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
/* 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!
+ stream fp, which should have been open for writing.
The EXPECTED argument is passed to show_progress() unchanged, but
otherwise ignored.
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.
+ (RESTart VALue) is the position from which the download starts,
+ needed for progress display.
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.
+ write to the output stream, respectively. */
- 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 *len, long restval, long expected,
+ int use_expected, double *elapsed)
{
int res = 0;
static char dlbuf[16384];
int dlbufsize = sizeof (dlbuf);
- void *progress = NULL;
struct wget_timer *timer = wtimer_allocate ();
+ double last_successful_read_tm;
+
+ /* The progress gauge, set according to the user preferences. */
+ void *progress = NULL;
+
+ /* 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;
*len = restval;
if (opt.verbose)
- progress = progress_create (restval, expected);
-
- if (rbuf && RBUF_FD (rbuf) == fd)
{
- int sz = 0;
- while ((res = rbuf_flush (rbuf, dlbuf, sizeof (dlbuf))) != 0)
- {
- fwrite (dlbuf, 1, res, fp);
- *len += res;
- sz += res;
- }
- if (sz)
- fflush (fp);
- if (ferror (fp))
- {
- res = -2;
- goto out;
- }
- if (progress)
- progress_update (progress, sz, 0);
+ progress = progress_create (restval, expected);
+ progress_interactive = progress_interactive_p (progress);
}
if (opt.limit_rate)
limit_bandwidth_reset ();
wtimer_reset (timer);
+ 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
if (opt.limit_rate && opt.limit_rate < dlbufsize)
dlbufsize = opt.limit_rate;
- /* Read from fd while there is available data.
+ /* Read from FD while there is available data.
Normally, if expected is 0, it means that it is not known how
much data is expected. However, if use_expected is specified,
{
int amount_to_read = (use_expected
? MIN (expected - *len, dlbufsize) : dlbufsize);
- res = xread (fd, dlbuf, amount_to_read, -1);
+ double tmout = opt.read_timeout;
+ if (progress_interactive)
+ {
+ double waittm;
+ /* 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;
+ waittm = (wtimer_read (timer) - last_successful_read_tm) / 1000;
+ if (waittm + tmout > opt.read_timeout)
+ {
+ /* Don't allow waiting time to exceed read timeout. */
+ tmout = opt.read_timeout - waittm;
+ if (tmout < 0)
+ {
+ /* We've already exceeded the timeout. */
+ res = -1, errno = ETIMEDOUT;
+ break;
+ }
+ }
+ }
+ res = fd_read (fd, dlbuf, amount_to_read, tmout);
- if (res <= 0)
+ if (res == 0 || (res < 0 && errno != ETIMEDOUT))
break;
+ else if (res < 0)
+ res = 0; /* timeout */
- 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 by disk performance. */
- fflush (fp);
- if (ferror (fp))
+ wtimer_update (timer);
+ if (res > 0)
{
- res = -2;
- goto out;
+ if (out)
+ {
+ fwrite (dlbuf, 1, res, out);
+ /* 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 by disk performance. */
+ fflush (out);
+ if (ferror (out))
+ {
+ res = -2;
+ goto out;
+ }
+ }
+ last_successful_read_tm = wtimer_read (timer);
}
- wtimer_update (timer);
if (opt.limit_rate)
limit_bandwidth (res, timer);
return res;
}
\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.
+
+ 2. Determine whether the peeked data, along with the previously
+ read data, includes the terminator.
+
+ 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)
+ {
+ const char *end;
+ int pklen, rdlen, remain;
+
+ /* First, peek at the available data. */
+
+ 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)
+ {
+ /* The data contains the terminator: we'll drain the data up
+ to the end of the terminator. */
+ remain = end - (hunk + tail);
+ if (remain == 0)
+ {
+ /* 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 (rdlen == 0)
+ {
+ if (tail == 0)
+ {
+ /* 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;
+
+ /* Keep looping until all the data arrives. */
+
+ if (tail == bufsize - 1)
+ {
+ bufsize <<= 1;
+ hunk = xrealloc (hunk, bufsize);
+ }
+ }
+}
+
+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;
+}
+
+/* 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
appropriate for the speed. If PAD is non-zero, strings will be
padded to the width of 7 characters (xxxx.xx). */
rewritten_url = rewrite_shorthand_url (proxy);
if (rewritten_url)
{
- strncpy (rewritten_storage, rewritten_url, sizeof(rewritten_storage));
+ strncpy (rewritten_storage, rewritten_url, sizeof (rewritten_storage));
rewritten_storage[sizeof (rewritten_storage) - 1] = '\0';
proxy = rewritten_storage;
}