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 DELTA
88 is the number of milliseconds it took to receive them. */
91 limit_bandwidth (long bytes, double *dltime, struct wget_timer *timer)
93 double delta_t = *dltime - 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));
118 t1 = wtimer_elapsed (timer);
120 /* Due to scheduling, we probably slept slightly longer (or
121 shorter) than desired. Calculate the difference between the
122 desired and the actual sleep, and adjust the next sleep by
124 limit_data.sleep_adjust = slp - (t1 - t0);
126 /* Since we've called wtimer_elapsed, we might as well update
127 the caller's dltime. */
131 limit_data.chunk_bytes = 0;
132 limit_data.chunk_start = *dltime;
135 #define MIN(i, j) ((i) <= (j) ? (i) : (j))
137 /* Reads the contents of file descriptor FD, until it is closed, or a
138 read error occurs. The data is read in 8K chunks, and stored to
139 stream fp, which should have been open for writing. If BUF is
140 non-NULL and its file descriptor is equal to FD, flush RBUF first.
141 This function will *not* use the rbuf_* functions!
143 The EXPECTED argument is passed to show_progress() unchanged, but
146 If opt.verbose is set, the progress is also shown. RESTVAL
147 represents a value from which to start downloading (which will be
148 shown accordingly). If RESTVAL is non-zero, the stream should have
149 been open for appending.
151 The function exits and returns codes of 0, -1 and -2 if the
152 connection was closed, there was a read error, or if it could not
153 write to the output stream, respectively.
155 IMPORTANT: The function flushes the contents of the buffer in
156 rbuf_flush() before actually reading from fd. If you wish to read
157 from fd immediately, flush or discard the buffer. */
159 get_contents (int fd, FILE *fp, long *len, long restval, long expected,
160 struct rbuf *rbuf, int use_expected, double *elapsed)
164 static char dlbuf[16384];
165 int dlbufsize = sizeof (dlbuf);
167 void *progress = NULL;
168 struct wget_timer *timer = wtimer_allocate ();
174 progress = progress_create (restval, expected);
176 if (rbuf && RBUF_FD (rbuf) == fd)
179 while ((res = rbuf_flush (rbuf, dlbuf, sizeof (dlbuf))) != 0)
181 fwrite (dlbuf, 1, res, fp);
193 progress_update (progress, sz, 0);
197 limit_bandwidth_reset ();
198 wtimer_reset (timer);
200 /* Use a smaller buffer for low requested bandwidths. For example,
201 with --limit-rate=2k, it doesn't make sense to slurp in 16K of
202 data and then sleep for 8s. With buffer size equal to the limit,
203 we never have to sleep for more than one second. */
204 if (opt.limit_rate && opt.limit_rate < dlbufsize)
205 dlbufsize = opt.limit_rate;
207 /* Read from fd while there is available data.
209 Normally, if expected is 0, it means that it is not known how
210 much data is expected. However, if use_expected is specified,
211 then expected being zero means exactly that. */
212 while (!use_expected || (*len < expected))
214 int amount_to_read = (use_expected
215 ? MIN (expected - *len, dlbufsize) : dlbufsize);
216 res = xread (fd, dlbuf, amount_to_read, -1);
221 fwrite (dlbuf, 1, res, fp);
222 /* Always flush the contents of the network packet. This should
223 not hinder performance: fast downloads will be received in
224 16K chunks (which stdio would write out anyway), and slow
225 downloads won't be limited with disk performance. */
233 dltime = wtimer_elapsed (timer);
235 limit_bandwidth (res, &dltime, timer);
239 progress_update (progress, res, dltime);
241 if (use_expected && expected > 0)
242 ws_percenttitle (100.0 * (double)(*len) / (double)expected);
250 progress_finish (progress, dltime);
253 wtimer_delete (timer);
258 /* Return a printed representation of the download rate, as
259 appropriate for the speed. If PAD is non-zero, strings will be
260 padded to the width of 7 characters (xxxx.xx). */
262 retr_rate (long bytes, double msecs, int pad)
265 static char *rate_names[] = {"B/s", "KB/s", "MB/s", "GB/s" };
268 double dlrate = calc_rate (bytes, msecs, &units);
269 sprintf (res, pad ? "%7.2f %s" : "%.2f %s", dlrate, rate_names[units]);
274 /* Calculate the download rate and trim it as appropriate for the
275 speed. Appropriate means that if rate is greater than 1K/s,
276 kilobytes are used, and if rate is greater than 1MB/s, megabytes
279 UNITS is zero for B/s, one for KB/s, two for MB/s, and three for
282 calc_rate (long bytes, double msecs, int *units)
290 /* If elapsed time is exactly zero, it means we're under the
291 granularity of the timer. This often happens on systems that
292 use time() for the timer. */
293 msecs = wtimer_granularity ();
295 dlrate = (double)1000 * bytes / msecs;
298 else if (dlrate < 1024.0 * 1024.0)
299 *units = 1, dlrate /= 1024.0;
300 else if (dlrate < 1024.0 * 1024.0 * 1024.0)
301 *units = 2, dlrate /= (1024.0 * 1024.0);
303 /* Maybe someone will need this, one day. */
304 *units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
309 /* Maximum number of allowed redirections. 20 was chosen as a
310 "reasonable" value, which is low enough to not cause havoc, yet
311 high enough to guarantee that normal retrievals will not be hurt by
314 #define MAX_REDIRECTIONS 20
316 #define SUSPEND_POST_DATA do { \
317 post_data_suspended = 1; \
318 saved_post_data = opt.post_data; \
319 saved_post_file_name = opt.post_file_name; \
320 opt.post_data = NULL; \
321 opt.post_file_name = NULL; \
324 #define RESTORE_POST_DATA do { \
325 if (post_data_suspended) \
327 opt.post_data = saved_post_data; \
328 opt.post_file_name = saved_post_file_name; \
329 post_data_suspended = 0; \
333 static char *getproxy PARAMS ((struct url *));
335 /* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
338 /* #### This function should be rewritten so it doesn't return from
342 retrieve_url (const char *origurl, char **file, char **newloc,
343 const char *refurl, int *dt)
347 int location_changed, dummy;
348 char *mynewloc, *proxy;
349 struct url *u, *proxy_url;
350 int up_error_code; /* url parse error code */
352 int redirection_count = 0;
354 int post_data_suspended = 0;
355 char *saved_post_data = NULL;
356 char *saved_post_file_name = NULL;
358 /* If dt is NULL, use local storage. */
364 url = xstrdup (origurl);
370 u = url_parse (url, &up_error_code);
373 logprintf (LOG_NOTQUIET, "%s: %s.\n", url, url_error (up_error_code));
379 refurl = opt.referer;
388 proxy = getproxy (u);
391 /* Parse the proxy URL. */
392 proxy_url = url_parse (proxy, &up_error_code);
395 logprintf (LOG_NOTQUIET, _("Error parsing proxy URL %s: %s.\n"),
396 proxy, url_error (up_error_code));
401 if (proxy_url->scheme != SCHEME_HTTP && proxy_url->scheme != u->scheme)
403 logprintf (LOG_NOTQUIET, _("Error in proxy URL %s: Must be HTTP.\n"), proxy);
404 url_free (proxy_url);
411 if (u->scheme == SCHEME_HTTP
413 || u->scheme == SCHEME_HTTPS
415 || (proxy_url && proxy_url->scheme == SCHEME_HTTP))
417 result = http_loop (u, &mynewloc, &local_file, refurl, dt, proxy_url);
419 else if (u->scheme == SCHEME_FTP)
421 /* If this is a redirection, we must not allow recursive FTP
422 retrieval, so we save recursion to oldrec, and restore it
424 int oldrec = opt.recursive;
425 if (redirection_count)
427 result = ftp_loop (u, dt, proxy_url);
428 opt.recursive = oldrec;
430 /* There is a possibility of having HTTP being redirected to
431 FTP. In these cases we must decide whether the text is HTML
432 according to the suffix. The HTML suffixes are `.html',
433 `.htm' and a few others, case-insensitive. */
434 if (redirection_count && local_file && u->scheme == SCHEME_FTP)
436 if (has_html_suffix_p (local_file))
443 url_free (proxy_url);
447 location_changed = (result == NEWLOCATION);
448 if (location_changed)
450 char *construced_newloc;
451 struct url *newloc_parsed;
453 assert (mynewloc != NULL);
458 /* The HTTP specs only allow absolute URLs to appear in
459 redirects, but a ton of boneheaded webservers and CGIs out
460 there break the rules and use relative URLs, and popular
461 browsers are lenient about this, so wget should be too. */
462 construced_newloc = uri_merge (url, mynewloc);
464 mynewloc = construced_newloc;
466 /* Now, see if this new location makes sense. */
467 newloc_parsed = url_parse (mynewloc, &up_error_code);
470 logprintf (LOG_NOTQUIET, "%s: %s.\n", mynewloc,
471 url_error (up_error_code));
479 /* Now mynewloc will become newloc_parsed->url, because if the
480 Location contained relative paths like .././something, we
481 don't want that propagating as url. */
483 mynewloc = xstrdup (newloc_parsed->url);
485 /* Check for max. number of redirections. */
486 if (++redirection_count > MAX_REDIRECTIONS)
488 logprintf (LOG_NOTQUIET, _("%d redirections exceeded.\n"),
490 url_free (newloc_parsed);
503 /* If we're being redirected from POST, we don't want to POST
504 again. Many requests answer POST with a redirection to an
505 index page; that redirection is clearly a GET. We "suspend"
506 POST data for the duration of the redirections, and restore
507 it when we're done. */
508 if (!post_data_suspended)
518 register_download (u->url, local_file);
519 if (redirection_count && 0 != strcmp (origurl, u->url))
520 register_redirection (origurl, u->url);
522 register_html (u->url, local_file);
527 *file = local_file ? local_file : NULL;
529 xfree_null (local_file);
533 if (redirection_count)
547 ++global_download_count;
553 /* Find the URLs in the file and call retrieve_url() for each of
554 them. If HTML is non-zero, treat the file as HTML, and construct
555 the URLs accordingly.
557 If opt.recursive is set, call retrieve_tree() for each file. */
560 retrieve_from_file (const char *file, int html, int *count)
563 struct urlpos *url_list, *cur_url;
565 url_list = (html ? get_urls_html (file, NULL, NULL)
566 : get_urls_file (file));
567 status = RETROK; /* Suppose everything is OK. */
568 *count = 0; /* Reset the URL count. */
570 for (cur_url = url_list; cur_url; cur_url = cur_url->next, ++*count)
572 char *filename = NULL, *new_file = NULL;
575 if (cur_url->ignore_when_downloading)
578 if (opt.quota && total_downloaded_bytes > opt.quota)
583 if ((opt.recursive || opt.page_requisites)
584 && cur_url->url->scheme != SCHEME_FTP)
585 status = retrieve_tree (cur_url->url->url);
587 status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
589 if (filename && opt.delete_after && file_exists_p (filename))
591 DEBUGP (("Removing file due to --delete-after in"
592 " retrieve_from_file():\n"));
593 logprintf (LOG_VERBOSE, _("Removing %s.\n"), filename);
594 if (unlink (filename))
595 logprintf (LOG_NOTQUIET, "unlink: %s\n", strerror (errno));
599 xfree_null (new_file);
600 xfree_null (filename);
603 /* Free the linked list of URL-s. */
604 free_urlpos (url_list);
609 /* Print `giving up', or `retrying', depending on the impending
610 action. N1 and N2 are the attempt number and the attempt limit. */
612 printwhat (int n1, int n2)
614 logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
617 /* If opt.wait or opt.waitretry are specified, and if certain
618 conditions are met, sleep the appropriate number of seconds. See
619 the documentation of --wait and --waitretry for more information.
621 COUNT is the count of current retrieval, beginning with 1. */
624 sleep_between_retrievals (int count)
626 static int first_retrieval = 1;
630 /* Don't sleep before the very first retrieval. */
635 if (opt.waitretry && count > 1)
637 /* If opt.waitretry is specified and this is a retry, wait for
638 COUNT-1 number of seconds, or for opt.waitretry seconds. */
639 if (count <= opt.waitretry)
642 xsleep (opt.waitretry);
646 if (!opt.random_wait || count > 1)
647 /* If random-wait is not specified, or if we are sleeping
648 between retries of the same download, sleep the fixed
653 /* Sleep a random amount of time averaging in opt.wait
654 seconds. The sleeping amount ranges from 0 to
655 opt.wait*2, inclusive. */
656 double waitsecs = 2 * opt.wait * random_float ();
657 DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
658 opt.wait, waitsecs));
664 /* Free the linked list of urlpos. */
666 free_urlpos (struct urlpos *l)
670 struct urlpos *next = l->next;
673 xfree_null (l->local_name);
679 /* Rotate FNAME opt.backups times */
681 rotate_backups(const char *fname)
683 int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
684 char *from = (char *)alloca (maxlen);
685 char *to = (char *)alloca (maxlen);
689 if (stat (fname, &sb) == 0)
690 if (S_ISREG (sb.st_mode) == 0)
693 for (i = opt.backups; i > 1; i--)
695 sprintf (from, "%s.%d", fname, i - 1);
696 sprintf (to, "%s.%d", fname, i);
700 sprintf (to, "%s.%d", fname, 1);
704 static int no_proxy_match PARAMS ((const char *, const char **));
706 /* Return the URL of the proxy appropriate for url U. */
709 getproxy (struct url *u)
713 static char rewritten_storage[1024];
717 if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
723 proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
727 proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
731 proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
736 if (!proxy || !*proxy)
739 /* Handle shorthands. `rewritten_storage' is a kludge to allow
740 getproxy() to return static storage. */
741 rewritten_url = rewrite_shorthand_url (proxy);
744 strncpy (rewritten_storage, rewritten_url, sizeof(rewritten_storage));
745 rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
746 proxy = rewritten_storage;
752 /* Should a host be accessed through proxy, concerning no_proxy? */
754 no_proxy_match (const char *host, const char **no_proxy)
759 return !sufmatch (no_proxy, host);