG not defined.");