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