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;
77 limit_bandwidth_reset (void)
79 limit_data.chunk_bytes = 0;
80 limit_data.chunk_start = 0;
83 /* Limit the bandwidth by pausing the download for an amount of time.
84 BYTES is the number of bytes received from the network, and DELTA
85 is the number of milliseconds it took to receive them. */
88 limit_bandwidth (long bytes, double *dltime, struct wget_timer *timer)
90 double delta_t = *dltime - limit_data.chunk_start;
93 limit_data.chunk_bytes += bytes;
95 /* Calculate the amount of time we expect downloading the chunk
96 should take. If in reality it took less time, sleep to
97 compensate for the difference. */
98 expected = 1000.0 * limit_data.chunk_bytes / opt.limit_rate;
100 if (expected > delta_t)
102 double slp = expected - delta_t + limit_data.sleep_adjust;
106 DEBUGP (("deferring a %.2f ms sleep (%ld/%.2f).\n",
107 slp, limit_data.chunk_bytes, delta_t));
110 DEBUGP (("\nsleeping %.2f ms for %ld bytes, adjust %.2f ms\n",
111 slp, limit_data.chunk_bytes, limit_data.sleep_adjust));
114 usleep ((unsigned long) (1000 * slp));
115 t1 = wtimer_elapsed (timer);
117 /* Due to scheduling, we probably slept slightly longer (or
118 shorter) than desired. Calculate the difference between the
119 desired and the actual sleep, and adjust the next sleep by
121 limit_data.sleep_adjust = slp - (t1 - t0);
123 /* Since we've called wtimer_elapsed, we might as well update
124 the caller's dltime. */
128 limit_data.chunk_bytes = 0;
129 limit_data.chunk_start = *dltime;
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 void *progress = NULL;
165 struct wget_timer *timer = wtimer_allocate ();
171 progress = progress_create (restval, expected);
173 if (rbuf && RBUF_FD (rbuf) == fd)
176 while ((res = rbuf_flush (rbuf, dlbuf, sizeof (dlbuf))) != 0)
178 fwrite (dlbuf, 1, res, fp);
190 progress_update (progress, sz, 0);
194 limit_bandwidth_reset ();
195 wtimer_reset (timer);
197 /* Use a smaller buffer for low requested bandwidths. For example,
198 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
199 data and then sleep for 8s. With buffer size equal to the limit,
200 we never have to sleep for more than one second. */
201 if (opt.limit_rate && opt.limit_rate < dlbufsize)
202 dlbufsize = opt.limit_rate;
204 /* Read from fd while there is available data.
206 Normally, if expected is 0, it means that it is not known how
207 much data is expected. However, if use_expected is specified,
208 then expected being zero means exactly that. */
209 while (!use_expected || (*len < expected))
211 int amount_to_read = (use_expected
212 ? MIN (expected - *len, dlbufsize) : dlbufsize);
215 res = ssl_iread (rbuf->ssl, dlbuf, amount_to_read);
217 #endif /* HAVE_SSL */
218 res = iread (fd, dlbuf, amount_to_read);
223 fwrite (dlbuf, 1, res, fp);
224 /* Always flush the contents of the network packet. This should
225 not hinder performance: fast downloads will be received in
226 16K chunks (which stdio would write out anyway), and slow
227 downloads won't be limited with disk performance. */
235 dltime = wtimer_elapsed (timer);
237 limit_bandwidth (res, &dltime, timer);
240 progress_update (progress, res, dltime);
248 progress_finish (progress, dltime);
251 wtimer_delete (timer);
256 /* Return a printed representation of the download rate, as
257 appropriate for the speed. If PAD is non-zero, strings will be
258 padded to the width of 7 characters (xxxx.xx). */
260 retr_rate (long bytes, double msecs, int pad)
263 static char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
266 double dlrate = calc_rate (bytes, msecs, &units);
267 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
272 /* Calculate the download rate and trim it as appropriate for the
273 speed. Appropriate means that if rate is greater than 1K/s,
274 kilobytes are used, and if rate is greater than 1MB/s, megabytes
277 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
280 calc_rate (long bytes, double msecs, int *units)
288 /* If elapsed time is exactly zero, it means we're under the
289 granularity of the timer. This often happens on systems that
290 use time() for the timer. */
291 msecs = wtimer_granularity ();
293 dlrate = (double)1000 * bytes / msecs;
296 else if (dlrate < 1024.0 * 1024.0)
297 *units = 1, dlrate /= 1024.0;
298 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
299 *units = 2, dlrate /= (1024.0 * 1024.0);
301 /* Maybe someone will need this, one day. */
302 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
307 /* Maximum number of allowed redirections. 20 was chosen as a
308 "reasonable" value, which is low enough to not cause havoc, yet
309 high enough to guarantee that normal retrievals will not be hurt by
312 #define MAX_REDIRECTIONS 20
314 #define SUSPEND_POST_DATA do { \
315 post_data_suspended = 1; \
316 saved_post_data = opt.post_data; \
317 saved_post_file_name = opt.post_file_name; \
318 opt.post_data = NULL; \
319 opt.post_file_name = NULL; \
322 #define RESTORE_POST_DATA do { \
323 if (post_data_suspended) \
325 opt.post_data = saved_post_data; \
326 opt.post_file_name = saved_post_file_name; \
327 post_data_suspended = 0; \
331 static char *getproxy PARAMS ((struct url *));
333 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
336 /* #### This function should be rewritten so it doesn't return from
340 retrieve_url (const char *origurl, char **file, char **newloc,
341 const char *refurl, int *dt)
345 int location_changed, dummy;
346 char *mynewloc, *proxy;
347 struct url *u, *proxy_url;
348 int up_error_code; /* url parse error code */
350 int redirection_count = 0;
352 int post_data_suspended = 0;
353 char *saved_post_data = NULL;
354 char *saved_post_file_name = NULL;
356 /* If dt is NULL, just ignore it. */
359 url = xstrdup (origurl);
365 u = url_parse (url, &up_error_code);
368 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
374 refurl = opt.referer;
383 proxy = getproxy (u);
386 /* Parse the proxy URL. */
387 proxy_url = url_parse (proxy, &up_error_code);
390 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
391 proxy, url_error (up_error_code));
396 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
398 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
399 url_free (proxy_url);
406 if (u->scheme == SCHEME_HTTP
408 || u->scheme == SCHEME_HTTPS
410 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
412 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
414 else if (u->scheme == SCHEME_FTP)
416 /* If this is a redirection, we must not allow recursive FTP
417 retrieval, so we save recursion to oldrec, and restore it
419 int oldrec = opt.recursive;
420 if (redirection_count)
422 result = ftp_loop (u, dt, proxy_url);
423 opt.recursive = oldrec;
425 /* There is a possibility of having HTTP being redirected to
426 FTP. In these cases we must decide whether the text is HTML
427 according to the suffix. The HTML suffixes are `.html',
428 `.htm' and a few others, case-insensitive. */
429 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
431 if (has_html_suffix_p (local_file))
438 url_free (proxy_url);
442 location_changed = (result == NEWLOCATION);
443 if (location_changed)
445 char *construced_newloc;
446 struct url *newloc_parsed;
448 assert (mynewloc != NULL);
453 /* The HTTP specs only allow absolute URLs to appear in
454 redirects, but a ton of boneheaded webservers and CGIs out
455 there break the rules and use relative URLs, and popular
456 browsers are lenient about this, so wget should be too. */
457 construced_newloc = uri_merge (url, mynewloc);
459 mynewloc = construced_newloc;
461 /* Now, see if this new location makes sense. */
462 newloc_parsed = url_parse (mynewloc, &up_error_code);
465 logprintf (LOG_NOTQUIET, "%s: %s.\n", mynewloc,
466 url_error (up_error_code));
474 /* Now mynewloc will become newloc_parsed->url, because if the
475 Location contained relative paths like .././something, we
476 don't want that propagating as url. */
478 mynewloc = xstrdup (newloc_parsed->url);
480 /* Check for max. number of redirections. */
481 if (++redirection_count > MAX_REDIRECTIONS)
483 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
485 url_free (newloc_parsed);
498 /* If we're being redirected from POST, we don't want to POST
499 again. Many requests answer POST with a redirection to an
500 index page; that redirection is clearly a GET. We "suspend"
501 POST data for the duration of the redirections, and restore
502 it when we're done. */
503 if (!post_data_suspended)
513 register_download (u->url, local_file);
514 if (redirection_count && 0 != strcmp (origurl, u->url))
515 register_redirection (origurl, u->url);
517 register_html (u->url, local_file);
522 *file = local_file ? local_file : NULL;
524 FREE_MAYBE (local_file);
528 if (redirection_count)
542 ++global_download_count;
548 /* Find the URLs in the file and call retrieve_url() for each of
549 them. If HTML is non-zero, treat the file as HTML, and construct
550 the URLs accordingly.
552 If opt.recursive is set, call recursive_retrieve() for each file. */
554 retrieve_from_file (const char *file, int html, int *count)
557 struct urlpos *url_list, *cur_url;
559 url_list = (html ? get_urls_html (file, NULL, NULL)
560 : get_urls_file (file));
561 status = RETROK; /* Suppose everything is OK. */
562 *count = 0; /* Reset the URL count. */
564 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
566 char *filename = NULL, *new_file = NULL;
569 if (cur_url->ignore_when_downloading)
572 if (downloaded_exceeds_quota ())
577 if (opt.recursive && cur_url->url->scheme != SCHEME_FTP)
578 status = retrieve_tree (cur_url->url->url);
580 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
582 if (filename && opt.delete_after && file_exists_p (filename))
584 DEBUGP (("Removing file due to --delete-after in"
585 " retrieve_from_file():\n"));
586 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
587 if (unlink (filename))
588 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
592 FREE_MAYBE (new_file);
593 FREE_MAYBE (filename);
596 /* Free the linked list of URL-s. */
597 free_urlpos (url_list);
602 /* Print `giving up', or `retrying', depending on the impending
603 action. N1 and N2 are the attempt number and the attempt limit. */
605 printwhat (int n1, int n2)
607 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
610 /* Increment opt.downloaded by BY_HOW_MUCH. If an overflow occurs,
611 set opt.downloaded_overflow to 1. */
613 downloaded_increase (unsigned long by_how_much)
616 if (opt.downloaded_overflow)
618 old = opt.downloaded;
619 opt.downloaded += by_how_much;
620 if (opt.downloaded < old) /* carry flag, where are you when I
624 opt.downloaded_overflow = 1;
625 opt.downloaded = ~((VERY_LONG_TYPE)0);
629 /* Return non-zero if the downloaded amount of bytes exceeds the
630 desired quota. If quota is not set or if the amount overflowed, 0
633 downloaded_exceeds_quota (void)
637 if (opt.downloaded_overflow)
638 /* We don't really know. (Wildly) assume not. */
641 return opt.downloaded > opt.quota;
644 /* If opt.wait or opt.waitretry are specified, and if certain
645 conditions are met, sleep the appropriate number of seconds. See
646 the documentation of --wait and --waitretry for more information.
648 COUNT is the count of current retrieval, beginning with 1. */
651 sleep_between_retrievals (int count)
653 static int first_retrieval = 1;
657 /* Don't sleep before the very first retrieval. */
662 if (opt.waitretry && count > 1)
664 /* If opt.waitretry is specified and this is a retry, wait for
665 COUNT-1 number of seconds, or for opt.waitretry seconds. */
666 if (count <= opt.waitretry)
669 usleep (1000000L * opt.waitretry);
673 if (!opt.random_wait || count > 1)
674 /* If random-wait is not specified, or if we are sleeping
675 between retries of the same download, sleep the fixed
677 usleep (1000000L * opt.wait);
680 /* Sleep a random amount of time averaging in opt.wait
681 seconds. The sleeping amount ranges from 0 to
682 opt.wait*2, inclusive. */
683 double waitsecs = 2 * opt.wait * random_float ();
684 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
685 opt.wait, waitsecs));
686 usleep (1000000L * waitsecs);
691 /* Free the linked list of urlpos. */
693 free_urlpos (struct urlpos *l)
697 struct urlpos *next = l->next;
700 FREE_MAYBE (l->local_name);
706 /* Rotate FNAME opt.backups times */
708 rotate_backups(const char *fname)
710 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
711 char *from = (char *)alloca (maxlen);
712 char *to = (char *)alloca (maxlen);
716 if (stat (fname, &sb) == 0)
717 if (S_ISREG (sb.st_mode) == 0)
720 for (i = opt.backups; i > 1; i--)
722 sprintf (from, "%s.%d", fname, i - 1);
723 sprintf (to, "%s.%d", fname, i);
727 sprintf (to, "%s.%d", fname, 1);
731 static int no_proxy_match PARAMS ((const char *, const char **));
733 /* Return the URL of the proxy appropriate for url U. */
736 getproxy (struct url *u)
740 static char rewritten_storage[1024];
744 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
750 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
754 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
758 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
763 if (!proxy || !*proxy)
766 /* Handle shorthands. `rewritten_storage' is a kludge to allow
767 getproxy() to return static storage. */
768 rewritten_url = rewrite_shorthand_url (proxy);
771 strncpy (rewritten_storage, rewritten_url, sizeof(rewritten_storage));
772 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
773 proxy = rewritten_storage;
779 /* Should a host be accessed through proxy, concerning no_proxy? */
781 no_proxy_match (const char *host, const char **no_proxy)
786 return !sufmatch (no_proxy, host);