* configure: Rebuilt.
[official-gcc.git] / gcc / c-lex.c
blob27c65f3fa27f20f912a3b00b5f058f8f7b541de8
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)
9 any later version.
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. */
21 #include "config.h"
22 #include "system.h"
24 #include "rtl.h"
25 #include "tree.h"
26 #include "input.h"
27 #include "output.h"
28 #include "c-lex.h"
29 #include "c-tree.h"
30 #include "flags.h"
31 #include "c-parse.h"
32 #include "c-pragma.h"
33 #include "toplev.h"
34 #include "intl.h"
36 /* MULTIBYTE_CHARS support only works for native compilers.
37 ??? Ideally what we want is to model widechar support after
38 the current floating point support. */
39 #ifdef CROSS_COMPILE
40 #undef MULTIBYTE_CHARS
41 #endif
43 #ifdef MULTIBYTE_CHARS
44 #include "mbchar.h"
45 #include <locale.h>
46 #endif /* MULTIBYTE_CHARS */
48 #if USE_CPPLIB
49 #include "cpplib.h"
50 extern cpp_reader parse_in;
51 extern cpp_options parse_options;
52 #else
53 /* Stream for reading from the input file. */
54 FILE *finput;
55 #endif
57 extern void yyprint PROTO((FILE *, int, YYSTYPE));
59 /* The elements of `ridpointers' are identifier nodes
60 for the reserved type names and storage classes.
61 It is indexed by a RID_... value. */
62 tree ridpointers[(int) RID_MAX];
64 /* Cause the `yydebug' variable to be defined. */
65 #define YYDEBUG 1
67 #if USE_CPPLIB
68 extern unsigned char *yy_cur, *yy_lim;
70 extern int yy_get_token ();
72 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
73 #define UNGETC(c) ((c) == EOF ? 0 : yy_cur--)
74 #else
75 #define GETC() getc (finput)
76 #define UNGETC(c) ungetc (c, finput)
77 #endif
79 /* the declaration found for the last IDENTIFIER token read in.
80 yylex must look this up to detect typedefs, which get token type TYPENAME,
81 so it is left around in case the identifier is not a typedef but is
82 used in a context which makes it a reference to a variable. */
83 tree lastiddecl;
85 /* Nonzero enables objc features. */
87 int doing_objc_thang;
89 extern int yydebug;
91 /* File used for outputting assembler code. */
92 extern FILE *asm_out_file;
94 #ifndef WCHAR_TYPE_SIZE
95 #ifdef INT_TYPE_SIZE
96 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
97 #else
98 #define WCHAR_TYPE_SIZE BITS_PER_WORD
99 #endif
100 #endif
102 /* Number of bytes in a wide character. */
103 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
105 static int maxtoken; /* Current nominal length of token buffer. */
106 char *token_buffer; /* Pointer to token buffer.
107 Actual allocated length is maxtoken + 2.
108 This is not static because objc-parse.y uses it. */
110 static int indent_level = 0; /* Number of { minus number of }. */
112 /* Nonzero if end-of-file has been seen on input. */
113 static int end_of_file;
115 #if !USE_CPPLIB
116 /* Buffered-back input character; faster than using ungetc. */
117 static int nextchar = -1;
118 #endif
120 #ifdef HANDLE_GENERIC_PRAGMAS
121 static int handle_generic_pragma PROTO((int));
122 #endif /* HANDLE_GENERIC_PRAGMAS */
123 static int whitespace_cr PROTO((int));
124 static int skip_white_space PROTO((int));
125 static int skip_white_space_on_line PROTO((void));
126 static char *extend_token_buffer PROTO((const char *));
127 static int readescape PROTO((int *));
128 static void parse_float PROTO((PTR));
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. */
133 #include "c-gperf.h"
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. */
143 tree
144 make_pointer_declarator (type_quals, target)
145 tree type_quals, target;
147 return build1 (INDIRECT_REF, type_quals, target);
150 void
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 = "";
161 void
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";
181 char *
182 init_parse (filename)
183 char *filename;
185 #if !USE_CPPLIB
186 /* Open input file. */
187 if (filename == 0 || !strcmp (filename, "-"))
189 finput = stdin;
190 filename = "stdin";
192 else
193 finput = fopen (filename, "r");
194 if (finput == 0)
195 pfatal_with_name (filename);
197 #ifdef IO_BUFFER_SIZE
198 setvbuf (finput, (char *) xmalloc (IO_BUFFER_SIZE), _IOFBF, IO_BUFFER_SIZE);
199 #endif
200 #else /* !USE_CPPLIB */
201 parse_in.show_column = 1;
202 if (! cpp_start_read (&parse_in, filename))
203 abort ();
205 if (filename == 0 || !strcmp (filename, "-"))
206 filename = "stdin";
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);
212 #endif
214 init_lex ();
216 return filename;
219 void
220 finish_parse ()
222 #if USE_CPPLIB
223 cpp_finish (&parse_in);
224 #else
225 fclose (finput);
226 #endif
229 void
230 init_lex ()
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. */
237 lineno = 0;
239 #ifdef MULTIBYTE_CHARS
240 /* Change to the native locale for multibyte conversions. */
241 setlocale (LC_CTYPE, "");
242 literal_codeset = getenv ("LANG");
243 #endif
245 maxtoken = 40;
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");
300 if (flag_no_asm)
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");
310 void
311 reinit_parse_for_function ()
315 /* Function used when yydebug is set, to print a token in more detail. */
317 void
318 yyprint (file, yychar, yylval)
319 FILE *file;
320 int yychar;
321 YYSTYPE yylval;
323 tree t;
324 switch (yychar)
326 case IDENTIFIER:
327 case TYPENAME:
328 case OBJECTNAME:
329 t = yylval.ttype;
330 if (IDENTIFIER_POINTER (t))
331 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
332 break;
334 case CONSTANT:
335 t = yylval.ttype;
336 if (TREE_CODE (t) == INTEGER_CST)
337 fprintf (file,
338 #if HOST_BITS_PER_WIDE_INT == 64
339 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
340 " 0x%x%016x",
341 #else
342 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
343 " 0x%lx%016lx",
344 #else
345 " 0x%llx%016llx",
346 #endif
347 #endif
348 #else
349 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
350 " 0x%lx%08lx",
351 #else
352 " 0x%x%08x",
353 #endif
354 #endif
355 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
356 break;
360 /* Iff C is a carriage return, warn about it - if appropriate -
361 and return nonzero. */
362 static int
363 whitespace_cr (c)
364 int c;
366 static int newline_warning = 0;
368 if (c == '\r')
370 /* ANSI C says the effects of a carriage return in a source file
371 are undefined. */
372 if (pedantic && !newline_warning)
374 warning ("carriage return in source file");
375 warning ("(we only warn about the first carriage return)");
376 newline_warning = 1;
378 return 1;
380 return 0;
383 /* If C is not whitespace, return C.
384 Otherwise skip whitespace and return first nonwhite char read. */
386 static int
387 skip_white_space (c)
388 register int c;
390 for (;;)
392 switch (c)
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. */
398 case '\n':
399 c = check_newline ();
400 break;
402 case ' ':
403 case '\t':
404 case '\f':
405 case '\v':
406 case '\b':
407 c = GETC();
408 break;
410 case '\r':
411 whitespace_cr (c);
412 c = GETC();
413 break;
415 case '\\':
416 c = GETC();
417 if (c == '\n')
418 lineno++;
419 else
420 error ("stray '\\' in program");
421 c = GETC();
422 break;
424 default:
425 return (c);
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. */
433 void
434 position_after_white_space ()
436 register int c;
438 #if !USE_CPPLIB
439 if (nextchar != -1)
440 c = nextchar, nextchar = -1;
441 else
442 #endif
443 c = GETC();
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. */
450 static int
451 skip_white_space_on_line ()
453 register int c;
455 while (1)
457 c = GETC();
458 switch (c)
460 case '\n':
461 default:
462 break;
464 case ' ':
465 case '\t':
466 case '\f':
467 case '\v':
468 case '\b':
469 continue;
471 case '\r':
472 whitespace_cr (c);
473 continue;
475 break;
477 return c;
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. */
485 static char *
486 extend_token_buffer (p)
487 const char *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. */
499 static int
500 pragma_getc ()
502 return GETC();
505 static void
506 pragma_ungetc (arg)
507 int arg;
509 UNGETC (arg);
511 #endif
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. */
519 check_newline ()
521 register int c;
522 register int token;
524 lineno++;
526 /* Read first nonwhite char on the line. */
528 c = GETC();
529 while (c == ' ' || c == '\t')
530 c = GETC();
532 if (c != '#')
534 /* If not #, return it so caller will use it. */
535 return c;
538 /* Read first nonwhite char after the `#'. */
540 c = GETC();
541 while (c == ' ' || c == '\t')
542 c = GETC();
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'))
550 if (c == 'p')
552 if (GETC() == 'r'
553 && GETC() == 'a'
554 && GETC() == 'g'
555 && GETC() == 'm'
556 && GETC() == 'a'
557 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'
558 || whitespace_cr (c) ))
560 while (c == ' ' || c == '\t' || whitespace_cr (c))
561 c = GETC ();
562 if (c == '\n')
563 return c;
565 #if defined HANDLE_PRAGMA || defined HANDLE_GENERIC_PRAGMAS
566 UNGETC (c);
567 token = yylex ();
568 if (token != IDENTIFIER)
569 goto skipline;
570 #endif /* HANDLE_PRAGMA || HANDLE_GENERIC_PRAGMAS */
572 #ifdef HANDLE_PRAGMA
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. */
576 #if !USE_CPPLIB
577 if (nextchar >= 0)
579 c = nextchar, nextchar = -1;
580 UNGETC (c);
582 #endif /* !USE_CPPLIB */
584 if (TREE_CODE (yylval.ttype) != IDENTIFIER_NODE)
585 goto skipline;
587 if (HANDLE_PRAGMA (pragma_getc, pragma_ungetc,
588 IDENTIFIER_POINTER (yylval.ttype)))
589 return GETC ();
590 #endif /* HANDLE_PRAGMA */
592 #ifdef HANDLE_GENERIC_PRAGMAS
593 if (handle_generic_pragma (token))
594 return GETC ();
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);
604 goto skipline;
608 else if (c == 'd')
610 if (GETC() == 'e'
611 && GETC() == 'f'
612 && GETC() == 'i'
613 && GETC() == 'n'
614 && GETC() == 'e'
615 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
617 if (c != '\n')
618 debug_define (lineno, GET_DIRECTIVE_LINE ());
619 goto skipline;
622 else if (c == 'u')
624 if (GETC() == 'n'
625 && GETC() == 'd'
626 && GETC() == 'e'
627 && GETC() == 'f'
628 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
630 if (c != '\n')
631 debug_undef (lineno, GET_DIRECTIVE_LINE ());
632 goto skipline;
635 else if (c == 'l')
637 if (GETC() == 'i'
638 && GETC() == 'n'
639 && GETC() == 'e'
640 && ((c = GETC()) == ' ' || c == '\t'))
641 goto linenum;
643 else if (c == 'i')
645 if (GETC() == 'd'
646 && GETC() == 'e'
647 && GETC() == 'n'
648 && GETC() == '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. */
659 if (c == '\n')
660 return c;
662 UNGETC (c);
663 token = yylex ();
664 if (token != STRING
665 || TREE_CODE (yylval.ttype) != STRING_CST)
667 error ("invalid #ident");
668 goto skipline;
671 if (!flag_no_ident)
673 #ifdef ASM_OUTPUT_IDENT
674 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
675 #endif
678 /* Skip the rest of this line. */
679 goto skipline;
683 error ("undefined or invalid # directive");
684 goto skipline;
687 linenum:
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
693 calling yylex. */
694 UNGETC (c);
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. */
699 if (c == '\n')
700 return c;
702 /* Something follows the #; read a token. */
704 UNGETC (c);
705 token = yylex ();
707 if (token == CONSTANT
708 && TREE_CODE (yylval.ttype) == INTEGER_CST)
710 int old_lineno = lineno;
711 int used_up = 0;
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 ();
719 if (c == '\n')
721 /* No more: store the line number and check following line. */
722 lineno = l;
723 return c;
725 UNGETC (c);
727 /* More follows: it must be a string constant (filename). */
729 /* Read the string constant. */
730 token = yylex ();
732 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
734 error ("invalid #line");
735 goto skipline;
738 input_filename
739 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
740 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
741 lineno = l;
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 ();
752 if (c == '\n')
754 /* Update the name in the top element of input_file_stack. */
755 if (input_file_stack)
756 input_file_stack->name = input_filename;
758 return c;
760 UNGETC (c);
762 token = yylex ();
763 used_up = 0;
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. */
774 struct file_stack *p
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);
783 used_up = 1;
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;
800 free (p);
801 input_file_stack_tick++;
802 debug_end_source_file (input_file_stack->line);
804 else
805 error ("#-lines for entering and leaving files don't match");
807 used_up = 1;
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. */
818 if (used_up)
820 /* Is this the last nonwhite stuff on the line? */
821 c = skip_white_space_on_line ();
822 if (c == '\n')
823 return c;
824 UNGETC (c);
826 token = yylex ();
827 used_up = 0;
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;
837 if (used_up)
839 /* Is this the last nonwhite stuff on the line? */
840 c = skip_white_space_on_line ();
841 if (c == '\n')
842 return c;
843 UNGETC (c);
846 warning ("unrecognized text at end of #line");
848 else
849 error ("invalid #-line");
851 /* skip the rest of this line. */
852 skipline:
853 #if !USE_CPPLIB
854 if (c != '\n' && c != EOF && nextchar >= 0)
855 c = nextchar, nextchar = -1;
856 #endif
857 while (c != '\n' && c != EOF)
858 c = GETC();
859 return c;
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
869 the token types. */
871 static int
872 handle_generic_pragma (token)
873 register int token;
875 register int c;
877 for (;;)
879 switch (token)
881 case IDENTIFIER:
882 case TYPENAME:
883 case STRING:
884 case CONSTANT:
885 handle_pragma_token (token_buffer, yylval.ttype);
886 break;
887 default:
888 handle_pragma_token (token_buffer, NULL);
890 #if !USE_CPPLIB
891 if (nextchar >= 0)
892 c = nextchar, nextchar = -1;
893 else
894 #endif
895 c = GETC ();
897 while (c == ' ' || c == '\t')
898 c = GETC ();
899 UNGETC (c);
901 if (c == '\n' || c == EOF)
902 return handle_pragma_token (NULL, NULL);
904 token = yylex ();
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. */
915 static int
916 readescape (ignore_ptr)
917 int *ignore_ptr;
919 register int c = GETC();
920 register int code;
921 register unsigned count;
922 unsigned firstdig = 0;
923 int nonnull;
925 switch (c)
927 case 'x':
928 if (warn_traditional)
929 warning ("the meaning of `\\x' varies with -traditional");
931 if (flag_traditional)
932 return c;
934 code = 0;
935 count = 0;
936 nonnull = 0;
937 while (1)
939 c = GETC();
940 if (!(c >= 'a' && c <= 'f')
941 && !(c >= 'A' && c <= 'F')
942 && !(c >= '0' && c <= '9'))
944 UNGETC (c);
945 break;
947 code *= 16;
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')
953 code += c - '0';
954 if (code != 0 || count != 0)
956 if (count == 0)
957 firstdig = code;
958 count++;
960 nonnull = 1;
962 if (! nonnull)
963 error ("\\x used with no following hex digits");
964 else if (count == 0)
965 /* Digits are all 0's. Ok. */
967 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
968 || (count > 1
969 && (((unsigned)1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
970 <= firstdig)))
971 pedwarn ("hex escape out of range");
972 return code;
974 case '0': case '1': case '2': case '3': case '4':
975 case '5': case '6': case '7':
976 code = 0;
977 count = 0;
978 while ((c <= '7') && (c >= '0') && (count++ < 3))
980 code = (code * 8) + (c - '0');
981 c = GETC();
983 UNGETC (c);
984 return code;
986 case '\\': case '\'': case '"':
987 return c;
989 case '\n':
990 lineno++;
991 *ignore_ptr = 1;
992 return 0;
994 case 'n':
995 return TARGET_NEWLINE;
997 case 't':
998 return TARGET_TAB;
1000 case 'r':
1001 return TARGET_CR;
1003 case 'f':
1004 return TARGET_FF;
1006 case 'b':
1007 return TARGET_BS;
1009 case 'a':
1010 if (warn_traditional)
1011 warning ("the meaning of `\\a' varies with -traditional");
1013 if (flag_traditional)
1014 return c;
1015 return TARGET_BELL;
1017 case 'v':
1018 #if 0 /* Vertical tab is present in common usage compilers. */
1019 if (flag_traditional)
1020 return c;
1021 #endif
1022 return TARGET_VT;
1024 case 'e':
1025 case 'E':
1026 if (pedantic)
1027 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
1028 return 033;
1030 case '?':
1031 return c;
1033 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1034 case '(':
1035 case '{':
1036 case '[':
1037 /* `\%' is used to prevent SCCS from getting confused. */
1038 case '%':
1039 if (pedantic)
1040 pedwarn ("non-ANSI escape sequence `\\%c'", c);
1041 return c;
1043 if (c >= 040 && c < 0177)
1044 pedwarn ("unknown escape sequence `\\%c'", c);
1045 else
1046 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
1047 return c;
1050 void
1051 yyerror (msgid)
1052 const char *msgid;
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. */
1058 if (end_of_file)
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]);
1068 else
1069 error ("%s before `%s'", string, token_buffer);
1072 #if 0
1074 struct try_type
1076 tree *node_var;
1077 char unsigned_flag;
1078 char long_flag;
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}
1091 #endif /* 0 */
1093 struct pf_args
1095 /* Input */
1096 int base;
1097 char * p;
1098 /* I/O */
1099 int c;
1100 int imag;
1101 tree type;
1102 int conversion_errno;
1103 /* Output */
1104 REAL_VALUE_TYPE value;
1107 static void
1108 parse_float (data)
1109 PTR data;
1111 struct pf_args * args = (struct pf_args *) data;
1112 int fflag = 0, lflag = 0;
1113 /* Copy token_buffer now, while it has just the number
1114 and not the suffixes; once we add `f' or `i',
1115 REAL_VALUE_ATOF may not work any more. */
1116 char *copy = (char *) alloca (args->p - token_buffer + 1);
1117 bcopy (token_buffer, copy, args->p - token_buffer + 1);
1119 while (1)
1121 int lose = 0;
1123 /* Read the suffixes to choose a data type. */
1124 switch (args->c)
1126 case 'f': case 'F':
1127 if (fflag)
1128 error ("more than one `f' in numeric constant");
1129 fflag = 1;
1130 break;
1132 case 'l': case 'L':
1133 if (lflag)
1134 error ("more than one `l' in numeric constant");
1135 lflag = 1;
1136 break;
1138 case 'i': case 'I':
1139 if (args->imag)
1140 error ("more than one `i' or `j' in numeric constant");
1141 else if (pedantic)
1142 pedwarn ("ANSI C forbids imaginary numeric constants");
1143 args->imag = 1;
1144 break;
1146 default:
1147 lose = 1;
1150 if (lose)
1151 break;
1153 if (args->p >= token_buffer + maxtoken - 3)
1154 args->p = extend_token_buffer (args->p);
1155 *(args->p++) = args->c;
1156 *(args->p) = 0;
1157 args->c = GETC();
1160 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1161 tells the desired precision of the binary result
1162 of decimal-to-binary conversion. */
1164 if (fflag)
1166 if (lflag)
1167 error ("both `f' and `l' in floating constant");
1169 args->type = float_type_node;
1170 errno = 0;
1171 if (args->base == 16)
1172 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1173 else
1174 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1175 args->conversion_errno = errno;
1176 /* A diagnostic is required here by some ANSI C testsuites.
1177 This is not pedwarn, because some people don't want
1178 an error for this. */
1179 if (REAL_VALUE_ISINF (args->value) && pedantic)
1180 warning ("floating point number exceeds range of `float'");
1182 else if (lflag)
1184 args->type = long_double_type_node;
1185 errno = 0;
1186 if (args->base == 16)
1187 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1188 else
1189 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1190 args->conversion_errno = errno;
1191 if (REAL_VALUE_ISINF (args->value) && pedantic)
1192 warning ("floating point number exceeds range of `long double'");
1194 else
1196 errno = 0;
1197 if (args->base == 16)
1198 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1199 else
1200 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1201 args->conversion_errno = errno;
1202 if (REAL_VALUE_ISINF (args->value) && pedantic)
1203 warning ("floating point number exceeds range of `double'");
1208 yylex ()
1210 register int c;
1211 register char *p;
1212 register int value;
1213 int wide_flag = 0;
1214 int objc_flag = 0;
1216 #if !USE_CPPLIB
1217 if (nextchar >= 0)
1218 c = nextchar, nextchar = -1;
1219 else
1220 #endif
1221 c = GETC();
1223 /* Effectively do c = skip_white_space (c)
1224 but do it faster in the usual cases. */
1225 while (1)
1226 switch (c)
1228 case ' ':
1229 case '\t':
1230 case '\f':
1231 case '\v':
1232 case '\b':
1233 c = GETC();
1234 break;
1236 case '\r':
1237 /* Call skip_white_space so we can warn if appropriate. */
1239 case '\n':
1240 case '/':
1241 case '\\':
1242 c = skip_white_space (c);
1243 default:
1244 goto found_nonwhite;
1246 found_nonwhite:
1248 token_buffer[0] = c;
1249 token_buffer[1] = 0;
1251 /* yylloc.first_line = lineno; */
1253 switch (c)
1255 case EOF:
1256 end_of_file = 1;
1257 token_buffer[0] = 0;
1258 value = ENDFILE;
1259 break;
1261 case 'L':
1262 /* Capital L may start a wide-string or wide-character constant. */
1264 register int c = GETC();
1265 if (c == '\'')
1267 wide_flag = 1;
1268 goto char_constant;
1270 if (c == '"')
1272 wide_flag = 1;
1273 goto string_constant;
1275 UNGETC (c);
1277 goto letter;
1279 case '@':
1280 if (!doing_objc_thang)
1282 value = c;
1283 break;
1285 else
1287 /* '@' may start a constant string object. */
1288 register int c = GETC ();
1289 if (c == '"')
1291 objc_flag = 1;
1292 goto string_constant;
1294 UNGETC (c);
1295 /* Fall through to treat '@' as the start of an identifier. */
1298 case 'A': case 'B': case 'C': case 'D': case 'E':
1299 case 'F': case 'G': case 'H': case 'I': case 'J':
1300 case 'K': case 'M': case 'N': case 'O':
1301 case 'P': case 'Q': case 'R': case 'S': case 'T':
1302 case 'U': case 'V': case 'W': case 'X': case 'Y':
1303 case 'Z':
1304 case 'a': case 'b': case 'c': case 'd': case 'e':
1305 case 'f': case 'g': case 'h': case 'i': case 'j':
1306 case 'k': case 'l': case 'm': case 'n': case 'o':
1307 case 'p': case 'q': case 'r': case 's': case 't':
1308 case 'u': case 'v': case 'w': case 'x': case 'y':
1309 case 'z':
1310 case '_':
1311 case '$':
1312 letter:
1313 p = token_buffer;
1314 while (ISALNUM (c) || c == '_' || c == '$' || c == '@')
1316 /* Make sure this char really belongs in an identifier. */
1317 if (c == '$')
1319 if (! dollars_in_ident)
1320 error ("`$' in identifier");
1321 else if (pedantic)
1322 pedwarn ("`$' in identifier");
1325 if (p >= token_buffer + maxtoken)
1326 p = extend_token_buffer (p);
1328 *p++ = c;
1329 c = GETC();
1332 *p = 0;
1333 #if USE_CPPLIB
1334 UNGETC (c);
1335 #else
1336 nextchar = c;
1337 #endif
1339 value = IDENTIFIER;
1340 yylval.itype = 0;
1342 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1345 register struct resword *ptr;
1347 if ((ptr = is_reserved_word (token_buffer, p - token_buffer)))
1349 if (ptr->rid)
1350 yylval.ttype = ridpointers[(int) ptr->rid];
1351 value = (int) ptr->token;
1353 /* Only return OBJECTNAME if it is a typedef. */
1354 if (doing_objc_thang && value == OBJECTNAME)
1356 lastiddecl = lookup_name(yylval.ttype);
1358 if (lastiddecl == NULL_TREE
1359 || TREE_CODE (lastiddecl) != TYPE_DECL)
1360 value = IDENTIFIER;
1363 /* Even if we decided to recognize asm, still perhaps warn. */
1364 if (pedantic
1365 && (value == ASM_KEYWORD || value == TYPEOF
1366 || ptr->rid == RID_INLINE)
1367 && token_buffer[0] != '_')
1368 pedwarn ("ANSI does not permit the keyword `%s'",
1369 token_buffer);
1373 /* If we did not find a keyword, look for an identifier
1374 (or a typename). */
1376 if (value == IDENTIFIER)
1378 if (token_buffer[0] == '@')
1379 error("invalid identifier `%s'", token_buffer);
1381 yylval.ttype = get_identifier (token_buffer);
1382 lastiddecl = lookup_name (yylval.ttype);
1384 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1385 value = TYPENAME;
1386 /* A user-invisible read-only initialized variable
1387 should be replaced by its value.
1388 We handle only strings since that's the only case used in C. */
1389 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1390 && DECL_IGNORED_P (lastiddecl)
1391 && TREE_READONLY (lastiddecl)
1392 && DECL_INITIAL (lastiddecl) != 0
1393 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1395 tree stringval = DECL_INITIAL (lastiddecl);
1397 /* Copy the string value so that we won't clobber anything
1398 if we put something in the TREE_CHAIN of this one. */
1399 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1400 TREE_STRING_POINTER (stringval));
1401 value = STRING;
1403 else if (doing_objc_thang)
1405 tree objc_interface_decl = is_class_name (yylval.ttype);
1407 if (objc_interface_decl)
1409 value = CLASSNAME;
1410 yylval.ttype = objc_interface_decl;
1415 break;
1417 case '0': case '1':
1419 int next_c;
1420 /* Check first for common special case: single-digit 0 or 1. */
1422 next_c = GETC ();
1423 UNGETC (next_c); /* Always undo this lookahead. */
1424 if (!ISALNUM (next_c) && next_c != '.')
1426 token_buffer[0] = (char)c, token_buffer[1] = '\0';
1427 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1428 value = CONSTANT;
1429 break;
1431 /*FALLTHRU*/
1433 case '2': case '3': case '4':
1434 case '5': case '6': case '7': case '8': case '9':
1435 case '.':
1437 int base = 10;
1438 int count = 0;
1439 int largest_digit = 0;
1440 int numdigits = 0;
1441 /* for multi-precision arithmetic,
1442 we actually store only HOST_BITS_PER_CHAR bits in each part.
1443 The number of parts is chosen so as to be sufficient to hold
1444 the enough bits to fit into the two HOST_WIDE_INTs that contain
1445 the integer value (this is always at least as many bits as are
1446 in a target `long long' value, but may be wider). */
1447 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1448 int parts[TOTAL_PARTS];
1449 int overflow = 0;
1451 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS, AFTER_EXPON}
1452 floatflag = NOT_FLOAT;
1454 for (count = 0; count < TOTAL_PARTS; count++)
1455 parts[count] = 0;
1457 p = token_buffer;
1458 *p++ = c;
1460 if (c == '0')
1462 *p++ = (c = GETC());
1463 if ((c == 'x') || (c == 'X'))
1465 base = 16;
1466 *p++ = (c = GETC());
1468 /* Leading 0 forces octal unless the 0 is the only digit. */
1469 else if (c >= '0' && c <= '9')
1471 base = 8;
1472 numdigits++;
1474 else
1475 numdigits++;
1478 /* Read all the digits-and-decimal-points. */
1480 while (c == '.'
1481 || (ISALNUM (c) && c != 'l' && c != 'L'
1482 && c != 'u' && c != 'U'
1483 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1484 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1486 if (c == '.')
1488 if (base == 16 && pedantic)
1489 error ("floating constant may not be in radix 16");
1490 if (floatflag == TOO_MANY_POINTS)
1491 /* We have already emitted an error. Don't need another. */
1493 else if (floatflag == AFTER_POINT || floatflag == AFTER_EXPON)
1495 error ("malformed floating constant");
1496 floatflag = TOO_MANY_POINTS;
1497 /* Avoid another error from atof by forcing all characters
1498 from here on to be ignored. */
1499 p[-1] = '\0';
1501 else
1502 floatflag = AFTER_POINT;
1504 if (base == 8)
1505 base = 10;
1506 *p++ = c = GETC();
1507 /* Accept '.' as the start of a floating-point number
1508 only when it is followed by a digit.
1509 Otherwise, unread the following non-digit
1510 and use the '.' as a structural token. */
1511 if (p == token_buffer + 2 && !ISDIGIT (c))
1513 if (c == '.')
1515 c = GETC();
1516 if (c == '.')
1518 *p++ = c;
1519 *p = 0;
1520 return ELLIPSIS;
1522 error ("parse error at `..'");
1524 UNGETC (c);
1525 token_buffer[1] = 0;
1526 value = '.';
1527 goto done;
1530 else
1532 /* It is not a decimal point.
1533 It should be a digit (perhaps a hex digit). */
1535 if (ISDIGIT (c))
1537 c = c - '0';
1539 else if (base <= 10)
1541 if (c == 'e' || c == 'E')
1543 base = 10;
1544 floatflag = AFTER_EXPON;
1545 break; /* start of exponent */
1547 error ("nondigits in number and not hexadecimal");
1548 c = 0;
1550 else if (base == 16 && (c == 'p' || c == 'P'))
1552 floatflag = AFTER_EXPON;
1553 break; /* start of exponent */
1555 else if (c >= 'a')
1557 c = c - 'a' + 10;
1559 else
1561 c = c - 'A' + 10;
1563 if (c >= largest_digit)
1564 largest_digit = c;
1565 numdigits++;
1567 for (count = 0; count < TOTAL_PARTS; count++)
1569 parts[count] *= base;
1570 if (count)
1572 parts[count]
1573 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1574 parts[count-1]
1575 &= (1 << HOST_BITS_PER_CHAR) - 1;
1577 else
1578 parts[0] += c;
1581 /* If the extra highest-order part ever gets anything in it,
1582 the number is certainly too big. */
1583 if (parts[TOTAL_PARTS - 1] != 0)
1584 overflow = 1;
1586 if (p >= token_buffer + maxtoken - 3)
1587 p = extend_token_buffer (p);
1588 *p++ = (c = GETC());
1592 if (numdigits == 0)
1593 error ("numeric constant with no digits");
1595 if (largest_digit >= base)
1596 error ("numeric constant contains digits beyond the radix");
1598 /* Remove terminating char from the token buffer and delimit the string */
1599 *--p = 0;
1601 if (floatflag != NOT_FLOAT)
1603 tree type = double_type_node;
1604 int imag = 0;
1605 int conversion_errno = 0;
1606 REAL_VALUE_TYPE value;
1607 struct pf_args args;
1609 /* Read explicit exponent if any, and put it in tokenbuf. */
1611 if ((base == 10 && ((c == 'e') || (c == 'E')))
1612 || (base == 16 && (c == 'p' || c == 'P')))
1614 if (p >= token_buffer + maxtoken - 3)
1615 p = extend_token_buffer (p);
1616 *p++ = c;
1617 c = GETC();
1618 if ((c == '+') || (c == '-'))
1620 *p++ = c;
1621 c = GETC();
1623 /* Exponent is decimal, even if string is a hex float. */
1624 if (! ISDIGIT (c))
1625 error ("floating constant exponent has no digits");
1626 while (ISDIGIT (c))
1628 if (p >= token_buffer + maxtoken - 3)
1629 p = extend_token_buffer (p);
1630 *p++ = c;
1631 c = GETC();
1634 if (base == 16 && floatflag != AFTER_EXPON)
1635 error ("hexadecimal floating constant has no exponent");
1637 *p = 0;
1639 /* Setup input for parse_float() */
1640 args.base = base;
1641 args.p = p;
1642 args.c = c;
1643 args.imag = imag;
1644 args.type = type;
1645 args.conversion_errno = conversion_errno;
1647 /* Convert string to a double, checking for overflow. */
1648 if (do_float_handler (parse_float, (PTR) &args))
1650 /* Receive output from parse_float() */
1651 value = args.value;
1653 else
1655 /* We got an exception from parse_float() */
1656 error ("floating constant out of range");
1657 value = dconst0;
1660 /* Receive output from parse_float() */
1661 c = args.c;
1662 imag = args.imag;
1663 type = args.type;
1664 conversion_errno = args.conversion_errno;
1666 #ifdef ERANGE
1667 /* ERANGE is also reported for underflow,
1668 so test the value to distinguish overflow from that. */
1669 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1670 && (REAL_VALUES_LESS (dconst1, value)
1671 || REAL_VALUES_LESS (value, dconstm1)))
1672 warning ("floating point number exceeds range of `double'");
1673 #endif
1675 /* If the result is not a number, assume it must have been
1676 due to some error message above, so silently convert
1677 it to a zero. */
1678 if (REAL_VALUE_ISNAN (value))
1679 value = dconst0;
1681 /* Create a node with determined type and value. */
1682 if (imag)
1683 yylval.ttype = build_complex (NULL_TREE,
1684 convert (type, integer_zero_node),
1685 build_real (type, value));
1686 else
1687 yylval.ttype = build_real (type, value);
1689 else
1691 tree traditional_type, ansi_type, type;
1692 HOST_WIDE_INT high, low;
1693 int spec_unsigned = 0;
1694 int spec_long = 0;
1695 int spec_long_long = 0;
1696 int spec_imag = 0;
1697 int warn, i;
1699 traditional_type = ansi_type = type = NULL_TREE;
1700 while (1)
1702 if (c == 'u' || c == 'U')
1704 if (spec_unsigned)
1705 error ("two `u's in integer constant");
1706 spec_unsigned = 1;
1708 else if (c == 'l' || c == 'L')
1710 if (spec_long)
1712 if (spec_long_long)
1713 error ("three `l's in integer constant");
1714 else if (pedantic && ! in_system_header && warn_long_long)
1715 pedwarn ("ANSI C forbids long long integer constants");
1716 spec_long_long = 1;
1718 spec_long = 1;
1720 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1722 if (spec_imag)
1723 error ("more than one `i' or `j' in numeric constant");
1724 else if (pedantic)
1725 pedwarn ("ANSI C forbids imaginary numeric constants");
1726 spec_imag = 1;
1728 else
1729 break;
1730 if (p >= token_buffer + maxtoken - 3)
1731 p = extend_token_buffer (p);
1732 *p++ = c;
1733 c = GETC();
1736 /* If it won't fit in the host's representation for integers,
1737 then pedwarn. */
1739 warn = overflow;
1740 if (warn)
1741 pedwarn ("integer constant out of range");
1743 /* This is simplified by the fact that our constant
1744 is always positive. */
1746 high = low = 0;
1748 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1750 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1751 / HOST_BITS_PER_CHAR)]
1752 << (i * HOST_BITS_PER_CHAR));
1753 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1756 yylval.ttype = build_int_2 (low, high);
1757 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1759 /* If warn_traditional, calculate both the ANSI type and the
1760 traditional type, then see if they disagree.
1761 Otherwise, calculate only the type for the dialect in use. */
1762 if (warn_traditional || flag_traditional)
1764 /* Calculate the traditional type. */
1765 /* Traditionally, any constant is signed;
1766 but if unsigned is specified explicitly, obey that.
1767 Use the smallest size with the right number of bits,
1768 except for one special case with decimal constants. */
1769 if (! spec_long && base != 10
1770 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1771 traditional_type = (spec_unsigned ? unsigned_type_node
1772 : integer_type_node);
1773 /* A decimal constant must be long
1774 if it does not fit in type int.
1775 I think this is independent of whether
1776 the constant is signed. */
1777 else if (! spec_long && base == 10
1778 && int_fits_type_p (yylval.ttype, integer_type_node))
1779 traditional_type = (spec_unsigned ? unsigned_type_node
1780 : integer_type_node);
1781 else if (! spec_long_long)
1782 traditional_type = (spec_unsigned ? long_unsigned_type_node
1783 : long_integer_type_node);
1784 else
1785 traditional_type = (spec_unsigned
1786 ? long_long_unsigned_type_node
1787 : long_long_integer_type_node);
1789 if (warn_traditional || ! flag_traditional)
1791 /* Calculate the ANSI type. */
1792 if (! spec_long && ! spec_unsigned
1793 && int_fits_type_p (yylval.ttype, integer_type_node))
1794 ansi_type = integer_type_node;
1795 else if (! spec_long && (base != 10 || spec_unsigned)
1796 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1797 ansi_type = unsigned_type_node;
1798 else if (! spec_unsigned && !spec_long_long
1799 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1800 ansi_type = long_integer_type_node;
1801 else if (! spec_long_long
1802 && int_fits_type_p (yylval.ttype,
1803 long_unsigned_type_node))
1804 ansi_type = long_unsigned_type_node;
1805 else if (! spec_unsigned
1806 && int_fits_type_p (yylval.ttype,
1807 long_long_integer_type_node))
1808 ansi_type = long_long_integer_type_node;
1809 else
1810 ansi_type = long_long_unsigned_type_node;
1813 type = flag_traditional ? traditional_type : ansi_type;
1815 if (warn_traditional && traditional_type != ansi_type)
1817 if (TYPE_PRECISION (traditional_type)
1818 != TYPE_PRECISION (ansi_type))
1819 warning ("width of integer constant changes with -traditional");
1820 else if (TREE_UNSIGNED (traditional_type)
1821 != TREE_UNSIGNED (ansi_type))
1822 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1823 else
1824 warning ("width of integer constant may change on other systems with -traditional");
1827 if (pedantic && !flag_traditional && !spec_long_long && !warn
1828 && (TYPE_PRECISION (long_integer_type_node)
1829 < TYPE_PRECISION (type)))
1831 warn = 1;
1832 pedwarn ("integer constant out of range");
1835 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1836 warning ("decimal constant is so large that it is unsigned");
1838 if (spec_imag)
1840 if (TYPE_PRECISION (type)
1841 <= TYPE_PRECISION (integer_type_node))
1842 yylval.ttype
1843 = build_complex (NULL_TREE, integer_zero_node,
1844 convert (integer_type_node,
1845 yylval.ttype));
1846 else
1847 error ("complex integer constant is too wide for `complex int'");
1849 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1850 /* The traditional constant 0x80000000 is signed
1851 but doesn't fit in the range of int.
1852 This will change it to -0x80000000, which does fit. */
1854 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1855 yylval.ttype = convert (type, yylval.ttype);
1856 TREE_OVERFLOW (yylval.ttype)
1857 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1859 else
1860 TREE_TYPE (yylval.ttype) = type;
1863 /* If it's still an integer (not a complex), and it doesn't
1864 fit in the type we choose for it, then pedwarn. */
1866 if (! warn
1867 && TREE_CODE (TREE_TYPE (yylval.ttype)) == INTEGER_TYPE
1868 && ! int_fits_type_p (yylval.ttype, TREE_TYPE (yylval.ttype)))
1869 pedwarn ("integer constant out of range");
1872 UNGETC (c);
1873 *p = 0;
1875 if (ISALNUM (c) || c == '.' || c == '_' || c == '$'
1876 || (!flag_traditional && (c == '-' || c == '+')
1877 && (p[-1] == 'e' || p[-1] == 'E')))
1878 error ("missing white space after number `%s'", token_buffer);
1880 value = CONSTANT; break;
1883 case '\'':
1884 char_constant:
1886 register int result = 0;
1887 register int num_chars = 0;
1888 int chars_seen = 0;
1889 unsigned width = TYPE_PRECISION (char_type_node);
1890 int max_chars;
1891 #ifdef MULTIBYTE_CHARS
1892 int longest_char = local_mb_cur_max ();
1893 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1894 #endif
1896 max_chars = TYPE_PRECISION (integer_type_node) / width;
1897 if (wide_flag)
1898 width = WCHAR_TYPE_SIZE;
1900 while (1)
1902 tryagain:
1903 c = GETC();
1905 if (c == '\'' || c == EOF)
1906 break;
1908 ++chars_seen;
1909 if (c == '\\')
1911 int ignore = 0;
1912 c = readescape (&ignore);
1913 if (ignore)
1914 goto tryagain;
1915 if (width < HOST_BITS_PER_INT
1916 && (unsigned) c >= ((unsigned)1 << width))
1917 pedwarn ("escape sequence out of range for character");
1918 #ifdef MAP_CHARACTER
1919 if (ISPRINT (c))
1920 c = MAP_CHARACTER (c);
1921 #endif
1923 else if (c == '\n')
1925 if (pedantic)
1926 pedwarn ("ANSI C forbids newline in character constant");
1927 lineno++;
1929 else
1931 #ifdef MULTIBYTE_CHARS
1932 wchar_t wc;
1933 int i;
1934 int char_len = -1;
1935 for (i = 1; i <= longest_char; ++i)
1937 if (i > maxtoken - 4)
1938 extend_token_buffer (token_buffer);
1940 token_buffer[i] = c;
1941 char_len = local_mbtowc (& wc,
1942 token_buffer + 1,
1944 if (char_len != -1)
1945 break;
1946 c = GETC ();
1948 if (char_len > 1)
1950 /* mbtowc sometimes needs an extra char before accepting */
1951 if (char_len < i)
1952 UNGETC (c);
1953 if (! wide_flag)
1955 /* Merge character into result; ignore excess chars. */
1956 for (i = 1; i <= char_len; ++i)
1958 if (i > max_chars)
1959 break;
1960 if (width < HOST_BITS_PER_INT)
1961 result = (result << width)
1962 | (token_buffer[i]
1963 & ((1 << width) - 1));
1964 else
1965 result = token_buffer[i];
1967 num_chars += char_len;
1968 goto tryagain;
1970 c = wc;
1972 else
1974 if (char_len == -1)
1975 warning ("Ignoring invalid multibyte character");
1976 if (wide_flag)
1977 c = wc;
1978 #ifdef MAP_CHARACTER
1979 else
1980 c = MAP_CHARACTER (c);
1981 #endif
1983 #else /* ! MULTIBYTE_CHARS */
1984 #ifdef MAP_CHARACTER
1985 c = MAP_CHARACTER (c);
1986 #endif
1987 #endif /* ! MULTIBYTE_CHARS */
1990 if (wide_flag)
1992 if (chars_seen == 1) /* only keep the first one */
1993 result = c;
1994 goto tryagain;
1997 /* Merge character into result; ignore excess chars. */
1998 num_chars += (width / TYPE_PRECISION (char_type_node));
1999 if (num_chars < max_chars + 1)
2001 if (width < HOST_BITS_PER_INT)
2002 result = (result << width) | (c & ((1 << width) - 1));
2003 else
2004 result = c;
2008 if (c != '\'')
2009 error ("malformatted character constant");
2010 else if (chars_seen == 0)
2011 error ("empty character constant");
2012 else if (num_chars > max_chars)
2014 num_chars = max_chars;
2015 error ("character constant too long");
2017 else if (chars_seen != 1 && ! flag_traditional && warn_multichar)
2018 warning ("multi-character character constant");
2020 /* If char type is signed, sign-extend the constant. */
2021 if (! wide_flag)
2023 int num_bits = num_chars * width;
2024 if (num_bits == 0)
2025 /* We already got an error; avoid invalid shift. */
2026 yylval.ttype = build_int_2 (0, 0);
2027 else if (TREE_UNSIGNED (char_type_node)
2028 || ((result >> (num_bits - 1)) & 1) == 0)
2029 yylval.ttype
2030 = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
2031 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2033 else
2034 yylval.ttype
2035 = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
2036 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2037 -1);
2038 TREE_TYPE (yylval.ttype) = integer_type_node;
2040 else
2042 yylval.ttype = build_int_2 (result, 0);
2043 TREE_TYPE (yylval.ttype) = wchar_type_node;
2046 value = CONSTANT;
2047 break;
2050 case '"':
2051 string_constant:
2053 unsigned width = wide_flag ? WCHAR_TYPE_SIZE
2054 : TYPE_PRECISION (char_type_node);
2055 #ifdef MULTIBYTE_CHARS
2056 int longest_char = local_mb_cur_max ();
2057 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
2058 #endif
2059 c = GETC ();
2060 p = token_buffer + 1;
2062 while (c != '"' && c >= 0)
2064 if (c == '\\')
2066 int ignore = 0;
2067 c = readescape (&ignore);
2068 if (ignore)
2069 goto skipnewline;
2070 if (width < HOST_BITS_PER_INT
2071 && (unsigned) c >= ((unsigned)1 << width))
2072 pedwarn ("escape sequence out of range for character");
2074 else if (c == '\n')
2076 if (pedantic)
2077 pedwarn ("ANSI C forbids newline in string constant");
2078 lineno++;
2080 else
2082 #ifdef MULTIBYTE_CHARS
2083 wchar_t wc;
2084 int i;
2085 int char_len = -1;
2086 for (i = 0; i < longest_char; ++i)
2088 if (p + i >= token_buffer + maxtoken)
2089 p = extend_token_buffer (p);
2090 p[i] = c;
2092 char_len = local_mbtowc (& wc, p, i + 1);
2093 if (char_len != -1)
2094 break;
2095 c = GETC ();
2097 if (char_len == -1)
2098 warning ("Ignoring invalid multibyte character");
2099 else
2101 /* mbtowc sometimes needs an extra char before accepting */
2102 if (char_len <= i)
2103 UNGETC (c);
2104 if (! wide_flag)
2106 p += (i + 1);
2107 c = GETC ();
2108 continue;
2110 c = wc;
2112 #endif /* MULTIBYTE_CHARS */
2115 /* Add this single character into the buffer either as a wchar_t
2116 or as a single byte. */
2117 if (wide_flag)
2119 unsigned width = TYPE_PRECISION (char_type_node);
2120 unsigned bytemask = (1 << width) - 1;
2121 int byte;
2123 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2124 p = extend_token_buffer (p);
2126 for (byte = 0; byte < WCHAR_BYTES; ++byte)
2128 int value;
2129 if (byte >= (int) sizeof (c))
2130 value = 0;
2131 else
2132 value = (c >> (byte * width)) & bytemask;
2133 if (BYTES_BIG_ENDIAN)
2134 p[WCHAR_BYTES - byte - 1] = value;
2135 else
2136 p[byte] = value;
2138 p += WCHAR_BYTES;
2140 else
2142 if (p >= token_buffer + maxtoken)
2143 p = extend_token_buffer (p);
2144 *p++ = c;
2147 skipnewline:
2148 c = GETC ();
2151 /* Terminate the string value, either with a single byte zero
2152 or with a wide zero. */
2153 if (wide_flag)
2155 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2156 p = extend_token_buffer (p);
2157 bzero (p, WCHAR_BYTES);
2158 p += WCHAR_BYTES;
2160 else
2162 if (p >= token_buffer + maxtoken)
2163 p = extend_token_buffer (p);
2164 *p++ = 0;
2167 if (c < 0)
2168 error ("Unterminated string constant");
2170 /* We have read the entire constant.
2171 Construct a STRING_CST for the result. */
2173 if (wide_flag)
2175 yylval.ttype = build_string (p - (token_buffer + 1),
2176 token_buffer + 1);
2177 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2178 value = STRING;
2180 else if (objc_flag)
2182 /* Return an Objective-C @"..." constant string object. */
2183 yylval.ttype = build_objc_string (p - (token_buffer + 1),
2184 token_buffer + 1);
2185 TREE_TYPE (yylval.ttype) = char_array_type_node;
2186 value = OBJC_STRING;
2188 else
2190 yylval.ttype = build_string (p - (token_buffer + 1),
2191 token_buffer + 1);
2192 TREE_TYPE (yylval.ttype) = char_array_type_node;
2193 value = STRING;
2196 break;
2199 case '+':
2200 case '-':
2201 case '&':
2202 case '|':
2203 case ':':
2204 case '<':
2205 case '>':
2206 case '*':
2207 case '/':
2208 case '%':
2209 case '^':
2210 case '!':
2211 case '=':
2213 register int c1;
2215 combine:
2217 switch (c)
2219 case '+':
2220 yylval.code = PLUS_EXPR; break;
2221 case '-':
2222 yylval.code = MINUS_EXPR; break;
2223 case '&':
2224 yylval.code = BIT_AND_EXPR; break;
2225 case '|':
2226 yylval.code = BIT_IOR_EXPR; break;
2227 case '*':
2228 yylval.code = MULT_EXPR; break;
2229 case '/':
2230 yylval.code = TRUNC_DIV_EXPR; break;
2231 case '%':
2232 yylval.code = TRUNC_MOD_EXPR; break;
2233 case '^':
2234 yylval.code = BIT_XOR_EXPR; break;
2235 case LSHIFT:
2236 yylval.code = LSHIFT_EXPR; break;
2237 case RSHIFT:
2238 yylval.code = RSHIFT_EXPR; break;
2239 case '<':
2240 yylval.code = LT_EXPR; break;
2241 case '>':
2242 yylval.code = GT_EXPR; break;
2245 token_buffer[1] = c1 = GETC();
2246 token_buffer[2] = 0;
2248 if (c1 == '=')
2250 switch (c)
2252 case '<':
2253 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2254 case '>':
2255 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2256 case '!':
2257 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2258 case '=':
2259 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2261 value = ASSIGN; goto done;
2263 else if (c == c1)
2264 switch (c)
2266 case '+':
2267 value = PLUSPLUS; goto done;
2268 case '-':
2269 value = MINUSMINUS; goto done;
2270 case '&':
2271 value = ANDAND; goto done;
2272 case '|':
2273 value = OROR; goto done;
2274 case '<':
2275 c = LSHIFT;
2276 goto combine;
2277 case '>':
2278 c = RSHIFT;
2279 goto combine;
2281 else
2282 switch (c)
2284 case '-':
2285 if (c1 == '>')
2286 { value = POINTSAT; goto done; }
2287 break;
2288 case ':':
2289 if (c1 == '>')
2290 { value = ']'; goto done; }
2291 break;
2292 case '<':
2293 if (c1 == '%')
2294 { value = '{'; indent_level++; goto done; }
2295 if (c1 == ':')
2296 { value = '['; goto done; }
2297 break;
2298 case '%':
2299 if (c1 == '>')
2300 { value = '}'; indent_level--; goto done; }
2301 break;
2303 UNGETC (c1);
2304 token_buffer[1] = 0;
2306 if ((c == '<') || (c == '>'))
2307 value = ARITHCOMPARE;
2308 else value = c;
2309 goto done;
2312 case 0:
2313 /* Don't make yyparse think this is eof. */
2314 value = 1;
2315 break;
2317 case '{':
2318 indent_level++;
2319 value = c;
2320 break;
2322 case '}':
2323 indent_level--;
2324 value = c;
2325 break;
2327 default:
2328 value = c;
2331 done:
2332 /* yylloc.last_line = lineno; */
2334 return value;
2337 /* Sets the value of the 'yydebug' variable to VALUE.
2338 This is a function so we don't have to have YYDEBUG defined
2339 in order to build the compiler. */
2341 void
2342 set_yydebug (value)
2343 int value;
2345 #if YYDEBUG != 0
2346 yydebug = value;
2347 #else
2348 warning ("YYDEBUG not defined.");
2349 #endif