2 Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget 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 GNU Wget 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 Wget; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
34 #include <sys/types.h>
37 #endif /* HAVE_UNISTD_H */
43 #endif /* HAVE_STRING_H */
59 # include "gen_sslfunc.h" /* for ssl_iread */
66 /* See the comment in gethttp() why this is needed. */
67 int global_download_count;
69 /* Total size of downloaded files. Used to enforce quota. */
70 LARGE_INT total_downloaded_bytes;
80 limit_bandwidth_reset (void)
82 limit_data.chunk_bytes = 0;
83 limit_data.chunk_start = 0;
86 /* Limit the bandwidth by pausing the download for an amount of time.
87 BYTES is the number of bytes received from the network, and TIMER
88 is the timer that started at the beginning of download. */
91 limit_bandwidth (long bytes, struct wget_timer *timer)
93 double delta_t = wtimer_read (timer) - limit_data.chunk_start;
96 limit_data.chunk_bytes += bytes;
98 /* Calculate the amount of time we expect downloading the chunk
99 should take. If in reality it took less time, sleep to
100 compensate for the difference. */
101 expected = 1000.0 * limit_data.chunk_bytes / opt.limit_rate;
103 if (expected > delta_t)
105 double slp = expected - delta_t + limit_data.sleep_adjust;
109 DEBUGP (("deferring a %.2f ms sleep (%ld/%.2f).\n",
110 slp, limit_data.chunk_bytes, delta_t));
113 DEBUGP (("\nsleeping %.2f ms for %ld bytes, adjust %.2f ms\n",
114 slp, limit_data.chunk_bytes, limit_data.sleep_adjust));
116 t0 = wtimer_read (timer);
118 wtimer_update (timer);
119 t1 = wtimer_read (timer);
121 /* Due to scheduling, we probably slept slightly longer (or
122 shorter) than desired. Calculate the difference between the
123 desired and the actual sleep, and adjust the next sleep by
125 limit_data.sleep_adjust = slp - (t1 - t0);
128 limit_data.chunk_bytes = 0;
129 limit_data.chunk_start = wtimer_read (timer);
132 #define MIN(i, j) ((i) <= (j) ? (i) : (j))
134 /* Reads the contents of file descriptor FD, until it is closed, or a
135 read error occurs. The data is read in 8K chunks, and stored to
136 stream fp, which should have been open for writing. If BUF is
137 non-NULL and its file descriptor is equal to FD, flush RBUF first.
138 This function will *not* use the rbuf_* functions!
140 The EXPECTED argument is passed to show_progress() unchanged, but
143 If opt.verbose is set, the progress is also shown. RESTVAL
144 represents a value from which to start downloading (which will be
145 shown accordingly). If RESTVAL is non-zero, the stream should have
146 been open for appending.
148 The function exits and returns codes of 0, -1 and -2 if the
149 connection was closed, there was a read error, or if it could not
150 write to the output stream, respectively.
152 IMPORTANT: The function flushes the contents of the buffer in
153 rbuf_flush() before actually reading from fd. If you wish to read
154 from fd immediately, flush or discard the buffer. */
156 get_contents (int fd, FILE *fp, long *len, long restval, long expected,
157 struct rbuf *rbuf, int use_expected, double *elapsed)
161 static char dlbuf[16384];
162 int dlbufsize = sizeof (dlbuf);
164 struct wget_timer *timer = wtimer_allocate ();
165 double last_successful_read_tm;
167 /* The progress gauge, set according to the user preferences. */
168 void *progress = NULL;
170 /* Non-zero if the progress gauge is interactive, i.e. if it can
171 continually update the display. When true, smaller timeout
172 values are used so that the gauge can update the display when
173 data arrives slowly. */
174 int progress_interactive = 0;
180 progress = progress_create (restval, expected);
181 progress_interactive = progress_interactive_p (progress);
184 if (rbuf && RBUF_FD (rbuf) == fd)
187 while ((res = rbuf_flush (rbuf, dlbuf, sizeof (dlbuf))) != 0)
189 fwrite (dlbuf, 1, res, fp);
201 progress_update (progress, sz, 0);
205 limit_bandwidth_reset ();
206 wtimer_reset (timer);
207 last_successful_read_tm = 0;
209 /* Use a smaller buffer for low requested bandwidths. For example,
210 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
211 data and then sleep for 8s. With buffer size equal to the limit,
212 we never have to sleep for more than one second. */
213 if (opt.limit_rate && opt.limit_rate < dlbufsize)
214 dlbufsize = opt.limit_rate;
216 /* Read from fd while there is available data.
218 Normally, if expected is 0, it means that it is not known how
219 much data is expected. However, if use_expected is specified,
220 then expected being zero means exactly that. */
221 while (!use_expected || (*len < expected))
223 int amount_to_read = (use_expected
224 ? MIN (expected - *len, dlbufsize) : dlbufsize);
225 double tmout = opt.read_timeout;
226 if (progress_interactive)
229 /* For interactive progress gauges, always specify a ~1s
230 timeout, so that the gauge can be updated regularly even
231 when the data arrives very slowly or stalls. */
233 waittm = (wtimer_read (timer) - last_successful_read_tm) / 1000;
234 if (waittm + tmout > opt.read_timeout)
236 /* Don't allow waiting for data to exceed read timeout. */
237 tmout = opt.read_timeout - waittm;
240 /* We've already exceeded the timeout. */
247 res = xread (fd, dlbuf, amount_to_read, tmout);
249 if (res == 0 || (res < 0 && errno != ETIMEDOUT))
252 res = 0; /* timeout */
254 wtimer_update (timer);
257 fwrite (dlbuf, 1, res, fp);
258 /* Always flush the contents of the network packet. This
259 should not hinder performance: fast downloads will be
260 received in 16K chunks (which stdio would write out
261 anyway), and slow downloads won't be limited by disk
269 last_successful_read_tm = wtimer_read (timer);
273 limit_bandwidth (res, timer);
277 progress_update (progress, res, wtimer_read (timer));
279 if (use_expected && expected > 0)
280 ws_percenttitle (100.0 * (double)(*len) / (double)expected);
288 progress_finish (progress, wtimer_read (timer));
290 *elapsed = wtimer_read (timer);
291 wtimer_delete (timer);
296 /* Return a printed representation of the download rate, as
297 appropriate for the speed. If PAD is non-zero, strings will be
298 padded to the width of 7 characters (xxxx.xx). */
300 retr_rate (long bytes, double msecs, int pad)
303 static char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
306 double dlrate = calc_rate (bytes, msecs, &units);
307 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
312 /* Calculate the download rate and trim it as appropriate for the
313 speed. Appropriate means that if rate is greater than 1K/s,
314 kilobytes are used, and if rate is greater than 1MB/s, megabytes
317 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
320 calc_rate (long bytes, double msecs, int *units)
328 /* If elapsed time is exactly zero, it means we're under the
329 granularity of the timer. This often happens on systems that
330 use time() for the timer. */
331 msecs = wtimer_granularity ();
333 dlrate = (double)1000 * bytes / msecs;
336 else if (dlrate < 1024.0 * 1024.0)
337 *units = 1, dlrate /= 1024.0;
338 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
339 *units = 2, dlrate /= (1024.0 * 1024.0);
341 /* Maybe someone will need this, one day. */
342 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
347 /* Maximum number of allowed redirections. 20 was chosen as a
348 "reasonable" value, which is low enough to not cause havoc, yet
349 high enough to guarantee that normal retrievals will not be hurt by
352 #define MAX_REDIRECTIONS 20
354 #define SUSPEND_POST_DATA do { \
355 post_data_suspended = 1; \
356 saved_post_data = opt.post_data; \
357 saved_post_file_name = opt.post_file_name; \
358 opt.post_data = NULL; \
359 opt.post_file_name = NULL; \
362 #define RESTORE_POST_DATA do { \
363 if (post_data_suspended) \
365 opt.post_data = saved_post_data; \
366 opt.post_file_name = saved_post_file_name; \
367 post_data_suspended = 0; \
371 static char *getproxy PARAMS ((struct url *));
373 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
376 /* #### This function should be rewritten so it doesn't return from
380 retrieve_url (const char *origurl, char **file, char **newloc,
381 const char *refurl, int *dt)
385 int location_changed, dummy;
386 char *mynewloc, *proxy;
387 struct url *u, *proxy_url;
388 int up_error_code; /* url parse error code */
390 int redirection_count = 0;
392 int post_data_suspended = 0;
393 char *saved_post_data = NULL;
394 char *saved_post_file_name = NULL;
396 /* If dt is NULL, use local storage. */
402 url = xstrdup (origurl);
408 u = url_parse (url, &up_error_code);
411 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
417 refurl = opt.referer;
426 proxy = getproxy (u);
429 /* Parse the proxy URL. */
430 proxy_url = url_parse (proxy, &up_error_code);
433 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
434 proxy, url_error (up_error_code));
439 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
441 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
442 url_free (proxy_url);
449 if (u->scheme == SCHEME_HTTP
451 || u->scheme == SCHEME_HTTPS
453 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
455 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
457 else if (u->scheme == SCHEME_FTP)
459 /* If this is a redirection, we must not allow recursive FTP
460 retrieval, so we save recursion to oldrec, and restore it
462 int oldrec = opt.recursive;
463 if (redirection_count)
465 result = ftp_loop (u, dt, proxy_url);
466 opt.recursive = oldrec;
468 /* There is a possibility of having HTTP being redirected to
469 FTP. In these cases we must decide whether the text is HTML
470 according to the suffix. The HTML suffixes are `.html',
471 `.htm' and a few others, case-insensitive. */
472 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
474 if (has_html_suffix_p (local_file))
481 url_free (proxy_url);
485 location_changed = (result == NEWLOCATION);
486 if (location_changed)
488 char *construced_newloc;
489 struct url *newloc_parsed;
491 assert (mynewloc != NULL);
496 /* The HTTP specs only allow absolute URLs to appear in
497 redirects, but a ton of boneheaded webservers and CGIs out
498 there break the rules and use relative URLs, and popular
499 browsers are lenient about this, so wget should be too. */
500 construced_newloc = uri_merge (url, mynewloc);
502 mynewloc = construced_newloc;
504 /* Now, see if this new location makes sense. */
505 newloc_parsed = url_parse (mynewloc, &up_error_code);
508 logprintf (LOG_NOTQUIET, "%s: %s.\n", mynewloc,
509 url_error (up_error_code));
517 /* Now mynewloc will become newloc_parsed->url, because if the
518 Location contained relative paths like .././something, we
519 don't want that propagating as url. */
521 mynewloc = xstrdup (newloc_parsed->url);
523 /* Check for max. number of redirections. */
524 if (++redirection_count > MAX_REDIRECTIONS)
526 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
528 url_free (newloc_parsed);
541 /* If we're being redirected from POST, we don't want to POST
542 again. Many requests answer POST with a redirection to an
543 index page; that redirection is clearly a GET. We "suspend"
544 POST data for the duration of the redirections, and restore
545 it when we're done. */
546 if (!post_data_suspended)
556 register_download (u->url, local_file);
557 if (redirection_count && 0 != strcmp (origurl, u->url))
558 register_redirection (origurl, u->url);
560 register_html (u->url, local_file);
565 *file = local_file ? local_file : NULL;
567 xfree_null (local_file);
571 if (redirection_count)
585 ++global_download_count;
591 /* Find the URLs in the file and call retrieve_url() for each of
592 them. If HTML is non-zero, treat the file as HTML, and construct
593 the URLs accordingly.
595 If opt.recursive is set, call retrieve_tree() for each file. */
598 retrieve_from_file (const char *file, int html, int *count)
601 struct urlpos *url_list, *cur_url;
603 url_list = (html ? get_urls_html (file, NULL, NULL)
604 : get_urls_file (file));
605 status = RETROK; /* Suppose everything is OK. */
606 *count = 0; /* Reset the URL count. */
608 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
610 char *filename = NULL, *new_file = NULL;
613 if (cur_url->ignore_when_downloading)
616 if (opt.quota && total_downloaded_bytes > opt.quota)
621 if ((opt.recursive || opt.page_requisites)
622 && cur_url->url->scheme != SCHEME_FTP)
623 status = retrieve_tree (cur_url->url->url);
625 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
627 if (filename && opt.delete_after && file_exists_p (filename))
629 DEBUGP (("Removing file due to --delete-after in"
630 " retrieve_from_file():\n"));
631 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
632 if (unlink (filename))
633 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
637 xfree_null (new_file);
638 xfree_null (filename);
641 /* Free the linked list of URL-s. */
642 free_urlpos (url_list);
647 /* Print `giving up', or `retrying', depending on the impending
648 action. N1 and N2 are the attempt number and the attempt limit. */
650 printwhat (int n1, int n2)
652 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
655 /* If opt.wait or opt.waitretry are specified, and if certain
656 conditions are met, sleep the appropriate number of seconds. See
657 the documentation of --wait and --waitretry for more information.
659 COUNT is the count of current retrieval, beginning with 1. */
662 sleep_between_retrievals (int count)
664 static int first_retrieval = 1;
668 /* Don't sleep before the very first retrieval. */
673 if (opt.waitretry && count > 1)
675 /* If opt.waitretry is specified and this is a retry, wait for
676 COUNT-1 number of seconds, or for opt.waitretry seconds. */
677 if (count <= opt.waitretry)
680 xsleep (opt.waitretry);
684 if (!opt.random_wait || count > 1)
685 /* If random-wait is not specified, or if we are sleeping
686 between retries of the same download, sleep the fixed
691 /* Sleep a random amount of time averaging in opt.wait
692 seconds. The sleeping amount ranges from 0 to
693 opt.wait*2, inclusive. */
694 double waitsecs = 2 * opt.wait * random_float ();
695 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
696 opt.wait, waitsecs));
702 /* Free the linked list of urlpos. */
704 free_urlpos (struct urlpos *l)
708 struct urlpos *next = l->next;
711 xfree_null (l->local_name);
717 /* Rotate FNAME opt.backups times */
719 rotate_backups(const char *fname)
721 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
722 char *from = (char *)alloca (maxlen);
723 char *to = (char *)alloca (maxlen);
727 if (stat (fname, &sb) == 0)
728 if (S_ISREG (sb.st_mode) == 0)
731 for (i = opt.backups; i > 1; i--)
733 sprintf (from, "%s.%d", fname, i - 1);
734 sprintf (to, "%s.%d", fname, i);
738 sprintf (to, "%s.%d", fname, 1);
742 static int no_proxy_match PARAMS ((const char *, const char **));
744 /* Return the URL of the proxy appropriate for url U. */
747 getproxy (struct url *u)
751 static char rewritten_storage[1024];
755 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
761 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
765 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
769 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
774 if (!proxy || !*proxy)
777 /* Handle shorthands. `rewritten_storage' is a kludge to allow
778 getproxy() to return static storage. */
779 rewritten_url = rewrite_shorthand_url (proxy);
782 strncpy (rewritten_storage, rewritten_url, sizeof(rewritten_storage));
783 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
784 proxy = rewritten_storage;
790 /* Should a host be accessed through proxy, concerning no_proxy? */
792 no_proxy_match (const char *host, const char **no_proxy)
797 return !sufmatch (no_proxy, host);