]> gcc.gnu.org Git - gcc.git/blob - gcc/c-lex.c
# Fix misspellings in comments.
[gcc.git] / gcc / c-lex.c
1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 1988, 1989, 1992 Free Software Foundation, Inc.
3
4 This file is part of GNU CC.
5
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.
10
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.
15
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, 675 Mass Ave, Cambridge, MA 02139, USA. */
19
20
21 #include <stdio.h>
22 #include <errno.h>
23 #include <setjmp.h>
24
25 #include "config.h"
26 #include "rtl.h"
27 #include "tree.h"
28 #include "input.h"
29 #include "c-lex.h"
30 #include "c-tree.h"
31 #include "flags.h"
32 #include "c-parse.h"
33
34 #ifdef MULTIBYTE_CHARS
35 #include <stdlib.h>
36 #include <locale.h>
37 #endif
38
39 #ifndef errno
40 extern int errno;
41 #endif
42
43 /* The elements of `ridpointers' are identifier nodes
44 for the reserved type names and storage classes.
45 It is indexed by a RID_... value. */
46 tree ridpointers[(int) RID_MAX];
47
48 /* Cause the `yydebug' variable to be defined. */
49 #define YYDEBUG 1
50
51 /* the declaration found for the last IDENTIFIER token read in.
52 yylex must look this up to detect typedefs, which get token type TYPENAME,
53 so it is left around in case the identifier is not a typedef but is
54 used in a context which makes it a reference to a variable. */
55 tree lastiddecl;
56
57 /* Nonzero enables objc features. */
58
59 int doing_objc_thang;
60
61 extern tree lookup_interface ();
62
63 extern int yydebug;
64
65 /* File used for outputting assembler code. */
66 extern FILE *asm_out_file;
67
68 #ifndef WCHAR_TYPE_SIZE
69 #ifdef INT_TYPE_SIZE
70 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
71 #else
72 #define WCHAR_TYPE_SIZE BITS_PER_WORD
73 #endif
74 #endif
75
76 /* Number of bytes in a wide character. */
77 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
78
79 static int maxtoken; /* Current nominal length of token buffer. */
80 char *token_buffer; /* Pointer to token buffer.
81 Actual allocated length is maxtoken + 2.
82 This is not static because objc-parse.y uses it. */
83
84 /* Nonzero if end-of-file has been seen on input. */
85 static int end_of_file;
86
87 /* Buffered-back input character; faster than using ungetc. */
88 static int nextchar = -1;
89
90 int check_newline ();
91
92 /* Nonzero tells yylex to ignore \ in string constants. */
93 static int ignore_escape_flag = 0;
94 \f
95 /* C code produced by gperf version 2.5 (GNU C++ version) */
96 /* Command-line: gperf -p -j1 -i 1 -g -o -t -N is_reserved_word -k1,3,$ c-parse.gperf */
97 struct resword { char *name; short token; enum rid rid; };
98
99 #define TOTAL_KEYWORDS 53
100 #define MIN_WORD_LENGTH 2
101 #define MAX_WORD_LENGTH 13
102 #define MIN_HASH_VALUE 7
103 #define MAX_HASH_VALUE 102
104 /* maximum key range = 96, duplicates = 0 */
105
106 #ifdef __GNUC__
107 __inline
108 #endif
109 static unsigned int
110 hash (str, len)
111 register char *str;
112 register int unsigned len;
113 {
114 static unsigned char asso_values[] =
115 {
116 103, 103, 103, 103, 103, 103, 103, 103, 103, 103,
117 103, 103, 103, 103, 103, 103, 103, 103, 103, 103,
118 103, 103, 103, 103, 103, 103, 103, 103, 103, 103,
119 103, 103, 103, 103, 103, 103, 103, 103, 103, 103,
120 103, 103, 103, 103, 103, 103, 103, 103, 103, 103,
121 103, 103, 103, 103, 103, 103, 103, 103, 103, 103,
122 103, 103, 103, 103, 103, 103, 103, 103, 103, 103,
123 103, 103, 103, 103, 103, 103, 103, 103, 103, 103,
124 103, 103, 103, 103, 103, 103, 103, 103, 103, 103,
125 103, 103, 103, 103, 103, 1, 103, 2, 1, 24,
126 1, 5, 19, 39, 16, 13, 103, 1, 25, 1,
127 34, 34, 24, 103, 13, 12, 1, 45, 24, 7,
128 103, 103, 2, 103, 103, 103, 103, 103,
129 };
130 register int hval = len;
131
132 switch (hval)
133 {
134 default:
135 case 3:
136 hval += asso_values[str[2]];
137 case 2:
138 case 1:
139 hval += asso_values[str[0]];
140 }
141 return hval + asso_values[str[len - 1]];
142 }
143
144 #ifdef __GNUC__
145 __inline
146 #endif
147 struct resword *
148 is_reserved_word (str, len)
149 register char *str;
150 register unsigned int len;
151 {
152 static struct resword wordlist[] =
153 {
154 {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",},
155 {"asm", ASM_KEYWORD, NORID},
156 {"",},
157 {"__asm", ASM_KEYWORD, NORID},
158 {"",},
159 {"__asm__", ASM_KEYWORD, NORID},
160 {"break", BREAK, NORID},
161 {"__typeof__", TYPEOF, NORID},
162 {"",},
163 {"__alignof__", ALIGNOF, NORID},
164 {"",},
165 {"__attribute__", ATTRIBUTE, NORID},
166 {"int", TYPESPEC, RID_INT},
167 {"__attribute", ATTRIBUTE, NORID},
168 {"__extension__", EXTENSION, NORID},
169 {"",},
170 {"__signed", TYPESPEC, RID_SIGNED},
171 {"",},
172 {"__signed__", TYPESPEC, RID_SIGNED},
173 {"__inline__", SCSPEC, RID_INLINE},
174 {"else", ELSE, NORID},
175 {"__inline", SCSPEC, RID_INLINE},
176 {"default", DEFAULT, NORID},
177 {"__typeof", TYPEOF, NORID},
178 {"while", WHILE, NORID},
179 {"__alignof", ALIGNOF, NORID},
180 {"struct", STRUCT, NORID},
181 {"__const", TYPE_QUAL, RID_CONST},
182 {"if", IF, NORID},
183 {"__const__", TYPE_QUAL, RID_CONST},
184 {"__label__", LABEL, NORID},
185 {"do", DO, NORID},
186 {"__volatile__", TYPE_QUAL, RID_VOLATILE},
187 {"sizeof", SIZEOF, NORID},
188 {"__volatile", TYPE_QUAL, RID_VOLATILE},
189 {"auto", SCSPEC, RID_AUTO},
190 {"void", TYPESPEC, RID_VOID},
191 {"char", TYPESPEC, RID_CHAR},
192 {"static", SCSPEC, RID_STATIC},
193 {"case", CASE, NORID},
194 {"extern", SCSPEC, RID_EXTERN},
195 {"switch", SWITCH, NORID},
196 {"for", FOR, NORID},
197 {"inline", SCSPEC, RID_INLINE},
198 {"typeof", TYPEOF, NORID},
199 {"typedef", SCSPEC, RID_TYPEDEF},
200 {"short", TYPESPEC, RID_SHORT},
201 {"",},
202 {"return", RETURN, NORID},
203 {"enum", ENUM, NORID},
204 {"",},
205 {"double", TYPESPEC, RID_DOUBLE},
206 {"signed", TYPESPEC, RID_SIGNED},
207 {"float", TYPESPEC, RID_FLOAT},
208 {"",}, {"",},
209 {"volatile", TYPE_QUAL, RID_VOLATILE},
210 {"",},
211 {"const", TYPE_QUAL, RID_CONST},
212 {"",},
213 {"unsigned", TYPESPEC, RID_UNSIGNED},
214 {"",}, {"",}, {"",}, {"",},
215 {"continue", CONTINUE, NORID},
216 {"",},
217 {"register", SCSPEC, RID_REGISTER},
218 {"",}, {"",}, {"",}, {"",},
219 {"goto", GOTO, NORID},
220 {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",},
221 {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",},
222
223 {"union", UNION, NORID},
224 {"",}, {"",}, {"",}, {"",},
225 {"long", TYPESPEC, RID_LONG},
226 };
227
228 if (len <= MAX_WORD_LENGTH && len >= MIN_WORD_LENGTH)
229 {
230 register int key = hash (str, len);
231
232 if (key <= MAX_HASH_VALUE && key >= 0)
233 {
234 register char *s = wordlist[key].name;
235
236 if (*s == *str && !strcmp (str + 1, s + 1))
237 return &wordlist[key];
238 }
239 }
240 return 0;
241 }
242 \f
243 /* Return something to represent absolute declarators containing a *.
244 TARGET is the absolute declarator that the * contains.
245 TYPE_QUALS is a list of modifiers such as const or volatile
246 to apply to the pointer type, represented as identifiers.
247
248 We return an INDIRECT_REF whose "contents" are TARGET
249 and whose type is the modifier list. */
250
251 tree
252 make_pointer_declarator (type_quals, target)
253 tree type_quals, target;
254 {
255 return build1 (INDIRECT_REF, type_quals, target);
256 }
257 \f
258 void
259 init_lex ()
260 {
261 /* Make identifier nodes long enough for the language-specific slots. */
262 set_identifier_size (sizeof (struct lang_identifier));
263
264 /* Start it at 0, because check_newline is called at the very beginning
265 and will increment it to 1. */
266 lineno = 0;
267
268 #ifdef MULTIBYTE_CHARS
269 /* Change to the native locale for multibyte conversions. */
270 setlocale (LC_CTYPE, "");
271 #endif
272
273 maxtoken = 40;
274 token_buffer = (char *) xmalloc (maxtoken + 2);
275
276 ridpointers[(int) RID_INT] = get_identifier ("int");
277 ridpointers[(int) RID_CHAR] = get_identifier ("char");
278 ridpointers[(int) RID_VOID] = get_identifier ("void");
279 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
280 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
281 ridpointers[(int) RID_SHORT] = get_identifier ("short");
282 ridpointers[(int) RID_LONG] = get_identifier ("long");
283 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
284 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
285 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
286 ridpointers[(int) RID_CONST] = get_identifier ("const");
287 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
288 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
289 ridpointers[(int) RID_STATIC] = get_identifier ("static");
290 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
291 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
292 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
293
294 /* Some options inhibit certain reserved words.
295 Clear those words out of the hash table so they won't be recognized. */
296 #define UNSET_RESERVED_WORD(STRING) \
297 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
298 if (s) s->name = ""; } while (0)
299
300 if (flag_traditional)
301 {
302 UNSET_RESERVED_WORD ("const");
303 UNSET_RESERVED_WORD ("volatile");
304 UNSET_RESERVED_WORD ("typeof");
305 UNSET_RESERVED_WORD ("signed");
306 UNSET_RESERVED_WORD ("inline");
307 }
308 if (flag_no_asm)
309 {
310 UNSET_RESERVED_WORD ("asm");
311 UNSET_RESERVED_WORD ("typeof");
312 UNSET_RESERVED_WORD ("inline");
313 }
314 }
315
316 void
317 reinit_parse_for_function ()
318 {
319 }
320 \f
321 /* Function used when yydebug is set, to print a token in more detail. */
322
323 void
324 yyprint (file, yychar, yylval)
325 FILE *file;
326 int yychar;
327 YYSTYPE yylval;
328 {
329 tree t;
330 switch (yychar)
331 {
332 case IDENTIFIER:
333 case TYPENAME:
334 t = yylval.ttype;
335 if (IDENTIFIER_POINTER (t))
336 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
337 break;
338
339 case CONSTANT:
340 t = yylval.ttype;
341 if (TREE_CODE (t) == INTEGER_CST)
342 fprintf (file,
343 #if HOST_BITS_PER_WIDE_INT == 64
344 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
345 " 0x%lx%016lx",
346 #else
347 " 0x%x%016x",
348 #endif
349 #else
350 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
351 " 0x%lx%08lx",
352 #else
353 " 0x%x%08x",
354 #endif
355 #endif
356 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
357 break;
358 }
359 }
360
361 \f
362 /* If C is not whitespace, return C.
363 Otherwise skip whitespace and return first nonwhite char read. */
364
365 static int
366 skip_white_space (c)
367 register int c;
368 {
369 static int newline_warning = 0;
370
371 for (;;)
372 {
373 switch (c)
374 {
375 /* We don't recognize comments here, because
376 cpp output can include / and * consecutively as operators.
377 Also, there's no need, since cpp removes all comments. */
378
379 case '\n':
380 c = check_newline ();
381 break;
382
383 case ' ':
384 case '\t':
385 case '\f':
386 case '\v':
387 case '\b':
388 c = getc (finput);
389 break;
390
391 case '\r':
392 /* ANSI C says the effects of a carriage return in a source file
393 are undefined. */
394 if (pedantic && !newline_warning)
395 {
396 warning ("carriage return in source file");
397 warning ("(we only warn about the first carriage return)");
398 newline_warning = 1;
399 }
400 c = getc (finput);
401 break;
402
403 case '\\':
404 c = getc (finput);
405 if (c == '\n')
406 lineno++;
407 else
408 error ("stray '\\' in program");
409 c = getc (finput);
410 break;
411
412 default:
413 return (c);
414 }
415 }
416 }
417
418 /* Skips all of the white space at the current location in the input file.
419 Must use and reset nextchar if it has the next character. */
420
421 void
422 position_after_white_space ()
423 {
424 register int c;
425
426 if (nextchar != -1)
427 c = nextchar, nextchar = -1;
428 else
429 c = getc (finput);
430
431 ungetc (skip_white_space (c), finput);
432 }
433
434 /* Make the token buffer longer, preserving the data in it.
435 P should point to just beyond the last valid character in the old buffer.
436 The value we return is a pointer to the new buffer
437 at a place corresponding to P. */
438
439 static char *
440 extend_token_buffer (p)
441 char *p;
442 {
443 int offset = p - token_buffer;
444
445 maxtoken = maxtoken * 2 + 10;
446 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
447
448 return token_buffer + offset;
449 }
450 \f
451 /* At the beginning of a line, increment the line number
452 and process any #-directive on this line.
453 If the line is a #-directive, read the entire line and return a newline.
454 Otherwise, return the line's first non-whitespace character. */
455
456 int
457 check_newline ()
458 {
459 register int c;
460 register int token;
461
462 lineno++;
463
464 /* Read first nonwhite char on the line. */
465
466 c = getc (finput);
467 while (c == ' ' || c == '\t')
468 c = getc (finput);
469
470 if (c != '#')
471 {
472 /* If not #, return it so caller will use it. */
473 return c;
474 }
475
476 /* Read first nonwhite char after the `#'. */
477
478 c = getc (finput);
479 while (c == ' ' || c == '\t')
480 c = getc (finput);
481
482 /* If a letter follows, then if the word here is `line', skip
483 it and ignore it; otherwise, ignore the line, with an error
484 if the word isn't `pragma', `ident', `define', or `undef'. */
485
486 if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
487 {
488 if (c == 'p')
489 {
490 if (getc (finput) == 'r'
491 && getc (finput) == 'a'
492 && getc (finput) == 'g'
493 && getc (finput) == 'm'
494 && getc (finput) == 'a'
495 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
496 {
497 #ifdef HANDLE_SYSV_PRAGMA
498 return handle_sysv_pragma (finput, c);
499 #endif /* HANDLE_SYSV_PRAGMA */
500 #ifdef HANDLE_PRAGMA
501 HANDLE_PRAGMA (finput);
502 #endif /* HANDLE_PRAGMA */
503 goto skipline;
504 }
505 }
506
507 else if (c == 'd')
508 {
509 if (getc (finput) == 'e'
510 && getc (finput) == 'f'
511 && getc (finput) == 'i'
512 && getc (finput) == 'n'
513 && getc (finput) == 'e'
514 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
515 {
516 #ifdef DWARF_DEBUGGING_INFO
517 if ((debug_info_level == DINFO_LEVEL_VERBOSE)
518 && (write_symbols == DWARF_DEBUG))
519 dwarfout_define (lineno, get_directive_line (finput));
520 #endif /* DWARF_DEBUGGING_INFO */
521 goto skipline;
522 }
523 }
524 else if (c == 'u')
525 {
526 if (getc (finput) == 'n'
527 && getc (finput) == 'd'
528 && getc (finput) == 'e'
529 && getc (finput) == 'f'
530 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
531 {
532 #ifdef DWARF_DEBUGGING_INFO
533 if ((debug_info_level == DINFO_LEVEL_VERBOSE)
534 && (write_symbols == DWARF_DEBUG))
535 dwarfout_undef (lineno, get_directive_line (finput));
536 #endif /* DWARF_DEBUGGING_INFO */
537 goto skipline;
538 }
539 }
540 else if (c == 'l')
541 {
542 if (getc (finput) == 'i'
543 && getc (finput) == 'n'
544 && getc (finput) == 'e'
545 && ((c = getc (finput)) == ' ' || c == '\t'))
546 goto linenum;
547 }
548 else if (c == 'i')
549 {
550 if (getc (finput) == 'd'
551 && getc (finput) == 'e'
552 && getc (finput) == 'n'
553 && getc (finput) == 't'
554 && ((c = getc (finput)) == ' ' || c == '\t'))
555 {
556 /* #ident. The pedantic warning is now in cccp.c. */
557
558 /* Here we have just seen `#ident '.
559 A string constant should follow. */
560
561 while (c == ' ' || c == '\t')
562 c = getc (finput);
563
564 /* If no argument, ignore the line. */
565 if (c == '\n')
566 return c;
567
568 ungetc (c, finput);
569 token = yylex ();
570 if (token != STRING
571 || TREE_CODE (yylval.ttype) != STRING_CST)
572 {
573 error ("invalid #ident");
574 goto skipline;
575 }
576
577 if (!flag_no_ident)
578 {
579 #ifdef ASM_OUTPUT_IDENT
580 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
581 #endif
582 }
583
584 /* Skip the rest of this line. */
585 goto skipline;
586 }
587 }
588
589 error ("undefined or invalid # directive");
590 goto skipline;
591 }
592
593 linenum:
594 /* Here we have either `#line' or `# <nonletter>'.
595 In either case, it should be a line number; a digit should follow. */
596
597 while (c == ' ' || c == '\t')
598 c = getc (finput);
599
600 /* If the # is the only nonwhite char on the line,
601 just ignore it. Check the new newline. */
602 if (c == '\n')
603 return c;
604
605 /* Something follows the #; read a token. */
606
607 ungetc (c, finput);
608 token = yylex ();
609
610 if (token == CONSTANT
611 && TREE_CODE (yylval.ttype) == INTEGER_CST)
612 {
613 int old_lineno = lineno;
614 int used_up = 0;
615 /* subtract one, because it is the following line that
616 gets the specified number */
617
618 int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
619
620 /* Is this the last nonwhite stuff on the line? */
621 c = getc (finput);
622 while (c == ' ' || c == '\t')
623 c = getc (finput);
624 if (c == '\n')
625 {
626 /* No more: store the line number and check following line. */
627 lineno = l;
628 return c;
629 }
630 ungetc (c, finput);
631
632 /* More follows: it must be a string constant (filename). */
633
634 /* Read the string constant, but don't treat \ as special. */
635 ignore_escape_flag = 1;
636 token = yylex ();
637 ignore_escape_flag = 0;
638
639 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
640 {
641 error ("invalid #line");
642 goto skipline;
643 }
644
645 input_filename
646 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
647 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
648 lineno = l;
649
650 /* Each change of file name
651 reinitializes whether we are now in a system header. */
652 in_system_header = 0;
653
654 if (main_input_filename == 0)
655 main_input_filename = input_filename;
656
657 /* Is this the last nonwhite stuff on the line? */
658 c = getc (finput);
659 while (c == ' ' || c == '\t')
660 c = getc (finput);
661 if (c == '\n')
662 return c;
663 ungetc (c, finput);
664
665 token = yylex ();
666 used_up = 0;
667
668 /* `1' after file name means entering new file.
669 `2' after file name means just left a file. */
670
671 if (token == CONSTANT
672 && TREE_CODE (yylval.ttype) == INTEGER_CST)
673 {
674 if (TREE_INT_CST_LOW (yylval.ttype) == 1)
675 {
676 /* Pushing to a new file. */
677 struct file_stack *p
678 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
679 input_file_stack->line = old_lineno;
680 p->next = input_file_stack;
681 p->name = input_filename;
682 input_file_stack = p;
683 input_file_stack_tick++;
684 #ifdef DWARF_DEBUGGING_INFO
685 if (debug_info_level == DINFO_LEVEL_VERBOSE
686 && write_symbols == DWARF_DEBUG)
687 dwarfout_start_new_source_file (input_filename);
688 #endif /* DWARF_DEBUGGING_INFO */
689
690 used_up = 1;
691 }
692 else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
693 {
694 /* Popping out of a file. */
695 if (input_file_stack->next)
696 {
697 struct file_stack *p = input_file_stack;
698 input_file_stack = p->next;
699 free (p);
700 input_file_stack_tick++;
701 #ifdef DWARF_DEBUGGING_INFO
702 if (debug_info_level == DINFO_LEVEL_VERBOSE
703 && write_symbols == DWARF_DEBUG)
704 dwarfout_resume_previous_source_file (input_file_stack->line);
705 #endif /* DWARF_DEBUGGING_INFO */
706 }
707 else
708 error ("#-lines for entering and leaving files don't match");
709
710 used_up = 1;
711 }
712 }
713
714 /* If we have handled a `1' or a `2',
715 see if there is another number to read. */
716 if (used_up)
717 {
718 /* Is this the last nonwhite stuff on the line? */
719 c = getc (finput);
720 while (c == ' ' || c == '\t')
721 c = getc (finput);
722 if (c == '\n')
723 return c;
724 ungetc (c, finput);
725
726 token = yylex ();
727 used_up = 0;
728 }
729
730 /* `3' after file name means this is a system header file. */
731
732 if (token == CONSTANT
733 && TREE_CODE (yylval.ttype) == INTEGER_CST
734 && TREE_INT_CST_LOW (yylval.ttype) == 3)
735 in_system_header = 1;
736 }
737 else
738 error ("invalid #-line");
739
740 /* skip the rest of this line. */
741 skipline:
742 if (c == '\n')
743 return c;
744 while ((c = getc (finput)) != EOF && c != '\n');
745 return c;
746 }
747 \f
748 #ifdef HANDLE_SYSV_PRAGMA
749
750 /* Handle a #pragma directive. INPUT is the current input stream,
751 and C is a character to reread. Processes the entire input line
752 and returns a character for the caller to reread: either \n or EOF. */
753
754 /* This function has to be in this file, in order to get at
755 the token types. */
756
757 int
758 handle_sysv_pragma (input, c)
759 FILE *input;
760 int c;
761 {
762 for (;;)
763 {
764 while (c == ' ' || c == '\t')
765 c = getc (input);
766 if (c == '\n' || c == EOF)
767 {
768 handle_pragma_token (0, 0);
769 return c;
770 }
771 ungetc (c, input);
772 switch (yylex ())
773 {
774 case IDENTIFIER:
775 case TYPENAME:
776 case STRING:
777 case CONSTANT:
778 handle_pragma_token (token_buffer, yylval.ttype);
779 break;
780 default:
781 handle_pragma_token (token_buffer, 0);
782 }
783 if (nextchar >= 0)
784 c = nextchar, nextchar = -1;
785 else
786 c = getc (input);
787 }
788 }
789
790 #endif /* HANDLE_SYSV_PRAGMA */
791 \f
792 #define isalnum(char) ((char >= 'a' && char <= 'z') || (char >= 'A' && char <= 'Z') || (char >= '0' && char <= '9'))
793 #define isdigit(char) (char >= '0' && char <= '9')
794 #define ENDFILE -1 /* token that represents end-of-file */
795
796 /* Read an escape sequence, returning its equivalent as a character,
797 or store 1 in *ignore_ptr if it is backslash-newline. */
798
799 static int
800 readescape (ignore_ptr)
801 int *ignore_ptr;
802 {
803 register int c = getc (finput);
804 register int code;
805 register unsigned count;
806 unsigned firstdig;
807 int nonnull;
808
809 switch (c)
810 {
811 case 'x':
812 if (warn_traditional)
813 warning ("the meaning of `\\x' varies with -traditional");
814
815 if (flag_traditional)
816 return c;
817
818 code = 0;
819 count = 0;
820 nonnull = 0;
821 while (1)
822 {
823 c = getc (finput);
824 if (!(c >= 'a' && c <= 'f')
825 && !(c >= 'A' && c <= 'F')
826 && !(c >= '0' && c <= '9'))
827 {
828 ungetc (c, finput);
829 break;
830 }
831 code *= 16;
832 if (c >= 'a' && c <= 'f')
833 code += c - 'a' + 10;
834 if (c >= 'A' && c <= 'F')
835 code += c - 'A' + 10;
836 if (c >= '0' && c <= '9')
837 code += c - '0';
838 if (code != 0 || count != 0)
839 {
840 if (count == 0)
841 firstdig = code;
842 count++;
843 }
844 nonnull = 1;
845 }
846 if (! nonnull)
847 error ("\\x used with no following hex digits");
848 else if (count == 0)
849 /* Digits are all 0's. Ok. */
850 ;
851 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
852 || (count > 1
853 && ((1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
854 <= firstdig)))
855 pedwarn ("hex escape out of range");
856 return code;
857
858 case '0': case '1': case '2': case '3': case '4':
859 case '5': case '6': case '7':
860 code = 0;
861 count = 0;
862 while ((c <= '7') && (c >= '0') && (count++ < 3))
863 {
864 code = (code * 8) + (c - '0');
865 c = getc (finput);
866 }
867 ungetc (c, finput);
868 return code;
869
870 case '\\': case '\'': case '"':
871 return c;
872
873 case '\n':
874 lineno++;
875 *ignore_ptr = 1;
876 return 0;
877
878 case 'n':
879 return TARGET_NEWLINE;
880
881 case 't':
882 return TARGET_TAB;
883
884 case 'r':
885 return TARGET_CR;
886
887 case 'f':
888 return TARGET_FF;
889
890 case 'b':
891 return TARGET_BS;
892
893 case 'a':
894 if (warn_traditional)
895 warning ("the meaning of `\\a' varies with -traditional");
896
897 if (flag_traditional)
898 return c;
899 return TARGET_BELL;
900
901 case 'v':
902 #if 0 /* Vertical tab is present in common usage compilers. */
903 if (flag_traditional)
904 return c;
905 #endif
906 return TARGET_VT;
907
908 case 'e':
909 case 'E':
910 if (pedantic)
911 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
912 return 033;
913
914 case '?':
915 return c;
916
917 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
918 case '(':
919 case '{':
920 case '[':
921 if (pedantic)
922 pedwarn ("non-ANSI escape sequence `\\%c'", c);
923 return c;
924 }
925 if (c >= 040 && c < 0177)
926 pedwarn ("unknown escape sequence `\\%c'", c);
927 else
928 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
929 return c;
930 }
931 \f
932 void
933 yyerror (string)
934 char *string;
935 {
936 char buf[200];
937
938 strcpy (buf, string);
939
940 /* We can't print string and character constants well
941 because the token_buffer contains the result of processing escapes. */
942 if (end_of_file)
943 strcat (buf, " at end of input");
944 else if (token_buffer[0] == 0)
945 strcat (buf, " at null character");
946 else if (token_buffer[0] == '"')
947 strcat (buf, " before string constant");
948 else if (token_buffer[0] == '\'')
949 strcat (buf, " before character constant");
950 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
951 sprintf (buf + strlen (buf), " before character 0%o",
952 (unsigned char) token_buffer[0]);
953 else
954 strcat (buf, " before `%s'");
955
956 error (buf, token_buffer);
957 }
958
959 #if 0
960
961 struct try_type
962 {
963 tree *node_var;
964 char unsigned_flag;
965 char long_flag;
966 char long_long_flag;
967 };
968
969 struct try_type type_sequence[] =
970 {
971 { &integer_type_node, 0, 0, 0},
972 { &unsigned_type_node, 1, 0, 0},
973 { &long_integer_type_node, 0, 1, 0},
974 { &long_unsigned_type_node, 1, 1, 0},
975 { &long_long_integer_type_node, 0, 1, 1},
976 { &long_long_unsigned_type_node, 1, 1, 1}
977 };
978 #endif /* 0 */
979 \f
980 int
981 yylex ()
982 {
983 register int c;
984 register char *p;
985 register int value;
986 int wide_flag = 0;
987
988 if (nextchar >= 0)
989 c = nextchar, nextchar = -1;
990 else
991 c = getc (finput);
992
993 /* Effectively do c = skip_white_space (c)
994 but do it faster in the usual cases. */
995 while (1)
996 switch (c)
997 {
998 case ' ':
999 case '\t':
1000 case '\f':
1001 case '\v':
1002 case '\b':
1003 c = getc (finput);
1004 break;
1005
1006 case '\r':
1007 /* Call skip_white_space so we can warn if appropriate. */
1008
1009 case '\n':
1010 case '/':
1011 case '\\':
1012 c = skip_white_space (c);
1013 default:
1014 goto found_nonwhite;
1015 }
1016 found_nonwhite:
1017
1018 token_buffer[0] = c;
1019 token_buffer[1] = 0;
1020
1021 /* yylloc.first_line = lineno; */
1022
1023 switch (c)
1024 {
1025 case EOF:
1026 end_of_file = 1;
1027 token_buffer[0] = 0;
1028 value = ENDFILE;
1029 break;
1030
1031 case '$':
1032 if (dollars_in_ident)
1033 goto letter;
1034 return '$';
1035
1036 case 'L':
1037 /* Capital L may start a wide-string or wide-character constant. */
1038 {
1039 register int c = getc (finput);
1040 if (c == '\'')
1041 {
1042 wide_flag = 1;
1043 goto char_constant;
1044 }
1045 if (c == '"')
1046 {
1047 wide_flag = 1;
1048 goto string_constant;
1049 }
1050 ungetc (c, finput);
1051 }
1052 goto letter;
1053
1054 case '@':
1055 if (!doing_objc_thang)
1056 {
1057 value = c;
1058 break;
1059 }
1060 p = token_buffer;
1061 *p++ = '@';
1062 c = getc (finput);
1063 while (isalnum (c) || c == '_')
1064 {
1065 if (p >= token_buffer + maxtoken)
1066 p = extend_token_buffer (p);
1067
1068 *p++ = c;
1069 c = getc (finput);
1070 }
1071
1072 *p = 0;
1073 nextchar = c;
1074 value = recognize_objc_keyword (token_buffer + 1);
1075 if (value != 0)
1076 break;
1077 error ("invalid Objective C keyword `%s'", token_buffer);
1078 /* Cause a syntax error--1 is not a valid token type. */
1079 value = 1;
1080 break;
1081
1082 case 'A': case 'B': case 'C': case 'D': case 'E':
1083 case 'F': case 'G': case 'H': case 'I': case 'J':
1084 case 'K': case 'M': case 'N': case 'O':
1085 case 'P': case 'Q': case 'R': case 'S': case 'T':
1086 case 'U': case 'V': case 'W': case 'X': case 'Y':
1087 case 'Z':
1088 case 'a': case 'b': case 'c': case 'd': case 'e':
1089 case 'f': case 'g': case 'h': case 'i': case 'j':
1090 case 'k': case 'l': case 'm': case 'n': case 'o':
1091 case 'p': case 'q': case 'r': case 's': case 't':
1092 case 'u': case 'v': case 'w': case 'x': case 'y':
1093 case 'z':
1094 case '_':
1095 letter:
1096 p = token_buffer;
1097 while (isalnum (c) || c == '_' || c == '$' || c == '@')
1098 {
1099 /* Make sure this char really belongs in an identifier. */
1100 if (c == '@' && ! doing_objc_thang)
1101 break;
1102 if (c == '$' && ! dollars_in_ident)
1103 break;
1104
1105 if (p >= token_buffer + maxtoken)
1106 p = extend_token_buffer (p);
1107
1108 *p++ = c;
1109 c = getc (finput);
1110 }
1111
1112 *p = 0;
1113 nextchar = c;
1114
1115 value = IDENTIFIER;
1116 yylval.itype = 0;
1117
1118 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1119
1120 {
1121 register struct resword *ptr;
1122
1123 if (ptr = is_reserved_word (token_buffer, p - token_buffer))
1124 {
1125 if (ptr->rid)
1126 yylval.ttype = ridpointers[(int) ptr->rid];
1127 value = (int) ptr->token;
1128
1129 /* Even if we decided to recognize asm, still perhaps warn. */
1130 if (pedantic
1131 && (value == ASM_KEYWORD || value == TYPEOF
1132 || ptr->rid == RID_INLINE)
1133 && token_buffer[0] != '_')
1134 pedwarn ("ANSI does not permit the keyword `%s'",
1135 token_buffer);
1136 }
1137 }
1138
1139 /* If we did not find a keyword, look for an identifier
1140 (or a typename). */
1141
1142 if (value == IDENTIFIER)
1143 {
1144 yylval.ttype = get_identifier (token_buffer);
1145 lastiddecl = lookup_name (yylval.ttype);
1146
1147 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1148 value = TYPENAME;
1149 /* A user-invisible read-only initialized variable
1150 should be replaced by its value.
1151 We handle only strings since that's the only case used in C. */
1152 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1153 && DECL_IGNORED_P (lastiddecl)
1154 && TREE_READONLY (lastiddecl)
1155 && DECL_INITIAL (lastiddecl) != 0
1156 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1157 {
1158 yylval.ttype = DECL_INITIAL (lastiddecl);
1159 value = STRING;
1160 }
1161 else if (doing_objc_thang)
1162 {
1163 tree objc_interface_decl = lookup_interface (yylval.ttype);
1164
1165 if (objc_interface_decl)
1166 {
1167 value = CLASSNAME;
1168 yylval.ttype = objc_interface_decl;
1169 }
1170 }
1171 }
1172
1173 break;
1174
1175 case '0': case '1': case '2': case '3': case '4':
1176 case '5': case '6': case '7': case '8': case '9':
1177 case '.':
1178 {
1179 int base = 10;
1180 int count = 0;
1181 int largest_digit = 0;
1182 int numdigits = 0;
1183 /* for multi-precision arithmetic,
1184 we actually store only HOST_BITS_PER_CHAR bits in each part.
1185 The number of parts is chosen so as to be sufficient to hold
1186 the enough bits to fit into the two HOST_WIDE_INTs that contain
1187 the integer value (this is always at least as many bits as are
1188 in a target `long long' value, but may be wider). */
1189 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1190 int parts[TOTAL_PARTS];
1191 int overflow = 0;
1192
1193 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag
1194 = NOT_FLOAT;
1195
1196 for (count = 0; count < TOTAL_PARTS; count++)
1197 parts[count] = 0;
1198
1199 p = token_buffer;
1200 *p++ = c;
1201
1202 if (c == '0')
1203 {
1204 *p++ = (c = getc (finput));
1205 if ((c == 'x') || (c == 'X'))
1206 {
1207 base = 16;
1208 *p++ = (c = getc (finput));
1209 }
1210 /* Leading 0 forces octal unless the 0 is the only digit. */
1211 else if (c >= '0' && c <= '9')
1212 {
1213 base = 8;
1214 numdigits++;
1215 }
1216 else
1217 numdigits++;
1218 }
1219
1220 /* Read all the digits-and-decimal-points. */
1221
1222 while (c == '.'
1223 || (isalnum (c) && (c != 'l') && (c != 'L')
1224 && (c != 'u') && (c != 'U')
1225 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1226 {
1227 if (c == '.')
1228 {
1229 if (base == 16)
1230 error ("floating constant may not be in radix 16");
1231 if (floatflag == AFTER_POINT)
1232 {
1233 error ("malformed floating constant");
1234 floatflag = TOO_MANY_POINTS;
1235 }
1236 else
1237 floatflag = AFTER_POINT;
1238
1239 base = 10;
1240 *p++ = c = getc (finput);
1241 /* Accept '.' as the start of a floating-point number
1242 only when it is followed by a digit.
1243 Otherwise, unread the following non-digit
1244 and use the '.' as a structural token. */
1245 if (p == token_buffer + 2 && !isdigit (c))
1246 {
1247 if (c == '.')
1248 {
1249 c = getc (finput);
1250 if (c == '.')
1251 {
1252 *p++ = c;
1253 *p = 0;
1254 return ELLIPSIS;
1255 }
1256 error ("parse error at `..'");
1257 }
1258 ungetc (c, finput);
1259 token_buffer[1] = 0;
1260 value = '.';
1261 goto done;
1262 }
1263 }
1264 else
1265 {
1266 /* It is not a decimal point.
1267 It should be a digit (perhaps a hex digit). */
1268
1269 if (isdigit (c))
1270 {
1271 c = c - '0';
1272 }
1273 else if (base <= 10)
1274 {
1275 if ((c&~040) == 'E')
1276 {
1277 base = 10;
1278 floatflag = AFTER_POINT;
1279 break; /* start of exponent */
1280 }
1281 error ("nondigits in number and not hexadecimal");
1282 c = 0;
1283 }
1284 else if (c >= 'a')
1285 {
1286 c = c - 'a' + 10;
1287 }
1288 else
1289 {
1290 c = c - 'A' + 10;
1291 }
1292 if (c >= largest_digit)
1293 largest_digit = c;
1294 numdigits++;
1295
1296 for (count = 0; count < TOTAL_PARTS; count++)
1297 {
1298 parts[count] *= base;
1299 if (count)
1300 {
1301 parts[count]
1302 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1303 parts[count-1]
1304 &= (1 << HOST_BITS_PER_CHAR) - 1;
1305 }
1306 else
1307 parts[0] += c;
1308 }
1309
1310 /* If the extra highest-order part ever gets anything in it,
1311 the number is certainly too big. */
1312 if (parts[TOTAL_PARTS - 1] != 0)
1313 overflow = 1;
1314
1315 if (p >= token_buffer + maxtoken - 3)
1316 p = extend_token_buffer (p);
1317 *p++ = (c = getc (finput));
1318 }
1319 }
1320
1321 if (numdigits == 0)
1322 error ("numeric constant with no digits");
1323
1324 if (largest_digit >= base)
1325 error ("numeric constant contains digits beyond the radix");
1326
1327 /* Remove terminating char from the token buffer and delimit the string */
1328 *--p = 0;
1329
1330 if (floatflag != NOT_FLOAT)
1331 {
1332 tree type = double_type_node;
1333 int garbage_chars = 0, exceeds_double = 0;
1334 REAL_VALUE_TYPE value;
1335 jmp_buf handler;
1336
1337 /* Read explicit exponent if any, and put it in tokenbuf. */
1338
1339 if ((c == 'e') || (c == 'E'))
1340 {
1341 if (p >= token_buffer + maxtoken - 3)
1342 p = extend_token_buffer (p);
1343 *p++ = c;
1344 c = getc (finput);
1345 if ((c == '+') || (c == '-'))
1346 {
1347 *p++ = c;
1348 c = getc (finput);
1349 }
1350 if (! isdigit (c))
1351 error ("floating constant exponent has no digits");
1352 while (isdigit (c))
1353 {
1354 if (p >= token_buffer + maxtoken - 3)
1355 p = extend_token_buffer (p);
1356 *p++ = c;
1357 c = getc (finput);
1358 }
1359 }
1360
1361 *p = 0;
1362 errno = 0;
1363
1364 /* Convert string to a double, checking for overflow. */
1365 if (setjmp (handler))
1366 {
1367 error ("floating constant out of range");
1368 value = dconst0;
1369 }
1370 else
1371 {
1372 set_float_handler (handler);
1373 value = REAL_VALUE_ATOF (token_buffer);
1374 set_float_handler (NULL_PTR);
1375 }
1376 #ifdef ERANGE
1377 if (errno == ERANGE && !flag_traditional && pedantic)
1378 {
1379 /* ERANGE is also reported for underflow,
1380 so test the value to distinguish overflow from that. */
1381 if (REAL_VALUES_LESS (dconst1, value)
1382 || REAL_VALUES_LESS (value, dconstm1))
1383 {
1384 pedwarn ("floating point number exceeds range of `double'");
1385 exceeds_double = 1;
1386 }
1387 }
1388 #endif
1389
1390 /* Read the suffixes to choose a data type. */
1391 switch (c)
1392 {
1393 case 'f': case 'F':
1394 type = float_type_node;
1395 value = REAL_VALUE_TRUNCATE (TYPE_MODE (type), value);
1396 if (REAL_VALUE_ISINF (value) && ! exceeds_double && pedantic)
1397 pedwarn ("floating point number exceeds range of `float'");
1398 garbage_chars = -1;
1399 break;
1400
1401 case 'l': case 'L':
1402 type = long_double_type_node;
1403 garbage_chars = -1;
1404 break;
1405 }
1406 /* Note: garbage_chars is -1 if first char is *not* garbage. */
1407 while (isalnum (c))
1408 {
1409 if (p >= token_buffer + maxtoken - 3)
1410 p = extend_token_buffer (p);
1411 *p++ = c;
1412 c = getc (finput);
1413 garbage_chars++;
1414 }
1415 if (garbage_chars > 0)
1416 error ("garbage at end of number");
1417
1418 /* Create a node with determined type and value. */
1419 yylval.ttype = build_real (type, value);
1420
1421 ungetc (c, finput);
1422 *p = 0;
1423 }
1424 else
1425 {
1426 tree traditional_type, ansi_type, type;
1427 HOST_WIDE_INT high, low;
1428 int spec_unsigned = 0;
1429 int spec_long = 0;
1430 int spec_long_long = 0;
1431 int bytes, warn, i;
1432
1433 while (1)
1434 {
1435 if (c == 'u' || c == 'U')
1436 {
1437 if (spec_unsigned)
1438 error ("two `u's in integer constant");
1439 spec_unsigned = 1;
1440 }
1441 else if (c == 'l' || c == 'L')
1442 {
1443 if (spec_long)
1444 {
1445 if (spec_long_long)
1446 error ("three `l's in integer constant");
1447 else if (pedantic)
1448 pedwarn ("ANSI C forbids long long integer constants");
1449 spec_long_long = 1;
1450 }
1451 spec_long = 1;
1452 }
1453 else
1454 {
1455 if (isalnum (c))
1456 {
1457 error ("garbage at end of number");
1458 while (isalnum (c))
1459 {
1460 if (p >= token_buffer + maxtoken - 3)
1461 p = extend_token_buffer (p);
1462 *p++ = c;
1463 c = getc (finput);
1464 }
1465 }
1466 break;
1467 }
1468 if (p >= token_buffer + maxtoken - 3)
1469 p = extend_token_buffer (p);
1470 *p++ = c;
1471 c = getc (finput);
1472 }
1473
1474 ungetc (c, finput);
1475
1476 /* If the constant is not long long and it won't fit in an
1477 unsigned long, or if the constant is long long and won't fit
1478 in an unsigned long long, then warn that the constant is out
1479 of range. */
1480
1481 /* ??? This assumes that long long and long integer types are
1482 a multiple of 8 bits. This better than the original code
1483 though which assumed that long was exactly 32 bits and long
1484 long was exactly 64 bits. */
1485
1486 if (spec_long_long)
1487 bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1488 else
1489 bytes = TYPE_PRECISION (long_integer_type_node) / 8;
1490
1491 warn = overflow;
1492 for (i = bytes; i < TOTAL_PARTS; i++)
1493 if (parts[i])
1494 warn = 1;
1495 if (warn)
1496 pedwarn ("integer constant out of range");
1497
1498 /* This is simplified by the fact that our constant
1499 is always positive. */
1500
1501 high = low = 0;
1502
1503 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1504 {
1505 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1506 / HOST_BITS_PER_CHAR)]
1507 << (i * HOST_BITS_PER_CHAR));
1508 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1509 }
1510
1511 yylval.ttype = build_int_2 (low, high);
1512 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1513
1514 /* If warn_traditional, calculate both the ANSI type and the
1515 traditional type, then see if they disagree.
1516 Otherwise, calculate only the type for the dialect in use. */
1517 if (warn_traditional || flag_traditional)
1518 {
1519 /* Calculate the traditional type. */
1520 /* Traditionally, any constant is signed;
1521 but if unsigned is specified explicitly, obey that.
1522 Use the smallest size with the right number of bits,
1523 except for one special case with decimal constants. */
1524 if (! spec_long && base != 10
1525 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1526 traditional_type = (spec_unsigned ? unsigned_type_node
1527 : integer_type_node);
1528 /* A decimal constant must be long
1529 if it does not fit in type int.
1530 I think this is independent of whether
1531 the constant is signed. */
1532 else if (! spec_long && base == 10
1533 && int_fits_type_p (yylval.ttype, integer_type_node))
1534 traditional_type = (spec_unsigned ? unsigned_type_node
1535 : integer_type_node);
1536 else if (! spec_long_long)
1537 traditional_type = (spec_unsigned ? long_unsigned_type_node
1538 : long_integer_type_node);
1539 else
1540 traditional_type = (spec_unsigned
1541 ? long_long_unsigned_type_node
1542 : long_long_integer_type_node);
1543 }
1544 if (warn_traditional || ! flag_traditional)
1545 {
1546 /* Calculate the ANSI type. */
1547 if (! spec_long && ! spec_unsigned
1548 && int_fits_type_p (yylval.ttype, integer_type_node))
1549 ansi_type = integer_type_node;
1550 else if (! spec_long && (base != 10 || spec_unsigned)
1551 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1552 ansi_type = unsigned_type_node;
1553 else if (! spec_unsigned && !spec_long_long
1554 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1555 ansi_type = long_integer_type_node;
1556 else if (! spec_long_long)
1557 ansi_type = long_unsigned_type_node;
1558 else if (! spec_unsigned
1559 /* Verify value does not overflow into sign bit. */
1560 && TREE_INT_CST_HIGH (yylval.ttype) >= 0
1561 && int_fits_type_p (yylval.ttype,
1562 long_long_integer_type_node))
1563 ansi_type = long_long_integer_type_node;
1564 else
1565 ansi_type = long_long_unsigned_type_node;
1566 }
1567
1568 type = flag_traditional ? traditional_type : ansi_type;
1569
1570 if (warn_traditional && traditional_type != ansi_type)
1571 {
1572 if (TYPE_PRECISION (traditional_type)
1573 != TYPE_PRECISION (ansi_type))
1574 warning ("width of integer constant changes with -traditional");
1575 else if (TREE_UNSIGNED (traditional_type)
1576 != TREE_UNSIGNED (ansi_type))
1577 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1578 else
1579 warning ("width of integer constant may change on other systems with -traditional");
1580 }
1581
1582 if (!flag_traditional && !int_fits_type_p (yylval.ttype, type)
1583 && !warn)
1584 pedwarn ("integer constant out of range");
1585
1586 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1587 warning ("decimal constant is so large that it is unsigned");
1588
1589 if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1590 /* The traditional constant 0x80000000 is signed
1591 but doesn't fit in the range of int.
1592 This will change it to -0x80000000, which does fit. */
1593 {
1594 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1595 yylval.ttype = convert (type, yylval.ttype);
1596 }
1597 else
1598 TREE_TYPE (yylval.ttype) = type;
1599
1600 *p = 0;
1601 }
1602
1603 value = CONSTANT; break;
1604 }
1605
1606 case '\'':
1607 char_constant:
1608 {
1609 register int result = 0;
1610 register int num_chars = 0;
1611 unsigned width = TYPE_PRECISION (char_type_node);
1612 int max_chars;
1613
1614 if (wide_flag)
1615 {
1616 width = WCHAR_TYPE_SIZE;
1617 #ifdef MULTIBYTE_CHARS
1618 max_chars = MB_CUR_MAX;
1619 #else
1620 max_chars = 1;
1621 #endif
1622 }
1623 else
1624 max_chars = TYPE_PRECISION (integer_type_node) / width;
1625
1626 while (1)
1627 {
1628 tryagain:
1629
1630 c = getc (finput);
1631
1632 if (c == '\'' || c == EOF)
1633 break;
1634
1635 if (c == '\\')
1636 {
1637 int ignore = 0;
1638 c = readescape (&ignore);
1639 if (ignore)
1640 goto tryagain;
1641 if (width < HOST_BITS_PER_INT
1642 && (unsigned) c >= (1 << width))
1643 pedwarn ("escape sequence out of range for character");
1644 }
1645 else if (c == '\n')
1646 {
1647 if (pedantic)
1648 pedwarn ("ANSI C forbids newline in character constant");
1649 lineno++;
1650 }
1651
1652 num_chars++;
1653 if (num_chars > maxtoken - 4)
1654 extend_token_buffer (token_buffer);
1655
1656 token_buffer[num_chars] = c;
1657
1658 /* Merge character into result; ignore excess chars. */
1659 if (num_chars < max_chars + 1)
1660 {
1661 if (width < HOST_BITS_PER_INT)
1662 result = (result << width) | (c & ((1 << width) - 1));
1663 else
1664 result = c;
1665 }
1666 }
1667
1668 token_buffer[num_chars + 1] = '\'';
1669 token_buffer[num_chars + 2] = 0;
1670
1671 if (c != '\'')
1672 error ("malformatted character constant");
1673 else if (num_chars == 0)
1674 error ("empty character constant");
1675 else if (num_chars > max_chars)
1676 {
1677 num_chars = max_chars;
1678 error ("character constant too long");
1679 }
1680 else if (num_chars != 1 && ! flag_traditional)
1681 warning ("multi-character character constant");
1682
1683 /* If char type is signed, sign-extend the constant. */
1684 if (! wide_flag)
1685 {
1686 int num_bits = num_chars * width;
1687 if (TREE_UNSIGNED (char_type_node)
1688 || ((result >> (num_bits - 1)) & 1) == 0)
1689 yylval.ttype
1690 = build_int_2 (result & ((unsigned HOST_WIDE_INT) ~0
1691 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1692 0);
1693 else
1694 yylval.ttype
1695 = build_int_2 (result | ~((unsigned HOST_WIDE_INT) ~0
1696 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1697 -1);
1698 }
1699 else
1700 {
1701 #ifdef MULTIBYTE_CHARS
1702 /* Set the initial shift state and convert the next sequence. */
1703 result = 0;
1704 /* In all locales L'\0' is zero and mbtowc will return zero,
1705 so don't use it. */
1706 if (num_chars > 1
1707 || (num_chars == 1 && token_buffer[1] != '\0'))
1708 {
1709 wchar_t wc;
1710 (void) mbtowc (NULL_PTR, NULL_PTR, 0);
1711 if (mbtowc (& wc, token_buffer + 1, num_chars) == num_chars)
1712 result = wc;
1713 else
1714 warning ("Ignoring invalid multibyte character");
1715 }
1716 #endif
1717 yylval.ttype = build_int_2 (result, 0);
1718 }
1719
1720 TREE_TYPE (yylval.ttype) = integer_type_node;
1721 value = CONSTANT;
1722 break;
1723 }
1724
1725 case '"':
1726 string_constant:
1727 {
1728 c = getc (finput);
1729 p = token_buffer + 1;
1730
1731 while (c != '"' && c >= 0)
1732 {
1733 /* ignore_escape_flag is set for reading the filename in #line. */
1734 if (!ignore_escape_flag && c == '\\')
1735 {
1736 int ignore = 0;
1737 c = readescape (&ignore);
1738 if (ignore)
1739 goto skipnewline;
1740 if (!wide_flag
1741 && TYPE_PRECISION (char_type_node) < HOST_BITS_PER_INT
1742 && c >= (1 << TYPE_PRECISION (char_type_node)))
1743 pedwarn ("escape sequence out of range for character");
1744 }
1745 else if (c == '\n')
1746 {
1747 if (pedantic)
1748 pedwarn ("ANSI C forbids newline in string constant");
1749 lineno++;
1750 }
1751
1752 if (p == token_buffer + maxtoken)
1753 p = extend_token_buffer (p);
1754 *p++ = c;
1755
1756 skipnewline:
1757 c = getc (finput);
1758 }
1759 *p = 0;
1760
1761 /* We have read the entire constant.
1762 Construct a STRING_CST for the result. */
1763
1764 if (wide_flag)
1765 {
1766 /* If this is a L"..." wide-string, convert the multibyte string
1767 to a wide character string. */
1768 char *widep = (char *) alloca ((p - token_buffer) * WCHAR_BYTES);
1769 int len;
1770
1771 #ifdef MULTIBYTE_CHARS
1772 len = mbstowcs ((wchar_t *) widep, token_buffer + 1, p - token_buffer);
1773 if ((unsigned) len >= (p - token_buffer))
1774 {
1775 warning ("Ignoring invalid multibyte string");
1776 len = 0;
1777 }
1778 bzero (widep + (len * WCHAR_BYTES), WCHAR_BYTES);
1779 #else
1780 {
1781 union { long l; char c[sizeof (long)]; } u;
1782 int big_endian;
1783 char *wp, *cp;
1784
1785 /* Determine whether host is little or big endian. */
1786 u.l = 1;
1787 big_endian = u.c[sizeof (long) - 1];
1788 wp = widep + (big_endian ? WCHAR_BYTES - 1 : 0);
1789
1790 bzero (widep, (p - token_buffer) * WCHAR_BYTES);
1791 for (cp = token_buffer + 1; cp < p; cp++)
1792 *wp = *cp, wp += WCHAR_BYTES;
1793 len = p - token_buffer - 1;
1794 }
1795 #endif
1796 yylval.ttype = build_string ((len + 1) * WCHAR_BYTES, widep);
1797 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
1798 }
1799 else
1800 {
1801 yylval.ttype = build_string (p - token_buffer, token_buffer + 1);
1802 TREE_TYPE (yylval.ttype) = char_array_type_node;
1803 }
1804
1805 *p++ = '"';
1806 *p = 0;
1807
1808 value = STRING; break;
1809 }
1810
1811 case '+':
1812 case '-':
1813 case '&':
1814 case '|':
1815 case '<':
1816 case '>':
1817 case '*':
1818 case '/':
1819 case '%':
1820 case '^':
1821 case '!':
1822 case '=':
1823 {
1824 register int c1;
1825
1826 combine:
1827
1828 switch (c)
1829 {
1830 case '+':
1831 yylval.code = PLUS_EXPR; break;
1832 case '-':
1833 yylval.code = MINUS_EXPR; break;
1834 case '&':
1835 yylval.code = BIT_AND_EXPR; break;
1836 case '|':
1837 yylval.code = BIT_IOR_EXPR; break;
1838 case '*':
1839 yylval.code = MULT_EXPR; break;
1840 case '/':
1841 yylval.code = TRUNC_DIV_EXPR; break;
1842 case '%':
1843 yylval.code = TRUNC_MOD_EXPR; break;
1844 case '^':
1845 yylval.code = BIT_XOR_EXPR; break;
1846 case LSHIFT:
1847 yylval.code = LSHIFT_EXPR; break;
1848 case RSHIFT:
1849 yylval.code = RSHIFT_EXPR; break;
1850 case '<':
1851 yylval.code = LT_EXPR; break;
1852 case '>':
1853 yylval.code = GT_EXPR; break;
1854 }
1855
1856 token_buffer[1] = c1 = getc (finput);
1857 token_buffer[2] = 0;
1858
1859 if (c1 == '=')
1860 {
1861 switch (c)
1862 {
1863 case '<':
1864 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
1865 case '>':
1866 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
1867 case '!':
1868 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
1869 case '=':
1870 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
1871 }
1872 value = ASSIGN; goto done;
1873 }
1874 else if (c == c1)
1875 switch (c)
1876 {
1877 case '+':
1878 value = PLUSPLUS; goto done;
1879 case '-':
1880 value = MINUSMINUS; goto done;
1881 case '&':
1882 value = ANDAND; goto done;
1883 case '|':
1884 value = OROR; goto done;
1885 case '<':
1886 c = LSHIFT;
1887 goto combine;
1888 case '>':
1889 c = RSHIFT;
1890 goto combine;
1891 }
1892 else if ((c == '-') && (c1 == '>'))
1893 { value = POINTSAT; goto done; }
1894 ungetc (c1, finput);
1895 token_buffer[1] = 0;
1896
1897 if ((c == '<') || (c == '>'))
1898 value = ARITHCOMPARE;
1899 else value = c;
1900 goto done;
1901 }
1902
1903 case 0:
1904 /* Don't make yyparse think this is eof. */
1905 value = 1;
1906 break;
1907
1908 default:
1909 value = c;
1910 }
1911
1912 done:
1913 /* yylloc.last_line = lineno; */
1914
1915 return value;
1916 }
1917
1918 /* Sets the value of the 'yydebug' variable to VALUE.
1919 This is a function so we don't have to have YYDEBUG defined
1920 in order to build the compiler. */
1921
1922 void
1923 set_yydebug (value)
1924 int value;
1925 {
1926 #if YYDEBUG != 0
1927 yydebug = value;
1928 #else
1929 warning ("YYDEBUG not defined.");
1930 #endif
1931 }
This page took 0.123114 seconds and 5 git commands to generate.