2 Copyright (C) 1995, 1996, 1997, 2000 Free Software Foundation, Inc.
4 This file is part of Wget.
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or (at
9 your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
30 #include <sys/types.h>
46 /* Default port definitions */
47 #define DEFAULT_HTTP_PORT 80
48 #define DEFAULT_FTP_PORT 21
49 #define DEFAULT_HTTPS_PORT 443
51 /* Table of Unsafe chars. This is intialized in
52 init_unsafe_char_table. */
54 static char unsafe_char_table[256];
56 #define UNSAFE_CHAR(c) (unsafe_char_table[(unsigned char)(c)])
58 /* If S contains unsafe characters, free it and replace it with a
59 version that doesn't. */
60 #define URL_CLEANSE(s) do \
62 if (contains_unsafe (s)) \
64 char *uc_tmp = encode_string (s); \
70 /* Is a directory "."? */
71 #define DOTP(x) ((*(x) == '.') && (!*(x + 1)))
72 /* Is a directory ".."? */
73 #define DDOTP(x) ((*(x) == '.') && (*(x + 1) == '.') && (!*(x + 2)))
76 static void path_simplify_with_kludge PARAMS ((char *));
78 static int urlpath_length PARAMS ((const char *));
80 /* NULL-terminated list of strings to be recognized as prototypes (URL
81 schemes). Note that recognized doesn't mean supported -- only HTTP,
82 HTTPS and FTP are currently supported .
84 However, a string that does not match anything in the list will be
85 considered a relative URL. Thus it's important that this list has
86 anything anyone could think of being legal.
88 There are wild things here. :-) Take a look at
89 <URL:http://www.w3.org/pub/WWW/Addressing/schemes.html> for more
91 static char *protostrings[] =
133 /* Similar to former, but for supported protocols: */
134 static struct proto sup_protos[] =
136 { "http://", URLHTTP, DEFAULT_HTTP_PORT },
138 { "https://",URLHTTPS, DEFAULT_HTTPS_PORT},
140 { "ftp://", URLFTP, DEFAULT_FTP_PORT },
141 /*{ "file://", URLFILE, DEFAULT_FTP_PORT },*/
144 static void parse_dir PARAMS ((const char *, char **, char **));
145 static uerr_t parse_uname PARAMS ((const char *, char **, char **));
146 static char *construct PARAMS ((const char *, const char *, int , int));
147 static char *construct_relative PARAMS ((const char *, const char *));
148 static char process_ftp_type PARAMS ((char *));
151 /* Returns the number of characters to be skipped if the first thing
152 in a URL is URL: (which is 0 or 4+). The optional spaces after
153 URL: are also skipped. */
155 skip_url (const char *url)
159 if (TOUPPER (url[0]) == 'U'
160 && TOUPPER (url[1]) == 'R'
161 && TOUPPER (url[2]) == 'L'
165 for (i = 4; url[i] && ISSPACE (url[i]); i++);
174 - stuff from rfc1738 ("<>\"#%{}|\\^~[]`");
175 - @ and :, for user/password encoding.
176 - everything over 127 (but we don't bother with recording those. */
178 init_unsafe_char_table (void)
181 for (i = 0; i < 256; i++)
182 if (i < 32 || i >= 127
198 unsafe_char_table[i] = 1;
201 /* Returns 1 if the string contains unsafe characters, 0 otherwise. */
203 contains_unsafe (const char *s)
206 if (UNSAFE_CHAR (*s))
211 /* Decodes the forms %xy in a URL to the character the hexadecimal
212 code of which is xy. xy are hexadecimal digits from
213 [0123456789ABCDEF] (case-insensitive). If x or y are not
214 hex-digits or `%' precedes `\0', the sequence is inserted
218 decode_string (char *s)
228 /* Do nothing if at the end of the string, or if the chars
229 are not hex-digits. */
230 if (!*(s + 1) || !*(s + 2)
231 || !(ISXDIGIT (*(s + 1)) && ISXDIGIT (*(s + 2))))
236 *p = (ASC2HEXD (*(s + 1)) << 4) + ASC2HEXD (*(s + 2));
243 /* Encode the unsafe characters (as determined by URL_UNSAFE) in a
244 given string, returning a malloc-ed %XX encoded string. */
246 encode_string (const char *s)
253 for (i = 0; *s; s++, i++)
254 if (UNSAFE_CHAR (*s))
255 i += 2; /* Two more characters (hex digits) */
256 res = (char *)xmalloc (i + 1);
258 for (p = res; *s; s++)
259 if (UNSAFE_CHAR (*s))
261 const unsigned char c = *s;
263 *p++ = HEXD2ASC (c >> 4);
264 *p++ = HEXD2ASC (c & 0xf);
272 /* Returns the proto-type if URL's protocol is supported, or
273 URLUNKNOWN if not. */
275 urlproto (const char *url)
279 url += skip_url (url);
280 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
281 if (!strncasecmp (url, sup_protos[i].name, strlen (sup_protos[i].name)))
282 return sup_protos[i].ind;
283 for (i = 0; url[i] && url[i] != ':' && url[i] != '/'; i++);
286 for (++i; url[i] && url[i] != '/'; i++)
287 if (!ISDIGIT (url[i]))
289 if (url[i - 1] == ':')
298 /* Skip the protocol part of the URL, e.g. `http://'. If no protocol
299 part is found, returns 0. */
301 skip_proto (const char *url)
306 for (s = protostrings; *s; s++)
307 if (!strncasecmp (*s, url, strlen (*s)))
312 /* HTTP and FTP protocols are expected to yield exact host names
313 (i.e. the `//' part must be skipped, too). */
314 if (!strcmp (*s, "http:") || !strcmp (*s, "ftp:"))
319 /* Returns 1 if the URL begins with a protocol (supported or
320 unsupported), 0 otherwise. */
322 has_proto (const char *url)
326 url += skip_url (url);
327 for (s = protostrings; *s; s++)
328 if (strncasecmp (url, *s, strlen (*s)) == 0)
333 /* Skip the username and password, if present here. The function
334 should be called *not* with the complete URL, but with the part
335 right after the protocol.
337 If no username and password are found, return 0. */
339 skip_uname (const char *url)
342 for (p = url; *p && *p != '/'; p++)
345 /* If a `@' was found before the first occurrence of `/', skip
353 /* Allocate a new urlinfo structure, fill it with default values and
354 return a pointer to it. */
360 u = (struct urlinfo *)xmalloc (sizeof (struct urlinfo));
361 memset (u, 0, sizeof (*u));
362 u->proto = URLUNKNOWN;
366 /* Perform a "deep" free of the urlinfo structure. The structure
367 should have been created with newurl, but need not have been used.
368 If free_pointer is non-0, free the pointer itself. */
370 freeurl (struct urlinfo *u, int complete)
374 FREE_MAYBE (u->host);
375 FREE_MAYBE (u->path);
376 FREE_MAYBE (u->file);
378 FREE_MAYBE (u->user);
379 FREE_MAYBE (u->passwd);
380 FREE_MAYBE (u->local);
381 FREE_MAYBE (u->referer);
383 freeurl (u->proxy, 1);
389 /* Extract the given URL of the form
390 (http:|ftp:)// (user (:password)?@)?hostname (:port)? (/path)?
391 1. hostname (terminated with `/' or `:')
392 2. port number (terminated with `/'), or chosen for the protocol
393 3. dirname (everything after hostname)
394 Most errors are handled. No allocation is done, you must supply
395 pointers to allocated memory.
396 ...and a host of other stuff :-)
398 - Recognizes hostname:dir/file for FTP and
399 hostname (:portnum)?/dir/file for HTTP.
400 - Parses the path to yield directory and file
401 - Parses the URL to yield the username and passwd (if present)
402 - Decodes the strings, in case they contain "forbidden" characters
403 - Writes the result to struct urlinfo
405 If the argument STRICT is set, it recognizes only the canonical
408 parseurl (const char *url, struct urlinfo *u, int strict)
411 int recognizable; /* Recognizable URL is the one where
412 the protocol name was explicitly
413 named, i.e. it wasn't deduced from
417 DEBUGP (("parseurl (\"%s\") -> ", url));
418 url += skip_url (url);
419 recognizable = has_proto (url);
420 if (strict && !recognizable)
422 for (i = 0, l = 0; i < ARRAY_SIZE (sup_protos); i++)
424 l = strlen (sup_protos[i].name);
425 if (!strncasecmp (sup_protos[i].name, url, l))
428 /* If protocol is recognizable, but unsupported, bail out, else
430 if (recognizable && i == ARRAY_SIZE (sup_protos))
432 else if (i == ARRAY_SIZE (sup_protos))
435 u->proto = type = sup_protos[i].ind;
437 if (type == URLUNKNOWN)
439 /* Allow a username and password to be specified (i.e. just skip
442 l += skip_uname (url + l);
443 for (i = l; url[i] && url[i] != ':' && url[i] != '/'; i++);
446 /* Get the hostname. */
447 u->host = strdupdelim (url + l, url + i);
448 DEBUGP (("host %s -> ", u->host));
450 /* Assume no port has been given. */
454 /* We have a colon delimiting the hostname. It could mean that
455 a port number is following it, or a directory. */
456 if (ISDIGIT (url[++i])) /* A port number */
458 if (type == URLUNKNOWN)
459 u->proto = type = URLHTTP;
460 for (; url[i] && url[i] != '/'; i++)
461 if (ISDIGIT (url[i]))
462 u->port = 10 * u->port + (url[i] - '0');
467 DEBUGP (("port %hu -> ", u->port));
469 else if (type == URLUNKNOWN) /* or a directory */
470 u->proto = type = URLFTP;
471 else /* or just a misformed port number */
474 else if (type == URLUNKNOWN)
475 u->proto = type = URLHTTP;
479 for (ind = 0; ind < ARRAY_SIZE (sup_protos); ind++)
480 if (sup_protos[ind].ind == type)
482 if (ind == ARRAY_SIZE (sup_protos))
484 u->port = sup_protos[ind].port;
486 /* Some delimiter troubles... */
487 if (url[i] == '/' && url[i - 1] != ':')
490 while (url[i] && url[i] == '/')
492 u->path = (char *)xmalloc (strlen (url + i) + 8);
493 strcpy (u->path, url + i);
496 u->ftp_type = process_ftp_type (u->path);
497 /* #### We don't handle type `d' correctly yet. */
498 if (!u->ftp_type || TOUPPER (u->ftp_type) == 'D')
501 DEBUGP (("opath %s -> ", u->path));
502 /* Parse the username and password (if existing). */
503 parse_uname (url, &u->user, &u->passwd);
504 /* Decode the strings, as per RFC 1738. */
505 decode_string (u->host);
506 decode_string (u->path);
508 decode_string (u->user);
510 decode_string (u->passwd);
511 /* Parse the directory. */
512 parse_dir (u->path, &u->dir, &u->file);
513 DEBUGP (("dir %s -> file %s -> ", u->dir, u->file));
514 /* Simplify the directory. */
515 path_simplify (u->dir);
516 /* Remove the leading `/' in HTTP. */
517 if (type == URLHTTP && *u->dir == '/')
518 strcpy (u->dir, u->dir + 1);
519 DEBUGP (("ndir %s\n", u->dir));
520 /* Strip trailing `/'. */
522 if (l && u->dir[l - 1] == '/')
523 u->dir[l - 1] = '\0';
524 /* Re-create the path: */
525 abs_ftp = (u->proto == URLFTP && *u->dir == '/');
526 /* sprintf (u->path, "%s%s%s%s", abs_ftp ? "%2F": "/",
527 abs_ftp ? (u->dir + 1) : u->dir, *u->dir ? "/" : "", u->file); */
528 strcpy (u->path, abs_ftp ? "%2F" : "/");
529 strcat (u->path, abs_ftp ? (u->dir + 1) : u->dir);
530 strcat (u->path, *u->dir ? "/" : "");
531 strcat (u->path, u->file);
532 URL_CLEANSE (u->path);
533 DEBUGP (("newpath: %s\n", u->path));
534 /* Create the clean URL. */
535 u->url = str_url (u, 0);
539 /* Special versions of DOTP and DDOTP for parse_dir(). */
541 #define PD_DOTP(x) ((*(x) == '.') && (!*((x) + 1) || *((x) + 1) == '?'))
542 #define PD_DDOTP(x) ((*(x) == '.') && (*(x) == '.') \
543 && (!*((x) + 2) || *((x) + 2) == '?'))
545 /* Build the directory and filename components of the path. Both
546 components are *separately* malloc-ed strings! It does not change
547 the contents of path.
549 If the path ends with "." or "..", they are (correctly) counted as
552 parse_dir (const char *path, char **dir, char **file)
556 l = urlpath_length (path);
557 for (i = l; i && path[i] != '/'; i--);
559 if (!i && *path != '/') /* Just filename */
561 if (PD_DOTP (path) || PD_DDOTP (path))
563 *dir = strdupdelim (path, path + l);
564 *file = xstrdup (path + l); /* normally empty, but could
569 *dir = xstrdup (""); /* This is required because of FTP */
570 *file = xstrdup (path);
573 else if (!i) /* /filename */
575 if (PD_DOTP (path + 1) || PD_DDOTP (path + 1))
577 *dir = strdupdelim (path, path + l);
578 *file = xstrdup (path + l); /* normally empty, but could
583 *dir = xstrdup ("/");
584 *file = xstrdup (path + 1);
587 else /* Nonempty directory with or without a filename */
589 if (PD_DOTP (path + i + 1) || PD_DDOTP (path + i + 1))
591 *dir = strdupdelim (path, path + l);
592 *file = xstrdup (path + l); /* normally empty, but could
597 *dir = strdupdelim (path, path + i);
598 *file = xstrdup (path + i + 1);
603 /* Find the optional username and password within the URL, as per
604 RFC1738. The returned user and passwd char pointers are
607 parse_uname (const char *url, char **user, char **passwd)
615 url += skip_url (url);
616 /* Look for end of protocol string. */
617 l = skip_proto (url);
620 /* Add protocol offset. */
622 /* Is there an `@' character? */
623 for (p = url; *p && *p != '/'; p++)
626 /* If not, return. */
629 /* Else find the username and password. */
630 for (p = col = url; *p != '@'; p++)
632 if (*p == ':' && !*user)
634 *user = (char *)xmalloc (p - url + 1);
635 memcpy (*user, url, p - url);
636 (*user)[p - url] = '\0';
640 /* Decide whether you have only the username or both. */
641 where = *user ? passwd : user;
642 *where = (char *)xmalloc (p - col + 1);
643 memcpy (*where, col, p - col);
644 (*where)[p - col] = '\0';
648 /* If PATH ends with `;type=X', return the character X. */
650 process_ftp_type (char *path)
652 int len = strlen (path);
655 && !memcmp (path + len - 7, ";type=", 6))
657 path[len - 7] = '\0';
658 return path[len - 1];
664 /* Return the URL as fine-formed string, with a proper protocol,
665 optional port number, directory and optional user/password. If
666 HIDE is non-zero, password will be hidden. The forbidden
667 characters in the URL will be cleansed. */
669 str_url (const struct urlinfo *u, int hide)
671 char *res, *host, *user, *passwd, *proto_name, *dir, *file;
672 int i, l, ln, lu, lh, lp, lf, ld;
673 unsigned short proto_default_port;
675 /* Look for the protocol name. */
676 for (i = 0; i < ARRAY_SIZE (sup_protos); i++)
677 if (sup_protos[i].ind == u->proto)
679 if (i == ARRAY_SIZE (sup_protos))
681 proto_name = sup_protos[i].name;
682 proto_default_port = sup_protos[i].port;
683 host = CLEANDUP (u->host);
684 dir = CLEANDUP (u->dir);
685 file = CLEANDUP (u->file);
686 user = passwd = NULL;
688 user = CLEANDUP (u->user);
692 passwd = CLEANDUP (u->passwd);
694 for (j = 0; passwd[j]; j++)
697 if (u->proto == URLFTP && *dir == '/')
699 char *tmp = (char *)xmalloc (strlen (dir) + 3);
700 /*sprintf (tmp, "%%2F%s", dir + 1);*/
704 strcpy (tmp + 3, dir + 1);
709 ln = strlen (proto_name);
710 lu = user ? strlen (user) : 0;
711 lp = passwd ? strlen (passwd) : 0;
715 res = (char *)xmalloc (ln + lu + lp + lh + ld + lf + 20); /* safe sex */
716 /* sprintf (res, "%s%s%s%s%s%s:%d/%s%s%s", proto_name,
717 (user ? user : ""), (passwd ? ":" : ""),
718 (passwd ? passwd : ""), (user ? "@" : ""),
719 host, u->port, dir, *dir ? "/" : "", file); */
721 memcpy (res, proto_name, ln);
725 memcpy (res + l, user, lu);
730 memcpy (res + l, passwd, lp);
735 memcpy (res + l, host, lh);
737 if (u->port != proto_default_port)
740 long_to_string (res + l, (long)u->port);
741 l += numdigit (u->port);
744 memcpy (res + l, dir, ld);
748 strcpy (res + l, file);
757 /* Check whether two URL-s are equivalent, i.e. pointing to the same
758 location. Uses parseurl to parse them, and compares the canonical
761 Returns 1 if the URL1 is equivalent to URL2, 0 otherwise. Also
762 return 0 on error. */
764 url_equal (const char *url1, const char *url2)
766 struct urlinfo *u1, *u2;
771 err = parseurl (url1, u1, 0);
778 err = parseurl (url2, u2, 0);
784 res = !strcmp (u1->url, u2->url);
791 get_urls_file (const char *file)
793 struct file_memory *fm;
795 const char *text, *text_end;
798 fm = read_file (file);
801 logprintf (LOG_NOTQUIET, "%s: %s\n", file, strerror (errno));
804 DEBUGP (("Loaded %s (size %ld).\n", file, fm->length));
807 text_end = fm->content + fm->length;
808 while (text < text_end)
810 const char *line_beg = text;
811 const char *line_end = memchr (text, '\n', text_end - text);
817 while (line_beg < line_end
818 && ISSPACE (*line_beg))
820 while (line_end > line_beg + 1
821 && ISSPACE (*(line_end - 1)))
823 if (line_end > line_beg)
825 urlpos *entry = (urlpos *)xmalloc (sizeof (urlpos));
826 memset (entry, 0, sizeof (*entry));
828 entry->url = strdupdelim (line_beg, line_end);
840 /* Free the linked list of urlpos. */
842 free_urlpos (urlpos *l)
846 urlpos *next = l->next;
848 FREE_MAYBE (l->local_name);
854 /* Rotate FNAME opt.backups times */
856 rotate_backups(const char *fname)
858 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
859 char *from = (char *)alloca (maxlen);
860 char *to = (char *)alloca (maxlen);
864 if (stat (fname, &sb) == 0)
865 if (S_ISREG (sb.st_mode) == 0)
868 for (i = opt.backups; i > 1; i--)
870 sprintf (from, "%s.%d", fname, i - 1);
871 sprintf (to, "%s.%d", fname, i);
872 /* #### This will fail on machines without the rename() system
877 sprintf (to, "%s.%d", fname, 1);
881 /* Create all the necessary directories for PATH (a file). Calls
882 mkdirhier() internally. */
884 mkalldirs (const char *path)
891 p = path + strlen (path);
892 for (; *p != '/' && p != path; p--);
893 /* Don't create if it's just a file. */
894 if ((p == path) && (*p != '/'))
896 t = strdupdelim (path, p);
897 /* Check whether the directory exists. */
898 if ((stat (t, &st) == 0))
900 if (S_ISDIR (st.st_mode))
907 /* If the dir exists as a file name, remove it first. This
908 is *only* for Wget to work with buggy old CERN http
909 servers. Here is the scenario: When Wget tries to
910 retrieve a directory without a slash, e.g.
911 http://foo/bar (bar being a directory), CERN server will
912 not redirect it too http://foo/bar/ -- it will generate a
913 directory listing containing links to bar/file1,
914 bar/file2, etc. Wget will lose because it saves this
915 HTML listing to a file `bar', so it cannot create the
916 directory. To work around this, if the file of the same
917 name exists, we just remove it and create the directory
919 DEBUGP (("Removing %s because of directory danger!\n", t));
923 res = make_directory (t);
925 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
931 count_slashes (const char *s)
940 /* Return the path name of the URL-equivalent file name, with a
941 remote-like structure of directories. */
943 mkstruct (const struct urlinfo *u)
945 char *host, *dir, *file, *res, *dirpref;
948 assert (u->dir != NULL);
949 assert (u->host != NULL);
953 char *ptr = u->dir + (*u->dir == '/');
954 int slash_count = 1 + count_slashes (ptr);
955 int cut = MINVAL (opt.cut_dirs, slash_count);
956 for (; cut && *ptr; ptr++)
959 STRDUP_ALLOCA (dir, ptr);
962 dir = u->dir + (*u->dir == '/');
964 host = xstrdup (u->host);
965 /* Check for the true name (or at least a consistent name for saving
966 to directory) of HOST, reusing the hlist if possible. */
967 if (opt.add_hostdir && !opt.simple_check)
969 char *nhost = realhost (host);
973 /* Add dir_prefix and hostname (if required) to the beginning of
977 if (!DOTP (opt.dir_prefix))
979 dirpref = (char *)alloca (strlen (opt.dir_prefix) + 1
980 + strlen (host) + 1);
981 sprintf (dirpref, "%s/%s", opt.dir_prefix, host);
984 STRDUP_ALLOCA (dirpref, host);
986 else /* not add_hostdir */
988 if (!DOTP (opt.dir_prefix))
989 dirpref = opt.dir_prefix;
995 /* If there is a prefix, prepend it. */
998 char *newdir = (char *)alloca (strlen (dirpref) + 1 + strlen (dir) + 2);
999 sprintf (newdir, "%s%s%s", dirpref, *dir == '/' ? "" : "/", dir);
1002 dir = xstrdup (dir);
1005 if (l && dir[l - 1] == '/')
1009 file = "index.html";
1013 /* Finally, construct the full name. */
1014 res = (char *)xmalloc (strlen (dir) + 1 + strlen (file) + 1);
1015 sprintf (res, "%s%s%s", dir, *dir ? "/" : "", file);
1020 /* Create a unique filename, corresponding to a given URL. Calls
1021 mkstruct if necessary. Does *not* actually create any directories. */
1023 url_filename (const struct urlinfo *u)
1026 int have_prefix = 0; /* whether we must prepend opt.dir_prefix */
1030 file = mkstruct (u);
1036 file = xstrdup ("index.html");
1038 file = xstrdup (u->file);
1043 /* Check whether the prefix directory is something other than "."
1044 before prepending it. */
1045 if (!DOTP (opt.dir_prefix))
1047 char *nfile = (char *)xmalloc (strlen (opt.dir_prefix)
1048 + 1 + strlen (file) + 1);
1049 sprintf (nfile, "%s/%s", opt.dir_prefix, file);
1054 /* DOS-ish file systems don't like `%' signs in them; we change it
1059 for (p = file; *p; p++)
1063 #endif /* WINDOWS */
1065 /* Check the cases in which the unique extensions are not used:
1066 1) Clobbering is turned off (-nc).
1067 2) Retrieval with regetting.
1068 3) Timestamping is used.
1069 4) Hierarchy is built.
1071 The exception is the case when file does exist and is a
1072 directory (actually support for bad httpd-s). */
1073 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1074 && !(file_exists_p (file) && !file_non_directory_p (file)))
1077 /* Find a unique name. */
1078 name = unique_name (file);
1083 /* Like strlen(), but allow the URL to be ended with '?'. */
1085 urlpath_length (const char *url)
1087 const char *q = strchr (url, '?');
1090 return strlen (url);
1093 /* Find the last occurrence of character C in the range [b, e), or
1094 NULL, if none are present. This is almost completely equivalent to
1095 { *e = '\0'; return strrchr(b); }, except that it doesn't change
1096 the contents of the string. */
1098 find_last_char (const char *b, const char *e, char c)
1106 /* Construct a URL by concatenating an absolute URL and a path, which
1107 may or may not be absolute. This tries to behave "reasonably" in
1108 all foreseeable cases. It employs little specific knowledge about
1109 protocols or URL-specific stuff -- it just works on strings. */
1111 construct (const char *url, const char *sub, int subsize, int no_proto)
1117 const char *end = url + urlpath_length (url);
1121 /* SUB is a relative URL: we need to replace everything
1122 after last slash (possibly empty) with SUB.
1124 So, if URL is "whatever/foo/bar", and SUB is "qux/xyzzy",
1125 our result should be "whatever/foo/qux/xyzzy". */
1126 int need_explicit_slash = 0;
1128 const char *start_insert;
1129 const char *last_slash = find_last_char (url, end, '/'); /* the last slash. */
1132 /* No slash found at all. Append SUB to what we have,
1133 but we'll need a slash as a separator.
1135 Example: if url == "foo" and sub == "qux/xyzzy", then
1136 we cannot just append sub to url, because we'd get
1137 "fooqux/xyzzy", whereas what we want is
1140 To make sure the / gets inserted, we set
1141 need_explicit_slash to 1. We also set start_insert
1142 to end + 1, so that the length calculations work out
1143 correctly for one more (slash) character. Accessing
1144 that character is fine, since it will be the
1145 delimiter, '\0' or '?'. */
1146 /* example: "foo?..." */
1147 /* ^ ('?' gets changed to '/') */
1148 start_insert = end + 1;
1149 need_explicit_slash = 1;
1151 else if (last_slash && last_slash != url && *(last_slash - 1) == '/')
1153 /* example: http://host" */
1155 start_insert = end + 1;
1156 need_explicit_slash = 1;
1160 /* example: "whatever/foo/bar" */
1162 start_insert = last_slash + 1;
1165 span = start_insert - url;
1166 constr = (char *)xmalloc (span + subsize + 1);
1168 memcpy (constr, url, span);
1169 if (need_explicit_slash)
1170 constr[span - 1] = '/';
1172 memcpy (constr + span, sub, subsize);
1173 constr[span + subsize] = '\0';
1175 else /* *sub == `/' */
1177 /* SUB is an absolute path: we need to replace everything
1178 after (and including) the FIRST slash with SUB.
1180 So, if URL is "http://host/whatever/foo/bar", and SUB is
1181 "/qux/xyzzy", our result should be
1182 "http://host/qux/xyzzy". */
1185 const char *start_insert = NULL; /* for gcc to shut up. */
1186 const char *pos = url;
1187 int seen_slash_slash = 0;
1188 /* We're looking for the first slash, but want to ignore
1191 slash = memchr (pos, '/', end - pos);
1192 if (slash && !seen_slash_slash)
1193 if (*(slash + 1) == '/')
1196 seen_slash_slash = 1;
1200 /* At this point, SLASH is the location of the first / after
1201 "//", or the first slash altogether. START_INSERT is the
1202 pointer to the location where SUB will be inserted. When
1203 examining the last two examples, keep in mind that SUB
1206 if (!slash && !seen_slash_slash)
1207 /* example: "foo" */
1210 else if (!slash && seen_slash_slash)
1211 /* example: "http://foo" */
1214 else if (slash && !seen_slash_slash)
1215 /* example: "foo/bar" */
1218 else if (slash && seen_slash_slash)
1219 /* example: "http://something/" */
1221 start_insert = slash;
1223 span = start_insert - url;
1224 constr = (char *)xmalloc (span + subsize + 1);
1226 memcpy (constr, url, span);
1228 memcpy (constr + span, sub, subsize);
1229 constr[span + subsize] = '\0';
1232 else /* !no_proto */
1234 constr = strdupdelim (sub, sub + subsize);
1239 /* Like the function above, but with a saner caller interface. */
1241 url_concat (const char *base_url, const char *new_url)
1243 return construct (base_url, new_url, strlen (new_url), !has_proto (new_url));
1246 /* Optimize URL by host, destructively replacing u->host with realhost
1247 (u->host). Do this regardless of opt.simple_check. */
1249 opt_url (struct urlinfo *u)
1251 /* Find the "true" host. */
1252 char *host = realhost (u->host);
1255 assert (u->dir != NULL); /* the URL must have been parsed */
1256 /* Refresh the printed representation. */
1258 u->url = str_url (u, 0);
1261 /* This beautiful kludge is fortunately not needed, as I've made
1262 parse_dir do the (almost) right thing, so that a query can never
1263 become a part of directory. */
1265 /* Call path_simplify, but make sure that the part after the
1266 question-mark, if any, is not destroyed by path_simplify's
1269 path_simplify_with_kludge (char *path)
1271 char *query = strchr (path, '?');
1273 /* path_simplify also works destructively, so we also have the
1274 license to write. */
1276 path_simplify (path);
1279 char *newend = path + strlen (path);
1281 if (newend != query)
1282 memmove (newend, query, strlen (query) + 1);
1287 /* Returns proxy host address, in accordance with PROTO. */
1289 getproxy (uerr_t proto)
1291 if (proto == URLHTTP)
1292 return opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
1293 else if (proto == URLFTP)
1294 return opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
1296 else if (proto == URLHTTPS)
1297 return opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
1298 #endif /* HAVE_SSL */
1303 /* Should a host be accessed through proxy, concerning no_proxy? */
1305 no_proxy_match (const char *host, const char **no_proxy)
1310 return !sufmatch (no_proxy, host);
1313 static void write_backup_file PARAMS ((const char *, downloaded_file_t));
1314 static void replace_attr PARAMS ((const char **, int, FILE *, const char *));
1316 /* Change the links in an HTML document. Accepts a structure that
1317 defines the positions of all the links. */
1319 convert_links (const char *file, urlpos *l)
1321 struct file_memory *fm;
1324 downloaded_file_t downloaded_file_return;
1326 logprintf (LOG_VERBOSE, _("Converting %s... "), file);
1329 /* First we do a "dry run": go through the list L and see whether
1330 any URL needs to be converted in the first place. If not, just
1331 leave the file alone. */
1334 for (dry = l; dry; dry = dry->next)
1335 if (dry->convert != CO_NOCONVERT)
1339 logputs (LOG_VERBOSE, _("nothing to do.\n"));
1344 fm = read_file (file);
1347 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1348 file, strerror (errno));
1352 downloaded_file_return = downloaded_file (CHECK_FOR_FILE, file);
1353 if (opt.backup_converted && downloaded_file_return)
1354 write_backup_file (file, downloaded_file_return);
1356 /* Before opening the file for writing, unlink the file. This is
1357 important if the data in FM is mmaped. In such case, nulling the
1358 file, which is what fopen() below does, would make us read all
1359 zeroes from the mmaped region. */
1360 if (unlink (file) < 0 && errno != ENOENT)
1362 logprintf (LOG_NOTQUIET, _("Unable to delete `%s': %s\n"),
1363 file, strerror (errno));
1364 read_file_free (fm);
1367 /* Now open the file for writing. */
1368 fp = fopen (file, "wb");
1371 logprintf (LOG_NOTQUIET, _("Cannot convert links in %s: %s\n"),
1372 file, strerror (errno));
1373 read_file_free (fm);
1376 /* Here we loop through all the URLs in file, replacing those of
1377 them that are downloaded with relative references. */
1379 for (; l; l = l->next)
1381 char *url_start = fm->content + l->pos;
1383 if (l->pos >= fm->length)
1385 DEBUGP (("Something strange is going on. Please investigate."));
1388 /* If the URL is not to be converted, skip it. */
1389 if (l->convert == CO_NOCONVERT)
1391 DEBUGP (("Skipping %s at position %d.\n", l->url, l->pos));
1395 /* Echo the file contents, up to the offending URL's opening
1396 quote, to the outfile. */
1397 fwrite (p, 1, url_start - p, fp);
1399 if (l->convert == CO_CONVERT_TO_RELATIVE)
1401 /* Convert absolute URL to relative. */
1402 char *newname = construct_relative (file, l->local_name);
1403 char *quoted_newname = html_quote_string (newname);
1404 replace_attr (&p, l->size, fp, quoted_newname);
1405 DEBUGP (("TO_RELATIVE: %s to %s at position %d in %s.\n",
1406 l->url, newname, l->pos, file));
1408 xfree (quoted_newname);
1410 else if (l->convert == CO_CONVERT_TO_COMPLETE)
1412 /* Convert the link to absolute URL. */
1413 char *newlink = l->url;
1414 char *quoted_newlink = html_quote_string (newlink);
1415 replace_attr (&p, l->size, fp, quoted_newlink);
1416 DEBUGP (("TO_COMPLETE: <something> to %s at position %d in %s.\n",
1417 newlink, l->pos, file));
1418 xfree (quoted_newlink);
1421 /* Output the rest of the file. */
1422 if (p - fm->content < fm->length)
1423 fwrite (p, 1, fm->length - (p - fm->content), fp);
1425 read_file_free (fm);
1426 logputs (LOG_VERBOSE, _("done.\n"));
1429 /* Construct and return a malloced copy of the relative link from two
1430 pieces of information: local name S1 of the referring file and
1431 local name S2 of the referred file.
1433 So, if S1 is "jagor.srce.hr/index.html" and S2 is
1434 "jagor.srce.hr/images/news.gif", the function will return
1437 Alternately, if S1 is "fly.cc.fer.hr/ioccc/index.html", and S2 is
1438 "fly.cc.fer.hr/images/fly.gif", the function will return
1439 "../images/fly.gif".
1441 Caveats: S1 should not begin with `/', unless S2 also begins with
1442 '/'. S1 should not contain things like ".." and such --
1443 construct_relative ("fly/ioccc/../index.html",
1444 "fly/images/fly.gif") will fail. (A workaround is to call
1445 something like path_simplify() on S1). */
1447 construct_relative (const char *s1, const char *s2)
1449 int i, cnt, sepdirs1;
1453 return xstrdup (s2);
1454 /* S1 should *not* be absolute, if S2 wasn't. */
1455 assert (*s1 != '/');
1457 /* Skip the directories common to both strings. */
1460 while (s1[i] && s2[i]
1465 if (s1[i] == '/' && s2[i] == '/')
1470 for (sepdirs1 = 0; s1[i]; i++)
1473 /* Now, construct the file as of:
1474 - ../ repeated sepdirs1 time
1475 - all the non-mutual directories of S2. */
1476 res = (char *)xmalloc (3 * sepdirs1 + strlen (s2 + cnt) + 1);
1477 for (i = 0; i < sepdirs1; i++)
1478 memcpy (res + 3 * i, "../", 3);
1479 strcpy (res + 3 * i, s2 + cnt);
1483 /* Add URL to the head of the list L. */
1485 add_url (urlpos *l, const char *url, const char *file)
1489 t = (urlpos *)xmalloc (sizeof (urlpos));
1490 memset (t, 0, sizeof (*t));
1491 t->url = xstrdup (url);
1492 t->local_name = xstrdup (file);
1498 write_backup_file (const char *file, downloaded_file_t downloaded_file_return)
1500 /* Rather than just writing over the original .html file with the
1501 converted version, save the former to *.orig. Note we only do
1502 this for files we've _successfully_ downloaded, so we don't
1503 clobber .orig files sitting around from previous invocations. */
1505 /* Construct the backup filename as the original name plus ".orig". */
1506 size_t filename_len = strlen(file);
1507 char* filename_plus_orig_suffix;
1508 boolean already_wrote_backup_file = FALSE;
1509 slist* converted_file_ptr;
1510 static slist* converted_files = NULL;
1512 if (downloaded_file_return == FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED)
1514 /* Just write "orig" over "html". We need to do it this way
1515 because when we're checking to see if we've downloaded the
1516 file before (to see if we can skip downloading it), we don't
1517 know if it's a text/html file. Therefore we don't know yet
1518 at that stage that -E is going to cause us to tack on
1519 ".html", so we need to compare vs. the original URL plus
1520 ".orig", not the original URL plus ".html.orig". */
1521 filename_plus_orig_suffix = alloca (filename_len + 1);
1522 strcpy(filename_plus_orig_suffix, file);
1523 strcpy((filename_plus_orig_suffix + filename_len) - 4, "orig");
1525 else /* downloaded_file_return == FILE_DOWNLOADED_NORMALLY */
1527 /* Append ".orig" to the name. */
1528 filename_plus_orig_suffix = alloca (filename_len + sizeof(".orig"));
1529 strcpy(filename_plus_orig_suffix, file);
1530 strcpy(filename_plus_orig_suffix + filename_len, ".orig");
1533 /* We can get called twice on the same URL thanks to the
1534 convert_all_links() call in main(). If we write the .orig file
1535 each time in such a case, it'll end up containing the first-pass
1536 conversion, not the original file. So, see if we've already been
1537 called on this file. */
1538 converted_file_ptr = converted_files;
1539 while (converted_file_ptr != NULL)
1540 if (strcmp(converted_file_ptr->string, file) == 0)
1542 already_wrote_backup_file = TRUE;
1546 converted_file_ptr = converted_file_ptr->next;
1548 if (!already_wrote_backup_file)
1550 /* Rename <file> to <file>.orig before former gets written over. */
1551 if (rename(file, filename_plus_orig_suffix) != 0)
1552 logprintf (LOG_NOTQUIET, _("Cannot back up %s as %s: %s\n"),
1553 file, filename_plus_orig_suffix, strerror (errno));
1555 /* Remember that we've already written a .orig backup for this file.
1556 Note that we never free this memory since we need it till the
1557 convert_all_links() call, which is one of the last things the
1558 program does before terminating. BTW, I'm not sure if it would be
1559 safe to just set 'converted_file_ptr->string' to 'file' below,
1560 rather than making a copy of the string... Another note is that I
1561 thought I could just add a field to the urlpos structure saying
1562 that we'd written a .orig file for this URL, but that didn't work,
1563 so I had to make this separate list.
1565 This [adding a field to the urlpos structure] didn't work
1566 because convert_file() is called twice: once after all its
1567 sublinks have been retrieved in recursive_retrieve(), and
1568 once at the end of the day in convert_all_links(). The
1569 original linked list collected in recursive_retrieve() is
1570 lost after the first invocation of convert_links(), and
1571 convert_all_links() makes a new one (it calls get_urls_html()
1572 for each file it covers.) That's why your approach didn't
1573 work. The way to make it work is perhaps to make this flag a
1574 field in the `urls_html' list. */
1576 converted_file_ptr = xmalloc(sizeof(*converted_file_ptr));
1577 converted_file_ptr->string = xstrdup(file); /* die on out-of-mem. */
1578 converted_file_ptr->next = converted_files;
1579 converted_files = converted_file_ptr;
1583 static int find_fragment PARAMS ((const char *, int, const char **,
1587 replace_attr (const char **pp, int raw_size, FILE *fp, const char *new_str)
1589 const char *p = *pp;
1591 int size = raw_size;
1592 char quote_char = '\"';
1593 const char *frag_beg, *frag_end;
1595 /* Structure of our string is:
1596 "...old-contents..."
1597 <--- l->size ---> (with quotes)
1600 <--- l->size --> (no quotes) */
1602 if (*p == '\"' || *p == '\'')
1607 size -= 2; /* disregard opening and closing quote */
1609 putc (quote_char, fp);
1610 fputs (new_str, fp);
1612 /* Look for fragment identifier, if any. */
1613 if (find_fragment (p, size, &frag_beg, &frag_end))
1614 fwrite (frag_beg, 1, frag_end - frag_beg, fp);
1618 putc (quote_char, fp);
1622 /* Find the first occurrence of '#' in [BEG, BEG+SIZE) that is not
1623 preceded by '&'. If the character is not found, return zero. If
1624 the character is found, return 1 and set BP and EP to point to the
1625 beginning and end of the region.
1627 This is used for finding the fragment indentifiers in URLs. */
1630 find_fragment (const char *beg, int size, const char **bp, const char **ep)
1632 const char *end = beg + size;
1634 for (; beg < end; beg++)
1656 typedef struct _downloaded_file_list {
1658 downloaded_file_t download_type;
1659 struct _downloaded_file_list* next;
1660 } downloaded_file_list;
1662 static downloaded_file_list *downloaded_files;
1664 /* Remembers which files have been downloaded. In the standard case, should be
1665 called with mode == FILE_DOWNLOADED_NORMALLY for each file we actually
1666 download successfully (i.e. not for ones we have failures on or that we skip
1669 When we've downloaded a file and tacked on a ".html" extension due to -E,
1670 call this function with FILE_DOWNLOADED_AND_HTML_EXTENSION_ADDED rather than
1671 FILE_DOWNLOADED_NORMALLY.
1673 If you just want to check if a file has been previously added without adding
1674 it, call with mode == CHECK_FOR_FILE. Please be sure to call this function
1675 with local filenames, not remote URLs. */
1677 downloaded_file (downloaded_file_t mode, const char* file)
1679 boolean found_file = FALSE;
1680 downloaded_file_list* rover = downloaded_files;
1682 while (rover != NULL)
1683 if (strcmp(rover->file, file) == 0)
1689 rover = rover->next;
1692 return rover->download_type; /* file had already been downloaded */
1695 if (mode != CHECK_FOR_FILE)
1697 rover = xmalloc(sizeof(*rover));
1698 rover->file = xstrdup(file); /* use xstrdup() so die on out-of-mem. */
1699 rover->download_type = mode;
1700 rover->next = downloaded_files;
1701 downloaded_files = rover;
1704 return FILE_NOT_ALREADY_DOWNLOADED;
1709 downloaded_files_free (void)
1711 downloaded_file_list* rover = downloaded_files;
1714 downloaded_file_list *next = rover->next;
1715 xfree (rover->file);
1721 /* Initialization of static stuff. */
1725 init_unsafe_char_table ();