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 * @(#)vfscanf.c 8.1 (Berkeley) 6/4/93 33 * $FreeBSD: src/lib/libc/stdio/vfscanf.c,v 1.43 2009/01/19 06:19:51 das Exp $ 34 * $DragonFly: src/lib/libc/stdio/vfscanf.c,v 1.11 2006/02/18 17:55:52 joerg Exp $ 35 */ 36 37 #include "namespace.h" 38 #include <ctype.h> 39 #include <inttypes.h> 40 #include <stddef.h> 41 #include <stdio.h> 42 #include <stdlib.h> 43 #include <stdarg.h> 44 #include <string.h> 45 #include <wchar.h> 46 #include <wctype.h> 47 #include "un-namespace.h" 48 49 #include "collate.h" 50 #include "libc_private.h" 51 #include "local.h" 52 #include "priv_stdio.h" 53 54 #ifndef NO_FLOATING_POINT 55 #include <locale.h> 56 #endif 57 58 #define BUF 513 /* Maximum length of numeric string. */ 59 60 /* 61 * Flags used during conversion. 62 */ 63 #define LONG 0x01 /* l: long or double */ 64 #define LONGDBL 0x02 /* L: long double */ 65 #define SHORT 0x04 /* h: short */ 66 #define SUPPRESS 0x08 /* *: suppress assignment */ 67 #define POINTER 0x10 /* p: void * (as hex) */ 68 #define NOSKIP 0x20 /* [ or c: do not skip blanks */ 69 #define LONGLONG 0x400 /* ll: long long (+ deprecated q: quad) */ 70 #define INTMAXT 0x800 /* j: intmax_t */ 71 #define PTRDIFFT 0x1000 /* t: ptrdiff_t */ 72 #define SIZET 0x2000 /* z: size_t */ 73 #define SHORTSHORT 0x4000 /* hh: char */ 74 #define UNSIGNED 0x8000 /* %[oupxX] conversions */ 75 76 /* 77 * The following are used in integral conversions only: 78 * SIGNOK, NDIGITS, PFXOK, and NZDIGITS 79 */ 80 #define SIGNOK 0x40 /* +/- is (still) legal */ 81 #define NDIGITS 0x80 /* no digits detected */ 82 #define PFXOK 0x100 /* 0x prefix is (still) legal */ 83 #define NZDIGITS 0x200 /* no zero digits detected */ 84 #define HAVESIGN 0x10000 /* sign detected */ 85 86 /* 87 * Conversion types. 88 */ 89 #define CT_CHAR 0 /* %c conversion */ 90 #define CT_CCL 1 /* %[...] conversion */ 91 #define CT_STRING 2 /* %s conversion */ 92 #define CT_INT 3 /* %[dioupxX] conversion */ 93 #define CT_FLOAT 4 /* %[efgEFG] conversion */ 94 95 static const u_char *__sccl(char *, const u_char *); 96 #ifndef NO_FLOATING_POINT 97 static int parsefloat(FILE *, char *, char *); 98 #endif 99 100 __weak_reference(__vfscanf, vfscanf); 101 102 /* 103 * __vfscanf - MT-safe version 104 */ 105 int 106 __vfscanf(FILE *fp, const char *fmt0, va_list ap) 107 { 108 int ret; 109 110 FLOCKFILE(fp); 111 ret = __svfscanf(fp, fmt0, ap); 112 FUNLOCKFILE(fp); 113 return (ret); 114 } 115 116 /* 117 * __svfscanf - non-MT-safe version of __vfscanf 118 */ 119 int 120 __svfscanf(FILE *fp, const char *fmt0, va_list ap) 121 { 122 const u_char *fmt = (const u_char *)fmt0; 123 int c; /* character from format, or conversion */ 124 size_t width; /* field width, or 0 */ 125 char *p; /* points into all kinds of strings */ 126 int n; /* handy integer */ 127 int flags; /* flags as defined above */ 128 char *p0; /* saves original value of p when necessary */ 129 int nassigned; /* number of fields assigned */ 130 int nconversions; /* number of conversions */ 131 int nread; /* number of characters consumed from fp */ 132 int base; /* base argument to conversion function */ 133 char ccltab[256]; /* character class table for %[...] */ 134 char buf[BUF]; /* buffer for numeric and mb conversions */ 135 wchar_t *wcp; /* handy wide character pointer */ 136 size_t nconv; /* length of multibyte sequence converted */ 137 static const mbstate_t initial; 138 mbstate_t mbs; 139 140 /* `basefix' is used to avoid `if' tests in the integer scanner */ 141 static short basefix[17] = 142 { 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 }; 143 144 _SET_ORIENTATION(fp, -1); 145 146 nassigned = 0; 147 nconversions = 0; 148 nread = 0; 149 for (;;) { 150 c = *fmt++; 151 if (c == 0) 152 return (nassigned); 153 if (isspace(c)) { 154 while ((fp->pub._r > 0 || __srefill(fp) == 0) && 155 isspace(*fp->pub._p)) 156 nread++, fp->pub._r--, fp->pub._p++; 157 continue; 158 } 159 if (c != '%') 160 goto literal; 161 width = 0; 162 flags = 0; 163 /* 164 * switch on the format. continue if done; 165 * break once format type is derived. 166 */ 167 again: c = *fmt++; 168 switch (c) { 169 case '%': 170 literal: 171 if (fp->pub._r <= 0 && __srefill(fp)) 172 goto input_failure; 173 if (*fp->pub._p != c) 174 goto match_failure; 175 fp->pub._r--, fp->pub._p++; 176 nread++; 177 continue; 178 179 case '*': 180 flags |= SUPPRESS; 181 goto again; 182 case 'j': 183 flags |= INTMAXT; 184 goto again; 185 case 'l': 186 if (flags & LONG) { 187 flags &= ~LONG; 188 flags |= LONGLONG; 189 } else { 190 flags |= LONG; 191 } 192 goto again; 193 case 'q': 194 flags |= LONGLONG; /* assume long long == quad */ 195 goto again; 196 case 't': 197 flags |= PTRDIFFT; 198 goto again; 199 case 'z': 200 flags |= SIZET; 201 goto again; 202 case 'L': 203 flags |= LONGDBL; 204 goto again; 205 case 'h': 206 if (flags & SHORT) { 207 flags &= ~SHORT; 208 flags |= SHORTSHORT; 209 } else { 210 flags |= SHORT; 211 } 212 goto again; 213 214 case '0': case '1': case '2': case '3': case '4': 215 case '5': case '6': case '7': case '8': case '9': 216 width = width * 10 + c - '0'; 217 goto again; 218 219 /* 220 * Conversions. 221 */ 222 case 'd': 223 c = CT_INT; 224 base = 10; 225 break; 226 227 case 'i': 228 c = CT_INT; 229 base = 0; 230 break; 231 232 case 'o': 233 c = CT_INT; 234 flags |= UNSIGNED; 235 base = 8; 236 break; 237 238 case 'u': 239 c = CT_INT; 240 flags |= UNSIGNED; 241 base = 10; 242 break; 243 244 case 'X': 245 case 'x': 246 flags |= PFXOK; /* enable 0x prefixing */ 247 c = CT_INT; 248 flags |= UNSIGNED; 249 base = 16; 250 break; 251 252 #ifndef NO_FLOATING_POINT 253 case 'A': case 'E': case 'F': case 'G': 254 case 'a': case 'e': case 'f': case 'g': 255 c = CT_FLOAT; 256 break; 257 #endif 258 259 case 'S': 260 flags |= LONG; 261 /* FALLTHROUGH */ 262 case 's': 263 c = CT_STRING; 264 break; 265 266 case '[': 267 fmt = __sccl(ccltab, fmt); 268 flags |= NOSKIP; 269 c = CT_CCL; 270 break; 271 272 case 'C': 273 flags |= LONG; 274 /* FALLTHROUGH */ 275 case 'c': 276 flags |= NOSKIP; 277 c = CT_CHAR; 278 break; 279 280 case 'p': /* pointer format is like hex */ 281 flags |= POINTER | PFXOK; 282 c = CT_INT; /* assumes sizeof(uintmax_t) */ 283 flags |= UNSIGNED; /* >= sizeof(uintptr_t) */ 284 base = 16; 285 break; 286 287 case 'n': 288 nconversions++; 289 if (flags & SUPPRESS) /* ??? */ 290 continue; 291 if (flags & SHORTSHORT) 292 *va_arg(ap, char *) = nread; 293 else if (flags & SHORT) 294 *va_arg(ap, short *) = nread; 295 else if (flags & LONG) 296 *va_arg(ap, long *) = nread; 297 else if (flags & LONGLONG) 298 *va_arg(ap, long long *) = nread; 299 else if (flags & INTMAXT) 300 *va_arg(ap, intmax_t *) = nread; 301 else if (flags & SIZET) 302 *va_arg(ap, size_t *) = nread; 303 else if (flags & PTRDIFFT) 304 *va_arg(ap, ptrdiff_t *) = nread; 305 else 306 *va_arg(ap, int *) = nread; 307 continue; 308 309 default: 310 goto match_failure; 311 312 /* 313 * Disgusting backwards compatibility hack. XXX 314 */ 315 case '\0': /* compat */ 316 return (EOF); 317 } 318 319 /* 320 * We have a conversion that requires input. 321 */ 322 if (fp->pub._r <= 0 && __srefill(fp)) 323 goto input_failure; 324 325 /* 326 * Consume leading white space, except for formats 327 * that suppress this. 328 */ 329 if ((flags & NOSKIP) == 0) { 330 while (isspace(*fp->pub._p)) { 331 nread++; 332 if (--fp->pub._r > 0) 333 fp->pub._p++; 334 else if (__srefill(fp)) 335 goto input_failure; 336 } 337 /* 338 * Note that there is at least one character in 339 * the buffer, so conversions that do not set NOSKIP 340 * ca no longer result in an input failure. 341 */ 342 } 343 344 /* 345 * Do the conversion. 346 */ 347 switch (c) { 348 349 case CT_CHAR: 350 /* scan arbitrary characters (sets NOSKIP) */ 351 if (width == 0) 352 width = 1; 353 if (flags & LONG) { 354 if ((flags & SUPPRESS) == 0) 355 wcp = va_arg(ap, wchar_t *); 356 else 357 wcp = NULL; 358 n = 0; 359 while (width != 0) { 360 if (n == MB_CUR_MAX) { 361 fp->pub._flags |= __SERR; 362 goto input_failure; 363 } 364 buf[n++] = *fp->pub._p; 365 fp->pub._p++; 366 fp->pub._r--; 367 mbs = initial; 368 nconv = mbrtowc(wcp, buf, n, &mbs); 369 if (nconv == (size_t)-1) { 370 fp->pub._flags |= __SERR; 371 goto input_failure; 372 } 373 if (nconv == 0 && !(flags & SUPPRESS)) 374 *wcp = L'\0'; 375 if (nconv != (size_t)-2) { 376 nread += n; 377 width--; 378 if (!(flags & SUPPRESS)) 379 wcp++; 380 n = 0; 381 } 382 if (fp->pub._r <= 0 && __srefill(fp)) { 383 if (n != 0) { 384 fp->pub._flags |= __SERR; 385 goto input_failure; 386 } 387 break; 388 } 389 } 390 if ((flags & SUPPRESS) == 0) 391 nassigned++; 392 } else if (flags & SUPPRESS) { 393 size_t sum = 0; 394 for (;;) { 395 if ((n = fp->pub._r) < width) { 396 sum += n; 397 width -= n; 398 fp->pub._p += n; 399 if (__srefill(fp)) { 400 if (sum == 0) 401 goto input_failure; 402 break; 403 } 404 } else { 405 sum += width; 406 fp->pub._r -= width; 407 fp->pub._p += width; 408 break; 409 } 410 } 411 nread += sum; 412 } else { 413 size_t r = __fread((void *)va_arg(ap, char *), 414 1, width, fp); 415 416 if (r == 0) 417 goto input_failure; 418 nread += r; 419 nassigned++; 420 } 421 nconversions++; 422 break; 423 424 case CT_CCL: 425 /* scan a (nonempty) character class (sets NOSKIP) */ 426 if (width == 0) 427 width = (size_t)~0; /* `infinity' */ 428 /* take only those things in the class */ 429 if (flags & LONG) { 430 wchar_t twc; 431 int nchars; 432 433 if ((flags & SUPPRESS) == 0) 434 wcp = va_arg(ap, wchar_t *); 435 else 436 wcp = &twc; 437 n = 0; 438 nchars = 0; 439 while (width != 0) { 440 if (n == MB_CUR_MAX) { 441 fp->pub._flags |= __SERR; 442 goto input_failure; 443 } 444 buf[n++] = *fp->pub._p; 445 fp->pub._p++; 446 fp->pub._r--; 447 mbs = initial; 448 nconv = mbrtowc(wcp, buf, n, &mbs); 449 if (nconv == (size_t)-1) { 450 fp->pub._flags |= __SERR; 451 goto input_failure; 452 } 453 if (nconv == 0) 454 *wcp = L'\0'; 455 if (nconv != (size_t)-2) { 456 if (wctob(*wcp) != EOF && 457 !ccltab[wctob(*wcp)]) { 458 while (n != 0) { 459 n--; 460 __ungetc(buf[n], 461 fp); 462 } 463 break; 464 } 465 nread += n; 466 width--; 467 if (!(flags & SUPPRESS)) 468 wcp++; 469 nchars++; 470 n = 0; 471 } 472 if (fp->pub._r <= 0 && __srefill(fp)) { 473 if (n != 0) { 474 fp->pub._flags |= __SERR; 475 goto input_failure; 476 } 477 break; 478 } 479 } 480 if (n != 0) { 481 fp->pub._flags |= __SERR; 482 goto input_failure; 483 } 484 n = nchars; 485 if (n == 0) 486 goto match_failure; 487 if (!(flags & SUPPRESS)) { 488 *wcp = L'\0'; 489 nassigned++; 490 } 491 } else if (flags & SUPPRESS) { 492 n = 0; 493 while (ccltab[*fp->pub._p]) { 494 n++, fp->pub._r--, fp->pub._p++; 495 if (--width == 0) 496 break; 497 if (fp->pub._r <= 0 && __srefill(fp)) { 498 if (n == 0) 499 goto input_failure; 500 break; 501 } 502 } 503 if (n == 0) 504 goto match_failure; 505 } else { 506 p0 = p = va_arg(ap, char *); 507 while (ccltab[*fp->pub._p]) { 508 fp->pub._r--; 509 *p++ = *fp->pub._p++; 510 if (--width == 0) 511 break; 512 if (fp->pub._r <= 0 && __srefill(fp)) { 513 if (p == p0) 514 goto input_failure; 515 break; 516 } 517 } 518 n = p - p0; 519 if (n == 0) 520 goto match_failure; 521 *p = 0; 522 nassigned++; 523 } 524 nread += n; 525 nconversions++; 526 break; 527 528 case CT_STRING: 529 /* like CCL, but zero-length string OK, & no NOSKIP */ 530 if (width == 0) 531 width = (size_t)~0; 532 if (flags & LONG) { 533 wchar_t twc; 534 535 if ((flags & SUPPRESS) == 0) 536 wcp = va_arg(ap, wchar_t *); 537 else 538 wcp = &twc; 539 n = 0; 540 while (!isspace(*fp->pub._p) && width != 0) { 541 if (n == MB_CUR_MAX) { 542 fp->pub._flags |= __SERR; 543 goto input_failure; 544 } 545 buf[n++] = *fp->pub._p; 546 fp->pub._p++; 547 fp->pub._r--; 548 mbs = initial; 549 nconv = mbrtowc(wcp, buf, n, &mbs); 550 if (nconv == (size_t)-1) { 551 fp->pub._flags |= __SERR; 552 goto input_failure; 553 } 554 if (nconv == 0) 555 *wcp = L'\0'; 556 if (nconv != (size_t)-2) { 557 if (iswspace(*wcp)) { 558 while (n != 0) { 559 n--; 560 __ungetc(buf[n], 561 fp); 562 } 563 break; 564 } 565 nread += n; 566 width--; 567 if (!(flags & SUPPRESS)) 568 wcp++; 569 n = 0; 570 } 571 if (fp->pub._r <= 0 && __srefill(fp)) { 572 if (n != 0) { 573 fp->pub._flags |= __SERR; 574 goto input_failure; 575 } 576 break; 577 } 578 } 579 if (!(flags & SUPPRESS)) { 580 *wcp = L'\0'; 581 nassigned++; 582 } 583 } else if (flags & SUPPRESS) { 584 n = 0; 585 while (!isspace(*fp->pub._p)) { 586 n++, fp->pub._r--, fp->pub._p++; 587 if (--width == 0) 588 break; 589 if (fp->pub._r <= 0 && __srefill(fp)) 590 break; 591 } 592 nread += n; 593 } else { 594 p0 = p = va_arg(ap, char *); 595 while (!isspace(*fp->pub._p)) { 596 fp->pub._r--; 597 *p++ = *fp->pub._p++; 598 if (--width == 0) 599 break; 600 if (fp->pub._r <= 0 && __srefill(fp)) 601 break; 602 } 603 *p = 0; 604 nread += p - p0; 605 nassigned++; 606 } 607 nconversions++; 608 continue; 609 610 case CT_INT: 611 /* scan an integer as if by the conversion function */ 612 #ifdef hardway 613 if (width == 0 || width > sizeof(buf) - 1) 614 width = sizeof(buf) - 1; 615 #else 616 /* size_t is unsigned, hence this optimisation */ 617 if (--width > sizeof(buf) - 2) 618 width = sizeof(buf) - 2; 619 width++; 620 #endif 621 flags |= SIGNOK | NDIGITS | NZDIGITS; 622 for (p = buf; width; width--) { 623 c = *fp->pub._p; 624 /* 625 * Switch on the character; `goto ok' 626 * if we accept it as a part of number. 627 */ 628 switch (c) { 629 630 /* 631 * The digit 0 is always legal, but is 632 * special. For %i conversions, if no 633 * digits (zero or nonzero) have been 634 * scanned (only signs), we will have 635 * base==0. In that case, we should set 636 * it to 8 and enable 0x prefixing. 637 * Also, if we have not scanned zero digits 638 * before this, do not turn off prefixing 639 * (someone else will turn it off if we 640 * have scanned any nonzero digits). 641 */ 642 case '0': 643 if (base == 0) { 644 base = 8; 645 flags |= PFXOK; 646 } 647 if (flags & NZDIGITS) 648 flags &= ~(SIGNOK|NZDIGITS|NDIGITS); 649 else 650 flags &= ~(SIGNOK|PFXOK|NDIGITS); 651 goto ok; 652 653 /* 1 through 7 always legal */ 654 case '1': case '2': case '3': 655 case '4': case '5': case '6': case '7': 656 base = basefix[base]; 657 flags &= ~(SIGNOK | PFXOK | NDIGITS); 658 goto ok; 659 660 /* digits 8 and 9 ok iff decimal or hex */ 661 case '8': case '9': 662 base = basefix[base]; 663 if (base <= 8) 664 break; /* not legal here */ 665 flags &= ~(SIGNOK | PFXOK | NDIGITS); 666 goto ok; 667 668 /* letters ok iff hex */ 669 case 'A': case 'B': case 'C': 670 case 'D': case 'E': case 'F': 671 case 'a': case 'b': case 'c': 672 case 'd': case 'e': case 'f': 673 /* no need to fix base here */ 674 if (base <= 10) 675 break; /* not legal here */ 676 flags &= ~(SIGNOK | PFXOK | NDIGITS); 677 goto ok; 678 679 /* sign ok only as first character */ 680 case '+': case '-': 681 if (flags & SIGNOK) { 682 flags &= ~SIGNOK; 683 flags |= HAVESIGN; 684 goto ok; 685 } 686 break; 687 688 /* 689 * x ok iff flag still set & 2nd char (or 690 * 3rd char if we have a sign). 691 */ 692 case 'x': case 'X': 693 if (flags & PFXOK && p == 694 buf + 1 + !!(flags & HAVESIGN)) { 695 base = 16; /* if %i */ 696 flags &= ~PFXOK; 697 goto ok; 698 } 699 break; 700 } 701 702 /* 703 * If we got here, c is not a legal character 704 * for a number. Stop accumulating digits. 705 */ 706 break; 707 ok: 708 /* 709 * c is legal: store it and look at the next. 710 */ 711 *p++ = c; 712 if (--fp->pub._r > 0) 713 fp->pub._p++; 714 else if (__srefill(fp)) 715 break; /* EOF */ 716 } 717 /* 718 * If we had only a sign, it is no good; push 719 * back the sign. If the number ends in `x', 720 * it was [sign] '0' 'x', so push back the x 721 * and treat it as [sign] '0'. 722 */ 723 if (flags & NDIGITS) { 724 if (p > buf) 725 __ungetc(*(u_char *)--p, fp); 726 goto match_failure; 727 } 728 c = ((u_char *)p)[-1]; 729 if (c == 'x' || c == 'X') { 730 --p; 731 __ungetc(c, fp); 732 } 733 if ((flags & SUPPRESS) == 0) { 734 uintmax_t res; 735 736 *p = 0; 737 if ((flags & UNSIGNED) == 0) 738 res = strtoimax(buf, NULL, base); 739 else 740 res = strtoumax(buf, NULL, base); 741 if (flags & POINTER) 742 *va_arg(ap, void **) = 743 (void *)(uintptr_t)res; 744 else if (flags & SHORTSHORT) 745 *va_arg(ap, char *) = res; 746 else if (flags & SHORT) 747 *va_arg(ap, short *) = res; 748 else if (flags & LONG) 749 *va_arg(ap, long *) = res; 750 else if (flags & LONGLONG) 751 *va_arg(ap, long long *) = res; 752 else if (flags & INTMAXT) 753 *va_arg(ap, intmax_t *) = res; 754 else if (flags & PTRDIFFT) 755 *va_arg(ap, ptrdiff_t *) = res; 756 else if (flags & SIZET) 757 *va_arg(ap, size_t *) = res; 758 else 759 *va_arg(ap, int *) = res; 760 nassigned++; 761 } 762 nread += p - buf; 763 nconversions++; 764 break; 765 766 #ifndef NO_FLOATING_POINT 767 case CT_FLOAT: 768 /* scan a floating point number as if by strtod */ 769 if (width == 0 || width > sizeof(buf) - 1) 770 width = sizeof(buf) - 1; 771 if ((width = parsefloat(fp, buf, buf + width)) == 0) 772 goto match_failure; 773 if ((flags & SUPPRESS) == 0) { 774 if (flags & LONGDBL) { 775 long double res = strtold(buf, &p); 776 *va_arg(ap, long double *) = res; 777 } else if (flags & LONG) { 778 double res = strtod(buf, &p); 779 *va_arg(ap, double *) = res; 780 } else { 781 float res = strtof(buf, &p); 782 *va_arg(ap, float *) = res; 783 } 784 nassigned++; 785 } 786 nread += width; 787 nconversions++; 788 break; 789 #endif /* !NO_FLOATING_POINT */ 790 } 791 } 792 input_failure: 793 return (nconversions != 0 ? nassigned : EOF); 794 match_failure: 795 return (nassigned); 796 } 797 798 /* 799 * Fill in the given table from the scanset at the given format 800 * (just after `['). Return a pointer to the character past the 801 * closing `]'. The table has a 1 wherever characters should be 802 * considered part of the scanset. 803 */ 804 static const u_char * 805 __sccl(char *tab, const u_char *fmt) 806 { 807 int c, n, v, i; 808 809 /* first `clear' the whole table */ 810 c = *fmt++; /* first char hat => negated scanset */ 811 if (c == '^') { 812 v = 1; /* default => accept */ 813 c = *fmt++; /* get new first char */ 814 } else 815 v = 0; /* default => reject */ 816 817 /* XXX: Will not work if sizeof(tab*) > sizeof(char) */ 818 memset(tab, v, 256); 819 820 if (c == 0) 821 return (fmt - 1);/* format ended before closing ] */ 822 823 /* 824 * Now set the entries corresponding to the actual scanset 825 * to the opposite of the above. 826 * 827 * The first character may be ']' (or '-') without being special; 828 * the last character may be '-'. 829 */ 830 v = 1 - v; 831 for (;;) { 832 tab[c] = v; /* take character c */ 833 doswitch: 834 n = *fmt++; /* and examine the next */ 835 switch (n) { 836 837 case 0: /* format ended too soon */ 838 return (fmt - 1); 839 840 case '-': 841 /* 842 * A scanset of the form 843 * [01+-] 844 * is defined as `the digit 0, the digit 1, 845 * the character +, the character -', but 846 * the effect of a scanset such as 847 * [a-zA-Z0-9] 848 * is implementation defined. The V7 Unix 849 * scanf treats `a-z' as `the letters a through 850 * z', but treats `a-a' as `the letter a, the 851 * character -, and the letter a'. 852 * 853 * For compatibility, the `-' is not considerd 854 * to define a range if the character following 855 * it is either a close bracket (required by ANSI) 856 * or is not numerically greater than the character 857 * we just stored in the table (c). 858 */ 859 n = *fmt; 860 if (n == ']' 861 || (__collate_load_error ? n < c : 862 __collate_range_cmp (n, c) < 0 863 ) 864 ) { 865 c = '-'; 866 break; /* resume the for(;;) */ 867 } 868 fmt++; 869 /* fill in the range */ 870 if (__collate_load_error) { 871 do { 872 tab[++c] = v; 873 } while (c < n); 874 } else { 875 for (i = 0; i < 256; i ++) 876 if ( __collate_range_cmp (c, i) < 0 877 && __collate_range_cmp (i, n) <= 0 878 ) 879 tab[i] = v; 880 } 881 #if 1 /* XXX another disgusting compatibility hack */ 882 c = n; 883 /* 884 * Alas, the V7 Unix scanf also treats formats 885 * such as [a-c-e] as `the letters a through e'. 886 * This too is permitted by the standard.... 887 */ 888 goto doswitch; 889 #else 890 c = *fmt++; 891 if (c == 0) 892 return (fmt - 1); 893 if (c == ']') 894 return (fmt); 895 #endif 896 break; 897 898 case ']': /* end of scanset */ 899 return (fmt); 900 901 default: /* just another character */ 902 c = n; 903 break; 904 } 905 } 906 /* NOTREACHED */ 907 } 908 909 #ifndef NO_FLOATING_POINT 910 static int 911 parsefloat(FILE *fp, char *buf, char *end) 912 { 913 char *commit, *p; 914 int infnanpos = 0, decptpos = 0; 915 enum { 916 S_START, S_GOTSIGN, S_INF, S_NAN, S_DONE, S_MAYBEHEX, 917 S_DIGITS, S_DECPT, S_FRAC, S_EXP, S_EXPDIGITS 918 } state = S_START; 919 unsigned char c; 920 const char *decpt = localeconv()->decimal_point; 921 _Bool gotmantdig = 0, ishex = 0; 922 923 /* 924 * We set commit = p whenever the string we have read so far 925 * constitutes a valid representation of a floating point 926 * number by itself. At some point, the parse will complete 927 * or fail, and we will ungetc() back to the last commit point. 928 * To ensure that the file offset gets updated properly, it is 929 * always necessary to read at least one character that doesn't 930 * match; thus, we can't short-circuit "infinity" or "nan(...)". 931 */ 932 commit = buf - 1; 933 for (p = buf; p < end; ) { 934 c = *fp->pub._p; 935 reswitch: 936 switch (state) { 937 case S_START: 938 state = S_GOTSIGN; 939 if (c == '-' || c == '+') 940 break; 941 else 942 goto reswitch; 943 case S_GOTSIGN: 944 switch (c) { 945 case '0': 946 state = S_MAYBEHEX; 947 commit = p; 948 break; 949 case 'I': 950 case 'i': 951 state = S_INF; 952 break; 953 case 'N': 954 case 'n': 955 state = S_NAN; 956 break; 957 default: 958 state = S_DIGITS; 959 goto reswitch; 960 } 961 break; 962 case S_INF: 963 if (infnanpos > 6 || 964 (c != "nfinity"[infnanpos] && 965 c != "NFINITY"[infnanpos])) 966 goto parsedone; 967 if (infnanpos == 1 || infnanpos == 6) 968 commit = p; /* inf or infinity */ 969 infnanpos++; 970 break; 971 case S_NAN: 972 switch (infnanpos) { 973 case 0: 974 if (c != 'A' && c != 'a') 975 goto parsedone; 976 break; 977 case 1: 978 if (c != 'N' && c != 'n') 979 goto parsedone; 980 else 981 commit = p; 982 break; 983 case 2: 984 if (c != '(') 985 goto parsedone; 986 break; 987 default: 988 if (c == ')') { 989 commit = p; 990 state = S_DONE; 991 } else if (!isalnum(c) && c != '_') 992 goto parsedone; 993 break; 994 } 995 infnanpos++; 996 break; 997 case S_DONE: 998 goto parsedone; 999 case S_MAYBEHEX: 1000 state = S_DIGITS; 1001 if (c == 'X' || c == 'x') { 1002 ishex = 1; 1003 break; 1004 } else { /* we saw a '0', but no 'x' */ 1005 gotmantdig = 1; 1006 goto reswitch; 1007 } 1008 case S_DIGITS: 1009 if ((ishex && isxdigit(c)) || isdigit(c)) { 1010 gotmantdig = 1; 1011 commit = p; 1012 break; 1013 } else { 1014 state = S_DECPT; 1015 goto reswitch; 1016 } 1017 case S_DECPT: 1018 if (c == decpt[decptpos]) { 1019 if (decpt[++decptpos] == '\0') { 1020 /* We read the complete decpt seq. */ 1021 state = S_FRAC; 1022 if (gotmantdig) 1023 commit = p; 1024 } 1025 break; 1026 } else if (!decptpos) { 1027 /* We didn't read any decpt characters. */ 1028 state = S_FRAC; 1029 goto reswitch; 1030 } else { 1031 /* 1032 * We read part of a multibyte decimal point, 1033 * but the rest is invalid, so bail. 1034 */ 1035 goto parsedone; 1036 } 1037 case S_FRAC: 1038 if (((c == 'E' || c == 'e') && !ishex) || 1039 ((c == 'P' || c == 'p') && ishex)) { 1040 if (!gotmantdig) 1041 goto parsedone; 1042 else 1043 state = S_EXP; 1044 } else if ((ishex && isxdigit(c)) || isdigit(c)) { 1045 commit = p; 1046 gotmantdig = 1; 1047 } else 1048 goto parsedone; 1049 break; 1050 case S_EXP: 1051 state = S_EXPDIGITS; 1052 if (c == '-' || c == '+') 1053 break; 1054 else 1055 goto reswitch; 1056 case S_EXPDIGITS: 1057 if (isdigit(c)) 1058 commit = p; 1059 else 1060 goto parsedone; 1061 break; 1062 default: 1063 abort(); 1064 } 1065 *p++ = c; 1066 if (--fp->pub._r > 0) 1067 fp->pub._p++; 1068 else if (__srefill(fp)) 1069 break; /* EOF */ 1070 } 1071 1072 parsedone: 1073 while (commit < --p) 1074 __ungetc(*(u_char *)p, fp); 1075 *++commit = '\0'; 1076 return (commit - buf); 1077 } 1078 #endif 1079