]>
Commit | Line | Data |
---|---|---|
1 | /*- | |
2 | * Copyright (c) 1990, 1993 | |
3 | * The Regents of the University of California. All rights reserved. | |
4 | * | |
5 | * This code is derived from software contributed to Berkeley by | |
6 | * Chris Torek. | |
7 | * | |
8 | * Redistribution and use in source and binary forms, with or without | |
9 | * modification, are permitted provided that the following conditions | |
10 | * are met: | |
11 | * 1. Redistributions of source code must retain the above copyright | |
12 | * notice, this list of conditions and the following disclaimer. | |
13 | * 2. Redistributions in binary form must reproduce the above copyright | |
14 | * notice, this list of conditions and the following disclaimer in the | |
15 | * documentation and/or other materials provided with the distribution. | |
16 | * 4. Neither the name of the University nor the names of its contributors | |
17 | * may be used to endorse or promote products derived from this software | |
18 | * without specific prior written permission. | |
19 | * | |
20 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | |
21 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | |
22 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | |
23 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
24 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | |
25 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | |
26 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | |
27 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | |
28 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | |
29 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | |
30 | * SUCH DAMAGE. | |
31 | */ | |
32 | ||
33 | #if defined(LIBC_SCCS) && !defined(lint) | |
34 | static char sccsid[] = "@(#)vfscanf.c 8.1 (Berkeley) 6/4/93"; | |
35 | #endif /* LIBC_SCCS and not lint */ | |
36 | #include <sys/cdefs.h> | |
37 | __FBSDID("$FreeBSD: src/lib/libc/stdio/vfscanf.c,v 1.43 2009/01/19 06:19:51 das Exp $"); | |
38 | ||
39 | #include "xlocale_private.h" | |
40 | ||
41 | #include "namespace.h" | |
42 | #include <ctype.h> | |
43 | #include <inttypes.h> | |
44 | #include <stdio.h> | |
45 | #include <stdlib.h> | |
46 | #include <stddef.h> | |
47 | #include <stdarg.h> | |
48 | #include <string.h> | |
49 | #include <wchar.h> | |
50 | #include <wctype.h> | |
51 | #include <pthread.h> | |
52 | #include "un-namespace.h" | |
53 | ||
54 | #include "collate.h" | |
55 | #include "libc_private.h" | |
56 | #include "local.h" | |
57 | ||
58 | #ifndef NO_FLOATING_POINT | |
59 | #include <locale.h> | |
60 | #endif | |
61 | ||
62 | #define BUF 513 /* Maximum length of numeric string. */ | |
63 | ||
64 | /* | |
65 | * Flags used during conversion. | |
66 | */ | |
67 | #define LONG 0x01 /* l: long or double */ | |
68 | #define LONGDBL 0x02 /* L: long double */ | |
69 | #define SHORT 0x04 /* h: short */ | |
70 | #define SUPPRESS 0x08 /* *: suppress assignment */ | |
71 | #define POINTER 0x10 /* p: void * (as hex) */ | |
72 | #define NOSKIP 0x20 /* [ or c: do not skip blanks */ | |
73 | #define LONGLONG 0x400 /* ll: long long (+ deprecated q: quad) */ | |
74 | #define INTMAXT 0x800 /* j: intmax_t */ | |
75 | #define PTRDIFFT 0x1000 /* t: ptrdiff_t */ | |
76 | #define SIZET 0x2000 /* z: size_t */ | |
77 | #define SHORTSHORT 0x4000 /* hh: char */ | |
78 | #define UNSIGNED 0x8000 /* %[oupxX] conversions */ | |
79 | ||
80 | /* | |
81 | * The following are used in integral conversions only: | |
82 | * SIGNOK, NDIGITS, PFXOK, and NZDIGITS | |
83 | */ | |
84 | #define SIGNOK 0x40 /* +/- is (still) legal */ | |
85 | #define NDIGITS 0x80 /* no digits detected */ | |
86 | #define PFXOK 0x100 /* 0x prefix is (still) legal */ | |
87 | #define NZDIGITS 0x200 /* no zero digits detected */ | |
88 | #define HAVESIGN 0x10000 /* sign detected */ | |
89 | ||
90 | /* | |
91 | * Conversion types. | |
92 | */ | |
93 | #define CT_CHAR 0 /* %c conversion */ | |
94 | #define CT_CCL 1 /* %[...] conversion */ | |
95 | #define CT_STRING 2 /* %s conversion */ | |
96 | #define CT_INT 3 /* %[dioupxX] conversion */ | |
97 | #define CT_FLOAT 4 /* %[efgEFG] conversion */ | |
98 | ||
99 | static const u_char *__sccl(char *, const u_char *, locale_t); | |
100 | #ifndef NO_FLOATING_POINT | |
101 | static int parsefloat(FILE *, char **, size_t, locale_t); | |
102 | #endif | |
103 | ||
104 | __weak_reference(__vfscanf, vfscanf); | |
105 | ||
106 | /* | |
107 | * __vfscanf - MT-safe version | |
108 | */ | |
109 | int | |
110 | __vfscanf(FILE * __restrict fp, char const * __restrict fmt0, va_list ap) | |
111 | { | |
112 | int ret; | |
113 | ||
114 | FLOCKFILE(fp); | |
115 | ret = __svfscanf_l(fp, __current_locale(), fmt0, ap); | |
116 | FUNLOCKFILE(fp); | |
117 | return (ret); | |
118 | } | |
119 | ||
120 | int | |
121 | vfscanf_l(FILE * __restrict fp, locale_t loc, char const * __restrict fmt0, va_list ap) | |
122 | { | |
123 | int ret; | |
124 | ||
125 | NORMALIZE_LOCALE(loc); | |
126 | FLOCKFILE(fp); | |
127 | ret = __svfscanf_l(fp, loc, fmt0, ap); | |
128 | FUNLOCKFILE(fp); | |
129 | return (ret); | |
130 | } | |
131 | ||
132 | /* | |
133 | * __svfscanf - non-MT-safe version of __vfscanf | |
134 | */ | |
135 | __private_extern__ int | |
136 | __svfscanf_l(FILE * __restrict fp, locale_t loc, const char * __restrict fmt0, va_list ap) | |
137 | { | |
138 | const u_char *fmt = (const u_char *)fmt0; | |
139 | int c; /* character from format, or conversion */ | |
140 | size_t width; /* field width, or 0 */ | |
141 | char *p; /* points into all kinds of strings */ | |
142 | int n; /* handy integer */ | |
143 | int flags; /* flags as defined above */ | |
144 | char *p0; /* saves original value of p when necessary */ | |
145 | int nassigned; /* number of fields assigned */ | |
146 | int nread; /* number of characters consumed from fp */ | |
147 | int base; /* base argument to conversion function */ | |
148 | char ccltab[256]; /* character class table for %[...] */ | |
149 | char buf[BUF]; /* buffer for numeric and mb conversions */ | |
150 | wchar_t *wcp; /* handy wide character pointer */ | |
151 | size_t nconv; /* length of multibyte sequence converted */ | |
152 | int index; /* %index$, zero if unset */ | |
153 | va_list ap_orig; /* to reset ap to first argument */ | |
154 | static const mbstate_t initial; | |
155 | mbstate_t mbs; | |
156 | int mb_cur_max; | |
157 | ||
158 | /* `basefix' is used to avoid `if' tests in the integer scanner */ | |
159 | static const short basefix[17] = | |
160 | { 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 }; | |
161 | ||
162 | NORMALIZE_LOCALE(loc); | |
163 | mb_cur_max = MB_CUR_MAX_L(loc); | |
164 | ORIENT(fp, -1); | |
165 | ||
166 | nassigned = 0; | |
167 | nread = 0; | |
168 | va_copy(ap_orig, ap); | |
169 | for (;;) { | |
170 | c = *fmt++; | |
171 | if (c == 0) | |
172 | return (nassigned); | |
173 | if (isspace_l(c, loc)) { | |
174 | while ((fp->_r > 0 || __srefill(fp) == 0) && isspace_l(*fp->_p, loc)) | |
175 | nread++, fp->_r--, fp->_p++; | |
176 | continue; | |
177 | } | |
178 | if (c != '%') { | |
179 | if (fp->_r <= 0 && __srefill(fp)) | |
180 | goto input_failure; | |
181 | goto literal; | |
182 | } | |
183 | width = 0; | |
184 | flags = 0; | |
185 | /* | |
186 | * switch on the format. continue if done; | |
187 | * break once format type is derived. | |
188 | */ | |
189 | again: c = *fmt++; | |
190 | switch (c) { | |
191 | case '%': | |
192 | /* Consume leading white space */ | |
193 | for(;;) { | |
194 | if (fp->_r <= 0 && __srefill(fp)) | |
195 | goto input_failure; | |
196 | if (!isspace_l(*fp->_p, loc)) | |
197 | break; | |
198 | nread++; | |
199 | fp->_r--; | |
200 | fp->_p++; | |
201 | } | |
202 | literal: | |
203 | if (*fp->_p != c) | |
204 | goto match_failure; | |
205 | fp->_r--, fp->_p++; | |
206 | nread++; | |
207 | continue; | |
208 | ||
209 | case '$': | |
210 | index = width; | |
211 | if (index < 1 || index > NL_ARGMAX || fmt[-3] != '%') { | |
212 | goto input_failure; | |
213 | } | |
214 | width = 0; | |
215 | va_end(ap); | |
216 | va_copy(ap, ap_orig); /* reset to %1$ */ | |
217 | for (; index > 1; index--) { | |
218 | va_arg(ap, void*); | |
219 | } | |
220 | goto again; | |
221 | case '*': | |
222 | flags |= SUPPRESS; | |
223 | goto again; | |
224 | case 'j': | |
225 | flags |= INTMAXT; | |
226 | goto again; | |
227 | case 'l': | |
228 | if (flags & LONG) { | |
229 | flags &= ~LONG; | |
230 | flags |= LONGLONG; | |
231 | } else | |
232 | flags |= LONG; | |
233 | goto again; | |
234 | case 'q': | |
235 | flags |= LONGLONG; /* not quite */ | |
236 | goto again; | |
237 | case 't': | |
238 | flags |= PTRDIFFT; | |
239 | goto again; | |
240 | case 'z': | |
241 | flags |= SIZET; | |
242 | goto again; | |
243 | case 'L': | |
244 | flags |= LONGDBL; | |
245 | goto again; | |
246 | case 'h': | |
247 | if (flags & SHORT) { | |
248 | flags &= ~SHORT; | |
249 | flags |= SHORTSHORT; | |
250 | } else | |
251 | flags |= SHORT; | |
252 | goto again; | |
253 | ||
254 | case '0': case '1': case '2': case '3': case '4': | |
255 | case '5': case '6': case '7': case '8': case '9': | |
256 | width = width * 10 + c - '0'; | |
257 | goto again; | |
258 | ||
259 | /* | |
260 | * Conversions. | |
261 | */ | |
262 | case 'd': | |
263 | c = CT_INT; | |
264 | base = 10; | |
265 | break; | |
266 | ||
267 | case 'i': | |
268 | c = CT_INT; | |
269 | base = 0; | |
270 | break; | |
271 | ||
272 | case 'o': | |
273 | c = CT_INT; | |
274 | flags |= UNSIGNED; | |
275 | base = 8; | |
276 | break; | |
277 | ||
278 | case 'u': | |
279 | c = CT_INT; | |
280 | flags |= UNSIGNED; | |
281 | base = 10; | |
282 | break; | |
283 | ||
284 | case 'X': | |
285 | case 'x': | |
286 | flags |= PFXOK; /* enable 0x prefixing */ | |
287 | c = CT_INT; | |
288 | flags |= UNSIGNED; | |
289 | base = 16; | |
290 | break; | |
291 | ||
292 | #ifndef NO_FLOATING_POINT | |
293 | case 'A': case 'E': case 'F': case 'G': | |
294 | case 'a': case 'e': case 'f': case 'g': | |
295 | c = CT_FLOAT; | |
296 | break; | |
297 | #endif | |
298 | ||
299 | case 'S': | |
300 | flags |= LONG; | |
301 | /* FALLTHROUGH */ | |
302 | case 's': | |
303 | c = CT_STRING; | |
304 | break; | |
305 | ||
306 | case '[': | |
307 | fmt = __sccl(ccltab, fmt, loc); | |
308 | flags |= NOSKIP; | |
309 | c = CT_CCL; | |
310 | break; | |
311 | ||
312 | case 'C': | |
313 | flags |= LONG; | |
314 | /* FALLTHROUGH */ | |
315 | case 'c': | |
316 | flags |= NOSKIP; | |
317 | c = CT_CHAR; | |
318 | break; | |
319 | ||
320 | case 'p': /* pointer format is like hex */ | |
321 | flags |= POINTER | PFXOK; | |
322 | c = CT_INT; /* assumes sizeof(uintmax_t) */ | |
323 | flags |= UNSIGNED; /* >= sizeof(uintptr_t) */ | |
324 | base = 16; | |
325 | break; | |
326 | ||
327 | case 'n': | |
328 | { | |
329 | if (flags & SUPPRESS) /* ??? */ | |
330 | continue; | |
331 | void *ptr = va_arg(ap, void *); | |
332 | if (ptr == NULL) | |
333 | continue; | |
334 | else if (flags & SHORTSHORT) | |
335 | *(char *)ptr = nread; | |
336 | else if (flags & SHORT) | |
337 | *(short *)ptr = nread; | |
338 | else if (flags & LONG) | |
339 | *(long *)ptr = nread; | |
340 | else if (flags & LONGLONG) | |
341 | *(long long *)ptr = nread; | |
342 | else if (flags & INTMAXT) | |
343 | *(intmax_t *)ptr = nread; | |
344 | else if (flags & SIZET) | |
345 | *(size_t *)ptr = nread; | |
346 | else if (flags & PTRDIFFT) | |
347 | *(ptrdiff_t *)ptr = nread; | |
348 | else | |
349 | *(int *)ptr = nread; | |
350 | continue; | |
351 | } | |
352 | default: | |
353 | goto match_failure; | |
354 | ||
355 | /* | |
356 | * Disgusting backwards compatibility hack. XXX | |
357 | */ | |
358 | case '\0': /* compat */ | |
359 | return (EOF); | |
360 | } | |
361 | ||
362 | /* | |
363 | * We have a conversion that requires input. | |
364 | */ | |
365 | if (fp->_r <= 0 && __srefill(fp)) | |
366 | goto input_failure; | |
367 | ||
368 | /* | |
369 | * Consume leading white space, except for formats | |
370 | * that suppress this. | |
371 | */ | |
372 | if ((flags & NOSKIP) == 0) { | |
373 | while (isspace_l(*fp->_p, loc)) { | |
374 | nread++; | |
375 | if (--fp->_r > 0) | |
376 | fp->_p++; | |
377 | else if (__srefill(fp)) | |
378 | goto input_failure; | |
379 | } | |
380 | /* | |
381 | * Note that there is at least one character in | |
382 | * the buffer, so conversions that do not set NOSKIP | |
383 | * ca no longer result in an input failure. | |
384 | */ | |
385 | } | |
386 | ||
387 | /* | |
388 | * Do the conversion. | |
389 | */ | |
390 | switch (c) { | |
391 | ||
392 | case CT_CHAR: | |
393 | /* scan arbitrary characters (sets NOSKIP) */ | |
394 | if (width == 0) | |
395 | width = 1; | |
396 | if (flags & LONG) { | |
397 | if ((flags & SUPPRESS) == 0) | |
398 | wcp = va_arg(ap, wchar_t *); | |
399 | else | |
400 | wcp = NULL; | |
401 | n = 0; | |
402 | while (width != 0) { | |
403 | if (n == mb_cur_max) { | |
404 | fp->_flags |= __SERR; | |
405 | goto input_failure; | |
406 | } | |
407 | buf[n++] = *fp->_p; | |
408 | fp->_p++; | |
409 | fp->_r--; | |
410 | mbs = initial; | |
411 | nconv = mbrtowc_l(wcp, buf, n, &mbs, loc); | |
412 | if (nconv == (size_t)-1) { | |
413 | fp->_flags |= __SERR; | |
414 | goto input_failure; | |
415 | } | |
416 | if (nconv == 0 && !(flags & SUPPRESS)) | |
417 | *wcp = L'\0'; | |
418 | if (nconv != (size_t)-2) { | |
419 | nread += n; | |
420 | width--; | |
421 | if (!(flags & SUPPRESS)) | |
422 | wcp++; | |
423 | n = 0; | |
424 | } | |
425 | if (fp->_r <= 0 && __srefill(fp)) { | |
426 | if (n != 0) { | |
427 | fp->_flags |= __SERR; | |
428 | goto input_failure; | |
429 | } | |
430 | break; | |
431 | } | |
432 | } | |
433 | if (!(flags & SUPPRESS)) | |
434 | nassigned++; | |
435 | } else if (flags & SUPPRESS) { | |
436 | size_t sum = 0; | |
437 | for (;;) { | |
438 | if ((n = fp->_r) < width) { | |
439 | sum += n; | |
440 | width -= n; | |
441 | fp->_p += n; | |
442 | if (__srefill(fp)) { | |
443 | if (sum == 0) | |
444 | goto input_failure; | |
445 | break; | |
446 | } | |
447 | } else { | |
448 | sum += width; | |
449 | fp->_r -= width; | |
450 | fp->_p += width; | |
451 | break; | |
452 | } | |
453 | } | |
454 | nread += sum; | |
455 | } else { | |
456 | size_t r = __fread((void *)va_arg(ap, char *), 1, | |
457 | width, fp); | |
458 | ||
459 | if (r == 0) | |
460 | goto input_failure; | |
461 | nread += r; | |
462 | nassigned++; | |
463 | } | |
464 | break; | |
465 | ||
466 | case CT_CCL: | |
467 | /* scan a (nonempty) character class (sets NOSKIP) */ | |
468 | if (width == 0) | |
469 | width = (size_t)~0; /* `infinity' */ | |
470 | /* take only those things in the class */ | |
471 | if (flags & LONG) { | |
472 | wchar_t twc; | |
473 | int nchars; | |
474 | ||
475 | if ((flags & SUPPRESS) == 0) | |
476 | wcp = va_arg(ap, wchar_t *); | |
477 | else | |
478 | wcp = &twc; | |
479 | n = 0; | |
480 | nchars = 0; | |
481 | while (width != 0) { | |
482 | if (n == mb_cur_max) { | |
483 | fp->_flags |= __SERR; | |
484 | goto input_failure; | |
485 | } | |
486 | buf[n++] = *fp->_p; | |
487 | fp->_p++; | |
488 | fp->_r--; | |
489 | mbs = initial; | |
490 | nconv = mbrtowc_l(wcp, buf, n, &mbs, loc); | |
491 | if (nconv == (size_t)-1) { | |
492 | fp->_flags |= __SERR; | |
493 | goto input_failure; | |
494 | } | |
495 | if (nconv == 0) | |
496 | *wcp = L'\0'; | |
497 | if (nconv != (size_t)-2) { | |
498 | if (wctob_l(*wcp, loc) != EOF && | |
499 | !ccltab[wctob_l(*wcp, loc)]) { | |
500 | while (n != 0) { | |
501 | n--; | |
502 | __ungetc(buf[n], | |
503 | fp); | |
504 | } | |
505 | break; | |
506 | } | |
507 | nread += n; | |
508 | width--; | |
509 | if (!(flags & SUPPRESS)) | |
510 | wcp++; | |
511 | nchars++; | |
512 | n = 0; | |
513 | } | |
514 | if (fp->_r <= 0 && __srefill(fp)) { | |
515 | if (n != 0) { | |
516 | fp->_flags |= __SERR; | |
517 | goto input_failure; | |
518 | } | |
519 | break; | |
520 | } | |
521 | } | |
522 | if (n != 0) { | |
523 | fp->_flags |= __SERR; | |
524 | goto input_failure; | |
525 | } | |
526 | n = nchars; | |
527 | if (n == 0) | |
528 | goto match_failure; | |
529 | if (!(flags & SUPPRESS)) { | |
530 | *wcp = L'\0'; | |
531 | nassigned++; | |
532 | } | |
533 | } else if (flags & SUPPRESS) { | |
534 | n = 0; | |
535 | while (ccltab[*fp->_p]) { | |
536 | n++, fp->_r--, fp->_p++; | |
537 | if (--width == 0) | |
538 | break; | |
539 | if (fp->_r <= 0 && __srefill(fp)) { | |
540 | if (n == 0) | |
541 | goto input_failure; | |
542 | break; | |
543 | } | |
544 | } | |
545 | if (n == 0) | |
546 | goto match_failure; | |
547 | } else { | |
548 | p0 = p = va_arg(ap, char *); | |
549 | while (ccltab[*fp->_p]) { | |
550 | fp->_r--; | |
551 | *p++ = *fp->_p++; | |
552 | if (--width == 0) | |
553 | break; | |
554 | if (fp->_r <= 0 && __srefill(fp)) { | |
555 | if (p == p0) | |
556 | goto input_failure; | |
557 | break; | |
558 | } | |
559 | } | |
560 | n = p - p0; | |
561 | if (n == 0) | |
562 | goto match_failure; | |
563 | *p = 0; | |
564 | nassigned++; | |
565 | } | |
566 | nread += n; | |
567 | break; | |
568 | ||
569 | case CT_STRING: | |
570 | /* like CCL, but zero-length string OK, & no NOSKIP */ | |
571 | if (width == 0) | |
572 | width = (size_t)~0; | |
573 | if (flags & LONG) { | |
574 | wchar_t twc; | |
575 | ||
576 | if ((flags & SUPPRESS) == 0) | |
577 | wcp = va_arg(ap, wchar_t *); | |
578 | else | |
579 | wcp = &twc; | |
580 | n = 0; | |
581 | while (width != 0) { | |
582 | if (n == mb_cur_max) { | |
583 | fp->_flags |= __SERR; | |
584 | goto input_failure; | |
585 | } | |
586 | buf[n++] = *fp->_p; | |
587 | fp->_p++; | |
588 | fp->_r--; | |
589 | mbs = initial; | |
590 | nconv = mbrtowc_l(wcp, buf, n, &mbs, loc); | |
591 | if (nconv == (size_t)-1) { | |
592 | fp->_flags |= __SERR; | |
593 | goto input_failure; | |
594 | } | |
595 | if (nconv == 0) | |
596 | *wcp = L'\0'; | |
597 | if (nconv != (size_t)-2) { | |
598 | if (iswspace_l(*wcp, loc)) { | |
599 | while (n != 0) { | |
600 | n--; | |
601 | __ungetc(buf[n], | |
602 | fp); | |
603 | } | |
604 | break; | |
605 | } | |
606 | nread += n; | |
607 | width--; | |
608 | if (!(flags & SUPPRESS)) | |
609 | wcp++; | |
610 | n = 0; | |
611 | } | |
612 | if (fp->_r <= 0 && __srefill(fp)) { | |
613 | if (n != 0) { | |
614 | fp->_flags |= __SERR; | |
615 | goto input_failure; | |
616 | } | |
617 | break; | |
618 | } | |
619 | } | |
620 | if (!(flags & SUPPRESS)) { | |
621 | *wcp = L'\0'; | |
622 | nassigned++; | |
623 | } | |
624 | } else if (flags & SUPPRESS) { | |
625 | n = 0; | |
626 | while (!isspace_l(*fp->_p, loc)) { | |
627 | n++, fp->_r--, fp->_p++; | |
628 | if (--width == 0) | |
629 | break; | |
630 | if (fp->_r <= 0 && __srefill(fp)) | |
631 | break; | |
632 | } | |
633 | nread += n; | |
634 | } else { | |
635 | p0 = p = va_arg(ap, char *); | |
636 | while (!isspace_l(*fp->_p, loc)) { | |
637 | fp->_r--; | |
638 | *p++ = *fp->_p++; | |
639 | if (--width == 0) | |
640 | break; | |
641 | if (fp->_r <= 0 && __srefill(fp)) | |
642 | break; | |
643 | } | |
644 | *p = 0; | |
645 | nread += p - p0; | |
646 | nassigned++; | |
647 | } | |
648 | continue; | |
649 | ||
650 | case CT_INT: | |
651 | /* scan an integer as if by the conversion function */ | |
652 | #ifdef hardway | |
653 | if (width == 0 || width > sizeof(buf) - 1) | |
654 | width = sizeof(buf) - 1; | |
655 | #else | |
656 | /* size_t is unsigned, hence this optimisation */ | |
657 | if (--width > sizeof(buf) - 2) | |
658 | width = sizeof(buf) - 2; | |
659 | width++; | |
660 | #endif | |
661 | flags |= SIGNOK | NDIGITS | NZDIGITS; | |
662 | for (p = buf; width; width--) { | |
663 | c = *fp->_p; | |
664 | /* | |
665 | * Switch on the character; `goto ok' | |
666 | * if we accept it as a part of number. | |
667 | */ | |
668 | switch (c) { | |
669 | ||
670 | /* | |
671 | * The digit 0 is always legal, but is | |
672 | * special. For %i conversions, if no | |
673 | * digits (zero or nonzero) have been | |
674 | * scanned (only signs), we will have | |
675 | * base==0. In that case, we should set | |
676 | * it to 8 and enable 0x prefixing. | |
677 | * Also, if we have not scanned zero digits | |
678 | * before this, do not turn off prefixing | |
679 | * (someone else will turn it off if we | |
680 | * have scanned any nonzero digits). | |
681 | */ | |
682 | case '0': | |
683 | if (base == 0) { | |
684 | base = 8; | |
685 | flags |= PFXOK; | |
686 | } | |
687 | if (flags & NZDIGITS) | |
688 | flags &= ~(SIGNOK|NZDIGITS|NDIGITS); | |
689 | else | |
690 | flags &= ~(SIGNOK|PFXOK|NDIGITS); | |
691 | goto ok; | |
692 | ||
693 | /* 1 through 7 always legal */ | |
694 | case '1': case '2': case '3': | |
695 | case '4': case '5': case '6': case '7': | |
696 | base = basefix[base]; | |
697 | flags &= ~(SIGNOK | PFXOK | NDIGITS); | |
698 | goto ok; | |
699 | ||
700 | /* digits 8 and 9 ok iff decimal or hex */ | |
701 | case '8': case '9': | |
702 | base = basefix[base]; | |
703 | if (base <= 8) | |
704 | break; /* not legal here */ | |
705 | flags &= ~(SIGNOK | PFXOK | NDIGITS); | |
706 | goto ok; | |
707 | ||
708 | /* letters ok iff hex */ | |
709 | case 'A': case 'B': case 'C': | |
710 | case 'D': case 'E': case 'F': | |
711 | case 'a': case 'b': case 'c': | |
712 | case 'd': case 'e': case 'f': | |
713 | /* no need to fix base here */ | |
714 | if (base <= 10) | |
715 | break; /* not legal here */ | |
716 | flags &= ~(SIGNOK | PFXOK | NDIGITS); | |
717 | goto ok; | |
718 | ||
719 | /* sign ok only as first character */ | |
720 | case '+': case '-': | |
721 | if (flags & SIGNOK) { | |
722 | flags &= ~SIGNOK; | |
723 | flags |= HAVESIGN; | |
724 | goto ok; | |
725 | } | |
726 | break; | |
727 | ||
728 | /* | |
729 | * x ok iff flag still set & 2nd char (or | |
730 | * 3rd char if we have a sign). | |
731 | */ | |
732 | case 'x': case 'X': | |
733 | if (flags & PFXOK && p == | |
734 | buf + 1 + !!(flags & HAVESIGN)) { | |
735 | base = 16; /* if %i */ | |
736 | flags &= ~PFXOK; | |
737 | goto ok; | |
738 | } | |
739 | break; | |
740 | } | |
741 | ||
742 | /* | |
743 | * If we got here, c is not a legal character | |
744 | * for a number. Stop accumulating digits. | |
745 | */ | |
746 | break; | |
747 | ok: | |
748 | /* | |
749 | * c is legal: store it and look at the next. | |
750 | */ | |
751 | *p++ = c; | |
752 | if (--fp->_r > 0) | |
753 | fp->_p++; | |
754 | else if (__srefill(fp)) | |
755 | break; /* EOF */ | |
756 | } | |
757 | /* | |
758 | * If we had only a sign, it is no good; push | |
759 | * back the sign. If the number ends in `x', | |
760 | * it was [sign] '0' 'x', so push back the x | |
761 | * and treat it as [sign] '0'. | |
762 | */ | |
763 | if (flags & NDIGITS) { | |
764 | if (p > buf) | |
765 | (void) __ungetc(*(u_char *)--p, fp); | |
766 | goto match_failure; | |
767 | } | |
768 | c = ((u_char *)p)[-1]; | |
769 | if (c == 'x' || c == 'X') { | |
770 | --p; | |
771 | (void) __ungetc(c, fp); | |
772 | } | |
773 | if ((flags & SUPPRESS) == 0) { | |
774 | uintmax_t res; | |
775 | ||
776 | *p = 0; | |
777 | if ((flags & UNSIGNED) == 0) | |
778 | res = strtoimax_l(buf, (char **)NULL, base, loc); | |
779 | else | |
780 | res = strtoumax_l(buf, (char **)NULL, base, loc); | |
781 | if (flags & POINTER) | |
782 | *va_arg(ap, void **) = | |
783 | (void *)(uintptr_t)res; | |
784 | else if (flags & SHORTSHORT) | |
785 | *va_arg(ap, char *) = res; | |
786 | else if (flags & SHORT) | |
787 | *va_arg(ap, short *) = res; | |
788 | else if (flags & LONG) | |
789 | *va_arg(ap, long *) = res; | |
790 | else if (flags & LONGLONG) | |
791 | *va_arg(ap, long long *) = res; | |
792 | else if (flags & INTMAXT) | |
793 | *va_arg(ap, intmax_t *) = res; | |
794 | else if (flags & PTRDIFFT) | |
795 | *va_arg(ap, ptrdiff_t *) = res; | |
796 | else if (flags & SIZET) | |
797 | *va_arg(ap, size_t *) = res; | |
798 | else | |
799 | *va_arg(ap, int *) = res; | |
800 | nassigned++; | |
801 | } | |
802 | nread += p - buf; | |
803 | break; | |
804 | ||
805 | #ifndef NO_FLOATING_POINT | |
806 | case CT_FLOAT: | |
807 | { | |
808 | char *pbuf; | |
809 | /* scan a floating point number as if by strtod */ | |
810 | if ((width = parsefloat(fp, &pbuf, width, loc)) == 0) | |
811 | goto match_failure; | |
812 | if ((flags & SUPPRESS) == 0) { | |
813 | if (flags & LONGDBL) { | |
814 | long double res = strtold_l(pbuf, &p, loc); | |
815 | *va_arg(ap, long double *) = res; | |
816 | } else if (flags & LONG) { | |
817 | double res = strtod_l(pbuf, &p, loc); | |
818 | *va_arg(ap, double *) = res; | |
819 | } else { | |
820 | float res = strtof_l(pbuf, &p, loc); | |
821 | *va_arg(ap, float *) = res; | |
822 | } | |
823 | nassigned++; | |
824 | } | |
825 | nread += width; | |
826 | break; | |
827 | } | |
828 | #endif /* !NO_FLOATING_POINT */ | |
829 | } | |
830 | } | |
831 | input_failure: | |
832 | return (nassigned ? nassigned : EOF); | |
833 | match_failure: | |
834 | return (nassigned); | |
835 | } | |
836 | ||
837 | int | |
838 | __svfscanf(FILE * __restrict fp, const char * __restrict fmt0, va_list ap) | |
839 | { | |
840 | return __svfscanf_l(fp, __current_locale(), fmt0, ap); | |
841 | } | |
842 | ||
843 | /* | |
844 | * Fill in the given table from the scanset at the given format | |
845 | * (just after `['). Return a pointer to the character past the | |
846 | * closing `]'. The table has a 1 wherever characters should be | |
847 | * considered part of the scanset. | |
848 | */ | |
849 | static const u_char * | |
850 | __sccl(tab, fmt, loc) | |
851 | char *tab; | |
852 | const u_char *fmt; | |
853 | locale_t loc; | |
854 | { | |
855 | int c, n, v, i; | |
856 | ||
857 | /* first `clear' the whole table */ | |
858 | c = *fmt++; /* first char hat => negated scanset */ | |
859 | if (c == '^') { | |
860 | v = 1; /* default => accept */ | |
861 | c = *fmt++; /* get new first char */ | |
862 | } else | |
863 | v = 0; /* default => reject */ | |
864 | ||
865 | /* XXX: Will not work if sizeof(tab*) > sizeof(char) */ | |
866 | (void) memset(tab, v, 256); | |
867 | ||
868 | if (c == 0) | |
869 | return (fmt - 1);/* format ended before closing ] */ | |
870 | ||
871 | /* | |
872 | * Now set the entries corresponding to the actual scanset | |
873 | * to the opposite of the above. | |
874 | * | |
875 | * The first character may be ']' (or '-') without being special; | |
876 | * the last character may be '-'. | |
877 | */ | |
878 | v = 1 - v; | |
879 | for (;;) { | |
880 | tab[c] = v; /* take character c */ | |
881 | doswitch: | |
882 | n = *fmt++; /* and examine the next */ | |
883 | switch (n) { | |
884 | ||
885 | case 0: /* format ended too soon */ | |
886 | return (fmt - 1); | |
887 | ||
888 | case '-': | |
889 | { | |
890 | /* | |
891 | * A scanset of the form | |
892 | * [01+-] | |
893 | * is defined as `the digit 0, the digit 1, | |
894 | * the character +, the character -', but | |
895 | * the effect of a scanset such as | |
896 | * [a-zA-Z0-9] | |
897 | * is implementation defined. The V7 Unix | |
898 | * scanf treats `a-z' as `the letters a through | |
899 | * z', but treats `a-a' as `the letter a, the | |
900 | * character -, and the letter a'. | |
901 | * | |
902 | * For compatibility, the `-' is not considerd | |
903 | * to define a range if the character following | |
904 | * it is either a close bracket (required by ANSI) | |
905 | * or is not numerically greater than the character | |
906 | * we just stored in the table (c). | |
907 | */ | |
908 | n = *fmt; | |
909 | if (n == ']' | |
910 | || (loc->__collate_load_error ? n < c : | |
911 | __collate_range_cmp (n, c, loc) < 0 | |
912 | ) | |
913 | ) { | |
914 | c = '-'; | |
915 | break; /* resume the for(;;) */ | |
916 | } | |
917 | fmt++; | |
918 | /* fill in the range */ | |
919 | if (loc->__collate_load_error) { | |
920 | do { | |
921 | tab[++c] = v; | |
922 | } while (c < n); | |
923 | } else { | |
924 | for (i = 0; i < 256; i ++) | |
925 | if ( __collate_range_cmp (c, i, loc) < 0 | |
926 | && __collate_range_cmp (i, n, loc) <= 0 | |
927 | ) | |
928 | tab[i] = v; | |
929 | } | |
930 | #if 1 /* XXX another disgusting compatibility hack */ | |
931 | c = n; | |
932 | /* | |
933 | * Alas, the V7 Unix scanf also treats formats | |
934 | * such as [a-c-e] as `the letters a through e'. | |
935 | * This too is permitted by the standard.... | |
936 | */ | |
937 | goto doswitch; | |
938 | #else | |
939 | c = *fmt++; | |
940 | if (c == 0) | |
941 | return (fmt - 1); | |
942 | if (c == ']') | |
943 | return (fmt); | |
944 | #endif | |
945 | break; | |
946 | } | |
947 | case ']': /* end of scanset */ | |
948 | return (fmt); | |
949 | ||
950 | default: /* just another character */ | |
951 | c = n; | |
952 | break; | |
953 | } | |
954 | } | |
955 | /* NOTREACHED */ | |
956 | } | |
957 | ||
958 | #ifndef NO_FLOATING_POINT | |
959 | /* | |
960 | * Maintain a per-thread parsefloat buffer, shared by __svfscanf_l and | |
961 | * __vfwscanf. | |
962 | */ | |
963 | #ifdef BUILDING_VARIANT | |
964 | extern char *__parsefloat_buf(size_t s); | |
965 | #else /* !BUILDING_VARIANT */ | |
966 | __private_extern__ char * | |
967 | __parsefloat_buf(size_t s) | |
968 | { | |
969 | char *b; | |
970 | static pthread_key_t parsefloat_tsd_key = (pthread_key_t)-1; | |
971 | static pthread_mutex_t parsefloat_tsd_lock = PTHREAD_MUTEX_INITIALIZER; | |
972 | static size_t bsiz = 0; | |
973 | ||
974 | if (parsefloat_tsd_key == (pthread_key_t)-1) { | |
975 | pthread_mutex_lock(&parsefloat_tsd_lock); | |
976 | if (parsefloat_tsd_key == (pthread_key_t)-1) { | |
977 | parsefloat_tsd_key = __LIBC_PTHREAD_KEY_PARSEFLOAT; | |
978 | pthread_key_init_np(parsefloat_tsd_key, free); | |
979 | } | |
980 | pthread_mutex_unlock(&parsefloat_tsd_lock); | |
981 | } | |
982 | if ((b = (char *)pthread_getspecific(parsefloat_tsd_key)) == NULL) { | |
983 | bsiz = s > BUF ? s : BUF; | |
984 | b = (char *)malloc(bsiz); | |
985 | if (b == NULL) { | |
986 | bsiz = 0; | |
987 | return NULL; | |
988 | } | |
989 | pthread_setspecific(parsefloat_tsd_key, b); | |
990 | return b; | |
991 | } | |
992 | if (s > bsiz) { | |
993 | b = (char *)reallocf(b, s); | |
994 | pthread_setspecific(parsefloat_tsd_key, b); | |
995 | if (b == NULL) { | |
996 | bsiz = 0; | |
997 | return NULL; | |
998 | } | |
999 | bsiz = s; | |
1000 | } | |
1001 | return b; | |
1002 | } | |
1003 | #endif /* BUILDING_VARIANT */ | |
1004 | ||
1005 | static int | |
1006 | parsefloat(FILE *fp, char **buf, size_t width, locale_t loc) | |
1007 | { | |
1008 | char *commit, *p; | |
1009 | int infnanpos = 0, decptpos = 0; | |
1010 | enum { | |
1011 | S_START, S_GOTSIGN, S_INF, S_NAN, S_DONE, S_MAYBEHEX, | |
1012 | S_DIGITS, S_DECPT, S_FRAC, S_EXP, S_EXPDIGITS | |
1013 | } state = S_START; | |
1014 | unsigned char c; | |
1015 | const char *decpt = localeconv_l(loc)->decimal_point; | |
1016 | _Bool gotmantdig = 0, ishex = 0; | |
1017 | char *b; | |
1018 | char *e; | |
1019 | size_t s; | |
1020 | ||
1021 | s = (width == 0 ? BUF : (width + 1)); | |
1022 | if ((b = __parsefloat_buf(s)) == NULL) { | |
1023 | *buf = NULL; | |
1024 | return 0; | |
1025 | } | |
1026 | e = b + (s - 1); | |
1027 | /* | |
1028 | * We set commit = p whenever the string we have read so far | |
1029 | * constitutes a valid representation of a floating point | |
1030 | * number by itself. At some point, the parse will complete | |
1031 | * or fail, and we will ungetc() back to the last commit point. | |
1032 | * To ensure that the file offset gets updated properly, it is | |
1033 | * always necessary to read at least one character that doesn't | |
1034 | * match; thus, we can't short-circuit "infinity" or "nan(...)". | |
1035 | */ | |
1036 | commit = b - 1; | |
1037 | for (p = b; width == 0 || p < e; ) { | |
1038 | c = *fp->_p; | |
1039 | reswitch: | |
1040 | switch (state) { | |
1041 | case S_START: | |
1042 | state = S_GOTSIGN; | |
1043 | if (c == '-' || c == '+') | |
1044 | break; | |
1045 | else | |
1046 | goto reswitch; | |
1047 | case S_GOTSIGN: | |
1048 | switch (c) { | |
1049 | case '0': | |
1050 | state = S_MAYBEHEX; | |
1051 | commit = p; | |
1052 | break; | |
1053 | case 'I': | |
1054 | case 'i': | |
1055 | state = S_INF; | |
1056 | break; | |
1057 | case 'N': | |
1058 | case 'n': | |
1059 | state = S_NAN; | |
1060 | break; | |
1061 | default: | |
1062 | state = S_DIGITS; | |
1063 | goto reswitch; | |
1064 | } | |
1065 | break; | |
1066 | case S_INF: | |
1067 | if (infnanpos > 6 || | |
1068 | (c != "nfinity"[infnanpos] && | |
1069 | c != "NFINITY"[infnanpos])) | |
1070 | goto parsedone; | |
1071 | if (infnanpos == 1 || infnanpos == 6) | |
1072 | commit = p; /* inf or infinity */ | |
1073 | infnanpos++; | |
1074 | break; | |
1075 | case S_NAN: | |
1076 | switch (infnanpos) { | |
1077 | case 0: | |
1078 | if (c != 'A' && c != 'a') | |
1079 | goto parsedone; | |
1080 | break; | |
1081 | case 1: | |
1082 | if (c != 'N' && c != 'n') | |
1083 | goto parsedone; | |
1084 | else | |
1085 | commit = p; | |
1086 | break; | |
1087 | case 2: | |
1088 | if (c != '(') | |
1089 | goto parsedone; | |
1090 | break; | |
1091 | default: | |
1092 | if (c == ')') { | |
1093 | commit = p; | |
1094 | state = S_DONE; | |
1095 | } else if (!isalnum_l(c, loc) && c != '_') | |
1096 | goto parsedone; | |
1097 | break; | |
1098 | } | |
1099 | infnanpos++; | |
1100 | break; | |
1101 | case S_DONE: | |
1102 | goto parsedone; | |
1103 | case S_MAYBEHEX: | |
1104 | state = S_DIGITS; | |
1105 | if (c == 'X' || c == 'x') { | |
1106 | ishex = 1; | |
1107 | break; | |
1108 | } else { /* we saw a '0', but no 'x' */ | |
1109 | gotmantdig = 1; | |
1110 | goto reswitch; | |
1111 | } | |
1112 | case S_DIGITS: | |
1113 | if ((ishex && isxdigit_l(c, loc)) || isdigit_l(c, loc)) { | |
1114 | gotmantdig = 1; | |
1115 | commit = p; | |
1116 | break; | |
1117 | } else { | |
1118 | state = S_DECPT; | |
1119 | goto reswitch; | |
1120 | } | |
1121 | case S_DECPT: | |
1122 | if (c == decpt[decptpos]) { | |
1123 | if (decpt[++decptpos] == '\0') { | |
1124 | /* We read the complete decpt seq. */ | |
1125 | state = S_FRAC; | |
1126 | if (gotmantdig) | |
1127 | commit = p; | |
1128 | } | |
1129 | break; | |
1130 | } else if (!decptpos) { | |
1131 | /* We didn't read any decpt characters. */ | |
1132 | state = S_FRAC; | |
1133 | goto reswitch; | |
1134 | } else { | |
1135 | /* | |
1136 | * We read part of a multibyte decimal point, | |
1137 | * but the rest is invalid, so bail. | |
1138 | */ | |
1139 | goto parsedone; | |
1140 | } | |
1141 | case S_FRAC: | |
1142 | if (((c == 'E' || c == 'e') && !ishex) || | |
1143 | ((c == 'P' || c == 'p') && ishex)) { | |
1144 | if (!gotmantdig) | |
1145 | goto parsedone; | |
1146 | else | |
1147 | state = S_EXP; | |
1148 | } else if ((ishex && isxdigit_l(c, loc)) || isdigit_l(c, loc)) { | |
1149 | commit = p; | |
1150 | gotmantdig = 1; | |
1151 | } else | |
1152 | goto parsedone; | |
1153 | break; | |
1154 | case S_EXP: | |
1155 | state = S_EXPDIGITS; | |
1156 | if (c == '-' || c == '+') | |
1157 | break; | |
1158 | else | |
1159 | goto reswitch; | |
1160 | case S_EXPDIGITS: | |
1161 | if (isdigit_l(c, loc)) | |
1162 | commit = p; | |
1163 | else | |
1164 | goto parsedone; | |
1165 | break; | |
1166 | default: | |
1167 | LIBC_ABORT("unknown state %d", state); | |
1168 | } | |
1169 | if (p >= e) { | |
1170 | ssize_t diff = (p - b); | |
1171 | ssize_t com = (commit - b); | |
1172 | s += BUF; | |
1173 | b = __parsefloat_buf(s); | |
1174 | if (b == NULL) { | |
1175 | *buf = NULL; | |
1176 | return 0; | |
1177 | } | |
1178 | e = b + (s - 1); | |
1179 | p = b + diff; | |
1180 | commit = b + com; | |
1181 | } | |
1182 | *p++ = c; | |
1183 | if (--fp->_r > 0) | |
1184 | fp->_p++; | |
1185 | else if (__srefill(fp)) | |
1186 | break; /* EOF */ | |
1187 | } | |
1188 | ||
1189 | parsedone: | |
1190 | while (commit < --p) | |
1191 | __ungetc(*(u_char *)p, fp); | |
1192 | *++commit = '\0'; | |
1193 | *buf = b; | |
1194 | return (commit - b); | |
1195 | } | |
1196 | #endif |