1 | /* Copyright (C) 1991-2018 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 <libc-diag.h> |
31 | #include <libc-lock.h> |
32 | #include <locale/localeinfo.h> |
33 | #include <scratch_buffer.h> |
34 | |
35 | #ifdef __GNUC__ |
36 | # define HAVE_LONGLONG |
37 | # define LONGLONG long long |
38 | #else |
39 | # define LONGLONG long |
40 | #endif |
41 | |
42 | /* Determine whether we have to handle `long long' at all. */ |
43 | #if LONG_MAX == LONG_LONG_MAX |
44 | # define need_longlong 0 |
45 | #else |
46 | # define need_longlong 1 |
47 | #endif |
48 | |
49 | /* Determine whether we have to handle `long'. */ |
50 | #if INT_MAX == LONG_MAX |
51 | # define need_long 0 |
52 | #else |
53 | # define need_long 1 |
54 | #endif |
55 | |
56 | /* Those are flags in the conversion format. */ |
57 | #define LONG 0x0001 /* l: long or double */ |
58 | #define LONGDBL 0x0002 /* L: long long or long double */ |
59 | #define SHORT 0x0004 /* h: short */ |
60 | #define SUPPRESS 0x0008 /* *: suppress assignment */ |
61 | #define POINTER 0x0010 /* weird %p pointer (`fake hex') */ |
62 | #define NOSKIP 0x0020 /* do not skip blanks */ |
63 | #define NUMBER_SIGNED 0x0040 /* signed integer */ |
64 | #define GROUP 0x0080 /* ': group numbers */ |
65 | #define GNU_MALLOC 0x0100 /* a: malloc strings */ |
66 | #define CHAR 0x0200 /* hh: char */ |
67 | #define I18N 0x0400 /* I: use locale's digits */ |
68 | #define HEXA_FLOAT 0x0800 /* hexadecimal float */ |
69 | #define READ_POINTER 0x1000 /* this is a pointer value */ |
70 | #define POSIX_MALLOC 0x2000 /* m: malloc strings */ |
71 | #define MALLOC (GNU_MALLOC | POSIX_MALLOC) |
72 | |
73 | #include <locale/localeinfo.h> |
74 | #include <libioP.h> |
75 | |
76 | #ifdef COMPILE_WSCANF |
77 | # define ungetc(c, s) ((void) (c == WEOF \ |
78 | || (--read_in, \ |
79 | _IO_sputbackwc (s, c)))) |
80 | # define ungetc_not_eof(c, s) ((void) (--read_in, \ |
81 | _IO_sputbackwc (s, c))) |
82 | # define inchar() (c == WEOF ? ((errno = inchar_errno), WEOF) \ |
83 | : ((c = _IO_getwc_unlocked (s)), \ |
84 | (void) (c != WEOF \ |
85 | ? ++read_in \ |
86 | : (size_t) (inchar_errno = errno)), c)) |
87 | |
88 | # define ISSPACE(Ch) iswspace (Ch) |
89 | # define ISDIGIT(Ch) iswdigit (Ch) |
90 | # define ISXDIGIT(Ch) iswxdigit (Ch) |
91 | # define TOLOWER(Ch) towlower (Ch) |
92 | # define ORIENT if (_IO_fwide (s, 1) != 1) return WEOF |
93 | # define __strtoll_internal __wcstoll_internal |
94 | # define __strtoull_internal __wcstoull_internal |
95 | # define __strtol_internal __wcstol_internal |
96 | # define __strtoul_internal __wcstoul_internal |
97 | # define __strtold_internal __wcstold_internal |
98 | # define __strtod_internal __wcstod_internal |
99 | # define __strtof_internal __wcstof_internal |
100 | |
101 | # define L_(Str) L##Str |
102 | # define CHAR_T wchar_t |
103 | # define UCHAR_T unsigned int |
104 | # define WINT_T wint_t |
105 | # undef EOF |
106 | # define EOF WEOF |
107 | #else |
108 | # define ungetc(c, s) ((void) ((int) c == EOF \ |
109 | || (--read_in, \ |
110 | _IO_sputbackc (s, (unsigned char) c)))) |
111 | # define ungetc_not_eof(c, s) ((void) (--read_in, \ |
112 | _IO_sputbackc (s, (unsigned char) c))) |
113 | # define inchar() (c == EOF ? ((errno = inchar_errno), EOF) \ |
114 | : ((c = _IO_getc_unlocked (s)), \ |
115 | (void) (c != EOF \ |
116 | ? ++read_in \ |
117 | : (size_t) (inchar_errno = errno)), c)) |
118 | # define ISSPACE(Ch) __isspace_l (Ch, loc) |
119 | # define ISDIGIT(Ch) __isdigit_l (Ch, loc) |
120 | # define ISXDIGIT(Ch) __isxdigit_l (Ch, loc) |
121 | # define TOLOWER(Ch) __tolower_l ((unsigned char) (Ch), loc) |
122 | # define ORIENT if (_IO_vtable_offset (s) == 0 \ |
123 | && _IO_fwide (s, -1) != -1) \ |
124 | return EOF |
125 | |
126 | # define L_(Str) Str |
127 | # define CHAR_T char |
128 | # define UCHAR_T unsigned char |
129 | # define WINT_T int |
130 | #endif |
131 | |
132 | #include "printf-parse.h" /* Use read_int. */ |
133 | |
134 | #define encode_error() do { \ |
135 | errval = 4; \ |
136 | __set_errno (EILSEQ); \ |
137 | goto errout; \ |
138 | } while (0) |
139 | #define conv_error() do { \ |
140 | errval = 2; \ |
141 | goto errout; \ |
142 | } while (0) |
143 | #define input_error() do { \ |
144 | errval = 1; \ |
145 | if (done == 0) done = EOF; \ |
146 | goto errout; \ |
147 | } while (0) |
148 | #define add_ptr_to_free(ptr) \ |
149 | do \ |
150 | { \ |
151 | if (ptrs_to_free == NULL \ |
152 | || ptrs_to_free->count == (sizeof (ptrs_to_free->ptrs) \ |
153 | / sizeof (ptrs_to_free->ptrs[0]))) \ |
154 | { \ |
155 | struct ptrs_to_free *new_ptrs = alloca (sizeof (*ptrs_to_free)); \ |
156 | new_ptrs->count = 0; \ |
157 | new_ptrs->next = ptrs_to_free; \ |
158 | ptrs_to_free = new_ptrs; \ |
159 | } \ |
160 | ptrs_to_free->ptrs[ptrs_to_free->count++] = (ptr); \ |
161 | } \ |
162 | while (0) |
163 | #define ARGCHECK(s, format) \ |
164 | do \ |
165 | { \ |
166 | /* Check file argument for consistence. */ \ |
167 | CHECK_FILE (s, EOF); \ |
168 | if (s->_flags & _IO_NO_READS) \ |
169 | { \ |
170 | __set_errno (EBADF); \ |
171 | return EOF; \ |
172 | } \ |
173 | else if (format == NULL) \ |
174 | { \ |
175 | __set_errno (EINVAL); \ |
176 | return EOF; \ |
177 | } \ |
178 | } while (0) |
179 | #define LOCK_STREAM(S) \ |
180 | __libc_cleanup_region_start (1, (void (*) (void *)) &_IO_funlockfile, (S)); \ |
181 | _IO_flockfile (S) |
182 | #define UNLOCK_STREAM(S) \ |
183 | _IO_funlockfile (S); \ |
184 | __libc_cleanup_region_end (0) |
185 | |
186 | struct ptrs_to_free |
187 | { |
188 | size_t count; |
189 | struct ptrs_to_free *next; |
190 | char **ptrs[32]; |
191 | }; |
192 | |
193 | struct char_buffer { |
194 | CHAR_T *current; |
195 | CHAR_T *end; |
196 | struct scratch_buffer scratch; |
197 | }; |
198 | |
199 | /* Returns a pointer to the first CHAR_T object in the buffer. Only |
200 | valid if char_buffer_add (BUFFER, CH) has been called and |
201 | char_buffer_error (BUFFER) is false. */ |
202 | static inline CHAR_T * |
203 | char_buffer_start (const struct char_buffer *buffer) |
204 | { |
205 | return (CHAR_T *) buffer->scratch.data; |
206 | } |
207 | |
208 | /* Returns the number of CHAR_T objects in the buffer. Only valid if |
209 | char_buffer_error (BUFFER) is false. */ |
210 | static inline size_t |
211 | char_buffer_size (const struct char_buffer *buffer) |
212 | { |
213 | return buffer->current - char_buffer_start (buffer); |
214 | } |
215 | |
216 | /* Reinitializes BUFFER->current and BUFFER->end to cover the entire |
217 | scratch buffer. */ |
218 | static inline void |
219 | char_buffer_rewind (struct char_buffer *buffer) |
220 | { |
221 | buffer->current = char_buffer_start (buffer); |
222 | buffer->end = buffer->current + buffer->scratch.length / sizeof (CHAR_T); |
223 | } |
224 | |
225 | /* Returns true if a previous call to char_buffer_add (BUFFER, CH) |
226 | failed. */ |
227 | static inline bool |
228 | char_buffer_error (const struct char_buffer *buffer) |
229 | { |
230 | return __glibc_unlikely (buffer->current == NULL); |
231 | } |
232 | |
233 | /* Slow path for char_buffer_add. */ |
234 | static void |
235 | char_buffer_add_slow (struct char_buffer *buffer, CHAR_T ch) |
236 | { |
237 | if (char_buffer_error (buffer)) |
238 | return; |
239 | size_t offset = buffer->end - (CHAR_T *) buffer->scratch.data; |
240 | if (!scratch_buffer_grow_preserve (&buffer->scratch)) |
241 | { |
242 | buffer->current = NULL; |
243 | buffer->end = NULL; |
244 | return; |
245 | } |
246 | char_buffer_rewind (buffer); |
247 | buffer->current += offset; |
248 | *buffer->current++ = ch; |
249 | } |
250 | |
251 | /* Adds CH to BUFFER. This function does not report any errors, check |
252 | for them with char_buffer_error. */ |
253 | static inline void |
254 | char_buffer_add (struct char_buffer *buffer, CHAR_T ch) |
255 | __attribute__ ((always_inline)); |
256 | static inline void |
257 | char_buffer_add (struct char_buffer *buffer, CHAR_T ch) |
258 | { |
259 | if (__glibc_unlikely (buffer->current == buffer->end)) |
260 | char_buffer_add_slow (buffer, ch); |
261 | else |
262 | *buffer->current++ = ch; |
263 | } |
264 | |
265 | /* Read formatted input from S according to the format string |
266 | FORMAT, using the argument list in ARG. |
267 | Return the number of assignments made, or -1 for an input error. */ |
268 | #ifdef COMPILE_WSCANF |
269 | int |
270 | _IO_vfwscanf (FILE *s, const wchar_t *format, va_list argptr, |
271 | int *errp) |
272 | #else |
273 | int |
274 | _IO_vfscanf_internal (FILE *s, const char *format, va_list argptr, |
275 | int *errp) |
276 | #endif |
277 | { |
278 | va_list arg; |
279 | const CHAR_T *f = format; |
280 | UCHAR_T fc; /* Current character of the format. */ |
281 | WINT_T done = 0; /* Assignments done. */ |
282 | size_t read_in = 0; /* Chars read in. */ |
283 | WINT_T c = 0; /* Last char read. */ |
284 | int width; /* Maximum field width. */ |
285 | int flags; /* Modifiers for current format element. */ |
286 | int errval = 0; |
287 | #ifndef COMPILE_WSCANF |
288 | locale_t loc = _NL_CURRENT_LOCALE; |
289 | struct __locale_data *const curctype = loc->__locales[LC_CTYPE]; |
290 | #endif |
291 | |
292 | /* Errno of last failed inchar call. */ |
293 | int inchar_errno = 0; |
294 | /* Status for reading F-P nums. */ |
295 | char got_digit, got_dot, got_e, got_sign; |
296 | /* If a [...] is a [^...]. */ |
297 | CHAR_T not_in; |
298 | #define exp_char not_in |
299 | /* Base for integral numbers. */ |
300 | int base; |
301 | /* Decimal point character. */ |
302 | #ifdef COMPILE_WSCANF |
303 | wint_t decimal; |
304 | #else |
305 | const char *decimal; |
306 | #endif |
307 | /* The thousands character of the current locale. */ |
308 | #ifdef COMPILE_WSCANF |
309 | wint_t thousands; |
310 | #else |
311 | const char *thousands; |
312 | #endif |
313 | struct ptrs_to_free *ptrs_to_free = NULL; |
314 | /* State for the conversions. */ |
315 | mbstate_t state; |
316 | /* Integral holding variables. */ |
317 | union |
318 | { |
319 | long long int q; |
320 | unsigned long long int uq; |
321 | long int l; |
322 | unsigned long int ul; |
323 | } num; |
324 | /* Character-buffer pointer. */ |
325 | char *str = NULL; |
326 | wchar_t *wstr = NULL; |
327 | char **strptr = NULL; |
328 | ssize_t strsize = 0; |
329 | /* We must not react on white spaces immediately because they can |
330 | possibly be matched even if in the input stream no character is |
331 | available anymore. */ |
332 | int skip_space = 0; |
333 | /* Workspace. */ |
334 | CHAR_T *tw; /* Temporary pointer. */ |
335 | struct char_buffer charbuf; |
336 | scratch_buffer_init (&charbuf.scratch); |
337 | |
338 | #ifdef __va_copy |
339 | __va_copy (arg, argptr); |
340 | #else |
341 | arg = (va_list) argptr; |
342 | #endif |
343 | |
344 | #ifdef ORIENT |
345 | ORIENT; |
346 | #endif |
347 | |
348 | ARGCHECK (s, format); |
349 | |
350 | { |
351 | #ifndef COMPILE_WSCANF |
352 | struct __locale_data *const curnumeric = loc->__locales[LC_NUMERIC]; |
353 | #endif |
354 | |
355 | /* Figure out the decimal point character. */ |
356 | #ifdef COMPILE_WSCANF |
357 | decimal = _NL_CURRENT_WORD (LC_NUMERIC, _NL_NUMERIC_DECIMAL_POINT_WC); |
358 | #else |
359 | decimal = curnumeric->values[_NL_ITEM_INDEX (DECIMAL_POINT)].string; |
360 | #endif |
361 | /* Figure out the thousands separator character. */ |
362 | #ifdef COMPILE_WSCANF |
363 | thousands = _NL_CURRENT_WORD (LC_NUMERIC, _NL_NUMERIC_THOUSANDS_SEP_WC); |
364 | #else |
365 | thousands = curnumeric->values[_NL_ITEM_INDEX (THOUSANDS_SEP)].string; |
366 | if (*thousands == '\0') |
367 | thousands = NULL; |
368 | #endif |
369 | } |
370 | |
371 | /* Lock the stream. */ |
372 | LOCK_STREAM (s); |
373 | |
374 | |
375 | #ifndef COMPILE_WSCANF |
376 | /* From now on we use `state' to convert the format string. */ |
377 | memset (&state, '\0', sizeof (state)); |
378 | #endif |
379 | |
380 | /* Run through the format string. */ |
381 | while (*f != '\0') |
382 | { |
383 | unsigned int argpos; |
384 | /* Extract the next argument, which is of type TYPE. |
385 | For a %N$... spec, this is the Nth argument from the beginning; |
386 | otherwise it is the next argument after the state now in ARG. */ |
387 | #ifdef __va_copy |
388 | # define ARG(type) (argpos == 0 ? va_arg (arg, type) : \ |
389 | ({ unsigned int pos = argpos; \ |
390 | va_list arg; \ |
391 | __va_copy (arg, argptr); \ |
392 | while (--pos > 0) \ |
393 | (void) va_arg (arg, void *); \ |
394 | va_arg (arg, type); \ |
395 | })) |
396 | #else |
397 | # if 0 |
398 | /* XXX Possible optimization. */ |
399 | # define ARG(type) (argpos == 0 ? va_arg (arg, type) : \ |
400 | ({ va_list arg = (va_list) argptr; \ |
401 | arg = (va_list) ((char *) arg \ |
402 | + (argpos - 1) \ |
403 | * __va_rounded_size (void *)); \ |
404 | va_arg (arg, type); \ |
405 | })) |
406 | # else |
407 | # define ARG(type) (argpos == 0 ? va_arg (arg, type) : \ |
408 | ({ unsigned int pos = argpos; \ |
409 | va_list arg = (va_list) argptr; \ |
410 | while (--pos > 0) \ |
411 | (void) va_arg (arg, void *); \ |
412 | va_arg (arg, type); \ |
413 | })) |
414 | # endif |
415 | #endif |
416 | |
417 | #ifndef COMPILE_WSCANF |
418 | if (!isascii ((unsigned char) *f)) |
419 | { |
420 | /* Non-ASCII, may be a multibyte. */ |
421 | int len = __mbrlen (f, strlen (f), &state); |
422 | if (len > 0) |
423 | { |
424 | do |
425 | { |
426 | c = inchar (); |
427 | if (__glibc_unlikely (c == EOF)) |
428 | input_error (); |
429 | else if (c != (unsigned char) *f++) |
430 | { |
431 | ungetc_not_eof (c, s); |
432 | conv_error (); |
433 | } |
434 | } |
435 | while (--len > 0); |
436 | continue; |
437 | } |
438 | } |
439 | #endif |
440 | |
441 | fc = *f++; |
442 | if (fc != '%') |
443 | { |
444 | /* Remember to skip spaces. */ |
445 | if (ISSPACE (fc)) |
446 | { |
447 | skip_space = 1; |
448 | continue; |
449 | } |
450 | |
451 | /* Read a character. */ |
452 | c = inchar (); |
453 | |
454 | /* Characters other than format specs must just match. */ |
455 | if (__glibc_unlikely (c == EOF)) |
456 | input_error (); |
457 | |
458 | /* We saw white space char as the last character in the format |
459 | string. Now it's time to skip all leading white space. */ |
460 | if (skip_space) |
461 | { |
462 | while (ISSPACE (c)) |
463 | if (__glibc_unlikely (inchar () == EOF)) |
464 | input_error (); |
465 | skip_space = 0; |
466 | } |
467 | |
468 | if (__glibc_unlikely (c != fc)) |
469 | { |
470 | ungetc (c, s); |
471 | conv_error (); |
472 | } |
473 | |
474 | continue; |
475 | } |
476 | |
477 | /* This is the start of the conversion string. */ |
478 | flags = 0; |
479 | |
480 | /* Initialize state of modifiers. */ |
481 | argpos = 0; |
482 | |
483 | /* Prepare temporary buffer. */ |
484 | char_buffer_rewind (&charbuf); |
485 | |
486 | /* Check for a positional parameter specification. */ |
487 | if (ISDIGIT ((UCHAR_T) *f)) |
488 | { |
489 | argpos = read_int ((const UCHAR_T **) &f); |
490 | if (*f == L_('$')) |
491 | ++f; |
492 | else |
493 | { |
494 | /* Oops; that was actually the field width. */ |
495 | width = argpos; |
496 | argpos = 0; |
497 | goto got_width; |
498 | } |
499 | } |
500 | |
501 | /* Check for the assignment-suppressing, the number grouping flag, |
502 | and the signal to use the locale's digit representation. */ |
503 | while (*f == L_('*') || *f == L_('\'') || *f == L_('I')) |
504 | switch (*f++) |
505 | { |
506 | case L_('*'): |
507 | flags |= SUPPRESS; |
508 | break; |
509 | case L_('\''): |
510 | #ifdef COMPILE_WSCANF |
511 | if (thousands != L'\0') |
512 | #else |
513 | if (thousands != NULL) |
514 | #endif |
515 | flags |= GROUP; |
516 | break; |
517 | case L_('I'): |
518 | flags |= I18N; |
519 | break; |
520 | } |
521 | |
522 | /* Find the maximum field width. */ |
523 | width = 0; |
524 | if (ISDIGIT ((UCHAR_T) *f)) |
525 | width = read_int ((const UCHAR_T **) &f); |
526 | got_width: |
527 | if (width == 0) |
528 | width = -1; |
529 | |
530 | /* Check for type modifiers. */ |
531 | switch (*f++) |
532 | { |
533 | case L_('h'): |
534 | /* ints are short ints or chars. */ |
535 | if (*f == L_('h')) |
536 | { |
537 | ++f; |
538 | flags |= CHAR; |
539 | } |
540 | else |
541 | flags |= SHORT; |
542 | break; |
543 | case L_('l'): |
544 | if (*f == L_('l')) |
545 | { |
546 | /* A double `l' is equivalent to an `L'. */ |
547 | ++f; |
548 | flags |= LONGDBL | LONG; |
549 | } |
550 | else |
551 | /* ints are long ints. */ |
552 | flags |= LONG; |
553 | break; |
554 | case L_('q'): |
555 | case L_('L'): |
556 | /* doubles are long doubles, and ints are long long ints. */ |
557 | flags |= LONGDBL | LONG; |
558 | break; |
559 | case L_('a'): |
560 | /* The `a' is used as a flag only if followed by `s', `S' or |
561 | `['. */ |
562 | if (*f != L_('s') && *f != L_('S') && *f != L_('[')) |
563 | { |
564 | --f; |
565 | break; |
566 | } |
567 | /* In __isoc99_*scanf %as, %aS and %a[ extension is not |
568 | supported at all. */ |
569 | if (s->_flags2 & _IO_FLAGS2_SCANF_STD) |
570 | { |
571 | --f; |
572 | break; |
573 | } |
574 | /* String conversions (%s, %[) take a `char **' |
575 | arg and fill it in with a malloc'd pointer. */ |
576 | flags |= GNU_MALLOC; |
577 | break; |
578 | case L_('m'): |
579 | flags |= POSIX_MALLOC; |
580 | if (*f == L_('l')) |
581 | { |
582 | ++f; |
583 | flags |= LONG; |
584 | } |
585 | break; |
586 | case L_('z'): |
587 | if (need_longlong && sizeof (size_t) > sizeof (unsigned long int)) |
588 | flags |= LONGDBL; |
589 | else if (sizeof (size_t) > sizeof (unsigned int)) |
590 | flags |= LONG; |
591 | break; |
592 | case L_('j'): |
593 | if (need_longlong && sizeof (uintmax_t) > sizeof (unsigned long int)) |
594 | flags |= LONGDBL; |
595 | else if (sizeof (uintmax_t) > sizeof (unsigned int)) |
596 | flags |= LONG; |
597 | break; |
598 | case L_('t'): |
599 | if (need_longlong && sizeof (ptrdiff_t) > sizeof (long int)) |
600 | flags |= LONGDBL; |
601 | else if (sizeof (ptrdiff_t) > sizeof (int)) |
602 | flags |= LONG; |
603 | break; |
604 | default: |
605 | /* Not a recognized modifier. Backup. */ |
606 | --f; |
607 | break; |
608 | } |
609 | |
610 | /* End of the format string? */ |
611 | if (__glibc_unlikely (*f == L_('\0'))) |
612 | conv_error (); |
613 | |
614 | /* Find the conversion specifier. */ |
615 | fc = *f++; |
616 | if (skip_space || (fc != L_('[') && fc != L_('c') |
617 | && fc != L_('C') && fc != L_('n'))) |
618 | { |
619 | /* Eat whitespace. */ |
620 | int save_errno = errno; |
621 | __set_errno (0); |
622 | do |
623 | /* We add the additional test for EOF here since otherwise |
624 | inchar will restore the old errno value which might be |
625 | EINTR but does not indicate an interrupt since nothing |
626 | was read at this time. */ |
627 | if (__builtin_expect ((c == EOF || inchar () == EOF) |
628 | && errno == EINTR, 0)) |
629 | input_error (); |
630 | while (ISSPACE (c)); |
631 | __set_errno (save_errno); |
632 | ungetc (c, s); |
633 | skip_space = 0; |
634 | } |
635 | |
636 | switch (fc) |
637 | { |
638 | case L_('%'): /* Must match a literal '%'. */ |
639 | c = inchar (); |
640 | if (__glibc_unlikely (c == EOF)) |
641 | input_error (); |
642 | if (__glibc_unlikely (c != fc)) |
643 | { |
644 | ungetc_not_eof (c, s); |
645 | conv_error (); |
646 | } |
647 | break; |
648 | |
649 | case L_('n'): /* Answer number of assignments done. */ |
650 | /* Corrigendum 1 to ISO C 1990 describes the allowed flags |
651 | with the 'n' conversion specifier. */ |
652 | if (!(flags & SUPPRESS)) |
653 | { |
654 | /* Don't count the read-ahead. */ |
655 | if (need_longlong && (flags & LONGDBL)) |
656 | *ARG (long long int *) = read_in; |
657 | else if (need_long && (flags & LONG)) |
658 | *ARG (long int *) = read_in; |
659 | else if (flags & SHORT) |
660 | *ARG (short int *) = read_in; |
661 | else if (!(flags & CHAR)) |
662 | *ARG (int *) = read_in; |
663 | else |
664 | *ARG (char *) = read_in; |
665 | |
666 | #ifdef NO_BUG_IN_ISO_C_CORRIGENDUM_1 |
667 | /* We have a severe problem here. The ISO C standard |
668 | contradicts itself in explaining the effect of the %n |
669 | format in `scanf'. While in ISO C:1990 and the ISO C |
670 | Amendement 1:1995 the result is described as |
671 | |
672 | Execution of a %n directive does not effect the |
673 | assignment count returned at the completion of |
674 | execution of the f(w)scanf function. |
675 | |
676 | in ISO C Corrigendum 1:1994 the following was added: |
677 | |
678 | Subclause 7.9.6.2 |
679 | Add the following fourth example: |
680 | In: |
681 | #include <stdio.h> |
682 | int d1, d2, n1, n2, i; |
683 | i = sscanf("123", "%d%n%n%d", &d1, &n1, &n2, &d2); |
684 | the value 123 is assigned to d1 and the value3 to n1. |
685 | Because %n can never get an input failure the value |
686 | of 3 is also assigned to n2. The value of d2 is not |
687 | affected. The value 3 is assigned to i. |
688 | |
689 | We go for now with the historically correct code from ISO C, |
690 | i.e., we don't count the %n assignments. When it ever |
691 | should proof to be wrong just remove the #ifdef above. */ |
692 | ++done; |
693 | #endif |
694 | } |
695 | break; |
696 | |
697 | case L_('c'): /* Match characters. */ |
698 | if ((flags & LONG) == 0) |
699 | { |
700 | if (width == -1) |
701 | width = 1; |
702 | |
703 | #define STRING_ARG(Str, Type, Width) \ |
704 | do if (!(flags & SUPPRESS)) \ |
705 | { \ |
706 | if (flags & MALLOC) \ |
707 | { \ |
708 | /* The string is to be stored in a malloc'd buffer. */ \ |
709 | /* For %mS using char ** is actually wrong, but \ |
710 | shouldn't make a difference on any arch glibc \ |
711 | supports and would unnecessarily complicate \ |
712 | things. */ \ |
713 | strptr = ARG (char **); \ |
714 | if (strptr == NULL) \ |
715 | conv_error (); \ |
716 | /* Allocate an initial buffer. */ \ |
717 | strsize = Width; \ |
718 | *strptr = (char *) malloc (strsize * sizeof (Type)); \ |
719 | Str = (Type *) *strptr; \ |
720 | if (Str != NULL) \ |
721 | add_ptr_to_free (strptr); \ |
722 | else if (flags & POSIX_MALLOC) \ |
723 | { \ |
724 | done = EOF; \ |
725 | goto errout; \ |
726 | } \ |
727 | } \ |
728 | else \ |
729 | Str = ARG (Type *); \ |
730 | if (Str == NULL) \ |
731 | conv_error (); \ |
732 | } while (0) |
733 | #ifdef COMPILE_WSCANF |
734 | STRING_ARG (str, char, 100); |
735 | #else |
736 | STRING_ARG (str, char, (width > 1024 ? 1024 : width)); |
737 | #endif |
738 | |
739 | c = inchar (); |
740 | if (__glibc_unlikely (c == EOF)) |
741 | input_error (); |
742 | |
743 | #ifdef COMPILE_WSCANF |
744 | /* We have to convert the wide character(s) into multibyte |
745 | characters and store the result. */ |
746 | memset (&state, '\0', sizeof (state)); |
747 | |
748 | do |
749 | { |
750 | size_t n; |
751 | |
752 | if (!(flags & SUPPRESS) && (flags & POSIX_MALLOC) |
753 | && *strptr + strsize - str <= MB_LEN_MAX) |
754 | { |
755 | /* We have to enlarge the buffer if the `m' flag |
756 | was given. */ |
757 | size_t strleng = str - *strptr; |
758 | char *newstr; |
759 | |
760 | newstr = (char *) realloc (*strptr, strsize * 2); |
761 | if (newstr == NULL) |
762 | { |
763 | /* Can't allocate that much. Last-ditch effort. */ |
764 | newstr = (char *) realloc (*strptr, |
765 | strleng + MB_LEN_MAX); |
766 | if (newstr == NULL) |
767 | { |
768 | /* c can't have `a' flag, only `m'. */ |
769 | done = EOF; |
770 | goto errout; |
771 | } |
772 | else |
773 | { |
774 | *strptr = newstr; |
775 | str = newstr + strleng; |
776 | strsize = strleng + MB_LEN_MAX; |
777 | } |
778 | } |
779 | else |
780 | { |
781 | *strptr = newstr; |
782 | str = newstr + strleng; |
783 | strsize *= 2; |
784 | } |
785 | } |
786 | |
787 | n = __wcrtomb (!(flags & SUPPRESS) ? str : NULL, c, &state); |
788 | if (__glibc_unlikely (n == (size_t) -1)) |
789 | /* No valid wide character. */ |
790 | input_error (); |
791 | |
792 | /* Increment the output pointer. Even if we don't |
793 | write anything. */ |
794 | str += n; |
795 | } |
796 | while (--width > 0 && inchar () != EOF); |
797 | #else |
798 | if (!(flags & SUPPRESS)) |
799 | { |
800 | do |
801 | { |
802 | if ((flags & MALLOC) |
803 | && (char *) str == *strptr + strsize) |
804 | { |
805 | /* Enlarge the buffer. */ |
806 | size_t newsize |
807 | = strsize |
808 | + (strsize >= width ? width - 1 : strsize); |
809 | |
810 | str = (char *) realloc (*strptr, newsize); |
811 | if (str == NULL) |
812 | { |
813 | /* Can't allocate that much. Last-ditch |
814 | effort. */ |
815 | str = (char *) realloc (*strptr, strsize + 1); |
816 | if (str == NULL) |
817 | { |
818 | /* c can't have `a' flag, only `m'. */ |
819 | done = EOF; |
820 | goto errout; |
821 | } |
822 | else |
823 | { |
824 | *strptr = (char *) str; |
825 | str += strsize; |
826 | ++strsize; |
827 | } |
828 | } |
829 | else |
830 | { |
831 | *strptr = (char *) str; |
832 | str += strsize; |
833 | strsize = newsize; |
834 | } |
835 | } |
836 | *str++ = c; |
837 | } |
838 | while (--width > 0 && inchar () != EOF); |
839 | } |
840 | else |
841 | while (--width > 0 && inchar () != EOF); |
842 | #endif |
843 | |
844 | if (!(flags & SUPPRESS)) |
845 | { |
846 | if ((flags & MALLOC) && str - *strptr != strsize) |
847 | { |
848 | char *cp = (char *) realloc (*strptr, str - *strptr); |
849 | if (cp != NULL) |
850 | *strptr = cp; |
851 | } |
852 | strptr = NULL; |
853 | ++done; |
854 | } |
855 | |
856 | break; |
857 | } |
858 | /* FALLTHROUGH */ |
859 | case L_('C'): |
860 | if (width == -1) |
861 | width = 1; |
862 | |
863 | STRING_ARG (wstr, wchar_t, (width > 1024 ? 1024 : width)); |
864 | |
865 | c = inchar (); |
866 | if (__glibc_unlikely (c == EOF)) |
867 | input_error (); |
868 | |
869 | #ifdef COMPILE_WSCANF |
870 | /* Just store the incoming wide characters. */ |
871 | if (!(flags & SUPPRESS)) |
872 | { |
873 | do |
874 | { |
875 | if ((flags & MALLOC) |
876 | && wstr == (wchar_t *) *strptr + strsize) |
877 | { |
878 | size_t newsize |
879 | = strsize + (strsize > width ? width - 1 : strsize); |
880 | /* Enlarge the buffer. */ |
881 | wstr = (wchar_t *) realloc (*strptr, |
882 | newsize * sizeof (wchar_t)); |
883 | if (wstr == NULL) |
884 | { |
885 | /* Can't allocate that much. Last-ditch effort. */ |
886 | wstr = (wchar_t *) realloc (*strptr, |
887 | (strsize + 1) |
888 | * sizeof (wchar_t)); |
889 | if (wstr == NULL) |
890 | { |
891 | /* C or lc can't have `a' flag, only `m' |
892 | flag. */ |
893 | done = EOF; |
894 | goto errout; |
895 | } |
896 | else |
897 | { |
898 | *strptr = (char *) wstr; |
899 | wstr += strsize; |
900 | ++strsize; |
901 | } |
902 | } |
903 | else |
904 | { |
905 | *strptr = (char *) wstr; |
906 | wstr += strsize; |
907 | strsize = newsize; |
908 | } |
909 | } |
910 | *wstr++ = c; |
911 | } |
912 | while (--width > 0 && inchar () != EOF); |
913 | } |
914 | else |
915 | while (--width > 0 && inchar () != EOF); |
916 | #else |
917 | { |
918 | /* We have to convert the multibyte input sequence to wide |
919 | characters. */ |
920 | char buf[1]; |
921 | mbstate_t cstate; |
922 | |
923 | memset (&cstate, '\0', sizeof (cstate)); |
924 | |
925 | do |
926 | { |
927 | /* This is what we present the mbrtowc function first. */ |
928 | buf[0] = c; |
929 | |
930 | if (!(flags & SUPPRESS) && (flags & MALLOC) |
931 | && wstr == (wchar_t *) *strptr + strsize) |
932 | { |
933 | size_t newsize |
934 | = strsize + (strsize > width ? width - 1 : strsize); |
935 | /* Enlarge the buffer. */ |
936 | wstr = (wchar_t *) realloc (*strptr, |
937 | newsize * sizeof (wchar_t)); |
938 | if (wstr == NULL) |
939 | { |
940 | /* Can't allocate that much. Last-ditch effort. */ |
941 | wstr = (wchar_t *) realloc (*strptr, |
942 | ((strsize + 1) |
943 | * sizeof (wchar_t))); |
944 | if (wstr == NULL) |
945 | { |
946 | /* C or lc can't have `a' flag, only `m' flag. */ |
947 | done = EOF; |
948 | goto errout; |
949 | } |
950 | else |
951 | { |
952 | *strptr = (char *) wstr; |
953 | wstr += strsize; |
954 | ++strsize; |
955 | } |
956 | } |
957 | else |
958 | { |
959 | *strptr = (char *) wstr; |
960 | wstr += strsize; |
961 | strsize = newsize; |
962 | } |
963 | } |
964 | |
965 | while (1) |
966 | { |
967 | size_t n; |
968 | |
969 | n = __mbrtowc (!(flags & SUPPRESS) ? wstr : NULL, |
970 | buf, 1, &cstate); |
971 | |
972 | if (n == (size_t) -2) |
973 | { |
974 | /* Possibly correct character, just not enough |
975 | input. */ |
976 | if (__glibc_unlikely (inchar () == EOF)) |
977 | encode_error (); |
978 | |
979 | buf[0] = c; |
980 | continue; |
981 | } |
982 | |
983 | if (__glibc_unlikely (n != 1)) |
984 | encode_error (); |
985 | |
986 | /* We have a match. */ |
987 | break; |
988 | } |
989 | |
990 | /* Advance the result pointer. */ |
991 | ++wstr; |
992 | } |
993 | while (--width > 0 && inchar () != EOF); |
994 | } |
995 | #endif |
996 | |
997 | if (!(flags & SUPPRESS)) |
998 | { |
999 | if ((flags & MALLOC) && wstr - (wchar_t *) *strptr != strsize) |
1000 | { |
1001 | wchar_t *cp = (wchar_t *) realloc (*strptr, |
1002 | ((wstr |
1003 | - (wchar_t *) *strptr) |
1004 | * sizeof (wchar_t))); |
1005 | if (cp != NULL) |
1006 | *strptr = (char *) cp; |
1007 | } |
1008 | strptr = NULL; |
1009 | |
1010 | ++done; |
1011 | } |
1012 | |
1013 | break; |
1014 | |
1015 | case L_('s'): /* Read a string. */ |
1016 | if (!(flags & LONG)) |
1017 | { |
1018 | STRING_ARG (str, char, 100); |
1019 | |
1020 | c = inchar (); |
1021 | if (__glibc_unlikely (c == EOF)) |
1022 | input_error (); |
1023 | |
1024 | #ifdef COMPILE_WSCANF |
1025 | memset (&state, '\0', sizeof (state)); |
1026 | #endif |
1027 | |
1028 | do |
1029 | { |
1030 | if (ISSPACE (c)) |
1031 | { |
1032 | ungetc_not_eof (c, s); |
1033 | break; |
1034 | } |
1035 | |
1036 | #ifdef COMPILE_WSCANF |
1037 | /* This is quite complicated. We have to convert the |
1038 | wide characters into multibyte characters and then |
1039 | store them. */ |
1040 | { |
1041 | size_t n; |
1042 | |
1043 | if (!(flags & SUPPRESS) && (flags & MALLOC) |
1044 | && *strptr + strsize - str <= MB_LEN_MAX) |
1045 | { |
1046 | /* We have to enlarge the buffer if the `a' or `m' |
1047 | flag was given. */ |
1048 | size_t strleng = str - *strptr; |
1049 | char *newstr; |
1050 | |
1051 | newstr = (char *) realloc (*strptr, strsize * 2); |
1052 | if (newstr == NULL) |
1053 | { |
1054 | /* Can't allocate that much. Last-ditch |
1055 | effort. */ |
1056 | newstr = (char *) realloc (*strptr, |
1057 | strleng + MB_LEN_MAX); |
1058 | if (newstr == NULL) |
1059 | { |
1060 | if (flags & POSIX_MALLOC) |
1061 | { |
1062 | done = EOF; |
1063 | goto errout; |
1064 | } |
1065 | /* We lose. Oh well. Terminate the |
1066 | string and stop converting, |
1067 | so at least we don't skip any input. */ |
1068 | ((char *) (*strptr))[strleng] = '\0'; |
1069 | strptr = NULL; |
1070 | ++done; |
1071 | conv_error (); |
1072 | } |
1073 | else |
1074 | { |
1075 | *strptr = newstr; |
1076 | str = newstr + strleng; |
1077 | strsize = strleng + MB_LEN_MAX; |
1078 | } |
1079 | } |
1080 | else |
1081 | { |
1082 | *strptr = newstr; |
1083 | str = newstr + strleng; |
1084 | strsize *= 2; |
1085 | } |
1086 | } |
1087 | |
1088 | n = __wcrtomb (!(flags & SUPPRESS) ? str : NULL, c, |
1089 | &state); |
1090 | if (__glibc_unlikely (n == (size_t) -1)) |
1091 | encode_error (); |
1092 | |
1093 | assert (n <= MB_LEN_MAX); |
1094 | str += n; |
1095 | } |
1096 | #else |
1097 | /* This is easy. */ |
1098 | if (!(flags & SUPPRESS)) |
1099 | { |
1100 | *str++ = c; |
1101 | if ((flags & MALLOC) |
1102 | && (char *) str == *strptr + strsize) |
1103 | { |
1104 | /* Enlarge the buffer. */ |
1105 | str = (char *) realloc (*strptr, 2 * strsize); |
1106 | if (str == NULL) |
1107 | { |
1108 | /* Can't allocate that much. Last-ditch |
1109 | effort. */ |
1110 | str = (char *) realloc (*strptr, strsize + 1); |
1111 | if (str == NULL) |
1112 | { |
1113 | if (flags & POSIX_MALLOC) |
1114 | { |
1115 | done = EOF; |
1116 | goto errout; |
1117 | } |
1118 | /* We lose. Oh well. Terminate the |
1119 | string and stop converting, |
1120 | so at least we don't skip any input. */ |
1121 | ((char *) (*strptr))[strsize - 1] = '\0'; |
1122 | strptr = NULL; |
1123 | ++done; |
1124 | conv_error (); |
1125 | } |
1126 | else |
1127 | { |
1128 | *strptr = (char *) str; |
1129 | str += strsize; |
1130 | ++strsize; |
1131 | } |
1132 | } |
1133 | else |
1134 | { |
1135 | *strptr = (char *) str; |
1136 | str += strsize; |
1137 | strsize *= 2; |
1138 | } |
1139 | } |
1140 | } |
1141 | #endif |
1142 | } |
1143 | while ((width <= 0 || --width > 0) && inchar () != EOF); |
1144 | |
1145 | if (!(flags & SUPPRESS)) |
1146 | { |
1147 | #ifdef COMPILE_WSCANF |
1148 | /* We have to emit the code to get into the initial |
1149 | state. */ |
1150 | char buf[MB_LEN_MAX]; |
1151 | size_t n = __wcrtomb (buf, L'\0', &state); |
1152 | if (n > 0 && (flags & MALLOC) |
1153 | && str + n >= *strptr + strsize) |
1154 | { |
1155 | /* Enlarge the buffer. */ |
1156 | size_t strleng = str - *strptr; |
1157 | char *newstr; |
1158 | |
1159 | newstr = (char *) realloc (*strptr, strleng + n + 1); |
1160 | if (newstr == NULL) |
1161 | { |
1162 | if (flags & POSIX_MALLOC) |
1163 | { |
1164 | done = EOF; |
1165 | goto errout; |
1166 | } |
1167 | /* We lose. Oh well. Terminate the string |
1168 | and stop converting, so at least we don't |
1169 | skip any input. */ |
1170 | ((char *) (*strptr))[strleng] = '\0'; |
1171 | strptr = NULL; |
1172 | ++done; |
1173 | conv_error (); |
1174 | } |
1175 | else |
1176 | { |
1177 | *strptr = newstr; |
1178 | str = newstr + strleng; |
1179 | strsize = strleng + n + 1; |
1180 | } |
1181 | } |
1182 | |
1183 | str = __mempcpy (str, buf, n); |
1184 | #endif |
1185 | *str++ = '\0'; |
1186 | |
1187 | if ((flags & MALLOC) && str - *strptr != strsize) |
1188 | { |
1189 | char *cp = (char *) realloc (*strptr, str - *strptr); |
1190 | if (cp != NULL) |
1191 | *strptr = cp; |
1192 | } |
1193 | strptr = NULL; |
1194 | |
1195 | ++done; |
1196 | } |
1197 | break; |
1198 | } |
1199 | /* FALLTHROUGH */ |
1200 | |
1201 | case L_('S'): |
1202 | { |
1203 | #ifndef COMPILE_WSCANF |
1204 | mbstate_t cstate; |
1205 | #endif |
1206 | |
1207 | /* Wide character string. */ |
1208 | STRING_ARG (wstr, wchar_t, 100); |
1209 | |
1210 | c = inchar (); |
1211 | if (__builtin_expect (c == EOF, 0)) |
1212 | input_error (); |
1213 | |
1214 | #ifndef COMPILE_WSCANF |
1215 | memset (&cstate, '\0', sizeof (cstate)); |
1216 | #endif |
1217 | |
1218 | do |
1219 | { |
1220 | if (ISSPACE (c)) |
1221 | { |
1222 | ungetc_not_eof (c, s); |
1223 | break; |
1224 | } |
1225 | |
1226 | #ifdef COMPILE_WSCANF |
1227 | /* This is easy. */ |
1228 | if (!(flags & SUPPRESS)) |
1229 | { |
1230 | *wstr++ = c; |
1231 | if ((flags & MALLOC) |
1232 | && wstr == (wchar_t *) *strptr + strsize) |
1233 | { |
1234 | /* Enlarge the buffer. */ |
1235 | wstr = (wchar_t *) realloc (*strptr, |
1236 | (2 * strsize) |
1237 | * sizeof (wchar_t)); |
1238 | if (wstr == NULL) |
1239 | { |
1240 | /* Can't allocate that much. Last-ditch |
1241 | effort. */ |
1242 | wstr = (wchar_t *) realloc (*strptr, |
1243 | (strsize + 1) |
1244 | * sizeof (wchar_t)); |
1245 | if (wstr == NULL) |
1246 | { |
1247 | if (flags & POSIX_MALLOC) |
1248 | { |
1249 | done = EOF; |
1250 | goto errout; |
1251 | } |
1252 | /* We lose. Oh well. Terminate the string |
1253 | and stop converting, so at least we don't |
1254 | skip any input. */ |
1255 | ((wchar_t *) (*strptr))[strsize - 1] = L'\0'; |
1256 | strptr = NULL; |
1257 | ++done; |
1258 | conv_error (); |
1259 | } |
1260 | else |
1261 | { |
1262 | *strptr = (char *) wstr; |
1263 | wstr += strsize; |
1264 | ++strsize; |
1265 | } |
1266 | } |
1267 | else |
1268 | { |
1269 | *strptr = (char *) wstr; |
1270 | wstr += strsize; |
1271 | strsize *= 2; |
1272 | } |
1273 | } |
1274 | } |
1275 | #else |
1276 | { |
1277 | char buf[1]; |
1278 | |
1279 | buf[0] = c; |
1280 | |
1281 | while (1) |
1282 | { |
1283 | size_t n; |
1284 | |
1285 | n = __mbrtowc (!(flags & SUPPRESS) ? wstr : NULL, |
1286 | buf, 1, &cstate); |
1287 | |
1288 | if (n == (size_t) -2) |
1289 | { |
1290 | /* Possibly correct character, just not enough |
1291 | input. */ |
1292 | if (__glibc_unlikely (inchar () == EOF)) |
1293 | encode_error (); |
1294 | |
1295 | buf[0] = c; |
1296 | continue; |
1297 | } |
1298 | |
1299 | if (__glibc_unlikely (n != 1)) |
1300 | encode_error (); |
1301 | |
1302 | /* We have a match. */ |
1303 | ++wstr; |
1304 | break; |
1305 | } |
1306 | |
1307 | if (!(flags & SUPPRESS) && (flags & MALLOC) |
1308 | && wstr == (wchar_t *) *strptr + strsize) |
1309 | { |
1310 | /* Enlarge the buffer. */ |
1311 | wstr = (wchar_t *) realloc (*strptr, |
1312 | (2 * strsize |
1313 | * sizeof (wchar_t))); |
1314 | if (wstr == NULL) |
1315 | { |
1316 | /* Can't allocate that much. Last-ditch effort. */ |
1317 | wstr = (wchar_t *) realloc (*strptr, |
1318 | ((strsize + 1) |
1319 | * sizeof (wchar_t))); |
1320 | if (wstr == NULL) |
1321 | { |
1322 | if (flags & POSIX_MALLOC) |
1323 | { |
1324 | done = EOF; |
1325 | goto errout; |
1326 | } |
1327 | /* We lose. Oh well. Terminate the |
1328 | string and stop converting, so at |
1329 | least we don't skip any input. */ |
1330 | ((wchar_t *) (*strptr))[strsize - 1] = L'\0'; |
1331 | strptr = NULL; |
1332 | ++done; |
1333 | conv_error (); |
1334 | } |
1335 | else |
1336 | { |
1337 | *strptr = (char *) wstr; |
1338 | wstr += strsize; |
1339 | ++strsize; |
1340 | } |
1341 | } |
1342 | else |
1343 | { |
1344 | *strptr = (char *) wstr; |
1345 | wstr += strsize; |
1346 | strsize *= 2; |
1347 | } |
1348 | } |
1349 | } |
1350 | #endif |
1351 | } |
1352 | while ((width <= 0 || --width > 0) && inchar () != EOF); |
1353 | |
1354 | if (!(flags & SUPPRESS)) |
1355 | { |
1356 | *wstr++ = L'\0'; |
1357 | |
1358 | if ((flags & MALLOC) && wstr - (wchar_t *) *strptr != strsize) |
1359 | { |
1360 | wchar_t *cp = (wchar_t *) realloc (*strptr, |
1361 | ((wstr |
1362 | - (wchar_t *) *strptr) |
1363 | * sizeof(wchar_t))); |
1364 | if (cp != NULL) |
1365 | *strptr = (char *) cp; |
1366 | } |
1367 | strptr = NULL; |
1368 | |
1369 | ++done; |
1370 | } |
1371 | } |
1372 | break; |
1373 | |
1374 | case L_('x'): /* Hexadecimal integer. */ |
1375 | case L_('X'): /* Ditto. */ |
1376 | base = 16; |
1377 | goto number; |
1378 | |
1379 | case L_('o'): /* Octal integer. */ |
1380 | base = 8; |
1381 | goto number; |
1382 | |
1383 | case L_('u'): /* Unsigned decimal integer. */ |
1384 | base = 10; |
1385 | goto number; |
1386 | |
1387 | case L_('d'): /* Signed decimal integer. */ |
1388 | base = 10; |
1389 | flags |= NUMBER_SIGNED; |
1390 | goto number; |
1391 | |
1392 | case L_('i'): /* Generic number. */ |
1393 | base = 0; |
1394 | flags |= NUMBER_SIGNED; |
1395 | |
1396 | number: |
1397 | c = inchar (); |
1398 | if (__glibc_unlikely (c == EOF)) |
1399 | input_error (); |
1400 | |
1401 | /* Check for a sign. */ |
1402 | if (c == L_('-') || c == L_('+')) |
1403 | { |
1404 | char_buffer_add (&charbuf, c); |
1405 | if (width > 0) |
1406 | --width; |
1407 | c = inchar (); |
1408 | } |
1409 | |
1410 | /* Look for a leading indication of base. */ |
1411 | if (width != 0 && c == L_('0')) |
1412 | { |
1413 | if (width > 0) |
1414 | --width; |
1415 | |
1416 | char_buffer_add (&charbuf, c); |
1417 | c = inchar (); |
1418 | |
1419 | if (width != 0 && TOLOWER (c) == L_('x')) |
1420 | { |
1421 | if (base == 0) |
1422 | base = 16; |
1423 | if (base == 16) |
1424 | { |
1425 | if (width > 0) |
1426 | --width; |
1427 | c = inchar (); |
1428 | } |
1429 | } |
1430 | else if (base == 0) |
1431 | base = 8; |
1432 | } |
1433 | |
1434 | if (base == 0) |
1435 | base = 10; |
1436 | |
1437 | if (base == 10 && __builtin_expect ((flags & I18N) != 0, 0)) |
1438 | { |
1439 | int from_level; |
1440 | int to_level; |
1441 | int level; |
1442 | #ifdef COMPILE_WSCANF |
1443 | const wchar_t *wcdigits[10]; |
1444 | const wchar_t *wcdigits_extended[10]; |
1445 | #else |
1446 | const char *mbdigits[10]; |
1447 | const char *mbdigits_extended[10]; |
1448 | #endif |
1449 | /* "to_inpunct" is a map from ASCII digits to their |
1450 | equivalent in locale. This is defined for locales |
1451 | which use an extra digits set. */ |
1452 | wctrans_t map = __wctrans ("to_inpunct" ); |
1453 | int n; |
1454 | |
1455 | from_level = 0; |
1456 | #ifdef COMPILE_WSCANF |
1457 | to_level = _NL_CURRENT_WORD (LC_CTYPE, |
1458 | _NL_CTYPE_INDIGITS_WC_LEN) - 1; |
1459 | #else |
1460 | to_level = (uint32_t) curctype->values[_NL_ITEM_INDEX (_NL_CTYPE_INDIGITS_MB_LEN)].word - 1; |
1461 | #endif |
1462 | |
1463 | /* Get the alternative digit forms if there are any. */ |
1464 | if (__glibc_unlikely (map != NULL)) |
1465 | { |
1466 | /* Adding new level for extra digits set in locale file. */ |
1467 | ++to_level; |
1468 | |
1469 | for (n = 0; n < 10; ++n) |
1470 | { |
1471 | #ifdef COMPILE_WSCANF |
1472 | wcdigits[n] = (const wchar_t *) |
1473 | _NL_CURRENT (LC_CTYPE, _NL_CTYPE_INDIGITS0_WC + n); |
1474 | |
1475 | wchar_t *wc_extended = (wchar_t *) |
1476 | alloca ((to_level + 2) * sizeof (wchar_t)); |
1477 | __wmemcpy (wc_extended, wcdigits[n], to_level); |
1478 | wc_extended[to_level] = __towctrans (L'0' + n, map); |
1479 | wc_extended[to_level + 1] = '\0'; |
1480 | wcdigits_extended[n] = wc_extended; |
1481 | #else |
1482 | mbdigits[n] |
1483 | = curctype->values[_NL_CTYPE_INDIGITS0_MB + n].string; |
1484 | |
1485 | /* Get the equivalent wide char in map. */ |
1486 | wint_t = __towctrans (L'0' + n, map); |
1487 | |
1488 | /* Convert it to multibyte representation. */ |
1489 | mbstate_t state; |
1490 | memset (&state, '\0', sizeof (state)); |
1491 | |
1492 | char [MB_LEN_MAX]; |
1493 | size_t mblen |
1494 | = __wcrtomb (extra_mbdigit, extra_wcdigit, &state); |
1495 | |
1496 | if (mblen == (size_t) -1) |
1497 | { |
1498 | /* Ignore this new level. */ |
1499 | map = NULL; |
1500 | break; |
1501 | } |
1502 | |
1503 | /* Calculate the length of mbdigits[n]. */ |
1504 | const char *last_char = mbdigits[n]; |
1505 | for (level = 0; level < to_level; ++level) |
1506 | last_char = strchr (last_char, '\0') + 1; |
1507 | |
1508 | size_t mbdigits_len = last_char - mbdigits[n]; |
1509 | |
1510 | /* Allocate memory for extended multibyte digit. */ |
1511 | char *mb_extended; |
1512 | mb_extended = (char *) alloca (mbdigits_len + mblen + 1); |
1513 | |
1514 | /* And get the mbdigits + extra_digit string. */ |
1515 | *(char *) __mempcpy (__mempcpy (mb_extended, mbdigits[n], |
1516 | mbdigits_len), |
1517 | extra_mbdigit, mblen) = '\0'; |
1518 | mbdigits_extended[n] = mb_extended; |
1519 | #endif |
1520 | } |
1521 | } |
1522 | |
1523 | /* Read the number into workspace. */ |
1524 | while (c != EOF && width != 0) |
1525 | { |
1526 | /* In this round we get the pointer to the digit strings |
1527 | and also perform the first round of comparisons. */ |
1528 | for (n = 0; n < 10; ++n) |
1529 | { |
1530 | /* Get the string for the digits with value N. */ |
1531 | #ifdef COMPILE_WSCANF |
1532 | |
1533 | /* wcdigits_extended[] is fully set in the loop |
1534 | above, but the test for "map != NULL" is done |
1535 | inside the loop here and outside the loop there. */ |
1536 | DIAG_PUSH_NEEDS_COMMENT; |
1537 | DIAG_IGNORE_NEEDS_COMMENT (4.7, "-Wmaybe-uninitialized" ); |
1538 | |
1539 | if (__glibc_unlikely (map != NULL)) |
1540 | wcdigits[n] = wcdigits_extended[n]; |
1541 | else |
1542 | wcdigits[n] = (const wchar_t *) |
1543 | _NL_CURRENT (LC_CTYPE, _NL_CTYPE_INDIGITS0_WC + n); |
1544 | wcdigits[n] += from_level; |
1545 | |
1546 | DIAG_POP_NEEDS_COMMENT; |
1547 | |
1548 | if (c == (wint_t) *wcdigits[n]) |
1549 | { |
1550 | to_level = from_level; |
1551 | break; |
1552 | } |
1553 | |
1554 | /* Advance the pointer to the next string. */ |
1555 | ++wcdigits[n]; |
1556 | #else |
1557 | const char *cmpp; |
1558 | int avail = width > 0 ? width : INT_MAX; |
1559 | |
1560 | if (__glibc_unlikely (map != NULL)) |
1561 | mbdigits[n] = mbdigits_extended[n]; |
1562 | else |
1563 | mbdigits[n] |
1564 | = curctype->values[_NL_CTYPE_INDIGITS0_MB + n].string; |
1565 | |
1566 | for (level = 0; level < from_level; level++) |
1567 | mbdigits[n] = strchr (mbdigits[n], '\0') + 1; |
1568 | |
1569 | cmpp = mbdigits[n]; |
1570 | while ((unsigned char) *cmpp == c && avail >= 0) |
1571 | { |
1572 | if (*++cmpp == '\0') |
1573 | break; |
1574 | else |
1575 | { |
1576 | if (avail == 0 || inchar () == EOF) |
1577 | break; |
1578 | --avail; |
1579 | } |
1580 | } |
1581 | |
1582 | if (*cmpp == '\0') |
1583 | { |
1584 | if (width > 0) |
1585 | width = avail; |
1586 | to_level = from_level; |
1587 | break; |
1588 | } |
1589 | |
1590 | /* We are pushing all read characters back. */ |
1591 | if (cmpp > mbdigits[n]) |
1592 | { |
1593 | ungetc (c, s); |
1594 | while (--cmpp > mbdigits[n]) |
1595 | ungetc_not_eof ((unsigned char) *cmpp, s); |
1596 | c = (unsigned char) *cmpp; |
1597 | } |
1598 | |
1599 | /* Advance the pointer to the next string. */ |
1600 | mbdigits[n] = strchr (mbdigits[n], '\0') + 1; |
1601 | #endif |
1602 | } |
1603 | |
1604 | if (n == 10) |
1605 | { |
1606 | /* Have not yet found the digit. */ |
1607 | for (level = from_level + 1; level <= to_level; ++level) |
1608 | { |
1609 | /* Search all ten digits of this level. */ |
1610 | for (n = 0; n < 10; ++n) |
1611 | { |
1612 | #ifdef COMPILE_WSCANF |
1613 | if (c == (wint_t) *wcdigits[n]) |
1614 | break; |
1615 | |
1616 | /* Advance the pointer to the next string. */ |
1617 | ++wcdigits[n]; |
1618 | #else |
1619 | const char *cmpp; |
1620 | int avail = width > 0 ? width : INT_MAX; |
1621 | |
1622 | cmpp = mbdigits[n]; |
1623 | while ((unsigned char) *cmpp == c && avail >= 0) |
1624 | { |
1625 | if (*++cmpp == '\0') |
1626 | break; |
1627 | else |
1628 | { |
1629 | if (avail == 0 || inchar () == EOF) |
1630 | break; |
1631 | --avail; |
1632 | } |
1633 | } |
1634 | |
1635 | if (*cmpp == '\0') |
1636 | { |
1637 | if (width > 0) |
1638 | width = avail; |
1639 | break; |
1640 | } |
1641 | |
1642 | /* We are pushing all read characters back. */ |
1643 | if (cmpp > mbdigits[n]) |
1644 | { |
1645 | ungetc (c, s); |
1646 | while (--cmpp > mbdigits[n]) |
1647 | ungetc_not_eof ((unsigned char) *cmpp, s); |
1648 | c = (unsigned char) *cmpp; |
1649 | } |
1650 | |
1651 | /* Advance the pointer to the next string. */ |
1652 | mbdigits[n] = strchr (mbdigits[n], '\0') + 1; |
1653 | #endif |
1654 | } |
1655 | |
1656 | if (n < 10) |
1657 | { |
1658 | /* Found it. */ |
1659 | from_level = level; |
1660 | to_level = level; |
1661 | break; |
1662 | } |
1663 | } |
1664 | } |
1665 | |
1666 | if (n < 10) |
1667 | c = L_('0') + n; |
1668 | else if (flags & GROUP) |
1669 | { |
1670 | /* Try matching against the thousands separator. */ |
1671 | #ifdef COMPILE_WSCANF |
1672 | if (c != thousands) |
1673 | break; |
1674 | #else |
1675 | const char *cmpp = thousands; |
1676 | int avail = width > 0 ? width : INT_MAX; |
1677 | |
1678 | while ((unsigned char) *cmpp == c && avail >= 0) |
1679 | { |
1680 | char_buffer_add (&charbuf, c); |
1681 | if (*++cmpp == '\0') |
1682 | break; |
1683 | else |
1684 | { |
1685 | if (avail == 0 || inchar () == EOF) |
1686 | break; |
1687 | --avail; |
1688 | } |
1689 | } |
1690 | |
1691 | if (char_buffer_error (&charbuf)) |
1692 | { |
1693 | __set_errno (ENOMEM); |
1694 | done = EOF; |
1695 | goto errout; |
1696 | } |
1697 | |
1698 | if (*cmpp != '\0') |
1699 | { |
1700 | /* We are pushing all read characters back. */ |
1701 | if (cmpp > thousands) |
1702 | { |
1703 | charbuf.current -= cmpp - thousands; |
1704 | ungetc (c, s); |
1705 | while (--cmpp > thousands) |
1706 | ungetc_not_eof ((unsigned char) *cmpp, s); |
1707 | c = (unsigned char) *cmpp; |
1708 | } |
1709 | break; |
1710 | } |
1711 | |
1712 | if (width > 0) |
1713 | width = avail; |
1714 | |
1715 | /* The last thousands character will be added back by |
1716 | the char_buffer_add below. */ |
1717 | --charbuf.current; |
1718 | #endif |
1719 | } |
1720 | else |
1721 | break; |
1722 | |
1723 | char_buffer_add (&charbuf, c); |
1724 | if (width > 0) |
1725 | --width; |
1726 | |
1727 | c = inchar (); |
1728 | } |
1729 | } |
1730 | else |
1731 | /* Read the number into workspace. */ |
1732 | while (c != EOF && width != 0) |
1733 | { |
1734 | if (base == 16) |
1735 | { |
1736 | if (!ISXDIGIT (c)) |
1737 | break; |
1738 | } |
1739 | else if (!ISDIGIT (c) || (int) (c - L_('0')) >= base) |
1740 | { |
1741 | if (base == 10 && (flags & GROUP)) |
1742 | { |
1743 | /* Try matching against the thousands separator. */ |
1744 | #ifdef COMPILE_WSCANF |
1745 | if (c != thousands) |
1746 | break; |
1747 | #else |
1748 | const char *cmpp = thousands; |
1749 | int avail = width > 0 ? width : INT_MAX; |
1750 | |
1751 | while ((unsigned char) *cmpp == c && avail >= 0) |
1752 | { |
1753 | char_buffer_add (&charbuf, c); |
1754 | if (*++cmpp == '\0') |
1755 | break; |
1756 | else |
1757 | { |
1758 | if (avail == 0 || inchar () == EOF) |
1759 | break; |
1760 | --avail; |
1761 | } |
1762 | } |
1763 | |
1764 | if (char_buffer_error (&charbuf)) |
1765 | { |
1766 | __set_errno (ENOMEM); |
1767 | done = EOF; |
1768 | goto errout; |
1769 | } |
1770 | |
1771 | if (*cmpp != '\0') |
1772 | { |
1773 | /* We are pushing all read characters back. */ |
1774 | if (cmpp > thousands) |
1775 | { |
1776 | charbuf.current -= cmpp - thousands; |
1777 | ungetc (c, s); |
1778 | while (--cmpp > thousands) |
1779 | ungetc_not_eof ((unsigned char) *cmpp, s); |
1780 | c = (unsigned char) *cmpp; |
1781 | } |
1782 | break; |
1783 | } |
1784 | |
1785 | if (width > 0) |
1786 | width = avail; |
1787 | |
1788 | /* The last thousands character will be added back by |
1789 | the char_buffer_add below. */ |
1790 | --charbuf.current; |
1791 | #endif |
1792 | } |
1793 | else |
1794 | break; |
1795 | } |
1796 | char_buffer_add (&charbuf, c); |
1797 | if (width > 0) |
1798 | --width; |
1799 | |
1800 | c = inchar (); |
1801 | } |
1802 | |
1803 | if (char_buffer_error (&charbuf)) |
1804 | { |
1805 | __set_errno (ENOMEM); |
1806 | done = EOF; |
1807 | goto errout; |
1808 | } |
1809 | |
1810 | if (char_buffer_size (&charbuf) == 0 |
1811 | || (char_buffer_size (&charbuf) == 1 |
1812 | && (char_buffer_start (&charbuf)[0] == L_('+') |
1813 | || char_buffer_start (&charbuf)[0] == L_('-')))) |
1814 | { |
1815 | /* There was no number. If we are supposed to read a pointer |
1816 | we must recognize "(nil)" as well. */ |
1817 | if (__builtin_expect (char_buffer_size (&charbuf) == 0 |
1818 | && (flags & READ_POINTER) |
1819 | && (width < 0 || width >= 5) |
1820 | && c == '(' |
1821 | && TOLOWER (inchar ()) == L_('n') |
1822 | && TOLOWER (inchar ()) == L_('i') |
1823 | && TOLOWER (inchar ()) == L_('l') |
1824 | && inchar () == L_(')'), 1)) |
1825 | /* We must produce the value of a NULL pointer. A single |
1826 | '0' digit is enough. */ |
1827 | char_buffer_add (&charbuf, L_('0')); |
1828 | else |
1829 | { |
1830 | /* The last read character is not part of the number |
1831 | anymore. */ |
1832 | ungetc (c, s); |
1833 | |
1834 | conv_error (); |
1835 | } |
1836 | } |
1837 | else |
1838 | /* The just read character is not part of the number anymore. */ |
1839 | ungetc (c, s); |
1840 | |
1841 | /* Convert the number. */ |
1842 | char_buffer_add (&charbuf, L_('\0')); |
1843 | if (char_buffer_error (&charbuf)) |
1844 | { |
1845 | __set_errno (ENOMEM); |
1846 | done = EOF; |
1847 | goto errout; |
1848 | } |
1849 | if (need_longlong && (flags & LONGDBL)) |
1850 | { |
1851 | if (flags & NUMBER_SIGNED) |
1852 | num.q = __strtoll_internal |
1853 | (char_buffer_start (&charbuf), &tw, base, flags & GROUP); |
1854 | else |
1855 | num.uq = __strtoull_internal |
1856 | (char_buffer_start (&charbuf), &tw, base, flags & GROUP); |
1857 | } |
1858 | else |
1859 | { |
1860 | if (flags & NUMBER_SIGNED) |
1861 | num.l = __strtol_internal |
1862 | (char_buffer_start (&charbuf), &tw, base, flags & GROUP); |
1863 | else |
1864 | num.ul = __strtoul_internal |
1865 | (char_buffer_start (&charbuf), &tw, base, flags & GROUP); |
1866 | } |
1867 | if (__glibc_unlikely (char_buffer_start (&charbuf) == tw)) |
1868 | conv_error (); |
1869 | |
1870 | if (!(flags & SUPPRESS)) |
1871 | { |
1872 | if (flags & NUMBER_SIGNED) |
1873 | { |
1874 | if (need_longlong && (flags & LONGDBL)) |
1875 | *ARG (LONGLONG int *) = num.q; |
1876 | else if (need_long && (flags & LONG)) |
1877 | *ARG (long int *) = num.l; |
1878 | else if (flags & SHORT) |
1879 | *ARG (short int *) = (short int) num.l; |
1880 | else if (!(flags & CHAR)) |
1881 | *ARG (int *) = (int) num.l; |
1882 | else |
1883 | *ARG (signed char *) = (signed char) num.ul; |
1884 | } |
1885 | else |
1886 | { |
1887 | if (need_longlong && (flags & LONGDBL)) |
1888 | *ARG (unsigned LONGLONG int *) = num.uq; |
1889 | else if (need_long && (flags & LONG)) |
1890 | *ARG (unsigned long int *) = num.ul; |
1891 | else if (flags & SHORT) |
1892 | *ARG (unsigned short int *) |
1893 | = (unsigned short int) num.ul; |
1894 | else if (!(flags & CHAR)) |
1895 | *ARG (unsigned int *) = (unsigned int) num.ul; |
1896 | else |
1897 | *ARG (unsigned char *) = (unsigned char) num.ul; |
1898 | } |
1899 | ++done; |
1900 | } |
1901 | break; |
1902 | |
1903 | case L_('e'): /* Floating-point numbers. */ |
1904 | case L_('E'): |
1905 | case L_('f'): |
1906 | case L_('F'): |
1907 | case L_('g'): |
1908 | case L_('G'): |
1909 | case L_('a'): |
1910 | case L_('A'): |
1911 | c = inchar (); |
1912 | if (width > 0) |
1913 | --width; |
1914 | if (__glibc_unlikely (c == EOF)) |
1915 | input_error (); |
1916 | |
1917 | got_digit = got_dot = got_e = got_sign = 0; |
1918 | |
1919 | /* Check for a sign. */ |
1920 | if (c == L_('-') || c == L_('+')) |
1921 | { |
1922 | got_sign = 1; |
1923 | char_buffer_add (&charbuf, c); |
1924 | if (__glibc_unlikely (width == 0 || inchar () == EOF)) |
1925 | /* EOF is only an input error before we read any chars. */ |
1926 | conv_error (); |
1927 | if (width > 0) |
1928 | --width; |
1929 | } |
1930 | |
1931 | /* Take care for the special arguments "nan" and "inf". */ |
1932 | if (TOLOWER (c) == L_('n')) |
1933 | { |
1934 | /* Maybe "nan". */ |
1935 | char_buffer_add (&charbuf, c); |
1936 | if (__builtin_expect (width == 0 |
1937 | || inchar () == EOF |
1938 | || TOLOWER (c) != L_('a'), 0)) |
1939 | conv_error (); |
1940 | if (width > 0) |
1941 | --width; |
1942 | char_buffer_add (&charbuf, c); |
1943 | if (__builtin_expect (width == 0 |
1944 | || inchar () == EOF |
1945 | || TOLOWER (c) != L_('n'), 0)) |
1946 | conv_error (); |
1947 | if (width > 0) |
1948 | --width; |
1949 | char_buffer_add (&charbuf, c); |
1950 | /* It is "nan". */ |
1951 | goto scan_float; |
1952 | } |
1953 | else if (TOLOWER (c) == L_('i')) |
1954 | { |
1955 | /* Maybe "inf" or "infinity". */ |
1956 | char_buffer_add (&charbuf, c); |
1957 | if (__builtin_expect (width == 0 |
1958 | || inchar () == EOF |
1959 | || TOLOWER (c) != L_('n'), 0)) |
1960 | conv_error (); |
1961 | if (width > 0) |
1962 | --width; |
1963 | char_buffer_add (&charbuf, c); |
1964 | if (__builtin_expect (width == 0 |
1965 | || inchar () == EOF |
1966 | || TOLOWER (c) != L_('f'), 0)) |
1967 | conv_error (); |
1968 | if (width > 0) |
1969 | --width; |
1970 | char_buffer_add (&charbuf, c); |
1971 | /* It is as least "inf". */ |
1972 | if (width != 0 && inchar () != EOF) |
1973 | { |
1974 | if (TOLOWER (c) == L_('i')) |
1975 | { |
1976 | if (width > 0) |
1977 | --width; |
1978 | /* Now we have to read the rest as well. */ |
1979 | char_buffer_add (&charbuf, c); |
1980 | if (__builtin_expect (width == 0 |
1981 | || inchar () == EOF |
1982 | || TOLOWER (c) != L_('n'), 0)) |
1983 | conv_error (); |
1984 | if (width > 0) |
1985 | --width; |
1986 | char_buffer_add (&charbuf, c); |
1987 | if (__builtin_expect (width == 0 |
1988 | || inchar () == EOF |
1989 | || TOLOWER (c) != L_('i'), 0)) |
1990 | conv_error (); |
1991 | if (width > 0) |
1992 | --width; |
1993 | char_buffer_add (&charbuf, c); |
1994 | if (__builtin_expect (width == 0 |
1995 | || inchar () == EOF |
1996 | || TOLOWER (c) != L_('t'), 0)) |
1997 | conv_error (); |
1998 | if (width > 0) |
1999 | --width; |
2000 | char_buffer_add (&charbuf, c); |
2001 | if (__builtin_expect (width == 0 |
2002 | || inchar () == EOF |
2003 | || TOLOWER (c) != L_('y'), 0)) |
2004 | conv_error (); |
2005 | if (width > 0) |
2006 | --width; |
2007 | char_buffer_add (&charbuf, c); |
2008 | } |
2009 | else |
2010 | /* Never mind. */ |
2011 | ungetc (c, s); |
2012 | } |
2013 | goto scan_float; |
2014 | } |
2015 | |
2016 | exp_char = L_('e'); |
2017 | if (width != 0 && c == L_('0')) |
2018 | { |
2019 | char_buffer_add (&charbuf, c); |
2020 | c = inchar (); |
2021 | if (width > 0) |
2022 | --width; |
2023 | if (width != 0 && TOLOWER (c) == L_('x')) |
2024 | { |
2025 | /* It is a number in hexadecimal format. */ |
2026 | char_buffer_add (&charbuf, c); |
2027 | |
2028 | flags |= HEXA_FLOAT; |
2029 | exp_char = L_('p'); |
2030 | |
2031 | /* Grouping is not allowed. */ |
2032 | flags &= ~GROUP; |
2033 | c = inchar (); |
2034 | if (width > 0) |
2035 | --width; |
2036 | } |
2037 | else |
2038 | got_digit = 1; |
2039 | } |
2040 | |
2041 | while (1) |
2042 | { |
2043 | if (char_buffer_error (&charbuf)) |
2044 | { |
2045 | __set_errno (ENOMEM); |
2046 | done = EOF; |
2047 | goto errout; |
2048 | } |
2049 | if (ISDIGIT (c)) |
2050 | { |
2051 | char_buffer_add (&charbuf, c); |
2052 | got_digit = 1; |
2053 | } |
2054 | else if (!got_e && (flags & HEXA_FLOAT) && ISXDIGIT (c)) |
2055 | { |
2056 | char_buffer_add (&charbuf, c); |
2057 | got_digit = 1; |
2058 | } |
2059 | else if (got_e && charbuf.current[-1] == exp_char |
2060 | && (c == L_('-') || c == L_('+'))) |
2061 | char_buffer_add (&charbuf, c); |
2062 | else if (got_digit && !got_e |
2063 | && (CHAR_T) TOLOWER (c) == exp_char) |
2064 | { |
2065 | char_buffer_add (&charbuf, exp_char); |
2066 | got_e = got_dot = 1; |
2067 | } |
2068 | else |
2069 | { |
2070 | #ifdef COMPILE_WSCANF |
2071 | if (! got_dot && c == decimal) |
2072 | { |
2073 | char_buffer_add (&charbuf, c); |
2074 | got_dot = 1; |
2075 | } |
2076 | else if ((flags & GROUP) != 0 && ! got_dot && c == thousands) |
2077 | char_buffer_add (&charbuf, c); |
2078 | else |
2079 | { |
2080 | /* The last read character is not part of the number |
2081 | anymore. */ |
2082 | ungetc (c, s); |
2083 | break; |
2084 | } |
2085 | #else |
2086 | const char *cmpp = decimal; |
2087 | int avail = width > 0 ? width : INT_MAX; |
2088 | |
2089 | if (! got_dot) |
2090 | { |
2091 | while ((unsigned char) *cmpp == c && avail >= 0) |
2092 | if (*++cmpp == '\0') |
2093 | break; |
2094 | else |
2095 | { |
2096 | if (avail == 0 || inchar () == EOF) |
2097 | break; |
2098 | --avail; |
2099 | } |
2100 | } |
2101 | |
2102 | if (*cmpp == '\0') |
2103 | { |
2104 | /* Add all the characters. */ |
2105 | for (cmpp = decimal; *cmpp != '\0'; ++cmpp) |
2106 | char_buffer_add (&charbuf, (unsigned char) *cmpp); |
2107 | if (width > 0) |
2108 | width = avail; |
2109 | got_dot = 1; |
2110 | } |
2111 | else |
2112 | { |
2113 | /* Figure out whether it is a thousands separator. |
2114 | There is one problem: we possibly read more than |
2115 | one character. We cannot push them back but since |
2116 | we know that parts of the `decimal' string matched, |
2117 | we can compare against it. */ |
2118 | const char *cmp2p = thousands; |
2119 | |
2120 | if ((flags & GROUP) != 0 && ! got_dot) |
2121 | { |
2122 | while (cmp2p - thousands < cmpp - decimal |
2123 | && *cmp2p == decimal[cmp2p - thousands]) |
2124 | ++cmp2p; |
2125 | if (cmp2p - thousands == cmpp - decimal) |
2126 | { |
2127 | while ((unsigned char) *cmp2p == c && avail >= 0) |
2128 | if (*++cmp2p == '\0') |
2129 | break; |
2130 | else |
2131 | { |
2132 | if (avail == 0 || inchar () == EOF) |
2133 | break; |
2134 | --avail; |
2135 | } |
2136 | } |
2137 | } |
2138 | |
2139 | if (cmp2p != NULL && *cmp2p == '\0') |
2140 | { |
2141 | /* Add all the characters. */ |
2142 | for (cmpp = thousands; *cmpp != '\0'; ++cmpp) |
2143 | char_buffer_add (&charbuf, (unsigned char) *cmpp); |
2144 | if (width > 0) |
2145 | width = avail; |
2146 | } |
2147 | else |
2148 | { |
2149 | /* The last read character is not part of the number |
2150 | anymore. */ |
2151 | ungetc (c, s); |
2152 | break; |
2153 | } |
2154 | } |
2155 | #endif |
2156 | } |
2157 | |
2158 | if (width == 0 || inchar () == EOF) |
2159 | break; |
2160 | |
2161 | if (width > 0) |
2162 | --width; |
2163 | } |
2164 | |
2165 | if (char_buffer_error (&charbuf)) |
2166 | { |
2167 | __set_errno (ENOMEM); |
2168 | done = EOF; |
2169 | goto errout; |
2170 | } |
2171 | |
2172 | wctrans_t map; |
2173 | if (__builtin_expect ((flags & I18N) != 0, 0) |
2174 | /* Hexadecimal floats make no sense, fixing localized |
2175 | digits with ASCII letters. */ |
2176 | && !(flags & HEXA_FLOAT) |
2177 | /* Minimum requirement. */ |
2178 | && (char_buffer_size (&charbuf) == got_sign || got_dot) |
2179 | && (map = __wctrans ("to_inpunct" )) != NULL) |
2180 | { |
2181 | /* Reget the first character. */ |
2182 | inchar (); |
2183 | |
2184 | /* Localized digits, decimal points, and thousands |
2185 | separator. */ |
2186 | wint_t wcdigits[12]; |
2187 | |
2188 | /* First get decimal equivalent to check if we read it |
2189 | or not. */ |
2190 | wcdigits[11] = __towctrans (L'.', map); |
2191 | |
2192 | /* If we have not read any character or have just read |
2193 | locale decimal point which matches the decimal point |
2194 | for localized FP numbers, then we may have localized |
2195 | digits. Note, we test GOT_DOT above. */ |
2196 | #ifdef COMPILE_WSCANF |
2197 | if (char_buffer_size (&charbuf) == got_sign |
2198 | || (char_buffer_size (&charbuf) == got_sign + 1 |
2199 | && wcdigits[11] == decimal)) |
2200 | #else |
2201 | char mbdigits[12][MB_LEN_MAX + 1]; |
2202 | |
2203 | mbstate_t state; |
2204 | memset (&state, '\0', sizeof (state)); |
2205 | |
2206 | bool match_so_far = char_buffer_size (&charbuf) == got_sign; |
2207 | size_t mblen = __wcrtomb (mbdigits[11], wcdigits[11], &state); |
2208 | if (mblen != (size_t) -1) |
2209 | { |
2210 | mbdigits[11][mblen] = '\0'; |
2211 | match_so_far |= |
2212 | (char_buffer_size (&charbuf) == strlen (decimal) + got_sign |
2213 | && strcmp (decimal, mbdigits[11]) == 0); |
2214 | } |
2215 | else |
2216 | { |
2217 | size_t decimal_len = strlen (decimal); |
2218 | /* This should always be the case but the data comes |
2219 | from a file. */ |
2220 | if (decimal_len <= MB_LEN_MAX) |
2221 | { |
2222 | match_so_far |= (char_buffer_size (&charbuf) |
2223 | == decimal_len + got_sign); |
2224 | memcpy (mbdigits[11], decimal, decimal_len + 1); |
2225 | } |
2226 | else |
2227 | match_so_far = false; |
2228 | } |
2229 | |
2230 | if (match_so_far) |
2231 | #endif |
2232 | { |
2233 | bool have_locthousands = (flags & GROUP) != 0; |
2234 | |
2235 | /* Now get the digits and the thousands-sep equivalents. */ |
2236 | for (int n = 0; n < 11; ++n) |
2237 | { |
2238 | if (n < 10) |
2239 | wcdigits[n] = __towctrans (L'0' + n, map); |
2240 | else if (n == 10) |
2241 | { |
2242 | wcdigits[10] = __towctrans (L',', map); |
2243 | have_locthousands &= wcdigits[10] != L'\0'; |
2244 | } |
2245 | |
2246 | #ifndef COMPILE_WSCANF |
2247 | memset (&state, '\0', sizeof (state)); |
2248 | |
2249 | size_t mblen = __wcrtomb (mbdigits[n], wcdigits[n], |
2250 | &state); |
2251 | if (mblen == (size_t) -1) |
2252 | { |
2253 | if (n == 10) |
2254 | { |
2255 | if (have_locthousands) |
2256 | { |
2257 | size_t thousands_len = strlen (thousands); |
2258 | if (thousands_len <= MB_LEN_MAX) |
2259 | memcpy (mbdigits[10], thousands, |
2260 | thousands_len + 1); |
2261 | else |
2262 | have_locthousands = false; |
2263 | } |
2264 | } |
2265 | else |
2266 | /* Ignore checking against localized digits. */ |
2267 | goto no_i18nflt; |
2268 | } |
2269 | else |
2270 | mbdigits[n][mblen] = '\0'; |
2271 | #endif |
2272 | } |
2273 | |
2274 | /* Start checking against localized digits, if |
2275 | conversion is done correctly. */ |
2276 | while (1) |
2277 | { |
2278 | if (char_buffer_error (&charbuf)) |
2279 | { |
2280 | __set_errno (ENOMEM); |
2281 | done = EOF; |
2282 | goto errout; |
2283 | } |
2284 | if (got_e && charbuf.current[-1] == exp_char |
2285 | && (c == L_('-') || c == L_('+'))) |
2286 | char_buffer_add (&charbuf, c); |
2287 | else if (char_buffer_size (&charbuf) > got_sign && !got_e |
2288 | && (CHAR_T) TOLOWER (c) == exp_char) |
2289 | { |
2290 | char_buffer_add (&charbuf, exp_char); |
2291 | got_e = got_dot = 1; |
2292 | } |
2293 | else |
2294 | { |
2295 | /* Check against localized digits, decimal point, |
2296 | and thousands separator. */ |
2297 | int n; |
2298 | for (n = 0; n < 12; ++n) |
2299 | { |
2300 | #ifdef COMPILE_WSCANF |
2301 | if (c == wcdigits[n]) |
2302 | { |
2303 | if (n < 10) |
2304 | char_buffer_add (&charbuf, L_('0') + n); |
2305 | else if (n == 11 && !got_dot) |
2306 | { |
2307 | char_buffer_add (&charbuf, decimal); |
2308 | got_dot = 1; |
2309 | } |
2310 | else if (n == 10 && have_locthousands |
2311 | && ! got_dot) |
2312 | char_buffer_add (&charbuf, thousands); |
2313 | else |
2314 | /* The last read character is not part |
2315 | of the number anymore. */ |
2316 | n = 12; |
2317 | |
2318 | break; |
2319 | } |
2320 | #else |
2321 | const char *cmpp = mbdigits[n]; |
2322 | int avail = width > 0 ? width : INT_MAX; |
2323 | |
2324 | while ((unsigned char) *cmpp == c && avail >= 0) |
2325 | if (*++cmpp == '\0') |
2326 | break; |
2327 | else |
2328 | { |
2329 | if (avail == 0 || inchar () == EOF) |
2330 | break; |
2331 | --avail; |
2332 | } |
2333 | if (*cmpp == '\0') |
2334 | { |
2335 | if (width > 0) |
2336 | width = avail; |
2337 | |
2338 | if (n < 10) |
2339 | char_buffer_add (&charbuf, L_('0') + n); |
2340 | else if (n == 11 && !got_dot) |
2341 | { |
2342 | /* Add all the characters. */ |
2343 | for (cmpp = decimal; *cmpp != '\0'; |
2344 | ++cmpp) |
2345 | char_buffer_add (&charbuf, |
2346 | (unsigned char) *cmpp); |
2347 | |
2348 | got_dot = 1; |
2349 | } |
2350 | else if (n == 10 && (flags & GROUP) != 0 |
2351 | && ! got_dot) |
2352 | { |
2353 | /* Add all the characters. */ |
2354 | for (cmpp = thousands; *cmpp != '\0'; |
2355 | ++cmpp) |
2356 | char_buffer_add (&charbuf, |
2357 | (unsigned char) *cmpp); |
2358 | } |
2359 | else |
2360 | /* The last read character is not part |
2361 | of the number anymore. */ |
2362 | n = 12; |
2363 | |
2364 | break; |
2365 | } |
2366 | |
2367 | /* We are pushing all read characters back. */ |
2368 | if (cmpp > mbdigits[n]) |
2369 | { |
2370 | ungetc (c, s); |
2371 | while (--cmpp > mbdigits[n]) |
2372 | ungetc_not_eof ((unsigned char) *cmpp, s); |
2373 | c = (unsigned char) *cmpp; |
2374 | } |
2375 | #endif |
2376 | } |
2377 | |
2378 | if (n >= 12) |
2379 | { |
2380 | /* The last read character is not part |
2381 | of the number anymore. */ |
2382 | ungetc (c, s); |
2383 | break; |
2384 | } |
2385 | } |
2386 | |
2387 | if (width == 0 || inchar () == EOF) |
2388 | break; |
2389 | |
2390 | if (width > 0) |
2391 | --width; |
2392 | } |
2393 | } |
2394 | |
2395 | #ifndef COMPILE_WSCANF |
2396 | no_i18nflt: |
2397 | ; |
2398 | #endif |
2399 | } |
2400 | |
2401 | if (char_buffer_error (&charbuf)) |
2402 | { |
2403 | __set_errno (ENOMEM); |
2404 | done = EOF; |
2405 | goto errout; |
2406 | } |
2407 | |
2408 | /* Have we read any character? If we try to read a number |
2409 | in hexadecimal notation and we have read only the `0x' |
2410 | prefix this is an error. */ |
2411 | if (__glibc_unlikely (char_buffer_size (&charbuf) == got_sign |
2412 | || ((flags & HEXA_FLOAT) |
2413 | && (char_buffer_size (&charbuf) |
2414 | == 2 + got_sign)))) |
2415 | conv_error (); |
2416 | |
2417 | scan_float: |
2418 | /* Convert the number. */ |
2419 | char_buffer_add (&charbuf, L_('\0')); |
2420 | if (char_buffer_error (&charbuf)) |
2421 | { |
2422 | __set_errno (ENOMEM); |
2423 | done = EOF; |
2424 | goto errout; |
2425 | } |
2426 | if ((flags & LONGDBL) && !__ldbl_is_dbl) |
2427 | { |
2428 | long double d = __strtold_internal |
2429 | (char_buffer_start (&charbuf), &tw, flags & GROUP); |
2430 | if (!(flags & SUPPRESS) && tw != char_buffer_start (&charbuf)) |
2431 | *ARG (long double *) = d; |
2432 | } |
2433 | else if (flags & (LONG | LONGDBL)) |
2434 | { |
2435 | double d = __strtod_internal |
2436 | (char_buffer_start (&charbuf), &tw, flags & GROUP); |
2437 | if (!(flags & SUPPRESS) && tw != char_buffer_start (&charbuf)) |
2438 | *ARG (double *) = d; |
2439 | } |
2440 | else |
2441 | { |
2442 | float d = __strtof_internal |
2443 | (char_buffer_start (&charbuf), &tw, flags & GROUP); |
2444 | if (!(flags & SUPPRESS) && tw != char_buffer_start (&charbuf)) |
2445 | *ARG (float *) = d; |
2446 | } |
2447 | |
2448 | if (__glibc_unlikely (tw == char_buffer_start (&charbuf))) |
2449 | conv_error (); |
2450 | |
2451 | if (!(flags & SUPPRESS)) |
2452 | ++done; |
2453 | break; |
2454 | |
2455 | case L_('['): /* Character class. */ |
2456 | if (flags & LONG) |
2457 | STRING_ARG (wstr, wchar_t, 100); |
2458 | else |
2459 | STRING_ARG (str, char, 100); |
2460 | |
2461 | if (*f == L_('^')) |
2462 | { |
2463 | ++f; |
2464 | not_in = 1; |
2465 | } |
2466 | else |
2467 | not_in = 0; |
2468 | |
2469 | if (width < 0) |
2470 | /* There is no width given so there is also no limit on the |
2471 | number of characters we read. Therefore we set width to |
2472 | a very high value to make the algorithm easier. */ |
2473 | width = INT_MAX; |
2474 | |
2475 | #ifdef COMPILE_WSCANF |
2476 | /* Find the beginning and the end of the scanlist. We are not |
2477 | creating a lookup table since it would have to be too large. |
2478 | Instead we search each time through the string. This is not |
2479 | a constant lookup time but who uses this feature deserves to |
2480 | be punished. */ |
2481 | tw = (wchar_t *) f; /* Marks the beginning. */ |
2482 | |
2483 | if (*f == L']') |
2484 | ++f; |
2485 | |
2486 | while ((fc = *f++) != L'\0' && fc != L']'); |
2487 | |
2488 | if (__glibc_unlikely (fc == L'\0')) |
2489 | conv_error (); |
2490 | wchar_t *twend = (wchar_t *) f - 1; |
2491 | #else |
2492 | /* Fill WP with byte flags indexed by character. |
2493 | We will use this flag map for matching input characters. */ |
2494 | if (!scratch_buffer_set_array_size |
2495 | (&charbuf.scratch, UCHAR_MAX + 1, 1)) |
2496 | { |
2497 | done = EOF; |
2498 | goto errout; |
2499 | } |
2500 | memset (charbuf.scratch.data, '\0', UCHAR_MAX + 1); |
2501 | |
2502 | fc = *f; |
2503 | if (fc == ']' || fc == '-') |
2504 | { |
2505 | /* If ] or - appears before any char in the set, it is not |
2506 | the terminator or separator, but the first char in the |
2507 | set. */ |
2508 | ((char *)charbuf.scratch.data)[fc] = 1; |
2509 | ++f; |
2510 | } |
2511 | |
2512 | while ((fc = *f++) != '\0' && fc != ']') |
2513 | if (fc == '-' && *f != '\0' && *f != ']' |
2514 | && (unsigned char) f[-2] <= (unsigned char) *f) |
2515 | { |
2516 | /* Add all characters from the one before the '-' |
2517 | up to (but not including) the next format char. */ |
2518 | for (fc = (unsigned char) f[-2]; fc < (unsigned char) *f; ++fc) |
2519 | ((char *)charbuf.scratch.data)[fc] = 1; |
2520 | } |
2521 | else |
2522 | /* Add the character to the flag map. */ |
2523 | ((char *)charbuf.scratch.data)[fc] = 1; |
2524 | |
2525 | if (__glibc_unlikely (fc == '\0')) |
2526 | conv_error(); |
2527 | #endif |
2528 | |
2529 | if (flags & LONG) |
2530 | { |
2531 | size_t now = read_in; |
2532 | #ifdef COMPILE_WSCANF |
2533 | if (__glibc_unlikely (inchar () == WEOF)) |
2534 | input_error (); |
2535 | |
2536 | do |
2537 | { |
2538 | wchar_t *runp; |
2539 | |
2540 | /* Test whether it's in the scanlist. */ |
2541 | runp = tw; |
2542 | while (runp < twend) |
2543 | { |
2544 | if (runp[0] == L'-' && runp[1] != '\0' |
2545 | && runp + 1 != twend |
2546 | && runp != tw |
2547 | && (unsigned int) runp[-1] <= (unsigned int) runp[1]) |
2548 | { |
2549 | /* Match against all characters in between the |
2550 | first and last character of the sequence. */ |
2551 | wchar_t wc; |
2552 | |
2553 | for (wc = runp[-1] + 1; wc <= runp[1]; ++wc) |
2554 | if ((wint_t) wc == c) |
2555 | break; |
2556 | |
2557 | if (wc <= runp[1] && !not_in) |
2558 | break; |
2559 | if (wc <= runp[1] && not_in) |
2560 | { |
2561 | /* The current character is not in the |
2562 | scanset. */ |
2563 | ungetc (c, s); |
2564 | goto out; |
2565 | } |
2566 | |
2567 | runp += 2; |
2568 | } |
2569 | else |
2570 | { |
2571 | if ((wint_t) *runp == c && !not_in) |
2572 | break; |
2573 | if ((wint_t) *runp == c && not_in) |
2574 | { |
2575 | ungetc (c, s); |
2576 | goto out; |
2577 | } |
2578 | |
2579 | ++runp; |
2580 | } |
2581 | } |
2582 | |
2583 | if (runp == twend && !not_in) |
2584 | { |
2585 | ungetc (c, s); |
2586 | goto out; |
2587 | } |
2588 | |
2589 | if (!(flags & SUPPRESS)) |
2590 | { |
2591 | *wstr++ = c; |
2592 | |
2593 | if ((flags & MALLOC) |
2594 | && wstr == (wchar_t *) *strptr + strsize) |
2595 | { |
2596 | /* Enlarge the buffer. */ |
2597 | wstr = (wchar_t *) realloc (*strptr, |
2598 | (2 * strsize) |
2599 | * sizeof (wchar_t)); |
2600 | if (wstr == NULL) |
2601 | { |
2602 | /* Can't allocate that much. Last-ditch |
2603 | effort. */ |
2604 | wstr = (wchar_t *) |
2605 | realloc (*strptr, (strsize + 1) |
2606 | * sizeof (wchar_t)); |
2607 | if (wstr == NULL) |
2608 | { |
2609 | if (flags & POSIX_MALLOC) |
2610 | { |
2611 | done = EOF; |
2612 | goto errout; |
2613 | } |
2614 | /* We lose. Oh well. Terminate the string |
2615 | and stop converting, so at least we don't |
2616 | skip any input. */ |
2617 | ((wchar_t *) (*strptr))[strsize - 1] = L'\0'; |
2618 | strptr = NULL; |
2619 | ++done; |
2620 | conv_error (); |
2621 | } |
2622 | else |
2623 | { |
2624 | *strptr = (char *) wstr; |
2625 | wstr += strsize; |
2626 | ++strsize; |
2627 | } |
2628 | } |
2629 | else |
2630 | { |
2631 | *strptr = (char *) wstr; |
2632 | wstr += strsize; |
2633 | strsize *= 2; |
2634 | } |
2635 | } |
2636 | } |
2637 | } |
2638 | while (--width > 0 && inchar () != WEOF); |
2639 | out: |
2640 | #else |
2641 | char buf[MB_LEN_MAX]; |
2642 | size_t cnt = 0; |
2643 | mbstate_t cstate; |
2644 | |
2645 | if (__glibc_unlikely (inchar () == EOF)) |
2646 | input_error (); |
2647 | |
2648 | memset (&cstate, '\0', sizeof (cstate)); |
2649 | |
2650 | do |
2651 | { |
2652 | if (((char *) charbuf.scratch.data)[c] == not_in) |
2653 | { |
2654 | ungetc_not_eof (c, s); |
2655 | break; |
2656 | } |
2657 | |
2658 | /* This is easy. */ |
2659 | if (!(flags & SUPPRESS)) |
2660 | { |
2661 | size_t n; |
2662 | |
2663 | /* Convert it into a wide character. */ |
2664 | buf[0] = c; |
2665 | n = __mbrtowc (wstr, buf, 1, &cstate); |
2666 | |
2667 | if (n == (size_t) -2) |
2668 | { |
2669 | /* Possibly correct character, just not enough |
2670 | input. */ |
2671 | ++cnt; |
2672 | assert (cnt < MB_LEN_MAX); |
2673 | continue; |
2674 | } |
2675 | cnt = 0; |
2676 | |
2677 | ++wstr; |
2678 | if ((flags & MALLOC) |
2679 | && wstr == (wchar_t *) *strptr + strsize) |
2680 | { |
2681 | /* Enlarge the buffer. */ |
2682 | wstr = (wchar_t *) realloc (*strptr, |
2683 | (2 * strsize |
2684 | * sizeof (wchar_t))); |
2685 | if (wstr == NULL) |
2686 | { |
2687 | /* Can't allocate that much. Last-ditch |
2688 | effort. */ |
2689 | wstr = (wchar_t *) |
2690 | realloc (*strptr, ((strsize + 1) |
2691 | * sizeof (wchar_t))); |
2692 | if (wstr == NULL) |
2693 | { |
2694 | if (flags & POSIX_MALLOC) |
2695 | { |
2696 | done = EOF; |
2697 | goto errout; |
2698 | } |
2699 | /* We lose. Oh well. Terminate the |
2700 | string and stop converting, |
2701 | so at least we don't skip any input. */ |
2702 | ((wchar_t *) (*strptr))[strsize - 1] = L'\0'; |
2703 | strptr = NULL; |
2704 | ++done; |
2705 | conv_error (); |
2706 | } |
2707 | else |
2708 | { |
2709 | *strptr = (char *) wstr; |
2710 | wstr += strsize; |
2711 | ++strsize; |
2712 | } |
2713 | } |
2714 | else |
2715 | { |
2716 | *strptr = (char *) wstr; |
2717 | wstr += strsize; |
2718 | strsize *= 2; |
2719 | } |
2720 | } |
2721 | } |
2722 | |
2723 | if (--width <= 0) |
2724 | break; |
2725 | } |
2726 | while (inchar () != EOF); |
2727 | |
2728 | if (__glibc_unlikely (cnt != 0)) |
2729 | /* We stopped in the middle of recognizing another |
2730 | character. That's a problem. */ |
2731 | encode_error (); |
2732 | #endif |
2733 | |
2734 | if (__glibc_unlikely (now == read_in)) |
2735 | /* We haven't succesfully read any character. */ |
2736 | conv_error (); |
2737 | |
2738 | if (!(flags & SUPPRESS)) |
2739 | { |
2740 | *wstr++ = L'\0'; |
2741 | |
2742 | if ((flags & MALLOC) |
2743 | && wstr - (wchar_t *) *strptr != strsize) |
2744 | { |
2745 | wchar_t *cp = (wchar_t *) |
2746 | realloc (*strptr, ((wstr - (wchar_t *) *strptr) |
2747 | * sizeof(wchar_t))); |
2748 | if (cp != NULL) |
2749 | *strptr = (char *) cp; |
2750 | } |
2751 | strptr = NULL; |
2752 | |
2753 | ++done; |
2754 | } |
2755 | } |
2756 | else |
2757 | { |
2758 | size_t now = read_in; |
2759 | |
2760 | if (__glibc_unlikely (inchar () == EOF)) |
2761 | input_error (); |
2762 | |
2763 | #ifdef COMPILE_WSCANF |
2764 | |
2765 | memset (&state, '\0', sizeof (state)); |
2766 | |
2767 | do |
2768 | { |
2769 | wchar_t *runp; |
2770 | size_t n; |
2771 | |
2772 | /* Test whether it's in the scanlist. */ |
2773 | runp = tw; |
2774 | while (runp < twend) |
2775 | { |
2776 | if (runp[0] == L'-' && runp[1] != '\0' |
2777 | && runp + 1 != twend |
2778 | && runp != tw |
2779 | && (unsigned int) runp[-1] <= (unsigned int) runp[1]) |
2780 | { |
2781 | /* Match against all characters in between the |
2782 | first and last character of the sequence. */ |
2783 | wchar_t wc; |
2784 | |
2785 | for (wc = runp[-1] + 1; wc <= runp[1]; ++wc) |
2786 | if ((wint_t) wc == c) |
2787 | break; |
2788 | |
2789 | if (wc <= runp[1] && !not_in) |
2790 | break; |
2791 | if (wc <= runp[1] && not_in) |
2792 | { |
2793 | /* The current character is not in the |
2794 | scanset. */ |
2795 | ungetc (c, s); |
2796 | goto out2; |
2797 | } |
2798 | |
2799 | runp += 2; |
2800 | } |
2801 | else |
2802 | { |
2803 | if ((wint_t) *runp == c && !not_in) |
2804 | break; |
2805 | if ((wint_t) *runp == c && not_in) |
2806 | { |
2807 | ungetc (c, s); |
2808 | goto out2; |
2809 | } |
2810 | |
2811 | ++runp; |
2812 | } |
2813 | } |
2814 | |
2815 | if (runp == twend && !not_in) |
2816 | { |
2817 | ungetc (c, s); |
2818 | goto out2; |
2819 | } |
2820 | |
2821 | if (!(flags & SUPPRESS)) |
2822 | { |
2823 | if ((flags & MALLOC) |
2824 | && *strptr + strsize - str <= MB_LEN_MAX) |
2825 | { |
2826 | /* Enlarge the buffer. */ |
2827 | size_t strleng = str - *strptr; |
2828 | char *newstr; |
2829 | |
2830 | newstr = (char *) realloc (*strptr, 2 * strsize); |
2831 | if (newstr == NULL) |
2832 | { |
2833 | /* Can't allocate that much. Last-ditch |
2834 | effort. */ |
2835 | newstr = (char *) realloc (*strptr, |
2836 | strleng + MB_LEN_MAX); |
2837 | if (newstr == NULL) |
2838 | { |
2839 | if (flags & POSIX_MALLOC) |
2840 | { |
2841 | done = EOF; |
2842 | goto errout; |
2843 | } |
2844 | /* We lose. Oh well. Terminate the string |
2845 | and stop converting, so at least we don't |
2846 | skip any input. */ |
2847 | ((char *) (*strptr))[strleng] = '\0'; |
2848 | strptr = NULL; |
2849 | ++done; |
2850 | conv_error (); |
2851 | } |
2852 | else |
2853 | { |
2854 | *strptr = newstr; |
2855 | str = newstr + strleng; |
2856 | strsize = strleng + MB_LEN_MAX; |
2857 | } |
2858 | } |
2859 | else |
2860 | { |
2861 | *strptr = newstr; |
2862 | str = newstr + strleng; |
2863 | strsize *= 2; |
2864 | } |
2865 | } |
2866 | } |
2867 | |
2868 | n = __wcrtomb (!(flags & SUPPRESS) ? str : NULL, c, &state); |
2869 | if (__glibc_unlikely (n == (size_t) -1)) |
2870 | encode_error (); |
2871 | |
2872 | assert (n <= MB_LEN_MAX); |
2873 | str += n; |
2874 | } |
2875 | while (--width > 0 && inchar () != WEOF); |
2876 | out2: |
2877 | #else |
2878 | do |
2879 | { |
2880 | if (((char *) charbuf.scratch.data)[c] == not_in) |
2881 | { |
2882 | ungetc_not_eof (c, s); |
2883 | break; |
2884 | } |
2885 | |
2886 | /* This is easy. */ |
2887 | if (!(flags & SUPPRESS)) |
2888 | { |
2889 | *str++ = c; |
2890 | if ((flags & MALLOC) |
2891 | && (char *) str == *strptr + strsize) |
2892 | { |
2893 | /* Enlarge the buffer. */ |
2894 | size_t newsize = 2 * strsize; |
2895 | |
2896 | allocagain: |
2897 | str = (char *) realloc (*strptr, newsize); |
2898 | if (str == NULL) |
2899 | { |
2900 | /* Can't allocate that much. Last-ditch |
2901 | effort. */ |
2902 | if (newsize > strsize + 1) |
2903 | { |
2904 | newsize = strsize + 1; |
2905 | goto allocagain; |
2906 | } |
2907 | if (flags & POSIX_MALLOC) |
2908 | { |
2909 | done = EOF; |
2910 | goto errout; |
2911 | } |
2912 | /* We lose. Oh well. Terminate the |
2913 | string and stop converting, |
2914 | so at least we don't skip any input. */ |
2915 | ((char *) (*strptr))[strsize - 1] = '\0'; |
2916 | strptr = NULL; |
2917 | ++done; |
2918 | conv_error (); |
2919 | } |
2920 | else |
2921 | { |
2922 | *strptr = (char *) str; |
2923 | str += strsize; |
2924 | strsize = newsize; |
2925 | } |
2926 | } |
2927 | } |
2928 | } |
2929 | while (--width > 0 && inchar () != EOF); |
2930 | #endif |
2931 | |
2932 | if (__glibc_unlikely (now == read_in)) |
2933 | /* We haven't succesfully read any character. */ |
2934 | conv_error (); |
2935 | |
2936 | if (!(flags & SUPPRESS)) |
2937 | { |
2938 | #ifdef COMPILE_WSCANF |
2939 | /* We have to emit the code to get into the initial |
2940 | state. */ |
2941 | char buf[MB_LEN_MAX]; |
2942 | size_t n = __wcrtomb (buf, L'\0', &state); |
2943 | if (n > 0 && (flags & MALLOC) |
2944 | && str + n >= *strptr + strsize) |
2945 | { |
2946 | /* Enlarge the buffer. */ |
2947 | size_t strleng = str - *strptr; |
2948 | char *newstr; |
2949 | |
2950 | newstr = (char *) realloc (*strptr, strleng + n + 1); |
2951 | if (newstr == NULL) |
2952 | { |
2953 | if (flags & POSIX_MALLOC) |
2954 | { |
2955 | done = EOF; |
2956 | goto errout; |
2957 | } |
2958 | /* We lose. Oh well. Terminate the string |
2959 | and stop converting, so at least we don't |
2960 | skip any input. */ |
2961 | ((char *) (*strptr))[strleng] = '\0'; |
2962 | strptr = NULL; |
2963 | ++done; |
2964 | conv_error (); |
2965 | } |
2966 | else |
2967 | { |
2968 | *strptr = newstr; |
2969 | str = newstr + strleng; |
2970 | strsize = strleng + n + 1; |
2971 | } |
2972 | } |
2973 | |
2974 | str = __mempcpy (str, buf, n); |
2975 | #endif |
2976 | *str++ = '\0'; |
2977 | |
2978 | if ((flags & MALLOC) && str - *strptr != strsize) |
2979 | { |
2980 | char *cp = (char *) realloc (*strptr, str - *strptr); |
2981 | if (cp != NULL) |
2982 | *strptr = cp; |
2983 | } |
2984 | strptr = NULL; |
2985 | |
2986 | ++done; |
2987 | } |
2988 | } |
2989 | break; |
2990 | |
2991 | case L_('p'): /* Generic pointer. */ |
2992 | base = 16; |
2993 | /* A PTR must be the same size as a `long int'. */ |
2994 | flags &= ~(SHORT|LONGDBL); |
2995 | if (need_long) |
2996 | flags |= LONG; |
2997 | flags |= READ_POINTER; |
2998 | goto number; |
2999 | |
3000 | default: |
3001 | /* If this is an unknown format character punt. */ |
3002 | conv_error (); |
3003 | } |
3004 | } |
3005 | |
3006 | /* The last thing we saw int the format string was a white space. |
3007 | Consume the last white spaces. */ |
3008 | if (skip_space) |
3009 | { |
3010 | do |
3011 | c = inchar (); |
3012 | while (ISSPACE (c)); |
3013 | ungetc (c, s); |
3014 | } |
3015 | |
3016 | errout: |
3017 | /* Unlock stream. */ |
3018 | UNLOCK_STREAM (s); |
3019 | |
3020 | scratch_buffer_free (&charbuf.scratch); |
3021 | if (errp != NULL) |
3022 | *errp |= errval; |
3023 | |
3024 | if (__glibc_unlikely (done == EOF)) |
3025 | { |
3026 | if (__glibc_unlikely (ptrs_to_free != NULL)) |
3027 | { |
3028 | struct ptrs_to_free *p = ptrs_to_free; |
3029 | while (p != NULL) |
3030 | { |
3031 | for (size_t cnt = 0; cnt < p->count; ++cnt) |
3032 | { |
3033 | free (*p->ptrs[cnt]); |
3034 | *p->ptrs[cnt] = NULL; |
3035 | } |
3036 | p = p->next; |
3037 | ptrs_to_free = p; |
3038 | } |
3039 | } |
3040 | } |
3041 | else if (__glibc_unlikely (strptr != NULL)) |
3042 | { |
3043 | free (*strptr); |
3044 | *strptr = NULL; |
3045 | } |
3046 | return done; |
3047 | } |
3048 | |
3049 | #ifdef COMPILE_WSCANF |
3050 | int |
3051 | __vfwscanf (FILE *s, const wchar_t *format, va_list argptr) |
3052 | { |
3053 | return _IO_vfwscanf (s, format, argptr, NULL); |
3054 | } |
3055 | ldbl_weak_alias (__vfwscanf, vfwscanf) |
3056 | #else |
3057 | int |
3058 | ___vfscanf (FILE *s, const char *format, va_list argptr) |
3059 | { |
3060 | return _IO_vfscanf_internal (s, format, argptr, NULL); |
3061 | } |
3062 | ldbl_strong_alias (_IO_vfscanf_internal, _IO_vfscanf) |
3063 | ldbl_hidden_def (_IO_vfscanf_internal, _IO_vfscanf) |
3064 | ldbl_strong_alias (___vfscanf, __vfscanf) |
3065 | ldbl_hidden_def (___vfscanf, __vfscanf) |
3066 | ldbl_weak_alias (___vfscanf, vfscanf) |
3067 | #endif |
3068 | |