* arm.c (arm_split_constant): Don't try to force a constant to
[official-gcc.git] / gcc / c-lex.c
blob8f63b2db7dc3e31b62ef59fb03ab4439a62f2045
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"
23 #include <setjmp.h>
25 #include "rtl.h"
26 #include "tree.h"
27 #include "input.h"
28 #include "output.h"
29 #include "c-lex.h"
30 #include "c-tree.h"
31 #include "flags.h"
32 #include "c-parse.h"
33 #include "c-pragma.h"
34 #include "toplev.h"
35 #include "intl.h"
37 /* MULTIBYTE_CHARS support only works for native compilers.
38 ??? Ideally what we want is to model widechar support after
39 the current floating point support. */
40 #ifdef CROSS_COMPILE
41 #undef MULTIBYTE_CHARS
42 #endif
44 #ifdef MULTIBYTE_CHARS
45 #include "mbchar.h"
46 #include <locale.h>
47 #endif /* MULTIBYTE_CHARS */
49 #if USE_CPPLIB
50 #include "cpplib.h"
51 extern cpp_reader parse_in;
52 extern cpp_options parse_options;
53 #else
54 /* Stream for reading from the input file. */
55 FILE *finput;
56 #endif
58 extern void yyprint PROTO((FILE *, int, YYSTYPE));
60 /* The elements of `ridpointers' are identifier nodes
61 for the reserved type names and storage classes.
62 It is indexed by a RID_... value. */
63 tree ridpointers[(int) RID_MAX];
65 /* Cause the `yydebug' variable to be defined. */
66 #define YYDEBUG 1
68 #if USE_CPPLIB
69 extern unsigned char *yy_cur, *yy_lim;
71 extern int yy_get_token ();
73 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
74 #define UNGETC(c) ((c) == EOF ? 0 : yy_cur--)
75 #else
76 #define GETC() getc (finput)
77 #define UNGETC(c) ungetc (c, finput)
78 #endif
80 /* the declaration found for the last IDENTIFIER token read in.
81 yylex must look this up to detect typedefs, which get token type TYPENAME,
82 so it is left around in case the identifier is not a typedef but is
83 used in a context which makes it a reference to a variable. */
84 tree lastiddecl;
86 /* Nonzero enables objc features. */
88 int doing_objc_thang;
90 extern int yydebug;
92 /* File used for outputting assembler code. */
93 extern FILE *asm_out_file;
95 #ifndef WCHAR_TYPE_SIZE
96 #ifdef INT_TYPE_SIZE
97 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
98 #else
99 #define WCHAR_TYPE_SIZE BITS_PER_WORD
100 #endif
101 #endif
103 /* Number of bytes in a wide character. */
104 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
106 static int maxtoken; /* Current nominal length of token buffer. */
107 char *token_buffer; /* Pointer to token buffer.
108 Actual allocated length is maxtoken + 2.
109 This is not static because objc-parse.y uses it. */
111 static int indent_level = 0; /* Number of { minus number of }. */
113 /* Nonzero if end-of-file has been seen on input. */
114 static int end_of_file;
116 #if !USE_CPPLIB
117 /* Buffered-back input character; faster than using ungetc. */
118 static int nextchar = -1;
119 #endif
121 #ifdef HANDLE_GENERIC_PRAGMAS
122 static int handle_generic_pragma PROTO((int));
123 #endif /* HANDLE_GENERIC_PRAGMAS */
124 static int whitespace_cr PROTO((int));
125 static int skip_white_space PROTO((int));
126 static int skip_white_space_on_line PROTO((void));
127 static char *extend_token_buffer PROTO((const char *));
128 static int readescape PROTO((int *));
130 /* Do not insert generated code into the source, instead, include it.
131 This allows us to build gcc automatically even for targets that
132 need to add or modify the reserved keyword lists. */
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 */
1094 yylex ()
1096 register int c;
1097 register char *p;
1098 register int value;
1099 int wide_flag = 0;
1100 int objc_flag = 0;
1102 #if !USE_CPPLIB
1103 if (nextchar >= 0)
1104 c = nextchar, nextchar = -1;
1105 else
1106 #endif
1107 c = GETC();
1109 /* Effectively do c = skip_white_space (c)
1110 but do it faster in the usual cases. */
1111 while (1)
1112 switch (c)
1114 case ' ':
1115 case '\t':
1116 case '\f':
1117 case '\v':
1118 case '\b':
1119 c = GETC();
1120 break;
1122 case '\r':
1123 /* Call skip_white_space so we can warn if appropriate. */
1125 case '\n':
1126 case '/':
1127 case '\\':
1128 c = skip_white_space (c);
1129 default:
1130 goto found_nonwhite;
1132 found_nonwhite:
1134 token_buffer[0] = c;
1135 token_buffer[1] = 0;
1137 /* yylloc.first_line = lineno; */
1139 switch (c)
1141 case EOF:
1142 end_of_file = 1;
1143 token_buffer[0] = 0;
1144 value = ENDFILE;
1145 break;
1147 case 'L':
1148 /* Capital L may start a wide-string or wide-character constant. */
1150 register int c = GETC();
1151 if (c == '\'')
1153 wide_flag = 1;
1154 goto char_constant;
1156 if (c == '"')
1158 wide_flag = 1;
1159 goto string_constant;
1161 UNGETC (c);
1163 goto letter;
1165 case '@':
1166 if (!doing_objc_thang)
1168 value = c;
1169 break;
1171 else
1173 /* '@' may start a constant string object. */
1174 register int c = GETC ();
1175 if (c == '"')
1177 objc_flag = 1;
1178 goto string_constant;
1180 UNGETC (c);
1181 /* Fall through to treat '@' as the start of an identifier. */
1184 case 'A': case 'B': case 'C': case 'D': case 'E':
1185 case 'F': case 'G': case 'H': case 'I': case 'J':
1186 case 'K': case 'M': case 'N': case 'O':
1187 case 'P': case 'Q': case 'R': case 'S': case 'T':
1188 case 'U': case 'V': case 'W': case 'X': case 'Y':
1189 case 'Z':
1190 case 'a': case 'b': case 'c': case 'd': case 'e':
1191 case 'f': case 'g': case 'h': case 'i': case 'j':
1192 case 'k': case 'l': case 'm': case 'n': case 'o':
1193 case 'p': case 'q': case 'r': case 's': case 't':
1194 case 'u': case 'v': case 'w': case 'x': case 'y':
1195 case 'z':
1196 case '_':
1197 case '$':
1198 letter:
1199 p = token_buffer;
1200 while (ISALNUM (c) || c == '_' || c == '$' || c == '@')
1202 /* Make sure this char really belongs in an identifier. */
1203 if (c == '$')
1205 if (! dollars_in_ident)
1206 error ("`$' in identifier");
1207 else if (pedantic)
1208 pedwarn ("`$' in identifier");
1211 if (p >= token_buffer + maxtoken)
1212 p = extend_token_buffer (p);
1214 *p++ = c;
1215 c = GETC();
1218 *p = 0;
1219 #if USE_CPPLIB
1220 UNGETC (c);
1221 #else
1222 nextchar = c;
1223 #endif
1225 value = IDENTIFIER;
1226 yylval.itype = 0;
1228 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1231 register struct resword *ptr;
1233 if ((ptr = is_reserved_word (token_buffer, p - token_buffer)))
1235 if (ptr->rid)
1236 yylval.ttype = ridpointers[(int) ptr->rid];
1237 value = (int) ptr->token;
1239 /* Only return OBJECTNAME if it is a typedef. */
1240 if (doing_objc_thang && value == OBJECTNAME)
1242 lastiddecl = lookup_name(yylval.ttype);
1244 if (lastiddecl == NULL_TREE
1245 || TREE_CODE (lastiddecl) != TYPE_DECL)
1246 value = IDENTIFIER;
1249 /* Even if we decided to recognize asm, still perhaps warn. */
1250 if (pedantic
1251 && (value == ASM_KEYWORD || value == TYPEOF
1252 || ptr->rid == RID_INLINE)
1253 && token_buffer[0] != '_')
1254 pedwarn ("ANSI does not permit the keyword `%s'",
1255 token_buffer);
1259 /* If we did not find a keyword, look for an identifier
1260 (or a typename). */
1262 if (value == IDENTIFIER)
1264 if (token_buffer[0] == '@')
1265 error("invalid identifier `%s'", token_buffer);
1267 yylval.ttype = get_identifier (token_buffer);
1268 lastiddecl = lookup_name (yylval.ttype);
1270 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1271 value = TYPENAME;
1272 /* A user-invisible read-only initialized variable
1273 should be replaced by its value.
1274 We handle only strings since that's the only case used in C. */
1275 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1276 && DECL_IGNORED_P (lastiddecl)
1277 && TREE_READONLY (lastiddecl)
1278 && DECL_INITIAL (lastiddecl) != 0
1279 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1281 tree stringval = DECL_INITIAL (lastiddecl);
1283 /* Copy the string value so that we won't clobber anything
1284 if we put something in the TREE_CHAIN of this one. */
1285 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1286 TREE_STRING_POINTER (stringval));
1287 value = STRING;
1289 else if (doing_objc_thang)
1291 tree objc_interface_decl = is_class_name (yylval.ttype);
1293 if (objc_interface_decl)
1295 value = CLASSNAME;
1296 yylval.ttype = objc_interface_decl;
1301 break;
1303 case '0': case '1':
1305 int next_c;
1306 /* Check first for common special case: single-digit 0 or 1. */
1308 next_c = GETC ();
1309 UNGETC (next_c); /* Always undo this lookahead. */
1310 if (!ISALNUM (next_c) && next_c != '.')
1312 token_buffer[0] = (char)c, token_buffer[1] = '\0';
1313 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1314 value = CONSTANT;
1315 break;
1317 /*FALLTHRU*/
1319 case '2': case '3': case '4':
1320 case '5': case '6': case '7': case '8': case '9':
1321 case '.':
1323 int base = 10;
1324 int count = 0;
1325 int largest_digit = 0;
1326 int numdigits = 0;
1327 /* for multi-precision arithmetic,
1328 we actually store only HOST_BITS_PER_CHAR bits in each part.
1329 The number of parts is chosen so as to be sufficient to hold
1330 the enough bits to fit into the two HOST_WIDE_INTs that contain
1331 the integer value (this is always at least as many bits as are
1332 in a target `long long' value, but may be wider). */
1333 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1334 int parts[TOTAL_PARTS];
1335 int overflow = 0;
1337 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS, AFTER_EXPON}
1338 floatflag = NOT_FLOAT;
1340 for (count = 0; count < TOTAL_PARTS; count++)
1341 parts[count] = 0;
1343 p = token_buffer;
1344 *p++ = c;
1346 if (c == '0')
1348 *p++ = (c = GETC());
1349 if ((c == 'x') || (c == 'X'))
1351 base = 16;
1352 *p++ = (c = GETC());
1354 /* Leading 0 forces octal unless the 0 is the only digit. */
1355 else if (c >= '0' && c <= '9')
1357 base = 8;
1358 numdigits++;
1360 else
1361 numdigits++;
1364 /* Read all the digits-and-decimal-points. */
1366 while (c == '.'
1367 || (ISALNUM (c) && c != 'l' && c != 'L'
1368 && c != 'u' && c != 'U'
1369 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1370 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1372 if (c == '.')
1374 if (base == 16 && pedantic)
1375 error ("floating constant may not be in radix 16");
1376 if (floatflag == TOO_MANY_POINTS)
1377 /* We have already emitted an error. Don't need another. */
1379 else if (floatflag == AFTER_POINT || floatflag == AFTER_EXPON)
1381 error ("malformed floating constant");
1382 floatflag = TOO_MANY_POINTS;
1383 /* Avoid another error from atof by forcing all characters
1384 from here on to be ignored. */
1385 p[-1] = '\0';
1387 else
1388 floatflag = AFTER_POINT;
1390 if (base == 8)
1391 base = 10;
1392 *p++ = c = GETC();
1393 /* Accept '.' as the start of a floating-point number
1394 only when it is followed by a digit.
1395 Otherwise, unread the following non-digit
1396 and use the '.' as a structural token. */
1397 if (p == token_buffer + 2 && !ISDIGIT (c))
1399 if (c == '.')
1401 c = GETC();
1402 if (c == '.')
1404 *p++ = c;
1405 *p = 0;
1406 return ELLIPSIS;
1408 error ("parse error at `..'");
1410 UNGETC (c);
1411 token_buffer[1] = 0;
1412 value = '.';
1413 goto done;
1416 else
1418 /* It is not a decimal point.
1419 It should be a digit (perhaps a hex digit). */
1421 if (ISDIGIT (c))
1423 c = c - '0';
1425 else if (base <= 10)
1427 if (c == 'e' || c == 'E')
1429 base = 10;
1430 floatflag = AFTER_EXPON;
1431 break; /* start of exponent */
1433 error ("nondigits in number and not hexadecimal");
1434 c = 0;
1436 else if (base == 16 && (c == 'p' || c == 'P'))
1438 floatflag = AFTER_EXPON;
1439 break; /* start of exponent */
1441 else if (c >= 'a')
1443 c = c - 'a' + 10;
1445 else
1447 c = c - 'A' + 10;
1449 if (c >= largest_digit)
1450 largest_digit = c;
1451 numdigits++;
1453 for (count = 0; count < TOTAL_PARTS; count++)
1455 parts[count] *= base;
1456 if (count)
1458 parts[count]
1459 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1460 parts[count-1]
1461 &= (1 << HOST_BITS_PER_CHAR) - 1;
1463 else
1464 parts[0] += c;
1467 /* If the extra highest-order part ever gets anything in it,
1468 the number is certainly too big. */
1469 if (parts[TOTAL_PARTS - 1] != 0)
1470 overflow = 1;
1472 if (p >= token_buffer + maxtoken - 3)
1473 p = extend_token_buffer (p);
1474 *p++ = (c = GETC());
1478 if (numdigits == 0)
1479 error ("numeric constant with no digits");
1481 if (largest_digit >= base)
1482 error ("numeric constant contains digits beyond the radix");
1484 /* Remove terminating char from the token buffer and delimit the string */
1485 *--p = 0;
1487 if (floatflag != NOT_FLOAT)
1489 tree type = double_type_node;
1490 int imag = 0;
1491 int conversion_errno = 0;
1492 REAL_VALUE_TYPE value;
1493 jmp_buf handler;
1495 /* Read explicit exponent if any, and put it in tokenbuf. */
1497 if ((base == 10 && ((c == 'e') || (c == 'E')))
1498 || (base == 16 && (c == 'p' || c == 'P')))
1500 if (p >= token_buffer + maxtoken - 3)
1501 p = extend_token_buffer (p);
1502 *p++ = c;
1503 c = GETC();
1504 if ((c == '+') || (c == '-'))
1506 *p++ = c;
1507 c = GETC();
1509 /* Exponent is decimal, even if string is a hex float. */
1510 if (! ISDIGIT (c))
1511 error ("floating constant exponent has no digits");
1512 while (ISDIGIT (c))
1514 if (p >= token_buffer + maxtoken - 3)
1515 p = extend_token_buffer (p);
1516 *p++ = c;
1517 c = GETC();
1520 if (base == 16 && floatflag != AFTER_EXPON)
1521 error ("hexadecimal floating constant has no exponent");
1523 *p = 0;
1525 /* Convert string to a double, checking for overflow. */
1526 if (setjmp (handler))
1528 error ("floating constant out of range");
1529 value = dconst0;
1531 else
1533 int fflag = 0, lflag = 0;
1534 /* Copy token_buffer now, while it has just the number
1535 and not the suffixes; once we add `f' or `i',
1536 REAL_VALUE_ATOF may not work any more. */
1537 char *copy = (char *) alloca (p - token_buffer + 1);
1538 bcopy (token_buffer, copy, p - token_buffer + 1);
1540 set_float_handler (handler);
1542 while (1)
1544 int lose = 0;
1546 /* Read the suffixes to choose a data type. */
1547 switch (c)
1549 case 'f': case 'F':
1550 if (fflag)
1551 error ("more than one `f' in numeric constant");
1552 fflag = 1;
1553 break;
1555 case 'l': case 'L':
1556 if (lflag)
1557 error ("more than one `l' in numeric constant");
1558 lflag = 1;
1559 break;
1561 case 'i': case 'I':
1562 if (imag)
1563 error ("more than one `i' or `j' in numeric constant");
1564 else if (pedantic)
1565 pedwarn ("ANSI C forbids imaginary numeric constants");
1566 imag = 1;
1567 break;
1569 default:
1570 lose = 1;
1573 if (lose)
1574 break;
1576 if (p >= token_buffer + maxtoken - 3)
1577 p = extend_token_buffer (p);
1578 *p++ = c;
1579 *p = 0;
1580 c = GETC();
1583 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1584 tells the desired precision of the binary result
1585 of decimal-to-binary conversion. */
1587 if (fflag)
1589 if (lflag)
1590 error ("both `f' and `l' in floating constant");
1592 type = float_type_node;
1593 errno = 0;
1594 if (base == 16)
1595 value = REAL_VALUE_HTOF (copy, TYPE_MODE (type));
1596 else
1597 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1598 conversion_errno = errno;
1599 /* A diagnostic is required here by some ANSI C testsuites.
1600 This is not pedwarn, because some people don't want
1601 an error for this. */
1602 if (REAL_VALUE_ISINF (value) && pedantic)
1603 warning ("floating point number exceeds range of `float'");
1605 else if (lflag)
1607 type = long_double_type_node;
1608 errno = 0;
1609 if (base == 16)
1610 value = REAL_VALUE_HTOF (copy, TYPE_MODE (type));
1611 else
1612 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1613 conversion_errno = errno;
1614 if (REAL_VALUE_ISINF (value) && pedantic)
1615 warning ("floating point number exceeds range of `long double'");
1617 else
1619 errno = 0;
1620 if (base == 16)
1621 value = REAL_VALUE_HTOF (copy, TYPE_MODE (type));
1622 else
1623 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1624 conversion_errno = errno;
1625 if (REAL_VALUE_ISINF (value) && pedantic)
1626 warning ("floating point number exceeds range of `double'");
1629 set_float_handler (NULL_PTR);
1631 #ifdef ERANGE
1632 /* ERANGE is also reported for underflow,
1633 so test the value to distinguish overflow from that. */
1634 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1635 && (REAL_VALUES_LESS (dconst1, value)
1636 || REAL_VALUES_LESS (value, dconstm1)))
1637 warning ("floating point number exceeds range of `double'");
1638 #endif
1640 /* If the result is not a number, assume it must have been
1641 due to some error message above, so silently convert
1642 it to a zero. */
1643 if (REAL_VALUE_ISNAN (value))
1644 value = dconst0;
1646 /* Create a node with determined type and value. */
1647 if (imag)
1648 yylval.ttype = build_complex (NULL_TREE,
1649 convert (type, integer_zero_node),
1650 build_real (type, value));
1651 else
1652 yylval.ttype = build_real (type, value);
1654 else
1656 tree traditional_type, ansi_type, type;
1657 HOST_WIDE_INT high, low;
1658 int spec_unsigned = 0;
1659 int spec_long = 0;
1660 int spec_long_long = 0;
1661 int spec_imag = 0;
1662 int bytes, warn, i;
1664 traditional_type = ansi_type = type = NULL_TREE;
1665 while (1)
1667 if (c == 'u' || c == 'U')
1669 if (spec_unsigned)
1670 error ("two `u's in integer constant");
1671 spec_unsigned = 1;
1673 else if (c == 'l' || c == 'L')
1675 if (spec_long)
1677 if (spec_long_long)
1678 error ("three `l's in integer constant");
1679 else if (pedantic && ! in_system_header && warn_long_long)
1680 pedwarn ("ANSI C forbids long long integer constants");
1681 spec_long_long = 1;
1683 spec_long = 1;
1685 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1687 if (spec_imag)
1688 error ("more than one `i' or `j' in numeric constant");
1689 else if (pedantic)
1690 pedwarn ("ANSI C forbids imaginary numeric constants");
1691 spec_imag = 1;
1693 else
1694 break;
1695 if (p >= token_buffer + maxtoken - 3)
1696 p = extend_token_buffer (p);
1697 *p++ = c;
1698 c = GETC();
1701 /* If the constant won't fit in the targets widest int,
1702 or it won't fit in the host's representation for ints,
1703 then warn that the constant is out of range. */
1705 #if HOST_BITS_PER_WIDE_INT >= 64
1706 bytes = TYPE_PRECISION (intTI_type_node) / HOST_BITS_PER_CHAR;
1707 #else
1708 bytes = TYPE_PRECISION (intDI_type_node) / HOST_BITS_PER_CHAR;
1709 #endif
1711 warn = overflow;
1712 for (i = bytes; i < TOTAL_PARTS; i++)
1713 if (parts[i])
1714 warn = 1;
1715 if (warn)
1716 pedwarn ("integer constant out of range");
1718 /* This is simplified by the fact that our constant
1719 is always positive. */
1721 high = low = 0;
1723 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1725 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1726 / HOST_BITS_PER_CHAR)]
1727 << (i * HOST_BITS_PER_CHAR));
1728 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1731 yylval.ttype = build_int_2 (low, high);
1732 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1734 /* If warn_traditional, calculate both the ANSI type and the
1735 traditional type, then see if they disagree.
1736 Otherwise, calculate only the type for the dialect in use. */
1737 if (warn_traditional || flag_traditional)
1739 /* Calculate the traditional type. */
1740 /* Traditionally, any constant is signed;
1741 but if unsigned is specified explicitly, obey that.
1742 Use the smallest size with the right number of bits,
1743 except for one special case with decimal constants. */
1744 if (! spec_long && base != 10
1745 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1746 traditional_type = (spec_unsigned ? unsigned_type_node
1747 : integer_type_node);
1748 /* A decimal constant must be long
1749 if it does not fit in type int.
1750 I think this is independent of whether
1751 the constant is signed. */
1752 else if (! spec_long && base == 10
1753 && int_fits_type_p (yylval.ttype, integer_type_node))
1754 traditional_type = (spec_unsigned ? unsigned_type_node
1755 : integer_type_node);
1756 else if (! spec_long_long)
1757 traditional_type = (spec_unsigned ? long_unsigned_type_node
1758 : long_integer_type_node);
1759 else
1760 traditional_type = (spec_unsigned
1761 ? long_long_unsigned_type_node
1762 : long_long_integer_type_node);
1764 if (warn_traditional || ! flag_traditional)
1766 /* Calculate the ANSI type. */
1767 if (! spec_long && ! spec_unsigned
1768 && int_fits_type_p (yylval.ttype, integer_type_node))
1769 ansi_type = integer_type_node;
1770 else if (! spec_long && (base != 10 || spec_unsigned)
1771 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1772 ansi_type = unsigned_type_node;
1773 else if (! spec_unsigned && !spec_long_long
1774 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1775 ansi_type = long_integer_type_node;
1776 else if (! spec_long_long
1777 && int_fits_type_p (yylval.ttype,
1778 long_unsigned_type_node))
1779 ansi_type = long_unsigned_type_node;
1780 else if (! spec_unsigned
1781 && int_fits_type_p (yylval.ttype,
1782 long_long_integer_type_node))
1783 ansi_type = long_long_integer_type_node;
1784 else
1785 ansi_type = long_long_unsigned_type_node;
1788 type = flag_traditional ? traditional_type : ansi_type;
1790 if (warn_traditional && traditional_type != ansi_type)
1792 if (TYPE_PRECISION (traditional_type)
1793 != TYPE_PRECISION (ansi_type))
1794 warning ("width of integer constant changes with -traditional");
1795 else if (TREE_UNSIGNED (traditional_type)
1796 != TREE_UNSIGNED (ansi_type))
1797 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1798 else
1799 warning ("width of integer constant may change on other systems with -traditional");
1802 if (pedantic && !flag_traditional && !spec_long_long && !warn
1803 && (TYPE_PRECISION (long_integer_type_node)
1804 < TYPE_PRECISION (type)))
1805 pedwarn ("integer constant out of range");
1807 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1808 warning ("decimal constant is so large that it is unsigned");
1810 if (spec_imag)
1812 if (TYPE_PRECISION (type)
1813 <= TYPE_PRECISION (integer_type_node))
1814 yylval.ttype
1815 = build_complex (NULL_TREE, integer_zero_node,
1816 convert (integer_type_node,
1817 yylval.ttype));
1818 else
1819 error ("complex integer constant is too wide for `complex int'");
1821 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1822 /* The traditional constant 0x80000000 is signed
1823 but doesn't fit in the range of int.
1824 This will change it to -0x80000000, which does fit. */
1826 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1827 yylval.ttype = convert (type, yylval.ttype);
1828 TREE_OVERFLOW (yylval.ttype)
1829 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1831 else
1832 TREE_TYPE (yylval.ttype) = type;
1835 UNGETC (c);
1836 *p = 0;
1838 if (ISALNUM (c) || c == '.' || c == '_' || c == '$'
1839 || (!flag_traditional && (c == '-' || c == '+')
1840 && (p[-1] == 'e' || p[-1] == 'E')))
1841 error ("missing white space after number `%s'", token_buffer);
1843 value = CONSTANT; break;
1846 case '\'':
1847 char_constant:
1849 register int result = 0;
1850 register int num_chars = 0;
1851 int chars_seen = 0;
1852 unsigned width = TYPE_PRECISION (char_type_node);
1853 int max_chars;
1854 #ifdef MULTIBYTE_CHARS
1855 int longest_char = local_mb_cur_max ();
1856 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1857 #endif
1859 max_chars = TYPE_PRECISION (integer_type_node) / width;
1860 if (wide_flag)
1861 width = WCHAR_TYPE_SIZE;
1863 while (1)
1865 tryagain:
1866 c = GETC();
1868 if (c == '\'' || c == EOF)
1869 break;
1871 ++chars_seen;
1872 if (c == '\\')
1874 int ignore = 0;
1875 c = readescape (&ignore);
1876 if (ignore)
1877 goto tryagain;
1878 if (width < HOST_BITS_PER_INT
1879 && (unsigned) c >= ((unsigned)1 << width))
1880 pedwarn ("escape sequence out of range for character");
1881 #ifdef MAP_CHARACTER
1882 if (ISPRINT (c))
1883 c = MAP_CHARACTER (c);
1884 #endif
1886 else if (c == '\n')
1888 if (pedantic)
1889 pedwarn ("ANSI C forbids newline in character constant");
1890 lineno++;
1892 else
1894 #ifdef MULTIBYTE_CHARS
1895 wchar_t wc;
1896 int i;
1897 int char_len = -1;
1898 for (i = 1; i <= longest_char; ++i)
1900 if (i > maxtoken - 4)
1901 extend_token_buffer (token_buffer);
1903 token_buffer[i] = c;
1904 char_len = local_mbtowc (& wc,
1905 token_buffer + 1,
1907 if (char_len != -1)
1908 break;
1909 c = GETC ();
1911 if (char_len > 1)
1913 /* mbtowc sometimes needs an extra char before accepting */
1914 if (char_len < i)
1915 UNGETC (c);
1916 if (! wide_flag)
1918 /* Merge character into result; ignore excess chars. */
1919 for (i = 1; i <= char_len; ++i)
1921 if (i > max_chars)
1922 break;
1923 if (width < HOST_BITS_PER_INT)
1924 result = (result << width)
1925 | (token_buffer[i]
1926 & ((1 << width) - 1));
1927 else
1928 result = token_buffer[i];
1930 num_chars += char_len;
1931 goto tryagain;
1933 c = wc;
1935 else
1937 if (char_len == -1)
1938 warning ("Ignoring invalid multibyte character");
1939 if (wide_flag)
1940 c = wc;
1941 #ifdef MAP_CHARACTER
1942 else
1943 c = MAP_CHARACTER (c);
1944 #endif
1946 #else /* ! MULTIBYTE_CHARS */
1947 #ifdef MAP_CHARACTER
1948 c = MAP_CHARACTER (c);
1949 #endif
1950 #endif /* ! MULTIBYTE_CHARS */
1953 if (wide_flag)
1955 if (chars_seen == 1) /* only keep the first one */
1956 result = c;
1957 goto tryagain;
1960 /* Merge character into result; ignore excess chars. */
1961 num_chars += (width / TYPE_PRECISION (char_type_node));
1962 if (num_chars < max_chars + 1)
1964 if (width < HOST_BITS_PER_INT)
1965 result = (result << width) | (c & ((1 << width) - 1));
1966 else
1967 result = c;
1971 if (c != '\'')
1972 error ("malformatted character constant");
1973 else if (chars_seen == 0)
1974 error ("empty character constant");
1975 else if (num_chars > max_chars)
1977 num_chars = max_chars;
1978 error ("character constant too long");
1980 else if (chars_seen != 1 && ! flag_traditional && warn_multichar)
1981 warning ("multi-character character constant");
1983 /* If char type is signed, sign-extend the constant. */
1984 if (! wide_flag)
1986 int num_bits = num_chars * width;
1987 if (num_bits == 0)
1988 /* We already got an error; avoid invalid shift. */
1989 yylval.ttype = build_int_2 (0, 0);
1990 else if (TREE_UNSIGNED (char_type_node)
1991 || ((result >> (num_bits - 1)) & 1) == 0)
1992 yylval.ttype
1993 = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
1994 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1996 else
1997 yylval.ttype
1998 = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
1999 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2000 -1);
2001 TREE_TYPE (yylval.ttype) = integer_type_node;
2003 else
2005 yylval.ttype = build_int_2 (result, 0);
2006 TREE_TYPE (yylval.ttype) = wchar_type_node;
2009 value = CONSTANT;
2010 break;
2013 case '"':
2014 string_constant:
2016 unsigned width = wide_flag ? WCHAR_TYPE_SIZE
2017 : TYPE_PRECISION (char_type_node);
2018 #ifdef MULTIBYTE_CHARS
2019 int longest_char = local_mb_cur_max ();
2020 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
2021 #endif
2022 c = GETC ();
2023 p = token_buffer + 1;
2025 while (c != '"' && c >= 0)
2027 if (c == '\\')
2029 int ignore = 0;
2030 c = readescape (&ignore);
2031 if (ignore)
2032 goto skipnewline;
2033 if (width < HOST_BITS_PER_INT
2034 && (unsigned) c >= ((unsigned)1 << width))
2035 pedwarn ("escape sequence out of range for character");
2037 else if (c == '\n')
2039 if (pedantic)
2040 pedwarn ("ANSI C forbids newline in string constant");
2041 lineno++;
2043 else
2045 #ifdef MULTIBYTE_CHARS
2046 wchar_t wc;
2047 int i;
2048 int char_len = -1;
2049 for (i = 0; i < longest_char; ++i)
2051 if (p + i >= token_buffer + maxtoken)
2052 p = extend_token_buffer (p);
2053 p[i] = c;
2055 char_len = local_mbtowc (& wc, p, i + 1);
2056 if (char_len != -1)
2057 break;
2058 c = GETC ();
2060 if (char_len == -1)
2061 warning ("Ignoring invalid multibyte character");
2062 else
2064 /* mbtowc sometimes needs an extra char before accepting */
2065 if (char_len <= i)
2066 UNGETC (c);
2067 if (! wide_flag)
2069 p += (i + 1);
2070 c = GETC ();
2071 continue;
2073 c = wc;
2075 #endif /* MULTIBYTE_CHARS */
2078 /* Add this single character into the buffer either as a wchar_t
2079 or as a single byte. */
2080 if (wide_flag)
2082 unsigned width = TYPE_PRECISION (char_type_node);
2083 unsigned bytemask = (1 << width) - 1;
2084 int byte;
2086 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2087 p = extend_token_buffer (p);
2089 for (byte = 0; byte < WCHAR_BYTES; ++byte)
2091 int value;
2092 if (byte >= (int) sizeof (c))
2093 value = 0;
2094 else
2095 value = (c >> (byte * width)) & bytemask;
2096 if (BYTES_BIG_ENDIAN)
2097 p[WCHAR_BYTES - byte - 1] = value;
2098 else
2099 p[byte] = value;
2101 p += WCHAR_BYTES;
2103 else
2105 if (p >= token_buffer + maxtoken)
2106 p = extend_token_buffer (p);
2107 *p++ = c;
2110 skipnewline:
2111 c = GETC ();
2114 /* Terminate the string value, either with a single byte zero
2115 or with a wide zero. */
2116 if (wide_flag)
2118 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2119 p = extend_token_buffer (p);
2120 bzero (p, WCHAR_BYTES);
2121 p += WCHAR_BYTES;
2123 else
2125 if (p >= token_buffer + maxtoken)
2126 p = extend_token_buffer (p);
2127 *p++ = 0;
2130 if (c < 0)
2131 error ("Unterminated string constant");
2133 /* We have read the entire constant.
2134 Construct a STRING_CST for the result. */
2136 if (wide_flag)
2138 yylval.ttype = build_string (p - (token_buffer + 1),
2139 token_buffer + 1);
2140 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2141 value = STRING;
2143 else if (objc_flag)
2145 /* Return an Objective-C @"..." constant string object. */
2146 yylval.ttype = build_objc_string (p - (token_buffer + 1),
2147 token_buffer + 1);
2148 TREE_TYPE (yylval.ttype) = char_array_type_node;
2149 value = OBJC_STRING;
2151 else
2153 yylval.ttype = build_string (p - (token_buffer + 1),
2154 token_buffer + 1);
2155 TREE_TYPE (yylval.ttype) = char_array_type_node;
2156 value = STRING;
2159 break;
2162 case '+':
2163 case '-':
2164 case '&':
2165 case '|':
2166 case ':':
2167 case '<':
2168 case '>':
2169 case '*':
2170 case '/':
2171 case '%':
2172 case '^':
2173 case '!':
2174 case '=':
2176 register int c1;
2178 combine:
2180 switch (c)
2182 case '+':
2183 yylval.code = PLUS_EXPR; break;
2184 case '-':
2185 yylval.code = MINUS_EXPR; break;
2186 case '&':
2187 yylval.code = BIT_AND_EXPR; break;
2188 case '|':
2189 yylval.code = BIT_IOR_EXPR; break;
2190 case '*':
2191 yylval.code = MULT_EXPR; break;
2192 case '/':
2193 yylval.code = TRUNC_DIV_EXPR; break;
2194 case '%':
2195 yylval.code = TRUNC_MOD_EXPR; break;
2196 case '^':
2197 yylval.code = BIT_XOR_EXPR; break;
2198 case LSHIFT:
2199 yylval.code = LSHIFT_EXPR; break;
2200 case RSHIFT:
2201 yylval.code = RSHIFT_EXPR; break;
2202 case '<':
2203 yylval.code = LT_EXPR; break;
2204 case '>':
2205 yylval.code = GT_EXPR; break;
2208 token_buffer[1] = c1 = GETC();
2209 token_buffer[2] = 0;
2211 if (c1 == '=')
2213 switch (c)
2215 case '<':
2216 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2217 case '>':
2218 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2219 case '!':
2220 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2221 case '=':
2222 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2224 value = ASSIGN; goto done;
2226 else if (c == c1)
2227 switch (c)
2229 case '+':
2230 value = PLUSPLUS; goto done;
2231 case '-':
2232 value = MINUSMINUS; goto done;
2233 case '&':
2234 value = ANDAND; goto done;
2235 case '|':
2236 value = OROR; goto done;
2237 case '<':
2238 c = LSHIFT;
2239 goto combine;
2240 case '>':
2241 c = RSHIFT;
2242 goto combine;
2244 else
2245 switch (c)
2247 case '-':
2248 if (c1 == '>')
2249 { value = POINTSAT; goto done; }
2250 break;
2251 case ':':
2252 if (c1 == '>')
2253 { value = ']'; goto done; }
2254 break;
2255 case '<':
2256 if (c1 == '%')
2257 { value = '{'; indent_level++; goto done; }
2258 if (c1 == ':')
2259 { value = '['; goto done; }
2260 break;
2261 case '%':
2262 if (c1 == '>')
2263 { value = '}'; indent_level--; goto done; }
2264 break;
2266 UNGETC (c1);
2267 token_buffer[1] = 0;
2269 if ((c == '<') || (c == '>'))
2270 value = ARITHCOMPARE;
2271 else value = c;
2272 goto done;
2275 case 0:
2276 /* Don't make yyparse think this is eof. */
2277 value = 1;
2278 break;
2280 case '{':
2281 indent_level++;
2282 value = c;
2283 break;
2285 case '}':
2286 indent_level--;
2287 value = c;
2288 break;
2290 default:
2291 value = c;
2294 done:
2295 /* yylloc.last_line = lineno; */
2297 return value;
2300 /* Sets the value of the 'yydebug' variable to VALUE.
2301 This is a function so we don't have to have YYDEBUG defined
2302 in order to build the compiler. */
2304 void
2305 set_yydebug (value)
2306 int value;
2308 #if YYDEBUG != 0
2309 yydebug = value;
2310 #else
2311 warning ("YYDEBUG not defined.");
2312 #endif