archive_read_support_format_warc.c revision 313570
1/*- 2 * Copyright (c) 2014 Sebastian Freundt 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 17 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 18 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 19 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 20 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 21 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 22 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 23 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 24 */ 25 26#include "archive_platform.h" 27__FBSDID("$FreeBSD: stable/11/contrib/libarchive/libarchive/archive_read_support_format_warc.c 313570 2017-02-11 00:54:16Z mm $"); 28 29/** 30 * WARC is standardised by ISO TC46/SC4/WG12 and currently available as 31 * ISO 28500:2009. 32 * For the purposes of this file we used the final draft from: 33 * http://bibnum.bnf.fr/warc/WARC_ISO_28500_version1_latestdraft.pdf 34 * 35 * Todo: 36 * [ ] real-world warcs can contain resources at endpoints ending in / 37 * e.g. http://bibnum.bnf.fr/warc/ 38 * if you're lucky their response contains a Content-Location: header 39 * pointing to a unix-compliant filename, in the example above it's 40 * Content-Location: http://bibnum.bnf.fr/warc/index.html 41 * however, that's not mandated and github for example doesn't follow 42 * this convention. 43 * We need a set of archive options to control what to do with 44 * entries like these, at the moment care is taken to skip them. 45 * 46 **/ 47 48#ifdef HAVE_SYS_STAT_H 49#include <sys/stat.h> 50#endif 51#ifdef HAVE_ERRNO_H 52#include <errno.h> 53#endif 54#ifdef HAVE_STDLIB_H 55#include <stdlib.h> 56#endif 57#ifdef HAVE_STRING_H 58#include <string.h> 59#endif 60#ifdef HAVE_LIMITS_H 61#include <limits.h> 62#endif 63#ifdef HAVE_CTYPE_H 64#include <ctype.h> 65#endif 66#ifdef HAVE_TIME_H 67#include <time.h> 68#endif 69 70#include "archive.h" 71#include "archive_entry.h" 72#include "archive_private.h" 73#include "archive_read_private.h" 74 75typedef enum { 76 WT_NONE, 77 /* warcinfo */ 78 WT_INFO, 79 /* metadata */ 80 WT_META, 81 /* resource */ 82 WT_RSRC, 83 /* request, unsupported */ 84 WT_REQ, 85 /* response, unsupported */ 86 WT_RSP, 87 /* revisit, unsupported */ 88 WT_RVIS, 89 /* conversion, unsupported */ 90 WT_CONV, 91 /* continuation, unsupported at the moment */ 92 WT_CONT, 93 /* invalid type */ 94 LAST_WT 95} warc_type_t; 96 97typedef struct { 98 size_t len; 99 const char *str; 100} warc_string_t; 101 102typedef struct { 103 size_t len; 104 char *str; 105} warc_strbuf_t; 106 107struct warc_s { 108 /* content length ahead */ 109 size_t cntlen; 110 /* and how much we've processed so far */ 111 size_t cntoff; 112 /* and how much we need to consume between calls */ 113 size_t unconsumed; 114 115 /* string pool */ 116 warc_strbuf_t pool; 117 /* previous version */ 118 unsigned int pver; 119 /* stringified format name */ 120 struct archive_string sver; 121}; 122 123static int _warc_bid(struct archive_read *a, int); 124static int _warc_cleanup(struct archive_read *a); 125static int _warc_read(struct archive_read*, const void**, size_t*, int64_t*); 126static int _warc_skip(struct archive_read *a); 127static int _warc_rdhdr(struct archive_read *a, struct archive_entry *e); 128 129/* private routines */ 130static unsigned int _warc_rdver(const char buf[10], size_t bsz); 131static unsigned int _warc_rdtyp(const char *buf, size_t bsz); 132static warc_string_t _warc_rduri(const char *buf, size_t bsz); 133static ssize_t _warc_rdlen(const char *buf, size_t bsz); 134static time_t _warc_rdrtm(const char *buf, size_t bsz); 135static time_t _warc_rdmtm(const char *buf, size_t bsz); 136static const char *_warc_find_eoh(const char *buf, size_t bsz); 137 138 139int 140archive_read_support_format_warc(struct archive *_a) 141{ 142 struct archive_read *a = (struct archive_read *)_a; 143 struct warc_s *w; 144 int r; 145 146 archive_check_magic(_a, ARCHIVE_READ_MAGIC, 147 ARCHIVE_STATE_NEW, "archive_read_support_format_warc"); 148 149 if ((w = calloc(1, sizeof(*w))) == NULL) { 150 archive_set_error(&a->archive, ENOMEM, 151 "Can't allocate warc data"); 152 return (ARCHIVE_FATAL); 153 } 154 155 r = __archive_read_register_format( 156 a, w, "warc", 157 _warc_bid, NULL, _warc_rdhdr, _warc_read, 158 _warc_skip, NULL, _warc_cleanup, NULL, NULL); 159 160 if (r != ARCHIVE_OK) { 161 free(w); 162 return (r); 163 } 164 return (ARCHIVE_OK); 165} 166 167static int 168_warc_cleanup(struct archive_read *a) 169{ 170 struct warc_s *w = a->format->data; 171 172 if (w->pool.len > 0U) { 173 free(w->pool.str); 174 } 175 archive_string_free(&w->sver); 176 free(w); 177 a->format->data = NULL; 178 return (ARCHIVE_OK); 179} 180 181static int 182_warc_bid(struct archive_read *a, int best_bid) 183{ 184 const char *hdr; 185 ssize_t nrd; 186 unsigned int ver; 187 188 (void)best_bid; /* UNUSED */ 189 190 /* check first line of file, it should be a record already */ 191 if ((hdr = __archive_read_ahead(a, 12U, &nrd)) == NULL) { 192 /* no idea what to do */ 193 return -1; 194 } else if (nrd < 12) { 195 /* nah, not for us, our magic cookie is at least 12 bytes */ 196 return -1; 197 } 198 199 /* otherwise snarf the record's version number */ 200 ver = _warc_rdver(hdr, nrd); 201 if (ver == 0U || ver > 10000U) { 202 /* oh oh oh, best not to wager ... */ 203 return -1; 204 } 205 206 /* otherwise be confident */ 207 return (64); 208} 209 210static int 211_warc_rdhdr(struct archive_read *a, struct archive_entry *entry) 212{ 213#define HDR_PROBE_LEN (12U) 214 struct warc_s *w = a->format->data; 215 unsigned int ver; 216 const char *buf; 217 ssize_t nrd; 218 const char *eoh; 219 /* for the file name, saves some strndup()'ing */ 220 warc_string_t fnam; 221 /* warc record type, not that we really use it a lot */ 222 warc_type_t ftyp; 223 /* content-length+error monad */ 224 ssize_t cntlen; 225 /* record time is the WARC-Date time we reinterpret it as ctime */ 226 time_t rtime; 227 /* mtime is the Last-Modified time which will be the entry's mtime */ 228 time_t mtime; 229 230start_over: 231 /* just use read_ahead() they keep track of unconsumed 232 * bits and bobs for us; no need to put an extra shift in 233 * and reproduce that functionality here */ 234 buf = __archive_read_ahead(a, HDR_PROBE_LEN, &nrd); 235 236 if (nrd < 0) { 237 /* no good */ 238 archive_set_error( 239 &a->archive, ARCHIVE_ERRNO_MISC, 240 "Bad record header"); 241 return (ARCHIVE_FATAL); 242 } else if (buf == NULL) { 243 /* there should be room for at least WARC/bla\r\n 244 * must be EOF therefore */ 245 return (ARCHIVE_EOF); 246 } 247 /* looks good so far, try and find the end of the header now */ 248 eoh = _warc_find_eoh(buf, nrd); 249 if (eoh == NULL) { 250 /* still no good, the header end might be beyond the 251 * probe we've requested, but then again who'd cram 252 * so much stuff into the header *and* be 28500-compliant */ 253 archive_set_error( 254 &a->archive, ARCHIVE_ERRNO_MISC, 255 "Bad record header"); 256 return (ARCHIVE_FATAL); 257 } else if ((ver = _warc_rdver(buf, eoh - buf)) > 10000U) { 258 /* nawww, I wish they promised backward compatibility 259 * anyhoo, in their infinite wisdom the 28500 guys might 260 * come up with something we can't possibly handle so 261 * best end things here */ 262 archive_set_error( 263 &a->archive, ARCHIVE_ERRNO_MISC, 264 "Unsupported record version"); 265 return (ARCHIVE_FATAL); 266 } else if ((cntlen = _warc_rdlen(buf, eoh - buf)) < 0) { 267 /* nightmare! the specs say content-length is mandatory 268 * so I don't feel overly bad stopping the reader here */ 269 archive_set_error( 270 &a->archive, EINVAL, 271 "Bad content length"); 272 return (ARCHIVE_FATAL); 273 } else if ((rtime = _warc_rdrtm(buf, eoh - buf)) == (time_t)-1) { 274 /* record time is mandatory as per WARC/1.0, 275 * so just barf here, fast and loud */ 276 archive_set_error( 277 &a->archive, EINVAL, 278 "Bad record time"); 279 return (ARCHIVE_FATAL); 280 } 281 282 /* let the world know we're a WARC archive */ 283 a->archive.archive_format = ARCHIVE_FORMAT_WARC; 284 if (ver != w->pver) { 285 /* stringify this entry's version */ 286 archive_string_sprintf(&w->sver, 287 "WARC/%u.%u", ver / 10000, ver % 10000); 288 /* remember the version */ 289 w->pver = ver; 290 } 291 /* start off with the type */ 292 ftyp = _warc_rdtyp(buf, eoh - buf); 293 /* and let future calls know about the content */ 294 w->cntlen = cntlen; 295 w->cntoff = 0U; 296 mtime = 0;/* Avoid compiling error on some platform. */ 297 298 switch (ftyp) { 299 case WT_RSRC: 300 case WT_RSP: 301 /* only try and read the filename in the cases that are 302 * guaranteed to have one */ 303 fnam = _warc_rduri(buf, eoh - buf); 304 /* check the last character in the URI to avoid creating 305 * directory endpoints as files, see Todo above */ 306 if (fnam.len == 0 || fnam.str[fnam.len - 1] == '/') { 307 /* break here for now */ 308 fnam.len = 0U; 309 fnam.str = NULL; 310 break; 311 } 312 /* bang to our string pool, so we save a 313 * malloc()+free() roundtrip */ 314 if (fnam.len + 1U > w->pool.len) { 315 w->pool.len = ((fnam.len + 64U) / 64U) * 64U; 316 w->pool.str = realloc(w->pool.str, w->pool.len); 317 } 318 memcpy(w->pool.str, fnam.str, fnam.len); 319 w->pool.str[fnam.len] = '\0'; 320 /* let no one else know about the pool, it's a secret, shhh */ 321 fnam.str = w->pool.str; 322 323 /* snarf mtime or deduce from rtime 324 * this is a custom header added by our writer, it's quite 325 * hard to believe anyone else would go through with it 326 * (apart from being part of some http responses of course) */ 327 if ((mtime = _warc_rdmtm(buf, eoh - buf)) == (time_t)-1) { 328 mtime = rtime; 329 } 330 break; 331 default: 332 fnam.len = 0U; 333 fnam.str = NULL; 334 break; 335 } 336 337 /* now eat some of those delicious buffer bits */ 338 __archive_read_consume(a, eoh - buf); 339 340 switch (ftyp) { 341 case WT_RSRC: 342 case WT_RSP: 343 if (fnam.len > 0U) { 344 /* populate entry object */ 345 archive_entry_set_filetype(entry, AE_IFREG); 346 archive_entry_copy_pathname(entry, fnam.str); 347 archive_entry_set_size(entry, cntlen); 348 archive_entry_set_perm(entry, 0644); 349 /* rtime is the new ctime, mtime stays mtime */ 350 archive_entry_set_ctime(entry, rtime, 0L); 351 archive_entry_set_mtime(entry, mtime, 0L); 352 break; 353 } 354 /* FALLTHROUGH */ 355 default: 356 /* consume the content and start over */ 357 _warc_skip(a); 358 goto start_over; 359 } 360 return (ARCHIVE_OK); 361} 362 363static int 364_warc_read(struct archive_read *a, const void **buf, size_t *bsz, int64_t *off) 365{ 366 struct warc_s *w = a->format->data; 367 const char *rab; 368 ssize_t nrd; 369 370 if (w->cntoff >= w->cntlen) { 371 eof: 372 /* it's our lucky day, no work, we can leave early */ 373 *buf = NULL; 374 *bsz = 0U; 375 *off = w->cntoff + 4U/*for \r\n\r\n separator*/; 376 w->unconsumed = 0U; 377 return (ARCHIVE_EOF); 378 } 379 380 rab = __archive_read_ahead(a, 1U, &nrd); 381 if (nrd < 0) { 382 *bsz = 0U; 383 /* big catastrophe */ 384 return (int)nrd; 385 } else if (nrd == 0) { 386 goto eof; 387 } else if ((size_t)nrd > w->cntlen - w->cntoff) { 388 /* clamp to content-length */ 389 nrd = w->cntlen - w->cntoff; 390 } 391 *off = w->cntoff; 392 *bsz = nrd; 393 *buf = rab; 394 395 w->cntoff += nrd; 396 w->unconsumed = (size_t)nrd; 397 return (ARCHIVE_OK); 398} 399 400static int 401_warc_skip(struct archive_read *a) 402{ 403 struct warc_s *w = a->format->data; 404 405 __archive_read_consume(a, w->cntlen + 4U/*\r\n\r\n separator*/); 406 w->cntlen = 0U; 407 w->cntoff = 0U; 408 return (ARCHIVE_OK); 409} 410 411 412/* private routines */ 413static void* 414deconst(const void *c) 415{ 416 return (char *)0x1 + (((const char *)c) - (const char *)0x1); 417} 418 419static char* 420xmemmem(const char *hay, const size_t haysize, 421 const char *needle, const size_t needlesize) 422{ 423 const char *const eoh = hay + haysize; 424 const char *const eon = needle + needlesize; 425 const char *hp; 426 const char *np; 427 const char *cand; 428 unsigned int hsum; 429 unsigned int nsum; 430 unsigned int eqp; 431 432 /* trivial checks first 433 * a 0-sized needle is defined to be found anywhere in haystack 434 * then run strchr() to find a candidate in HAYSTACK (i.e. a portion 435 * that happens to begin with *NEEDLE) */ 436 if (needlesize == 0UL) { 437 return deconst(hay); 438 } else if ((hay = memchr(hay, *needle, haysize)) == NULL) { 439 /* trivial */ 440 return NULL; 441 } 442 443 /* First characters of haystack and needle are the same now. Both are 444 * guaranteed to be at least one character long. Now computes the sum 445 * of characters values of needle together with the sum of the first 446 * needle_len characters of haystack. */ 447 for (hp = hay + 1U, np = needle + 1U, hsum = *hay, nsum = *hay, eqp = 1U; 448 hp < eoh && np < eon; 449 hsum ^= *hp, nsum ^= *np, eqp &= *hp == *np, hp++, np++); 450 451 /* HP now references the (NEEDLESIZE + 1)-th character. */ 452 if (np < eon) { 453 /* haystack is smaller than needle, :O */ 454 return NULL; 455 } else if (eqp) { 456 /* found a match */ 457 return deconst(hay); 458 } 459 460 /* now loop through the rest of haystack, 461 * updating the sum iteratively */ 462 for (cand = hay; hp < eoh; hp++) { 463 hsum ^= *cand++; 464 hsum ^= *hp; 465 466 /* Since the sum of the characters is already known to be 467 * equal at that point, it is enough to check just NEEDLESIZE - 1 468 * characters for equality, 469 * also CAND is by design < HP, so no need for range checks */ 470 if (hsum == nsum && memcmp(cand, needle, needlesize - 1U) == 0) { 471 return deconst(cand); 472 } 473 } 474 return NULL; 475} 476 477static int 478strtoi_lim(const char *str, const char **ep, int llim, int ulim) 479{ 480 int res = 0; 481 const char *sp; 482 /* we keep track of the number of digits via rulim */ 483 int rulim; 484 485 for (sp = str, rulim = ulim > 10 ? ulim : 10; 486 res * 10 <= ulim && rulim && *sp >= '0' && *sp <= '9'; 487 sp++, rulim /= 10) { 488 res *= 10; 489 res += *sp - '0'; 490 } 491 if (sp == str) { 492 res = -1; 493 } else if (res < llim || res > ulim) { 494 res = -2; 495 } 496 *ep = (const char*)sp; 497 return res; 498} 499 500static time_t 501time_from_tm(struct tm *t) 502{ 503#if HAVE_TIMEGM 504 /* Use platform timegm() if available. */ 505 return (timegm(t)); 506#elif HAVE__MKGMTIME64 507 return (_mkgmtime64(t)); 508#else 509 /* Else use direct calculation using POSIX assumptions. */ 510 /* First, fix up tm_yday based on the year/month/day. */ 511 if (mktime(t) == (time_t)-1) 512 return ((time_t)-1); 513 /* Then we can compute timegm() from first principles. */ 514 return (t->tm_sec 515 + t->tm_min * 60 516 + t->tm_hour * 3600 517 + t->tm_yday * 86400 518 + (t->tm_year - 70) * 31536000 519 + ((t->tm_year - 69) / 4) * 86400 520 - ((t->tm_year - 1) / 100) * 86400 521 + ((t->tm_year + 299) / 400) * 86400); 522#endif 523} 524 525static time_t 526xstrpisotime(const char *s, char **endptr) 527{ 528/** like strptime() but strictly for ISO 8601 Zulu strings */ 529 struct tm tm; 530 time_t res = (time_t)-1; 531 532 /* make sure tm is clean */ 533 memset(&tm, 0, sizeof(tm)); 534 535 /* as a courtesy to our callers, and since this is a non-standard 536 * routine, we skip leading whitespace */ 537 while (isblank((unsigned char)*s)) 538 ++s; 539 540 /* read year */ 541 if ((tm.tm_year = strtoi_lim(s, &s, 1583, 4095)) < 0 || *s++ != '-') { 542 goto out; 543 } 544 /* read month */ 545 if ((tm.tm_mon = strtoi_lim(s, &s, 1, 12)) < 0 || *s++ != '-') { 546 goto out; 547 } 548 /* read day-of-month */ 549 if ((tm.tm_mday = strtoi_lim(s, &s, 1, 31)) < 0 || *s++ != 'T') { 550 goto out; 551 } 552 /* read hour */ 553 if ((tm.tm_hour = strtoi_lim(s, &s, 0, 23)) < 0 || *s++ != ':') { 554 goto out; 555 } 556 /* read minute */ 557 if ((tm.tm_min = strtoi_lim(s, &s, 0, 59)) < 0 || *s++ != ':') { 558 goto out; 559 } 560 /* read second */ 561 if ((tm.tm_sec = strtoi_lim(s, &s, 0, 60)) < 0 || *s++ != 'Z') { 562 goto out; 563 } 564 565 /* massage TM to fulfill some of POSIX' constraints */ 566 tm.tm_year -= 1900; 567 tm.tm_mon--; 568 569 /* now convert our custom tm struct to a unix stamp using UTC */ 570 res = time_from_tm(&tm); 571 572out: 573 if (endptr != NULL) { 574 *endptr = deconst(s); 575 } 576 return res; 577} 578 579static unsigned int 580_warc_rdver(const char buf[10], size_t bsz) 581{ 582 static const char magic[] = "WARC/"; 583 unsigned int ver; 584 585 (void)bsz; /* UNUSED */ 586 587 if (memcmp(buf, magic, sizeof(magic) - 1U) != 0) { 588 /* nope */ 589 return 99999U; 590 } 591 /* looks good so far, read the version number for a laugh */ 592 buf += sizeof(magic) - 1U; 593 /* most common case gets a quick-check here */ 594 if (memcmp(buf, "1.0\r\n", 5U) == 0) { 595 ver = 10000U; 596 } else { 597 switch (*buf) { 598 case '0': 599 case '1': 600 case '2': 601 case '3': 602 case '4': 603 case '5': 604 case '6': 605 case '7': 606 case '8': 607 if (buf[1U] == '.') { 608 char *on; 609 610 /* set up major version */ 611 ver = (buf[0U] - '0') * 10000U; 612 /* minor version, anyone? */ 613 ver += (strtol(buf + 2U, &on, 10)) * 100U; 614 /* don't parse anything else */ 615 if (on > buf + 2U) { 616 break; 617 } 618 } 619 /* FALLTHROUGH */ 620 case '9': 621 default: 622 /* just make the version ridiculously high */ 623 ver = 999999U; 624 break; 625 } 626 } 627 return ver; 628} 629 630static unsigned int 631_warc_rdtyp(const char *buf, size_t bsz) 632{ 633 static const char _key[] = "\r\nWARC-Type:"; 634 const char *const eob = buf + bsz; 635 const char *val; 636 637 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) { 638 /* no bother */ 639 return WT_NONE; 640 } 641 /* overread whitespace */ 642 val += sizeof(_key) - 1U; 643 while (val < eob && isspace((unsigned char)*val)) 644 ++val; 645 646 if (val + 8U > eob) { 647 ; 648 } else if (memcmp(val, "resource", 8U) == 0) { 649 return WT_RSRC; 650 } else if (memcmp(val, "warcinfo", 8U) == 0) { 651 return WT_INFO; 652 } else if (memcmp(val, "metadata", 8U) == 0) { 653 return WT_META; 654 } else if (memcmp(val, "request", 7U) == 0) { 655 return WT_REQ; 656 } else if (memcmp(val, "response", 8U) == 0) { 657 return WT_RSP; 658 } else if (memcmp(val, "conversi", 8U) == 0) { 659 return WT_CONV; 660 } else if (memcmp(val, "continua", 8U) == 0) { 661 return WT_CONT; 662 } 663 return WT_NONE; 664} 665 666static warc_string_t 667_warc_rduri(const char *buf, size_t bsz) 668{ 669 static const char _key[] = "\r\nWARC-Target-URI:"; 670 const char *const eob = buf + bsz; 671 const char *val; 672 const char *uri; 673 const char *eol; 674 warc_string_t res = {0U, NULL}; 675 676 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) { 677 /* no bother */ 678 return res; 679 } 680 /* overread whitespace */ 681 val += sizeof(_key) - 1U; 682 while (val < eob && isspace((unsigned char)*val)) 683 ++val; 684 685 /* overread URL designators */ 686 if ((uri = xmemmem(val, eob - val, "://", 3U)) == NULL) { 687 /* not touching that! */ 688 return res; 689 } else if ((eol = memchr(uri, '\n', eob - uri)) == NULL) { 690 /* no end of line? :O */ 691 return res; 692 } 693 694 /* massage uri to point to after :// */ 695 uri += 3U; 696 /* also massage eol to point to the first whitespace 697 * after the last non-whitespace character before 698 * the end of the line */ 699 while (eol > uri && isspace((unsigned char)eol[-1])) 700 --eol; 701 702 /* now then, inspect the URI */ 703 if (memcmp(val, "file", 4U) == 0) { 704 /* perfect, nothing left to do here */ 705 706 } else if (memcmp(val, "http", 4U) == 0 || 707 memcmp(val, "ftp", 3U) == 0) { 708 /* overread domain, and the first / */ 709 while (uri < eol && *uri++ != '/'); 710 } else { 711 /* not sure what to do? best to bugger off */ 712 return res; 713 } 714 res.str = uri; 715 res.len = eol - uri; 716 return res; 717} 718 719static ssize_t 720_warc_rdlen(const char *buf, size_t bsz) 721{ 722 static const char _key[] = "\r\nContent-Length:"; 723 const char *val; 724 char *on = NULL; 725 long int len; 726 727 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) { 728 /* no bother */ 729 return -1; 730 } 731 732 /* strtol kindly overreads whitespace for us, so use that */ 733 val += sizeof(_key) - 1U; 734 len = strtol(val, &on, 10); 735 if (on == NULL || !isspace((unsigned char)*on)) { 736 /* hm, can we trust that number? Best not. */ 737 return -1; 738 } 739 return (size_t)len; 740} 741 742static time_t 743_warc_rdrtm(const char *buf, size_t bsz) 744{ 745 static const char _key[] = "\r\nWARC-Date:"; 746 const char *val; 747 char *on = NULL; 748 time_t res; 749 750 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) { 751 /* no bother */ 752 return (time_t)-1; 753 } 754 755 /* xstrpisotime() kindly overreads whitespace for us, so use that */ 756 val += sizeof(_key) - 1U; 757 res = xstrpisotime(val, &on); 758 if (on == NULL || !isspace((unsigned char)*on)) { 759 /* hm, can we trust that number? Best not. */ 760 return (time_t)-1; 761 } 762 return res; 763} 764 765static time_t 766_warc_rdmtm(const char *buf, size_t bsz) 767{ 768 static const char _key[] = "\r\nLast-Modified:"; 769 const char *val; 770 char *on = NULL; 771 time_t res; 772 773 if ((val = xmemmem(buf, bsz, _key, sizeof(_key) - 1U)) == NULL) { 774 /* no bother */ 775 return (time_t)-1; 776 } 777 778 /* xstrpisotime() kindly overreads whitespace for us, so use that */ 779 val += sizeof(_key) - 1U; 780 res = xstrpisotime(val, &on); 781 if (on == NULL || !isspace((unsigned char)*on)) { 782 /* hm, can we trust that number? Best not. */ 783 return (time_t)-1; 784 } 785 return res; 786} 787 788static const char* 789_warc_find_eoh(const char *buf, size_t bsz) 790{ 791 static const char _marker[] = "\r\n\r\n"; 792 const char *hit = xmemmem(buf, bsz, _marker, sizeof(_marker) - 1U); 793 794 if (hit != NULL) { 795 hit += sizeof(_marker) - 1U; 796 } 797 return hit; 798} 799 800/* archive_read_support_format_warc.c ends here */ 801