1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 88, 89, 92, 94-97, 1998 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. */
37 /* MULTIBYTE_CHARS support only works for native compilers.
38 ??? Ideally what we want is to model widechar support after
39 the current floating point support. */
41 #undef MULTIBYTE_CHARS
44 #ifdef MULTIBYTE_CHARS
47 #endif /* MULTIBYTE_CHARS */
51 extern cpp_reader parse_in;
52 extern cpp_options parse_options;
54 /* Stream for reading from the input file. */
58 extern void yyprint PROTO((FILE *, int, YYSTYPE));
60 /* The elements of `ridpointers' are identifier nodes
61 for the reserved type names and storage classes.
62 It is indexed by a RID_... value. */
63 tree ridpointers[(int) RID_MAX];
65 /* Cause the `yydebug' variable to be defined. */
69 extern unsigned char *yy_cur, *yy_lim;
71 extern int yy_get_token ();
73 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
74 #define UNGETC(c) ((c) == EOF ? 0 : yy_cur--)
76 #define GETC() getc (finput)
77 #define UNGETC(c) ungetc (c, finput)
80 /* the declaration found for the last IDENTIFIER token read in.
81 yylex must look this up to detect typedefs, which get token type TYPENAME,
82 so it is left around in case the identifier is not a typedef but is
83 used in a context which makes it a reference to a variable. */
86 /* Nonzero enables objc features. */
92 /* File used for outputting assembler code. */
93 extern FILE *asm_out_file;
95 #ifndef WCHAR_TYPE_SIZE
97 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
99 #define WCHAR_TYPE_SIZE BITS_PER_WORD
103 /* Number of bytes in a wide character. */
104 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
106 static int maxtoken; /* Current nominal length of token buffer. */
107 char *token_buffer; /* Pointer to token buffer.
108 Actual allocated length is maxtoken + 2.
109 This is not static because objc-parse.y uses it. */
111 static int indent_level = 0; /* Number of { minus number of }. */
113 /* Nonzero if end-of-file has been seen on input. */
114 static int end_of_file;
117 /* Buffered-back input character; faster than using ungetc. */
118 static int nextchar = -1;
121 #ifdef HANDLE_GENERIC_PRAGMAS
122 static int handle_generic_pragma PROTO((int));
123 #endif /* HANDLE_GENERIC_PRAGMAS */
124 static int whitespace_cr PROTO((int));
125 static int skip_white_space PROTO((int));
126 static int skip_white_space_on_line PROTO((void));
127 static char *extend_token_buffer PROTO((const char *));
128 static int readescape PROTO((int *));
130 /* Do not insert generated code into the source, instead, include it.
131 This allows us to build gcc automatically even for targets that
132 need to add or modify the reserved keyword lists. */
135 /* Return something to represent absolute declarators containing a *.
136 TARGET is the absolute declarator that the * contains.
137 TYPE_QUALS is a list of modifiers such as const or volatile
138 to apply to the pointer type, represented as identifiers.
140 We return an INDIRECT_REF whose "contents" are TARGET
141 and whose type is the modifier list. */
144 make_pointer_declarator (type_quals, target)
145 tree type_quals, target;
147 return build1 (INDIRECT_REF, type_quals, target);
151 forget_protocol_qualifiers ()
153 int i, n = sizeof wordlist / sizeof (struct resword);
155 for (i = 0; i < n; i++)
156 if ((int) wordlist[i].rid >= (int) RID_IN
157 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
158 wordlist[i].name = "";
162 remember_protocol_qualifiers ()
164 int i, n = sizeof wordlist / sizeof (struct resword);
166 for (i = 0; i < n; i++)
167 if (wordlist[i].rid == RID_IN)
168 wordlist[i].name = "in";
169 else if (wordlist[i].rid == RID_OUT)
170 wordlist[i].name = "out";
171 else if (wordlist[i].rid == RID_INOUT)
172 wordlist[i].name = "inout";
173 else if (wordlist[i].rid == RID_BYCOPY)
174 wordlist[i].name = "bycopy";
175 else if (wordlist[i].rid == RID_BYREF)
176 wordlist[i].name = "byref";
177 else if (wordlist[i].rid == RID_ONEWAY)
178 wordlist[i].name = "oneway";
182 init_parse (filename)
186 /* Open input file. */
187 if (filename == 0 || !strcmp (filename, "-"))
193 finput = fopen (filename, "r");
195 pfatal_with_name (filename);
197 #ifdef IO_BUFFER_SIZE
198 setvbuf (finput, (char *) xmalloc (IO_BUFFER_SIZE), _IOFBF, IO_BUFFER_SIZE);
200 #else /* !USE_CPPLIB */
201 parse_in.show_column = 1;
202 if (! cpp_start_read (&parse_in, filename))
205 if (filename == 0 || !strcmp (filename, "-"))
208 /* cpp_start_read always puts at least one line directive into the
209 token buffer. We must arrange to read it out here. */
210 yy_cur = parse_in.token_buffer;
211 yy_lim = CPP_PWRITTEN (&parse_in);
223 cpp_finish (&parse_in);
232 /* Make identifier nodes long enough for the language-specific slots. */
233 set_identifier_size (sizeof (struct lang_identifier));
235 /* Start it at 0, because check_newline is called at the very beginning
236 and will increment it to 1. */
239 #ifdef MULTIBYTE_CHARS
240 /* Change to the native locale for multibyte conversions. */
241 setlocale (LC_CTYPE, "");
242 literal_codeset = getenv ("LANG");
246 token_buffer = (char *) xmalloc (maxtoken + 2);
248 ridpointers[(int) RID_INT] = get_identifier ("int");
249 ridpointers[(int) RID_CHAR] = get_identifier ("char");
250 ridpointers[(int) RID_VOID] = get_identifier ("void");
251 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
252 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
253 ridpointers[(int) RID_SHORT] = get_identifier ("short");
254 ridpointers[(int) RID_LONG] = get_identifier ("long");
255 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
256 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
257 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
258 ridpointers[(int) RID_CONST] = get_identifier ("const");
259 ridpointers[(int) RID_RESTRICT] = get_identifier ("restrict");
260 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
261 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
262 ridpointers[(int) RID_STATIC] = get_identifier ("static");
263 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
264 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
265 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
266 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
267 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
268 ridpointers[(int) RID_ID] = get_identifier ("id");
269 ridpointers[(int) RID_IN] = get_identifier ("in");
270 ridpointers[(int) RID_OUT] = get_identifier ("out");
271 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
272 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
273 ridpointers[(int) RID_BYREF] = get_identifier ("byref");
274 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
275 forget_protocol_qualifiers();
277 /* Some options inhibit certain reserved words.
278 Clear those words out of the hash table so they won't be recognized. */
279 #define UNSET_RESERVED_WORD(STRING) \
280 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
281 if (s) s->name = ""; } while (0)
283 if (! doing_objc_thang)
284 UNSET_RESERVED_WORD ("id");
286 if (flag_traditional)
288 UNSET_RESERVED_WORD ("const");
289 UNSET_RESERVED_WORD ("restrict");
290 UNSET_RESERVED_WORD ("volatile");
291 UNSET_RESERVED_WORD ("typeof");
292 UNSET_RESERVED_WORD ("signed");
293 UNSET_RESERVED_WORD ("inline");
294 UNSET_RESERVED_WORD ("iterator");
295 UNSET_RESERVED_WORD ("complex");
297 else if (!flag_isoc9x)
298 UNSET_RESERVED_WORD ("restrict");
302 UNSET_RESERVED_WORD ("asm");
303 UNSET_RESERVED_WORD ("typeof");
304 UNSET_RESERVED_WORD ("inline");
305 UNSET_RESERVED_WORD ("iterator");
306 UNSET_RESERVED_WORD ("complex");
311 reinit_parse_for_function ()
315 /* Function used when yydebug is set, to print a token in more detail. */
318 yyprint (file, yychar, yylval)
330 if (IDENTIFIER_POINTER (t))
331 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
336 if (TREE_CODE (t) == INTEGER_CST)
338 #if HOST_BITS_PER_WIDE_INT == 64
339 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
342 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
349 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
355 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
360 /* Iff C is a carriage return, warn about it - if appropriate -
361 and return nonzero. */
366 static int newline_warning = 0;
370 /* ANSI C says the effects of a carriage return in a source file
372 if (pedantic && !newline_warning)
374 warning ("carriage return in source file");
375 warning ("(we only warn about the first carriage return)");
383 /* If C is not whitespace, return C.
384 Otherwise skip whitespace and return first nonwhite char read. */
394 /* We don't recognize comments here, because
395 cpp output can include / and * consecutively as operators.
396 Also, there's no need, since cpp removes all comments. */
399 c = check_newline ();
420 error ("stray '\\' in program");
430 /* Skips all of the white space at the current location in the input file.
431 Must use and reset nextchar if it has the next character. */
434 position_after_white_space ()
440 c = nextchar, nextchar = -1;
445 UNGETC (skip_white_space (c));
448 /* Like skip_white_space, but don't advance beyond the end of line.
449 Moreover, we don't get passed a character to start with. */
451 skip_white_space_on_line ()
480 /* Make the token buffer longer, preserving the data in it.
481 P should point to just beyond the last valid character in the old buffer.
482 The value we return is a pointer to the new buffer
483 at a place corresponding to P. */
486 extend_token_buffer (p)
489 int offset = p - token_buffer;
491 maxtoken = maxtoken * 2 + 10;
492 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
494 return token_buffer + offset;
497 #if defined HANDLE_PRAGMA
498 /* Local versions of these macros, that can be passed as function pointers. */
513 /* At the beginning of a line, increment the line number
514 and process any #-directive on this line.
515 If the line is a #-directive, read the entire line and return a newline.
516 Otherwise, return the line's first non-whitespace character. */
526 /* Read first nonwhite char on the line. */
529 while (c == ' ' || c == '\t')
534 /* If not #, return it so caller will use it. */
538 /* Read first nonwhite char after the `#'. */
541 while (c == ' ' || c == '\t')
544 /* If a letter follows, then if the word here is `line', skip
545 it and ignore it; otherwise, ignore the line, with an error
546 if the word isn't `pragma', `ident', `define', or `undef'. */
548 if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
557 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'
558 || whitespace_cr (c) ))
560 while (c == ' ' || c == '\t' || whitespace_cr (c))
565 #if defined HANDLE_PRAGMA || defined HANDLE_GENERIC_PRAGMAS
568 if (token != IDENTIFIER)
570 #endif /* HANDLE_PRAGMA || HANDLE_GENERIC_PRAGMAS */
573 /* We invoke HANDLE_PRAGMA before HANDLE_GENERIC_PRAGMAS (if
574 both are defined), in order to give the back end a chance to
575 override the interpretation of generic style pragmas. */
579 c = nextchar, nextchar = -1;
582 #endif /* !USE_CPPLIB */
584 if (TREE_CODE (yylval.ttype) != IDENTIFIER_NODE)
587 if (HANDLE_PRAGMA (pragma_getc, pragma_ungetc,
588 IDENTIFIER_POINTER (yylval.ttype)))
590 #endif /* HANDLE_PRAGMA */
592 #ifdef HANDLE_GENERIC_PRAGMAS
593 if (handle_generic_pragma (token))
595 #endif /* HANDLE_GENERIC_PRAGMAS */
597 /* Issue a warning message if we have been asked to do so.
598 Ignoring unknown pragmas in system header file unless
599 an explcit -Wunknown-pragmas has been given. */
600 if (warn_unknown_pragmas > 1
601 || (warn_unknown_pragmas && ! in_system_header))
602 warning ("ignoring pragma: %s", token_buffer);
615 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
618 debug_define (lineno, GET_DIRECTIVE_LINE ());
628 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
631 debug_undef (lineno, GET_DIRECTIVE_LINE ());
640 && ((c = GETC()) == ' ' || c == '\t'))
649 && ((c = GETC()) == ' ' || c == '\t'))
651 /* #ident. The pedantic warning is now in cccp.c. */
653 /* Here we have just seen `#ident '.
654 A string constant should follow. */
656 c = skip_white_space_on_line ();
658 /* If no argument, ignore the line. */
665 || TREE_CODE (yylval.ttype) != STRING_CST)
667 error ("invalid #ident");
673 #ifdef ASM_OUTPUT_IDENT
674 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
678 /* Skip the rest of this line. */
683 error ("undefined or invalid # directive");
688 /* Here we have either `#line' or `# <nonletter>'.
689 In either case, it should be a line number; a digit should follow. */
691 /* Can't use skip_white_space here, but must handle all whitespace
692 that is not '\n', lest we get a recursion for '\r' '\n' when
695 c = skip_white_space_on_line ();
697 /* If the # is the only nonwhite char on the line,
698 just ignore it. Check the new newline. */
702 /* Something follows the #; read a token. */
707 if (token == CONSTANT
708 && TREE_CODE (yylval.ttype) == INTEGER_CST)
710 int old_lineno = lineno;
712 /* subtract one, because it is the following line that
713 gets the specified number */
715 int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
717 /* Is this the last nonwhite stuff on the line? */
718 c = skip_white_space_on_line ();
721 /* No more: store the line number and check following line. */
727 /* More follows: it must be a string constant (filename). */
729 /* Read the string constant. */
732 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
734 error ("invalid #line");
739 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
740 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
743 /* Each change of file name
744 reinitializes whether we are now in a system header. */
745 in_system_header = 0;
747 if (main_input_filename == 0)
748 main_input_filename = input_filename;
750 /* Is this the last nonwhite stuff on the line? */
751 c = skip_white_space_on_line ();
754 /* Update the name in the top element of input_file_stack. */
755 if (input_file_stack)
756 input_file_stack->name = input_filename;
765 /* `1' after file name means entering new file.
766 `2' after file name means just left a file. */
768 if (token == CONSTANT
769 && TREE_CODE (yylval.ttype) == INTEGER_CST)
771 if (TREE_INT_CST_LOW (yylval.ttype) == 1)
773 /* Pushing to a new file. */
775 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
776 input_file_stack->line = old_lineno;
777 p->next = input_file_stack;
778 p->name = input_filename;
779 p->indent_level = indent_level;
780 input_file_stack = p;
781 input_file_stack_tick++;
782 debug_start_source_file (input_filename);
785 else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
787 /* Popping out of a file. */
788 if (input_file_stack->next)
790 struct file_stack *p = input_file_stack;
791 if (indent_level != p->indent_level)
793 warning_with_file_and_line
794 (p->name, old_lineno,
795 "This file contains more `%c's than `%c's.",
796 indent_level > p->indent_level ? '{' : '}',
797 indent_level > p->indent_level ? '}' : '{');
799 input_file_stack = p->next;
801 input_file_stack_tick++;
802 debug_end_source_file (input_file_stack->line);
805 error ("#-lines for entering and leaving files don't match");
811 /* Now that we've pushed or popped the input stack,
812 update the name in the top element. */
813 if (input_file_stack)
814 input_file_stack->name = input_filename;
816 /* If we have handled a `1' or a `2',
817 see if there is another number to read. */
820 /* Is this the last nonwhite stuff on the line? */
821 c = skip_white_space_on_line ();
830 /* `3' after file name means this is a system header file. */
832 if (token == CONSTANT
833 && TREE_CODE (yylval.ttype) == INTEGER_CST
834 && TREE_INT_CST_LOW (yylval.ttype) == 3)
835 in_system_header = 1, used_up = 1;
839 /* Is this the last nonwhite stuff on the line? */
840 c = skip_white_space_on_line ();
846 warning ("unrecognized text at end of #line");
849 error ("invalid #-line");
851 /* skip the rest of this line. */
854 if (c != '\n' && c != EOF && nextchar >= 0)
855 c = nextchar, nextchar = -1;
857 while (c != '\n' && c != EOF)
862 #ifdef HANDLE_GENERIC_PRAGMAS
864 /* Handle a #pragma directive.
865 TOKEN is the token we read after `#pragma'. Processes the entire input
866 line and return non-zero iff the pragma has been successfully parsed. */
868 /* This function has to be in this file, in order to get at
872 handle_generic_pragma (token)
885 handle_pragma_token (token_buffer, yylval.ttype);
888 handle_pragma_token (token_buffer, NULL);
892 c = nextchar, nextchar = -1;
897 while (c == ' ' || c == '\t')
901 if (c == '\n' || c == EOF)
902 return handle_pragma_token (NULL, NULL);
908 #endif /* HANDLE_GENERIC_PRAGMAS */
910 #define ENDFILE -1 /* token that represents end-of-file */
912 /* Read an escape sequence, returning its equivalent as a character,
913 or store 1 in *ignore_ptr if it is backslash-newline. */
916 readescape (ignore_ptr)
919 register int c = GETC();
921 register unsigned count;
922 unsigned firstdig = 0;
928 if (warn_traditional)
929 warning ("the meaning of `\\x' varies with -traditional");
931 if (flag_traditional)
940 if (!(c >= 'a' && c <= 'f')
941 && !(c >= 'A' && c <= 'F')
942 && !(c >= '0' && c <= '9'))
948 if (c >= 'a' && c <= 'f')
949 code += c - 'a' + 10;
950 if (c >= 'A' && c <= 'F')
951 code += c - 'A' + 10;
952 if (c >= '0' && c <= '9')
954 if (code != 0 || count != 0)
963 error ("\\x used with no following hex digits");
965 /* Digits are all 0's. Ok. */
967 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
969 && (((unsigned)1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
971 pedwarn ("hex escape out of range");
974 case '0': case '1': case '2': case '3': case '4':
975 case '5': case '6': case '7':
978 while ((c <= '7') && (c >= '0') && (count++ < 3))
980 code = (code * 8) + (c - '0');
986 case '\\': case '\'': case '"':
995 return TARGET_NEWLINE;
1010 if (warn_traditional)
1011 warning ("the meaning of `\\a' varies with -traditional");
1013 if (flag_traditional)
1018 #if 0 /* Vertical tab is present in common usage compilers. */
1019 if (flag_traditional)
1027 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
1033 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1037 /* `\%' is used to prevent SCCS from getting confused. */
1040 pedwarn ("non-ANSI escape sequence `\\%c'", c);
1043 if (c >= 040 && c < 0177)
1044 pedwarn ("unknown escape sequence `\\%c'", c);
1046 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
1054 const char *string = _(msgid);
1056 /* We can't print string and character constants well
1057 because the token_buffer contains the result of processing escapes. */
1059 error ("%s at end of input", string);
1060 else if (token_buffer[0] == 0)
1061 error ("%s at null character", string);
1062 else if (token_buffer[0] == '"')
1063 error ("%s before string constant", string);
1064 else if (token_buffer[0] == '\'')
1065 error ("%s before character constant", string);
1066 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
1067 error ("%s before character 0%o", string, (unsigned char) token_buffer[0]);
1069 error ("%s before `%s'", string, token_buffer);
1079 char long_long_flag;
1082 struct try_type type_sequence[] =
1084 { &integer_type_node, 0, 0, 0},
1085 { &unsigned_type_node, 1, 0, 0},
1086 { &long_integer_type_node, 0, 1, 0},
1087 { &long_unsigned_type_node, 1, 1, 0},
1088 { &long_long_integer_type_node, 0, 1, 1},
1089 { &long_long_unsigned_type_node, 1, 1, 1}
1104 c = nextchar, nextchar = -1;
1109 /* Effectively do c = skip_white_space (c)
1110 but do it faster in the usual cases. */
1123 /* Call skip_white_space so we can warn if appropriate. */
1128 c = skip_white_space (c);
1130 goto found_nonwhite;
1134 token_buffer[0] = c;
1135 token_buffer[1] = 0;
1137 /* yylloc.first_line = lineno; */
1143 token_buffer[0] = 0;
1148 /* Capital L may start a wide-string or wide-character constant. */
1150 register int c = GETC();
1159 goto string_constant;
1166 if (!doing_objc_thang)
1173 /* '@' may start a constant string object. */
1174 register int c = GETC ();
1178 goto string_constant;
1181 /* Fall through to treat '@' as the start of an identifier. */
1184 case 'A': case 'B': case 'C': case 'D': case 'E':
1185 case 'F': case 'G': case 'H': case 'I': case 'J':
1186 case 'K': case 'M': case 'N': case 'O':
1187 case 'P': case 'Q': case 'R': case 'S': case 'T':
1188 case 'U': case 'V': case 'W': case 'X': case 'Y':
1190 case 'a': case 'b': case 'c': case 'd': case 'e':
1191 case 'f': case 'g': case 'h': case 'i': case 'j':
1192 case 'k': case 'l': case 'm': case 'n': case 'o':
1193 case 'p': case 'q': case 'r': case 's': case 't':
1194 case 'u': case 'v': case 'w': case 'x': case 'y':
1200 while (ISALNUM (c) || c == '_' || c == '$' || c == '@')
1202 /* Make sure this char really belongs in an identifier. */
1205 if (! dollars_in_ident)
1206 error ("`$' in identifier");
1208 pedwarn ("`$' in identifier");
1211 if (p >= token_buffer + maxtoken)
1212 p = extend_token_buffer (p);
1228 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1231 register struct resword *ptr;
1233 if ((ptr = is_reserved_word (token_buffer, p - token_buffer)))
1236 yylval.ttype = ridpointers[(int) ptr->rid];
1237 value = (int) ptr->token;
1239 /* Only return OBJECTNAME if it is a typedef. */
1240 if (doing_objc_thang && value == OBJECTNAME)
1242 lastiddecl = lookup_name(yylval.ttype);
1244 if (lastiddecl == NULL_TREE
1245 || TREE_CODE (lastiddecl) != TYPE_DECL)
1249 /* Even if we decided to recognize asm, still perhaps warn. */
1251 && (value == ASM_KEYWORD || value == TYPEOF
1252 || ptr->rid == RID_INLINE)
1253 && token_buffer[0] != '_')
1254 pedwarn ("ANSI does not permit the keyword `%s'",
1259 /* If we did not find a keyword, look for an identifier
1262 if (value == IDENTIFIER)
1264 if (token_buffer[0] == '@')
1265 error("invalid identifier `%s'", token_buffer);
1267 yylval.ttype = get_identifier (token_buffer);
1268 lastiddecl = lookup_name (yylval.ttype);
1270 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1272 /* A user-invisible read-only initialized variable
1273 should be replaced by its value.
1274 We handle only strings since that's the only case used in C. */
1275 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1276 && DECL_IGNORED_P (lastiddecl)
1277 && TREE_READONLY (lastiddecl)
1278 && DECL_INITIAL (lastiddecl) != 0
1279 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1281 tree stringval = DECL_INITIAL (lastiddecl);
1283 /* Copy the string value so that we won't clobber anything
1284 if we put something in the TREE_CHAIN of this one. */
1285 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1286 TREE_STRING_POINTER (stringval));
1289 else if (doing_objc_thang)
1291 tree objc_interface_decl = is_class_name (yylval.ttype);
1293 if (objc_interface_decl)
1296 yylval.ttype = objc_interface_decl;
1306 /* Check first for common special case: single-digit 0 or 1. */
1309 UNGETC (next_c); /* Always undo this lookahead. */
1310 if (!ISALNUM (next_c) && next_c != '.')
1312 token_buffer[0] = (char)c, token_buffer[1] = '\0';
1313 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1319 case '2': case '3': case '4':
1320 case '5': case '6': case '7': case '8': case '9':
1325 int largest_digit = 0;
1327 /* for multi-precision arithmetic,
1328 we actually store only HOST_BITS_PER_CHAR bits in each part.
1329 The number of parts is chosen so as to be sufficient to hold
1330 the enough bits to fit into the two HOST_WIDE_INTs that contain
1331 the integer value (this is always at least as many bits as are
1332 in a target `long long' value, but may be wider). */
1333 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1334 int parts[TOTAL_PARTS];
1337 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS, AFTER_EXPON}
1338 floatflag = NOT_FLOAT;
1340 for (count = 0; count < TOTAL_PARTS; count++)
1348 *p++ = (c = GETC());
1349 if ((c == 'x') || (c == 'X'))
1352 *p++ = (c = GETC());
1354 /* Leading 0 forces octal unless the 0 is the only digit. */
1355 else if (c >= '0' && c <= '9')
1364 /* Read all the digits-and-decimal-points. */
1367 || (ISALNUM (c) && c != 'l' && c != 'L'
1368 && c != 'u' && c != 'U'
1369 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1370 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1374 if (base == 16 && pedantic)
1375 error ("floating constant may not be in radix 16");
1376 if (floatflag == TOO_MANY_POINTS)
1377 /* We have already emitted an error. Don't need another. */
1379 else if (floatflag == AFTER_POINT || floatflag == AFTER_EXPON)
1381 error ("malformed floating constant");
1382 floatflag = TOO_MANY_POINTS;
1383 /* Avoid another error from atof by forcing all characters
1384 from here on to be ignored. */
1388 floatflag = AFTER_POINT;
1393 /* Accept '.' as the start of a floating-point number
1394 only when it is followed by a digit.
1395 Otherwise, unread the following non-digit
1396 and use the '.' as a structural token. */
1397 if (p == token_buffer + 2 && !ISDIGIT (c))
1408 error ("parse error at `..'");
1411 token_buffer[1] = 0;
1418 /* It is not a decimal point.
1419 It should be a digit (perhaps a hex digit). */
1425 else if (base <= 10)
1427 if (c == 'e' || c == 'E')
1430 floatflag = AFTER_EXPON;
1431 break; /* start of exponent */
1433 error ("nondigits in number and not hexadecimal");
1436 else if (base == 16 && (c == 'p' || c == 'P'))
1438 floatflag = AFTER_EXPON;
1439 break; /* start of exponent */
1449 if (c >= largest_digit)
1453 for (count = 0; count < TOTAL_PARTS; count++)
1455 parts[count] *= base;
1459 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1461 &= (1 << HOST_BITS_PER_CHAR) - 1;
1467 /* If the extra highest-order part ever gets anything in it,
1468 the number is certainly too big. */
1469 if (parts[TOTAL_PARTS - 1] != 0)
1472 if (p >= token_buffer + maxtoken - 3)
1473 p = extend_token_buffer (p);
1474 *p++ = (c = GETC());
1479 error ("numeric constant with no digits");
1481 if (largest_digit >= base)
1482 error ("numeric constant contains digits beyond the radix");
1484 /* Remove terminating char from the token buffer and delimit the string */
1487 if (floatflag != NOT_FLOAT)
1489 tree type = double_type_node;
1491 int conversion_errno = 0;
1492 REAL_VALUE_TYPE value;
1495 /* Read explicit exponent if any, and put it in tokenbuf. */
1497 if ((base == 10 && ((c == 'e') || (c == 'E')))
1498 || (base == 16 && (c == 'p' || c == 'P')))
1500 if (p >= token_buffer + maxtoken - 3)
1501 p = extend_token_buffer (p);
1504 if ((c == '+') || (c == '-'))
1509 /* Exponent is decimal, even if string is a hex float. */
1511 error ("floating constant exponent has no digits");
1514 if (p >= token_buffer + maxtoken - 3)
1515 p = extend_token_buffer (p);
1520 if (base == 16 && floatflag != AFTER_EXPON)
1521 error ("hexadecimal floating constant has no exponent");
1525 /* Convert string to a double, checking for overflow. */
1526 if (setjmp (handler))
1528 error ("floating constant out of range");
1533 int fflag = 0, lflag = 0;
1534 /* Copy token_buffer now, while it has just the number
1535 and not the suffixes; once we add `f' or `i',
1536 REAL_VALUE_ATOF may not work any more. */
1537 char *copy = (char *) alloca (p - token_buffer + 1);
1538 bcopy (token_buffer, copy, p - token_buffer + 1);
1540 set_float_handler (handler);
1546 /* Read the suffixes to choose a data type. */
1551 error ("more than one `f' in numeric constant");
1557 error ("more than one `l' in numeric constant");
1563 error ("more than one `i' or `j' in numeric constant");
1565 pedwarn ("ANSI C forbids imaginary numeric constants");
1576 if (p >= token_buffer + maxtoken - 3)
1577 p = extend_token_buffer (p);
1583 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1584 tells the desired precision of the binary result
1585 of decimal-to-binary conversion. */
1590 error ("both `f' and `l' in floating constant");
1592 type = float_type_node;
1595 value = REAL_VALUE_HTOF (copy, TYPE_MODE (type));
1597 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1598 conversion_errno = errno;
1599 /* A diagnostic is required here by some ANSI C testsuites.
1600 This is not pedwarn, because some people don't want
1601 an error for this. */
1602 if (REAL_VALUE_ISINF (value) && pedantic)
1603 warning ("floating point number exceeds range of `float'");
1607 type = long_double_type_node;
1610 value = REAL_VALUE_HTOF (copy, TYPE_MODE (type));
1612 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1613 conversion_errno = errno;
1614 if (REAL_VALUE_ISINF (value) && pedantic)
1615 warning ("floating point number exceeds range of `long double'");
1621 value = REAL_VALUE_HTOF (copy, TYPE_MODE (type));
1623 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1624 conversion_errno = errno;
1625 if (REAL_VALUE_ISINF (value) && pedantic)
1626 warning ("floating point number exceeds range of `double'");
1629 set_float_handler (NULL_PTR);
1632 /* ERANGE is also reported for underflow,
1633 so test the value to distinguish overflow from that. */
1634 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1635 && (REAL_VALUES_LESS (dconst1, value)
1636 || REAL_VALUES_LESS (value, dconstm1)))
1637 warning ("floating point number exceeds range of `double'");
1640 /* If the result is not a number, assume it must have been
1641 due to some error message above, so silently convert
1643 if (REAL_VALUE_ISNAN (value))
1646 /* Create a node with determined type and value. */
1648 yylval.ttype = build_complex (NULL_TREE,
1649 convert (type, integer_zero_node),
1650 build_real (type, value));
1652 yylval.ttype = build_real (type, value);
1656 tree traditional_type, ansi_type, type;
1657 HOST_WIDE_INT high, low;
1658 int spec_unsigned = 0;
1660 int spec_long_long = 0;
1664 traditional_type = ansi_type = type = NULL_TREE;
1667 if (c == 'u' || c == 'U')
1670 error ("two `u's in integer constant");
1673 else if (c == 'l' || c == 'L')
1678 error ("three `l's in integer constant");
1679 else if (pedantic && ! in_system_header && warn_long_long)
1680 pedwarn ("ANSI C forbids long long integer constants");
1685 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1688 error ("more than one `i' or `j' in numeric constant");
1690 pedwarn ("ANSI C forbids imaginary numeric constants");
1695 if (p >= token_buffer + maxtoken - 3)
1696 p = extend_token_buffer (p);
1701 /* If it won't fit in the host's representation for integers,
1706 pedwarn ("integer constant out of range");
1708 /* This is simplified by the fact that our constant
1709 is always positive. */
1713 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1715 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1716 / HOST_BITS_PER_CHAR)]
1717 << (i * HOST_BITS_PER_CHAR));
1718 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1721 yylval.ttype = build_int_2 (low, high);
1722 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1724 /* If warn_traditional, calculate both the ANSI type and the
1725 traditional type, then see if they disagree.
1726 Otherwise, calculate only the type for the dialect in use. */
1727 if (warn_traditional || flag_traditional)
1729 /* Calculate the traditional type. */
1730 /* Traditionally, any constant is signed;
1731 but if unsigned is specified explicitly, obey that.
1732 Use the smallest size with the right number of bits,
1733 except for one special case with decimal constants. */
1734 if (! spec_long && base != 10
1735 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1736 traditional_type = (spec_unsigned ? unsigned_type_node
1737 : integer_type_node);
1738 /* A decimal constant must be long
1739 if it does not fit in type int.
1740 I think this is independent of whether
1741 the constant is signed. */
1742 else if (! spec_long && base == 10
1743 && int_fits_type_p (yylval.ttype, integer_type_node))
1744 traditional_type = (spec_unsigned ? unsigned_type_node
1745 : integer_type_node);
1746 else if (! spec_long_long)
1747 traditional_type = (spec_unsigned ? long_unsigned_type_node
1748 : long_integer_type_node);
1750 traditional_type = (spec_unsigned
1751 ? long_long_unsigned_type_node
1752 : long_long_integer_type_node);
1754 if (warn_traditional || ! flag_traditional)
1756 /* Calculate the ANSI type. */
1757 if (! spec_long && ! spec_unsigned
1758 && int_fits_type_p (yylval.ttype, integer_type_node))
1759 ansi_type = integer_type_node;
1760 else if (! spec_long && (base != 10 || spec_unsigned)
1761 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1762 ansi_type = unsigned_type_node;
1763 else if (! spec_unsigned && !spec_long_long
1764 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1765 ansi_type = long_integer_type_node;
1766 else if (! spec_long_long
1767 && int_fits_type_p (yylval.ttype,
1768 long_unsigned_type_node))
1769 ansi_type = long_unsigned_type_node;
1770 else if (! spec_unsigned
1771 && int_fits_type_p (yylval.ttype,
1772 long_long_integer_type_node))
1773 ansi_type = long_long_integer_type_node;
1775 ansi_type = long_long_unsigned_type_node;
1778 type = flag_traditional ? traditional_type : ansi_type;
1780 if (warn_traditional && traditional_type != ansi_type)
1782 if (TYPE_PRECISION (traditional_type)
1783 != TYPE_PRECISION (ansi_type))
1784 warning ("width of integer constant changes with -traditional");
1785 else if (TREE_UNSIGNED (traditional_type)
1786 != TREE_UNSIGNED (ansi_type))
1787 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1789 warning ("width of integer constant may change on other systems with -traditional");
1792 if (pedantic && !flag_traditional && !spec_long_long && !warn
1793 && (TYPE_PRECISION (long_integer_type_node)
1794 < TYPE_PRECISION (type)))
1797 pedwarn ("integer constant out of range");
1800 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1801 warning ("decimal constant is so large that it is unsigned");
1805 if (TYPE_PRECISION (type)
1806 <= TYPE_PRECISION (integer_type_node))
1808 = build_complex (NULL_TREE, integer_zero_node,
1809 convert (integer_type_node,
1812 error ("complex integer constant is too wide for `complex int'");
1814 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1815 /* The traditional constant 0x80000000 is signed
1816 but doesn't fit in the range of int.
1817 This will change it to -0x80000000, which does fit. */
1819 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1820 yylval.ttype = convert (type, yylval.ttype);
1821 TREE_OVERFLOW (yylval.ttype)
1822 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1825 TREE_TYPE (yylval.ttype) = type;
1828 /* If it's still an integer (not a complex), and it doesn't
1829 fit in the type we choose for it, then pedwarn. */
1832 && TREE_CODE (TREE_TYPE (yylval.ttype)) == INTEGER_TYPE
1833 && ! int_fits_type_p (yylval.ttype, TREE_TYPE (yylval.ttype)))
1834 pedwarn ("integer constant out of range");
1840 if (ISALNUM (c) || c == '.' || c == '_' || c == '$'
1841 || (!flag_traditional && (c == '-' || c == '+')
1842 && (p[-1] == 'e' || p[-1] == 'E')))
1843 error ("missing white space after number `%s'", token_buffer);
1845 value = CONSTANT; break;
1851 register int result = 0;
1852 register int num_chars = 0;
1854 unsigned width = TYPE_PRECISION (char_type_node);
1856 #ifdef MULTIBYTE_CHARS
1857 int longest_char = local_mb_cur_max ();
1858 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1861 max_chars = TYPE_PRECISION (integer_type_node) / width;
1863 width = WCHAR_TYPE_SIZE;
1870 if (c == '\'' || c == EOF)
1877 c = readescape (&ignore);
1880 if (width < HOST_BITS_PER_INT
1881 && (unsigned) c >= ((unsigned)1 << width))
1882 pedwarn ("escape sequence out of range for character");
1883 #ifdef MAP_CHARACTER
1885 c = MAP_CHARACTER (c);
1891 pedwarn ("ANSI C forbids newline in character constant");
1896 #ifdef MULTIBYTE_CHARS
1900 for (i = 1; i <= longest_char; ++i)
1902 if (i > maxtoken - 4)
1903 extend_token_buffer (token_buffer);
1905 token_buffer[i] = c;
1906 char_len = local_mbtowc (& wc,
1915 /* mbtowc sometimes needs an extra char before accepting */
1920 /* Merge character into result; ignore excess chars. */
1921 for (i = 1; i <= char_len; ++i)
1925 if (width < HOST_BITS_PER_INT)
1926 result = (result << width)
1928 & ((1 << width) - 1));
1930 result = token_buffer[i];
1932 num_chars += char_len;
1940 warning ("Ignoring invalid multibyte character");
1943 #ifdef MAP_CHARACTER
1945 c = MAP_CHARACTER (c);
1948 #else /* ! MULTIBYTE_CHARS */
1949 #ifdef MAP_CHARACTER
1950 c = MAP_CHARACTER (c);
1952 #endif /* ! MULTIBYTE_CHARS */
1957 if (chars_seen == 1) /* only keep the first one */
1962 /* Merge character into result; ignore excess chars. */
1963 num_chars += (width / TYPE_PRECISION (char_type_node));
1964 if (num_chars < max_chars + 1)
1966 if (width < HOST_BITS_PER_INT)
1967 result = (result << width) | (c & ((1 << width) - 1));
1974 error ("malformatted character constant");
1975 else if (chars_seen == 0)
1976 error ("empty character constant");
1977 else if (num_chars > max_chars)
1979 num_chars = max_chars;
1980 error ("character constant too long");
1982 else if (chars_seen != 1 && ! flag_traditional && warn_multichar)
1983 warning ("multi-character character constant");
1985 /* If char type is signed, sign-extend the constant. */
1988 int num_bits = num_chars * width;
1990 /* We already got an error; avoid invalid shift. */
1991 yylval.ttype = build_int_2 (0, 0);
1992 else if (TREE_UNSIGNED (char_type_node)
1993 || ((result >> (num_bits - 1)) & 1) == 0)
1995 = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
1996 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2000 = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
2001 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2003 TREE_TYPE (yylval.ttype) = integer_type_node;
2007 yylval.ttype = build_int_2 (result, 0);
2008 TREE_TYPE (yylval.ttype) = wchar_type_node;
2018 unsigned width = wide_flag ? WCHAR_TYPE_SIZE
2019 : TYPE_PRECISION (char_type_node);
2020 #ifdef MULTIBYTE_CHARS
2021 int longest_char = local_mb_cur_max ();
2022 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
2025 p = token_buffer + 1;
2027 while (c != '"' && c >= 0)
2032 c = readescape (&ignore);
2035 if (width < HOST_BITS_PER_INT
2036 && (unsigned) c >= ((unsigned)1 << width))
2037 pedwarn ("escape sequence out of range for character");
2042 pedwarn ("ANSI C forbids newline in string constant");
2047 #ifdef MULTIBYTE_CHARS
2051 for (i = 0; i < longest_char; ++i)
2053 if (p + i >= token_buffer + maxtoken)
2054 p = extend_token_buffer (p);
2057 char_len = local_mbtowc (& wc, p, i + 1);
2063 warning ("Ignoring invalid multibyte character");
2066 /* mbtowc sometimes needs an extra char before accepting */
2077 #endif /* MULTIBYTE_CHARS */
2080 /* Add this single character into the buffer either as a wchar_t
2081 or as a single byte. */
2084 unsigned width = TYPE_PRECISION (char_type_node);
2085 unsigned bytemask = (1 << width) - 1;
2088 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2089 p = extend_token_buffer (p);
2091 for (byte = 0; byte < WCHAR_BYTES; ++byte)
2094 if (byte >= (int) sizeof (c))
2097 value = (c >> (byte * width)) & bytemask;
2098 if (BYTES_BIG_ENDIAN)
2099 p[WCHAR_BYTES - byte - 1] = value;
2107 if (p >= token_buffer + maxtoken)
2108 p = extend_token_buffer (p);
2116 /* Terminate the string value, either with a single byte zero
2117 or with a wide zero. */
2120 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2121 p = extend_token_buffer (p);
2122 bzero (p, WCHAR_BYTES);
2127 if (p >= token_buffer + maxtoken)
2128 p = extend_token_buffer (p);
2133 error ("Unterminated string constant");
2135 /* We have read the entire constant.
2136 Construct a STRING_CST for the result. */
2140 yylval.ttype = build_string (p - (token_buffer + 1),
2142 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2147 /* Return an Objective-C @"..." constant string object. */
2148 yylval.ttype = build_objc_string (p - (token_buffer + 1),
2150 TREE_TYPE (yylval.ttype) = char_array_type_node;
2151 value = OBJC_STRING;
2155 yylval.ttype = build_string (p - (token_buffer + 1),
2157 TREE_TYPE (yylval.ttype) = char_array_type_node;
2185 yylval.code = PLUS_EXPR; break;
2187 yylval.code = MINUS_EXPR; break;
2189 yylval.code = BIT_AND_EXPR; break;
2191 yylval.code = BIT_IOR_EXPR; break;
2193 yylval.code = MULT_EXPR; break;
2195 yylval.code = TRUNC_DIV_EXPR; break;
2197 yylval.code = TRUNC_MOD_EXPR; break;
2199 yylval.code = BIT_XOR_EXPR; break;
2201 yylval.code = LSHIFT_EXPR; break;
2203 yylval.code = RSHIFT_EXPR; break;
2205 yylval.code = LT_EXPR; break;
2207 yylval.code = GT_EXPR; break;
2210 token_buffer[1] = c1 = GETC();
2211 token_buffer[2] = 0;
2218 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2220 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2222 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2224 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2226 value = ASSIGN; goto done;
2232 value = PLUSPLUS; goto done;
2234 value = MINUSMINUS; goto done;
2236 value = ANDAND; goto done;
2238 value = OROR; goto done;
2251 { value = POINTSAT; goto done; }
2255 { value = ']'; goto done; }
2259 { value = '{'; indent_level++; goto done; }
2261 { value = '['; goto done; }
2265 { value = '}'; indent_level--; goto done; }
2269 token_buffer[1] = 0;
2271 if ((c == '<') || (c == '>'))
2272 value = ARITHCOMPARE;
2278 /* Don't make yyparse think this is eof. */
2297 /* yylloc.last_line = lineno; */
2302 /* Sets the value of the 'yydebug' variable to VALUE.
2303 This is a function so we don't have to have YYDEBUG defined
2304 in order to build the compiler. */
2313 warning ("YYDEBUG not defined.");