#include "host.h"
#include "connect.h"
#include "hash.h"
+#include "convert.h"
#ifdef HAVE_SSL
# include "gen_sslfunc.h" /* for ssl_iread */
extern int errno;
#endif
-/* 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;
+/* If non-NULL, the stream to which output should be written. This
+ stream is initialized when `-O' is used. */
+FILE *output_stream;
+
+/* Whether output_document is a regular file we can manipulate,
+ i.e. not `-' or a device file. */
+int output_stream_regular;
\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 TIMER
+ is the timer that started at the beginning of download. */
static void
-limit_bandwidth (long bytes, double delta)
+limit_bandwidth (long bytes, struct wget_timer *timer)
{
+ double delta_t = wtimer_read (timer) - 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));
- usleep ((unsigned long) (1000 * slp));
+ DEBUGP (("\nsleeping %.2f ms for %ld bytes, adjust %.2f ms\n",
+ slp, limit_data.chunk_bytes, limit_data.sleep_adjust));
+
+ t0 = wtimer_read (timer);
+ xsleep (slp / 1000);
+ wtimer_update (timer);
+ t1 = wtimer_read (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);
}
- limit_data.bytes = 0;
- limit_data.dltime = 0;
+ limit_data.chunk_bytes = 0;
+ limit_data.chunk_start = wtimer_read (timer);
}
-#define MIN(i, j) ((i) <= (j) ? (i) : (j))
+#ifndef MIN
+# define MIN(i, j) ((i) <= (j) ? (i) : (j))
+#endif
+
+/* Write data in BUF to OUT. However, if *SKIP is non-zero, skip that
+ amount of data and decrease SKIP. Increment *TOTAL by the amount
+ of data written. */
+
+static int
+write_data (FILE *out, const char *buf, int bufsize, long *skip,
+ long *written)
+{
+ if (!out)
+ return 1;
+ if (*skip > bufsize)
+ {
+ *skip -= bufsize;
+ return 1;
+ }
+ if (*skip)
+ {
+ buf += *skip;
+ bufsize -= *skip;
+ *skip = 0;
+ if (bufsize == 0)
+ return 1;
+ }
+
+ fwrite (buf, 1, bufsize, out);
+ *written += bufsize;
+
+ /* Immediately flush the downloaded data. This should not hinder
+ performance: fast downloads will arrive in large 16K chunks
+ (which stdio would write out immediately anyway), and slow
+ downloads wouldn't be limited by disk speed. */
+ fflush (out);
+ return !ferror (out);
+}
-/* Reads the contents of file descriptor FD, until it is closed, or a
- read error occurs. The data is read in 8K chunks, and stored to
- stream fp, which should have been open for writing. If BUF is
- non-NULL and its file descriptor is equal to FD, flush RBUF first.
- This function will *not* use the rbuf_* functions!
+/* Read the contents of file descriptor FD until it the connection
+ terminates or a read error occurs. The data is read in portions of
+ up to 16K and written to OUT as it arrives. If opt.verbose is set,
+ the progress is shown.
- The EXPECTED argument is passed to show_progress() unchanged, but
- otherwise ignored.
+ TOREAD is the amount of data expected to arrive, normally only used
+ by the progress gauge.
- If opt.verbose is set, the progress is also shown. RESTVAL
- represents a value from which to start downloading (which will be
- shown accordingly). If RESTVAL is non-zero, the stream should have
- been open for appending.
+ STARTPOS is the position from which the download starts, used by
+ the progress gauge. If QTYREAD is non-NULL, the value it points to
+ is incremented by the amount of data read from the network. If
+ QTYWRITTEN is non-NULL, the value it points to is incremented by
+ the amount of data written to disk. The time it took to download
+ the data (in milliseconds) is stored to ELAPSED.
- The function exits and returns codes of 0, -1 and -2 if the
- connection was closed, there was a read error, or if it could not
- write to the output stream, respectively.
+ The function exits and returns the amount of data read. In case of
+ error while reading data, -1 is returned. In case of error while
+ writing data, -2 is returned. */
- IMPORTANT: The function flushes the contents of the buffer in
- rbuf_flush() before actually reading from fd. If you wish to read
- from fd immediately, flush or discard the buffer. */
int
-get_contents (int fd, FILE *fp, long *len, long restval, long expected,
- struct rbuf *rbuf, int use_expected, double *elapsed)
+fd_read_body (int fd, FILE *out, long toread, long startpos,
+ long *qtyread, long *qtywritten, double *elapsed, int flags)
{
- int res = 0;
+ int ret = 0;
static char dlbuf[16384];
int dlbufsize = sizeof (dlbuf);
+ struct wget_timer *timer = NULL;
+ double last_successful_read_tm = 0;
+
+ /* The progress gauge, set according to the user preferences. */
void *progress = NULL;
- struct wget_timer *timer = wtimer_allocate ();
- double dltime = 0, last_dltime = 0;
- *len = restval;
+ /* Non-zero if the progress gauge is interactive, i.e. if it can
+ continually update the display. When true, smaller timeout
+ values are used so that the gauge can update the display when
+ data arrives slowly. */
+ int progress_interactive = 0;
+
+ int exact = flags & rb_read_exactly;
+ long skip = 0;
+
+ /* How much data we've read/written. */
+ long sum_read = 0;
+ long sum_written = 0;
+
+ if (flags & rb_skip_startpos)
+ skip = startpos;
if (opt.verbose)
- progress = progress_create (restval, expected);
+ {
+ /* If we're skipping STARTPOS bytes, pass 0 as the INITIAL
+ argument to progress_create because the indicator doesn't
+ (yet) know about "skipping" data. */
+ progress = progress_create (skip ? 0 : startpos, startpos + toread);
+ progress_interactive = progress_interactive_p (progress);
+ }
- if (rbuf && RBUF_FD (rbuf) == fd)
+ if (opt.limit_rate)
+ limit_bandwidth_reset ();
+
+ /* A timer is needed for tracking progress, for throttling, and for
+ tracking elapsed time. If either of these are requested, start
+ the timer. */
+ if (progress || opt.limit_rate || elapsed)
+ {
+ timer = wtimer_new ();
+ last_successful_read_tm = 0;
+ }
+
+ /* 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 data to read. Normally toread==0
+ means that it is unknown how much data is to arrive. However, if
+ EXACT is set, then toread==0 means what it says: that no data
+ should be read. */
+ while (!exact || (sum_read < toread))
{
- int sz = 0;
- while ((res = rbuf_flush (rbuf, dlbuf, sizeof (dlbuf))) != 0)
+ int rdsize = exact ? MIN (toread - sum_read, dlbufsize) : dlbufsize;
+ double tmout = opt.read_timeout;
+ if (progress_interactive)
{
- fwrite (dlbuf, 1, res, fp);
- *len += res;
- sz += res;
+ /* For interactive progress gauges, always specify a ~1s
+ timeout, so that the gauge can be updated regularly even
+ when the data arrives very slowly or stalls. */
+ tmout = 0.95;
+ if (opt.read_timeout)
+ {
+ double waittm;
+ waittm = (wtimer_read (timer) - last_successful_read_tm) / 1000;
+ if (waittm + tmout > opt.read_timeout)
+ {
+ /* Don't let total idle time exceed read timeout. */
+ tmout = opt.read_timeout - waittm;
+ if (tmout < 0)
+ {
+ /* We've already exceeded the timeout. */
+ ret = -1, errno = ETIMEDOUT;
+ break;
+ }
+ }
+ }
}
- if (sz)
- fflush (fp);
- if (ferror (fp))
+ ret = fd_read (fd, dlbuf, rdsize, tmout);
+
+ if (ret == 0 || (ret < 0 && errno != ETIMEDOUT))
+ break; /* read error */
+ else if (ret < 0)
+ ret = 0; /* read timeout */
+
+ if (progress || opt.limit_rate)
{
- res = -2;
- goto out;
+ wtimer_update (timer);
+ if (ret > 0)
+ last_successful_read_tm = wtimer_read (timer);
}
+
+ if (ret > 0)
+ {
+ sum_read += ret;
+ if (!write_data (out, dlbuf, ret, &skip, &sum_written))
+ {
+ ret = -2;
+ goto out;
+ }
+ }
+
+ if (opt.limit_rate)
+ limit_bandwidth (ret, timer);
+
if (progress)
- progress_update (progress, sz, 0);
+ progress_update (progress, ret, wtimer_read (timer));
+#ifdef WINDOWS
+ if (toread > 0)
+ ws_percenttitle (100.0 *
+ (startpos + sum_read) / (startpos + toread));
+#endif
}
+ if (ret < -1)
+ ret = -1;
- if (opt.limit_rate)
- limit_bandwidth_reset ();
- wtimer_reset (timer);
+ out:
+ if (progress)
+ progress_finish (progress, wtimer_read (timer));
- /* If we're limiting the download, set our buffer size to the
- limit. */
- if (opt.limit_rate && opt.limit_rate < dlbufsize)
- dlbufsize = opt.limit_rate;
+ if (elapsed)
+ *elapsed = wtimer_read (timer);
+ if (timer)
+ wtimer_delete (timer);
+
+ if (qtyread)
+ *qtyread += sum_read;
+ if (qtywritten)
+ *qtywritten += sum_written;
+
+ return ret;
+}
+\f
+/* Read a hunk of data from FD, up until a terminator. The terminator
+ is whatever the TERMINATOR function determines it to be; for
+ example, it can be a line of data, or the head of an HTTP response.
+ The function returns the data read allocated with malloc.
+
+ In case of error, NULL is returned. In case of EOF and no data
+ read, NULL is returned and errno set to 0. In case of EOF with
+ data having been read, the data is returned, but it will
+ (obviously) not contain the terminator.
+
+ The idea is to be able to read a line of input, or otherwise a hunk
+ of text, such as the head of an HTTP request, without crossing the
+ boundary, so that the next call to fd_read etc. reads the data
+ after the hunk. To achieve that, this function does the following:
+
+ 1. Peek at available data.
- /* Read from fd while there is available data.
+ 2. Determine whether the peeked data, along with the previously
+ read data, includes the terminator.
- Normally, if expected is 0, it means that it is not known how
- much data is expected. However, if use_expected is specified,
- then expected being zero means exactly that. */
- while (!use_expected || (*len < expected))
+ 2a. If yes, read the data until the end of the terminator, and
+ exit.
+
+ 2b. If no, read the peeked data and goto 1.
+
+ The function is careful to assume as little as possible about the
+ implementation of peeking. For example, every peek is followed by
+ a read. If the read returns a different amount of data, the
+ process is retried until all data arrives safely.
+
+ BUFSIZE is the size of the initial buffer expected to read all the
+ data in the typical case.
+
+ This function should be used as a building block for other
+ functions -- see fd_read_line as a simple example. */
+
+char *
+fd_read_hunk (int fd, hunk_terminator_t hunk_terminator, int bufsize)
+{
+ char *hunk = xmalloc (bufsize);
+ int tail = 0; /* tail position in HUNK */
+
+ while (1)
{
- int amount_to_read = (use_expected
- ? MIN (expected - *len, dlbufsize) : dlbufsize);
-#ifdef HAVE_SSL
- if (rbuf->ssl!=NULL)
- res = ssl_iread (rbuf->ssl, dlbufsize, amount_to_read);
- else
-#endif /* HAVE_SSL */
- res = iread (fd, dlbuf, amount_to_read);
+ const char *end;
+ int pklen, rdlen, remain;
+
+ /* First, peek at the available data. */
- if (res > 0)
+ pklen = fd_peek (fd, hunk + tail, bufsize - 1 - tail, -1);
+ if (pklen < 0)
+ {
+ xfree (hunk);
+ return NULL;
+ }
+ end = hunk_terminator (hunk, tail, pklen);
+ if (end)
{
- fwrite (dlbuf, 1, 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))
+ /* The data contains the terminator: we'll drain the data up
+ to the end of the terminator. */
+ remain = end - (hunk + tail);
+ if (remain == 0)
{
- res = -2;
- goto out;
+ /* No more data needs to be read. */
+ hunk[tail] = '\0';
+ return hunk;
+ }
+ if (bufsize - 1 < tail + remain)
+ {
+ bufsize = tail + remain + 1;
+ hunk = xrealloc (hunk, bufsize);
}
+ }
+ else
+ /* No terminator: simply read the data we know is (or should
+ be) available. */
+ remain = pklen;
+
+ /* Now, read the data. Note that we make no assumptions about
+ how much data we'll get. (Some TCP stacks are notorious for
+ read returning less data than the previous MSG_PEEK.) */
+
+ rdlen = fd_read (fd, hunk + tail, remain, 0);
+ if (rdlen < 0)
+ {
+ xfree_null (hunk);
+ return NULL;
+ }
+ tail += rdlen;
+ hunk[tail] = '\0';
- /* If bandwidth is not limited, one call to wtimer_elapsed
- is sufficient. */
- dltime = wtimer_elapsed (timer);
- if (opt.limit_rate)
+ if (rdlen == 0)
+ {
+ if (tail == 0)
{
- limit_bandwidth (res, dltime - last_dltime);
- dltime = wtimer_elapsed (timer);
- last_dltime = dltime;
+ /* EOF without anything having been read */
+ xfree (hunk);
+ errno = 0;
+ return NULL;
}
+ else
+ /* EOF seen: return the data we've read. */
+ return hunk;
+ }
+ if (end && rdlen == remain)
+ /* The terminator was seen and the remaining data drained --
+ we got what we came for. */
+ return hunk;
- if (progress)
- progress_update (progress, res, dltime);
- *len += res;
+ /* Keep looping until all the data arrives. */
+
+ if (tail == bufsize - 1)
+ {
+ bufsize <<= 1;
+ hunk = xrealloc (hunk, bufsize);
}
- else
- break;
}
- if (res < -1)
- res = -1;
+}
- out:
- if (progress)
- progress_finish (progress, dltime);
- if (elapsed)
- *elapsed = dltime;
- wtimer_delete (timer);
+static const char *
+line_terminator (const char *hunk, int oldlen, int peeklen)
+{
+ const char *p = memchr (hunk + oldlen, '\n', peeklen);
+ if (p)
+ /* p+1 because we want the line to include '\n' */
+ return p + 1;
+ return NULL;
+}
- return res;
+/* Read one line from FD and return it. The line is allocated using
+ malloc.
+
+ If an error occurs, or if no data can be read, NULL is returned.
+ In the former case errno indicates the error condition, and in the
+ latter case, errno is NULL. */
+
+char *
+fd_read_line (int fd)
+{
+ return fd_read_hunk (fd, line_terminator, 128);
}
\f
/* Return a printed representation of the download rate, as
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);
xfree (url);
}
- ++global_download_count;
RESTORE_POST_DATA;
return result;
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 opt.waitretry is specified and this is a retry, wait for
COUNT-1 number of seconds, or for opt.waitretry seconds. */
if (count <= opt.waitretry)
- sleep (count - 1);
+ xsleep (count - 1);
else
- sleep (opt.waitretry);
+ xsleep (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);
+ xsleep (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));
+ xsleep (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);
}