]> gcc.gnu.org Git - gcc.git/blob - gcc/c-lex.c
(readescape): DEL isn't printable.
[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 -1 if it is backslash-newline. */
798
799 static int
800 readescape ()
801 {
802 register int c = getc (finput);
803 register int code;
804 register unsigned count;
805 unsigned firstdig;
806
807 switch (c)
808 {
809 case 'x':
810 if (warn_traditional)
811 warning ("the meaning of `\\x' varies with -traditional");
812
813 if (flag_traditional)
814 return c;
815
816 code = 0;
817 count = 0;
818 while (1)
819 {
820 c = getc (finput);
821 if (!(c >= 'a' && c <= 'f')
822 && !(c >= 'A' && c <= 'F')
823 && !(c >= '0' && c <= '9'))
824 {
825 ungetc (c, finput);
826 break;
827 }
828 code *= 16;
829 if (c >= 'a' && c <= 'f')
830 code += c - 'a' + 10;
831 if (c >= 'A' && c <= 'F')
832 code += c - 'A' + 10;
833 if (c >= '0' && c <= '9')
834 code += c - '0';
835 if (count == 0)
836 firstdig = code;
837 count++;
838 }
839 if (count == 0)
840 error ("\\x used with no following hex digits");
841 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
842 || (count > 1
843 && ((1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
844 <= firstdig)))
845 pedwarn ("hex escape out of range");
846 return code;
847
848 case '0': case '1': case '2': case '3': case '4':
849 case '5': case '6': case '7':
850 code = 0;
851 count = 0;
852 while ((c <= '7') && (c >= '0') && (count++ < 3))
853 {
854 code = (code * 8) + (c - '0');
855 c = getc (finput);
856 }
857 ungetc (c, finput);
858 return code;
859
860 case '\\': case '\'': case '"':
861 return c;
862
863 case '\n':
864 lineno++;
865 return -1;
866
867 case 'n':
868 return TARGET_NEWLINE;
869
870 case 't':
871 return TARGET_TAB;
872
873 case 'r':
874 return TARGET_CR;
875
876 case 'f':
877 return TARGET_FF;
878
879 case 'b':
880 return TARGET_BS;
881
882 case 'a':
883 if (warn_traditional)
884 warning ("the meaning of `\\a' varies with -traditional");
885
886 if (flag_traditional)
887 return c;
888 return TARGET_BELL;
889
890 case 'v':
891 #if 0 /* Vertical tab is present in common usage compilers. */
892 if (flag_traditional)
893 return c;
894 #endif
895 return TARGET_VT;
896
897 case 'E':
898 pedwarn ("non-ANSI-standard escape sequence, `\\E'");
899 return 033;
900
901 case '?':
902 return c;
903
904 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
905 case '(':
906 case '{':
907 case '[':
908 if (pedantic)
909 pedwarn ("non-ANSI escape sequence `\\%c'", c);
910 return c;
911 }
912 if (c >= 040 && c < 0177)
913 pedwarn ("unknown escape sequence `\\%c'", c);
914 else
915 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
916 return c;
917 }
918 \f
919 void
920 yyerror (string)
921 char *string;
922 {
923 char buf[200];
924
925 strcpy (buf, string);
926
927 /* We can't print string and character constants well
928 because the token_buffer contains the result of processing escapes. */
929 if (end_of_file)
930 strcat (buf, " at end of input");
931 else if (token_buffer[0] == 0)
932 strcat (buf, " at null character");
933 else if (token_buffer[0] == '"')
934 strcat (buf, " before string constant");
935 else if (token_buffer[0] == '\'')
936 strcat (buf, " before character constant");
937 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
938 sprintf (buf + strlen (buf), " before character 0%o",
939 (unsigned char) token_buffer[0]);
940 else
941 strcat (buf, " before `%s'");
942
943 error (buf, token_buffer);
944 }
945
946 #if 0
947
948 struct try_type
949 {
950 tree *node_var;
951 char unsigned_flag;
952 char long_flag;
953 char long_long_flag;
954 };
955
956 struct try_type type_sequence[] =
957 {
958 { &integer_type_node, 0, 0, 0},
959 { &unsigned_type_node, 1, 0, 0},
960 { &long_integer_type_node, 0, 1, 0},
961 { &long_unsigned_type_node, 1, 1, 0},
962 { &long_long_integer_type_node, 0, 1, 1},
963 { &long_long_unsigned_type_node, 1, 1, 1}
964 };
965 #endif /* 0 */
966 \f
967 int
968 yylex ()
969 {
970 register int c;
971 register char *p;
972 register int value;
973 int wide_flag = 0;
974
975 if (nextchar >= 0)
976 c = nextchar, nextchar = -1;
977 else
978 c = getc (finput);
979
980 /* Effectively do c = skip_white_space (c)
981 but do it faster in the usual cases. */
982 while (1)
983 switch (c)
984 {
985 case ' ':
986 case '\t':
987 case '\f':
988 case '\v':
989 case '\b':
990 c = getc (finput);
991 break;
992
993 case '\r':
994 /* Call skip_white_space so we can warn if appropriate. */
995
996 case '\n':
997 case '/':
998 case '\\':
999 c = skip_white_space (c);
1000 default:
1001 goto found_nonwhite;
1002 }
1003 found_nonwhite:
1004
1005 token_buffer[0] = c;
1006 token_buffer[1] = 0;
1007
1008 /* yylloc.first_line = lineno; */
1009
1010 switch (c)
1011 {
1012 case EOF:
1013 end_of_file = 1;
1014 token_buffer[0] = 0;
1015 value = ENDFILE;
1016 break;
1017
1018 case '$':
1019 if (dollars_in_ident)
1020 goto letter;
1021 return '$';
1022
1023 case 'L':
1024 /* Capital L may start a wide-string or wide-character constant. */
1025 {
1026 register int c = getc (finput);
1027 if (c == '\'')
1028 {
1029 wide_flag = 1;
1030 goto char_constant;
1031 }
1032 if (c == '"')
1033 {
1034 wide_flag = 1;
1035 goto string_constant;
1036 }
1037 ungetc (c, finput);
1038 }
1039 goto letter;
1040
1041 case '@':
1042 if (!doing_objc_thang)
1043 {
1044 value = c;
1045 break;
1046 }
1047 p = token_buffer;
1048 *p++ = '@';
1049 c = getc (finput);
1050 while (isalnum (c) || c == '_')
1051 {
1052 if (p >= token_buffer + maxtoken)
1053 p = extend_token_buffer (p);
1054
1055 *p++ = c;
1056 c = getc (finput);
1057 }
1058
1059 *p = 0;
1060 nextchar = c;
1061 value = recognize_objc_keyword (token_buffer + 1);
1062 if (value != 0)
1063 break;
1064 error ("invalid Objective C keyword `%s'", token_buffer);
1065 /* Cause a syntax error--1 is not a valid token type. */
1066 value = 1;
1067 break;
1068
1069 case 'A': case 'B': case 'C': case 'D': case 'E':
1070 case 'F': case 'G': case 'H': case 'I': case 'J':
1071 case 'K': case 'M': case 'N': case 'O':
1072 case 'P': case 'Q': case 'R': case 'S': case 'T':
1073 case 'U': case 'V': case 'W': case 'X': case 'Y':
1074 case 'Z':
1075 case 'a': case 'b': case 'c': case 'd': case 'e':
1076 case 'f': case 'g': case 'h': case 'i': case 'j':
1077 case 'k': case 'l': case 'm': case 'n': case 'o':
1078 case 'p': case 'q': case 'r': case 's': case 't':
1079 case 'u': case 'v': case 'w': case 'x': case 'y':
1080 case 'z':
1081 case '_':
1082 letter:
1083 p = token_buffer;
1084 while (isalnum (c) || c == '_' || c == '$' || c == '@')
1085 {
1086 if (p >= token_buffer + maxtoken)
1087 p = extend_token_buffer (p);
1088 if (c == '$' && ! dollars_in_ident)
1089 break;
1090
1091 *p++ = c;
1092 c = getc (finput);
1093 }
1094
1095 *p = 0;
1096 nextchar = c;
1097
1098 value = IDENTIFIER;
1099 yylval.itype = 0;
1100
1101 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1102
1103 {
1104 register struct resword *ptr;
1105
1106 if (ptr = is_reserved_word (token_buffer, p - token_buffer))
1107 {
1108 if (ptr->rid)
1109 yylval.ttype = ridpointers[(int) ptr->rid];
1110 value = (int) ptr->token;
1111
1112 /* Even if we decided to recognize asm, still perhaps warn. */
1113 if (pedantic
1114 && (value == ASM_KEYWORD || value == TYPEOF
1115 || ptr->rid == RID_INLINE)
1116 && token_buffer[0] != '_')
1117 pedwarn ("ANSI does not permit the keyword `%s'",
1118 token_buffer);
1119 }
1120 }
1121
1122 /* If we did not find a keyword, look for an identifier
1123 (or a typename). */
1124
1125 if (value == IDENTIFIER)
1126 {
1127 yylval.ttype = get_identifier (token_buffer);
1128 lastiddecl = lookup_name (yylval.ttype);
1129
1130 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1131 value = TYPENAME;
1132 /* A user-invisible read-only initialized variable
1133 should be replaced by its value.
1134 We handle only strings since that's the only case used in C. */
1135 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1136 && DECL_IGNORED_P (lastiddecl)
1137 && TREE_READONLY (lastiddecl)
1138 && DECL_INITIAL (lastiddecl) != 0
1139 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1140 {
1141 yylval.ttype = DECL_INITIAL (lastiddecl);
1142 value = STRING;
1143 }
1144 else if (doing_objc_thang)
1145 {
1146 tree objc_interface_decl = lookup_interface (yylval.ttype);
1147
1148 if (objc_interface_decl)
1149 {
1150 value = CLASSNAME;
1151 yylval.ttype = objc_interface_decl;
1152 }
1153 }
1154 }
1155
1156 break;
1157
1158 case '0': case '1': case '2': case '3': case '4':
1159 case '5': case '6': case '7': case '8': case '9':
1160 case '.':
1161 {
1162 int base = 10;
1163 int count = 0;
1164 int largest_digit = 0;
1165 int numdigits = 0;
1166 /* for multi-precision arithmetic,
1167 we actually store only HOST_BITS_PER_CHAR bits in each part.
1168 The number of parts is chosen so as to be sufficient to hold
1169 the enough bits to fit into the two HOST_WIDE_INTs that contain
1170 the integer value (this is always at least as many bits as are
1171 in a target `long long' value, but may be wider). */
1172 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1173 int parts[TOTAL_PARTS];
1174 int overflow = 0;
1175
1176 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag
1177 = NOT_FLOAT;
1178
1179 for (count = 0; count < TOTAL_PARTS; count++)
1180 parts[count] = 0;
1181
1182 p = token_buffer;
1183 *p++ = c;
1184
1185 if (c == '0')
1186 {
1187 *p++ = (c = getc (finput));
1188 if ((c == 'x') || (c == 'X'))
1189 {
1190 base = 16;
1191 *p++ = (c = getc (finput));
1192 }
1193 /* Leading 0 forces octal unless the 0 is the only digit. */
1194 else if (c >= '0' && c <= '9')
1195 {
1196 base = 8;
1197 numdigits++;
1198 }
1199 else
1200 numdigits++;
1201 }
1202
1203 /* Read all the digits-and-decimal-points. */
1204
1205 while (c == '.'
1206 || (isalnum (c) && (c != 'l') && (c != 'L')
1207 && (c != 'u') && (c != 'U')
1208 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1209 {
1210 if (c == '.')
1211 {
1212 if (base == 16)
1213 error ("floating constant may not be in radix 16");
1214 if (floatflag == AFTER_POINT)
1215 {
1216 error ("malformed floating constant");
1217 floatflag = TOO_MANY_POINTS;
1218 }
1219 else
1220 floatflag = AFTER_POINT;
1221
1222 base = 10;
1223 *p++ = c = getc (finput);
1224 /* Accept '.' as the start of a floating-point number
1225 only when it is followed by a digit.
1226 Otherwise, unread the following non-digit
1227 and use the '.' as a structural token. */
1228 if (p == token_buffer + 2 && !isdigit (c))
1229 {
1230 if (c == '.')
1231 {
1232 c = getc (finput);
1233 if (c == '.')
1234 {
1235 *p++ = c;
1236 *p = 0;
1237 return ELLIPSIS;
1238 }
1239 error ("parse error at `..'");
1240 }
1241 ungetc (c, finput);
1242 token_buffer[1] = 0;
1243 value = '.';
1244 goto done;
1245 }
1246 }
1247 else
1248 {
1249 /* It is not a decimal point.
1250 It should be a digit (perhaps a hex digit). */
1251
1252 if (isdigit (c))
1253 {
1254 c = c - '0';
1255 }
1256 else if (base <= 10)
1257 {
1258 if ((c&~040) == 'E')
1259 {
1260 base = 10;
1261 floatflag = AFTER_POINT;
1262 break; /* start of exponent */
1263 }
1264 error ("nondigits in number and not hexadecimal");
1265 c = 0;
1266 }
1267 else if (c >= 'a')
1268 {
1269 c = c - 'a' + 10;
1270 }
1271 else
1272 {
1273 c = c - 'A' + 10;
1274 }
1275 if (c >= largest_digit)
1276 largest_digit = c;
1277 numdigits++;
1278
1279 for (count = 0; count < TOTAL_PARTS; count++)
1280 {
1281 parts[count] *= base;
1282 if (count)
1283 {
1284 parts[count]
1285 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1286 parts[count-1]
1287 &= (1 << HOST_BITS_PER_CHAR) - 1;
1288 }
1289 else
1290 parts[0] += c;
1291 }
1292
1293 /* If the extra highest-order part ever gets anything in it,
1294 the number is certainly too big. */
1295 if (parts[TOTAL_PARTS - 1] != 0)
1296 overflow = 1;
1297
1298 if (p >= token_buffer + maxtoken - 3)
1299 p = extend_token_buffer (p);
1300 *p++ = (c = getc (finput));
1301 }
1302 }
1303
1304 if (numdigits == 0)
1305 error ("numeric constant with no digits");
1306
1307 if (largest_digit >= base)
1308 error ("numeric constant contains digits beyond the radix");
1309
1310 /* Remove terminating char from the token buffer and delimit the string */
1311 *--p = 0;
1312
1313 if (floatflag != NOT_FLOAT)
1314 {
1315 tree type = double_type_node;
1316 int garbage_chars = 0, exceeds_double = 0;
1317 REAL_VALUE_TYPE value;
1318 jmp_buf handler;
1319
1320 /* Read explicit exponent if any, and put it in tokenbuf. */
1321
1322 if ((c == 'e') || (c == 'E'))
1323 {
1324 if (p >= token_buffer + maxtoken - 3)
1325 p = extend_token_buffer (p);
1326 *p++ = c;
1327 c = getc (finput);
1328 if ((c == '+') || (c == '-'))
1329 {
1330 *p++ = c;
1331 c = getc (finput);
1332 }
1333 if (! isdigit (c))
1334 error ("floating constant exponent has no digits");
1335 while (isdigit (c))
1336 {
1337 if (p >= token_buffer + maxtoken - 3)
1338 p = extend_token_buffer (p);
1339 *p++ = c;
1340 c = getc (finput);
1341 }
1342 }
1343
1344 *p = 0;
1345 errno = 0;
1346
1347 /* Convert string to a double, checking for overflow. */
1348 if (setjmp (handler))
1349 {
1350 error ("floating constant out of range");
1351 value = dconst0;
1352 }
1353 else
1354 {
1355 set_float_handler (handler);
1356 value = REAL_VALUE_ATOF (token_buffer);
1357 set_float_handler (NULL_PTR);
1358 }
1359 #ifdef ERANGE
1360 if (errno == ERANGE && !flag_traditional && pedantic)
1361 {
1362 char *p1 = token_buffer;
1363 /* Check for "0.0" and variants;
1364 SunOS 4 spuriously returns ERANGE for them. */
1365 while (*p1 == '0') p1++;
1366 if (*p1 == '.')
1367 {
1368 p1++;
1369 while (*p1 == '0') p1++;
1370 }
1371 if (*p1 == 'e' || *p1 == 'E')
1372 {
1373 /* with significand==0, ignore the exponent */
1374 p1++;
1375 while (*p1 != 0) p1++;
1376 }
1377 /* ERANGE is also reported for underflow,
1378 so test the value to distinguish overflow from that. */
1379 if (*p1 != 0 && (value > 1.0 || value < -1.0))
1380 {
1381 pedwarn ("floating point number exceeds range of `double'");
1382 exceeds_double = 1;
1383 }
1384 }
1385 #endif
1386
1387 /* Read the suffixes to choose a data type. */
1388 switch (c)
1389 {
1390 case 'f': case 'F':
1391 type = float_type_node;
1392 value = REAL_VALUE_TRUNCATE (TYPE_MODE (type), value);
1393 if (REAL_VALUE_ISINF (value) && ! exceeds_double && pedantic)
1394 pedwarn ("floating point number exceeds range of `float'");
1395 garbage_chars = -1;
1396 break;
1397
1398 case 'l': case 'L':
1399 type = long_double_type_node;
1400 garbage_chars = -1;
1401 break;
1402 }
1403 /* Note: garbage_chars is -1 if first char is *not* garbage. */
1404 while (isalnum (c))
1405 {
1406 if (p >= token_buffer + maxtoken - 3)
1407 p = extend_token_buffer (p);
1408 *p++ = c;
1409 c = getc (finput);
1410 garbage_chars++;
1411 }
1412 if (garbage_chars > 0)
1413 error ("garbage at end of number");
1414
1415 /* Create a node with determined type and value. */
1416 yylval.ttype = build_real (type, value);
1417
1418 ungetc (c, finput);
1419 *p = 0;
1420 }
1421 else
1422 {
1423 tree traditional_type, ansi_type, type;
1424 HOST_WIDE_INT high, low;
1425 int spec_unsigned = 0;
1426 int spec_long = 0;
1427 int spec_long_long = 0;
1428 int bytes, warn, i;
1429
1430 while (1)
1431 {
1432 if (c == 'u' || c == 'U')
1433 {
1434 if (spec_unsigned)
1435 error ("two `u's in integer constant");
1436 spec_unsigned = 1;
1437 }
1438 else if (c == 'l' || c == 'L')
1439 {
1440 if (spec_long)
1441 {
1442 if (spec_long_long)
1443 error ("three `l's in integer constant");
1444 else if (pedantic)
1445 pedwarn ("ANSI C forbids long long integer constants");
1446 spec_long_long = 1;
1447 }
1448 spec_long = 1;
1449 }
1450 else
1451 {
1452 if (isalnum (c))
1453 {
1454 error ("garbage at end of number");
1455 while (isalnum (c))
1456 {
1457 if (p >= token_buffer + maxtoken - 3)
1458 p = extend_token_buffer (p);
1459 *p++ = c;
1460 c = getc (finput);
1461 }
1462 }
1463 break;
1464 }
1465 if (p >= token_buffer + maxtoken - 3)
1466 p = extend_token_buffer (p);
1467 *p++ = c;
1468 c = getc (finput);
1469 }
1470
1471 ungetc (c, finput);
1472
1473 /* If the constant is not long long and it won't fit in an
1474 unsigned long, or if the constant is long long and won't fit
1475 in an unsigned long long, then warn that the constant is out
1476 of range. */
1477
1478 /* ??? This assumes that long long and long integer types are
1479 a multiple of 8 bits. This better than the original code
1480 though which assumed that long was exactly 32 bits and long
1481 long was exactly 64 bits. */
1482
1483 if (spec_long_long)
1484 bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1485 else
1486 bytes = TYPE_PRECISION (long_integer_type_node) / 8;
1487
1488 warn = overflow;
1489 for (i = bytes; i < TOTAL_PARTS; i++)
1490 if (parts[i])
1491 warn = 1;
1492 if (warn)
1493 pedwarn ("integer constant out of range");
1494
1495 /* This is simplified by the fact that our constant
1496 is always positive. */
1497
1498 high = low = 0;
1499
1500 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1501 {
1502 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1503 / HOST_BITS_PER_CHAR)]
1504 << (i * HOST_BITS_PER_CHAR));
1505 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1506 }
1507
1508 yylval.ttype = build_int_2 (low, high);
1509
1510 /* If warn_traditional, calculate both the ANSI type and the
1511 traditional type, then see if they disagree.
1512 Otherwise, calculate only the type for the dialect in use. */
1513 if (warn_traditional || flag_traditional)
1514 {
1515 /* Calculate the traditional type. */
1516 /* Traditionally, any constant is signed;
1517 but if unsigned is specified explicitly, obey that.
1518 Use the smallest size with the right number of bits,
1519 except for one special case with decimal constants. */
1520 if (! spec_long && base != 10
1521 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1522 traditional_type = (spec_unsigned ? unsigned_type_node
1523 : integer_type_node);
1524 /* A decimal constant must be long
1525 if it does not fit in type int.
1526 I think this is independent of whether
1527 the constant is signed. */
1528 else if (! spec_long && base == 10
1529 && int_fits_type_p (yylval.ttype, integer_type_node))
1530 traditional_type = (spec_unsigned ? unsigned_type_node
1531 : integer_type_node);
1532 else if (! spec_long_long)
1533 traditional_type = (spec_unsigned ? long_unsigned_type_node
1534 : long_integer_type_node);
1535 else
1536 traditional_type = (spec_unsigned
1537 ? long_long_unsigned_type_node
1538 : long_long_integer_type_node);
1539 }
1540 if (warn_traditional || ! flag_traditional)
1541 {
1542 /* Calculate the ANSI type. */
1543 if (! spec_long && ! spec_unsigned
1544 && int_fits_type_p (yylval.ttype, integer_type_node))
1545 ansi_type = integer_type_node;
1546 else if (! spec_long && (base != 10 || spec_unsigned)
1547 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1548 ansi_type = unsigned_type_node;
1549 else if (! spec_unsigned && !spec_long_long
1550 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1551 ansi_type = long_integer_type_node;
1552 else if (! spec_long_long)
1553 ansi_type = long_unsigned_type_node;
1554 else if (! spec_unsigned
1555 && int_fits_type_p (yylval.ttype,
1556 long_long_integer_type_node))
1557 ansi_type = long_long_integer_type_node;
1558 else
1559 ansi_type = long_long_unsigned_type_node;
1560 }
1561
1562 type = flag_traditional ? traditional_type : ansi_type;
1563
1564 if (warn_traditional && traditional_type != ansi_type)
1565 {
1566 if (TYPE_PRECISION (traditional_type)
1567 != TYPE_PRECISION (ansi_type))
1568 warning ("width of integer constant changes with -traditional");
1569 else if (TREE_UNSIGNED (traditional_type)
1570 != TREE_UNSIGNED (ansi_type))
1571 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1572 else
1573 warning ("width of integer constant may change on other systems with -traditional");
1574 }
1575
1576 if (!flag_traditional && !int_fits_type_p (yylval.ttype, type)
1577 && !warn)
1578 pedwarn ("integer constant out of range");
1579
1580 if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1581 /* The traditional constant 0x80000000 is signed
1582 but doesn't fit in the range of int.
1583 This will change it to -0x80000000, which does fit. */
1584 {
1585 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1586 yylval.ttype = convert (type, yylval.ttype);
1587 }
1588 else
1589 TREE_TYPE (yylval.ttype) = type;
1590
1591 *p = 0;
1592 }
1593
1594 value = CONSTANT; break;
1595 }
1596
1597 case '\'':
1598 char_constant:
1599 {
1600 register int result = 0;
1601 register int num_chars = 0;
1602 unsigned width = TYPE_PRECISION (char_type_node);
1603 int max_chars;
1604
1605 if (wide_flag)
1606 {
1607 width = WCHAR_TYPE_SIZE;
1608 #ifdef MULTIBYTE_CHARS
1609 max_chars = MB_CUR_MAX;
1610 #else
1611 max_chars = 1;
1612 #endif
1613 }
1614 else
1615 max_chars = TYPE_PRECISION (integer_type_node) / width;
1616
1617 while (1)
1618 {
1619 tryagain:
1620
1621 c = getc (finput);
1622
1623 if (c == '\'' || c == EOF)
1624 break;
1625
1626 if (c == '\\')
1627 {
1628 c = readescape ();
1629 if (c < 0)
1630 goto tryagain;
1631 if (width < HOST_BITS_PER_INT
1632 && (unsigned) c >= (1 << width))
1633 pedwarn ("escape sequence out of range for character");
1634 }
1635 else if (c == '\n')
1636 {
1637 if (pedantic)
1638 pedwarn ("ANSI C forbids newline in character constant");
1639 lineno++;
1640 }
1641
1642 num_chars++;
1643 if (num_chars > maxtoken - 4)
1644 extend_token_buffer (token_buffer);
1645
1646 token_buffer[num_chars] = c;
1647
1648 /* Merge character into result; ignore excess chars. */
1649 if (num_chars < max_chars + 1)
1650 {
1651 if (width < HOST_BITS_PER_INT)
1652 result = (result << width) | (c & ((1 << width) - 1));
1653 else
1654 result = c;
1655 }
1656 }
1657
1658 token_buffer[num_chars + 1] = '\'';
1659 token_buffer[num_chars + 2] = 0;
1660
1661 if (c != '\'')
1662 error ("malformatted character constant");
1663 else if (num_chars == 0)
1664 error ("empty character constant");
1665 else if (num_chars > max_chars)
1666 {
1667 num_chars = max_chars;
1668 error ("character constant too long");
1669 }
1670 else if (num_chars != 1 && ! flag_traditional)
1671 warning ("multi-character character constant");
1672
1673 /* If char type is signed, sign-extend the constant. */
1674 if (! wide_flag)
1675 {
1676 int num_bits = num_chars * width;
1677 if (TREE_UNSIGNED (char_type_node)
1678 || ((result >> (num_bits - 1)) & 1) == 0)
1679 yylval.ttype
1680 = build_int_2 (result & ((unsigned HOST_WIDE_INT) ~0
1681 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1682 0);
1683 else
1684 yylval.ttype
1685 = build_int_2 (result | ~((unsigned HOST_WIDE_INT) ~0
1686 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1687 -1);
1688 }
1689 else
1690 {
1691 #ifdef MULTIBYTE_CHARS
1692 /* Set the initial shift state and convert the next sequence. */
1693 result = 0;
1694 /* In all locales L'\0' is zero and mbtowc will return zero,
1695 so don't use it. */
1696 if (num_chars > 1
1697 || (num_chars == 1 && token_buffer[1] != '\0'))
1698 {
1699 wchar_t wc;
1700 (void) mbtowc (NULL_PTR, NULL_PTR, 0);
1701 if (mbtowc (& wc, token_buffer + 1, num_chars) == num_chars)
1702 result = wc;
1703 else
1704 warning ("Ignoring invalid multibyte character");
1705 }
1706 #endif
1707 yylval.ttype = build_int_2 (result, 0);
1708 }
1709
1710 TREE_TYPE (yylval.ttype) = integer_type_node;
1711 value = CONSTANT;
1712 break;
1713 }
1714
1715 case '"':
1716 string_constant:
1717 {
1718 c = getc (finput);
1719 p = token_buffer + 1;
1720
1721 while (c != '"' && c >= 0)
1722 {
1723 /* ignore_escape_flag is set for reading the filename in #line. */
1724 if (!ignore_escape_flag && c == '\\')
1725 {
1726 c = readescape ();
1727 if (c < 0)
1728 goto skipnewline;
1729 if (!wide_flag
1730 && TYPE_PRECISION (char_type_node) < HOST_BITS_PER_INT
1731 && c >= (1 << TYPE_PRECISION (char_type_node)))
1732 pedwarn ("escape sequence out of range for character");
1733 }
1734 else if (c == '\n')
1735 {
1736 if (pedantic)
1737 pedwarn ("ANSI C forbids newline in string constant");
1738 lineno++;
1739 }
1740
1741 if (p == token_buffer + maxtoken)
1742 p = extend_token_buffer (p);
1743 *p++ = c;
1744
1745 skipnewline:
1746 c = getc (finput);
1747 }
1748 *p = 0;
1749
1750 /* We have read the entire constant.
1751 Construct a STRING_CST for the result. */
1752
1753 if (wide_flag)
1754 {
1755 /* If this is a L"..." wide-string, convert the multibyte string
1756 to a wide character string. */
1757 char *widep = (char *) alloca ((p - token_buffer) * WCHAR_BYTES);
1758 int len;
1759
1760 #ifdef MULTIBYTE_CHARS
1761 len = mbstowcs ((wchar_t *) widep, token_buffer + 1, p - token_buffer);
1762 if ((unsigned) len >= (p - token_buffer))
1763 {
1764 warning ("Ignoring invalid multibyte string");
1765 len = 0;
1766 }
1767 bzero (widep + (len * WCHAR_BYTES), WCHAR_BYTES);
1768 #else
1769 {
1770 union { long l; char c[sizeof (long)]; } u;
1771 int big_endian;
1772 char *wp, *cp;
1773
1774 /* Determine whether host is little or big endian. */
1775 u.l = 1;
1776 big_endian = u.c[sizeof (long) - 1];
1777 wp = widep + (big_endian ? WCHAR_BYTES - 1 : 0);
1778
1779 bzero (widep, (p - token_buffer) * WCHAR_BYTES);
1780 for (cp = token_buffer + 1; cp < p; cp++)
1781 *wp = *cp, wp += WCHAR_BYTES;
1782 len = p - token_buffer - 1;
1783 }
1784 #endif
1785 yylval.ttype = build_string ((len + 1) * WCHAR_BYTES, widep);
1786 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
1787 }
1788 else
1789 {
1790 yylval.ttype = build_string (p - token_buffer, token_buffer + 1);
1791 TREE_TYPE (yylval.ttype) = char_array_type_node;
1792 }
1793
1794 *p++ = '"';
1795 *p = 0;
1796
1797 value = STRING; break;
1798 }
1799
1800 case '+':
1801 case '-':
1802 case '&':
1803 case '|':
1804 case '<':
1805 case '>':
1806 case '*':
1807 case '/':
1808 case '%':
1809 case '^':
1810 case '!':
1811 case '=':
1812 {
1813 register int c1;
1814
1815 combine:
1816
1817 switch (c)
1818 {
1819 case '+':
1820 yylval.code = PLUS_EXPR; break;
1821 case '-':
1822 yylval.code = MINUS_EXPR; break;
1823 case '&':
1824 yylval.code = BIT_AND_EXPR; break;
1825 case '|':
1826 yylval.code = BIT_IOR_EXPR; break;
1827 case '*':
1828 yylval.code = MULT_EXPR; break;
1829 case '/':
1830 yylval.code = TRUNC_DIV_EXPR; break;
1831 case '%':
1832 yylval.code = TRUNC_MOD_EXPR; break;
1833 case '^':
1834 yylval.code = BIT_XOR_EXPR; break;
1835 case LSHIFT:
1836 yylval.code = LSHIFT_EXPR; break;
1837 case RSHIFT:
1838 yylval.code = RSHIFT_EXPR; break;
1839 case '<':
1840 yylval.code = LT_EXPR; break;
1841 case '>':
1842 yylval.code = GT_EXPR; break;
1843 }
1844
1845 token_buffer[1] = c1 = getc (finput);
1846 token_buffer[2] = 0;
1847
1848 if (c1 == '=')
1849 {
1850 switch (c)
1851 {
1852 case '<':
1853 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
1854 case '>':
1855 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
1856 case '!':
1857 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
1858 case '=':
1859 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
1860 }
1861 value = ASSIGN; goto done;
1862 }
1863 else if (c == c1)
1864 switch (c)
1865 {
1866 case '+':
1867 value = PLUSPLUS; goto done;
1868 case '-':
1869 value = MINUSMINUS; goto done;
1870 case '&':
1871 value = ANDAND; goto done;
1872 case '|':
1873 value = OROR; goto done;
1874 case '<':
1875 c = LSHIFT;
1876 goto combine;
1877 case '>':
1878 c = RSHIFT;
1879 goto combine;
1880 }
1881 else if ((c == '-') && (c1 == '>'))
1882 { value = POINTSAT; goto done; }
1883 ungetc (c1, finput);
1884 token_buffer[1] = 0;
1885
1886 if ((c == '<') || (c == '>'))
1887 value = ARITHCOMPARE;
1888 else value = c;
1889 goto done;
1890 }
1891
1892 case 0:
1893 /* Don't make yyparse think this is eof. */
1894 value = 1;
1895 break;
1896
1897 default:
1898 value = c;
1899 }
1900
1901 done:
1902 /* yylloc.last_line = lineno; */
1903
1904 return value;
1905 }
1906
1907 /* Sets the value of the 'yydebug' variable to VALUE.
1908 This is a function so we don't have to have YYDEBUG defined
1909 in order to build the compiler. */
1910
1911 void
1912 set_yydebug (value)
1913 int value;
1914 {
1915 #if YYDEBUG != 0
1916 yydebug = value;
1917 #else
1918 warning ("YYDEBUG not defined.");
1919 #endif
1920 }
This page took 0.129711 seconds and 6 git commands to generate.