?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
m68k: Consolidate NPTL/non versions of vfork
[glibc.git] / stdio-common / vfscanf.c
blobe0d224530cc67498c3fa5b37838d5bdcdf08eb28
1 /* Copyright (C) 1991-2014 Free Software Foundation, Inc.
2 This file is part of the GNU C Library.
3
4 The GNU C Library is free software; you can redistribute it and/or
5 modify it under the terms of the GNU Lesser General Public
6 License as published by the Free Software Foundation; either
7 version 2.1 of the License, or (at your option) any later version.
8
9 The GNU C Library is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 Lesser General Public License for more details.
13
14 You should have received a copy of the GNU Lesser General Public
15 License along with the GNU C Library; if not, see
16 <http://www.gnu.org/licenses/>. */
17
18 #include <assert.h>
19 #include <errno.h>
20 #include <limits.h>
21 #include <ctype.h>
22 #include <stdarg.h>
23 #include <stdbool.h>
24 #include <stdio.h>
25 #include <stdint.h>
26 #include <stdlib.h>
27 #include <string.h>
28 #include <wchar.h>
29 #include <wctype.h>
30 #include <bits/libc-lock.h>
31 #include <locale/localeinfo.h>
32
33 #ifdef __GNUC__
34 # define HAVE_LONGLONG
35 # define LONGLONG long long
36 #else
37 # define LONGLONG long
38 #endif
39
40 /* Determine whether we have to handle `long long' at all. */
41 #if LONG_MAX == LONG_LONG_MAX
42 # define need_longlong 0
43 #else
44 # define need_longlong 1
45 #endif
46
47 /* Determine whether we have to handle `long'. */
48 #if INT_MAX == LONG_MAX
49 # define need_long 0
50 #else
51 # define need_long 1
52 #endif
53
54 /* Those are flags in the conversion format. */
55 #define LONG 0x0001 /* l: long or double */
56 #define LONGDBL 0x0002 /* L: long long or long double */
57 #define SHORT 0x0004 /* h: short */
58 #define SUPPRESS 0x0008 /* *: suppress assignment */
59 #define POINTER 0x0010 /* weird %p pointer (`fake hex') */
60 #define NOSKIP 0x0020 /* do not skip blanks */
61 #define NUMBER_SIGNED 0x0040 /* signed integer */
62 #define GROUP 0x0080 /* ': group numbers */
63 #define GNU_MALLOC 0x0100 /* a: malloc strings */
64 #define CHAR 0x0200 /* hh: char */
65 #define I18N 0x0400 /* I: use locale's digits */
66 #define HEXA_FLOAT 0x0800 /* hexadecimal float */
67 #define READ_POINTER 0x1000 /* this is a pointer value */
68 #define POSIX_MALLOC 0x2000 /* m: malloc strings */
69 #define MALLOC (GNU_MALLOC | POSIX_MALLOC)
70
71 #include <locale/localeinfo.h>
72 #include <libioP.h>
73 #include <libio.h>
74
75 #undef va_list
76 #define va_list _IO_va_list
77
78 #ifdef COMPILE_WSCANF
79 # define ungetc(c, s) ((void) (c == WEOF \
80 || (--read_in, \
81 _IO_sputbackwc (s, c))))
82 # define ungetc_not_eof(c, s) ((void) (--read_in, \
83 _IO_sputbackwc (s, c)))
84 # define inchar() (c == WEOF ? ((errno = inchar_errno), WEOF) \
85 : ((c = _IO_getwc_unlocked (s)), \
86 (void) (c != WEOF \
87 ? ++read_in \
88 : (size_t) (inchar_errno = errno)), c))
89
90 # define MEMCPY(d, s, n) __wmemcpy (d, s, n)
91 # define ISSPACE(Ch) iswspace (Ch)
92 # define ISDIGIT(Ch) iswdigit (Ch)
93 # define ISXDIGIT(Ch) iswxdigit (Ch)
94 # define TOLOWER(Ch) towlower (Ch)
95 # define ORIENT if (_IO_fwide (s, 1) != 1) return WEOF
96 # define __strtoll_internal __wcstoll_internal
97 # define __strtoull_internal __wcstoull_internal
98 # define __strtol_internal __wcstol_internal
99 # define __strtoul_internal __wcstoul_internal
100 # define __strtold_internal __wcstold_internal
101 # define __strtod_internal __wcstod_internal
102 # define __strtof_internal __wcstof_internal
103
104 # define L_(Str) L##Str
105 # define CHAR_T wchar_t
106 # define UCHAR_T unsigned int
107 # define WINT_T wint_t
108 # undef EOF
109 # define EOF WEOF
110 #else
111 # define ungetc(c, s) ((void) ((int) c == EOF \
112 || (--read_in, \
113 _IO_sputbackc (s, (unsigned char) c))))
114 # define ungetc_not_eof(c, s) ((void) (--read_in, \
115 _IO_sputbackc (s, (unsigned char) c)))
116 # define inchar() (c == EOF ? ((errno = inchar_errno), EOF) \
117 : ((c = _IO_getc_unlocked (s)), \
118 (void) (c != EOF \
119 ? ++read_in \
120 : (size_t) (inchar_errno = errno)), c))
121 # define MEMCPY(d, s, n) memcpy (d, s, n)
122 # define ISSPACE(Ch) __isspace_l (Ch, loc)
123 # define ISDIGIT(Ch) __isdigit_l (Ch, loc)
124 # define ISXDIGIT(Ch) __isxdigit_l (Ch, loc)
125 # define TOLOWER(Ch) __tolower_l ((unsigned char) (Ch), loc)
126 # define ORIENT if (_IO_vtable_offset (s) == 0 \
127 && _IO_fwide (s, -1) != -1) \
128 return EOF
129
130 # define L_(Str) Str
131 # define CHAR_T char
132 # define UCHAR_T unsigned char
133 # define WINT_T int
134 #endif
135
136 #define encode_error() do { \
137 errval = 4; \
138 __set_errno (EILSEQ); \
139 goto errout; \
140 } while (0)
141 #define conv_error() do { \
142 errval = 2; \
143 goto errout; \
144 } while (0)
145 #define input_error() do { \
146 errval = 1; \
147 if (done == 0) done = EOF; \
148 goto errout; \
149 } while (0)
150 #define add_ptr_to_free(ptr) \
151 do \
152 { \
153 if (ptrs_to_free == NULL \
154 || ptrs_to_free->count == (sizeof (ptrs_to_free->ptrs) \
155 / sizeof (ptrs_to_free->ptrs[0]))) \
156 { \
157 struct ptrs_to_free *new_ptrs = alloca (sizeof (*ptrs_to_free)); \
158 new_ptrs->count = 0; \
159 new_ptrs->next = ptrs_to_free; \
160 ptrs_to_free = new_ptrs; \
161 } \
162 ptrs_to_free->ptrs[ptrs_to_free->count++] = (ptr); \
163 } \
164 while (0)
165 #define ARGCHECK(s, format) \
166 do \
167 { \
168 /* Check file argument for consistence. */ \
169 CHECK_FILE (s, EOF); \
170 if (s->_flags & _IO_NO_READS) \
171 { \
172 __set_errno (EBADF); \
173 return EOF; \
174 } \
175 else if (format == NULL) \
176 { \
177 MAYBE_SET_EINVAL; \
178 return EOF; \
179 } \
180 } while (0)
181 #define LOCK_STREAM(S) \
182 __libc_cleanup_region_start (1, (void (*) (void *)) &_IO_funlockfile, (S)); \
183 _IO_flockfile (S)
184 #define UNLOCK_STREAM(S) \
185 _IO_funlockfile (S); \
186 __libc_cleanup_region_end (0)
187
188 struct ptrs_to_free
189 {
190 size_t count;
191 struct ptrs_to_free *next;
192 char **ptrs[32];
193 };
194
195 /* Read formatted input from S according to the format string
196 FORMAT, using the argument list in ARG.
197 Return the number of assignments made, or -1 for an input error. */
198 #ifdef COMPILE_WSCANF
199 int
200 _IO_vfwscanf (_IO_FILE *s, const wchar_t *format, _IO_va_list argptr,
201 int *errp)
202 #else
203 int
204 _IO_vfscanf_internal (_IO_FILE *s, const char *format, _IO_va_list argptr,
205 int *errp)
206 #endif
207 {
208 va_list arg;
209 const CHAR_T *f = format;
210 UCHAR_T fc; /* Current character of the format. */
211 WINT_T done = 0; /* Assignments done. */
212 size_t read_in = 0; /* Chars read in. */
213 WINT_T c = 0; /* Last char read. */
214 int width; /* Maximum field width. */
215 int flags; /* Modifiers for current format element. */
216 int errval = 0;
217 #ifndef COMPILE_WSCANF
218 __locale_t loc = _NL_CURRENT_LOCALE;
219 struct __locale_data *const curctype = loc->__locales[LC_CTYPE];
220 #endif
221
222 /* Errno of last failed inchar call. */
223 int inchar_errno = 0;
224 /* Status for reading F-P nums. */
225 char got_digit, got_dot, got_e, negative;
226 /* If a [...] is a [^...]. */
227 CHAR_T not_in;
228 #define exp_char not_in
229 /* Base for integral numbers. */
230 int base;
231 /* Decimal point character. */
232 #ifdef COMPILE_WSCANF
233 wint_t decimal;
234 #else
235 const char *decimal;
236 #endif
237 /* The thousands character of the current locale. */
238 #ifdef COMPILE_WSCANF
239 wint_t thousands;
240 #else
241 const char *thousands;
242 #endif
243 struct ptrs_to_free *ptrs_to_free = NULL;
244 /* State for the conversions. */
245 mbstate_t state;
246 /* Integral holding variables. */
247 union
248 {
249 long long int q;
250 unsigned long long int uq;
251 long int l;
252 unsigned long int ul;
253 } num;
254 /* Character-buffer pointer. */
255 char *str = NULL;
256 wchar_t *wstr = NULL;
257 char **strptr = NULL;
258 ssize_t strsize = 0;
259 /* We must not react on white spaces immediately because they can
260 possibly be matched even if in the input stream no character is
261 available anymore. */
262 int skip_space = 0;
263 /* Workspace. */
264 CHAR_T *tw; /* Temporary pointer. */
265 CHAR_T *wp = NULL; /* Workspace. */
266 size_t wpmax = 0; /* Maximal size of workspace. */
267 size_t wpsize; /* Currently used bytes in workspace. */
268 bool use_malloc = false;
269 #define ADDW(Ch) \
270 do \
271 { \
272 if (__glibc_unlikely (wpsize == wpmax)) \
273 { \
274 CHAR_T *old = wp; \
275 size_t newsize = (UCHAR_MAX + 1 > 2 * wpmax \
276 ? UCHAR_MAX + 1 : 2 * wpmax); \
277 if (use_malloc || !__libc_use_alloca (newsize)) \
278 { \
279 wp = realloc (use_malloc ? wp : NULL, newsize); \
280 if (wp == NULL) \
281 { \
282 if (use_malloc) \
283 free (old); \
284 done = EOF; \
285 goto errout; \
286 } \
287 if (! use_malloc) \
288 MEMCPY (wp, old, wpsize); \
289 wpmax = newsize; \
290 use_malloc = true; \
291 } \
292 else \
293 { \
294 size_t s = wpmax * sizeof (CHAR_T); \
295 wp = (CHAR_T *) extend_alloca (wp, s, \
296 newsize * sizeof (CHAR_T)); \
297 wpmax = s / sizeof (CHAR_T); \
298 if (old != NULL) \
299 MEMCPY (wp, old, wpsize); \
300 } \
301 } \
302 wp[wpsize++] = (Ch); \
303 } \
304 while (0)
305
306 #ifdef __va_copy
307 __va_copy (arg, argptr);
308 #else
309 arg = (va_list) argptr;
310 #endif
311
312 #ifdef ORIENT
313 ORIENT;
314 #endif
315
316 ARGCHECK (s, format);
317
318 {
319 #ifndef COMPILE_WSCANF
320 struct __locale_data *const curnumeric = loc->__locales[LC_NUMERIC];
321 #endif
322
323 /* Figure out the decimal point character. */
324 #ifdef COMPILE_WSCANF
325 decimal = _NL_CURRENT_WORD (LC_NUMERIC, _NL_NUMERIC_DECIMAL_POINT_WC);
326 #else
327 decimal = curnumeric->values[_NL_ITEM_INDEX (DECIMAL_POINT)].string;
328 #endif
329 /* Figure out the thousands separator character. */
330 #ifdef COMPILE_WSCANF
331 thousands = _NL_CURRENT_WORD (LC_NUMERIC, _NL_NUMERIC_THOUSANDS_SEP_WC);
332 #else
333 thousands = curnumeric->values[_NL_ITEM_INDEX (THOUSANDS_SEP)].string;
334 if (*thousands == '\0')
335 thousands = NULL;
336 #endif
337 }
338
339 /* Lock the stream. */
340 LOCK_STREAM (s);
341
342
343 #ifndef COMPILE_WSCANF
344 /* From now on we use `state' to convert the format string. */
345 memset (&state, '\0', sizeof (state));
346 #endif
347
348 /* Run through the format string. */
349 while (*f != '\0')
350 {
351 unsigned int argpos;
352 /* Extract the next argument, which is of type TYPE.
353 For a %N$... spec, this is the Nth argument from the beginning;
354 otherwise it is the next argument after the state now in ARG. */
355 #ifdef __va_copy
356 # define ARG(type) (argpos == 0 ? va_arg (arg, type) : \
357 ({ unsigned int pos = argpos; \
358 va_list arg; \
359 __va_copy (arg, argptr); \
360 while (--pos > 0) \
361 (void) va_arg (arg, void *); \
362 va_arg (arg, type); \
363 }))
364 #else
365 # if 0
366 /* XXX Possible optimization. */
367 # define ARG(type) (argpos == 0 ? va_arg (arg, type) : \
368 ({ va_list arg = (va_list) argptr; \
369 arg = (va_list) ((char *) arg \
370 + (argpos - 1) \
371 * __va_rounded_size (void *)); \
372 va_arg (arg, type); \
373 }))
374 # else
375 # define ARG(type) (argpos == 0 ? va_arg (arg, type) : \
376 ({ unsigned int pos = argpos; \
377 va_list arg = (va_list) argptr; \
378 while (--pos > 0) \
379 (void) va_arg (arg, void *); \
380 va_arg (arg, type); \
381 }))
382 # endif
383 #endif
384
385 #ifndef COMPILE_WSCANF
386 if (!isascii ((unsigned char) *f))
387 {
388 /* Non-ASCII, may be a multibyte. */
389 int len = __mbrlen (f, strlen (f), &state);
390 if (len > 0)
391 {
392 do
393 {
394 c = inchar ();
395 if (__glibc_unlikely (c == EOF))
396 input_error ();
397 else if (c != (unsigned char) *f++)
398 {
399 ungetc_not_eof (c, s);
400 conv_error ();
401 }
402 }
403 while (--len > 0);
404 continue;
405 }
406 }
407 #endif
408
409 fc = *f++;
410 if (fc != '%')
411 {
412 /* Remember to skip spaces. */
413 if (ISSPACE (fc))
414 {
415 skip_space = 1;
416 continue;
417 }
418
419 /* Read a character. */
420 c = inchar ();
421
422 /* Characters other than format specs must just match. */
423 if (__glibc_unlikely (c == EOF))
424 input_error ();
425
426 /* We saw white space char as the last character in the format
427 string. Now it's time to skip all leading white space. */
428 if (skip_space)
429 {
430 while (ISSPACE (c))
431 if (__glibc_unlikely (inchar () == EOF))
432 input_error ();
433 skip_space = 0;
434 }
435
436 if (__glibc_unlikely (c != fc))
437 {
438 ungetc (c, s);
439 conv_error ();
440 }
441
442 continue;
443 }
444
445 /* This is the start of the conversion string. */
446 flags = 0;
447
448 /* Initialize state of modifiers. */
449 argpos = 0;
450
451 /* Prepare temporary buffer. */
452 wpsize = 0;
453
454 /* Check for a positional parameter specification. */
455 if (ISDIGIT ((UCHAR_T) *f))
456 {
457 argpos = (UCHAR_T) *f++ - L_('0');
458 while (ISDIGIT ((UCHAR_T) *f))
459 argpos = argpos * 10 + ((UCHAR_T) *f++ - L_('0'));
460 if (*f == L_('$'))
461 ++f;
462 else
463 {
464 /* Oops; that was actually the field width. */
465 width = argpos;
466 argpos = 0;
467 goto got_width;
468 }
469 }
470
471 /* Check for the assignment-suppressing, the number grouping flag,
472 and the signal to use the locale's digit representation. */
473 while (*f == L_('*') || *f == L_('\'') || *f == L_('I'))
474 switch (*f++)
475 {
476 case L_('*'):
477 flags |= SUPPRESS;
478 break;
479 case L_('\''):
480 #ifdef COMPILE_WSCANF
481 if (thousands != L'\0')
482 #else
483 if (thousands != NULL)
484 #endif
485 flags |= GROUP;
486 break;
487 case L_('I'):
488 flags |= I18N;
489 break;
490 }
491
492 /* Find the maximum field width. */
493 width = 0;
494 while (ISDIGIT ((UCHAR_T) *f))
495 {
496 width *= 10;
497 width += (UCHAR_T) *f++ - L_('0');
498 }
499 got_width:
500 if (width == 0)
501 width = -1;
502
503 /* Check for type modifiers. */
504 switch (*f++)
505 {
506 case L_('h'):
507 /* ints are short ints or chars. */
508 if (*f == L_('h'))
509 {
510 ++f;
511 flags |= CHAR;
512 }
513 else
514 flags |= SHORT;
515 break;
516 case L_('l'):
517 if (*f == L_('l'))
518 {
519 /* A double `l' is equivalent to an `L'. */
520 ++f;
521 flags |= LONGDBL | LONG;
522 }
523 else
524 /* ints are long ints. */
525 flags |= LONG;
526 break;
527 case L_('q'):
528 case L_('L'):
529 /* doubles are long doubles, and ints are long long ints. */
530 flags |= LONGDBL | LONG;
531 break;
532 case L_('a'):
533 /* The `a' is used as a flag only if followed by `s', `S' or
534 `['. */
535 if (*f != L_('s') && *f != L_('S') && *f != L_('['))
536 {
537 --f;
538 break;
539 }
540 /* In __isoc99_*scanf %as, %aS and %a[ extension is not
541 supported at all. */
542 if (s->_flags2 & _IO_FLAGS2_SCANF_STD)
543 {
544 --f;
545 break;
546 }
547 /* String conversions (%s, %[) take a `char **'
548 arg and fill it in with a malloc'd pointer. */
549 flags |= GNU_MALLOC;
550 break;
551 case L_('m'):
552 flags |= POSIX_MALLOC;
553 if (*f == L_('l'))
554 {
555 ++f;
556 flags |= LONG;
557 }
558 break;
559 case L_('z'):
560 if (need_longlong && sizeof (size_t) > sizeof (unsigned long int))
561 flags |= LONGDBL;
562 else if (sizeof (size_t) > sizeof (unsigned int))
563 flags |= LONG;
564 break;
565 case L_('j'):
566 if (need_longlong && sizeof (uintmax_t) > sizeof (unsigned long int))
567 flags |= LONGDBL;
568 else if (sizeof (uintmax_t) > sizeof (unsigned int))
569 flags |= LONG;
570 break;
571 case L_('t'):
572 if (need_longlong && sizeof (ptrdiff_t) > sizeof (long int))
573 flags |= LONGDBL;
574 else if (sizeof (ptrdiff_t) > sizeof (int))
575 flags |= LONG;
576 break;
577 default:
578 /* Not a recognized modifier. Backup. */
579 --f;
580 break;
581 }
582
583 /* End of the format string? */
584 if (__glibc_unlikely (*f == L_('\0')))
585 conv_error ();
586
587 /* Find the conversion specifier. */
588 fc = *f++;
589 if (skip_space || (fc != L_('[') && fc != L_('c')
590 && fc != L_('C') && fc != L_('n')))
591 {
592 /* Eat whitespace. */
593 int save_errno = errno;
594 __set_errno (0);
595 do
596 /* We add the additional test for EOF here since otherwise
597 inchar will restore the old errno value which might be
598 EINTR but does not indicate an interrupt since nothing
599 was read at this time. */
600 if (__builtin_expect ((c == EOF || inchar () == EOF)
601 && errno == EINTR, 0))
602 input_error ();
603 while (ISSPACE (c));
604 __set_errno (save_errno);
605 ungetc (c, s);
606 skip_space = 0;
607 }
608
609 switch (fc)
610 {
611 case L_('%'): /* Must match a literal '%'. */
612 c = inchar ();
613 if (__glibc_unlikely (c == EOF))
614 input_error ();
615 if (__glibc_unlikely (c != fc))
616 {
617 ungetc_not_eof (c, s);
618 conv_error ();
619 }
620 break;
621
622 case L_('n'): /* Answer number of assignments done. */
623 /* Corrigendum 1 to ISO C 1990 describes the allowed flags
624 with the 'n' conversion specifier. */
625 if (!(flags & SUPPRESS))
626 {
627 /* Don't count the read-ahead. */
628 if (need_longlong && (flags & LONGDBL))
629 *ARG (long long int *) = read_in;
630 else if (need_long && (flags & LONG))
631 *ARG (long int *) = read_in;
632 else if (flags & SHORT)
633 *ARG (short int *) = read_in;
634 else if (!(flags & CHAR))
635 *ARG (int *) = read_in;
636 else
637 *ARG (char *) = read_in;
638
639 #ifdef NO_BUG_IN_ISO_C_CORRIGENDUM_1
640 /* We have a severe problem here. The ISO C standard
641 contradicts itself in explaining the effect of the %n
642 format in `scanf'. While in ISO C:1990 and the ISO C
643 Amendement 1:1995 the result is described as
644
645 Execution of a %n directive does not effect the
646 assignment count returned at the completion of
647 execution of the f(w)scanf function.
648
649 in ISO C Corrigendum 1:1994 the following was added:
650
651 Subclause 7.9.6.2
652 Add the following fourth example:
653 In:
654 #include <stdio.h>
655 int d1, d2, n1, n2, i;
656 i = sscanf("123", "%d%n%n%d", &d1, &n1, &n2, &d2);
657 the value 123 is assigned to d1 and the value3 to n1.
658 Because %n can never get an input failure the value
659 of 3 is also assigned to n2. The value of d2 is not
660 affected. The value 3 is assigned to i.
661
662 We go for now with the historically correct code from ISO C,
663 i.e., we don't count the %n assignments. When it ever
664 should proof to be wrong just remove the #ifdef above. */
665 ++done;
666 #endif
667 }
668 break;
669
670 case L_('c'): /* Match characters. */
671 if ((flags & LONG) == 0)
672 {
673 if (width == -1)
674 width = 1;
675
676 #define STRING_ARG(Str, Type, Width) \
677 do if (!(flags & SUPPRESS)) \
678 { \
679 if (flags & MALLOC) \
680 { \
681 /* The string is to be stored in a malloc'd buffer. */ \
682 /* For %mS using char ** is actually wrong, but \
683 shouldn't make a difference on any arch glibc \
684 supports and would unnecessarily complicate \
685 things. */ \
686 strptr = ARG (char **); \
687 if (strptr == NULL) \
688 conv_error (); \
689 /* Allocate an initial buffer. */ \
690 strsize = Width; \
691 *strptr = (char *) malloc (strsize * sizeof (Type)); \
692 Str = (Type *) *strptr; \
693 if (Str != NULL) \
694 add_ptr_to_free (strptr); \
695 else if (flags & POSIX_MALLOC) \
696 { \
697 done = EOF; \
698 goto errout; \
699 } \
700 } \
701 else \
702 Str = ARG (Type *); \
703 if (Str == NULL) \
704 conv_error (); \
705 } while (0)
706 #ifdef COMPILE_WSCANF
707 STRING_ARG (str, char, 100);
708 #else
709 STRING_ARG (str, char, (width > 1024 ? 1024 : width));
710 #endif
711
712 c = inchar ();
713 if (__glibc_unlikely (c == EOF))
714 input_error ();
715
716 #ifdef COMPILE_WSCANF
717 /* We have to convert the wide character(s) into multibyte
718 characters and store the result. */
719 memset (&state, '\0', sizeof (state));
720
721 do
722 {
723 size_t n;
724
725 if (!(flags & SUPPRESS) && (flags & POSIX_MALLOC)
726 && str + MB_CUR_MAX >= *strptr + strsize)
727 {
728 /* We have to enlarge the buffer if the `m' flag
729 was given. */
730 size_t strleng = str - *strptr;
731 char *newstr;
732
733 newstr = (char *) realloc (*strptr, strsize * 2);
734 if (newstr == NULL)
735 {
736 /* Can't allocate that much. Last-ditch effort. */
737 newstr = (char *) realloc (*strptr,
738 strleng + MB_CUR_MAX);
739 if (newstr == NULL)
740 {
741 /* c can't have `a' flag, only `m'. */
742 done = EOF;
743 goto errout;
744 }
745 else
746 {
747 *strptr = newstr;
748 str = newstr + strleng;
749 strsize = strleng + MB_CUR_MAX;
750 }
751 }
752 else
753 {
754 *strptr = newstr;
755 str = newstr + strleng;
756 strsize *= 2;
757 }
758 }
759
760 n = __wcrtomb (!(flags & SUPPRESS) ? str : NULL, c, &state);
761 if (__glibc_unlikely (n == (size_t) -1))
762 /* No valid wide character. */
763 input_error ();
764
765 /* Increment the output pointer. Even if we don't
766 write anything. */
767 str += n;
768 }
769 while (--width > 0 && inchar () != EOF);
770 #else
771 if (!(flags & SUPPRESS))
772 {
773 do
774 {
775 if ((flags & MALLOC)
776 && (char *) str == *strptr + strsize)
777 {
778 /* Enlarge the buffer. */
779 size_t newsize
780 = strsize
781 + (strsize >= width ? width - 1 : strsize);
782
783 str = (char *) realloc (*strptr, newsize);
784 if (str == NULL)
785 {
786 /* Can't allocate that much. Last-ditch
787 effort. */
788 str = (char *) realloc (*strptr, strsize + 1);
789 if (str == NULL)
790 {
791 /* c can't have `a' flag, only `m'. */
792 done = EOF;
793 goto errout;
794 }
795 else
796 {
797 *strptr = (char *) str;
798 str += strsize;
799 ++strsize;
800 }
801 }
802 else
803 {
804 *strptr = (char *) str;
805 str += strsize;
806 strsize = newsize;
807 }
808 }
809 *str++ = c;
810 }
811 while (--width > 0 && inchar () != EOF);
812 }
813 else
814 while (--width > 0 && inchar () != EOF);
815 #endif
816
817 if (!(flags & SUPPRESS))
818 {
819 if ((flags & MALLOC) && str - *strptr != strsize)
820 {
821 char *cp = (char *) realloc (*strptr, str - *strptr);
822 if (cp != NULL)
823 *strptr = cp;
824 }
825 strptr = NULL;
826 ++done;
827 }
828
829 break;
830 }
831 /* FALLTHROUGH */
832 case L_('C'):
833 if (width == -1)
834 width = 1;
835
836 STRING_ARG (wstr, wchar_t, (width > 1024 ? 1024 : width));
837
838 c = inchar ();
839 if (__glibc_unlikely (c == EOF))
840 input_error ();
841
842 #ifdef COMPILE_WSCANF
843 /* Just store the incoming wide characters. */
844 if (!(flags & SUPPRESS))
845 {
846 do
847 {
848 if ((flags & MALLOC)
849 && wstr == (wchar_t *) *strptr + strsize)
850 {
851 size_t newsize
852 = strsize + (strsize > width ? width - 1 : strsize);
853 /* Enlarge the buffer. */
854 wstr = (wchar_t *) realloc (*strptr,
855 newsize * sizeof (wchar_t));
856 if (wstr == NULL)
857 {
858 /* Can't allocate that much. Last-ditch effort. */
859 wstr = (wchar_t *) realloc (*strptr,
860 (strsize + 1)
861 * sizeof (wchar_t));
862 if (wstr == NULL)
863 {
864 /* C or lc can't have `a' flag, only `m'
865 flag. */
866 done = EOF;
867 goto errout;
868 }
869 else
870 {
871 *strptr = (char *) wstr;
872 wstr += strsize;
873 ++strsize;
874 }
875 }
876 else
877 {
878 *strptr = (char *) wstr;
879 wstr += strsize;
880 strsize = newsize;
881 }
882 }
883 *wstr++ = c;
884 }
885 while (--width > 0 && inchar () != EOF);
886 }
887 else
888 while (--width > 0 && inchar () != EOF);
889 #else
890 {
891 /* We have to convert the multibyte input sequence to wide
892 characters. */
893 char buf[1];
894 mbstate_t cstate;
895
896 memset (&cstate, '\0', sizeof (cstate));
897
898 do
899 {
900 /* This is what we present the mbrtowc function first. */
901 buf[0] = c;
902
903 if (!(flags & SUPPRESS) && (flags & MALLOC)
904 && wstr == (wchar_t *) *strptr + strsize)
905 {
906 size_t newsize
907 = strsize + (strsize > width ? width - 1 : strsize);
908 /* Enlarge the buffer. */
909 wstr = (wchar_t *) realloc (*strptr,
910 newsize * sizeof (wchar_t));
911 if (wstr == NULL)
912 {
913 /* Can't allocate that much. Last-ditch effort. */
914 wstr = (wchar_t *) realloc (*strptr,
915 ((strsize + 1)
916 * sizeof (wchar_t)));
917 if (wstr == NULL)
918 {
919 /* C or lc can't have `a' flag, only `m' flag. */
920 done = EOF;
921 goto errout;
922 }
923 else
924 {
925 *strptr = (char *) wstr;
926 wstr += strsize;
927 ++strsize;
928 }
929 }
930 else
931 {
932 *strptr = (char *) wstr;
933 wstr += strsize;
934 strsize = newsize;
935 }
936 }
937
938 while (1)
939 {
940 size_t n;
941
942 n = __mbrtowc (!(flags & SUPPRESS) ? wstr : NULL,
943 buf, 1, &cstate);
944
945 if (n == (size_t) -2)
946 {
947 /* Possibly correct character, just not enough
948 input. */
949 if (__glibc_unlikely (inchar () == EOF))
950 encode_error ();
951
952 buf[0] = c;
953 continue;
954 }
955
956 if (__glibc_unlikely (n != 1))
957 encode_error ();
958
959 /* We have a match. */
960 break;
961 }
962
963 /* Advance the result pointer. */
964 ++wstr;
965 }
966 while (--width > 0 && inchar () != EOF);
967 }
968 #endif
969
970 if (!(flags & SUPPRESS))
971 {
972 if ((flags & MALLOC) && wstr - (wchar_t *) *strptr != strsize)
973 {
974 wchar_t *cp = (wchar_t *) realloc (*strptr,
975 ((wstr
976 - (wchar_t *) *strptr)
977 * sizeof (wchar_t)));
978 if (cp != NULL)
979 *strptr = (char *) cp;
980 }
981 strptr = NULL;
982
983 ++done;
984 }
985
986 break;
987
988 case L_('s'): /* Read a string. */
989 if (!(flags & LONG))
990 {
991 STRING_ARG (str, char, 100);
992
993 c = inchar ();
994 if (__glibc_unlikely (c == EOF))
995 input_error ();
996
997 #ifdef COMPILE_WSCANF
998 memset (&state, '\0', sizeof (state));
999 #endif
1000
1001 do
1002 {
1003 if (ISSPACE (c))
1004 {
1005 ungetc_not_eof (c, s);
1006 break;
1007 }
1008
1009 #ifdef COMPILE_WSCANF
1010 /* This is quite complicated. We have to convert the
1011 wide characters into multibyte characters and then
1012 store them. */
1013 {
1014 size_t n;
1015
1016 if (!(flags & SUPPRESS) && (flags & MALLOC)
1017 && str + MB_CUR_MAX >= *strptr + strsize)
1018 {
1019 /* We have to enlarge the buffer if the `a' or `m'
1020 flag was given. */
1021 size_t strleng = str - *strptr;
1022 char *newstr;
1023
1024 newstr = (char *) realloc (*strptr, strsize * 2);
1025 if (newstr == NULL)
1026 {
1027 /* Can't allocate that much. Last-ditch
1028 effort. */
1029 newstr = (char *) realloc (*strptr,
1030 strleng + MB_CUR_MAX);
1031 if (newstr == NULL)
1032 {
1033 if (flags & POSIX_MALLOC)
1034 {
1035 done = EOF;
1036 goto errout;
1037 }
1038 /* We lose. Oh well. Terminate the
1039 string and stop converting,
1040 so at least we don't skip any input. */
1041 ((char *) (*strptr))[strleng] = '\0';
1042 strptr = NULL;
1043 ++done;
1044 conv_error ();
1045 }
1046 else
1047 {
1048 *strptr = newstr;
1049 str = newstr + strleng;
1050 strsize = strleng + MB_CUR_MAX;
1051 }
1052 }
1053 else
1054 {
1055 *strptr = newstr;
1056 str = newstr + strleng;
1057 strsize *= 2;
1058 }
1059 }
1060
1061 n = __wcrtomb (!(flags & SUPPRESS) ? str : NULL, c,
1062 &state);
1063 if (__glibc_unlikely (n == (size_t) -1))
1064 encode_error ();
1065
1066 assert (n <= MB_CUR_MAX);
1067 str += n;
1068 }
1069 #else
1070 /* This is easy. */
1071 if (!(flags & SUPPRESS))
1072 {
1073 *str++ = c;
1074 if ((flags & MALLOC)
1075 && (char *) str == *strptr + strsize)
1076 {
1077 /* Enlarge the buffer. */
1078 str = (char *) realloc (*strptr, 2 * strsize);
1079 if (str == NULL)
1080 {
1081 /* Can't allocate that much. Last-ditch
1082 effort. */
1083 str = (char *) realloc (*strptr, strsize + 1);
1084 if (str == NULL)
1085 {
1086 if (flags & POSIX_MALLOC)
1087 {
1088 done = EOF;
1089 goto errout;
1090 }
1091 /* We lose. Oh well. Terminate the
1092 string and stop converting,
1093 so at least we don't skip any input. */
1094 ((char *) (*strptr))[strsize - 1] = '\0';
1095 strptr = NULL;
1096 ++done;
1097 conv_error ();
1098 }
1099 else
1100 {
1101 *strptr = (char *) str;
1102 str += strsize;
1103 ++strsize;
1104 }
1105 }
1106 else
1107 {
1108 *strptr = (char *) str;
1109 str += strsize;
1110 strsize *= 2;
1111 }
1112 }
1113 }
1114 #endif
1115 }
1116 while ((width <= 0 || --width > 0) && inchar () != EOF);
1117
1118 if (!(flags & SUPPRESS))
1119 {
1120 #ifdef COMPILE_WSCANF
1121 /* We have to emit the code to get into the initial
1122 state. */
1123 char buf[MB_LEN_MAX];
1124 size_t n = __wcrtomb (buf, L'\0', &state);
1125 if (n > 0 && (flags & MALLOC)
1126 && str + n >= *strptr + strsize)
1127 {
1128 /* Enlarge the buffer. */
1129 size_t strleng = str - *strptr;
1130 char *newstr;
1131
1132 newstr = (char *) realloc (*strptr, strleng + n + 1);
1133 if (newstr == NULL)
1134 {
1135 if (flags & POSIX_MALLOC)
1136 {
1137 done = EOF;
1138 goto errout;
1139 }
1140 /* We lose. Oh well. Terminate the string
1141 and stop converting, so at least we don't
1142 skip any input. */
1143 ((char *) (*strptr))[strleng] = '\0';
1144 strptr = NULL;
1145 ++done;
1146 conv_error ();
1147 }
1148 else
1149 {
1150 *strptr = newstr;
1151 str = newstr + strleng;
1152 strsize = strleng + n + 1;
1153 }
1154 }
1155
1156 str = __mempcpy (str, buf, n);
1157 #endif
1158 *str++ = '\0';
1159
1160 if ((flags & MALLOC) && str - *strptr != strsize)
1161 {
1162 char *cp = (char *) realloc (*strptr, str - *strptr);
1163 if (cp != NULL)
1164 *strptr = cp;
1165 }
1166 strptr = NULL;
1167
1168 ++done;
1169 }
1170 break;
1171 }
1172 /* FALLTHROUGH */
1173
1174 case L_('S'):
1175 {
1176 #ifndef COMPILE_WSCANF
1177 mbstate_t cstate;
1178 #endif
1179
1180 /* Wide character string. */
1181 STRING_ARG (wstr, wchar_t, 100);
1182
1183 c = inchar ();
1184 if (__builtin_expect (c == EOF, 0))
1185 input_error ();
1186
1187 #ifndef COMPILE_WSCANF
1188 memset (&cstate, '\0', sizeof (cstate));
1189 #endif
1190
1191 do
1192 {
1193 if (ISSPACE (c))
1194 {
1195 ungetc_not_eof (c, s);
1196 break;
1197 }
1198
1199 #ifdef COMPILE_WSCANF
1200 /* This is easy. */
1201 if (!(flags & SUPPRESS))
1202 {
1203 *wstr++ = c;
1204 if ((flags & MALLOC)
1205 && wstr == (wchar_t *) *strptr + strsize)
1206 {
1207 /* Enlarge the buffer. */
1208 wstr = (wchar_t *) realloc (*strptr,
1209 (2 * strsize)
1210 * sizeof (wchar_t));
1211 if (wstr == NULL)
1212 {
1213 /* Can't allocate that much. Last-ditch
1214 effort. */
1215 wstr = (wchar_t *) realloc (*strptr,
1216 (strsize + 1)
1217 * sizeof (wchar_t));
1218 if (wstr == NULL)
1219 {
1220 if (flags & POSIX_MALLOC)
1221 {
1222 done = EOF;
1223 goto errout;
1224 }
1225 /* We lose. Oh well. Terminate the string
1226 and stop converting, so at least we don't
1227 skip any input. */
1228 ((wchar_t *) (*strptr))[strsize - 1] = L'\0';
1229 strptr = NULL;
1230 ++done;
1231 conv_error ();
1232 }
1233 else
1234 {
1235 *strptr = (char *) wstr;
1236 wstr += strsize;
1237 ++strsize;
1238 }
1239 }
1240 else
1241 {
1242 *strptr = (char *) wstr;
1243 wstr += strsize;
1244 strsize *= 2;
1245 }
1246 }
1247 }
1248 #else
1249 {
1250 char buf[1];
1251
1252 buf[0] = c;
1253
1254 while (1)
1255 {
1256 size_t n;
1257
1258 n = __mbrtowc (!(flags & SUPPRESS) ? wstr : NULL,
1259 buf, 1, &cstate);
1260
1261 if (n == (size_t) -2)
1262 {
1263 /* Possibly correct character, just not enough
1264 input. */
1265 if (__glibc_unlikely (inchar () == EOF))
1266 encode_error ();
1267
1268 buf[0] = c;
1269 continue;
1270 }
1271
1272 if (__glibc_unlikely (n != 1))
1273 encode_error ();
1274
1275 /* We have a match. */
1276 ++wstr;
1277 break;
1278 }
1279
1280 if (!(flags & SUPPRESS) && (flags & MALLOC)
1281 && wstr == (wchar_t *) *strptr + strsize)
1282 {
1283 /* Enlarge the buffer. */
1284 wstr = (wchar_t *) realloc (*strptr,
1285 (2 * strsize
1286 * sizeof (wchar_t)));
1287 if (wstr == NULL)
1288 {
1289 /* Can't allocate that much. Last-ditch effort. */
1290 wstr = (wchar_t *) realloc (*strptr,
1291 ((strsize + 1)
1292 * sizeof (wchar_t)));
1293 if (wstr == NULL)
1294 {
1295 if (flags & POSIX_MALLOC)
1296 {
1297 done = EOF;
1298 goto errout;
1299 }
1300 /* We lose. Oh well. Terminate the
1301 string and stop converting, so at
1302 least we don't skip any input. */
1303 ((wchar_t *) (*strptr))[strsize - 1] = L'\0';
1304 strptr = NULL;
1305 ++done;
1306 conv_error ();
1307 }
1308 else
1309 {
1310 *strptr = (char *) wstr;
1311 wstr += strsize;
1312 ++strsize;
1313 }
1314 }
1315 else
1316 {
1317 *strptr = (char *) wstr;
1318 wstr += strsize;
1319 strsize *= 2;
1320 }
1321 }
1322 }
1323 #endif
1324 }
1325 while ((width <= 0 || --width > 0) && inchar () != EOF);
1326
1327 if (!(flags & SUPPRESS))
1328 {
1329 *wstr++ = L'\0';
1330
1331 if ((flags & MALLOC) && wstr - (wchar_t *) *strptr != strsize)
1332 {
1333 wchar_t *cp = (wchar_t *) realloc (*strptr,
1334 ((wstr
1335 - (wchar_t *) *strptr)
1336 * sizeof(wchar_t)));
1337 if (cp != NULL)
1338 *strptr = (char *) cp;
1339 }
1340 strptr = NULL;
1341
1342 ++done;
1343 }
1344 }
1345 break;
1346
1347 case L_('x'): /* Hexadecimal integer. */
1348 case L_('X'): /* Ditto. */
1349 base = 16;
1350 goto number;
1351
1352 case L_('o'): /* Octal integer. */
1353 base = 8;
1354 goto number;
1355
1356 case L_('u'): /* Unsigned decimal integer. */
1357 base = 10;
1358 goto number;
1359
1360 case L_('d'): /* Signed decimal integer. */
1361 base = 10;
1362 flags |= NUMBER_SIGNED;
1363 goto number;
1364
1365 case L_('i'): /* Generic number. */
1366 base = 0;
1367 flags |= NUMBER_SIGNED;
1368
1369 number:
1370 c = inchar ();
1371 if (__glibc_unlikely (c == EOF))
1372 input_error ();
1373
1374 /* Check for a sign. */
1375 if (c == L_('-') || c == L_('+'))
1376 {
1377 ADDW (c);
1378 if (width > 0)
1379 --width;
1380 c = inchar ();
1381 }
1382
1383 /* Look for a leading indication of base. */
1384 if (width != 0 && c == L_('0'))
1385 {
1386 if (width > 0)
1387 --width;
1388
1389 ADDW (c);
1390 c = inchar ();
1391
1392 if (width != 0 && TOLOWER (c) == L_('x'))
1393 {
1394 if (base == 0)
1395 base = 16;
1396 if (base == 16)
1397 {
1398 if (width > 0)
1399 --width;
1400 c = inchar ();
1401 }
1402 }
1403 else if (base == 0)
1404 base = 8;
1405 }
1406
1407 if (base == 0)
1408 base = 10;
1409
1410 if (base == 10 && __builtin_expect ((flags & I18N) != 0, 0))
1411 {
1412 int from_level;
1413 int to_level;
1414 int level;
1415 #ifdef COMPILE_WSCANF
1416 const wchar_t *wcdigits[10];
1417 const wchar_t *wcdigits_extended[10];
1418 #else
1419 const char *mbdigits[10];
1420 const char *mbdigits_extended[10];
1421 #endif
1422 /* "to_inpunct" is a map from ASCII digits to their
1423 equivalent in locale. This is defined for locales
1424 which use an extra digits set. */
1425 wctrans_t map = __wctrans ("to_inpunct");
1426 int n;
1427
1428 from_level = 0;
1429 #ifdef COMPILE_WSCANF
1430 to_level = _NL_CURRENT_WORD (LC_CTYPE,
1431 _NL_CTYPE_INDIGITS_WC_LEN) - 1;
1432 #else
1433 to_level = (uint32_t) curctype->values[_NL_ITEM_INDEX (_NL_CTYPE_INDIGITS_MB_LEN)].word - 1;
1434 #endif
1435
1436 /* Get the alternative digit forms if there are any. */
1437 if (__glibc_unlikely (map != NULL))
1438 {
1439 /* Adding new level for extra digits set in locale file. */
1440 ++to_level;
1441
1442 for (n = 0; n < 10; ++n)
1443 {
1444 #ifdef COMPILE_WSCANF
1445 wcdigits[n] = (const wchar_t *)
1446 _NL_CURRENT (LC_CTYPE, _NL_CTYPE_INDIGITS0_WC + n);
1447
1448 wchar_t *wc_extended = (wchar_t *)
1449 alloca ((to_level + 2) * sizeof (wchar_t));
1450 __wmemcpy (wc_extended, wcdigits[n], to_level);
1451 wc_extended[to_level] = __towctrans (L'0' + n, map);
1452 wc_extended[to_level + 1] = '\0';
1453 wcdigits_extended[n] = wc_extended;
1454 #else
1455 mbdigits[n]
1456 = curctype->values[_NL_CTYPE_INDIGITS0_MB + n].string;
1457
1458 /* Get the equivalent wide char in map. */
1459 wint_t extra_wcdigit = __towctrans (L'0' + n, map);
1460
1461 /* Convert it to multibyte representation. */
1462 mbstate_t state;
1463 memset (&state, '\0', sizeof (state));
1464
1465 char extra_mbdigit[MB_LEN_MAX];
1466 size_t mblen
1467 = __wcrtomb (extra_mbdigit, extra_wcdigit, &state);
1468
1469 if (mblen == (size_t) -1)
1470 {
1471 /* Ignore this new level. */
1472 map = NULL;
1473 break;
1474 }
1475
1476 /* Calculate the length of mbdigits[n]. */
1477 const char *last_char = mbdigits[n];
1478 for (level = 0; level < to_level; ++level)
1479 last_char = strchr (last_char, '\0') + 1;
1480
1481 size_t mbdigits_len = last_char - mbdigits[n];
1482
1483 /* Allocate memory for extended multibyte digit. */
1484 char *mb_extended;
1485 mb_extended = (char *) alloca (mbdigits_len + mblen + 1);
1486
1487 /* And get the mbdigits + extra_digit string. */
1488 *(char *) __mempcpy (__mempcpy (mb_extended, mbdigits[n],
1489 mbdigits_len),
1490 extra_mbdigit, mblen) = '\0';
1491 mbdigits_extended[n] = mb_extended;
1492 #endif
1493 }
1494 }
1495
1496 /* Read the number into workspace. */
1497 while (c != EOF && width != 0)
1498 {
1499 /* In this round we get the pointer to the digit strings
1500 and also perform the first round of comparisons. */
1501 for (n = 0; n < 10; ++n)
1502 {
1503 /* Get the string for the digits with value N. */
1504 #ifdef COMPILE_WSCANF
1505 if (__glibc_unlikely (map != NULL))
1506 wcdigits[n] = wcdigits_extended[n];
1507 else
1508 wcdigits[n] = (const wchar_t *)
1509 _NL_CURRENT (LC_CTYPE, _NL_CTYPE_INDIGITS0_WC + n);
1510 wcdigits[n] += from_level;
1511
1512 if (c == (wint_t) *wcdigits[n])
1513 {
1514 to_level = from_level;
1515 break;
1516 }
1517
1518 /* Advance the pointer to the next string. */
1519 ++wcdigits[n];
1520 #else
1521 const char *cmpp;
1522 int avail = width > 0 ? width : INT_MAX;
1523
1524 if (__glibc_unlikely (map != NULL))
1525 mbdigits[n] = mbdigits_extended[n];
1526 else
1527 mbdigits[n]
1528 = curctype->values[_NL_CTYPE_INDIGITS0_MB + n].string;
1529
1530 for (level = 0; level < from_level; level++)
1531 mbdigits[n] = strchr (mbdigits[n], '\0') + 1;
1532
1533 cmpp = mbdigits[n];
1534 while ((unsigned char) *cmpp == c && avail >= 0)
1535 {
1536 if (*++cmpp == '\0')
1537 break;
1538 else
1539 {
1540 if (avail == 0 || inchar () == EOF)
1541 break;
1542 --avail;
1543 }
1544 }
1545
1546 if (*cmpp == '\0')
1547 {
1548 if (width > 0)
1549 width = avail;
1550 to_level = from_level;
1551 break;
1552 }
1553
1554 /* We are pushing all read characters back. */
1555 if (cmpp > mbdigits[n])
1556 {
1557 ungetc (c, s);
1558 while (--cmpp > mbdigits[n])
1559 ungetc_not_eof ((unsigned char) *cmpp, s);
1560 c = (unsigned char) *cmpp;
1561 }
1562
1563 /* Advance the pointer to the next string. */
1564 mbdigits[n] = strchr (mbdigits[n], '\0') + 1;
1565 #endif
1566 }
1567
1568 if (n == 10)
1569 {
1570 /* Have not yet found the digit. */
1571 for (level = from_level + 1; level <= to_level; ++level)
1572 {
1573 /* Search all ten digits of this level. */
1574 for (n = 0; n < 10; ++n)
1575 {
1576 #ifdef COMPILE_WSCANF
1577 if (c == (wint_t) *wcdigits[n])
1578 break;
1579
1580 /* Advance the pointer to the next string. */
1581 ++wcdigits[n];
1582 #else
1583 const char *cmpp;
1584 int avail = width > 0 ? width : INT_MAX;
1585
1586 cmpp = mbdigits[n];
1587 while ((unsigned char) *cmpp == c && avail >= 0)
1588 {
1589 if (*++cmpp == '\0')
1590 break;
1591 else
1592 {
1593 if (avail == 0 || inchar () == EOF)
1594 break;
1595 --avail;
1596 }
1597 }
1598
1599 if (*cmpp == '\0')
1600 {
1601 if (width > 0)
1602 width = avail;
1603 break;
1604 }
1605
1606 /* We are pushing all read characters back. */
1607 if (cmpp > mbdigits[n])
1608 {
1609 ungetc (c, s);
1610 while (--cmpp > mbdigits[n])
1611 ungetc_not_eof ((unsigned char) *cmpp, s);
1612 c = (unsigned char) *cmpp;
1613 }
1614
1615 /* Advance the pointer to the next string. */
1616 mbdigits[n] = strchr (mbdigits[n], '\0') + 1;
1617 #endif
1618 }
1619
1620 if (n < 10)
1621 {
1622 /* Found it. */
1623 from_level = level;
1624 to_level = level;
1625 break;
1626 }
1627 }
1628 }
1629
1630 if (n < 10)
1631 c = L_('0') + n;
1632 else if (flags & GROUP)
1633 {
1634 /* Try matching against the thousands separator. */
1635 #ifdef COMPILE_WSCANF
1636 if (c != thousands)
1637 break;
1638 #else
1639 const char *cmpp = thousands;
1640 int avail = width > 0 ? width : INT_MAX;
1641
1642 while ((unsigned char) *cmpp == c && avail >= 0)
1643 {
1644 ADDW (c);
1645 if (*++cmpp == '\0')
1646 break;
1647 else
1648 {
1649 if (avail == 0 || inchar () == EOF)
1650 break;
1651 --avail;
1652 }
1653 }
1654
1655 if (*cmpp != '\0')
1656 {
1657 /* We are pushing all read characters back. */
1658 if (cmpp > thousands)
1659 {
1660 wpsize -= cmpp - thousands;
1661 ungetc (c, s);
1662 while (--cmpp > thousands)
1663 ungetc_not_eof ((unsigned char) *cmpp, s);
1664 c = (unsigned char) *cmpp;
1665 }
1666 break;
1667 }
1668
1669 if (width > 0)
1670 width = avail;
1671
1672 /* The last thousands character will be added back by
1673 the ADDW below. */
1674 --wpsize;
1675 #endif
1676 }
1677 else
1678 break;
1679
1680 ADDW (c);
1681 if (width > 0)
1682 --width;
1683
1684 c = inchar ();
1685 }
1686 }
1687 else
1688 /* Read the number into workspace. */
1689 while (c != EOF && width != 0)
1690 {
1691 if (base == 16)
1692 {
1693 if (!ISXDIGIT (c))
1694 break;
1695 }
1696 else if (!ISDIGIT (c) || (int) (c - L_('0')) >= base)
1697 {
1698 if (base == 10 && (flags & GROUP))
1699 {
1700 /* Try matching against the thousands separator. */
1701 #ifdef COMPILE_WSCANF
1702 if (c != thousands)
1703 break;
1704 #else
1705 const char *cmpp = thousands;
1706 int avail = width > 0 ? width : INT_MAX;
1707
1708 while ((unsigned char) *cmpp == c && avail >= 0)
1709 {
1710 ADDW (c);
1711 if (*++cmpp == '\0')
1712 break;
1713 else
1714 {
1715 if (avail == 0 || inchar () == EOF)
1716 break;
1717 --avail;
1718 }
1719 }
1720
1721 if (*cmpp != '\0')
1722 {
1723 /* We are pushing all read characters back. */
1724 if (cmpp > thousands)
1725 {
1726 wpsize -= cmpp - thousands;
1727 ungetc (c, s);
1728 while (--cmpp > thousands)
1729 ungetc_not_eof ((unsigned char) *cmpp, s);
1730 c = (unsigned char) *cmpp;
1731 }
1732 break;
1733 }
1734
1735 if (width > 0)
1736 width = avail;
1737
1738 /* The last thousands character will be added back by
1739 the ADDW below. */
1740 --wpsize;
1741 #endif
1742 }
1743 else
1744 break;
1745 }
1746 ADDW (c);
1747 if (width > 0)
1748 --width;
1749
1750 c = inchar ();
1751 }
1752
1753 if (wpsize == 0
1754 || (wpsize == 1 && (wp[0] == L_('+') || wp[0] == L_('-'))))
1755 {
1756 /* There was no number. If we are supposed to read a pointer
1757 we must recognize "(nil)" as well. */
1758 if (__builtin_expect (wpsize == 0
1759 && (flags & READ_POINTER)
1760 && (width < 0 || width >= 5)
1761 && c == '('
1762 && TOLOWER (inchar ()) == L_('n')
1763 && TOLOWER (inchar ()) == L_('i')
1764 && TOLOWER (inchar ()) == L_('l')
1765 && inchar () == L_(')'), 1))
1766 /* We must produce the value of a NULL pointer. A single
1767 '0' digit is enough. */
1768 ADDW (L_('0'));
1769 else
1770 {
1771 /* The last read character is not part of the number
1772 anymore. */
1773 ungetc (c, s);
1774
1775 conv_error ();
1776 }
1777 }
1778 else
1779 /* The just read character is not part of the number anymore. */
1780 ungetc (c, s);
1781
1782 /* Convert the number. */
1783 ADDW (L_('\0'));
1784 if (need_longlong && (flags & LONGDBL))
1785 {
1786 if (flags & NUMBER_SIGNED)
1787 num.q = __strtoll_internal (wp, &tw, base, flags & GROUP);
1788 else
1789 num.uq = __strtoull_internal (wp, &tw, base, flags & GROUP);
1790 }
1791 else
1792 {
1793 if (flags & NUMBER_SIGNED)
1794 num.l = __strtol_internal (wp, &tw, base, flags & GROUP);
1795 else
1796 num.ul = __strtoul_internal (wp, &tw, base, flags & GROUP);
1797 }
1798 if (__glibc_unlikely (wp == tw))
1799 conv_error ();
1800
1801 if (!(flags & SUPPRESS))
1802 {
1803 if (flags & NUMBER_SIGNED)
1804 {
1805 if (need_longlong && (flags & LONGDBL))
1806 *ARG (LONGLONG int *) = num.q;
1807 else if (need_long && (flags & LONG))
1808 *ARG (long int *) = num.l;
1809 else if (flags & SHORT)
1810 *ARG (short int *) = (short int) num.l;
1811 else if (!(flags & CHAR))
1812 *ARG (int *) = (int) num.l;
1813 else
1814 *ARG (signed char *) = (signed char) num.ul;
1815 }
1816 else
1817 {
1818 if (need_longlong && (flags & LONGDBL))
1819 *ARG (unsigned LONGLONG int *) = num.uq;
1820 else if (need_long && (flags & LONG))
1821 *ARG (unsigned long int *) = num.ul;
1822 else if (flags & SHORT)
1823 *ARG (unsigned short int *)
1824 = (unsigned short int) num.ul;
1825 else if (!(flags & CHAR))
1826 *ARG (unsigned int *) = (unsigned int) num.ul;
1827 else
1828 *ARG (unsigned char *) = (unsigned char) num.ul;
1829 }
1830 ++done;
1831 }
1832 break;
1833
1834 case L_('e'): /* Floating-point numbers. */
1835 case L_('E'):
1836 case L_('f'):
1837 case L_('F'):
1838 case L_('g'):
1839 case L_('G'):
1840 case L_('a'):
1841 case L_('A'):
1842 c = inchar ();
1843 if (width > 0)
1844 --width;
1845 if (__glibc_unlikely (c == EOF))
1846 input_error ();
1847
1848 got_digit = got_dot = got_e = 0;
1849
1850 /* Check for a sign. */
1851 if (c == L_('-') || c == L_('+'))
1852 {
1853 negative = c == L_('-');
1854 if (__glibc_unlikely (width == 0 || inchar () == EOF))
1855 /* EOF is only an input error before we read any chars. */
1856 conv_error ();
1857 if (width > 0)
1858 --width;
1859 }
1860 else
1861 negative = 0;
1862
1863 /* Take care for the special arguments "nan" and "inf". */
1864 if (TOLOWER (c) == L_('n'))
1865 {
1866 /* Maybe "nan". */
1867 ADDW (c);
1868 if (__builtin_expect (width == 0
1869 || inchar () == EOF
1870 || TOLOWER (c) != L_('a'), 0))
1871 conv_error ();
1872 if (width > 0)
1873 --width;
1874 ADDW (c);
1875 if (__builtin_expect (width == 0
1876 || inchar () == EOF
1877 || TOLOWER (c) != L_('n'), 0))
1878 conv_error ();
1879 if (width > 0)
1880 --width;
1881 ADDW (c);
1882 /* It is "nan". */
1883 goto scan_float;
1884 }
1885 else if (TOLOWER (c) == L_('i'))
1886 {
1887 /* Maybe "inf" or "infinity". */
1888 ADDW (c);
1889 if (__builtin_expect (width == 0
1890 || inchar () == EOF
1891 || TOLOWER (c) != L_('n'), 0))
1892 conv_error ();
1893 if (width > 0)
1894 --width;
1895 ADDW (c);
1896 if (__builtin_expect (width == 0
1897 || inchar () == EOF
1898 || TOLOWER (c) != L_('f'), 0))
1899 conv_error ();
1900 if (width > 0)
1901 --width;
1902 ADDW (c);
1903 /* It is as least "inf". */
1904 if (width != 0 && inchar () != EOF)
1905 {
1906 if (TOLOWER (c) == L_('i'))
1907 {
1908 if (width > 0)
1909 --width;
1910 /* Now we have to read the rest as well. */
1911 ADDW (c);
1912 if (__builtin_expect (width == 0
1913 || inchar () == EOF
1914 || TOLOWER (c) != L_('n'), 0))
1915 conv_error ();
1916 if (width > 0)
1917 --width;
1918 ADDW (c);
1919 if (__builtin_expect (width == 0
1920 || inchar () == EOF
1921 || TOLOWER (c) != L_('i'), 0))
1922 conv_error ();
1923 if (width > 0)
1924 --width;
1925 ADDW (c);
1926 if (__builtin_expect (width == 0
1927 || inchar () == EOF
1928 || TOLOWER (c) != L_('t'), 0))
1929 conv_error ();
1930 if (width > 0)
1931 --width;
1932 ADDW (c);
1933 if (__builtin_expect (width == 0
1934 || inchar () == EOF
1935 || TOLOWER (c) != L_('y'), 0))
1936 conv_error ();
1937 if (width > 0)
1938 --width;
1939 ADDW (c);
1940 }
1941 else
1942 /* Never mind. */
1943 ungetc (c, s);
1944 }
1945 goto scan_float;
1946 }
1947
1948 exp_char = L_('e');
1949 if (width != 0 && c == L_('0'))
1950 {
1951 ADDW (c);
1952 c = inchar ();
1953 if (width > 0)
1954 --width;
1955 if (width != 0 && TOLOWER (c) == L_('x'))
1956 {
1957 /* It is a number in hexadecimal format. */
1958 ADDW (c);
1959
1960 flags |= HEXA_FLOAT;
1961 exp_char = L_('p');
1962
1963 /* Grouping is not allowed. */
1964 flags &= ~GROUP;
1965 c = inchar ();
1966 if (width > 0)
1967 --width;
1968 }
1969 else
1970 got_digit = 1;
1971 }
1972
1973 while (1)
1974 {
1975 if (ISDIGIT (c))
1976 {
1977 ADDW (c);
1978 got_digit = 1;
1979 }
1980 else if (!got_e && (flags & HEXA_FLOAT) && ISXDIGIT (c))
1981 {
1982 ADDW (c);
1983 got_digit = 1;
1984 }
1985 else if (got_e && wp[wpsize - 1] == exp_char
1986 && (c == L_('-') || c == L_('+')))
1987 ADDW (c);
1988 else if (got_digit && !got_e
1989 && (CHAR_T) TOLOWER (c) == exp_char)
1990 {
1991 ADDW (exp_char);
1992 got_e = got_dot = 1;
1993 }
1994 else
1995 {
1996 #ifdef COMPILE_WSCANF
1997 if (! got_dot && c == decimal)
1998 {
1999 ADDW (c);
2000 got_dot = 1;
2001 }
2002 else if ((flags & GROUP) != 0 && ! got_dot && c == thousands)
2003 ADDW (c);
2004 else
2005 {
2006 /* The last read character is not part of the number
2007 anymore. */
2008 ungetc (c, s);
2009 break;
2010 }
2011 #else
2012 const char *cmpp = decimal;
2013 int avail = width > 0 ? width : INT_MAX;
2014
2015 if (! got_dot)
2016 {
2017 while ((unsigned char) *cmpp == c && avail >= 0)
2018 if (*++cmpp == '\0')
2019 break;
2020 else
2021 {
2022 if (avail == 0 || inchar () == EOF)
2023 break;
2024 --avail;
2025 }
2026 }
2027
2028 if (*cmpp == '\0')
2029 {
2030 /* Add all the characters. */
2031 for (cmpp = decimal; *cmpp != '\0'; ++cmpp)
2032 ADDW ((unsigned char) *cmpp);
2033 if (width > 0)
2034 width = avail;
2035 got_dot = 1;
2036 }
2037 else
2038 {
2039 /* Figure out whether it is a thousands separator.
2040 There is one problem: we possibly read more than
2041 one character. We cannot push them back but since
2042 we know that parts of the `decimal' string matched,
2043 we can compare against it. */
2044 const char *cmp2p = thousands;
2045
2046 if ((flags & GROUP) != 0 && ! got_dot)
2047 {
2048 while (cmp2p - thousands < cmpp - decimal
2049 && *cmp2p == decimal[cmp2p - thousands])
2050 ++cmp2p;
2051 if (cmp2p - thousands == cmpp - decimal)
2052 {
2053 while ((unsigned char) *cmp2p == c && avail >= 0)
2054 if (*++cmp2p == '\0')
2055 break;
2056 else
2057 {
2058 if (avail == 0 || inchar () == EOF)
2059 break;
2060 --avail;
2061 }
2062 }
2063 }
2064
2065 if (cmp2p != NULL && *cmp2p == '\0')
2066 {
2067 /* Add all the characters. */
2068 for (cmpp = thousands; *cmpp != '\0'; ++cmpp)
2069 ADDW ((unsigned char) *cmpp);
2070 if (width > 0)
2071 width = avail;
2072 }
2073 else
2074 {
2075 /* The last read character is not part of the number
2076 anymore. */
2077 ungetc (c, s);
2078 break;
2079 }
2080 }
2081 #endif
2082 }
2083
2084 if (width == 0 || inchar () == EOF)
2085 break;
2086
2087 if (width > 0)
2088 --width;
2089 }
2090
2091 wctrans_t map;
2092 if (__builtin_expect ((flags & I18N) != 0, 0)
2093 /* Hexadecimal floats make no sense, fixing localized
2094 digits with ASCII letters. */
2095 && !(flags & HEXA_FLOAT)
2096 /* Minimum requirement. */
2097 && (wpsize == 0 || got_dot)
2098 && (map = __wctrans ("to_inpunct")) != NULL)
2099 {
2100 /* Reget the first character. */
2101 inchar ();
2102
2103 /* Localized digits, decimal points, and thousands
2104 separator. */
2105 wint_t wcdigits[12];
2106
2107 /* First get decimal equivalent to check if we read it
2108 or not. */
2109 wcdigits[11] = __towctrans (L'.', map);
2110
2111 /* If we have not read any character or have just read
2112 locale decimal point which matches the decimal point
2113 for localized FP numbers, then we may have localized
2114 digits. Note, we test GOT_DOT above. */
2115 #ifdef COMPILE_WSCANF
2116 if (wpsize == 0 || (wpsize == 1 && wcdigits[11] == decimal))
2117 #else
2118 char mbdigits[12][MB_LEN_MAX + 1];
2119
2120 mbstate_t state;
2121 memset (&state, '\0', sizeof (state));
2122
2123 bool match_so_far = wpsize == 0;
2124 size_t mblen = __wcrtomb (mbdigits[11], wcdigits[11], &state);
2125 if (mblen != (size_t) -1)
2126 {
2127 mbdigits[11][mblen] = '\0';
2128 match_so_far |= (wpsize == strlen (decimal)
2129 && strcmp (decimal, mbdigits[11]) == 0);
2130 }
2131 else
2132 {
2133 size_t decimal_len = strlen (decimal);
2134 /* This should always be the case but the data comes
2135 from a file. */
2136 if (decimal_len <= MB_LEN_MAX)
2137 {
2138 match_so_far |= wpsize == decimal_len;
2139 memcpy (mbdigits[11], decimal, decimal_len + 1);
2140 }
2141 else
2142 match_so_far = false;
2143 }
2144
2145 if (match_so_far)
2146 #endif
2147 {
2148 bool have_locthousands = (flags & GROUP) != 0;
2149
2150 /* Now get the digits and the thousands-sep equivalents. */
2151 for (int n = 0; n < 11; ++n)
2152 {
2153 if (n < 10)
2154 wcdigits[n] = __towctrans (L'0' + n, map);
2155 else if (n == 10)
2156 {
2157 wcdigits[10] = __towctrans (L',', map);
2158 have_locthousands &= wcdigits[10] != L'\0';
2159 }
2160
2161 #ifndef COMPILE_WSCANF
2162 memset (&state, '\0', sizeof (state));
2163
2164 size_t mblen = __wcrtomb (mbdigits[n], wcdigits[n],
2165 &state);
2166 if (mblen == (size_t) -1)
2167 {
2168 if (n == 10)
2169 {
2170 if (have_locthousands)
2171 {
2172 size_t thousands_len = strlen (thousands);
2173 if (thousands_len <= MB_LEN_MAX)
2174 memcpy (mbdigits[10], thousands,
2175 thousands_len + 1);
2176 else
2177 have_locthousands = false;
2178 }
2179 }
2180 else
2181 /* Ignore checking against localized digits. */
2182 goto no_i18nflt;
2183 }
2184 else
2185 mbdigits[n][mblen] = '\0';
2186 #endif
2187 }
2188
2189 /* Start checking against localized digits, if
2190 conversion is done correctly. */
2191 while (1)
2192 {
2193 if (got_e && wp[wpsize - 1] == exp_char
2194 && (c == L_('-') || c == L_('+')))
2195 ADDW (c);
2196 else if (wpsize > 0 && !got_e
2197 && (CHAR_T) TOLOWER (c) == exp_char)
2198 {
2199 ADDW (exp_char);
2200 got_e = got_dot = 1;
2201 }
2202 else
2203 {
2204 /* Check against localized digits, decimal point,
2205 and thousands separator. */
2206 int n;
2207 for (n = 0; n < 12; ++n)
2208 {
2209 #ifdef COMPILE_WSCANF
2210 if (c == wcdigits[n])
2211 {
2212 if (n < 10)
2213 ADDW (L_('0') + n);
2214 else if (n == 11 && !got_dot)
2215 {
2216 ADDW (decimal);
2217 got_dot = 1;
2218 }
2219 else if (n == 10 && have_locthousands
2220 && ! got_dot)
2221 ADDW (thousands);
2222 else
2223 /* The last read character is not part
2224 of the number anymore. */
2225 n = 12;
2226
2227 break;
2228 }
2229 #else
2230 const char *cmpp = mbdigits[n];
2231 int avail = width > 0 ? width : INT_MAX;
2232
2233 while ((unsigned char) *cmpp == c && avail >= 0)
2234 if (*++cmpp == '\0')
2235 break;
2236 else
2237 {
2238 if (avail == 0 || inchar () == EOF)
2239 break;
2240 --avail;
2241 }
2242 if (*cmpp == '\0')
2243 {
2244 if (width > 0)
2245 width = avail;
2246
2247 if (n < 10)
2248 ADDW (L_('0') + n);
2249 else if (n == 11 && !got_dot)
2250 {
2251 /* Add all the characters. */
2252 for (cmpp = decimal; *cmpp != '\0';
2253 ++cmpp)
2254 ADDW ((unsigned char) *cmpp);
2255
2256 got_dot = 1;
2257 }
2258 else if (n == 10 && (flags & GROUP) != 0
2259 && ! got_dot)
2260 {
2261 /* Add all the characters. */
2262 for (cmpp = thousands; *cmpp != '\0';
2263 ++cmpp)
2264 ADDW ((unsigned char) *cmpp);
2265 }
2266 else
2267 /* The last read character is not part
2268 of the number anymore. */
2269 n = 12;
2270
2271 break;
2272 }
2273
2274 /* We are pushing all read characters back. */
2275 if (cmpp > mbdigits[n])
2276 {
2277 ungetc (c, s);
2278 while (--cmpp > mbdigits[n])
2279 ungetc_not_eof ((unsigned char) *cmpp, s);
2280 c = (unsigned char) *cmpp;
2281 }
2282 #endif
2283 }
2284
2285 if (n >= 12)
2286 {
2287 /* The last read character is not part
2288 of the number anymore. */
2289 ungetc (c, s);
2290 break;
2291 }
2292 }
2293
2294 if (width == 0 || inchar () == EOF)
2295 break;
2296
2297 if (width > 0)
2298 --width;
2299 }
2300 }
2301
2302 #ifndef COMPILE_WSCANF
2303 no_i18nflt:
2304 ;
2305 #endif
2306 }
2307
2308 /* Have we read any character? If we try to read a number
2309 in hexadecimal notation and we have read only the `0x'
2310 prefix this is an error. */
2311 if (__builtin_expect (wpsize == 0
2312 || ((flags & HEXA_FLOAT) && wpsize == 2), 0))
2313 conv_error ();
2314
2315 scan_float:
2316 /* Convert the number. */
2317 ADDW (L_('\0'));
2318 if ((flags & LONGDBL) && !__ldbl_is_dbl)
2319 {
2320 long double d = __strtold_internal (wp, &tw, flags & GROUP);
2321 if (!(flags & SUPPRESS) && tw != wp)
2322 *ARG (long double *) = negative ? -d : d;
2323 }
2324 else if (flags & (LONG | LONGDBL))
2325 {
2326 double d = __strtod_internal (wp, &tw, flags & GROUP);
2327 if (!(flags & SUPPRESS) && tw != wp)
2328 *ARG (double *) = negative ? -d : d;
2329 }
2330 else
2331 {
2332 float d = __strtof_internal (wp, &tw, flags & GROUP);
2333 if (!(flags & SUPPRESS) && tw != wp)
2334 *ARG (float *) = negative ? -d : d;
2335 }
2336
2337 if (__glibc_unlikely (tw == wp))
2338 conv_error ();
2339
2340 if (!(flags & SUPPRESS))
2341 ++done;
2342 break;
2343
2344 case L_('['): /* Character class. */
2345 if (flags & LONG)
2346 STRING_ARG (wstr, wchar_t, 100);
2347 else
2348 STRING_ARG (str, char, 100);
2349
2350 if (*f == L_('^'))
2351 {
2352 ++f;
2353 not_in = 1;
2354 }
2355 else
2356 not_in = 0;
2357
2358 if (width < 0)
2359 /* There is no width given so there is also no limit on the
2360 number of characters we read. Therefore we set width to
2361 a very high value to make the algorithm easier. */
2362 width = INT_MAX;
2363
2364 #ifdef COMPILE_WSCANF
2365 /* Find the beginning and the end of the scanlist. We are not
2366 creating a lookup table since it would have to be too large.
2367 Instead we search each time through the string. This is not
2368 a constant lookup time but who uses this feature deserves to
2369 be punished. */
2370 tw = (wchar_t *) f; /* Marks the beginning. */
2371
2372 if (*f == L']')
2373 ++f;
2374
2375 while ((fc = *f++) != L'\0' && fc != L']');
2376
2377 if (__glibc_unlikely (fc == L'\0'))
2378 conv_error ();
2379 wchar_t *twend = (wchar_t *) f - 1;
2380 #else
2381 /* Fill WP with byte flags indexed by character.
2382 We will use this flag map for matching input characters. */
2383 if (wpmax < UCHAR_MAX + 1)
2384 {
2385 wpmax = UCHAR_MAX + 1;
2386 wp = (char *) alloca (wpmax);
2387 }
2388 memset (wp, '\0', UCHAR_MAX + 1);
2389
2390 fc = *f;
2391 if (fc == ']' || fc == '-')
2392 {
2393 /* If ] or - appears before any char in the set, it is not
2394 the terminator or separator, but the first char in the
2395 set. */
2396 wp[fc] = 1;
2397 ++f;
2398 }
2399
2400 while ((fc = *f++) != '\0' && fc != ']')
2401 if (fc == '-' && *f != '\0' && *f != ']'
2402 && (unsigned char) f[-2] <= (unsigned char) *f)
2403 {
2404 /* Add all characters from the one before the '-'
2405 up to (but not including) the next format char. */
2406 for (fc = (unsigned char) f[-2]; fc < (unsigned char) *f; ++fc)
2407 wp[fc] = 1;
2408 }
2409 else
2410 /* Add the character to the flag map. */
2411 wp[fc] = 1;
2412
2413 if (__glibc_unlikely (fc == '\0'))
2414 conv_error();
2415 #endif
2416
2417 if (flags & LONG)
2418 {
2419 size_t now = read_in;
2420 #ifdef COMPILE_WSCANF
2421 if (__glibc_unlikely (inchar () == WEOF))
2422 input_error ();
2423
2424 do
2425 {
2426 wchar_t *runp;
2427
2428 /* Test whether it's in the scanlist. */
2429 runp = tw;
2430 while (runp < twend)
2431 {
2432 if (runp[0] == L'-' && runp[1] != '\0'
2433 && runp + 1 != twend
2434 && runp != tw
2435 && (unsigned int) runp[-1] <= (unsigned int) runp[1])
2436 {
2437 /* Match against all characters in between the
2438 first and last character of the sequence. */
2439 wchar_t wc;
2440
2441 for (wc = runp[-1] + 1; wc <= runp[1]; ++wc)
2442 if ((wint_t) wc == c)
2443 break;
2444
2445 if (wc <= runp[1] && !not_in)
2446 break;
2447 if (wc <= runp[1] && not_in)
2448 {
2449 /* The current character is not in the
2450 scanset. */
2451 ungetc (c, s);
2452 goto out;
2453 }
2454
2455 runp += 2;
2456 }
2457 else
2458 {
2459 if ((wint_t) *runp == c && !not_in)
2460 break;
2461 if ((wint_t) *runp == c && not_in)
2462 {
2463 ungetc (c, s);
2464 goto out;
2465 }
2466
2467 ++runp;
2468 }
2469 }
2470
2471 if (runp == twend && !not_in)
2472 {
2473 ungetc (c, s);
2474 goto out;
2475 }
2476
2477 if (!(flags & SUPPRESS))
2478 {
2479 *wstr++ = c;
2480
2481 if ((flags & MALLOC)
2482 && wstr == (wchar_t *) *strptr + strsize)
2483 {
2484 /* Enlarge the buffer. */
2485 wstr = (wchar_t *) realloc (*strptr,
2486 (2 * strsize)
2487 * sizeof (wchar_t));
2488 if (wstr == NULL)
2489 {
2490 /* Can't allocate that much. Last-ditch
2491 effort. */
2492 wstr = (wchar_t *)
2493 realloc (*strptr, (strsize + 1)
2494 * sizeof (wchar_t));
2495 if (wstr == NULL)
2496 {
2497 if (flags & POSIX_MALLOC)
2498 {
2499 done = EOF;
2500 goto errout;
2501 }
2502 /* We lose. Oh well. Terminate the string
2503 and stop converting, so at least we don't
2504 skip any input. */
2505 ((wchar_t *) (*strptr))[strsize - 1] = L'\0';
2506 strptr = NULL;
2507 ++done;
2508 conv_error ();
2509 }
2510 else
2511 {
2512 *strptr = (char *) wstr;
2513 wstr += strsize;
2514 ++strsize;
2515 }
2516 }
2517 else
2518 {
2519 *strptr = (char *) wstr;
2520 wstr += strsize;
2521 strsize *= 2;
2522 }
2523 }
2524 }
2525 }
2526 while (--width > 0 && inchar () != WEOF);
2527 out:
2528 #else
2529 char buf[MB_LEN_MAX];
2530 size_t cnt = 0;
2531 mbstate_t cstate;
2532
2533 if (__glibc_unlikely (inchar () == EOF))
2534 input_error ();
2535
2536 memset (&cstate, '\0', sizeof (cstate));
2537
2538 do
2539 {
2540 if (wp[c] == not_in)
2541 {
2542 ungetc_not_eof (c, s);
2543 break;
2544 }
2545
2546 /* This is easy. */
2547 if (!(flags & SUPPRESS))
2548 {
2549 size_t n;
2550
2551 /* Convert it into a wide character. */
2552 buf[0] = c;
2553 n = __mbrtowc (wstr, buf, 1, &cstate);
2554
2555 if (n == (size_t) -2)
2556 {
2557 /* Possibly correct character, just not enough
2558 input. */
2559 ++cnt;
2560 assert (cnt < MB_CUR_MAX);
2561 continue;
2562 }
2563 cnt = 0;
2564
2565 ++wstr;
2566 if ((flags & MALLOC)
2567 && wstr == (wchar_t *) *strptr + strsize)
2568 {
2569 /* Enlarge the buffer. */
2570 wstr = (wchar_t *) realloc (*strptr,
2571 (2 * strsize
2572 * sizeof (wchar_t)));
2573 if (wstr == NULL)
2574 {
2575 /* Can't allocate that much. Last-ditch
2576 effort. */
2577 wstr = (wchar_t *)
2578 realloc (*strptr, ((strsize + 1)
2579 * sizeof (wchar_t)));
2580 if (wstr == NULL)
2581 {
2582 if (flags & POSIX_MALLOC)
2583 {
2584 done = EOF;
2585 goto errout;
2586 }
2587 /* We lose. Oh well. Terminate the
2588 string and stop converting,
2589 so at least we don't skip any input. */
2590 ((wchar_t *) (*strptr))[strsize - 1] = L'\0';
2591 strptr = NULL;
2592 ++done;
2593 conv_error ();
2594 }
2595 else
2596 {
2597 *strptr = (char *) wstr;
2598 wstr += strsize;
2599 ++strsize;
2600 }
2601 }
2602 else
2603 {
2604 *strptr = (char *) wstr;
2605 wstr += strsize;
2606 strsize *= 2;
2607 }
2608 }
2609 }
2610
2611 if (--width <= 0)
2612 break;
2613 }
2614 while (inchar () != EOF);
2615
2616 if (__glibc_unlikely (cnt != 0))
2617 /* We stopped in the middle of recognizing another
2618 character. That's a problem. */
2619 encode_error ();
2620 #endif
2621
2622 if (__glibc_unlikely (now == read_in))
2623 /* We haven't succesfully read any character. */
2624 conv_error ();
2625
2626 if (!(flags & SUPPRESS))
2627 {
2628 *wstr++ = L'\0';
2629
2630 if ((flags & MALLOC)
2631 && wstr - (wchar_t *) *strptr != strsize)
2632 {
2633 wchar_t *cp = (wchar_t *)
2634 realloc (*strptr, ((wstr - (wchar_t *) *strptr)
2635 * sizeof(wchar_t)));
2636 if (cp != NULL)
2637 *strptr = (char *) cp;
2638 }
2639 strptr = NULL;
2640
2641 ++done;
2642 }
2643 }
2644 else
2645 {
2646 size_t now = read_in;
2647
2648 if (__glibc_unlikely (inchar () == EOF))
2649 input_error ();
2650
2651 #ifdef COMPILE_WSCANF
2652
2653 memset (&state, '\0', sizeof (state));
2654
2655 do
2656 {
2657 wchar_t *runp;
2658 size_t n;
2659
2660 /* Test whether it's in the scanlist. */
2661 runp = tw;
2662 while (runp < twend)
2663 {
2664 if (runp[0] == L'-' && runp[1] != '\0'
2665 && runp + 1 != twend
2666 && runp != tw
2667 && (unsigned int) runp[-1] <= (unsigned int) runp[1])
2668 {
2669 /* Match against all characters in between the
2670 first and last character of the sequence. */
2671 wchar_t wc;
2672
2673 for (wc = runp[-1] + 1; wc <= runp[1]; ++wc)
2674 if ((wint_t) wc == c)
2675 break;
2676
2677 if (wc <= runp[1] && !not_in)
2678 break;
2679 if (wc <= runp[1] && not_in)
2680 {
2681 /* The current character is not in the
2682 scanset. */
2683 ungetc (c, s);
2684 goto out2;
2685 }
2686
2687 runp += 2;
2688 }
2689 else
2690 {
2691 if ((wint_t) *runp == c && !not_in)
2692 break;
2693 if ((wint_t) *runp == c && not_in)
2694 {
2695 ungetc (c, s);
2696 goto out2;
2697 }
2698
2699 ++runp;
2700 }
2701 }
2702
2703 if (runp == twend && !not_in)
2704 {
2705 ungetc (c, s);
2706 goto out2;
2707 }
2708
2709 if (!(flags & SUPPRESS))
2710 {
2711 if ((flags & MALLOC)
2712 && str + MB_CUR_MAX >= *strptr + strsize)
2713 {
2714 /* Enlarge the buffer. */
2715 size_t strleng = str - *strptr;
2716 char *newstr;
2717
2718 newstr = (char *) realloc (*strptr, 2 * strsize);
2719 if (newstr == NULL)
2720 {
2721 /* Can't allocate that much. Last-ditch
2722 effort. */
2723 newstr = (char *) realloc (*strptr,
2724 strleng + MB_CUR_MAX);
2725 if (newstr == NULL)
2726 {
2727 if (flags & POSIX_MALLOC)
2728 {
2729 done = EOF;
2730 goto errout;
2731 }
2732 /* We lose. Oh well. Terminate the string
2733 and stop converting, so at least we don't
2734 skip any input. */
2735 ((char *) (*strptr))[strleng] = '\0';
2736 strptr = NULL;
2737 ++done;
2738 conv_error ();
2739 }
2740 else
2741 {
2742 *strptr = newstr;
2743 str = newstr + strleng;
2744 strsize = strleng + MB_CUR_MAX;
2745 }
2746 }
2747 else
2748 {
2749 *strptr = newstr;
2750 str = newstr + strleng;
2751 strsize *= 2;
2752 }
2753 }
2754 }
2755
2756 n = __wcrtomb (!(flags & SUPPRESS) ? str : NULL, c, &state);
2757 if (__glibc_unlikely (n == (size_t) -1))
2758 encode_error ();
2759
2760 assert (n <= MB_CUR_MAX);
2761 str += n;
2762 }
2763 while (--width > 0 && inchar () != WEOF);
2764 out2:
2765 #else
2766 do
2767 {
2768 if (wp[c] == not_in)
2769 {
2770 ungetc_not_eof (c, s);
2771 break;
2772 }
2773
2774 /* This is easy. */
2775 if (!(flags & SUPPRESS))
2776 {
2777 *str++ = c;
2778 if ((flags & MALLOC)
2779 && (char *) str == *strptr + strsize)
2780 {
2781 /* Enlarge the buffer. */
2782 size_t newsize = 2 * strsize;
2783
2784 allocagain:
2785 str = (char *) realloc (*strptr, newsize);
2786 if (str == NULL)
2787 {
2788 /* Can't allocate that much. Last-ditch
2789 effort. */
2790 if (newsize > strsize + 1)
2791 {
2792 newsize = strsize + 1;
2793 goto allocagain;
2794 }
2795 if (flags & POSIX_MALLOC)
2796 {
2797 done = EOF;
2798 goto errout;
2799 }
2800 /* We lose. Oh well. Terminate the
2801 string and stop converting,
2802 so at least we don't skip any input. */
2803 ((char *) (*strptr))[strsize - 1] = '\0';
2804 strptr = NULL;
2805 ++done;
2806 conv_error ();
2807 }
2808 else
2809 {
2810 *strptr = (char *) str;
2811 str += strsize;
2812 strsize = newsize;
2813 }
2814 }
2815 }
2816 }
2817 while (--width > 0 && inchar () != EOF);
2818 #endif
2819
2820 if (__glibc_unlikely (now == read_in))
2821 /* We haven't succesfully read any character. */
2822 conv_error ();
2823
2824 if (!(flags & SUPPRESS))
2825 {
2826 #ifdef COMPILE_WSCANF
2827 /* We have to emit the code to get into the initial
2828 state. */
2829 char buf[MB_LEN_MAX];
2830 size_t n = __wcrtomb (buf, L'\0', &state);
2831 if (n > 0 && (flags & MALLOC)
2832 && str + n >= *strptr + strsize)
2833 {
2834 /* Enlarge the buffer. */
2835 size_t strleng = str - *strptr;
2836 char *newstr;
2837
2838 newstr = (char *) realloc (*strptr, strleng + n + 1);
2839 if (newstr == NULL)
2840 {
2841 if (flags & POSIX_MALLOC)
2842 {
2843 done = EOF;
2844 goto errout;
2845 }
2846 /* We lose. Oh well. Terminate the string
2847 and stop converting, so at least we don't
2848 skip any input. */
2849 ((char *) (*strptr))[strleng] = '\0';
2850 strptr = NULL;
2851 ++done;
2852 conv_error ();
2853 }
2854 else
2855 {
2856 *strptr = newstr;
2857 str = newstr + strleng;
2858 strsize = strleng + n + 1;
2859 }
2860 }
2861
2862 str = __mempcpy (str, buf, n);
2863 #endif
2864 *str++ = '\0';
2865
2866 if ((flags & MALLOC) && str - *strptr != strsize)
2867 {
2868 char *cp = (char *) realloc (*strptr, str - *strptr);
2869 if (cp != NULL)
2870 *strptr = cp;
2871 }
2872 strptr = NULL;
2873
2874 ++done;
2875 }
2876 }
2877 break;
2878
2879 case L_('p'): /* Generic pointer. */
2880 base = 16;
2881 /* A PTR must be the same size as a `long int'. */
2882 flags &= ~(SHORT|LONGDBL);
2883 if (need_long)
2884 flags |= LONG;
2885 flags |= READ_POINTER;
2886 goto number;
2887
2888 default:
2889 /* If this is an unknown format character punt. */
2890 conv_error ();
2891 }
2892 }
2893
2894 /* The last thing we saw int the format string was a white space.
2895 Consume the last white spaces. */
2896 if (skip_space)
2897 {
2898 do
2899 c = inchar ();
2900 while (ISSPACE (c));
2901 ungetc (c, s);
2902 }
2903
2904 errout:
2905 /* Unlock stream. */
2906 UNLOCK_STREAM (s);
2907
2908 if (use_malloc)
2909 free (wp);
2910
2911 if (errp != NULL)
2912 *errp |= errval;
2913
2914 if (__glibc_unlikely (done == EOF))
2915 {
2916 if (__glibc_unlikely (ptrs_to_free != NULL))
2917 {
2918 struct ptrs_to_free *p = ptrs_to_free;
2919 while (p != NULL)
2920 {
2921 for (size_t cnt = 0; cnt < p->count; ++cnt)
2922 {
2923 free (*p->ptrs[cnt]);
2924 *p->ptrs[cnt] = NULL;
2925 }
2926 p = p->next;
2927 ptrs_to_free = p;
2928 }
2929 }
2930 }
2931 else if (__glibc_unlikely (strptr != NULL))
2932 {
2933 free (*strptr);
2934 *strptr = NULL;
2935 }
2936 return done;
2937 }
2938
2939 #ifdef COMPILE_WSCANF
2940 int
2941 __vfwscanf (FILE *s, const wchar_t *format, va_list argptr)
2942 {
2943 return _IO_vfwscanf (s, format, argptr, NULL);
2944 }
2945 ldbl_weak_alias (__vfwscanf, vfwscanf)
2946 #else
2947 int
2948 ___vfscanf (FILE *s, const char *format, va_list argptr)
2949 {
2950 return _IO_vfscanf_internal (s, format, argptr, NULL);
2951 }
2952 ldbl_strong_alias (_IO_vfscanf_internal, _IO_vfscanf)
2953 ldbl_hidden_def (_IO_vfscanf_internal, _IO_vfscanf)
2954 ldbl_strong_alias (___vfscanf, __vfscanf)
2955 ldbl_hidden_def (___vfscanf, __vfscanf)
2956 ldbl_weak_alias (___vfscanf, vfscanf)
2957 #endif
sources.redhat.com git mirror of glibc CVS