1 /*-
2 * Copyright (c) 1990, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Copyright (c) 2011 The FreeBSD Foundation
6 * All rights reserved.
7 * Portions of this software were developed by David Chisnall
8 * under sponsorship from the FreeBSD Foundation.
9 *
10 * This code is derived from software contributed to Berkeley by
11 * Chris Torek.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * @(#)vfscanf.c 8.1 (Berkeley) 6/4/93
38 * $FreeBSD: head/lib/libc/stdio/vfscanf.c 249808 2013-04-23 13:33:13Z emaste $
39 */
40
41
42 #include "namespace.h"
43 #include <ctype.h>
44 #include <inttypes.h>
45 #include <stdio.h>
46 #include <stdlib.h>
47 #include <stddef.h>
48 #include <stdarg.h>
49 #include <string.h>
50 #include <wchar.h>
51 #include <wctype.h>
52 #include "un-namespace.h"
53
54 #include "collate.h"
55 #include "libc_private.h"
56 #include "local.h"
57 #include "xlocale_private.h"
58
59 #ifndef NO_FLOATING_POINT
60 #include <locale.h>
61 #endif
62
63 #define BUF 513 /* Maximum length of numeric string. */
64
65 /*
66 * Flags used during conversion.
67 */
68 #define LONG 0x01 /* l: long or double */
69 #define LONGDBL 0x02 /* L: long double */
70 #define SHORT 0x04 /* h: short */
71 #define SUPPRESS 0x08 /* *: suppress assignment */
72 #define POINTER 0x10 /* p: void * (as hex) */
73 #define NOSKIP 0x20 /* [ or c: do not skip blanks */
74 #ifdef __DragonFly__ /* Non-standard extension: L and ll are equivalent */
75 #define LONGLONG LONGDBL /* ll/L: long long, long double */
76 /* (+ deprecated q: quad) */
77 #else
78 #define LONGLONG 0x400 /* ll: long long (+ deprecated q: quad) */
79 #endif
80 #define INTMAXT 0x800 /* j: intmax_t */
81 #define PTRDIFFT 0x1000 /* t: ptrdiff_t */
82 #define SIZET 0x2000 /* z: size_t */
83 #define SHORTSHORT 0x4000 /* hh: char */
84 #define UNSIGNED 0x8000 /* %[oupxX] conversions */
85
86 /*
87 * The following are used in integral conversions only:
88 * SIGNOK, NDIGITS, PFXOK, and NZDIGITS
89 */
90 #define SIGNOK 0x40 /* +/- is (still) legal */
91 #define NDIGITS 0x80 /* no digits detected */
92 #define PFXOK 0x100 /* 0x prefix is (still) legal */
93 #define NZDIGITS 0x200 /* no zero digits detected */
94 #define HAVESIGN 0x10000 /* sign detected */
95
96 /*
97 * Conversion types.
98 */
99 #define CT_CHAR 0 /* %c conversion */
100 #define CT_CCL 1 /* %[...] conversion */
101 #define CT_STRING 2 /* %s conversion */
102 #define CT_INT 3 /* %[dioupxX] conversion */
103 #define CT_FLOAT 4 /* %[efgEFG] conversion */
104
105 static const u_char *__sccl(char *, const u_char *);
106 #ifndef NO_FLOATING_POINT
107 static int parsefloat(FILE *, char *, char *, locale_t);
108 #endif
109
110 /*
111 * Conversion functions are passed a pointer to this object instead of
112 * a real parameter to indicate that the assignment-suppression (*)
113 * flag was specified. We could use a NULL pointer to indicate this,
114 * but that would mask bugs in applications that call scanf() with a
115 * NULL pointer.
116 */
117 static const int suppress;
118 #define SUPPRESS_PTR ((void *)&suppress)
119
120 static const mbstate_t initial_mbs;
121
122 /*
123 * The following conversion functions return the number of characters consumed,
124 * or -1 on input failure. Character class conversion returns 0 on match
125 * failure.
126 */
127
128 static __inline int
convert_char(FILE * fp,char * p,int width)129 convert_char(FILE *fp, char * p, int width)
130 {
131 int n;
132
133 if (p == SUPPRESS_PTR) {
134 size_t sum = 0;
135 for (;;) {
136 if ((n = fp->pub._r) < width) {
137 sum += n;
138 width -= n;
139 fp->pub._p += n;
140 if (__srefill(fp)) {
141 if (sum == 0)
142 return (-1);
143 break;
144 }
145 } else {
146 sum += width;
147 fp->pub._r -= width;
148 fp->pub._p += width;
149 break;
150 }
151 }
152 return (sum);
153 } else {
154 size_t r = __fread(p, 1, width, fp);
155
156 if (r == 0)
157 return (-1);
158 return (r);
159 }
160 }
161
162 static __inline int
convert_wchar(FILE * fp,wchar_t * wcp,int width,locale_t locale)163 convert_wchar(FILE *fp, wchar_t *wcp, int width, locale_t locale)
164 {
165 mbstate_t mbs;
166 int n, nread;
167 wint_t wi;
168
169 mbs = initial_mbs;
170 n = 0;
171 while (width-- != 0 &&
172 (wi = __fgetwc_mbs(fp, &mbs, &nread, locale)) != WEOF) {
173 if (wcp != SUPPRESS_PTR)
174 *wcp++ = (wchar_t)wi;
175 n += nread;
176 }
177 if (n == 0)
178 return (-1);
179 return (n);
180 }
181
182 static __inline int
convert_ccl(FILE * fp,char * p,int width,const char * ccltab)183 convert_ccl(FILE *fp, char * p, int width, const char *ccltab)
184 {
185 char *p0;
186 int n;
187
188 if (p == SUPPRESS_PTR) {
189 n = 0;
190 while (ccltab[*fp->pub._p]) {
191 n++, fp->pub._r--, fp->pub._p++;
192 if (--width == 0)
193 break;
194 if (fp->pub._r <= 0 && __srefill(fp)) {
195 if (n == 0)
196 return (-1);
197 break;
198 }
199 }
200 } else {
201 p0 = p;
202 while (ccltab[*fp->pub._p]) {
203 fp->pub._r--;
204 *p++ = *fp->pub._p++;
205 if (--width == 0)
206 break;
207 if (fp->pub._r <= 0 && __srefill(fp)) {
208 if (p == p0)
209 return (-1);
210 break;
211 }
212 }
213 n = p - p0;
214 if (n == 0)
215 return (0);
216 *p = 0;
217 }
218 return (n);
219 }
220
221 static __inline int
convert_wccl(FILE * fp,wchar_t * wcp,int width,const char * ccltab,locale_t locale)222 convert_wccl(FILE *fp, wchar_t *wcp, int width, const char *ccltab,
223 locale_t locale)
224 {
225 mbstate_t mbs;
226 wint_t wi;
227 int n, nread;
228
229 mbs = initial_mbs;
230 n = 0;
231 if (wcp == SUPPRESS_PTR) {
232 while ((wi = __fgetwc_mbs(fp, &mbs, &nread, locale)) != WEOF &&
233 width-- != 0 && ccltab[wctob(wi)])
234 n += nread;
235 if (wi != WEOF)
236 __ungetwc(wi, fp, __get_locale());
237 } else {
238 while ((wi = __fgetwc_mbs(fp, &mbs, &nread, locale)) != WEOF &&
239 width-- != 0 && ccltab[wctob(wi)]) {
240 *wcp++ = (wchar_t)wi;
241 n += nread;
242 }
243 if (wi != WEOF)
244 __ungetwc(wi, fp, __get_locale());
245 if (n == 0)
246 return (0);
247 *wcp = 0;
248 }
249 return (n);
250 }
251
252 static __inline int
convert_string(FILE * fp,char * p,int width)253 convert_string(FILE *fp, char * p, int width)
254 {
255 char *p0;
256 int n;
257
258 if (p == SUPPRESS_PTR) {
259 n = 0;
260 while (!isspace(*fp->pub._p)) {
261 n++, fp->pub._r--, fp->pub._p++;
262 if (--width == 0)
263 break;
264 if (fp->pub._r <= 0 && __srefill(fp))
265 break;
266 }
267 } else {
268 p0 = p;
269 while (!isspace(*fp->pub._p)) {
270 fp->pub._r--;
271 *p++ = *fp->pub._p++;
272 if (--width == 0)
273 break;
274 if (fp->pub._r <= 0 && __srefill(fp))
275 break;
276 }
277 *p = 0;
278 n = p - p0;
279 }
280 return (n);
281 }
282
283 static __inline int
convert_wstring(FILE * fp,wchar_t * wcp,int width,locale_t locale)284 convert_wstring(FILE *fp, wchar_t *wcp, int width, locale_t locale)
285 {
286 mbstate_t mbs;
287 wint_t wi;
288 int n, nread;
289
290 mbs = initial_mbs;
291 n = 0;
292 if (wcp == SUPPRESS_PTR) {
293 while ((wi = __fgetwc_mbs(fp, &mbs, &nread, locale)) != WEOF &&
294 width-- != 0 && !iswspace(wi))
295 n += nread;
296 if (wi != WEOF)
297 __ungetwc(wi, fp, __get_locale());
298 } else {
299 while ((wi = __fgetwc_mbs(fp, &mbs, &nread, locale)) != WEOF &&
300 width-- != 0 && !iswspace(wi)) {
301 *wcp++ = (wchar_t)wi;
302 n += nread;
303 }
304 if (wi != WEOF)
305 __ungetwc(wi, fp, __get_locale());
306 *wcp = '\0';
307 }
308 return (n);
309 }
310
311 /*
312 * Read an integer, storing it in buf. The only relevant bit in the
313 * flags argument is PFXOK.
314 *
315 * Return 0 on a match failure, and the number of characters read
316 * otherwise.
317 */
318 static __inline int
parseint(FILE * fp,char * __restrict buf,int width,int base,int flags)319 parseint(FILE *fp, char * __restrict buf, int width, int base, int flags)
320 {
321 /* `basefix' is used to avoid `if' tests */
322 static const short basefix[17] =
323 { 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 };
324 char *p;
325 int c;
326
327 flags |= SIGNOK | NDIGITS | NZDIGITS;
328 for (p = buf; width; width--) {
329 c = *fp->pub._p;
330 /*
331 * Switch on the character; `goto ok' if we accept it
332 * as a part of number.
333 */
334 switch (c) {
335
336 /*
337 * The digit 0 is always legal, but is special. For
338 * %i conversions, if no digits (zero or nonzero) have
339 * been scanned (only signs), we will have base==0.
340 * In that case, we should set it to 8 and enable 0x
341 * prefixing. Also, if we have not scanned zero
342 * digits before this, do not turn off prefixing
343 * (someone else will turn it off if we have scanned
344 * any nonzero digits).
345 */
346 case '0':
347 if (base == 0) {
348 base = 8;
349 flags |= PFXOK;
350 }
351 if (flags & NZDIGITS)
352 flags &= ~(SIGNOK|NZDIGITS|NDIGITS);
353 else
354 flags &= ~(SIGNOK|PFXOK|NDIGITS);
355 goto ok;
356
357 /* 1 through 7 always legal */
358 case '1': case '2': case '3':
359 case '4': case '5': case '6': case '7':
360 base = basefix[base];
361 flags &= ~(SIGNOK | PFXOK | NDIGITS);
362 goto ok;
363
364 /* digits 8 and 9 ok iff decimal or hex */
365 case '8': case '9':
366 base = basefix[base];
367 if (base <= 8)
368 break; /* not legal here */
369 flags &= ~(SIGNOK | PFXOK | NDIGITS);
370 goto ok;
371
372 /* letters ok iff hex */
373 case 'A': case 'B': case 'C':
374 case 'D': case 'E': case 'F':
375 case 'a': case 'b': case 'c':
376 case 'd': case 'e': case 'f':
377 /* no need to fix base here */
378 if (base <= 10)
379 break; /* not legal here */
380 flags &= ~(SIGNOK | PFXOK | NDIGITS);
381 goto ok;
382
383 /* sign ok only as first character */
384 case '+': case '-':
385 if (flags & SIGNOK) {
386 flags &= ~SIGNOK;
387 flags |= HAVESIGN;
388 goto ok;
389 }
390 break;
391
392 /*
393 * x ok iff flag still set & 2nd char (or 3rd char if
394 * we have a sign).
395 */
396 case 'x': case 'X':
397 if (flags & PFXOK && p ==
398 buf + 1 + !!(flags & HAVESIGN)) {
399 base = 16; /* if %i */
400 flags &= ~PFXOK;
401 goto ok;
402 }
403 break;
404 }
405
406 /*
407 * If we got here, c is not a legal character for a
408 * number. Stop accumulating digits.
409 */
410 break;
411 ok:
412 /*
413 * c is legal: store it and look at the next.
414 */
415 *p++ = c;
416 if (--fp->pub._r > 0)
417 fp->pub._p++;
418 else if (__srefill(fp))
419 break; /* EOF */
420 }
421 /*
422 * If we had only a sign, it is no good; push back the sign.
423 * If the number ends in `x', it was [sign] '0' 'x', so push
424 * back the x and treat it as [sign] '0'.
425 */
426 if (flags & NDIGITS) {
427 if (p > buf)
428 (void) __ungetc(*(u_char *)--p, fp);
429 return (0);
430 }
431 c = ((u_char *)p)[-1];
432 if (c == 'x' || c == 'X') {
433 --p;
434 (void) __ungetc(c, fp);
435 }
436 return (p - buf);
437 }
438
439 /*
440 * __vfscanf - MT-safe version
441 */
442 int
__vfscanf(FILE * fp,char const * fmt0,va_list ap)443 __vfscanf(FILE *fp, char const *fmt0, va_list ap)
444 {
445 int ret;
446
447 FLOCKFILE(fp);
448 ret = __svfscanf(fp, __get_locale(), fmt0, ap);
449 FUNLOCKFILE(fp);
450 return (ret);
451 }
452
453 __weak_reference(__vfscanf, vfscanf);
454
455 int
vfscanf_l(FILE * fp,locale_t locale,char const * fmt0,va_list ap)456 vfscanf_l(FILE *fp, locale_t locale, char const *fmt0, va_list ap)
457 {
458 int ret;
459 FIX_LOCALE(locale);
460
461 FLOCKFILE(fp);
462 ret = __svfscanf(fp, locale, fmt0, ap);
463 FUNLOCKFILE(fp);
464 return (ret);
465 }
466
467 /*
468 * __svfscanf - non-MT-safe version of __vfscanf
469 */
470 int
__svfscanf(FILE * fp,locale_t locale,const char * fmt0,va_list ap)471 __svfscanf(FILE *fp, locale_t locale, const char *fmt0, va_list ap)
472 {
473 #define GETARG(type) ((flags & SUPPRESS) ? SUPPRESS_PTR : va_arg(ap, type))
474 const u_char *fmt = (const u_char *)fmt0;
475 int c; /* character from format, or conversion */
476 size_t width; /* field width, or 0 */
477 int flags; /* flags as defined above */
478 int nassigned; /* number of fields assigned */
479 int nconversions; /* number of conversions */
480 int nr; /* characters read by the current conversion */
481 int nread; /* number of characters consumed from fp */
482 int base; /* base argument to conversion function */
483 char ccltab[256]; /* character class table for %[...] */
484 char buf[BUF]; /* buffer for numeric conversions */
485
486 ORIENT(fp, -1);
487
488 nassigned = 0;
489 nconversions = 0;
490 nread = 0;
491 for (;;) {
492 c = *fmt++;
493 if (c == 0)
494 return (nassigned);
495 if (isspace(c)) {
496 while ((fp->pub._r > 0 || __srefill(fp) == 0) && isspace(*fp->pub._p))
497 nread++, fp->pub._r--, fp->pub._p++;
498 continue;
499 }
500 if (c != '%')
501 goto literal;
502 width = 0;
503 flags = 0;
504 /*
505 * switch on the format. continue if done;
506 * break once format type is derived.
507 */
508 again: c = *fmt++;
509 switch (c) {
510 case '%':
511 literal:
512 if (fp->pub._r <= 0 && __srefill(fp))
513 goto input_failure;
514 if (*fp->pub._p != c)
515 goto match_failure;
516 fp->pub._r--, fp->pub._p++;
517 nread++;
518 continue;
519
520 case '*':
521 flags |= SUPPRESS;
522 goto again;
523 case 'j':
524 flags |= INTMAXT;
525 goto again;
526 case 'l':
527 if (flags & LONG) {
528 flags &= ~LONG;
529 flags |= LONGLONG;
530 } else
531 flags |= LONG;
532 goto again;
533 case 'q':
534 flags |= LONGLONG; /* not quite */
535 goto again;
536 case 't':
537 flags |= PTRDIFFT;
538 goto again;
539 case 'z':
540 flags |= SIZET;
541 goto again;
542 case 'L':
543 flags |= LONGDBL;
544 goto again;
545 case 'h':
546 if (flags & SHORT) {
547 flags &= ~SHORT;
548 flags |= SHORTSHORT;
549 } else
550 flags |= SHORT;
551 goto again;
552
553 case '0': case '1': case '2': case '3': case '4':
554 case '5': case '6': case '7': case '8': case '9':
555 width = width * 10 + c - '0';
556 goto again;
557
558 /*
559 * Conversions.
560 */
561 case 'd':
562 c = CT_INT;
563 base = 10;
564 break;
565
566 case 'i':
567 c = CT_INT;
568 base = 0;
569 break;
570
571 case 'o':
572 c = CT_INT;
573 flags |= UNSIGNED;
574 base = 8;
575 break;
576
577 case 'u':
578 c = CT_INT;
579 flags |= UNSIGNED;
580 base = 10;
581 break;
582
583 case 'X':
584 case 'x':
585 flags |= PFXOK; /* enable 0x prefixing */
586 c = CT_INT;
587 flags |= UNSIGNED;
588 base = 16;
589 break;
590
591 #ifndef NO_FLOATING_POINT
592 case 'A': case 'E': case 'F': case 'G':
593 case 'a': case 'e': case 'f': case 'g':
594 c = CT_FLOAT;
595 break;
596 #endif
597
598 case 'S':
599 flags |= LONG;
600 /* FALLTHROUGH */
601 case 's':
602 c = CT_STRING;
603 break;
604
605 case '[':
606 fmt = __sccl(ccltab, fmt);
607 flags |= NOSKIP;
608 c = CT_CCL;
609 break;
610
611 case 'C':
612 flags |= LONG;
613 /* FALLTHROUGH */
614 case 'c':
615 flags |= NOSKIP;
616 c = CT_CHAR;
617 break;
618
619 case 'p': /* pointer format is like hex */
620 flags |= POINTER | PFXOK;
621 c = CT_INT; /* assumes sizeof(uintmax_t) */
622 flags |= UNSIGNED; /* >= sizeof(uintptr_t) */
623 base = 16;
624 break;
625
626 case 'n':
627 if (flags & SUPPRESS) /* ??? */
628 continue;
629 if (flags & SHORTSHORT)
630 *va_arg(ap, char *) = nread;
631 else if (flags & SHORT)
632 *va_arg(ap, short *) = nread;
633 else if (flags & LONG)
634 *va_arg(ap, long *) = nread;
635 else if (flags & LONGLONG)
636 *va_arg(ap, long long *) = nread;
637 else if (flags & INTMAXT)
638 *va_arg(ap, intmax_t *) = nread;
639 else if (flags & SIZET)
640 *va_arg(ap, size_t *) = nread;
641 else if (flags & PTRDIFFT)
642 *va_arg(ap, ptrdiff_t *) = nread;
643 else
644 *va_arg(ap, int *) = nread;
645 continue;
646
647 default:
648 goto match_failure;
649
650 /*
651 * Disgusting backwards compatibility hack. XXX
652 */
653 case '\0': /* compat */
654 return (EOF);
655 }
656
657 /*
658 * We have a conversion that requires input.
659 */
660 if (fp->pub._r <= 0 && __srefill(fp))
661 goto input_failure;
662
663 /*
664 * Consume leading white space, except for formats
665 * that suppress this.
666 */
667 if ((flags & NOSKIP) == 0) {
668 while (isspace(*fp->pub._p)) {
669 nread++;
670 if (--fp->pub._r > 0)
671 fp->pub._p++;
672 else if (__srefill(fp))
673 goto input_failure;
674 }
675 /*
676 * Note that there is at least one character in
677 * the buffer, so conversions that do not set NOSKIP
678 * ca no longer result in an input failure.
679 */
680 }
681
682 /*
683 * Do the conversion.
684 */
685 switch (c) {
686
687 case CT_CHAR:
688 /* scan arbitrary characters (sets NOSKIP) */
689 if (width == 0)
690 width = 1;
691 if (flags & LONG) {
692 nr = convert_wchar(fp, GETARG(wchar_t *),
693 width, locale);
694 } else {
695 nr = convert_char(fp, GETARG(char *), width);
696 }
697 if (nr < 0)
698 goto input_failure;
699 break;
700
701 case CT_CCL:
702 /* scan a (nonempty) character class (sets NOSKIP) */
703 if (width == 0)
704 width = (size_t)~0; /* `infinity' */
705 if (flags & LONG) {
706 nr = convert_wccl(fp, GETARG(wchar_t *), width,
707 ccltab, locale);
708 } else {
709 nr = convert_ccl(fp, GETARG(char *), width,
710 ccltab);
711 }
712 if (nr <= 0) {
713 if (nr < 0)
714 goto input_failure;
715 else /* nr == 0 */
716 goto match_failure;
717 }
718 break;
719
720 case CT_STRING:
721 /* like CCL, but zero-length string OK, & no NOSKIP */
722 if (width == 0)
723 width = (size_t)~0;
724 if (flags & LONG) {
725 nr = convert_wstring(fp, GETARG(wchar_t *),
726 width, locale);
727 } else {
728 nr = convert_string(fp, GETARG(char *), width);
729 }
730 if (nr < 0)
731 goto input_failure;
732 break;
733
734 case CT_INT:
735 /* scan an integer as if by the conversion function */
736 #ifdef hardway
737 if (width == 0 || width > sizeof(buf) - 1)
738 width = sizeof(buf) - 1;
739 #else
740 /* size_t is unsigned, hence this optimisation */
741 if (--width > sizeof(buf) - 2)
742 width = sizeof(buf) - 2;
743 width++;
744 #endif
745 nr = parseint(fp, buf, width, base, flags);
746 if (nr == 0)
747 goto match_failure;
748 if ((flags & SUPPRESS) == 0) {
749 uintmax_t res;
750
751 buf[nr] = '\0';
752 if ((flags & UNSIGNED) == 0)
753 res = strtoimax_l(buf, NULL, base, locale);
754 else
755 res = strtoumax_l(buf, NULL, base, locale);
756 if (flags & POINTER)
757 *va_arg(ap, void **) =
758 (void *)(uintptr_t)res;
759 else if (flags & SHORTSHORT)
760 *va_arg(ap, char *) = res;
761 else if (flags & SHORT)
762 *va_arg(ap, short *) = res;
763 else if (flags & LONG)
764 *va_arg(ap, long *) = res;
765 else if (flags & LONGLONG)
766 *va_arg(ap, long long *) = res;
767 else if (flags & INTMAXT)
768 *va_arg(ap, intmax_t *) = res;
769 else if (flags & PTRDIFFT)
770 *va_arg(ap, ptrdiff_t *) = res;
771 else if (flags & SIZET)
772 *va_arg(ap, size_t *) = res;
773 else
774 *va_arg(ap, int *) = res;
775 }
776 break;
777
778 #ifndef NO_FLOATING_POINT
779 case CT_FLOAT:
780 /* scan a floating point number as if by strtod */
781 if (width == 0 || width > sizeof(buf) - 1)
782 width = sizeof(buf) - 1;
783 nr = parsefloat(fp, buf, buf + width, locale);
784 if (nr == 0)
785 goto match_failure;
786 if ((flags & SUPPRESS) == 0) {
787 if (flags & LONGDBL) {
788 long double res = strtold_l(buf, NULL,
789 locale);
790 *va_arg(ap, long double *) = res;
791 } else if (flags & LONG) {
792 double res = strtod_l(buf, NULL,
793 locale);
794 *va_arg(ap, double *) = res;
795 } else {
796 float res = strtof_l(buf, NULL, locale);
797 *va_arg(ap, float *) = res;
798 }
799 }
800 break;
801 #endif /* !NO_FLOATING_POINT */
802 }
803 if (!(flags & SUPPRESS))
804 nassigned++;
805 nread += nr;
806 nconversions++;
807 }
808 input_failure:
809 return (nconversions != 0 ? nassigned : EOF);
810 match_failure:
811 return (nassigned);
812 }
813
814 /*
815 * Fill in the given table from the scanset at the given format
816 * (just after `['). Return a pointer to the character past the
817 * closing `]'. The table has a 1 wherever characters should be
818 * considered part of the scanset.
819 */
820 static const u_char *
__sccl(char * tab,const u_char * fmt)821 __sccl(char *tab, const u_char *fmt)
822 {
823 int c, n, v, i;
824 struct xlocale_collate *table =
825 (struct xlocale_collate*)__get_locale()->components[XLC_COLLATE];
826
827 /* first `clear' the whole table */
828 c = *fmt++; /* first char hat => negated scanset */
829 if (c == '^') {
830 v = 1; /* default => accept */
831 c = *fmt++; /* get new first char */
832 } else
833 v = 0; /* default => reject */
834
835 /* XXX: Will not work if sizeof(tab*) > sizeof(char) */
836 (void) memset(tab, v, 256);
837
838 if (c == 0)
839 return (fmt - 1);/* format ended before closing ] */
840
841 /*
842 * Now set the entries corresponding to the actual scanset
843 * to the opposite of the above.
844 *
845 * The first character may be ']' (or '-') without being special;
846 * the last character may be '-'.
847 */
848 v = 1 - v;
849 for (;;) {
850 tab[c] = v; /* take character c */
851 doswitch:
852 n = *fmt++; /* and examine the next */
853 switch (n) {
854
855 case 0: /* format ended too soon */
856 return (fmt - 1);
857
858 case '-':
859 /*
860 * A scanset of the form
861 * [01+-]
862 * is defined as `the digit 0, the digit 1,
863 * the character +, the character -', but
864 * the effect of a scanset such as
865 * [a-zA-Z0-9]
866 * is implementation defined. The V7 Unix
867 * scanf treats `a-z' as `the letters a through
868 * z', but treats `a-a' as `the letter a, the
869 * character -, and the letter a'.
870 *
871 * For compatibility, the `-' is not considered
872 * to define a range if the character following
873 * it is either a close bracket (required by ANSI)
874 * or is not numerically greater than the character
875 * we just stored in the table (c).
876 */
877 n = *fmt;
878 if (n == ']'
879 || (table->__collate_load_error ? n < c :
880 __collate_range_cmp(n, c) < 0
881 )
882 ) {
883 c = '-';
884 break; /* resume the for(;;) */
885 }
886 fmt++;
887 /* fill in the range */
888 if (table->__collate_load_error) {
889 do {
890 tab[++c] = v;
891 } while (c < n);
892 } else {
893 for (i = 0; i < 256; i ++)
894 if (__collate_range_cmp(c, i) <= 0 &&
895 __collate_range_cmp(i, n) <= 0
896 )
897 tab[i] = v;
898 }
899 #if 1 /* XXX another disgusting compatibility hack */
900 c = n;
901 /*
902 * Alas, the V7 Unix scanf also treats formats
903 * such as [a-c-e] as `the letters a through e'.
904 * This too is permitted by the standard....
905 */
906 goto doswitch;
907 #else
908 c = *fmt++;
909 if (c == 0)
910 return (fmt - 1);
911 if (c == ']')
912 return (fmt);
913 #endif
914 break;
915
916 case ']': /* end of scanset */
917 return (fmt);
918
919 default: /* just another character */
920 c = n;
921 break;
922 }
923 }
924 /* NOTREACHED */
925 }
926
927 #ifndef NO_FLOATING_POINT
928 static int
parsefloat(FILE * fp,char * buf,char * end,locale_t locale)929 parsefloat(FILE *fp, char *buf, char *end, locale_t locale)
930 {
931 char *commit, *p;
932 int infnanpos = 0, decptpos = 0;
933 enum {
934 S_START, S_GOTSIGN, S_INF, S_NAN, S_DONE, S_MAYBEHEX,
935 S_DIGITS, S_DECPT, S_FRAC, S_EXP, S_EXPDIGITS
936 } state = S_START;
937 unsigned char c;
938 const char *decpt = localeconv_l(locale)->decimal_point;
939 _Bool gotmantdig = 0, ishex = 0;
940
941 /*
942 * We set commit = p whenever the string we have read so far
943 * constitutes a valid representation of a floating point
944 * number by itself. At some point, the parse will complete
945 * or fail, and we will ungetc() back to the last commit point.
946 * To ensure that the file offset gets updated properly, it is
947 * always necessary to read at least one character that doesn't
948 * match; thus, we can't short-circuit "infinity" or "nan(...)".
949 */
950 commit = buf - 1;
951 for (p = buf; p < end; ) {
952 c = *fp->pub._p;
953 reswitch:
954 switch (state) {
955 case S_START:
956 state = S_GOTSIGN;
957 if (c == '-' || c == '+')
958 break;
959 else
960 goto reswitch;
961 case S_GOTSIGN:
962 switch (c) {
963 case '0':
964 state = S_MAYBEHEX;
965 commit = p;
966 break;
967 case 'I':
968 case 'i':
969 state = S_INF;
970 break;
971 case 'N':
972 case 'n':
973 state = S_NAN;
974 break;
975 default:
976 state = S_DIGITS;
977 goto reswitch;
978 }
979 break;
980 case S_INF:
981 if (infnanpos > 6 ||
982 (c != "nfinity"[infnanpos] &&
983 c != "NFINITY"[infnanpos]))
984 goto parsedone;
985 if (infnanpos == 1 || infnanpos == 6)
986 commit = p; /* inf or infinity */
987 infnanpos++;
988 break;
989 case S_NAN:
990 switch (infnanpos) {
991 case 0:
992 if (c != 'A' && c != 'a')
993 goto parsedone;
994 break;
995 case 1:
996 if (c != 'N' && c != 'n')
997 goto parsedone;
998 else
999 commit = p;
1000 break;
1001 case 2:
1002 if (c != '(')
1003 goto parsedone;
1004 break;
1005 default:
1006 if (c == ')') {
1007 commit = p;
1008 state = S_DONE;
1009 } else if (!isalnum(c) && c != '_')
1010 goto parsedone;
1011 break;
1012 }
1013 infnanpos++;
1014 break;
1015 case S_DONE:
1016 goto parsedone;
1017 case S_MAYBEHEX:
1018 state = S_DIGITS;
1019 if (c == 'X' || c == 'x') {
1020 ishex = 1;
1021 break;
1022 } else { /* we saw a '0', but no 'x' */
1023 gotmantdig = 1;
1024 goto reswitch;
1025 }
1026 case S_DIGITS:
1027 if ((ishex && isxdigit(c)) || isdigit(c)) {
1028 gotmantdig = 1;
1029 commit = p;
1030 break;
1031 } else {
1032 state = S_DECPT;
1033 goto reswitch;
1034 }
1035 case S_DECPT:
1036 if (c == decpt[decptpos]) {
1037 if (decpt[++decptpos] == '\0') {
1038 /* We read the complete decpt seq. */
1039 state = S_FRAC;
1040 if (gotmantdig)
1041 commit = p;
1042 }
1043 break;
1044 } else if (!decptpos) {
1045 /* We didn't read any decpt characters. */
1046 state = S_FRAC;
1047 goto reswitch;
1048 } else {
1049 /*
1050 * We read part of a multibyte decimal point,
1051 * but the rest is invalid, so bail.
1052 */
1053 goto parsedone;
1054 }
1055 case S_FRAC:
1056 if (((c == 'E' || c == 'e') && !ishex) ||
1057 ((c == 'P' || c == 'p') && ishex)) {
1058 if (!gotmantdig)
1059 goto parsedone;
1060 else
1061 state = S_EXP;
1062 } else if ((ishex && isxdigit(c)) || isdigit(c)) {
1063 commit = p;
1064 gotmantdig = 1;
1065 } else
1066 goto parsedone;
1067 break;
1068 case S_EXP:
1069 state = S_EXPDIGITS;
1070 if (c == '-' || c == '+')
1071 break;
1072 else
1073 goto reswitch;
1074 case S_EXPDIGITS:
1075 if (isdigit(c))
1076 commit = p;
1077 else
1078 goto parsedone;
1079 break;
1080 default:
1081 abort();
1082 }
1083 *p++ = c;
1084 if (--fp->pub._r > 0)
1085 fp->pub._p++;
1086 else if (__srefill(fp))
1087 break; /* EOF */
1088 }
1089
1090 parsedone:
1091 while (commit < --p)
1092 __ungetc(*(u_char *)p, fp);
1093 *++commit = '\0';
1094 return (commit - buf);
1095 }
1096 #endif
1097