#include "host.h"
#include "connect.h"
#include "hash.h"
+#include "convert.h"
#ifdef HAVE_SSL
# include "gen_sslfunc.h" /* for ssl_iread */
/* See the comment in gethttp() why this is needed. */
int global_download_count;
+/* Total size of downloaded files. Used to enforce quota. */
+LARGE_INT total_downloaded_bytes;
+
\f
static struct {
- long bytes;
- double dltime;
+ long chunk_bytes;
+ double chunk_start;
+ double sleep_adjust;
} limit_data;
static void
limit_bandwidth_reset (void)
{
- limit_data.bytes = 0;
- limit_data.dltime = 0;
+ limit_data.chunk_bytes = 0;
+ limit_data.chunk_start = 0;
}
/* Limit the bandwidth by pausing the download for an amount of time.
- BYTES is the number of bytes received from the network, DELTA is
- how long it took to receive them, DLTIME the current download time,
- TIMER the timer, and ADJUSTMENT the previous. */
+ BYTES is the number of bytes received from the network, and DELTA
+ is the number of milliseconds it took to receive them. */
static void
-limit_bandwidth (long bytes, double delta)
+limit_bandwidth (long bytes, double *dltime, struct wget_timer *timer)
{
+ double delta_t = *dltime - limit_data.chunk_start;
double expected;
- limit_data.bytes += bytes;
- limit_data.dltime += delta;
+ limit_data.chunk_bytes += bytes;
- expected = 1000.0 * limit_data.bytes / opt.limit_rate;
+ /* Calculate the amount of time we expect downloading the chunk
+ should take. If in reality it took less time, sleep to
+ compensate for the difference. */
+ expected = 1000.0 * limit_data.chunk_bytes / opt.limit_rate;
- if (expected > limit_data.dltime)
+ if (expected > delta_t)
{
- double slp = expected - limit_data.dltime;
+ double slp = expected - delta_t + limit_data.sleep_adjust;
+ double t0, t1;
if (slp < 200)
{
DEBUGP (("deferring a %.2f ms sleep (%ld/%.2f).\n",
- slp, limit_data.bytes, limit_data.dltime));
+ slp, limit_data.chunk_bytes, delta_t));
return;
}
- DEBUGP (("sleeping %.2f ms\n", slp));
+ DEBUGP (("\nsleeping %.2f ms for %ld bytes, adjust %.2f ms\n",
+ slp, limit_data.chunk_bytes, limit_data.sleep_adjust));
+
+ t0 = *dltime;
usleep ((unsigned long) (1000 * slp));
+ t1 = wtimer_elapsed (timer);
+
+ /* Due to scheduling, we probably slept slightly longer (or
+ shorter) than desired. Calculate the difference between the
+ desired and the actual sleep, and adjust the next sleep by
+ that amount. */
+ limit_data.sleep_adjust = slp - (t1 - t0);
+
+ /* Since we've called wtimer_elapsed, we might as well update
+ the caller's dltime. */
+ *dltime = t1;
}
- limit_data.bytes = 0;
- limit_data.dltime = 0;
+ limit_data.chunk_bytes = 0;
+ limit_data.chunk_start = *dltime;
}
#define MIN(i, j) ((i) <= (j) ? (i) : (j))
struct rbuf *rbuf, int use_expected, double *elapsed)
{
int res = 0;
- static char c[16384];
+
+ static char dlbuf[16384];
+ int dlbufsize = sizeof (dlbuf);
+
void *progress = NULL;
struct wget_timer *timer = wtimer_allocate ();
- double dltime = 0, last_dltime = 0;
+ double dltime = 0;
*len = restval;
if (rbuf && RBUF_FD (rbuf) == fd)
{
int sz = 0;
- while ((res = rbuf_flush (rbuf, c, sizeof (c))) != 0)
+ while ((res = rbuf_flush (rbuf, dlbuf, sizeof (dlbuf))) != 0)
{
- fwrite (c, sizeof (char), res, fp);
+ fwrite (dlbuf, 1, res, fp);
*len += res;
sz += res;
}
limit_bandwidth_reset ();
wtimer_reset (timer);
+ /* Use a smaller buffer for low requested bandwidths. For example,
+ with --limit-rate=2k, it doesn't make sense to slurp in 16K of
+ data and then sleep for 8s. With buffer size equal to the limit,
+ we never have to sleep for more than one second. */
+ if (opt.limit_rate && opt.limit_rate < dlbufsize)
+ dlbufsize = opt.limit_rate;
+
/* Read from fd while there is available data.
Normally, if expected is 0, it means that it is not known how
while (!use_expected || (*len < expected))
{
int amount_to_read = (use_expected
- ? MIN (expected - *len, sizeof (c))
- : sizeof (c));
+ ? MIN (expected - *len, dlbufsize) : dlbufsize);
#ifdef HAVE_SSL
if (rbuf->ssl!=NULL)
- res = ssl_iread (rbuf->ssl, c, amount_to_read);
+ res = ssl_iread (rbuf->ssl, dlbuf, amount_to_read);
else
#endif /* HAVE_SSL */
- res = iread (fd, c, amount_to_read);
+ res = iread (fd, dlbuf, amount_to_read);
+
+ if (res <= 0)
+ break;
- if (res > 0)
+ fwrite (dlbuf, 1, res, fp);
+ /* Always flush the contents of the network packet. This should
+ not hinder performance: fast downloads will be received in
+ 16K chunks (which stdio would write out anyway), and slow
+ downloads won't be limited with disk performance. */
+ fflush (fp);
+ if (ferror (fp))
{
- fwrite (c, sizeof (char), res, fp);
- /* Always flush the contents of the network packet. This
- should not be adverse to performance, as the network
- packets typically won't be too tiny anyway. */
- fflush (fp);
- if (ferror (fp))
- {
- res = -2;
- goto out;
- }
-
- /* If bandwidth is not limited, one call to wtimer_elapsed
- is sufficient. */
- dltime = wtimer_elapsed (timer);
- if (opt.limit_rate)
- {
- limit_bandwidth (res, dltime - last_dltime);
- dltime = wtimer_elapsed (timer);
- last_dltime = dltime;
- }
-
- if (progress)
- progress_update (progress, res, dltime);
- *len += res;
+ res = -2;
+ goto out;
}
- else
- break;
+
+ dltime = wtimer_elapsed (timer);
+ if (opt.limit_rate)
+ limit_bandwidth (res, &dltime, timer);
+
+ *len += res;
+ if (progress)
+ progress_update (progress, res, dltime);
+#ifdef WINDOWS
+ if (use_expected && expected > 0)
+ ws_percenttitle (100.0 * (double)(*len) / (double)expected);
+#endif
}
if (res < -1)
res = -1;
else if (dlrate < 1024.0 * 1024.0 * 1024.0)
*units = 2, dlrate /= (1024.0 * 1024.0);
else
- /* Maybe someone will need this one day. More realistically, it
- will get tickled by buggy timers. */
+ /* Maybe someone will need this, one day. */
*units = 3, dlrate /= (1024.0 * 1024.0 * 1024.0);
return dlrate;
} \
} while (0)
+static char *getproxy PARAMS ((struct url *));
+
/* Retrieve the given URL. Decides which loop to call -- HTTP, FTP,
FTP, proxy, etc. */
char *saved_post_data = NULL;
char *saved_post_file_name = NULL;
- /* If dt is NULL, just ignore it. */
+ /* If dt is NULL, use local storage. */
if (!dt)
- dt = &dummy;
+ {
+ dt = &dummy;
+ dummy = 0;
+ }
url = xstrdup (origurl);
if (newloc)
*newloc = NULL;
if (file)
*file = local_file ? local_file : NULL;
else
- FREE_MAYBE (local_file);
+ xfree_null (local_file);
url_free (u);
them. If HTML is non-zero, treat the file as HTML, and construct
the URLs accordingly.
- If opt.recursive is set, call recursive_retrieve() for each file. */
+ If opt.recursive is set, call retrieve_tree() for each file. */
+
uerr_t
retrieve_from_file (const char *file, int html, int *count)
{
if (cur_url->ignore_when_downloading)
continue;
- if (downloaded_exceeds_quota ())
+ if (opt.quota && total_downloaded_bytes > opt.quota)
{
status = QUOTEXC;
break;
}
- if (opt.recursive && cur_url->url->scheme != SCHEME_FTP)
+ if ((opt.recursive || opt.page_requisites)
+ && cur_url->url->scheme != SCHEME_FTP)
status = retrieve_tree (cur_url->url->url);
else
status = retrieve_url (cur_url->url->url, &filename, &new_file, NULL, &dt);
dt &= ~RETROKF;
}
- FREE_MAYBE (new_file);
- FREE_MAYBE (filename);
+ xfree_null (new_file);
+ xfree_null (filename);
}
/* Free the linked list of URL-s. */
logputs (LOG_VERBOSE, (n1 == n2) ? _("Giving up.\n\n") : _("Retrying.\n\n"));
}
-/* Increment opt.downloaded by BY_HOW_MUCH. If an overflow occurs,
- set opt.downloaded_overflow to 1. */
-void
-downloaded_increase (unsigned long by_how_much)
-{
- VERY_LONG_TYPE old;
- if (opt.downloaded_overflow)
- return;
- old = opt.downloaded;
- opt.downloaded += by_how_much;
- if (opt.downloaded < old) /* carry flag, where are you when I
- need you? */
- {
- /* Overflow. */
- opt.downloaded_overflow = 1;
- opt.downloaded = ~((VERY_LONG_TYPE)0);
- }
-}
-
-/* Return non-zero if the downloaded amount of bytes exceeds the
- desired quota. If quota is not set or if the amount overflowed, 0
- is returned. */
-int
-downloaded_exceeds_quota (void)
-{
- if (!opt.quota)
- return 0;
- if (opt.downloaded_overflow)
- /* We don't really know. (Wildly) assume not. */
- return 0;
-
- return opt.downloaded > opt.quota;
-}
-
/* If opt.wait or opt.waitretry are specified, and if certain
conditions are met, sleep the appropriate number of seconds. See
the documentation of --wait and --waitretry for more information.
if (count <= opt.waitretry)
sleep (count - 1);
else
- sleep (opt.waitretry);
+ usleep (1000000L * opt.waitretry);
}
else if (opt.wait)
{
/* If random-wait is not specified, or if we are sleeping
between retries of the same download, sleep the fixed
interval. */
- sleep (opt.wait);
+ usleep (1000000L * opt.wait);
else
{
/* Sleep a random amount of time averaging in opt.wait
seconds. The sleeping amount ranges from 0 to
opt.wait*2, inclusive. */
- int waitsecs = random_number (opt.wait * 2 + 1);
+ double waitsecs = 2 * opt.wait * random_float ();
+ DEBUGP (("sleep_between_retrievals: avg=%f,sleep=%f\n",
+ opt.wait, waitsecs));
+ usleep (1000000L * waitsecs);
+ }
+ }
+}
- DEBUGP (("sleep_between_retrievals: norm=%ld,fuzz=%ld,sleep=%d\n",
- opt.wait, waitsecs - opt.wait, waitsecs));
+/* Free the linked list of urlpos. */
+void
+free_urlpos (struct urlpos *l)
+{
+ while (l)
+ {
+ struct urlpos *next = l->next;
+ if (l->url)
+ url_free (l->url);
+ xfree_null (l->local_name);
+ xfree (l);
+ l = next;
+ }
+}
- if (waitsecs)
- sleep (waitsecs);
- }
+/* Rotate FNAME opt.backups times */
+void
+rotate_backups(const char *fname)
+{
+ int maxlen = strlen (fname) + 1 + numdigit (opt.backups) + 1;
+ char *from = (char *)alloca (maxlen);
+ char *to = (char *)alloca (maxlen);
+ struct stat sb;
+ int i;
+
+ if (stat (fname, &sb) == 0)
+ if (S_ISREG (sb.st_mode) == 0)
+ return;
+
+ for (i = opt.backups; i > 1; i--)
+ {
+ sprintf (from, "%s.%d", fname, i - 1);
+ sprintf (to, "%s.%d", fname, i);
+ rename (from, to);
+ }
+
+ sprintf (to, "%s.%d", fname, 1);
+ rename(fname, to);
+}
+
+static int no_proxy_match PARAMS ((const char *, const char **));
+
+/* Return the URL of the proxy appropriate for url U. */
+
+static char *
+getproxy (struct url *u)
+{
+ char *proxy = NULL;
+ char *rewritten_url;
+ static char rewritten_storage[1024];
+
+ if (!opt.use_proxy)
+ return NULL;
+ if (!no_proxy_match (u->host, (const char **)opt.no_proxy))
+ return NULL;
+
+ switch (u->scheme)
+ {
+ case SCHEME_HTTP:
+ proxy = opt.http_proxy ? opt.http_proxy : getenv ("http_proxy");
+ break;
+#ifdef HAVE_SSL
+ case SCHEME_HTTPS:
+ proxy = opt.https_proxy ? opt.https_proxy : getenv ("https_proxy");
+ break;
+#endif
+ case SCHEME_FTP:
+ proxy = opt.ftp_proxy ? opt.ftp_proxy : getenv ("ftp_proxy");
+ break;
+ case SCHEME_INVALID:
+ break;
}
+ if (!proxy || !*proxy)
+ return NULL;
+
+ /* Handle shorthands. `rewritten_storage' is a kludge to allow
+ getproxy() to return static storage. */
+ rewritten_url = rewrite_shorthand_url (proxy);
+ if (rewritten_url)
+ {
+ strncpy (rewritten_storage, rewritten_url, sizeof(rewritten_storage));
+ rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
+ proxy = rewritten_storage;
+ }
+
+ return proxy;
+}
+
+/* Should a host be accessed through proxy, concerning no_proxy? */
+int
+no_proxy_match (const char *host, const char **no_proxy)
+{
+ if (!no_proxy)
+ return 1;
+ else
+ return !sufmatch (no_proxy, host);
}
projects / wget / blobdiff