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. */
45 /* MULTIBYTE_CHARS support only works for native compilers.
46 ??? Ideally what we want is to model widechar support after
47 the current floating point support. */
49 #undef MULTIBYTE_CHARS
52 #ifdef MULTIBYTE_CHARS
55 #endif /* MULTIBYTE_CHARS */
59 extern cpp_reader parse_in;
60 extern cpp_options parse_options;
62 /* Stream for reading from the input file. */
66 extern void yyprint PROTO((FILE *, int, YYSTYPE));
68 /* The elements of `ridpointers' are identifier nodes
69 for the reserved type names and storage classes.
70 It is indexed by a RID_... value. */
71 tree ridpointers[(int) RID_MAX];
73 /* Cause the `yydebug' variable to be defined. */
77 extern unsigned char *yy_cur, *yy_lim;
79 extern int yy_get_token ();
81 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
82 #define UNGETC(c) ((void)(c), yy_cur--)
84 #define GETC() getc (finput)
85 #define UNGETC(c) ungetc (c, finput)
88 /* the declaration found for the last IDENTIFIER token read in.
89 yylex must look this up to detect typedefs, which get token type TYPENAME,
90 so it is left around in case the identifier is not a typedef but is
91 used in a context which makes it a reference to a variable. */
94 /* Nonzero enables objc features. */
100 /* File used for outputting assembler code. */
101 extern FILE *asm_out_file;
103 #ifndef WCHAR_TYPE_SIZE
105 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
107 #define WCHAR_TYPE_SIZE BITS_PER_WORD
111 /* Number of bytes in a wide character. */
112 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
114 static int maxtoken; /* Current nominal length of token buffer. */
115 char *token_buffer; /* Pointer to token buffer.
116 Actual allocated length is maxtoken + 2.
117 This is not static because objc-parse.y uses it. */
119 static int indent_level = 0; /* Number of { minus number of }. */
121 /* Nonzero if end-of-file has been seen on input. */
122 static int end_of_file;
125 /* Buffered-back input character; faster than using ungetc. */
126 static int nextchar = -1;
129 #ifdef HANDLE_GENERIC_PRAGMAS
130 static int handle_generic_pragma PROTO((int));
131 #endif /* HANDLE_GENERIC_PRAGMAS */
132 static int whitespace_cr PROTO((int));
133 static int skip_white_space PROTO((int));
134 static int skip_white_space_on_line PROTO((void));
135 static char *extend_token_buffer PROTO((char *));
136 static int readescape PROTO((int *));
138 /* Do not insert generated code into the source, instead, include it.
139 This allows us to build gcc automatically even for targets that
140 need to add or modify the reserved keyword lists. */
143 /* Return something to represent absolute declarators containing a *.
144 TARGET is the absolute declarator that the * contains.
145 TYPE_QUALS is a list of modifiers such as const or volatile
146 to apply to the pointer type, represented as identifiers.
148 We return an INDIRECT_REF whose "contents" are TARGET
149 and whose type is the modifier list. */
152 make_pointer_declarator (type_quals, target)
153 tree type_quals, target;
155 return build1 (INDIRECT_REF, type_quals, target);
159 forget_protocol_qualifiers ()
161 int i, n = sizeof wordlist / sizeof (struct resword);
163 for (i = 0; i < n; i++)
164 if ((int) wordlist[i].rid >= (int) RID_IN
165 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
166 wordlist[i].name = "";
170 remember_protocol_qualifiers ()
172 int i, n = sizeof wordlist / sizeof (struct resword);
174 for (i = 0; i < n; i++)
175 if (wordlist[i].rid == RID_IN)
176 wordlist[i].name = "in";
177 else if (wordlist[i].rid == RID_OUT)
178 wordlist[i].name = "out";
179 else if (wordlist[i].rid == RID_INOUT)
180 wordlist[i].name = "inout";
181 else if (wordlist[i].rid == RID_BYCOPY)
182 wordlist[i].name = "bycopy";
183 else if (wordlist[i].rid == RID_BYREF)
184 wordlist[i].name = "byref";
185 else if (wordlist[i].rid == RID_ONEWAY)
186 wordlist[i].name = "oneway";
190 init_parse (filename)
194 /* Open input file. */
195 if (filename == 0 || !strcmp (filename, "-"))
201 finput = fopen (filename, "r");
203 pfatal_with_name (filename);
205 #ifdef IO_BUFFER_SIZE
206 setvbuf (finput, (char *) xmalloc (IO_BUFFER_SIZE), _IOFBF, IO_BUFFER_SIZE);
208 #else /* !USE_CPPLIB */
209 parse_in.show_column = 1;
210 if (! cpp_start_read (&parse_in, filename))
213 if (filename == 0 || !strcmp (filename, "-"))
216 /* cpp_start_read always puts at least one line directive into the
217 token buffer. We must arrange to read it out here. */
218 yy_cur = parse_in.token_buffer;
219 yy_lim = CPP_PWRITTEN (&parse_in);
231 cpp_finish (&parse_in);
242 /* Make identifier nodes long enough for the language-specific slots. */
243 set_identifier_size (sizeof (struct lang_identifier));
245 /* Start it at 0, because check_newline is called at the very beginning
246 and will increment it to 1. */
249 #ifdef MULTIBYTE_CHARS
250 /* Change to the native locale for multibyte conversions. */
251 setlocale (LC_CTYPE, "");
252 literal_codeset = getenv ("LANG");
256 token_buffer = (char *) xmalloc (maxtoken + 2);
258 ridpointers[(int) RID_INT] = get_identifier ("int");
259 ridpointers[(int) RID_CHAR] = get_identifier ("char");
260 ridpointers[(int) RID_VOID] = get_identifier ("void");
261 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
262 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
263 ridpointers[(int) RID_SHORT] = get_identifier ("short");
264 ridpointers[(int) RID_LONG] = get_identifier ("long");
265 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
266 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
267 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
268 ridpointers[(int) RID_CONST] = get_identifier ("const");
269 ridpointers[(int) RID_RESTRICT] = get_identifier ("restrict");
270 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
271 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
272 ridpointers[(int) RID_STATIC] = get_identifier ("static");
273 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
274 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
275 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
276 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
277 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
278 ridpointers[(int) RID_ID] = get_identifier ("id");
279 ridpointers[(int) RID_IN] = get_identifier ("in");
280 ridpointers[(int) RID_OUT] = get_identifier ("out");
281 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
282 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
283 ridpointers[(int) RID_BYREF] = get_identifier ("byref");
284 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
285 forget_protocol_qualifiers();
287 /* Some options inhibit certain reserved words.
288 Clear those words out of the hash table so they won't be recognized. */
289 #define UNSET_RESERVED_WORD(STRING) \
290 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
291 if (s) s->name = ""; } while (0)
293 if (! doing_objc_thang)
294 UNSET_RESERVED_WORD ("id");
296 if (flag_traditional)
298 UNSET_RESERVED_WORD ("const");
299 UNSET_RESERVED_WORD ("restrict");
300 UNSET_RESERVED_WORD ("volatile");
301 UNSET_RESERVED_WORD ("typeof");
302 UNSET_RESERVED_WORD ("signed");
303 UNSET_RESERVED_WORD ("inline");
304 UNSET_RESERVED_WORD ("iterator");
305 UNSET_RESERVED_WORD ("complex");
307 else if (!flag_isoc9x)
308 UNSET_RESERVED_WORD ("restrict");
312 UNSET_RESERVED_WORD ("asm");
313 UNSET_RESERVED_WORD ("typeof");
314 UNSET_RESERVED_WORD ("inline");
315 UNSET_RESERVED_WORD ("iterator");
316 UNSET_RESERVED_WORD ("complex");
321 reinit_parse_for_function ()
325 /* Function used when yydebug is set, to print a token in more detail. */
328 yyprint (file, yychar, yylval)
340 if (IDENTIFIER_POINTER (t))
341 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
346 if (TREE_CODE (t) == INTEGER_CST)
348 #if HOST_BITS_PER_WIDE_INT == 64
349 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
352 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
359 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
365 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
370 /* Iff C is a carriage return, warn about it - if appropriate -
371 and return nonzero. */
376 static int newline_warning = 0;
380 /* ANSI C says the effects of a carriage return in a source file
382 if (pedantic && !newline_warning)
384 warning ("carriage return in source file");
385 warning ("(we only warn about the first carriage return)");
393 /* If C is not whitespace, return C.
394 Otherwise skip whitespace and return first nonwhite char read. */
404 /* We don't recognize comments here, because
405 cpp output can include / and * consecutively as operators.
406 Also, there's no need, since cpp removes all comments. */
409 c = check_newline ();
430 error ("stray '\\' in program");
440 /* Skips all of the white space at the current location in the input file.
441 Must use and reset nextchar if it has the next character. */
444 position_after_white_space ()
450 c = nextchar, nextchar = -1;
455 UNGETC (skip_white_space (c));
458 /* Like skip_white_space, but don't advance beyond the end of line.
459 Moreover, we don't get passed a character to start with. */
461 skip_white_space_on_line ()
490 /* Make the token buffer longer, preserving the data in it.
491 P should point to just beyond the last valid character in the old buffer.
492 The value we return is a pointer to the new buffer
493 at a place corresponding to P. */
496 extend_token_buffer (p)
499 int offset = p - token_buffer;
501 maxtoken = maxtoken * 2 + 10;
502 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
504 return token_buffer + offset;
507 #if defined HANDLE_PRAGMA
508 /* Local versions of these macros, that can be passed as function pointers. */
523 /* At the beginning of a line, increment the line number
524 and process any #-directive on this line.
525 If the line is a #-directive, read the entire line and return a newline.
526 Otherwise, return the line's first non-whitespace character. */
536 /* Read first nonwhite char on the line. */
539 while (c == ' ' || c == '\t')
544 /* If not #, return it so caller will use it. */
548 /* Read first nonwhite char after the `#'. */
551 while (c == ' ' || c == '\t')
554 /* If a letter follows, then if the word here is `line', skip
555 it and ignore it; otherwise, ignore the line, with an error
556 if the word isn't `pragma', `ident', `define', or `undef'. */
558 if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
567 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'
568 || whitespace_cr (c) ))
570 while (c == ' ' || c == '\t' || whitespace_cr (c))
575 #if defined HANDLE_PRAGMA || defined HANDLE_GENERIC_PRAGMAS
578 if (token != IDENTIFIER)
580 #endif /* HANDLE_PRAGMA || HANDLE_GENERIC_PRAGMAS */
583 /* We invoke HANDLE_PRAGMA before HANDLE_GENERIC_PRAGMAS (if
584 both are defined), in order to give the back end a chance to
585 override the interpretation of generic style pragmas. */
589 c = nextchar, nextchar = -1;
592 #endif /* !USE_CPPLIB */
594 if (TREE_CODE (yylval.ttype) != IDENTIFIER_NODE)
597 if (HANDLE_PRAGMA (pragma_getc, pragma_ungetc,
598 IDENTIFIER_POINTER (yylval.ttype)))
600 #endif /* HANDLE_PRAGMA */
602 #ifdef HANDLE_GENERIC_PRAGMAS
603 if (handle_generic_pragma (token))
605 #endif /* HANDLE_GENERIC_PRAGMAS */
607 /* Issue a warning message if we have been asked to do so.
608 Ignoring unknown pragmas in system header file unless
609 an explcit -Wunknown-pragmas has been given. */
610 if (warn_unknown_pragmas > 1
611 || (warn_unknown_pragmas && ! in_system_header))
612 warning ("ignoring pragma: %s", token_buffer);
625 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
628 debug_define (lineno, GET_DIRECTIVE_LINE ());
638 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
641 debug_undef (lineno, GET_DIRECTIVE_LINE ());
650 && ((c = GETC()) == ' ' || c == '\t'))
659 && ((c = GETC()) == ' ' || c == '\t'))
661 /* #ident. The pedantic warning is now in cccp.c. */
663 /* Here we have just seen `#ident '.
664 A string constant should follow. */
666 c = skip_white_space_on_line ();
668 /* If no argument, ignore the line. */
675 || TREE_CODE (yylval.ttype) != STRING_CST)
677 error ("invalid #ident");
683 #ifdef ASM_OUTPUT_IDENT
684 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
688 /* Skip the rest of this line. */
693 error ("undefined or invalid # directive");
698 /* Here we have either `#line' or `# <nonletter>'.
699 In either case, it should be a line number; a digit should follow. */
701 /* Can't use skip_white_space here, but must handle all whitespace
702 that is not '\n', lest we get a recursion for '\r' '\n' when
705 c = skip_white_space_on_line ();
707 /* If the # is the only nonwhite char on the line,
708 just ignore it. Check the new newline. */
712 /* Something follows the #; read a token. */
717 if (token == CONSTANT
718 && TREE_CODE (yylval.ttype) == INTEGER_CST)
720 int old_lineno = lineno;
722 /* subtract one, because it is the following line that
723 gets the specified number */
725 int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
727 /* Is this the last nonwhite stuff on the line? */
728 c = skip_white_space_on_line ();
731 /* No more: store the line number and check following line. */
737 /* More follows: it must be a string constant (filename). */
739 /* Read the string constant. */
742 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
744 error ("invalid #line");
749 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
750 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
753 /* Each change of file name
754 reinitializes whether we are now in a system header. */
755 in_system_header = 0;
757 if (main_input_filename == 0)
758 main_input_filename = input_filename;
760 /* Is this the last nonwhite stuff on the line? */
761 c = skip_white_space_on_line ();
764 /* Update the name in the top element of input_file_stack. */
765 if (input_file_stack)
766 input_file_stack->name = input_filename;
775 /* `1' after file name means entering new file.
776 `2' after file name means just left a file. */
778 if (token == CONSTANT
779 && TREE_CODE (yylval.ttype) == INTEGER_CST)
781 if (TREE_INT_CST_LOW (yylval.ttype) == 1)
783 /* Pushing to a new file. */
785 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
786 input_file_stack->line = old_lineno;
787 p->next = input_file_stack;
788 p->name = input_filename;
789 p->indent_level = indent_level;
790 input_file_stack = p;
791 input_file_stack_tick++;
792 debug_start_source_file (input_filename);
795 else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
797 /* Popping out of a file. */
798 if (input_file_stack->next)
800 struct file_stack *p = input_file_stack;
801 if (indent_level != p->indent_level)
803 warning_with_file_and_line
804 (p->name, old_lineno,
805 "This file contains more `%c's than `%c's.",
806 indent_level > p->indent_level ? '{' : '}',
807 indent_level > p->indent_level ? '}' : '{');
809 input_file_stack = p->next;
811 input_file_stack_tick++;
812 debug_end_source_file (input_file_stack->line);
815 error ("#-lines for entering and leaving files don't match");
821 /* Now that we've pushed or popped the input stack,
822 update the name in the top element. */
823 if (input_file_stack)
824 input_file_stack->name = input_filename;
826 /* If we have handled a `1' or a `2',
827 see if there is another number to read. */
830 /* Is this the last nonwhite stuff on the line? */
831 c = skip_white_space_on_line ();
840 /* `3' after file name means this is a system header file. */
842 if (token == CONSTANT
843 && TREE_CODE (yylval.ttype) == INTEGER_CST
844 && TREE_INT_CST_LOW (yylval.ttype) == 3)
845 in_system_header = 1, used_up = 1;
849 /* Is this the last nonwhite stuff on the line? */
850 c = skip_white_space_on_line ();
856 warning ("unrecognized text at end of #line");
859 error ("invalid #-line");
861 /* skip the rest of this line. */
864 if (c != '\n' && c != EOF && nextchar >= 0)
865 c = nextchar, nextchar = -1;
867 while (c != '\n' && c != EOF)
872 #ifdef HANDLE_GENERIC_PRAGMAS
874 /* Handle a #pragma directive.
875 TOKEN is the token we read after `#pragma'. Processes the entire input
876 line and return non-zero iff the pragma has been successfully parsed. */
878 /* This function has to be in this file, in order to get at
882 handle_generic_pragma (token)
895 handle_pragma_token (token_buffer, yylval.ttype);
898 handle_pragma_token (token_buffer, NULL);
902 c = nextchar, nextchar = -1;
907 while (c == ' ' || c == '\t')
911 if (c == '\n' || c == EOF)
912 return handle_pragma_token (NULL, NULL);
918 #endif /* HANDLE_GENERIC_PRAGMAS */
920 #define ENDFILE -1 /* token that represents end-of-file */
922 /* Read an escape sequence, returning its equivalent as a character,
923 or store 1 in *ignore_ptr if it is backslash-newline. */
926 readescape (ignore_ptr)
929 register int c = GETC();
931 register unsigned count;
932 unsigned firstdig = 0;
938 if (warn_traditional)
939 warning ("the meaning of `\\x' varies with -traditional");
941 if (flag_traditional)
950 if (!(c >= 'a' && c <= 'f')
951 && !(c >= 'A' && c <= 'F')
952 && !(c >= '0' && c <= '9'))
958 if (c >= 'a' && c <= 'f')
959 code += c - 'a' + 10;
960 if (c >= 'A' && c <= 'F')
961 code += c - 'A' + 10;
962 if (c >= '0' && c <= '9')
964 if (code != 0 || count != 0)
973 error ("\\x used with no following hex digits");
975 /* Digits are all 0's. Ok. */
977 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
979 && (((unsigned)1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
981 pedwarn ("hex escape out of range");
984 case '0': case '1': case '2': case '3': case '4':
985 case '5': case '6': case '7':
988 while ((c <= '7') && (c >= '0') && (count++ < 3))
990 code = (code * 8) + (c - '0');
996 case '\\': case '\'': case '"':
1005 return TARGET_NEWLINE;
1020 if (warn_traditional)
1021 warning ("the meaning of `\\a' varies with -traditional");
1023 if (flag_traditional)
1028 #if 0 /* Vertical tab is present in common usage compilers. */
1029 if (flag_traditional)
1037 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
1043 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1047 /* `\%' is used to prevent SCCS from getting confused. */
1050 pedwarn ("non-ANSI escape sequence `\\%c'", c);
1053 if (c >= 040 && c < 0177)
1054 pedwarn ("unknown escape sequence `\\%c'", c);
1056 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
1064 char *string = _(msgid);
1066 /* We can't print string and character constants well
1067 because the token_buffer contains the result of processing escapes. */
1069 error ("%s at end of input", string);
1070 else if (token_buffer[0] == 0)
1071 error ("%s at null character", string);
1072 else if (token_buffer[0] == '"')
1073 error ("%s before string constant", string);
1074 else if (token_buffer[0] == '\'')
1075 error ("%s before character constant", string);
1076 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
1077 error ("%s before character 0%o", string, (unsigned char) token_buffer[0]);
1079 error ("%s before `%s'", string, token_buffer);
1089 char long_long_flag;
1092 struct try_type type_sequence[] =
1094 { &integer_type_node, 0, 0, 0},
1095 { &unsigned_type_node, 1, 0, 0},
1096 { &long_integer_type_node, 0, 1, 0},
1097 { &long_unsigned_type_node, 1, 1, 0},
1098 { &long_long_integer_type_node, 0, 1, 1},
1099 { &long_long_unsigned_type_node, 1, 1, 1}
1114 c = nextchar, nextchar = -1;
1119 /* Effectively do c = skip_white_space (c)
1120 but do it faster in the usual cases. */
1133 /* Call skip_white_space so we can warn if appropriate. */
1138 c = skip_white_space (c);
1140 goto found_nonwhite;
1144 token_buffer[0] = c;
1145 token_buffer[1] = 0;
1147 /* yylloc.first_line = lineno; */
1153 token_buffer[0] = 0;
1158 /* Capital L may start a wide-string or wide-character constant. */
1160 register int c = GETC();
1169 goto string_constant;
1176 if (!doing_objc_thang)
1183 /* '@' may start a constant string object. */
1184 register int c = GETC ();
1188 goto string_constant;
1191 /* Fall through to treat '@' as the start of an identifier. */
1194 case 'A': case 'B': case 'C': case 'D': case 'E':
1195 case 'F': case 'G': case 'H': case 'I': case 'J':
1196 case 'K': case 'M': case 'N': case 'O':
1197 case 'P': case 'Q': case 'R': case 'S': case 'T':
1198 case 'U': case 'V': case 'W': case 'X': case 'Y':
1200 case 'a': case 'b': case 'c': case 'd': case 'e':
1201 case 'f': case 'g': case 'h': case 'i': case 'j':
1202 case 'k': case 'l': case 'm': case 'n': case 'o':
1203 case 'p': case 'q': case 'r': case 's': case 't':
1204 case 'u': case 'v': case 'w': case 'x': case 'y':
1210 while (ISALNUM (c) || c == '_' || c == '$' || c == '@')
1212 /* Make sure this char really belongs in an identifier. */
1215 if (! dollars_in_ident)
1216 error ("`$' in identifier");
1218 pedwarn ("`$' in identifier");
1221 if (p >= token_buffer + maxtoken)
1222 p = extend_token_buffer (p);
1238 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1241 register struct resword *ptr;
1243 if ((ptr = is_reserved_word (token_buffer, p - token_buffer)))
1246 yylval.ttype = ridpointers[(int) ptr->rid];
1247 value = (int) ptr->token;
1249 /* Only return OBJECTNAME if it is a typedef. */
1250 if (doing_objc_thang && value == OBJECTNAME)
1252 lastiddecl = lookup_name(yylval.ttype);
1254 if (lastiddecl == NULL_TREE
1255 || TREE_CODE (lastiddecl) != TYPE_DECL)
1259 /* Even if we decided to recognize asm, still perhaps warn. */
1261 && (value == ASM_KEYWORD || value == TYPEOF
1262 || ptr->rid == RID_INLINE)
1263 && token_buffer[0] != '_')
1264 pedwarn ("ANSI does not permit the keyword `%s'",
1269 /* If we did not find a keyword, look for an identifier
1272 if (value == IDENTIFIER)
1274 if (token_buffer[0] == '@')
1275 error("invalid identifier `%s'", token_buffer);
1277 yylval.ttype = get_identifier (token_buffer);
1278 lastiddecl = lookup_name (yylval.ttype);
1280 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1282 /* A user-invisible read-only initialized variable
1283 should be replaced by its value.
1284 We handle only strings since that's the only case used in C. */
1285 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1286 && DECL_IGNORED_P (lastiddecl)
1287 && TREE_READONLY (lastiddecl)
1288 && DECL_INITIAL (lastiddecl) != 0
1289 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1291 tree stringval = DECL_INITIAL (lastiddecl);
1293 /* Copy the string value so that we won't clobber anything
1294 if we put something in the TREE_CHAIN of this one. */
1295 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1296 TREE_STRING_POINTER (stringval));
1299 else if (doing_objc_thang)
1301 tree objc_interface_decl = is_class_name (yylval.ttype);
1303 if (objc_interface_decl)
1306 yylval.ttype = objc_interface_decl;
1316 /* Check first for common special case: single-digit 0 or 1. */
1319 UNGETC (next_c); /* Always undo this lookahead. */
1320 if (!ISALNUM (next_c) && next_c != '.')
1322 token_buffer[0] = (char)c, token_buffer[1] = '\0';
1323 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1329 case '2': case '3': case '4':
1330 case '5': case '6': case '7': case '8': case '9':
1335 int largest_digit = 0;
1337 /* for multi-precision arithmetic,
1338 we actually store only HOST_BITS_PER_CHAR bits in each part.
1339 The number of parts is chosen so as to be sufficient to hold
1340 the enough bits to fit into the two HOST_WIDE_INTs that contain
1341 the integer value (this is always at least as many bits as are
1342 in a target `long long' value, but may be wider). */
1343 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1344 int parts[TOTAL_PARTS];
1347 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS, AFTER_EXPON}
1348 floatflag = NOT_FLOAT;
1350 for (count = 0; count < TOTAL_PARTS; count++)
1358 *p++ = (c = GETC());
1359 if ((c == 'x') || (c == 'X'))
1362 *p++ = (c = GETC());
1364 /* Leading 0 forces octal unless the 0 is the only digit. */
1365 else if (c >= '0' && c <= '9')
1374 /* Read all the digits-and-decimal-points. */
1377 || (ISALNUM (c) && c != 'l' && c != 'L'
1378 && c != 'u' && c != 'U'
1379 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1380 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1384 if (base == 16 && pedantic)
1385 error ("floating constant may not be in radix 16");
1386 if (floatflag == TOO_MANY_POINTS)
1387 /* We have already emitted an error. Don't need another. */
1389 else if (floatflag == AFTER_POINT || floatflag == AFTER_EXPON)
1391 error ("malformed floating constant");
1392 floatflag = TOO_MANY_POINTS;
1393 /* Avoid another error from atof by forcing all characters
1394 from here on to be ignored. */
1398 floatflag = AFTER_POINT;
1403 /* Accept '.' as the start of a floating-point number
1404 only when it is followed by a digit.
1405 Otherwise, unread the following non-digit
1406 and use the '.' as a structural token. */
1407 if (p == token_buffer + 2 && !ISDIGIT (c))
1418 error ("parse error at `..'");
1421 token_buffer[1] = 0;
1428 /* It is not a decimal point.
1429 It should be a digit (perhaps a hex digit). */
1435 else if (base <= 10)
1437 if (c == 'e' || c == 'E')
1440 floatflag = AFTER_EXPON;
1441 break; /* start of exponent */
1443 error ("nondigits in number and not hexadecimal");
1446 else if (base == 16 && (c == 'p' || c == 'P'))
1448 floatflag = AFTER_EXPON;
1449 break; /* start of exponent */
1459 if (c >= largest_digit)
1463 for (count = 0; count < TOTAL_PARTS; count++)
1465 parts[count] *= base;
1469 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1471 &= (1 << HOST_BITS_PER_CHAR) - 1;
1477 /* If the extra highest-order part ever gets anything in it,
1478 the number is certainly too big. */
1479 if (parts[TOTAL_PARTS - 1] != 0)
1482 if (p >= token_buffer + maxtoken - 3)
1483 p = extend_token_buffer (p);
1484 *p++ = (c = GETC());
1489 error ("numeric constant with no digits");
1491 if (largest_digit >= base)
1492 error ("numeric constant contains digits beyond the radix");
1494 /* Remove terminating char from the token buffer and delimit the string */
1497 if (floatflag != NOT_FLOAT)
1499 tree type = double_type_node;
1501 int conversion_errno = 0;
1502 REAL_VALUE_TYPE value;
1505 /* Read explicit exponent if any, and put it in tokenbuf. */
1507 if ((base == 10 && ((c == 'e') || (c == 'E')))
1508 || (base == 16 && (c == 'p' || c == 'P')))
1510 if (p >= token_buffer + maxtoken - 3)
1511 p = extend_token_buffer (p);
1514 if ((c == '+') || (c == '-'))
1519 /* Exponent is decimal, even if string is a hex float. */
1521 error ("floating constant exponent has no digits");
1524 if (p >= token_buffer + maxtoken - 3)
1525 p = extend_token_buffer (p);
1530 if (base == 16 && floatflag != AFTER_EXPON)
1531 error ("hexadecimal floating constant has no exponent");
1535 /* Convert string to a double, checking for overflow. */
1536 if (setjmp (handler))
1538 error ("floating constant out of range");
1543 int fflag = 0, lflag = 0;
1544 /* Copy token_buffer now, while it has just the number
1545 and not the suffixes; once we add `f' or `i',
1546 REAL_VALUE_ATOF may not work any more. */
1547 char *copy = (char *) alloca (p - token_buffer + 1);
1548 bcopy (token_buffer, copy, p - token_buffer + 1);
1550 set_float_handler (handler);
1556 /* Read the suffixes to choose a data type. */
1561 error ("more than one `f' in numeric constant");
1567 error ("more than one `l' in numeric constant");
1573 error ("more than one `i' or `j' in numeric constant");
1575 pedwarn ("ANSI C forbids imaginary numeric constants");
1586 if (p >= token_buffer + maxtoken - 3)
1587 p = extend_token_buffer (p);
1593 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1594 tells the desired precision of the binary result
1595 of decimal-to-binary conversion. */
1600 error ("both `f' and `l' in floating constant");
1602 type = float_type_node;
1605 value = REAL_VALUE_HTOF (copy, TYPE_MODE (type));
1607 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1608 conversion_errno = errno;
1609 /* A diagnostic is required here by some ANSI C testsuites.
1610 This is not pedwarn, because some people don't want
1611 an error for this. */
1612 if (REAL_VALUE_ISINF (value) && pedantic)
1613 warning ("floating point number exceeds range of `float'");
1617 type = long_double_type_node;
1620 value = REAL_VALUE_HTOF (copy, TYPE_MODE (type));
1622 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1623 conversion_errno = errno;
1624 if (REAL_VALUE_ISINF (value) && pedantic)
1625 warning ("floating point number exceeds range of `long double'");
1631 value = REAL_VALUE_HTOF (copy, TYPE_MODE (type));
1633 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1634 conversion_errno = errno;
1635 if (REAL_VALUE_ISINF (value) && pedantic)
1636 warning ("floating point number exceeds range of `double'");
1639 set_float_handler (NULL_PTR);
1642 /* ERANGE is also reported for underflow,
1643 so test the value to distinguish overflow from that. */
1644 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1645 && (REAL_VALUES_LESS (dconst1, value)
1646 || REAL_VALUES_LESS (value, dconstm1)))
1647 warning ("floating point number exceeds range of `double'");
1650 /* If the result is not a number, assume it must have been
1651 due to some error message above, so silently convert
1653 if (REAL_VALUE_ISNAN (value))
1656 /* Create a node with determined type and value. */
1658 yylval.ttype = build_complex (NULL_TREE,
1659 convert (type, integer_zero_node),
1660 build_real (type, value));
1662 yylval.ttype = build_real (type, value);
1666 tree traditional_type, ansi_type, type;
1667 HOST_WIDE_INT high, low;
1668 int spec_unsigned = 0;
1670 int spec_long_long = 0;
1674 traditional_type = ansi_type = type = NULL_TREE;
1677 if (c == 'u' || c == 'U')
1680 error ("two `u's in integer constant");
1683 else if (c == 'l' || c == 'L')
1688 error ("three `l's in integer constant");
1689 else if (pedantic && ! in_system_header && warn_long_long)
1690 pedwarn ("ANSI C forbids long long integer constants");
1695 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1698 error ("more than one `i' or `j' in numeric constant");
1700 pedwarn ("ANSI C forbids imaginary numeric constants");
1705 if (p >= token_buffer + maxtoken - 3)
1706 p = extend_token_buffer (p);
1711 /* If the constant won't fit in an unsigned long long,
1712 then warn that the constant is out of range. */
1714 /* ??? This assumes that long long and long integer types are
1715 a multiple of 8 bits. This better than the original code
1716 though which assumed that long was exactly 32 bits and long
1717 long was exactly 64 bits. */
1719 bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1722 for (i = bytes; i < TOTAL_PARTS; i++)
1726 pedwarn ("integer constant out of range");
1728 /* This is simplified by the fact that our constant
1729 is always positive. */
1733 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1735 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1736 / HOST_BITS_PER_CHAR)]
1737 << (i * HOST_BITS_PER_CHAR));
1738 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1741 yylval.ttype = build_int_2 (low, high);
1742 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1744 /* If warn_traditional, calculate both the ANSI type and the
1745 traditional type, then see if they disagree.
1746 Otherwise, calculate only the type for the dialect in use. */
1747 if (warn_traditional || flag_traditional)
1749 /* Calculate the traditional type. */
1750 /* Traditionally, any constant is signed;
1751 but if unsigned is specified explicitly, obey that.
1752 Use the smallest size with the right number of bits,
1753 except for one special case with decimal constants. */
1754 if (! spec_long && base != 10
1755 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1756 traditional_type = (spec_unsigned ? unsigned_type_node
1757 : integer_type_node);
1758 /* A decimal constant must be long
1759 if it does not fit in type int.
1760 I think this is independent of whether
1761 the constant is signed. */
1762 else if (! spec_long && base == 10
1763 && int_fits_type_p (yylval.ttype, integer_type_node))
1764 traditional_type = (spec_unsigned ? unsigned_type_node
1765 : integer_type_node);
1766 else if (! spec_long_long)
1767 traditional_type = (spec_unsigned ? long_unsigned_type_node
1768 : long_integer_type_node);
1770 traditional_type = (spec_unsigned
1771 ? long_long_unsigned_type_node
1772 : long_long_integer_type_node);
1774 if (warn_traditional || ! flag_traditional)
1776 /* Calculate the ANSI type. */
1777 if (! spec_long && ! spec_unsigned
1778 && int_fits_type_p (yylval.ttype, integer_type_node))
1779 ansi_type = integer_type_node;
1780 else if (! spec_long && (base != 10 || spec_unsigned)
1781 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1782 ansi_type = unsigned_type_node;
1783 else if (! spec_unsigned && !spec_long_long
1784 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1785 ansi_type = long_integer_type_node;
1786 else if (! spec_long_long
1787 && int_fits_type_p (yylval.ttype,
1788 long_unsigned_type_node))
1789 ansi_type = long_unsigned_type_node;
1790 else if (! spec_unsigned
1791 && int_fits_type_p (yylval.ttype,
1792 long_long_integer_type_node))
1793 ansi_type = long_long_integer_type_node;
1795 ansi_type = long_long_unsigned_type_node;
1798 type = flag_traditional ? traditional_type : ansi_type;
1800 if (warn_traditional && traditional_type != ansi_type)
1802 if (TYPE_PRECISION (traditional_type)
1803 != TYPE_PRECISION (ansi_type))
1804 warning ("width of integer constant changes with -traditional");
1805 else if (TREE_UNSIGNED (traditional_type)
1806 != TREE_UNSIGNED (ansi_type))
1807 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1809 warning ("width of integer constant may change on other systems with -traditional");
1812 if (pedantic && !flag_traditional && !spec_long_long && !warn
1813 && (TYPE_PRECISION (long_integer_type_node)
1814 < TYPE_PRECISION (type)))
1815 pedwarn ("integer constant out of range");
1817 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1818 warning ("decimal constant is so large that it is unsigned");
1822 if (TYPE_PRECISION (type)
1823 <= TYPE_PRECISION (integer_type_node))
1825 = build_complex (NULL_TREE, integer_zero_node,
1826 convert (integer_type_node,
1829 error ("complex integer constant is too wide for `complex int'");
1831 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1832 /* The traditional constant 0x80000000 is signed
1833 but doesn't fit in the range of int.
1834 This will change it to -0x80000000, which does fit. */
1836 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1837 yylval.ttype = convert (type, yylval.ttype);
1838 TREE_OVERFLOW (yylval.ttype)
1839 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1842 TREE_TYPE (yylval.ttype) = type;
1848 if (ISALNUM (c) || c == '.' || c == '_' || c == '$'
1849 || (!flag_traditional && (c == '-' || c == '+')
1850 && (p[-1] == 'e' || p[-1] == 'E')))
1851 error ("missing white space after number `%s'", token_buffer);
1853 value = CONSTANT; break;
1859 register int result = 0;
1860 register int num_chars = 0;
1862 unsigned width = TYPE_PRECISION (char_type_node);
1864 #ifdef MULTIBYTE_CHARS
1865 int longest_char = local_mb_cur_max ();
1866 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1869 max_chars = TYPE_PRECISION (integer_type_node) / width;
1871 width = WCHAR_TYPE_SIZE;
1878 if (c == '\'' || c == EOF)
1885 c = readescape (&ignore);
1888 if (width < HOST_BITS_PER_INT
1889 && (unsigned) c >= ((unsigned)1 << width))
1890 pedwarn ("escape sequence out of range for character");
1891 #ifdef MAP_CHARACTER
1893 c = MAP_CHARACTER (c);
1899 pedwarn ("ANSI C forbids newline in character constant");
1904 #ifdef MULTIBYTE_CHARS
1908 for (i = 1; i <= longest_char; ++i)
1910 if (i > maxtoken - 4)
1911 extend_token_buffer (token_buffer);
1913 token_buffer[i] = c;
1914 char_len = local_mbtowc (& wc,
1923 /* mbtowc sometimes needs an extra char before accepting */
1928 /* Merge character into result; ignore excess chars. */
1929 for (i = 1; i <= char_len; ++i)
1933 if (width < HOST_BITS_PER_INT)
1934 result = (result << width)
1936 & ((1 << width) - 1));
1938 result = token_buffer[i];
1940 num_chars += char_len;
1948 warning ("Ignoring invalid multibyte character");
1951 #ifdef MAP_CHARACTER
1953 c = MAP_CHARACTER (c);
1956 #else /* ! MULTIBYTE_CHARS */
1957 #ifdef MAP_CHARACTER
1958 c = MAP_CHARACTER (c);
1960 #endif /* ! MULTIBYTE_CHARS */
1965 if (chars_seen == 1) /* only keep the first one */
1970 /* Merge character into result; ignore excess chars. */
1971 num_chars += (width / TYPE_PRECISION (char_type_node));
1972 if (num_chars < max_chars + 1)
1974 if (width < HOST_BITS_PER_INT)
1975 result = (result << width) | (c & ((1 << width) - 1));
1982 error ("malformatted character constant");
1983 else if (chars_seen == 0)
1984 error ("empty character constant");
1985 else if (num_chars > max_chars)
1987 num_chars = max_chars;
1988 error ("character constant too long");
1990 else if (chars_seen != 1 && ! flag_traditional && warn_multichar)
1991 warning ("multi-character character constant");
1993 /* If char type is signed, sign-extend the constant. */
1996 int num_bits = num_chars * width;
1998 /* We already got an error; avoid invalid shift. */
1999 yylval.ttype = build_int_2 (0, 0);
2000 else if (TREE_UNSIGNED (char_type_node)
2001 || ((result >> (num_bits - 1)) & 1) == 0)
2003 = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
2004 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2008 = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
2009 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2011 TREE_TYPE (yylval.ttype) = integer_type_node;
2015 yylval.ttype = build_int_2 (result, 0);
2016 TREE_TYPE (yylval.ttype) = wchar_type_node;
2026 unsigned width = wide_flag ? WCHAR_TYPE_SIZE
2027 : TYPE_PRECISION (char_type_node);
2028 #ifdef MULTIBYTE_CHARS
2029 int longest_char = local_mb_cur_max ();
2030 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
2033 p = token_buffer + 1;
2035 while (c != '"' && c >= 0)
2040 c = readescape (&ignore);
2043 if (width < HOST_BITS_PER_INT
2044 && (unsigned) c >= ((unsigned)1 << width))
2045 pedwarn ("escape sequence out of range for character");
2050 pedwarn ("ANSI C forbids newline in string constant");
2055 #ifdef MULTIBYTE_CHARS
2059 for (i = 0; i < longest_char; ++i)
2061 if (p + i >= token_buffer + maxtoken)
2062 p = extend_token_buffer (p);
2065 char_len = local_mbtowc (& wc, p, i + 1);
2071 warning ("Ignoring invalid multibyte character");
2074 /* mbtowc sometimes needs an extra char before accepting */
2085 #endif /* MULTIBYTE_CHARS */
2088 /* Add this single character into the buffer either as a wchar_t
2089 or as a single byte. */
2092 unsigned width = TYPE_PRECISION (char_type_node);
2093 unsigned bytemask = (1 << width) - 1;
2096 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2097 p = extend_token_buffer (p);
2099 for (byte = 0; byte < WCHAR_BYTES; ++byte)
2102 if (byte >= (int) sizeof (c))
2105 value = (c >> (byte * width)) & bytemask;
2106 if (BYTES_BIG_ENDIAN)
2107 p[WCHAR_BYTES - byte - 1] = value;
2115 if (p >= token_buffer + maxtoken)
2116 p = extend_token_buffer (p);
2124 /* Terminate the string value, either with a single byte zero
2125 or with a wide zero. */
2128 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2129 p = extend_token_buffer (p);
2130 bzero (p, WCHAR_BYTES);
2135 if (p >= token_buffer + maxtoken)
2136 p = extend_token_buffer (p);
2141 error ("Unterminated string constant");
2143 /* We have read the entire constant.
2144 Construct a STRING_CST for the result. */
2148 yylval.ttype = build_string (p - (token_buffer + 1),
2150 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2155 /* Return an Objective-C @"..." constant string object. */
2156 yylval.ttype = build_objc_string (p - (token_buffer + 1),
2158 TREE_TYPE (yylval.ttype) = char_array_type_node;
2159 value = OBJC_STRING;
2163 yylval.ttype = build_string (p - (token_buffer + 1),
2165 TREE_TYPE (yylval.ttype) = char_array_type_node;
2193 yylval.code = PLUS_EXPR; break;
2195 yylval.code = MINUS_EXPR; break;
2197 yylval.code = BIT_AND_EXPR; break;
2199 yylval.code = BIT_IOR_EXPR; break;
2201 yylval.code = MULT_EXPR; break;
2203 yylval.code = TRUNC_DIV_EXPR; break;
2205 yylval.code = TRUNC_MOD_EXPR; break;
2207 yylval.code = BIT_XOR_EXPR; break;
2209 yylval.code = LSHIFT_EXPR; break;
2211 yylval.code = RSHIFT_EXPR; break;
2213 yylval.code = LT_EXPR; break;
2215 yylval.code = GT_EXPR; break;
2218 token_buffer[1] = c1 = GETC();
2219 token_buffer[2] = 0;
2226 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2228 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2230 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2232 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2234 value = ASSIGN; goto done;
2240 value = PLUSPLUS; goto done;
2242 value = MINUSMINUS; goto done;
2244 value = ANDAND; goto done;
2246 value = OROR; goto done;
2259 { value = POINTSAT; goto done; }
2263 { value = ']'; goto done; }
2267 { value = '{'; indent_level++; goto done; }
2269 { value = '['; goto done; }
2273 { value = '}'; indent_level--; goto done; }
2277 token_buffer[1] = 0;
2279 if ((c == '<') || (c == '>'))
2280 value = ARITHCOMPARE;
2286 /* Don't make yyparse think this is eof. */
2305 /* yylloc.last_line = lineno; */
2310 /* Sets the value of the 'yydebug' variable to VALUE.
2311 This is a function so we don't have to have YYDEBUG defined
2312 in order to build the compiler. */
2321 warning ("YYDEBUG not defined.");