1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 1988, 1989, 1992, 1994 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, 675 Mass Ave, Cambridge, MA 02139, USA. */
36 #ifdef MULTIBYTE_CHARS
45 /* The elements of `ridpointers' are identifier nodes
46 for the reserved type names and storage classes.
47 It is indexed by a RID_... value. */
48 tree ridpointers[(int) RID_MAX];
50 /* Cause the `yydebug' variable to be defined. */
53 /* the declaration found for the last IDENTIFIER token read in.
54 yylex must look this up to detect typedefs, which get token type TYPENAME,
55 so it is left around in case the identifier is not a typedef but is
56 used in a context which makes it a reference to a variable. */
59 /* Nonzero enables objc features. */
63 extern tree is_class_name ();
67 /* File used for outputting assembler code. */
68 extern FILE *asm_out_file;
70 #ifndef WCHAR_TYPE_SIZE
72 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
74 #define WCHAR_TYPE_SIZE BITS_PER_WORD
78 /* Number of bytes in a wide character. */
79 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
81 static int maxtoken; /* Current nominal length of token buffer. */
82 char *token_buffer; /* Pointer to token buffer.
83 Actual allocated length is maxtoken + 2.
84 This is not static because objc-parse.y uses it. */
86 /* Nonzero if end-of-file has been seen on input. */
87 static int end_of_file;
89 /* Buffered-back input character; faster than using ungetc. */
90 static int nextchar = -1;
94 /* Do not insert generated code into the source, instead, include it.
95 This allows us to build gcc automatically even for targets that
96 need to add or modify the reserved keyword lists. */
99 /* Return something to represent absolute declarators containing a *.
100 TARGET is the absolute declarator that the * contains.
101 TYPE_QUALS is a list of modifiers such as const or volatile
102 to apply to the pointer type, represented as identifiers.
104 We return an INDIRECT_REF whose "contents" are TARGET
105 and whose type is the modifier list. */
108 make_pointer_declarator (type_quals, target)
109 tree type_quals, target;
111 return build1 (INDIRECT_REF, type_quals, target);
115 forget_protocol_qualifiers ()
117 int i, n = sizeof wordlist / sizeof (struct resword);
119 for (i = 0; i < n; i++)
120 if ((int) wordlist[i].rid >= (int) RID_IN
121 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
122 wordlist[i].name = "";
126 remember_protocol_qualifiers ()
128 int i, n = sizeof wordlist / sizeof (struct resword);
130 for (i = 0; i < n; i++)
131 if (wordlist[i].rid == RID_IN)
132 wordlist[i].name = "in";
133 else if (wordlist[i].rid == RID_OUT)
134 wordlist[i].name = "out";
135 else if (wordlist[i].rid == RID_INOUT)
136 wordlist[i].name = "inout";
137 else if (wordlist[i].rid == RID_BYCOPY)
138 wordlist[i].name = "bycopy";
139 else if (wordlist[i].rid == RID_ONEWAY)
140 wordlist[i].name = "oneway";
146 /* Make identifier nodes long enough for the language-specific slots. */
147 set_identifier_size (sizeof (struct lang_identifier));
149 /* Start it at 0, because check_newline is called at the very beginning
150 and will increment it to 1. */
153 #ifdef MULTIBYTE_CHARS
154 /* Change to the native locale for multibyte conversions. */
155 setlocale (LC_CTYPE, "");
159 token_buffer = (char *) xmalloc (maxtoken + 2);
161 ridpointers[(int) RID_INT] = get_identifier ("int");
162 ridpointers[(int) RID_CHAR] = get_identifier ("char");
163 ridpointers[(int) RID_VOID] = get_identifier ("void");
164 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
165 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
166 ridpointers[(int) RID_SHORT] = get_identifier ("short");
167 ridpointers[(int) RID_LONG] = get_identifier ("long");
168 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
169 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
170 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
171 ridpointers[(int) RID_CONST] = get_identifier ("const");
172 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
173 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
174 ridpointers[(int) RID_STATIC] = get_identifier ("static");
175 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
176 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
177 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
178 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
179 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
180 ridpointers[(int) RID_ID] = get_identifier ("id");
181 ridpointers[(int) RID_IN] = get_identifier ("in");
182 ridpointers[(int) RID_OUT] = get_identifier ("out");
183 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
184 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
185 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
186 forget_protocol_qualifiers();
188 /* Some options inhibit certain reserved words.
189 Clear those words out of the hash table so they won't be recognized. */
190 #define UNSET_RESERVED_WORD(STRING) \
191 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
192 if (s) s->name = ""; } while (0)
194 if (! doing_objc_thang)
195 UNSET_RESERVED_WORD ("id");
197 if (flag_traditional)
199 UNSET_RESERVED_WORD ("const");
200 UNSET_RESERVED_WORD ("volatile");
201 UNSET_RESERVED_WORD ("typeof");
202 UNSET_RESERVED_WORD ("signed");
203 UNSET_RESERVED_WORD ("inline");
204 UNSET_RESERVED_WORD ("iterator");
205 UNSET_RESERVED_WORD ("complex");
209 UNSET_RESERVED_WORD ("asm");
210 UNSET_RESERVED_WORD ("typeof");
211 UNSET_RESERVED_WORD ("inline");
212 UNSET_RESERVED_WORD ("iterator");
213 UNSET_RESERVED_WORD ("complex");
218 reinit_parse_for_function ()
222 /* Function used when yydebug is set, to print a token in more detail. */
225 yyprint (file, yychar, yylval)
237 if (IDENTIFIER_POINTER (t))
238 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
243 if (TREE_CODE (t) == INTEGER_CST)
245 #if HOST_BITS_PER_WIDE_INT == 64
246 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
252 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
258 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
264 /* If C is not whitespace, return C.
265 Otherwise skip whitespace and return first nonwhite char read. */
271 static int newline_warning = 0;
277 /* We don't recognize comments here, because
278 cpp output can include / and * consecutively as operators.
279 Also, there's no need, since cpp removes all comments. */
282 c = check_newline ();
294 /* ANSI C says the effects of a carriage return in a source file
296 if (pedantic && !newline_warning)
298 warning ("carriage return in source file");
299 warning ("(we only warn about the first carriage return)");
310 error ("stray '\\' in program");
320 /* Skips all of the white space at the current location in the input file.
321 Must use and reset nextchar if it has the next character. */
324 position_after_white_space ()
329 c = nextchar, nextchar = -1;
333 ungetc (skip_white_space (c), finput);
336 /* Make the token buffer longer, preserving the data in it.
337 P should point to just beyond the last valid character in the old buffer.
338 The value we return is a pointer to the new buffer
339 at a place corresponding to P. */
342 extend_token_buffer (p)
345 int offset = p - token_buffer;
347 maxtoken = maxtoken * 2 + 10;
348 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
350 return token_buffer + offset;
353 /* At the beginning of a line, increment the line number
354 and process any #-directive on this line.
355 If the line is a #-directive, read the entire line and return a newline.
356 Otherwise, return the line's first non-whitespace character. */
366 /* Read first nonwhite char on the line. */
369 while (c == ' ' || c == '\t')
374 /* If not #, return it so caller will use it. */
378 /* Read first nonwhite char after the `#'. */
381 while (c == ' ' || c == '\t')
384 /* If a letter follows, then if the word here is `line', skip
385 it and ignore it; otherwise, ignore the line, with an error
386 if the word isn't `pragma', `ident', `define', or `undef'. */
388 if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
392 if (getc (finput) == 'r'
393 && getc (finput) == 'a'
394 && getc (finput) == 'g'
395 && getc (finput) == 'm'
396 && getc (finput) == 'a'
397 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
399 #ifdef HANDLE_SYSV_PRAGMA
400 return handle_sysv_pragma (finput, c);
401 #endif /* HANDLE_SYSV_PRAGMA */
403 HANDLE_PRAGMA (finput);
404 #endif /* HANDLE_PRAGMA */
411 if (getc (finput) == 'e'
412 && getc (finput) == 'f'
413 && getc (finput) == 'i'
414 && getc (finput) == 'n'
415 && getc (finput) == 'e'
416 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
418 #ifdef DWARF_DEBUGGING_INFO
419 if ((debug_info_level == DINFO_LEVEL_VERBOSE)
420 && (write_symbols == DWARF_DEBUG))
421 dwarfout_define (lineno, get_directive_line (finput));
422 #endif /* DWARF_DEBUGGING_INFO */
428 if (getc (finput) == 'n'
429 && getc (finput) == 'd'
430 && getc (finput) == 'e'
431 && getc (finput) == 'f'
432 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
434 #ifdef DWARF_DEBUGGING_INFO
435 if ((debug_info_level == DINFO_LEVEL_VERBOSE)
436 && (write_symbols == DWARF_DEBUG))
437 dwarfout_undef (lineno, get_directive_line (finput));
438 #endif /* DWARF_DEBUGGING_INFO */
444 if (getc (finput) == 'i'
445 && getc (finput) == 'n'
446 && getc (finput) == 'e'
447 && ((c = getc (finput)) == ' ' || c == '\t'))
452 if (getc (finput) == 'd'
453 && getc (finput) == 'e'
454 && getc (finput) == 'n'
455 && getc (finput) == 't'
456 && ((c = getc (finput)) == ' ' || c == '\t'))
458 /* #ident. The pedantic warning is now in cccp.c. */
460 /* Here we have just seen `#ident '.
461 A string constant should follow. */
463 while (c == ' ' || c == '\t')
466 /* If no argument, ignore the line. */
473 || TREE_CODE (yylval.ttype) != STRING_CST)
475 error ("invalid #ident");
481 #ifdef ASM_OUTPUT_IDENT
482 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
486 /* Skip the rest of this line. */
491 error ("undefined or invalid # directive");
496 /* Here we have either `#line' or `# <nonletter>'.
497 In either case, it should be a line number; a digit should follow. */
499 while (c == ' ' || c == '\t')
502 /* If the # is the only nonwhite char on the line,
503 just ignore it. Check the new newline. */
507 /* Something follows the #; read a token. */
512 if (token == CONSTANT
513 && TREE_CODE (yylval.ttype) == INTEGER_CST)
515 int old_lineno = lineno;
517 /* subtract one, because it is the following line that
518 gets the specified number */
520 int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
522 /* Is this the last nonwhite stuff on the line? */
524 while (c == ' ' || c == '\t')
528 /* No more: store the line number and check following line. */
534 /* More follows: it must be a string constant (filename). */
536 /* Read the string constant. */
539 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
541 error ("invalid #line");
546 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
547 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
550 /* Each change of file name
551 reinitializes whether we are now in a system header. */
552 in_system_header = 0;
554 if (main_input_filename == 0)
555 main_input_filename = input_filename;
557 /* Is this the last nonwhite stuff on the line? */
559 while (c == ' ' || c == '\t')
563 /* Update the name in the top element of input_file_stack. */
564 if (input_file_stack)
565 input_file_stack->name = input_filename;
574 /* `1' after file name means entering new file.
575 `2' after file name means just left a file. */
577 if (token == CONSTANT
578 && TREE_CODE (yylval.ttype) == INTEGER_CST)
580 if (TREE_INT_CST_LOW (yylval.ttype) == 1)
582 /* Pushing to a new file. */
584 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
585 input_file_stack->line = old_lineno;
586 p->next = input_file_stack;
587 p->name = input_filename;
588 input_file_stack = p;
589 input_file_stack_tick++;
590 #ifdef DWARF_DEBUGGING_INFO
591 if (debug_info_level == DINFO_LEVEL_VERBOSE
592 && write_symbols == DWARF_DEBUG)
593 dwarfout_start_new_source_file (input_filename);
594 #endif /* DWARF_DEBUGGING_INFO */
598 else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
600 /* Popping out of a file. */
601 if (input_file_stack->next)
603 struct file_stack *p = input_file_stack;
604 input_file_stack = p->next;
606 input_file_stack_tick++;
607 #ifdef DWARF_DEBUGGING_INFO
608 if (debug_info_level == DINFO_LEVEL_VERBOSE
609 && write_symbols == DWARF_DEBUG)
610 dwarfout_resume_previous_source_file (input_file_stack->line);
611 #endif /* DWARF_DEBUGGING_INFO */
614 error ("#-lines for entering and leaving files don't match");
620 /* Now that we've pushed or popped the input stack,
621 update the name in the top element. */
622 if (input_file_stack)
623 input_file_stack->name = input_filename;
625 /* If we have handled a `1' or a `2',
626 see if there is another number to read. */
629 /* Is this the last nonwhite stuff on the line? */
631 while (c == ' ' || c == '\t')
641 /* `3' after file name means this is a system header file. */
643 if (token == CONSTANT
644 && TREE_CODE (yylval.ttype) == INTEGER_CST
645 && TREE_INT_CST_LOW (yylval.ttype) == 3)
646 in_system_header = 1;
649 error ("invalid #-line");
651 /* skip the rest of this line. */
655 while ((c = getc (finput)) != EOF && c != '\n');
659 #ifdef HANDLE_SYSV_PRAGMA
661 /* Handle a #pragma directive. INPUT is the current input stream,
662 and C is a character to reread. Processes the entire input line
663 and returns a character for the caller to reread: either \n or EOF. */
665 /* This function has to be in this file, in order to get at
669 handle_sysv_pragma (input, c)
675 while (c == ' ' || c == '\t')
677 if (c == '\n' || c == EOF)
679 handle_pragma_token (0, 0);
689 handle_pragma_token (token_buffer, yylval.ttype);
692 handle_pragma_token (token_buffer, 0);
695 c = nextchar, nextchar = -1;
701 #endif /* HANDLE_SYSV_PRAGMA */
703 #define ENDFILE -1 /* token that represents end-of-file */
705 /* Read an escape sequence, returning its equivalent as a character,
706 or store 1 in *ignore_ptr if it is backslash-newline. */
709 readescape (ignore_ptr)
712 register int c = getc (finput);
714 register unsigned count;
715 unsigned firstdig = 0;
721 if (warn_traditional)
722 warning ("the meaning of `\\x' varies with -traditional");
724 if (flag_traditional)
733 if (!(c >= 'a' && c <= 'f')
734 && !(c >= 'A' && c <= 'F')
735 && !(c >= '0' && c <= '9'))
741 if (c >= 'a' && c <= 'f')
742 code += c - 'a' + 10;
743 if (c >= 'A' && c <= 'F')
744 code += c - 'A' + 10;
745 if (c >= '0' && c <= '9')
747 if (code != 0 || count != 0)
756 error ("\\x used with no following hex digits");
758 /* Digits are all 0's. Ok. */
760 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
762 && ((1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
764 pedwarn ("hex escape out of range");
767 case '0': case '1': case '2': case '3': case '4':
768 case '5': case '6': case '7':
771 while ((c <= '7') && (c >= '0') && (count++ < 3))
773 code = (code * 8) + (c - '0');
779 case '\\': case '\'': case '"':
788 return TARGET_NEWLINE;
803 if (warn_traditional)
804 warning ("the meaning of `\\a' varies with -traditional");
806 if (flag_traditional)
811 #if 0 /* Vertical tab is present in common usage compilers. */
812 if (flag_traditional)
820 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
826 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
830 /* `\%' is used to prevent SCCS from getting confused. */
833 pedwarn ("non-ANSI escape sequence `\\%c'", c);
836 if (c >= 040 && c < 0177)
837 pedwarn ("unknown escape sequence `\\%c'", c);
839 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
849 strcpy (buf, string);
851 /* We can't print string and character constants well
852 because the token_buffer contains the result of processing escapes. */
854 strcat (buf, " at end of input");
855 else if (token_buffer[0] == 0)
856 strcat (buf, " at null character");
857 else if (token_buffer[0] == '"')
858 strcat (buf, " before string constant");
859 else if (token_buffer[0] == '\'')
860 strcat (buf, " before character constant");
861 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
862 sprintf (buf + strlen (buf), " before character 0%o",
863 (unsigned char) token_buffer[0]);
865 strcat (buf, " before `%s'");
867 error (buf, token_buffer);
880 struct try_type type_sequence[] =
882 { &integer_type_node, 0, 0, 0},
883 { &unsigned_type_node, 1, 0, 0},
884 { &long_integer_type_node, 0, 1, 0},
885 { &long_unsigned_type_node, 1, 1, 0},
886 { &long_long_integer_type_node, 0, 1, 1},
887 { &long_long_unsigned_type_node, 1, 1, 1}
901 c = nextchar, nextchar = -1;
905 /* Effectively do c = skip_white_space (c)
906 but do it faster in the usual cases. */
919 /* Call skip_white_space so we can warn if appropriate. */
924 c = skip_white_space (c);
933 /* yylloc.first_line = lineno; */
944 if (dollars_in_ident)
949 /* Capital L may start a wide-string or wide-character constant. */
951 register int c = getc (finput);
960 goto string_constant;
967 if (!doing_objc_thang)
974 /* '@' may start a constant string object. */
975 register int c = getc(finput);
979 goto string_constant;
982 /* Fall through to treat '@' as the start of an indentifier. */
985 case 'A': case 'B': case 'C': case 'D': case 'E':
986 case 'F': case 'G': case 'H': case 'I': case 'J':
987 case 'K': case 'M': case 'N': case 'O':
988 case 'P': case 'Q': case 'R': case 'S': case 'T':
989 case 'U': case 'V': case 'W': case 'X': case 'Y':
991 case 'a': case 'b': case 'c': case 'd': case 'e':
992 case 'f': case 'g': case 'h': case 'i': case 'j':
993 case 'k': case 'l': case 'm': case 'n': case 'o':
994 case 'p': case 'q': case 'r': case 's': case 't':
995 case 'u': case 'v': case 'w': case 'x': case 'y':
1000 while (isalnum (c) || c == '_' || c == '$' || c == '@')
1002 /* Make sure this char really belongs in an identifier. */
1003 if (c == '@' && ! doing_objc_thang)
1005 if (c == '$' && ! dollars_in_ident)
1008 if (p >= token_buffer + maxtoken)
1009 p = extend_token_buffer (p);
1021 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1024 register struct resword *ptr;
1026 if (ptr = is_reserved_word (token_buffer, p - token_buffer))
1029 yylval.ttype = ridpointers[(int) ptr->rid];
1030 value = (int) ptr->token;
1032 /* Only return OBJECTNAME if it is a typedef. */
1033 if (doing_objc_thang && value == OBJECTNAME)
1035 lastiddecl = lookup_name(yylval.ttype);
1037 if (lastiddecl == NULL_TREE
1038 || TREE_CODE (lastiddecl) != TYPE_DECL)
1042 /* Even if we decided to recognize asm, still perhaps warn. */
1044 && (value == ASM_KEYWORD || value == TYPEOF
1045 || ptr->rid == RID_INLINE)
1046 && token_buffer[0] != '_')
1047 pedwarn ("ANSI does not permit the keyword `%s'",
1052 /* If we did not find a keyword, look for an identifier
1055 if (value == IDENTIFIER)
1057 if (token_buffer[0] == '@')
1058 error("invalid identifier `%s'", token_buffer);
1060 yylval.ttype = get_identifier (token_buffer);
1061 lastiddecl = lookup_name (yylval.ttype);
1063 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1065 /* A user-invisible read-only initialized variable
1066 should be replaced by its value.
1067 We handle only strings since that's the only case used in C. */
1068 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1069 && DECL_IGNORED_P (lastiddecl)
1070 && TREE_READONLY (lastiddecl)
1071 && DECL_INITIAL (lastiddecl) != 0
1072 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1074 tree stringval = DECL_INITIAL (lastiddecl);
1076 /* Copy the string value so that we won't clobber anything
1077 if we put something in the TREE_CHAIN of this one. */
1078 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1079 TREE_STRING_POINTER (stringval));
1082 else if (doing_objc_thang)
1084 tree objc_interface_decl = is_class_name (yylval.ttype);
1086 if (objc_interface_decl)
1089 yylval.ttype = objc_interface_decl;
1096 case '0': case '1': case '2': case '3': case '4':
1097 case '5': case '6': case '7': case '8': case '9':
1102 int largest_digit = 0;
1104 /* for multi-precision arithmetic,
1105 we actually store only HOST_BITS_PER_CHAR bits in each part.
1106 The number of parts is chosen so as to be sufficient to hold
1107 the enough bits to fit into the two HOST_WIDE_INTs that contain
1108 the integer value (this is always at least as many bits as are
1109 in a target `long long' value, but may be wider). */
1110 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1111 int parts[TOTAL_PARTS];
1114 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag
1117 for (count = 0; count < TOTAL_PARTS; count++)
1125 *p++ = (c = getc (finput));
1126 if ((c == 'x') || (c == 'X'))
1129 *p++ = (c = getc (finput));
1131 /* Leading 0 forces octal unless the 0 is the only digit. */
1132 else if (c >= '0' && c <= '9')
1141 /* Read all the digits-and-decimal-points. */
1144 || (isalnum (c) && c != 'l' && c != 'L'
1145 && c != 'u' && c != 'U'
1146 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1147 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1152 error ("floating constant may not be in radix 16");
1153 if (floatflag == TOO_MANY_POINTS)
1154 /* We have already emitted an error. Don't need another. */
1156 else if (floatflag == AFTER_POINT)
1158 error ("malformed floating constant");
1159 floatflag = TOO_MANY_POINTS;
1160 /* Avoid another error from atof by forcing all characters
1161 from here on to be ignored. */
1165 floatflag = AFTER_POINT;
1168 *p++ = c = getc (finput);
1169 /* Accept '.' as the start of a floating-point number
1170 only when it is followed by a digit.
1171 Otherwise, unread the following non-digit
1172 and use the '.' as a structural token. */
1173 if (p == token_buffer + 2 && !isdigit (c))
1184 error ("parse error at `..'");
1187 token_buffer[1] = 0;
1194 /* It is not a decimal point.
1195 It should be a digit (perhaps a hex digit). */
1201 else if (base <= 10)
1203 if (c == 'e' || c == 'E')
1206 floatflag = AFTER_POINT;
1207 break; /* start of exponent */
1209 error ("nondigits in number and not hexadecimal");
1220 if (c >= largest_digit)
1224 for (count = 0; count < TOTAL_PARTS; count++)
1226 parts[count] *= base;
1230 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1232 &= (1 << HOST_BITS_PER_CHAR) - 1;
1238 /* If the extra highest-order part ever gets anything in it,
1239 the number is certainly too big. */
1240 if (parts[TOTAL_PARTS - 1] != 0)
1243 if (p >= token_buffer + maxtoken - 3)
1244 p = extend_token_buffer (p);
1245 *p++ = (c = getc (finput));
1250 error ("numeric constant with no digits");
1252 if (largest_digit >= base)
1253 error ("numeric constant contains digits beyond the radix");
1255 /* Remove terminating char from the token buffer and delimit the string */
1258 if (floatflag != NOT_FLOAT)
1260 tree type = double_type_node;
1261 int garbage_chars = 0, exceeds_double = 0;
1263 REAL_VALUE_TYPE value;
1266 /* Read explicit exponent if any, and put it in tokenbuf. */
1268 if ((c == 'e') || (c == 'E'))
1270 if (p >= token_buffer + maxtoken - 3)
1271 p = extend_token_buffer (p);
1274 if ((c == '+') || (c == '-'))
1280 error ("floating constant exponent has no digits");
1283 if (p >= token_buffer + maxtoken - 3)
1284 p = extend_token_buffer (p);
1293 /* Convert string to a double, checking for overflow. */
1294 if (setjmp (handler))
1296 error ("floating constant out of range");
1301 int fflag = 0, lflag = 0;
1302 /* Copy token_buffer now, while it has just the number
1303 and not the suffixes; once we add `f' or `i',
1304 REAL_VALUE_ATOF may not work any more. */
1305 char *copy = (char *) alloca (p - token_buffer + 1);
1306 bcopy (token_buffer, copy, p - token_buffer + 1);
1308 set_float_handler (handler);
1314 /* Read the suffixes to choose a data type. */
1319 error ("more than one `f' in numeric constant");
1325 error ("more than one `l' in numeric constant");
1331 error ("more than one `i' or `j' in numeric constant");
1342 if (p >= token_buffer + maxtoken - 3)
1343 p = extend_token_buffer (p);
1349 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1350 tells the desired precision of the binary result
1351 of decimal-to-binary conversion. */
1356 error ("both `f' and `l' in floating constant");
1358 type = float_type_node;
1359 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1360 if (TARGET_FLOAT_FORMAT != IEEE_FLOAT_FORMAT
1361 && REAL_VALUE_ISINF (value) && pedantic)
1362 pedwarn ("floating point number exceeds range of `float'");
1366 type = long_double_type_node;
1367 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1368 if (TARGET_FLOAT_FORMAT != IEEE_FLOAT_FORMAT
1369 && REAL_VALUE_ISINF (value) && pedantic)
1370 pedwarn ("floating point number exceeds range of `long double'");
1374 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1375 if (TARGET_FLOAT_FORMAT != IEEE_FLOAT_FORMAT
1376 && REAL_VALUE_ISINF (value) && pedantic)
1377 pedwarn ("floating point number exceeds range of `double'");
1380 set_float_handler (NULL_PTR);
1383 if (errno == ERANGE && !flag_traditional && pedantic)
1385 /* ERANGE is also reported for underflow,
1386 so test the value to distinguish overflow from that. */
1387 if (TARGET_FLOAT_FORMAT != IEEE_FLOAT_FORMAT
1388 && (REAL_VALUES_LESS (dconst1, value)
1389 || REAL_VALUES_LESS (value, dconstm1)))
1391 pedwarn ("floating point number exceeds range of `double'");
1397 while (isalnum (c) || c == '.' || c == '_'
1398 || (!flag_traditional && (c == '+' || c == '-')
1399 && (p[-1] == 'e' || p[-1] == 'E')))
1401 if (p >= token_buffer + maxtoken - 3)
1402 p = extend_token_buffer (p);
1407 if (garbage_chars > 0)
1408 error ("garbage at end of number");
1410 /* If the result is not a number, assume it must have been
1411 due to some error message above, so silently convert
1413 if (REAL_VALUE_ISNAN (value))
1416 /* Create a node with determined type and value. */
1418 yylval.ttype = build_complex (convert (type, integer_zero_node),
1419 build_real (type, value));
1421 yylval.ttype = build_real (type, value);
1428 tree traditional_type, ansi_type, type;
1429 HOST_WIDE_INT high, low;
1430 int spec_unsigned = 0;
1432 int spec_long_long = 0;
1438 if (c == 'u' || c == 'U')
1441 error ("two `u's in integer constant");
1444 else if (c == 'l' || c == 'L')
1449 error ("three `l's in integer constant");
1451 pedwarn ("ANSI C forbids long long integer constants");
1456 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1459 error ("more than one `i' or `j' in numeric constant");
1464 if (isalnum (c) || c == '.' || c == '_'
1465 || (!flag_traditional && (c == '+' || c == '-')
1466 && (p[-1] == 'e' || p[-1] == 'E')))
1468 error ("garbage at end of number");
1469 while (isalnum (c) || c == '.' || c == '_'
1470 || (!flag_traditional && (c == '+' || c == '-')
1471 && (p[-1] == 'e' || p[-1] == 'E')))
1473 if (p >= token_buffer + maxtoken - 3)
1474 p = extend_token_buffer (p);
1481 if (p >= token_buffer + maxtoken - 3)
1482 p = extend_token_buffer (p);
1489 /* If the constant is not long long and it won't fit in an
1490 unsigned long, or if the constant is long long and won't fit
1491 in an unsigned long long, then warn that the constant is out
1494 /* ??? This assumes that long long and long integer types are
1495 a multiple of 8 bits. This better than the original code
1496 though which assumed that long was exactly 32 bits and long
1497 long was exactly 64 bits. */
1500 bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1502 bytes = TYPE_PRECISION (long_integer_type_node) / 8;
1505 for (i = bytes; i < TOTAL_PARTS; i++)
1509 pedwarn ("integer constant out of range");
1511 /* This is simplified by the fact that our constant
1512 is always positive. */
1516 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1518 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1519 / HOST_BITS_PER_CHAR)]
1520 << (i * HOST_BITS_PER_CHAR));
1521 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1524 yylval.ttype = build_int_2 (low, high);
1525 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1527 /* If warn_traditional, calculate both the ANSI type and the
1528 traditional type, then see if they disagree.
1529 Otherwise, calculate only the type for the dialect in use. */
1530 if (warn_traditional || flag_traditional)
1532 /* Calculate the traditional type. */
1533 /* Traditionally, any constant is signed;
1534 but if unsigned is specified explicitly, obey that.
1535 Use the smallest size with the right number of bits,
1536 except for one special case with decimal constants. */
1537 if (! spec_long && base != 10
1538 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1539 traditional_type = (spec_unsigned ? unsigned_type_node
1540 : integer_type_node);
1541 /* A decimal constant must be long
1542 if it does not fit in type int.
1543 I think this is independent of whether
1544 the constant is signed. */
1545 else if (! spec_long && base == 10
1546 && int_fits_type_p (yylval.ttype, integer_type_node))
1547 traditional_type = (spec_unsigned ? unsigned_type_node
1548 : integer_type_node);
1549 else if (! spec_long_long)
1550 traditional_type = (spec_unsigned ? long_unsigned_type_node
1551 : long_integer_type_node);
1553 traditional_type = (spec_unsigned
1554 ? long_long_unsigned_type_node
1555 : long_long_integer_type_node);
1557 if (warn_traditional || ! flag_traditional)
1559 /* Calculate the ANSI type. */
1560 if (! spec_long && ! spec_unsigned
1561 && int_fits_type_p (yylval.ttype, integer_type_node))
1562 ansi_type = integer_type_node;
1563 else if (! spec_long && (base != 10 || spec_unsigned)
1564 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1565 ansi_type = unsigned_type_node;
1566 else if (! spec_unsigned && !spec_long_long
1567 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1568 ansi_type = long_integer_type_node;
1569 else if (! spec_long_long)
1570 ansi_type = long_unsigned_type_node;
1571 else if (! spec_unsigned
1572 /* Verify value does not overflow into sign bit. */
1573 && TREE_INT_CST_HIGH (yylval.ttype) >= 0
1574 && int_fits_type_p (yylval.ttype,
1575 long_long_integer_type_node))
1576 ansi_type = long_long_integer_type_node;
1578 ansi_type = long_long_unsigned_type_node;
1581 type = flag_traditional ? traditional_type : ansi_type;
1583 if (warn_traditional && traditional_type != ansi_type)
1585 if (TYPE_PRECISION (traditional_type)
1586 != TYPE_PRECISION (ansi_type))
1587 warning ("width of integer constant changes with -traditional");
1588 else if (TREE_UNSIGNED (traditional_type)
1589 != TREE_UNSIGNED (ansi_type))
1590 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1592 warning ("width of integer constant may change on other systems with -traditional");
1595 if (!flag_traditional && !int_fits_type_p (yylval.ttype, type)
1597 pedwarn ("integer constant out of range");
1599 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1600 warning ("decimal constant is so large that it is unsigned");
1604 if (TYPE_PRECISION (type)
1605 <= TYPE_PRECISION (integer_type_node))
1607 = build_complex (integer_zero_node,
1608 convert (integer_type_node, yylval.ttype));
1610 error ("complex integer constant is too wide for `complex int'");
1612 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1613 /* The traditional constant 0x80000000 is signed
1614 but doesn't fit in the range of int.
1615 This will change it to -0x80000000, which does fit. */
1617 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1618 yylval.ttype = convert (type, yylval.ttype);
1619 TREE_OVERFLOW (yylval.ttype)
1620 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1623 TREE_TYPE (yylval.ttype) = type;
1628 value = CONSTANT; break;
1634 register int result = 0;
1635 register int num_chars = 0;
1636 unsigned width = TYPE_PRECISION (char_type_node);
1641 width = WCHAR_TYPE_SIZE;
1642 #ifdef MULTIBYTE_CHARS
1643 max_chars = MB_CUR_MAX;
1649 max_chars = TYPE_PRECISION (integer_type_node) / width;
1657 if (c == '\'' || c == EOF)
1663 c = readescape (&ignore);
1666 if (width < HOST_BITS_PER_INT
1667 && (unsigned) c >= (1 << width))
1668 pedwarn ("escape sequence out of range for character");
1669 #ifdef MAP_CHARACTER
1671 c = MAP_CHARACTER (c);
1677 pedwarn ("ANSI C forbids newline in character constant");
1680 #ifdef MAP_CHARACTER
1682 c = MAP_CHARACTER (c);
1686 if (num_chars > maxtoken - 4)
1687 extend_token_buffer (token_buffer);
1689 token_buffer[num_chars] = c;
1691 /* Merge character into result; ignore excess chars. */
1692 if (num_chars < max_chars + 1)
1694 if (width < HOST_BITS_PER_INT)
1695 result = (result << width) | (c & ((1 << width) - 1));
1701 token_buffer[num_chars + 1] = '\'';
1702 token_buffer[num_chars + 2] = 0;
1705 error ("malformatted character constant");
1706 else if (num_chars == 0)
1707 error ("empty character constant");
1708 else if (num_chars > max_chars)
1710 num_chars = max_chars;
1711 error ("character constant too long");
1713 else if (num_chars != 1 && ! flag_traditional)
1714 warning ("multi-character character constant");
1716 /* If char type is signed, sign-extend the constant. */
1719 int num_bits = num_chars * width;
1721 /* We already got an error; avoid invalid shift. */
1722 yylval.ttype = build_int_2 (0, 0);
1723 else if (TREE_UNSIGNED (char_type_node)
1724 || ((result >> (num_bits - 1)) & 1) == 0)
1726 = build_int_2 (result & ((unsigned HOST_WIDE_INT) ~0
1727 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1731 = build_int_2 (result | ~((unsigned HOST_WIDE_INT) ~0
1732 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1734 TREE_TYPE (yylval.ttype) = integer_type_node;
1738 #ifdef MULTIBYTE_CHARS
1739 /* Set the initial shift state and convert the next sequence. */
1741 /* In all locales L'\0' is zero and mbtowc will return zero,
1744 || (num_chars == 1 && token_buffer[1] != '\0'))
1747 (void) mbtowc (NULL_PTR, NULL_PTR, 0);
1748 if (mbtowc (& wc, token_buffer + 1, num_chars) == num_chars)
1751 warning ("Ignoring invalid multibyte character");
1754 yylval.ttype = build_int_2 (result, 0);
1755 TREE_TYPE (yylval.ttype) = wchar_type_node;
1766 p = token_buffer + 1;
1768 while (c != '"' && c >= 0)
1773 c = readescape (&ignore);
1777 && TYPE_PRECISION (char_type_node) < HOST_BITS_PER_INT
1778 && c >= (1 << TYPE_PRECISION (char_type_node)))
1779 pedwarn ("escape sequence out of range for character");
1784 pedwarn ("ANSI C forbids newline in string constant");
1788 if (p == token_buffer + maxtoken)
1789 p = extend_token_buffer (p);
1797 /* We have read the entire constant.
1798 Construct a STRING_CST for the result. */
1802 /* If this is a L"..." wide-string, convert the multibyte string
1803 to a wide character string. */
1804 char *widep = (char *) alloca ((p - token_buffer) * WCHAR_BYTES);
1807 #ifdef MULTIBYTE_CHARS
1808 len = mbstowcs ((wchar_t *) widep, token_buffer + 1, p - token_buffer);
1809 if (len < 0 || len >= (p - token_buffer))
1811 warning ("Ignoring invalid multibyte string");
1814 bzero (widep + (len * WCHAR_BYTES), WCHAR_BYTES);
1817 union { long l; char c[sizeof (long)]; } u;
1821 /* Determine whether host is little or big endian. */
1823 big_endian = u.c[sizeof (long) - 1];
1824 wp = widep + (big_endian ? WCHAR_BYTES - 1 : 0);
1826 bzero (widep, (p - token_buffer) * WCHAR_BYTES);
1827 for (cp = token_buffer + 1; cp < p; cp++)
1828 *wp = *cp, wp += WCHAR_BYTES;
1829 len = p - token_buffer - 1;
1832 yylval.ttype = build_string ((len + 1) * WCHAR_BYTES, widep);
1833 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
1838 extern tree build_objc_string();
1839 /* Return an Objective-C @"..." constant string object. */
1840 yylval.ttype = build_objc_string (p - token_buffer,
1842 TREE_TYPE (yylval.ttype) = char_array_type_node;
1843 value = OBJC_STRING;
1847 yylval.ttype = build_string (p - token_buffer, token_buffer + 1);
1848 TREE_TYPE (yylval.ttype) = char_array_type_node;
1878 yylval.code = PLUS_EXPR; break;
1880 yylval.code = MINUS_EXPR; break;
1882 yylval.code = BIT_AND_EXPR; break;
1884 yylval.code = BIT_IOR_EXPR; break;
1886 yylval.code = MULT_EXPR; break;
1888 yylval.code = TRUNC_DIV_EXPR; break;
1890 yylval.code = TRUNC_MOD_EXPR; break;
1892 yylval.code = BIT_XOR_EXPR; break;
1894 yylval.code = LSHIFT_EXPR; break;
1896 yylval.code = RSHIFT_EXPR; break;
1898 yylval.code = LT_EXPR; break;
1900 yylval.code = GT_EXPR; break;
1903 token_buffer[1] = c1 = getc (finput);
1904 token_buffer[2] = 0;
1911 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
1913 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
1915 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
1917 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
1919 value = ASSIGN; goto done;
1925 value = PLUSPLUS; goto done;
1927 value = MINUSMINUS; goto done;
1929 value = ANDAND; goto done;
1931 value = OROR; goto done;
1939 else if ((c == '-') && (c1 == '>'))
1940 { value = POINTSAT; goto done; }
1941 ungetc (c1, finput);
1942 token_buffer[1] = 0;
1944 if ((c == '<') || (c == '>'))
1945 value = ARITHCOMPARE;
1951 /* Don't make yyparse think this is eof. */
1960 /* yylloc.last_line = lineno; */
1965 /* Sets the value of the 'yydebug' variable to VALUE.
1966 This is a function so we don't have to have YYDEBUG defined
1967 in order to build the compiler. */
1976 warning ("YYDEBUG not defined.");