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. */
36 #ifdef MULTIBYTE_CHARS
39 #endif /* MULTIBYTE_CHARS */
43 extern cpp_reader parse_in;
44 extern cpp_options parse_options;
46 /* Stream for reading from the input file. */
50 extern void yyprint PROTO((FILE *, int, YYSTYPE));
52 /* The elements of `ridpointers' are identifier nodes
53 for the reserved type names and storage classes.
54 It is indexed by a RID_... value. */
55 tree ridpointers[(int) RID_MAX];
57 /* Cause the `yydebug' variable to be defined. */
61 extern unsigned char *yy_cur, *yy_lim;
63 extern int yy_get_token ();
65 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
66 #define UNGETC(c) ((c), yy_cur--)
68 #define GETC() getc (finput)
69 #define UNGETC(c) ungetc (c, finput)
72 /* the declaration found for the last IDENTIFIER token read in.
73 yylex must look this up to detect typedefs, which get token type TYPENAME,
74 so it is left around in case the identifier is not a typedef but is
75 used in a context which makes it a reference to a variable. */
78 /* Nonzero enables objc features. */
84 /* File used for outputting assembler code. */
85 extern FILE *asm_out_file;
87 #ifndef WCHAR_TYPE_SIZE
89 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
91 #define WCHAR_TYPE_SIZE BITS_PER_WORD
95 /* Number of bytes in a wide character. */
96 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
98 static int maxtoken; /* Current nominal length of token buffer. */
99 char *token_buffer; /* Pointer to token buffer.
100 Actual allocated length is maxtoken + 2.
101 This is not static because objc-parse.y uses it. */
103 static int indent_level = 0; /* Number of { minus number of }. */
105 /* Nonzero if end-of-file has been seen on input. */
106 static int end_of_file;
109 /* Buffered-back input character; faster than using ungetc. */
110 static int nextchar = -1;
113 #ifdef HANDLE_GENERIC_PRAGMAS
114 static int handle_generic_pragma PROTO((int));
115 #endif /* HANDLE_GENERIC_PRAGMAS */
116 static int whitespace_cr PROTO((int));
117 static int skip_white_space PROTO((int));
118 static int skip_white_space_on_line PROTO((void));
119 static char *extend_token_buffer PROTO((char *));
120 static int readescape PROTO((int *));
122 /* Do not insert generated code into the source, instead, include it.
123 This allows us to build gcc automatically even for targets that
124 need to add or modify the reserved keyword lists. */
127 /* Return something to represent absolute declarators containing a *.
128 TARGET is the absolute declarator that the * contains.
129 TYPE_QUALS is a list of modifiers such as const or volatile
130 to apply to the pointer type, represented as identifiers.
132 We return an INDIRECT_REF whose "contents" are TARGET
133 and whose type is the modifier list. */
136 make_pointer_declarator (type_quals, target)
137 tree type_quals, target;
139 return build1 (INDIRECT_REF, type_quals, target);
143 forget_protocol_qualifiers ()
145 int i, n = sizeof wordlist / sizeof (struct resword);
147 for (i = 0; i < n; i++)
148 if ((int) wordlist[i].rid >= (int) RID_IN
149 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
150 wordlist[i].name = "";
154 remember_protocol_qualifiers ()
156 int i, n = sizeof wordlist / sizeof (struct resword);
158 for (i = 0; i < n; i++)
159 if (wordlist[i].rid == RID_IN)
160 wordlist[i].name = "in";
161 else if (wordlist[i].rid == RID_OUT)
162 wordlist[i].name = "out";
163 else if (wordlist[i].rid == RID_INOUT)
164 wordlist[i].name = "inout";
165 else if (wordlist[i].rid == RID_BYCOPY)
166 wordlist[i].name = "bycopy";
167 else if (wordlist[i].rid == RID_BYREF)
168 wordlist[i].name = "byref";
169 else if (wordlist[i].rid == RID_ONEWAY)
170 wordlist[i].name = "oneway";
174 init_parse (filename)
178 /* Open input file. */
179 if (filename == 0 || !strcmp (filename, "-"))
185 finput = fopen (filename, "r");
187 pfatal_with_name (filename);
189 #ifdef IO_BUFFER_SIZE
190 setvbuf (finput, (char *) xmalloc (IO_BUFFER_SIZE), _IOFBF, IO_BUFFER_SIZE);
192 #endif /* !USE_CPPLIB */
197 parse_in.show_column = 1;
198 if (! cpp_start_read (&parse_in, filename))
201 /* cpp_start_read always puts at least one line directive into the
202 token buffer. We must arrange to read it out here. */
203 yy_cur = parse_in.token_buffer;
204 yy_lim = CPP_PWRITTEN (&parse_in);
214 cpp_finish (&parse_in);
223 /* Make identifier nodes long enough for the language-specific slots. */
224 set_identifier_size (sizeof (struct lang_identifier));
226 /* Start it at 0, because check_newline is called at the very beginning
227 and will increment it to 1. */
230 #ifdef MULTIBYTE_CHARS
231 /* Change to the native locale for multibyte conversions. */
232 setlocale (LC_CTYPE, "");
233 literal_codeset = getenv ("LANG");
237 token_buffer = (char *) xmalloc (maxtoken + 2);
239 ridpointers[(int) RID_INT] = get_identifier ("int");
240 ridpointers[(int) RID_CHAR] = get_identifier ("char");
241 ridpointers[(int) RID_VOID] = get_identifier ("void");
242 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
243 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
244 ridpointers[(int) RID_SHORT] = get_identifier ("short");
245 ridpointers[(int) RID_LONG] = get_identifier ("long");
246 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
247 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
248 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
249 ridpointers[(int) RID_CONST] = get_identifier ("const");
250 ridpointers[(int) RID_RESTRICT] = get_identifier ("restrict");
251 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
252 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
253 ridpointers[(int) RID_STATIC] = get_identifier ("static");
254 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
255 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
256 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
257 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
258 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
259 ridpointers[(int) RID_ID] = get_identifier ("id");
260 ridpointers[(int) RID_IN] = get_identifier ("in");
261 ridpointers[(int) RID_OUT] = get_identifier ("out");
262 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
263 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
264 ridpointers[(int) RID_BYREF] = get_identifier ("byref");
265 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
266 forget_protocol_qualifiers();
268 /* Some options inhibit certain reserved words.
269 Clear those words out of the hash table so they won't be recognized. */
270 #define UNSET_RESERVED_WORD(STRING) \
271 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
272 if (s) s->name = ""; } while (0)
274 if (! doing_objc_thang)
275 UNSET_RESERVED_WORD ("id");
277 if (flag_traditional)
279 UNSET_RESERVED_WORD ("const");
280 UNSET_RESERVED_WORD ("restrict");
281 UNSET_RESERVED_WORD ("volatile");
282 UNSET_RESERVED_WORD ("typeof");
283 UNSET_RESERVED_WORD ("signed");
284 UNSET_RESERVED_WORD ("inline");
285 UNSET_RESERVED_WORD ("iterator");
286 UNSET_RESERVED_WORD ("complex");
288 else if (!flag_isoc9x)
289 UNSET_RESERVED_WORD ("restrict");
293 UNSET_RESERVED_WORD ("asm");
294 UNSET_RESERVED_WORD ("typeof");
295 UNSET_RESERVED_WORD ("inline");
296 UNSET_RESERVED_WORD ("iterator");
297 UNSET_RESERVED_WORD ("complex");
302 reinit_parse_for_function ()
306 /* Function used when yydebug is set, to print a token in more detail. */
309 yyprint (file, yychar, yylval)
321 if (IDENTIFIER_POINTER (t))
322 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
327 if (TREE_CODE (t) == INTEGER_CST)
329 #if HOST_BITS_PER_WIDE_INT == 64
330 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
333 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
340 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
346 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
351 /* Iff C is a carriage return, warn about it - if appropriate -
352 and return nonzero. */
357 static int newline_warning = 0;
361 /* ANSI C says the effects of a carriage return in a source file
363 if (pedantic && !newline_warning)
365 warning ("carriage return in source file");
366 warning ("(we only warn about the first carriage return)");
374 /* If C is not whitespace, return C.
375 Otherwise skip whitespace and return first nonwhite char read. */
385 /* We don't recognize comments here, because
386 cpp output can include / and * consecutively as operators.
387 Also, there's no need, since cpp removes all comments. */
390 c = check_newline ();
411 error ("stray '\\' in program");
421 /* Skips all of the white space at the current location in the input file.
422 Must use and reset nextchar if it has the next character. */
425 position_after_white_space ()
431 c = nextchar, nextchar = -1;
436 UNGETC (skip_white_space (c));
439 /* Like skip_white_space, but don't advance beyond the end of line.
440 Moreover, we don't get passed a character to start with. */
442 skip_white_space_on_line ()
471 /* Make the token buffer longer, preserving the data in it.
472 P should point to just beyond the last valid character in the old buffer.
473 The value we return is a pointer to the new buffer
474 at a place corresponding to P. */
477 extend_token_buffer (p)
480 int offset = p - token_buffer;
482 maxtoken = maxtoken * 2 + 10;
483 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
485 return token_buffer + offset;
488 #if defined HANDLE_PRAGMA
489 /* Local versions of these macros, that can be passed as function pointers. */
504 /* At the beginning of a line, increment the line number
505 and process any #-directive on this line.
506 If the line is a #-directive, read the entire line and return a newline.
507 Otherwise, return the line's first non-whitespace character. */
517 /* Read first nonwhite char on the line. */
520 while (c == ' ' || c == '\t')
525 /* If not #, return it so caller will use it. */
529 /* Read first nonwhite char after the `#'. */
532 while (c == ' ' || c == '\t')
535 /* If a letter follows, then if the word here is `line', skip
536 it and ignore it; otherwise, ignore the line, with an error
537 if the word isn't `pragma', `ident', `define', or `undef'. */
539 if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
548 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'
549 || whitespace_cr (c) ))
551 while (c == ' ' || c == '\t' || whitespace_cr (c))
556 #if defined HANDLE_PRAGMA || defined HANDLE_GENERIC_PRAGMAS
559 if (token != IDENTIFIER)
561 #endif /* HANDLE_PRAGMA || HANDLE_GENERIC_PRAGMAS */
564 /* We invoke HANDLE_PRAGMA before HANDLE_GENERIC_PRAGMAS (if
565 both are defined), in order to give the back end a chance to
566 override the interpretation of generic style pragmas. */
570 c = nextchar, nextchar = -1;
573 #endif /* !USE_CPPLIB */
575 if (TREE_CODE (yylval.ttype) != IDENTIFIER_NODE)
578 if (HANDLE_PRAGMA (pragma_getc, pragma_ungetc,
579 IDENTIFIER_POINTER (yylval.ttype)))
581 #endif /* HANDLE_PRAGMA */
583 #ifdef HANDLE_GENERIC_PRAGMAS
584 if (handle_generic_pragma (token))
586 #endif /* HANDLE_GENERIC_PRAGMAS */
588 /* Issue a warning message if we have been asked to do so.
589 Ignoring unknown pragmas in system header file unless
590 an explcit -Wunknown-pragmas has been given. */
591 if (warn_unknown_pragmas > 1
592 || (warn_unknown_pragmas && ! in_system_header))
593 warning ("ignoring pragma: %s", token_buffer);
606 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
609 debug_define (lineno, GET_DIRECTIVE_LINE ());
619 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
622 debug_undef (lineno, GET_DIRECTIVE_LINE ());
631 && ((c = GETC()) == ' ' || c == '\t'))
640 && ((c = GETC()) == ' ' || c == '\t'))
642 /* #ident. The pedantic warning is now in cccp.c. */
644 /* Here we have just seen `#ident '.
645 A string constant should follow. */
647 c = skip_white_space_on_line ();
649 /* If no argument, ignore the line. */
656 || TREE_CODE (yylval.ttype) != STRING_CST)
658 error ("invalid #ident");
664 #ifdef ASM_OUTPUT_IDENT
665 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
669 /* Skip the rest of this line. */
674 error ("undefined or invalid # directive");
679 /* Here we have either `#line' or `# <nonletter>'.
680 In either case, it should be a line number; a digit should follow. */
682 /* Can't use skip_white_space here, but must handle all whitespace
683 that is not '\n', lest we get a recursion for '\r' '\n' when
686 c = skip_white_space_on_line ();
688 /* If the # is the only nonwhite char on the line,
689 just ignore it. Check the new newline. */
693 /* Something follows the #; read a token. */
698 if (token == CONSTANT
699 && TREE_CODE (yylval.ttype) == INTEGER_CST)
701 int old_lineno = lineno;
703 /* subtract one, because it is the following line that
704 gets the specified number */
706 int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
708 /* Is this the last nonwhite stuff on the line? */
709 c = skip_white_space_on_line ();
712 /* No more: store the line number and check following line. */
718 /* More follows: it must be a string constant (filename). */
720 /* Read the string constant. */
723 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
725 error ("invalid #line");
730 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
731 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
734 /* Each change of file name
735 reinitializes whether we are now in a system header. */
736 in_system_header = 0;
738 if (main_input_filename == 0)
739 main_input_filename = input_filename;
741 /* Is this the last nonwhite stuff on the line? */
742 c = skip_white_space_on_line ();
745 /* Update the name in the top element of input_file_stack. */
746 if (input_file_stack)
747 input_file_stack->name = input_filename;
756 /* `1' after file name means entering new file.
757 `2' after file name means just left a file. */
759 if (token == CONSTANT
760 && TREE_CODE (yylval.ttype) == INTEGER_CST)
762 if (TREE_INT_CST_LOW (yylval.ttype) == 1)
764 /* Pushing to a new file. */
766 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
767 input_file_stack->line = old_lineno;
768 p->next = input_file_stack;
769 p->name = input_filename;
770 p->indent_level = indent_level;
771 input_file_stack = p;
772 input_file_stack_tick++;
773 debug_start_source_file (input_filename);
776 else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
778 /* Popping out of a file. */
779 if (input_file_stack->next)
781 struct file_stack *p = input_file_stack;
782 if (indent_level != p->indent_level)
784 warning_with_file_and_line
785 (p->name, old_lineno,
786 "This file contains more `%c's than `%c's.",
787 indent_level > p->indent_level ? '{' : '}',
788 indent_level > p->indent_level ? '}' : '{');
790 input_file_stack = p->next;
792 input_file_stack_tick++;
793 debug_end_source_file (input_file_stack->line);
796 error ("#-lines for entering and leaving files don't match");
802 /* Now that we've pushed or popped the input stack,
803 update the name in the top element. */
804 if (input_file_stack)
805 input_file_stack->name = input_filename;
807 /* If we have handled a `1' or a `2',
808 see if there is another number to read. */
811 /* Is this the last nonwhite stuff on the line? */
812 c = skip_white_space_on_line ();
821 /* `3' after file name means this is a system header file. */
823 if (token == CONSTANT
824 && TREE_CODE (yylval.ttype) == INTEGER_CST
825 && TREE_INT_CST_LOW (yylval.ttype) == 3)
826 in_system_header = 1, used_up = 1;
830 /* Is this the last nonwhite stuff on the line? */
831 c = skip_white_space_on_line ();
837 warning ("unrecognized text at end of #line");
840 error ("invalid #-line");
842 /* skip the rest of this line. */
845 if (c != '\n' && c != EOF && nextchar >= 0)
846 c = nextchar, nextchar = -1;
848 while (c != '\n' && c != EOF)
853 #ifdef HANDLE_GENERIC_PRAGMAS
855 /* Handle a #pragma directive.
856 TOKEN is the token we read after `#pragma'. Processes the entire input
857 line and return non-zero iff the pragma has been successfully parsed. */
859 /* This function has to be in this file, in order to get at
863 handle_generic_pragma (token)
876 handle_pragma_token (token_buffer, yylval.ttype);
879 handle_pragma_token (token_buffer, NULL);
883 c = nextchar, nextchar = -1;
888 while (c == ' ' || c == '\t')
892 if (c == '\n' || c == EOF)
893 return handle_pragma_token (NULL, NULL);
899 #endif /* HANDLE_GENERIC_PRAGMAS */
901 #define ENDFILE -1 /* token that represents end-of-file */
903 /* Read an escape sequence, returning its equivalent as a character,
904 or store 1 in *ignore_ptr if it is backslash-newline. */
907 readescape (ignore_ptr)
910 register int c = GETC();
912 register unsigned count;
913 unsigned firstdig = 0;
919 if (warn_traditional)
920 warning ("the meaning of `\\x' varies with -traditional");
922 if (flag_traditional)
931 if (!(c >= 'a' && c <= 'f')
932 && !(c >= 'A' && c <= 'F')
933 && !(c >= '0' && c <= '9'))
939 if (c >= 'a' && c <= 'f')
940 code += c - 'a' + 10;
941 if (c >= 'A' && c <= 'F')
942 code += c - 'A' + 10;
943 if (c >= '0' && c <= '9')
945 if (code != 0 || count != 0)
954 error ("\\x used with no following hex digits");
956 /* Digits are all 0's. Ok. */
958 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
960 && (((unsigned)1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
962 pedwarn ("hex escape out of range");
965 case '0': case '1': case '2': case '3': case '4':
966 case '5': case '6': case '7':
969 while ((c <= '7') && (c >= '0') && (count++ < 3))
971 code = (code * 8) + (c - '0');
977 case '\\': case '\'': case '"':
986 return TARGET_NEWLINE;
1001 if (warn_traditional)
1002 warning ("the meaning of `\\a' varies with -traditional");
1004 if (flag_traditional)
1009 #if 0 /* Vertical tab is present in common usage compilers. */
1010 if (flag_traditional)
1018 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
1024 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1028 /* `\%' is used to prevent SCCS from getting confused. */
1031 pedwarn ("non-ANSI escape sequence `\\%c'", c);
1034 if (c >= 040 && c < 0177)
1035 pedwarn ("unknown escape sequence `\\%c'", c);
1037 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
1047 strcpy (buf, string);
1049 /* We can't print string and character constants well
1050 because the token_buffer contains the result of processing escapes. */
1052 strcat (buf, " at end of input");
1053 else if (token_buffer[0] == 0)
1054 strcat (buf, " at null character");
1055 else if (token_buffer[0] == '"')
1056 strcat (buf, " before string constant");
1057 else if (token_buffer[0] == '\'')
1058 strcat (buf, " before character constant");
1059 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
1060 sprintf (buf + strlen (buf), " before character 0%o",
1061 (unsigned char) token_buffer[0]);
1063 strcat (buf, " before `%s'");
1065 error (buf, token_buffer);
1075 char long_long_flag;
1078 struct try_type type_sequence[] =
1080 { &integer_type_node, 0, 0, 0},
1081 { &unsigned_type_node, 1, 0, 0},
1082 { &long_integer_type_node, 0, 1, 0},
1083 { &long_unsigned_type_node, 1, 1, 0},
1084 { &long_long_integer_type_node, 0, 1, 1},
1085 { &long_long_unsigned_type_node, 1, 1, 1}
1100 c = nextchar, nextchar = -1;
1105 /* Effectively do c = skip_white_space (c)
1106 but do it faster in the usual cases. */
1119 /* Call skip_white_space so we can warn if appropriate. */
1124 c = skip_white_space (c);
1126 goto found_nonwhite;
1130 token_buffer[0] = c;
1131 token_buffer[1] = 0;
1133 /* yylloc.first_line = lineno; */
1139 token_buffer[0] = 0;
1144 /* Capital L may start a wide-string or wide-character constant. */
1146 register int c = GETC();
1155 goto string_constant;
1162 if (!doing_objc_thang)
1169 /* '@' may start a constant string object. */
1170 register int c = GETC ();
1174 goto string_constant;
1177 /* Fall through to treat '@' as the start of an identifier. */
1180 case 'A': case 'B': case 'C': case 'D': case 'E':
1181 case 'F': case 'G': case 'H': case 'I': case 'J':
1182 case 'K': case 'M': case 'N': case 'O':
1183 case 'P': case 'Q': case 'R': case 'S': case 'T':
1184 case 'U': case 'V': case 'W': case 'X': case 'Y':
1186 case 'a': case 'b': case 'c': case 'd': case 'e':
1187 case 'f': case 'g': case 'h': case 'i': case 'j':
1188 case 'k': case 'l': case 'm': case 'n': case 'o':
1189 case 'p': case 'q': case 'r': case 's': case 't':
1190 case 'u': case 'v': case 'w': case 'x': case 'y':
1196 while (ISALNUM (c) || c == '_' || c == '$' || c == '@')
1198 /* Make sure this char really belongs in an identifier. */
1199 if (c == '@' && ! doing_objc_thang)
1203 if (! dollars_in_ident)
1204 error ("`$' in identifier");
1206 pedwarn ("`$' in identifier");
1209 if (p >= token_buffer + maxtoken)
1210 p = extend_token_buffer (p);
1226 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1229 register struct resword *ptr;
1231 if ((ptr = is_reserved_word (token_buffer, p - token_buffer)))
1234 yylval.ttype = ridpointers[(int) ptr->rid];
1235 value = (int) ptr->token;
1237 /* Only return OBJECTNAME if it is a typedef. */
1238 if (doing_objc_thang && value == OBJECTNAME)
1240 lastiddecl = lookup_name(yylval.ttype);
1242 if (lastiddecl == NULL_TREE
1243 || TREE_CODE (lastiddecl) != TYPE_DECL)
1247 /* Even if we decided to recognize asm, still perhaps warn. */
1249 && (value == ASM_KEYWORD || value == TYPEOF
1250 || ptr->rid == RID_INLINE)
1251 && token_buffer[0] != '_')
1252 pedwarn ("ANSI does not permit the keyword `%s'",
1257 /* If we did not find a keyword, look for an identifier
1260 if (value == IDENTIFIER)
1262 if (token_buffer[0] == '@')
1263 error("invalid identifier `%s'", token_buffer);
1265 yylval.ttype = get_identifier (token_buffer);
1266 lastiddecl = lookup_name (yylval.ttype);
1268 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1270 /* A user-invisible read-only initialized variable
1271 should be replaced by its value.
1272 We handle only strings since that's the only case used in C. */
1273 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1274 && DECL_IGNORED_P (lastiddecl)
1275 && TREE_READONLY (lastiddecl)
1276 && DECL_INITIAL (lastiddecl) != 0
1277 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1279 tree stringval = DECL_INITIAL (lastiddecl);
1281 /* Copy the string value so that we won't clobber anything
1282 if we put something in the TREE_CHAIN of this one. */
1283 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1284 TREE_STRING_POINTER (stringval));
1287 else if (doing_objc_thang)
1289 tree objc_interface_decl = is_class_name (yylval.ttype);
1291 if (objc_interface_decl)
1294 yylval.ttype = objc_interface_decl;
1304 /* Check first for common special case: single-digit 0 or 1. */
1307 UNGETC (next_c); /* Always undo this lookahead. */
1308 if (!ISALNUM (next_c) && next_c != '.')
1310 token_buffer[0] = (char)c, token_buffer[1] = '\0';
1311 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1317 case '2': case '3': case '4':
1318 case '5': case '6': case '7': case '8': case '9':
1323 int largest_digit = 0;
1325 /* for multi-precision arithmetic,
1326 we actually store only HOST_BITS_PER_CHAR bits in each part.
1327 The number of parts is chosen so as to be sufficient to hold
1328 the enough bits to fit into the two HOST_WIDE_INTs that contain
1329 the integer value (this is always at least as many bits as are
1330 in a target `long long' value, but may be wider). */
1331 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1332 int parts[TOTAL_PARTS];
1335 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag
1338 for (count = 0; count < TOTAL_PARTS; count++)
1346 *p++ = (c = GETC());
1347 if ((c == 'x') || (c == 'X'))
1350 *p++ = (c = GETC());
1352 /* Leading 0 forces octal unless the 0 is the only digit. */
1353 else if (c >= '0' && c <= '9')
1362 /* Read all the digits-and-decimal-points. */
1365 || (ISALNUM (c) && c != 'l' && c != 'L'
1366 && c != 'u' && c != 'U'
1367 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1368 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1373 error ("floating constant may not be in radix 16");
1374 if (floatflag == TOO_MANY_POINTS)
1375 /* We have already emitted an error. Don't need another. */
1377 else if (floatflag == AFTER_POINT)
1379 error ("malformed floating constant");
1380 floatflag = TOO_MANY_POINTS;
1381 /* Avoid another error from atof by forcing all characters
1382 from here on to be ignored. */
1386 floatflag = AFTER_POINT;
1390 /* Accept '.' as the start of a floating-point number
1391 only when it is followed by a digit.
1392 Otherwise, unread the following non-digit
1393 and use the '.' as a structural token. */
1394 if (p == token_buffer + 2 && !ISDIGIT (c))
1405 error ("parse error at `..'");
1408 token_buffer[1] = 0;
1415 /* It is not a decimal point.
1416 It should be a digit (perhaps a hex digit). */
1422 else if (base <= 10)
1424 if (c == 'e' || c == 'E')
1427 floatflag = AFTER_POINT;
1428 break; /* start of exponent */
1430 error ("nondigits in number and not hexadecimal");
1441 if (c >= largest_digit)
1445 for (count = 0; count < TOTAL_PARTS; count++)
1447 parts[count] *= base;
1451 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1453 &= (1 << HOST_BITS_PER_CHAR) - 1;
1459 /* If the extra highest-order part ever gets anything in it,
1460 the number is certainly too big. */
1461 if (parts[TOTAL_PARTS - 1] != 0)
1464 if (p >= token_buffer + maxtoken - 3)
1465 p = extend_token_buffer (p);
1466 *p++ = (c = GETC());
1471 error ("numeric constant with no digits");
1473 if (largest_digit >= base)
1474 error ("numeric constant contains digits beyond the radix");
1476 /* Remove terminating char from the token buffer and delimit the string */
1479 if (floatflag != NOT_FLOAT)
1481 tree type = double_type_node;
1483 int conversion_errno = 0;
1484 REAL_VALUE_TYPE value;
1487 /* Read explicit exponent if any, and put it in tokenbuf. */
1489 if ((c == 'e') || (c == 'E'))
1491 if (p >= token_buffer + maxtoken - 3)
1492 p = extend_token_buffer (p);
1495 if ((c == '+') || (c == '-'))
1501 error ("floating constant exponent has no digits");
1504 if (p >= token_buffer + maxtoken - 3)
1505 p = extend_token_buffer (p);
1513 /* Convert string to a double, checking for overflow. */
1514 if (setjmp (handler))
1516 error ("floating constant out of range");
1521 int fflag = 0, lflag = 0;
1522 /* Copy token_buffer now, while it has just the number
1523 and not the suffixes; once we add `f' or `i',
1524 REAL_VALUE_ATOF may not work any more. */
1525 char *copy = (char *) alloca (p - token_buffer + 1);
1526 bcopy (token_buffer, copy, p - token_buffer + 1);
1528 set_float_handler (handler);
1534 /* Read the suffixes to choose a data type. */
1539 error ("more than one `f' in numeric constant");
1545 error ("more than one `l' in numeric constant");
1551 error ("more than one `i' or `j' in numeric constant");
1553 pedwarn ("ANSI C forbids imaginary numeric constants");
1564 if (p >= token_buffer + maxtoken - 3)
1565 p = extend_token_buffer (p);
1571 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1572 tells the desired precision of the binary result
1573 of decimal-to-binary conversion. */
1578 error ("both `f' and `l' in floating constant");
1580 type = float_type_node;
1582 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1583 conversion_errno = errno;
1584 /* A diagnostic is required here by some ANSI C testsuites.
1585 This is not pedwarn, become some people don't want
1586 an error for this. */
1587 if (REAL_VALUE_ISINF (value) && pedantic)
1588 warning ("floating point number exceeds range of `float'");
1592 type = long_double_type_node;
1594 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1595 conversion_errno = errno;
1596 if (REAL_VALUE_ISINF (value) && pedantic)
1597 warning ("floating point number exceeds range of `long double'");
1602 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1603 conversion_errno = errno;
1604 if (REAL_VALUE_ISINF (value) && pedantic)
1605 warning ("floating point number exceeds range of `double'");
1608 set_float_handler (NULL_PTR);
1611 /* ERANGE is also reported for underflow,
1612 so test the value to distinguish overflow from that. */
1613 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1614 && (REAL_VALUES_LESS (dconst1, value)
1615 || REAL_VALUES_LESS (value, dconstm1)))
1616 warning ("floating point number exceeds range of `double'");
1619 /* If the result is not a number, assume it must have been
1620 due to some error message above, so silently convert
1622 if (REAL_VALUE_ISNAN (value))
1625 /* Create a node with determined type and value. */
1627 yylval.ttype = build_complex (NULL_TREE,
1628 convert (type, integer_zero_node),
1629 build_real (type, value));
1631 yylval.ttype = build_real (type, value);
1635 tree traditional_type, ansi_type, type;
1636 HOST_WIDE_INT high, low;
1637 int spec_unsigned = 0;
1639 int spec_long_long = 0;
1643 traditional_type = ansi_type = type = NULL_TREE;
1646 if (c == 'u' || c == 'U')
1649 error ("two `u's in integer constant");
1652 else if (c == 'l' || c == 'L')
1657 error ("three `l's in integer constant");
1658 else if (pedantic && ! in_system_header && warn_long_long)
1659 pedwarn ("ANSI C forbids long long integer constants");
1664 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1667 error ("more than one `i' or `j' in numeric constant");
1669 pedwarn ("ANSI C forbids imaginary numeric constants");
1674 if (p >= token_buffer + maxtoken - 3)
1675 p = extend_token_buffer (p);
1680 /* If the constant won't fit in an unsigned long long,
1681 then warn that the constant is out of range. */
1683 /* ??? This assumes that long long and long integer types are
1684 a multiple of 8 bits. This better than the original code
1685 though which assumed that long was exactly 32 bits and long
1686 long was exactly 64 bits. */
1688 bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1691 for (i = bytes; i < TOTAL_PARTS; i++)
1695 pedwarn ("integer constant out of range");
1697 /* This is simplified by the fact that our constant
1698 is always positive. */
1702 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1704 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1705 / HOST_BITS_PER_CHAR)]
1706 << (i * HOST_BITS_PER_CHAR));
1707 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1710 yylval.ttype = build_int_2 (low, high);
1711 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1713 /* If warn_traditional, calculate both the ANSI type and the
1714 traditional type, then see if they disagree.
1715 Otherwise, calculate only the type for the dialect in use. */
1716 if (warn_traditional || flag_traditional)
1718 /* Calculate the traditional type. */
1719 /* Traditionally, any constant is signed;
1720 but if unsigned is specified explicitly, obey that.
1721 Use the smallest size with the right number of bits,
1722 except for one special case with decimal constants. */
1723 if (! spec_long && base != 10
1724 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1725 traditional_type = (spec_unsigned ? unsigned_type_node
1726 : integer_type_node);
1727 /* A decimal constant must be long
1728 if it does not fit in type int.
1729 I think this is independent of whether
1730 the constant is signed. */
1731 else if (! spec_long && base == 10
1732 && int_fits_type_p (yylval.ttype, integer_type_node))
1733 traditional_type = (spec_unsigned ? unsigned_type_node
1734 : integer_type_node);
1735 else if (! spec_long_long)
1736 traditional_type = (spec_unsigned ? long_unsigned_type_node
1737 : long_integer_type_node);
1739 traditional_type = (spec_unsigned
1740 ? long_long_unsigned_type_node
1741 : long_long_integer_type_node);
1743 if (warn_traditional || ! flag_traditional)
1745 /* Calculate the ANSI type. */
1746 if (! spec_long && ! spec_unsigned
1747 && int_fits_type_p (yylval.ttype, integer_type_node))
1748 ansi_type = integer_type_node;
1749 else if (! spec_long && (base != 10 || spec_unsigned)
1750 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1751 ansi_type = unsigned_type_node;
1752 else if (! spec_unsigned && !spec_long_long
1753 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1754 ansi_type = long_integer_type_node;
1755 else if (! spec_long_long
1756 && int_fits_type_p (yylval.ttype,
1757 long_unsigned_type_node))
1758 ansi_type = long_unsigned_type_node;
1759 else if (! spec_unsigned
1760 && int_fits_type_p (yylval.ttype,
1761 long_long_integer_type_node))
1762 ansi_type = long_long_integer_type_node;
1764 ansi_type = long_long_unsigned_type_node;
1767 type = flag_traditional ? traditional_type : ansi_type;
1769 if (warn_traditional && traditional_type != ansi_type)
1771 if (TYPE_PRECISION (traditional_type)
1772 != TYPE_PRECISION (ansi_type))
1773 warning ("width of integer constant changes with -traditional");
1774 else if (TREE_UNSIGNED (traditional_type)
1775 != TREE_UNSIGNED (ansi_type))
1776 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1778 warning ("width of integer constant may change on other systems with -traditional");
1781 if (pedantic && !flag_traditional && !spec_long_long && !warn
1782 && (TYPE_PRECISION (long_integer_type_node)
1783 < TYPE_PRECISION (type)))
1784 pedwarn ("integer constant out of range");
1786 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1787 warning ("decimal constant is so large that it is unsigned");
1791 if (TYPE_PRECISION (type)
1792 <= TYPE_PRECISION (integer_type_node))
1794 = build_complex (NULL_TREE, integer_zero_node,
1795 convert (integer_type_node,
1798 error ("complex integer constant is too wide for `complex int'");
1800 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1801 /* The traditional constant 0x80000000 is signed
1802 but doesn't fit in the range of int.
1803 This will change it to -0x80000000, which does fit. */
1805 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1806 yylval.ttype = convert (type, yylval.ttype);
1807 TREE_OVERFLOW (yylval.ttype)
1808 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1811 TREE_TYPE (yylval.ttype) = type;
1817 if (ISALNUM (c) || c == '.' || c == '_' || c == '$'
1818 || (!flag_traditional && (c == '-' || c == '+')
1819 && (p[-1] == 'e' || p[-1] == 'E')))
1820 error ("missing white space after number `%s'", token_buffer);
1822 value = CONSTANT; break;
1828 register int result = 0;
1829 register int num_chars = 0;
1831 unsigned width = TYPE_PRECISION (char_type_node);
1833 #ifdef MULTIBYTE_CHARS
1834 int longest_char = local_mb_cur_max ();
1835 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1838 max_chars = TYPE_PRECISION (integer_type_node) / width;
1840 width = WCHAR_TYPE_SIZE;
1847 if (c == '\'' || c == EOF)
1854 c = readescape (&ignore);
1857 if (width < HOST_BITS_PER_INT
1858 && (unsigned) c >= ((unsigned)1 << width))
1859 pedwarn ("escape sequence out of range for character");
1860 #ifdef MAP_CHARACTER
1862 c = MAP_CHARACTER (c);
1868 pedwarn ("ANSI C forbids newline in character constant");
1873 #ifdef MULTIBYTE_CHARS
1877 for (i = 1; i <= longest_char; ++i)
1879 if (i > maxtoken - 4)
1880 extend_token_buffer (token_buffer);
1882 token_buffer[i] = c;
1883 char_len = local_mbtowc (& wc,
1892 /* mbtowc sometimes needs an extra char before accepting */
1897 /* Merge character into result; ignore excess chars. */
1898 for (i = 1; i <= char_len; ++i)
1902 if (width < HOST_BITS_PER_INT)
1903 result = (result << width)
1905 & ((1 << width) - 1));
1907 result = token_buffer[i];
1909 num_chars += char_len;
1917 warning ("Ignoring invalid multibyte character");
1920 #ifdef MAP_CHARACTER
1922 c = MAP_CHARACTER (c);
1925 #else /* ! MULTIBYTE_CHARS */
1926 #ifdef MAP_CHARACTER
1927 c = MAP_CHARACTER (c);
1929 #endif /* ! MULTIBYTE_CHARS */
1934 if (chars_seen == 1) /* only keep the first one */
1939 /* Merge character into result; ignore excess chars. */
1940 num_chars += (width / TYPE_PRECISION (char_type_node));
1941 if (num_chars < max_chars + 1)
1943 if (width < HOST_BITS_PER_INT)
1944 result = (result << width) | (c & ((1 << width) - 1));
1951 error ("malformatted character constant");
1952 else if (chars_seen == 0)
1953 error ("empty character constant");
1954 else if (num_chars > max_chars)
1956 num_chars = max_chars;
1957 error ("character constant too long");
1959 else if (chars_seen != 1 && ! flag_traditional && warn_multichar)
1960 warning ("multi-character character constant");
1962 /* If char type is signed, sign-extend the constant. */
1965 int num_bits = num_chars * width;
1967 /* We already got an error; avoid invalid shift. */
1968 yylval.ttype = build_int_2 (0, 0);
1969 else if (TREE_UNSIGNED (char_type_node)
1970 || ((result >> (num_bits - 1)) & 1) == 0)
1972 = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
1973 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1977 = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
1978 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1980 TREE_TYPE (yylval.ttype) = integer_type_node;
1984 yylval.ttype = build_int_2 (result, 0);
1985 TREE_TYPE (yylval.ttype) = wchar_type_node;
1995 unsigned width = wide_flag ? WCHAR_TYPE_SIZE
1996 : TYPE_PRECISION (char_type_node);
1997 #ifdef MULTIBYTE_CHARS
1998 int longest_char = local_mb_cur_max ();
1999 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
2002 p = token_buffer + 1;
2004 while (c != '"' && c >= 0)
2009 c = readescape (&ignore);
2012 if (width < HOST_BITS_PER_INT
2013 && (unsigned) c >= ((unsigned)1 << width))
2014 pedwarn ("escape sequence out of range for character");
2019 pedwarn ("ANSI C forbids newline in string constant");
2024 #ifdef MULTIBYTE_CHARS
2028 for (i = 0; i < longest_char; ++i)
2030 if (p + i >= token_buffer + maxtoken)
2031 p = extend_token_buffer (p);
2034 char_len = local_mbtowc (& wc, p, i + 1);
2040 warning ("Ignoring invalid multibyte character");
2043 /* mbtowc sometimes needs an extra char before accepting */
2054 #endif /* MULTIBYTE_CHARS */
2057 /* Add this single character into the buffer either as a wchar_t
2058 or as a single byte. */
2061 unsigned width = TYPE_PRECISION (char_type_node);
2062 unsigned bytemask = (1 << width) - 1;
2065 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2066 p = extend_token_buffer (p);
2068 for (byte = 0; byte < WCHAR_BYTES; ++byte)
2071 if (byte >= (int) sizeof (c))
2074 value = (c >> (byte * width)) & bytemask;
2075 if (BYTES_BIG_ENDIAN)
2076 p[WCHAR_BYTES - byte - 1] = value;
2084 if (p >= token_buffer + maxtoken)
2085 p = extend_token_buffer (p);
2093 /* Terminate the string value, either with a single byte zero
2094 or with a wide zero. */
2097 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2098 p = extend_token_buffer (p);
2099 bzero (p, WCHAR_BYTES);
2104 if (p >= token_buffer + maxtoken)
2105 p = extend_token_buffer (p);
2110 error ("Unterminated string constant");
2112 /* We have read the entire constant.
2113 Construct a STRING_CST for the result. */
2117 yylval.ttype = build_string (p - (token_buffer + 1),
2119 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2124 /* Return an Objective-C @"..." constant string object. */
2125 yylval.ttype = build_objc_string (p - (token_buffer + 1),
2127 TREE_TYPE (yylval.ttype) = char_array_type_node;
2128 value = OBJC_STRING;
2132 yylval.ttype = build_string (p - (token_buffer + 1),
2134 TREE_TYPE (yylval.ttype) = char_array_type_node;
2162 yylval.code = PLUS_EXPR; break;
2164 yylval.code = MINUS_EXPR; break;
2166 yylval.code = BIT_AND_EXPR; break;
2168 yylval.code = BIT_IOR_EXPR; break;
2170 yylval.code = MULT_EXPR; break;
2172 yylval.code = TRUNC_DIV_EXPR; break;
2174 yylval.code = TRUNC_MOD_EXPR; break;
2176 yylval.code = BIT_XOR_EXPR; break;
2178 yylval.code = LSHIFT_EXPR; break;
2180 yylval.code = RSHIFT_EXPR; break;
2182 yylval.code = LT_EXPR; break;
2184 yylval.code = GT_EXPR; break;
2187 token_buffer[1] = c1 = GETC();
2188 token_buffer[2] = 0;
2195 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2197 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2199 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2201 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2203 value = ASSIGN; goto done;
2209 value = PLUSPLUS; goto done;
2211 value = MINUSMINUS; goto done;
2213 value = ANDAND; goto done;
2215 value = OROR; goto done;
2228 { value = POINTSAT; goto done; }
2232 { value = ']'; goto done; }
2236 { value = '{'; indent_level++; goto done; }
2238 { value = '['; goto done; }
2242 { value = '}'; indent_level--; goto done; }
2246 token_buffer[1] = 0;
2248 if ((c == '<') || (c == '>'))
2249 value = ARITHCOMPARE;
2255 /* Don't make yyparse think this is eof. */
2274 /* yylloc.last_line = lineno; */
2279 /* Sets the value of the 'yydebug' variable to VALUE.
2280 This is a function so we don't have to have YYDEBUG defined
2281 in order to build the compiler. */
2290 warning ("YYDEBUG not defined.");