/* HTTP support.
- Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001, 2002
- Free Software Foundation, Inc.
+ Copyright (C) 2005 Free Software Foundation, Inc.
This file is part of GNU Wget.
#include <stdio.h>
#include <stdlib.h>
-#include <sys/types.h>
-#ifdef HAVE_STRING_H
-# include <string.h>
-#else
-# include <strings.h>
-#endif
+#include <string.h>
#ifdef HAVE_UNISTD_H
# include <unistd.h>
#endif
#include <assert.h>
#include <errno.h>
-#if TIME_WITH_SYS_TIME
-# include <sys/time.h>
-# include <time.h>
-#else
-# if HAVE_SYS_TIME_H
-# include <sys/time.h>
-# else
-# include <time.h>
-# endif
-#endif
-#ifndef errno
-extern int errno;
-#endif
+#include <time.h>
#include "wget.h"
#include "utils.h"
#include "connect.h"
#include "netrc.h"
#ifdef HAVE_SSL
-# include "gen_sslfunc.h"
-#endif /* HAVE_SSL */
+# include "ssl.h"
+#endif
+#ifdef ENABLE_NTLM
+# include "http-ntlm.h"
+#endif
#include "cookies.h"
-#ifdef USE_DIGEST
+#ifdef ENABLE_DIGEST
# include "gen-md5.h"
#endif
#include "convert.h"
extern char *version_string;
-extern LARGE_INT total_downloaded_bytes;
+
+extern FILE *output_stream;
+extern bool output_stream_regular;
#ifndef MIN
# define MIN(x, y) ((x) > (y) ? (y) : (x))
#endif
\f
-static int cookies_loaded_p;
-struct cookie_jar *wget_cookie_jar;
+static bool cookies_loaded_p;
+static struct cookie_jar *wget_cookie_jar;
#define TEXTHTML_S "text/html"
#define TEXTXHTML_S "application/xhtml+xml"
-#define HTTP_ACCEPT "*/*"
/* Some status code validation macros: */
#define H_20X(x) (((x) >= 200) && ((x) < 300))
#define H_PARTIAL(x) ((x) == HTTP_STATUS_PARTIAL_CONTENTS)
-#define H_REDIRECTED(x) ((x) == HTTP_STATUS_MOVED_PERMANENTLY \
- || (x) == HTTP_STATUS_MOVED_TEMPORARILY \
+#define H_REDIRECTED(x) ((x) == HTTP_STATUS_MOVED_PERMANENTLY \
+ || (x) == HTTP_STATUS_MOVED_TEMPORARILY \
+ || (x) == HTTP_STATUS_SEE_OTHER \
|| (x) == HTTP_STATUS_TEMPORARY_REDIRECT)
/* HTTP/1.0 status codes from RFC1945, provided for reference. */
#define HTTP_STATUS_MULTIPLE_CHOICES 300
#define HTTP_STATUS_MOVED_PERMANENTLY 301
#define HTTP_STATUS_MOVED_TEMPORARILY 302
+#define HTTP_STATUS_SEE_OTHER 303 /* from HTTP/1.1 */
#define HTTP_STATUS_NOT_MODIFIED 304
-#define HTTP_STATUS_TEMPORARY_REDIRECT 307
+#define HTTP_STATUS_TEMPORARY_REDIRECT 307 /* from HTTP/1.1 */
/* Client error 4xx. */
#define HTTP_STATUS_BAD_REQUEST 400
#define HTTP_STATUS_UNAUTHORIZED 401
#define HTTP_STATUS_FORBIDDEN 403
#define HTTP_STATUS_NOT_FOUND 404
+#define HTTP_STATUS_RANGE_NOT_SATISFIABLE 416
/* Server errors 5xx. */
#define HTTP_STATUS_INTERNAL 500
#define HTTP_STATUS_NOT_IMPLEMENTED 501
#define HTTP_STATUS_BAD_GATEWAY 502
#define HTTP_STATUS_UNAVAILABLE 503
+\f
+enum rp {
+ rel_none, rel_name, rel_value, rel_both
+};
+
+struct request {
+ const char *method;
+ char *arg;
+
+ struct request_header {
+ char *name, *value;
+ enum rp release_policy;
+ } *headers;
+ int hcount, hcapacity;
+};
+
+/* Create a new, empty request. At least request_set_method must be
+ called before the request can be used. */
+
+static struct request *
+request_new (void)
+{
+ struct request *req = xnew0 (struct request);
+ req->hcapacity = 8;
+ req->headers = xnew_array (struct request_header, req->hcapacity);
+ return req;
+}
+
+/* Set the request's method and its arguments. METH should be a
+ literal string (or it should outlive the request) because it will
+ not be freed. ARG will be freed by request_free. */
+
+static void
+request_set_method (struct request *req, const char *meth, char *arg)
+{
+ req->method = meth;
+ req->arg = arg;
+}
+
+/* Return the method string passed with the last call to
+ request_set_method. */
+
+static const char *
+request_method (const struct request *req)
+{
+ return req->method;
+}
+
+/* Free one header according to the release policy specified with
+ request_set_header. */
+
+static void
+release_header (struct request_header *hdr)
+{
+ switch (hdr->release_policy)
+ {
+ case rel_none:
+ break;
+ case rel_name:
+ xfree (hdr->name);
+ break;
+ case rel_value:
+ xfree (hdr->value);
+ break;
+ case rel_both:
+ xfree (hdr->name);
+ xfree (hdr->value);
+ break;
+ }
+}
+
+/* Set the request named NAME to VALUE. Specifically, this means that
+ a "NAME: VALUE\r\n" header line will be used in the request. If a
+ header with the same name previously existed in the request, its
+ value will be replaced by this one. A NULL value means do nothing.
+
+ RELEASE_POLICY determines whether NAME and VALUE should be released
+ (freed) with request_free. Allowed values are:
+
+ - rel_none - don't free NAME or VALUE
+ - rel_name - free NAME when done
+ - rel_value - free VALUE when done
+ - rel_both - free both NAME and VALUE when done
+
+ Setting release policy is useful when arguments come from different
+ sources. For example:
+
+ // Don't free literal strings!
+ request_set_header (req, "Pragma", "no-cache", rel_none);
+
+ // Don't free a global variable, we'll need it later.
+ request_set_header (req, "Referer", opt.referer, rel_none);
+
+ // Value freshly allocated, free it when done.
+ request_set_header (req, "Range",
+ aprintf ("bytes=%s-", number_to_static_string (hs->restval)),
+ rel_value);
+ */
+
+static void
+request_set_header (struct request *req, char *name, char *value,
+ enum rp release_policy)
+{
+ struct request_header *hdr;
+ int i;
+
+ if (!value)
+ {
+ /* A NULL value is a no-op; if freeing the name is requested,
+ free it now to avoid leaks. */
+ if (release_policy == rel_name || release_policy == rel_both)
+ xfree (name);
+ return;
+ }
+
+ for (i = 0; i < req->hcount; i++)
+ {
+ hdr = &req->headers[i];
+ if (0 == strcasecmp (name, hdr->name))
+ {
+ /* Replace existing header. */
+ release_header (hdr);
+ hdr->name = name;
+ hdr->value = value;
+ hdr->release_policy = release_policy;
+ return;
+ }
+ }
+
+ /* Install new header. */
+
+ if (req->hcount >= req->hcapacity)
+ {
+ req->hcapacity <<= 1;
+ req->headers = xrealloc (req->headers, req->hcapacity * sizeof (*hdr));
+ }
+ hdr = &req->headers[req->hcount++];
+ hdr->name = name;
+ hdr->value = value;
+ hdr->release_policy = release_policy;
+}
+
+/* Like request_set_header, but sets the whole header line, as
+ provided by the user using the `--header' option. For example,
+ request_set_user_header (req, "Foo: bar") works just like
+ request_set_header (req, "Foo", "bar"). */
+
+static void
+request_set_user_header (struct request *req, const char *header)
+{
+ char *name;
+ const char *p = strchr (header, ':');
+ if (!p)
+ return;
+ BOUNDED_TO_ALLOCA (header, p, name);
+ ++p;
+ while (ISSPACE (*p))
+ ++p;
+ request_set_header (req, xstrdup (name), (char *) p, rel_name);
+}
+
+/* Remove the header with specified name from REQ. Returns true if
+ the header was actually removed, false otherwise. */
+
+static bool
+request_remove_header (struct request *req, char *name)
+{
+ int i;
+ for (i = 0; i < req->hcount; i++)
+ {
+ struct request_header *hdr = &req->headers[i];
+ if (0 == strcasecmp (name, hdr->name))
+ {
+ release_header (hdr);
+ /* Move the remaining headers by one. */
+ if (i < req->hcount - 1)
+ memmove (hdr, hdr + 1, (req->hcount - i - 1) * sizeof (*hdr));
+ --req->hcount;
+ return true;
+ }
+ }
+ return false;
+}
+
+#define APPEND(p, str) do { \
+ int A_len = strlen (str); \
+ memcpy (p, str, A_len); \
+ p += A_len; \
+} while (0)
+
+/* Construct the request and write it to FD using fd_write. */
+
+static int
+request_send (const struct request *req, int fd)
+{
+ char *request_string, *p;
+ int i, size, write_error;
+
+ /* Count the request size. */
+ size = 0;
+
+ /* METHOD " " ARG " " "HTTP/1.0" "\r\n" */
+ size += strlen (req->method) + 1 + strlen (req->arg) + 1 + 8 + 2;
+
+ for (i = 0; i < req->hcount; i++)
+ {
+ struct request_header *hdr = &req->headers[i];
+ /* NAME ": " VALUE "\r\n" */
+ size += strlen (hdr->name) + 2 + strlen (hdr->value) + 2;
+ }
+
+ /* "\r\n\0" */
+ size += 3;
+
+ p = request_string = alloca_array (char, size);
+
+ /* Generate the request. */
+
+ APPEND (p, req->method); *p++ = ' ';
+ APPEND (p, req->arg); *p++ = ' ';
+ memcpy (p, "HTTP/1.0\r\n", 10); p += 10;
+
+ for (i = 0; i < req->hcount; i++)
+ {
+ struct request_header *hdr = &req->headers[i];
+ APPEND (p, hdr->name);
+ *p++ = ':', *p++ = ' ';
+ APPEND (p, hdr->value);
+ *p++ = '\r', *p++ = '\n';
+ }
+
+ *p++ = '\r', *p++ = '\n', *p++ = '\0';
+ assert (p - request_string == size);
+
+#undef APPEND
+
+ DEBUGP (("\n---request begin---\n%s---request end---\n", request_string));
+
+ /* Send the request to the server. */
+
+ write_error = fd_write (fd, request_string, size - 1, -1);
+ if (write_error < 0)
+ logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
+ strerror (errno));
+ return write_error;
+}
+
+/* Release the resources used by REQ. */
+
+static void
+request_free (struct request *req)
+{
+ int i;
+ xfree_null (req->arg);
+ for (i = 0; i < req->hcount; i++)
+ release_header (&req->headers[i]);
+ xfree_null (req->headers);
+ xfree (req);
+}
+
+/* Send the contents of FILE_NAME to SOCK. Make sure that exactly
+ PROMISED_SIZE bytes are sent over the wire -- if the file is
+ longer, read only that much; if the file is shorter, report an error. */
+static int
+post_file (int sock, const char *file_name, wgint promised_size)
+{
+ static char chunk[8192];
+ wgint written = 0;
+ int write_error;
+ FILE *fp;
+
+ DEBUGP (("[writing POST file %s ... ", file_name));
+
+ fp = fopen (file_name, "rb");
+ if (!fp)
+ return -1;
+ while (!feof (fp) && written < promised_size)
+ {
+ int towrite;
+ int length = fread (chunk, 1, sizeof (chunk), fp);
+ if (length == 0)
+ break;
+ towrite = MIN (promised_size - written, length);
+ write_error = fd_write (sock, chunk, towrite, -1);
+ if (write_error < 0)
+ {
+ fclose (fp);
+ return -1;
+ }
+ written += towrite;
+ }
+ fclose (fp);
+
+ /* If we've written less than was promised, report a (probably
+ nonsensical) error rather than break the promise. */
+ if (written < promised_size)
+ {
+ errno = EINVAL;
+ return -1;
+ }
+
+ assert (written == promised_size);
+ DEBUGP (("done]\n"));
+ return 0;
+}
+\f
static const char *
-head_terminator (const char *hunk, int oldlen, int peeklen)
+response_head_terminator (const char *hunk, int oldlen, int peeklen)
{
const char *start, *end;
return NULL;
}
+/* The maximum size of a single HTTP response we care to read. This
+ is not meant to impose an arbitrary limit, but to protect the user
+ from Wget slurping up available memory upon encountering malicious
+ or buggy server output. Define it to 0 to remove the limit. */
+
+#define HTTP_RESPONSE_MAX_SIZE 65536
+
/* Read the HTTP request head from FD and return it. The error
conditions are the same as with fd_read_hunk.
data can be treated as body. */
static char *
-fd_read_http_head (int fd)
+read_http_response_head (int fd)
{
- return fd_read_hunk (fd, head_terminator, 512);
+ return fd_read_hunk (fd, response_head_terminator, 512,
+ HTTP_RESPONSE_MAX_SIZE);
}
struct response {
const char *data;
/* The array of pointers that indicate where each header starts.
- For example, given three headers "foo", "bar", and "baz":
- foo: value\r\nbar: value\r\nbaz: value\r\n\r\n
- 0 1 2 3
- I.e. headers[0] points to the beginning of foo, headers[1] points
- to the end of foo and the beginning of bar, etc. */
+ For example, given this HTTP response:
+
+ HTTP/1.0 200 Ok
+ Description: some
+ text
+ Etag: x
+
+ The headers are located like this:
+
+ "HTTP/1.0 200 Ok\r\nDescription: some\r\n text\r\nEtag: x\r\n\r\n"
+ ^ ^ ^ ^
+ headers[0] headers[1] headers[2] headers[3]
+
+ I.e. headers[0] points to the beginning of the request,
+ headers[1] points to the end of the first header and the
+ beginning of the second one, etc. */
+
const char **headers;
};
+/* Create a new response object from the text of the HTTP response,
+ available in HEAD. That text is automatically split into
+ constituent header lines for fast retrieval using
+ resp_header_*. */
+
static struct response *
-response_new (const char *head)
+resp_new (const char *head)
{
const char *hdr;
int count, size;
return resp;
}
- /* Split HEAD into header lines, so that response_header_* functions
+ /* Split HEAD into header lines, so that resp_header_* functions
don't need to do this over and over again. */
size = count = 0;
while (*hdr == ' ' || *hdr == '\t');
}
DO_REALLOC (resp->headers, size, count + 1, const char *);
- resp->headers[count++] = NULL;
+ resp->headers[count] = NULL;
return resp;
}
+/* Locate the header named NAME in the request data, starting with
+ position START. This allows the code to loop through the request
+ data, filtering for all requests of a given name. Returns the
+ found position, or -1 for failure. The code that uses this
+ function typically looks like this:
+
+ for (pos = 0; (pos = resp_header_locate (...)) != -1; pos++)
+ ... do something with header ...
+
+ If you only care about one header, use resp_header_get instead of
+ this function. */
+
static int
-response_header_bounds (const struct response *resp, const char *name,
- const char **begptr, const char **endptr)
+resp_header_locate (const struct response *resp, const char *name, int start,
+ const char **begptr, const char **endptr)
{
int i;
const char **headers = resp->headers;
int name_len;
if (!headers || !headers[1])
- return 0;
+ return -1;
name_len = strlen (name);
+ if (start > 0)
+ i = start;
+ else
+ i = 1;
- for (i = 1; headers[i + 1]; i++)
+ for (; headers[i + 1]; i++)
{
const char *b = headers[i];
const char *e = headers[i + 1];
--e;
*begptr = b;
*endptr = e;
- return 1;
+ return i;
}
}
- return 0;
+ return -1;
}
-static int
-response_header_copy (const struct response *resp, const char *name,
- char *buf, int bufsize)
+/* Find and retrieve the header named NAME in the request data. If
+ found, set *BEGPTR to its starting, and *ENDPTR to its ending
+ position, and return true. Otherwise return false.
+
+ This function is used as a building block for resp_header_copy
+ and resp_header_strdup. */
+
+static bool
+resp_header_get (const struct response *resp, const char *name,
+ const char **begptr, const char **endptr)
+{
+ int pos = resp_header_locate (resp, name, 0, begptr, endptr);
+ return pos != -1;
+}
+
+/* Copy the response header named NAME to buffer BUF, no longer than
+ BUFSIZE (BUFSIZE includes the terminating 0). If the header
+ exists, true is returned, false otherwise. If there should be no
+ limit on the size of the header, use resp_header_strdup instead.
+
+ If BUFSIZE is 0, no data is copied, but the boolean indication of
+ whether the header is present is still returned. */
+
+static bool
+resp_header_copy (const struct response *resp, const char *name,
+ char *buf, int bufsize)
{
const char *b, *e;
- if (!response_header_bounds (resp, name, &b, &e))
- return 0;
+ if (!resp_header_get (resp, name, &b, &e))
+ return false;
if (bufsize)
{
- int len = MIN (e - b, bufsize);
- strncpy (buf, b, len);
+ int len = MIN (e - b, bufsize - 1);
+ memcpy (buf, b, len);
buf[len] = '\0';
}
- return 1;
+ return true;
}
+/* Return the value of header named NAME in RESP, allocated with
+ malloc. If such a header does not exist in RESP, return NULL. */
+
static char *
-response_header_strdup (const struct response *resp, const char *name)
+resp_header_strdup (const struct response *resp, const char *name)
{
const char *b, *e;
- if (!response_header_bounds (resp, name, &b, &e))
+ if (!resp_header_get (resp, name, &b, &e))
return NULL;
return strdupdelim (b, e);
}
returned in *MESSAGE. */
static int
-response_status (const struct response *resp, char **message)
+resp_status (const struct response *resp, char **message)
{
int status;
const char *p, *end;
if (!resp->headers)
{
- /* For a HTTP/0.9 response, always assume 200 response. */
+ /* For a HTTP/0.9 response, assume status 200. */
if (message)
- *message = xstrdup ("OK");
+ *message = xstrdup (_("No headers, assuming HTTP/0.9"));
return 200;
}
return -1;
p += 4;
- /* "/x.x" (optional because some Gnutella servers have been reported
- as not sending the "/x.x" part. */
+ /* Match the HTTP version. This is optional because Gnutella
+ servers have been reported to not specify HTTP version. */
if (p < end && *p == '/')
{
++p;
return status;
}
+/* Release the resources used by RESP. */
+
static void
-response_free (struct response *resp)
+resp_free (struct response *resp)
{
xfree_null (resp->headers);
xfree (resp);
}
-static void
-print_server_response_1 (const char *b, const char *e)
-{
- char *ln;
- if (b < e && e[-1] == '\n')
- --e;
- if (b < e && e[-1] == '\r')
- --e;
- BOUNDED_TO_ALLOCA (b, e, ln);
- logprintf (LOG_VERBOSE, " %s\n", ln);
-}
+/* Print the server response, line by line, omitting the trailing CRLF
+ from individual header lines, and prefixed with PREFIX. */
static void
-print_server_response (const struct response *resp)
+print_server_response (const struct response *resp, const char *prefix)
{
int i;
if (!resp->headers)
return;
for (i = 0; resp->headers[i + 1]; i++)
- print_server_response_1 (resp->headers[i], resp->headers[i + 1]);
+ {
+ const char *b = resp->headers[i];
+ const char *e = resp->headers[i + 1];
+ /* Skip CRLF */
+ if (b < e && e[-1] == '\n')
+ --e;
+ if (b < e && e[-1] == '\r')
+ --e;
+ /* This is safe even on printfs with broken handling of "%.<n>s"
+ because resp->headers ends with \0. */
+ logprintf (LOG_VERBOSE, "%s%.*s\n", prefix, e - b, b);
+ }
}
/* Parse the `Content-Range' header and extract the information it
- contains. Returns 1 if successful, -1 otherwise. */
-static int
-parse_content_range (const char *hdr, long *first_byte_ptr,
- long *last_byte_ptr, long *entity_length_ptr)
+ contains. Returns true if successful, false otherwise. */
+static bool
+parse_content_range (const char *hdr, wgint *first_byte_ptr,
+ wgint *last_byte_ptr, wgint *entity_length_ptr)
{
- long num;
+ wgint num;
/* Ancient versions of Netscape proxy server, presumably predating
rfc2068, sent out `Content-Range' without the "bytes"
specifier. */
- if (!strncasecmp (hdr, "bytes", 5))
+ if (0 == strncasecmp (hdr, "bytes", 5))
{
hdr += 5;
/* "JavaWebServer/1.1.1" sends "bytes: x-y/z", contrary to the
while (ISSPACE (*hdr))
++hdr;
if (!*hdr)
- return 0;
+ return false;
}
if (!ISDIGIT (*hdr))
- return 0;
+ return false;
for (num = 0; ISDIGIT (*hdr); hdr++)
num = 10 * num + (*hdr - '0');
if (*hdr != '-' || !ISDIGIT (*(hdr + 1)))
- return 0;
+ return false;
*first_byte_ptr = num;
++hdr;
for (num = 0; ISDIGIT (*hdr); hdr++)
num = 10 * num + (*hdr - '0');
if (*hdr != '/' || !ISDIGIT (*(hdr + 1)))
- return 0;
+ return false;
*last_byte_ptr = num;
++hdr;
for (num = 0; ISDIGIT (*hdr); hdr++)
num = 10 * num + (*hdr - '0');
*entity_length_ptr = num;
- return 1;
+ return true;
}
-\f
-/* Send the contents of FILE_NAME to SOCK/SSL. Make sure that exactly
- PROMISED_SIZE bytes are sent over the wire -- if the file is
- longer, read only that much; if the file is shorter, report an error. */
-static int
-post_file (int sock, const char *file_name, long promised_size)
+/* Read the body of the request, but don't store it anywhere and don't
+ display a progress gauge. This is useful for reading the bodies of
+ administrative responses to which we will soon issue another
+ request. The response is not useful to the user, but reading it
+ allows us to continue using the same connection to the server.
+
+ If reading fails, false is returned, true otherwise. In debug
+ mode, the body is displayed for debugging purposes. */
+
+static bool
+skip_short_body (int fd, wgint contlen)
{
- static char chunk[8192];
- long written = 0;
- int write_error;
- FILE *fp;
+ enum {
+ SKIP_SIZE = 512, /* size of the download buffer */
+ SKIP_THRESHOLD = 4096 /* the largest size we read */
+ };
+ char dlbuf[SKIP_SIZE + 1];
+ dlbuf[SKIP_SIZE] = '\0'; /* so DEBUGP can safely print it */
- DEBUGP (("[writing POST file %s ... ", file_name));
+ /* We shouldn't get here with unknown contlen. (This will change
+ with HTTP/1.1, which supports "chunked" transfer.) */
+ assert (contlen != -1);
- fp = fopen (file_name, "rb");
- if (!fp)
- return -1;
- while (!feof (fp) && written < promised_size)
+ /* If the body is too large, it makes more sense to simply close the
+ connection than to try to read the body. */
+ if (contlen > SKIP_THRESHOLD)
+ return false;
+
+ DEBUGP (("Skipping %s bytes of body: [", number_to_static_string (contlen)));
+
+ while (contlen > 0)
{
- int towrite;
- int length = fread (chunk, 1, sizeof (chunk), fp);
- if (length == 0)
- break;
- towrite = MIN (promised_size - written, length);
- write_error = fd_write (sock, chunk, towrite, -1);
- if (write_error < 0)
+ int ret = fd_read (fd, dlbuf, MIN (contlen, SKIP_SIZE), -1);
+ if (ret <= 0)
{
- fclose (fp);
- return -1;
+ /* Don't normally report the error since this is an
+ optimization that should be invisible to the user. */
+ DEBUGP (("] aborting (%s).\n",
+ ret < 0 ? strerror (errno) : "EOF received"));
+ return false;
}
- written += towrite;
+ contlen -= ret;
+ /* Safe even if %.*s bogusly expects terminating \0 because
+ we've zero-terminated dlbuf above. */
+ DEBUGP (("%.*s", ret, dlbuf));
}
- fclose (fp);
- /* If we've written less than was promised, report a (probably
- nonsensical) error rather than break the promise. */
- if (written < promised_size)
- {
- errno = EINVAL;
- return -1;
- }
-
- assert (written == promised_size);
- DEBUGP (("done]\n"));
- return 0;
+ DEBUGP (("] done.\n"));
+ return true;
}
\f
/* Persistent connections. Currently, we cache the most recently used
number of these connections. */
/* Whether a persistent connection is active. */
-static int pconn_active;
+static bool pconn_active;
static struct {
/* The socket of the connection. */
int port;
/* Whether a ssl handshake has occoured on this connection. */
- int ssl;
+ bool ssl;
+
+ /* Whether the connection was authorized. This is only done by
+ NTLM, which authorizes *connections* rather than individual
+ requests. (That practice is peculiar for HTTP, but it is a
+ useful optimization.) */
+ bool authorized;
+
+#ifdef ENABLE_NTLM
+ /* NTLM data of the current connection. */
+ struct ntlmdata ntlm;
+#endif
} pconn;
/* Mark the persistent connection as invalid and free the resources it
invalidate_persistent (void)
{
DEBUGP (("Disabling further reuse of socket %d.\n", pconn.socket));
- pconn_active = 0;
+ pconn_active = false;
fd_close (pconn.socket);
xfree (pconn.host);
xzero (pconn);
If a previous connection was persistent, it is closed. */
static void
-register_persistent (const char *host, int port, int fd, int ssl)
+register_persistent (const char *host, int port, int fd, bool ssl)
{
if (pconn_active)
{
}
}
- pconn_active = 1;
+ pconn_active = true;
pconn.socket = fd;
pconn.host = xstrdup (host);
pconn.port = port;
pconn.ssl = ssl;
+ pconn.authorized = false;
DEBUGP (("Registered socket %d for persistent reuse.\n", fd));
}
-/* Return non-zero if a persistent connection is available for
- connecting to HOST:PORT. */
+/* Return true if a persistent connection is available for connecting
+ to HOST:PORT. */
-static int
-persistent_available_p (const char *host, int port, int ssl,
- int *host_lookup_failed)
+static bool
+persistent_available_p (const char *host, int port, bool ssl,
+ bool *host_lookup_failed)
{
/* First, check whether a persistent connection is active at all. */
if (!pconn_active)
- return 0;
+ return false;
/* If we want SSL and the last connection wasn't or vice versa,
don't use it. Checking for host and port is not enough because
HTTP and HTTPS can apparently coexist on the same port. */
if (ssl != pconn.ssl)
- return 0;
+ return false;
/* If we're not connecting to the same port, we're not interested. */
if (port != pconn.port)
- return 0;
+ return false;
/* If the host is the same, we're in business. If not, there is
still hope -- read below. */
if (0 != strcasecmp (host, pconn.host))
{
- /* If pconn.socket is already talking to HOST, we needn't
- reconnect. This happens often when both sites are virtual
- hosts distinguished only by name and served by the same
- network interface, and hence the same web server (possibly
- set up by the ISP and serving many different web sites).
- This admittedly non-standard optimization does not contradict
+ /* Check if pconn.socket is talking to HOST under another name.
+ This happens often when both sites are virtual hosts
+ distinguished only by name and served by the same network
+ interface, and hence the same web server (possibly set up by
+ the ISP and serving many different web sites). This
+ admittedly unconventional optimization does not contradict
HTTP and works well with popular server software. */
- int found;
+ bool found;
ip_address ip;
struct address_list *al;
if (ssl)
/* Don't try to talk to two different SSL sites over the same
- secure connection! (Besides, it's not clear if name-based
- virtual hosting is even possible with SSL.) */
- return 0;
+ secure connection! (Besides, it's not clear that
+ name-based virtual hosting is even possible with SSL.) */
+ return false;
/* If pconn.socket's peer is one of the IP addresses HOST
resolves to, pconn.socket is for all intents and purposes
/* Can't get the peer's address -- something must be very
wrong with the connection. */
invalidate_persistent ();
- return 0;
+ return false;
}
al = lookup_host (host, 0);
if (!al)
{
- *host_lookup_failed = 1;
- return 0;
+ *host_lookup_failed = true;
+ return false;
}
found = address_list_contains (al, &ip);
address_list_release (al);
if (!found)
- return 0;
+ return false;
/* The persistent connection's peer address was found among the
addresses HOST resolved to; therefore, pconn.sock is in fact
let's invalidate the persistent connection before returning
0. */
invalidate_persistent ();
- return 0;
+ return false;
}
- return 1;
+ return true;
}
/* The idea behind these two CLOSE macros is to distinguish between
if (pconn_active && (fd) == pconn.socket) \
invalidate_persistent (); \
else \
- fd_close (fd); \
+ { \
+ fd_close (fd); \
+ fd = -1; \
+ } \
} \
} while (0)
invalidate_persistent (); \
else \
fd_close (fd); \
+ fd = -1; \
} while (0)
\f
struct http_stat
{
- long len; /* received length */
- long contlen; /* expected length */
- long restval; /* the restart value */
+ wgint len; /* received length */
+ wgint contlen; /* expected length */
+ wgint restval; /* the restart value */
int res; /* the result of last read */
char *newloc; /* new location (redirection) */
char *remote_time; /* remote time-stamp string */
char *error; /* textual HTTP error */
int statcode; /* status code */
- double dltime; /* time of the download in msecs */
- int no_truncate; /* whether truncating the file is
- forbidden. */
+ wgint rd_size; /* amount of data read from socket */
+ double dltime; /* time it took to download the data */
const char *referer; /* value of the referer header. */
char **local_file; /* local file. */
};
hs->error = NULL;
}
-static char *create_authorization_line PARAMS ((const char *, const char *,
- const char *, const char *,
- const char *));
-static char *basic_authentication_encode PARAMS ((const char *, const char *,
- const char *));
-static int known_authentication_scheme_p PARAMS ((const char *));
+static char *create_authorization_line (const char *, const char *,
+ const char *, const char *,
+ const char *, bool *);
+static char *basic_authentication_encode (const char *, const char *);
+static bool known_authentication_scheme_p (const char *, const char *);
-time_t http_atotm PARAMS ((const char *));
+time_t http_atotm (const char *);
#define BEGINS_WITH(line, string_constant) \
(!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
&& (ISSPACE (line[sizeof (string_constant) - 1]) \
|| !line[sizeof (string_constant) - 1]))
+#define SET_USER_AGENT(req) do { \
+ if (!opt.useragent) \
+ request_set_header (req, "User-Agent", \
+ aprintf ("Wget/%s", version_string), rel_value); \
+ else if (*opt.useragent) \
+ request_set_header (req, "User-Agent", opt.useragent, rel_none); \
+} while (0)
+
+/* The flags that allow clobbering the file (opening with "wb").
+ Defined here to avoid repetition later. #### This will require
+ rework. */
+#define ALLOW_CLOBBER (opt.noclobber || opt.always_rest || opt.timestamping \
+ || opt.dirstruct || opt.output_document)
+
/* Retrieve a document through HTTP protocol. It recognizes status
code, and correctly handles redirections. It closes the network
socket. If it receives an error from the functions below it, it
static uerr_t
gethttp (struct url *u, struct http_stat *hs, int *dt, struct url *proxy)
{
- char *request, *type, *command, *full_path;
+ struct request *req;
+
+ char *type;
char *user, *passwd;
- char *pragma_h, *referer, *useragent, *range, *wwwauth;
- char *authenticate_h;
char *proxyauth;
- char *port_maybe;
- char *request_keep_alive;
- int sock, statcode;
+ int statcode;
int write_error;
- long contlen, contrange;
+ wgint contlen, contrange;
struct url *conn;
FILE *fp;
- int auth_tried_already;
- int using_ssl = 0;
- char *cookies = NULL;
+
+ int sock = -1;
+ int flags;
+
+ /* Set to 1 when the authorization has failed permanently and should
+ not be tried again. */
+ bool auth_finished = false;
+
+ /* Whether NTLM authentication is used for this request. */
+ bool ntlm_seen = false;
+
+ /* Whether our connection to the remote host is through SSL. */
+ bool using_ssl = false;
+
+ /* Whether a HEAD request will be issued (as opposed to GET or
+ POST). */
+ bool head_only = !!(*dt & HEAD_ONLY);
char *head;
struct response *resp;
char hdrval[256];
char *message;
- char *set_cookie;
/* Whether this connection will be kept alive after the HTTP request
is done. */
- int keep_alive;
-
- /* Flag that detects having received a keep-alive response. */
- int keep_alive_confirmed;
+ bool keep_alive;
- /* Whether keep-alive should be inhibited. */
- int inhibit_keep_alive;
+ /* Whether keep-alive should be inhibited.
- /* Whether we need to print the host header with braces around host,
- e.g. "Host: [3ffe:8100:200:2::2]:1234" instead of the usual
- "Host: symbolic-name:1234". */
- int squares_around_host = 0;
+ RFC 2068 requests that 1.0 clients not send keep-alive requests
+ to proxies. This is because many 1.0 proxies do not interpret
+ the Connection header and transfer it to the remote server,
+ causing it to not close the connection and leave both the proxy
+ and the client hanging. */
+ bool inhibit_keep_alive =
+ !opt.http_keep_alive || opt.ignore_length || proxy != NULL;
/* Headers sent when using POST. */
- char *post_content_type, *post_content_length;
- long post_data_size = 0;
+ wgint post_data_size = 0;
- int host_lookup_failed;
+ bool host_lookup_failed = false;
#ifdef HAVE_SSL
- /* Initialize the SSL context. After the first run, this is a
- no-op. */
- switch (ssl_init ())
+ if (u->scheme == SCHEME_HTTPS)
{
- case SSLERRCTXCREATE:
- /* this is fatal */
- logprintf (LOG_NOTQUIET, _("Failed to set up an SSL context\n"));
- return SSLERRCTXCREATE;
- case SSLERRCERTFILE:
- /* try without certfile */
- logprintf (LOG_NOTQUIET,
- _("Failed to load certificates from %s\n"),
- opt.sslcertfile);
- logprintf (LOG_NOTQUIET,
- _("Trying without the specified certificate\n"));
- break;
- case SSLERRCERTKEY:
- logprintf (LOG_NOTQUIET,
- _("Failed to get certificate key from %s\n"),
- opt.sslcertkey);
- logprintf (LOG_NOTQUIET,
- _("Trying without the specified certificate\n"));
- break;
- default:
- break;
+ /* Initialize the SSL context. After this has once been done,
+ it becomes a no-op. */
+ if (!ssl_init ())
+ {
+ scheme_disable (SCHEME_HTTPS);
+ logprintf (LOG_NOTQUIET,
+ _("Disabling SSL due to encountered errors.\n"));
+ return SSLINITFAILED;
+ }
}
#endif /* HAVE_SSL */
- if (!(*dt & HEAD_ONLY))
+ if (!head_only)
/* If we're doing a GET on the URL, as opposed to just a HEAD, we need to
know the local filename so we can save to it. */
assert (*hs->local_file != NULL);
- authenticate_h = NULL;
- auth_tried_already = 0;
-
- inhibit_keep_alive = !opt.http_keep_alive;
-
- again:
- /* We need to come back here when the initial attempt to retrieve
- without authorization header fails. (Expected to happen at least
- for the Digest authorization scheme.) */
-
- keep_alive = 0;
- keep_alive_confirmed = 0;
-
- post_content_type = NULL;
- post_content_length = NULL;
-
/* Initialize certain elements of struct http_stat. */
- hs->len = 0L;
+ hs->len = 0;
hs->contlen = -1;
hs->res = -1;
hs->newloc = NULL;
conn = u;
+ /* Prepare the request to send. */
+
+ req = request_new ();
+ {
+ char *meth_arg;
+ const char *meth = "GET";
+ if (head_only)
+ meth = "HEAD";
+ else if (opt.post_file_name || opt.post_data)
+ meth = "POST";
+ /* Use the full path, i.e. one that includes the leading slash and
+ the query string. E.g. if u->path is "foo/bar" and u->query is
+ "param=value", full_path will be "/foo/bar?param=value". */
+ if (proxy
+#ifdef HAVE_SSL
+ /* When using SSL over proxy, CONNECT establishes a direct
+ connection to the HTTPS server. Therefore use the same
+ argument as when talking to the server directly. */
+ && u->scheme != SCHEME_HTTPS
+#endif
+ )
+ meth_arg = xstrdup (u->url);
+ else
+ meth_arg = url_full_path (u);
+ request_set_method (req, meth, meth_arg);
+ }
+
+ request_set_header (req, "Referer", (char *) hs->referer, rel_none);
+ if (*dt & SEND_NOCACHE)
+ request_set_header (req, "Pragma", "no-cache", rel_none);
+ if (hs->restval)
+ request_set_header (req, "Range",
+ aprintf ("bytes=%s-",
+ number_to_static_string (hs->restval)),
+ rel_value);
+ SET_USER_AGENT (req);
+ request_set_header (req, "Accept", "*/*", rel_none);
+
+ /* Find the username and password for authentication. */
+ user = u->user;
+ passwd = u->passwd;
+ search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
+ user = user ? user : (opt.http_user ? opt.http_user : opt.user);
+ passwd = passwd ? passwd : (opt.http_passwd ? opt.http_passwd : opt.passwd);
+
+ if (user && passwd)
+ {
+ /* We have the username and the password, but haven't tried
+ any authorization yet. Let's see if the "Basic" method
+ works. If not, we'll come back here and construct a
+ proper authorization method with the right challenges.
+
+ If we didn't employ this kind of logic, every URL that
+ requires authorization would have to be processed twice,
+ which is very suboptimal and generates a bunch of false
+ "unauthorized" errors in the server log.
+
+ #### But this logic also has a serious problem when used
+ with stronger authentications: we *first* transmit the
+ username and the password in clear text, and *then* attempt a
+ stronger authentication scheme. That cannot be right! We
+ are only fortunate that almost everyone still uses the
+ `Basic' scheme anyway.
+
+ There should be an option to prevent this from happening, for
+ those who use strong authentication schemes and value their
+ passwords. */
+ request_set_header (req, "Authorization",
+ basic_authentication_encode (user, passwd),
+ rel_value);
+ }
+
proxyauth = NULL;
if (proxy)
{
proxy_user = proxy->user;
proxy_passwd = proxy->passwd;
}
- /* #### This does not appear right. Can't the proxy request,
- say, `Digest' authentication? */
- if (proxy_user && proxy_passwd)
- proxyauth = basic_authentication_encode (proxy_user, proxy_passwd,
- "Proxy-Authorization");
+ /* #### This does not appear right. Can't the proxy request,
+ say, `Digest' authentication? */
+ if (proxy_user && proxy_passwd)
+ proxyauth = basic_authentication_encode (proxy_user, proxy_passwd);
+
+ /* If we're using a proxy, we will be connecting to the proxy
+ server. */
+ conn = proxy;
+
+ /* Proxy authorization over SSL is handled below. */
+#ifdef HAVE_SSL
+ if (u->scheme != SCHEME_HTTPS)
+#endif
+ request_set_header (req, "Proxy-Authorization", proxyauth, rel_value);
+ }
+
+ {
+ /* Whether we need to print the host header with braces around
+ host, e.g. "Host: [3ffe:8100:200:2::2]:1234" instead of the
+ usual "Host: symbolic-name:1234". */
+ bool squares = strchr (u->host, ':') != NULL;
+ if (u->port == scheme_default_port (u->scheme))
+ request_set_header (req, "Host",
+ aprintf (squares ? "[%s]" : "%s", u->host),
+ rel_value);
+ else
+ request_set_header (req, "Host",
+ aprintf (squares ? "[%s]:%d" : "%s:%d",
+ u->host, u->port),
+ rel_value);
+ }
+
+ if (!inhibit_keep_alive)
+ request_set_header (req, "Connection", "Keep-Alive", rel_none);
+
+ if (opt.cookies)
+ request_set_header (req, "Cookie",
+ cookie_header (wget_cookie_jar,
+ u->host, u->port, u->path,
+#ifdef HAVE_SSL
+ u->scheme == SCHEME_HTTPS
+#else
+ 0
+#endif
+ ),
+ rel_value);
+
+ if (opt.post_data || opt.post_file_name)
+ {
+ request_set_header (req, "Content-Type",
+ "application/x-www-form-urlencoded", rel_none);
+ if (opt.post_data)
+ post_data_size = strlen (opt.post_data);
+ else
+ {
+ post_data_size = file_size (opt.post_file_name);
+ if (post_data_size == -1)
+ {
+ logprintf (LOG_NOTQUIET, _("POST data file missing: %s (%s)\n"),
+ opt.post_file_name, strerror (errno));
+ post_data_size = 0;
+ }
+ }
+ request_set_header (req, "Content-Length",
+ xstrdup (number_to_static_string (post_data_size)),
+ rel_value);
+ }
- /* If we're using a proxy, we will be connecting to the proxy
- server. */
- conn = proxy;
+ /* Add the user headers. */
+ if (opt.user_headers)
+ {
+ int i;
+ for (i = 0; opt.user_headers[i]; i++)
+ request_set_user_header (req, opt.user_headers[i]);
}
- host_lookup_failed = 0;
- sock = -1;
+ retry_with_auth:
+ /* We need to come back here when the initial attempt to retrieve
+ without authorization header fails. (Expected to happen at least
+ for the Digest authorization scheme.) */
+
+ keep_alive = false;
- /* First: establish the connection. */
+ /* Establish the connection. */
if (!inhibit_keep_alive)
{
sock = pconn.socket;
using_ssl = pconn.ssl;
logprintf (LOG_VERBOSE, _("Reusing existing connection to %s:%d.\n"),
- pconn.host, pconn.port);
+ escnonprint (pconn.host), pconn.port);
DEBUGP (("Reusing fd %d.\n", sock));
+ if (pconn.authorized)
+ /* If the connection is already authorized, the "Basic"
+ authorization added by code above is unnecessary and
+ only hurts us. */
+ request_remove_header (req, "Authorization");
}
}
look up conn->host in some cases. If that lookup failed, we
don't need to bother with connect_to_host. */
if (host_lookup_failed)
- return HOSTERR;
+ {
+ request_free (req);
+ return HOSTERR;
+ }
sock = connect_to_host (conn->host, conn->port);
if (sock == E_HOST)
- return HOSTERR;
+ {
+ request_free (req);
+ return HOSTERR;
+ }
else if (sock < 0)
- return (retryable_socket_connect_error (errno)
- ? CONERROR : CONIMPOSSIBLE);
+ {
+ request_free (req);
+ return (retryable_socket_connect_error (errno)
+ ? CONERROR : CONIMPOSSIBLE);
+ }
#ifdef HAVE_SSL
if (proxy && u->scheme == SCHEME_HTTPS)
{
/* When requesting SSL URLs through proxies, use the
CONNECT method to request passthrough. */
- char *connect =
- (char *) alloca (64
- + strlen (u->host)
- + (proxyauth ? strlen (proxyauth) : 0));
- sprintf (connect, "CONNECT %s:%d HTTP/1.0\r\n%s\r\n",
- u->host, u->port, proxyauth ? proxyauth : "");
- /* Now that PROXYAUTH is part of the CONNECT request, zero
- it out so we don't send proxy authorization with the
- regular request below. */
- proxyauth = NULL;
-
- DEBUGP (("Writing to proxy: [%s]\n", connect));
- write_error = fd_write (sock, connect, strlen (connect), -1);
+ struct request *connreq = request_new ();
+ request_set_method (connreq, "CONNECT",
+ aprintf ("%s:%d", u->host, u->port));
+ SET_USER_AGENT (connreq);
+ if (proxyauth)
+ {
+ request_set_header (connreq, "Proxy-Authorization",
+ proxyauth, rel_value);
+ /* Now that PROXYAUTH is part of the CONNECT request,
+ zero it out so we don't send proxy authorization with
+ the regular request below. */
+ proxyauth = NULL;
+ }
+ /* Examples in rfc2817 use the Host header in CONNECT
+ requests. I don't see how that gains anything, given
+ that the contents of Host would be exactly the same as
+ the contents of CONNECT. */
+
+ write_error = request_send (connreq, sock);
+ request_free (connreq);
if (write_error < 0)
{
logprintf (LOG_VERBOSE, _("Failed writing to proxy: %s.\n"),
return WRITEFAILED;
}
- head = fd_read_http_head (sock);
+ head = read_http_response_head (sock);
if (!head)
{
logprintf (LOG_VERBOSE, _("Failed reading proxy response: %s\n"),
}
DEBUGP (("proxy responded with: [%s]\n", head));
- resp = response_new (head);
- statcode = response_status (resp, &message);
- response_free (resp);
+ resp = resp_new (head);
+ statcode = resp_status (resp, &message);
+ resp_free (resp);
+ xfree (head);
if (statcode != 200)
{
failed_tunnel:
logprintf (LOG_NOTQUIET, _("Proxy tunneling failed: %s"),
- message ? message : "?");
+ message ? escnonprint (message) : "?");
xfree_null (message);
return CONSSLERR;
}
- xfree (message);
+ xfree_null (message);
/* SOCK is now *really* connected to u->host, so update CONN
to reflect this. That way register_persistent will
if (conn->scheme == SCHEME_HTTPS)
{
- if (!ssl_connect (sock))
+ if (!ssl_connect (sock) || !ssl_check_certificate (sock, u->host))
{
fd_close (sock);
return CONSSLERR;
}
- using_ssl = 1;
+ using_ssl = true;
}
#endif /* HAVE_SSL */
}
- if (*dt & HEAD_ONLY)
- command = "HEAD";
- else if (opt.post_file_name || opt.post_data)
- command = "POST";
- else
- command = "GET";
-
- referer = NULL;
- if (hs->referer)
- {
- referer = (char *)alloca (9 + strlen (hs->referer) + 3);
- sprintf (referer, "Referer: %s\r\n", hs->referer);
- }
-
- if (*dt & SEND_NOCACHE)
- pragma_h = "Pragma: no-cache\r\n";
- else
- pragma_h = "";
-
- if (hs->restval)
- {
- range = (char *)alloca (13 + numdigit (hs->restval) + 4);
- /* Gag me! Some servers (e.g. WebSitePro) have been known to
- respond to the following `Range' format by generating a
- multipart/x-byte-ranges MIME document! This MIME type was
- present in an old draft of the byteranges specification.
- HTTP/1.1 specifies a multipart/byte-ranges MIME type, but
- only if multiple non-overlapping ranges are requested --
- which Wget never does. */
- sprintf (range, "Range: bytes=%ld-\r\n", hs->restval);
- }
- else
- range = NULL;
- if (opt.useragent)
- STRDUP_ALLOCA (useragent, opt.useragent);
- else
- {
- useragent = (char *)alloca (10 + strlen (version_string));
- sprintf (useragent, "Wget/%s", version_string);
- }
- /* Construct the authentication, if userid is present. */
- user = u->user;
- passwd = u->passwd;
- search_netrc (u->host, (const char **)&user, (const char **)&passwd, 0);
- user = user ? user : opt.http_user;
- passwd = passwd ? passwd : opt.http_passwd;
-
- wwwauth = NULL;
- if (user && passwd)
- {
- if (!authenticate_h)
- {
- /* We have the username and the password, but haven't tried
- any authorization yet. Let's see if the "Basic" method
- works. If not, we'll come back here and construct a
- proper authorization method with the right challenges.
-
- If we didn't employ this kind of logic, every URL that
- requires authorization would have to be processed twice,
- which is very suboptimal and generates a bunch of false
- "unauthorized" errors in the server log.
-
- #### But this logic also has a serious problem when used
- with stronger authentications: we *first* transmit the
- username and the password in clear text, and *then*
- attempt a stronger authentication scheme. That cannot be
- right! We are only fortunate that almost everyone still
- uses the `Basic' scheme anyway.
-
- There should be an option to prevent this from happening,
- for those who use strong authentication schemes and value
- their passwords. */
- wwwauth = basic_authentication_encode (user, passwd, "Authorization");
- }
- else
- {
- /* Use the full path, i.e. one that includes the leading
- slash and the query string, but is independent of proxy
- setting. */
- char *pth = url_full_path (u);
- wwwauth = create_authorization_line (authenticate_h, user, passwd,
- command, pth);
- xfree (pth);
- }
- }
-
- /* String of the form :PORT. Used only for non-standard ports. */
- port_maybe = NULL;
- if (u->port != scheme_default_port (u->scheme))
- {
- port_maybe = (char *)alloca (numdigit (u->port) + 2);
- sprintf (port_maybe, ":%d", u->port);
- }
-
- if (!inhibit_keep_alive)
- request_keep_alive = "Connection: Keep-Alive\r\n";
- else
- request_keep_alive = NULL;
-
- if (opt.cookies)
- cookies = cookie_header (wget_cookie_jar, u->host, u->port, u->path,
-#ifdef HAVE_SSL
- u->scheme == SCHEME_HTTPS
-#else
- 0
-#endif
- );
-
- if (opt.post_data || opt.post_file_name)
- {
- post_content_type = "Content-Type: application/x-www-form-urlencoded\r\n";
- if (opt.post_data)
- post_data_size = strlen (opt.post_data);
- else
- {
- post_data_size = file_size (opt.post_file_name);
- if (post_data_size == -1)
- {
- logprintf (LOG_NOTQUIET, "POST data file missing: %s\n",
- opt.post_file_name);
- post_data_size = 0;
- }
- }
- post_content_length = xmalloc (16 + numdigit (post_data_size) + 2 + 1);
- sprintf (post_content_length,
- "Content-Length: %ld\r\n", post_data_size);
- }
-
- if (proxy)
- full_path = xstrdup (u->url);
- else
- /* Use the full path, i.e. one that includes the leading slash and
- the query string. E.g. if u->path is "foo/bar" and u->query is
- "param=value", full_path will be "/foo/bar?param=value". */
- full_path = url_full_path (u);
-
- if (strchr (u->host, ':'))
- squares_around_host = 1;
-
- /* Allocate the memory for the request. */
- request = (char *)alloca (strlen (command)
- + strlen (full_path)
- + strlen (useragent)
- + strlen (u->host)
- + (port_maybe ? strlen (port_maybe) : 0)
- + strlen (HTTP_ACCEPT)
- + (request_keep_alive
- ? strlen (request_keep_alive) : 0)
- + (referer ? strlen (referer) : 0)
- + (cookies ? strlen (cookies) : 0)
- + (wwwauth ? strlen (wwwauth) : 0)
- + (proxyauth ? strlen (proxyauth) : 0)
- + (range ? strlen (range) : 0)
- + strlen (pragma_h)
- + (post_content_type
- ? strlen (post_content_type) : 0)
- + (post_content_length
- ? strlen (post_content_length) : 0)
- + (opt.user_header ? strlen (opt.user_header) : 0)
- + 64);
- /* Construct the request. */
- sprintf (request, "\
-%s %s HTTP/1.0\r\n\
-User-Agent: %s\r\n\
-Host: %s%s%s%s\r\n\
-Accept: %s\r\n\
-%s%s%s%s%s%s%s%s%s%s\r\n",
- command, full_path,
- useragent,
- squares_around_host ? "[" : "", u->host, squares_around_host ? "]" : "",
- port_maybe ? port_maybe : "",
- HTTP_ACCEPT,
- request_keep_alive ? request_keep_alive : "",
- referer ? referer : "",
- cookies ? cookies : "",
- wwwauth ? wwwauth : "",
- proxyauth ? proxyauth : "",
- range ? range : "",
- pragma_h,
- post_content_type ? post_content_type : "",
- post_content_length ? post_content_length : "",
- opt.user_header ? opt.user_header : "");
- DEBUGP (("\n---request begin---\n%s", request));
-
- /* Free the temporary memory. */
- xfree_null (wwwauth);
- xfree_null (proxyauth);
- xfree_null (cookies);
- xfree (full_path);
-
/* Send the request to server. */
- write_error = fd_write (sock, request, strlen (request), -1);
+ write_error = request_send (req, sock);
if (write_error >= 0)
{
else if (opt.post_file_name && post_data_size != 0)
write_error = post_file (sock, opt.post_file_name, post_data_size);
}
- DEBUGP (("---request end---\n"));
if (write_error < 0)
{
logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"),
strerror (errno));
CLOSE_INVALIDATE (sock);
+ request_free (req);
return WRITEFAILED;
}
logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "),
proxy ? "Proxy" : "HTTP");
- contlen = contrange = -1;
- type = NULL;
- statcode = -1;
+ contlen = -1;
+ contrange = 0;
*dt &= ~RETROKF;
- head = fd_read_http_head (sock);
+ head = read_http_response_head (sock);
if (!head)
{
- logputs (LOG_VERBOSE, "\n");
if (errno == 0)
{
logputs (LOG_NOTQUIET, _("No data received.\n"));
CLOSE_INVALIDATE (sock);
+ request_free (req);
return HEOF;
}
else
logprintf (LOG_NOTQUIET, _("Read error (%s) in headers.\n"),
strerror (errno));
CLOSE_INVALIDATE (sock);
+ request_free (req);
return HERR;
}
}
DEBUGP (("\n---response begin---\n%s---response end---\n", head));
- resp = response_new (head);
+ resp = resp_new (head);
/* Check for status line. */
message = NULL;
- statcode = response_status (resp, &message);
+ statcode = resp_status (resp, &message);
if (!opt.server_response)
- logprintf (LOG_VERBOSE, "%2d %s\n", statcode, message ? message : "");
+ logprintf (LOG_VERBOSE, "%2d %s\n", statcode,
+ message ? escnonprint (message) : "");
else
{
logprintf (LOG_VERBOSE, "\n");
- print_server_response (resp);
+ print_server_response (resp, " ");
+ }
+
+ if (!opt.ignore_length
+ && resp_header_copy (resp, "Content-Length", hdrval, sizeof (hdrval)))
+ {
+ wgint parsed;
+ errno = 0;
+ parsed = str_to_wgint (hdrval, NULL, 10);
+ if (parsed == WGINT_MAX && errno == ERANGE)
+ /* Out of range.
+ #### If Content-Length is out of range, it most likely
+ means that the file is larger than 2G and that we're
+ compiled without LFS. In that case we should probably
+ refuse to even attempt to download the file. */
+ contlen = -1;
+ else
+ contlen = parsed;
+ }
+
+ /* Check for keep-alive related responses. */
+ if (!inhibit_keep_alive && contlen != -1)
+ {
+ if (resp_header_copy (resp, "Keep-Alive", NULL, 0))
+ keep_alive = true;
+ else if (resp_header_copy (resp, "Connection", hdrval, sizeof (hdrval)))
+ {
+ if (0 == strcasecmp (hdrval, "Keep-Alive"))
+ keep_alive = true;
+ }
+ }
+ if (keep_alive)
+ /* The server has promised that it will not close the connection
+ when we're done. This means that we can register it. */
+ register_persistent (conn->host, conn->port, sock, using_ssl);
+
+ if (statcode == HTTP_STATUS_UNAUTHORIZED)
+ {
+ /* Authorization is required. */
+ if (keep_alive && !head_only && skip_short_body (sock, contlen))
+ CLOSE_FINISH (sock);
+ else
+ CLOSE_INVALIDATE (sock);
+ pconn.authorized = false;
+ if (!auth_finished && (user && passwd))
+ {
+ /* IIS sends multiple copies of WWW-Authenticate, one with
+ the value "negotiate", and other(s) with data. Loop over
+ all the occurrences and pick the one we recognize. */
+ int wapos;
+ const char *wabeg, *waend;
+ char *www_authenticate = NULL;
+ for (wapos = 0;
+ (wapos = resp_header_locate (resp, "WWW-Authenticate", wapos,
+ &wabeg, &waend)) != -1;
+ ++wapos)
+ if (known_authentication_scheme_p (wabeg, waend))
+ {
+ BOUNDED_TO_ALLOCA (wabeg, waend, www_authenticate);
+ break;
+ }
+
+ if (!www_authenticate)
+ /* If the authentication header is missing or
+ unrecognized, there's no sense in retrying. */
+ logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
+ else if (BEGINS_WITH (www_authenticate, "Basic"))
+ /* If the authentication scheme is "Basic", which we send
+ by default, there's no sense in retrying either. (This
+ should be changed when we stop sending "Basic" data by
+ default.) */
+ ;
+ else
+ {
+ char *pth;
+ pth = url_full_path (u);
+ request_set_header (req, "Authorization",
+ create_authorization_line (www_authenticate,
+ user, passwd,
+ request_method (req),
+ pth,
+ &auth_finished),
+ rel_value);
+ if (BEGINS_WITH (www_authenticate, "NTLM"))
+ ntlm_seen = true;
+ xfree (pth);
+ goto retry_with_auth;
+ }
+ }
+ logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
+ request_free (req);
+ return AUTHFAILED;
+ }
+ else /* statcode != HTTP_STATUS_UNAUTHORIZED */
+ {
+ /* Kludge: if NTLM is used, mark the TCP connection as authorized. */
+ if (ntlm_seen)
+ pconn.authorized = true;
}
+ request_free (req);
hs->statcode = statcode;
if (statcode == -1)
hs->error = xstrdup (_("(no description)"));
else
hs->error = xstrdup (message);
+ xfree (message);
- if (response_header_copy (resp, "Content-Length", hdrval, sizeof (hdrval)))
- contlen = strtol (hdrval, NULL, 10);
- type = response_header_strdup (resp, "Content-Type");
+ type = resp_header_strdup (resp, "Content-Type");
if (type)
{
char *tmp = strchr (type, ';');
*tmp = '\0';
}
}
- hs->newloc = response_header_strdup (resp, "Location");
- hs->remote_time = response_header_strdup (resp, "Last-Modified");
- set_cookie = response_header_strdup (resp, "Set-Cookie");
- if (set_cookie)
+ hs->newloc = resp_header_strdup (resp, "Location");
+ hs->remote_time = resp_header_strdup (resp, "Last-Modified");
+
+ /* Handle (possibly multiple instances of) the Set-Cookie header. */
+ if (opt.cookies)
{
+ int scpos;
+ const char *scbeg, *scend;
/* The jar should have been created by now. */
assert (wget_cookie_jar != NULL);
- cookie_handle_set_cookie (wget_cookie_jar, u->host, u->port, u->path,
- set_cookie);
- xfree (set_cookie);
- }
- authenticate_h = response_header_strdup (resp, "WWW-Authenticate");
- if (response_header_copy (resp, "Content-Range", hdrval, sizeof (hdrval)))
- {
- long first_byte_pos, last_byte_pos, entity_length;
- if (parse_content_range (hdrval, &first_byte_pos, &last_byte_pos,
- &entity_length))
- contrange = first_byte_pos;
- }
-
- /* Check for keep-alive related responses. */
- if (!inhibit_keep_alive && contlen != -1)
- {
- if (response_header_copy (resp, "Keep-Alive", NULL, 0))
- keep_alive = 1;
- else if (response_header_copy (resp, "Connection", hdrval,
- sizeof (hdrval)))
+ for (scpos = 0;
+ (scpos = resp_header_locate (resp, "Set-Cookie", scpos,
+ &scbeg, &scend)) != -1;
+ ++scpos)
{
- if (0 == strcasecmp (hdrval, "Keep-Alive"))
- keep_alive = 1;
+ char *set_cookie; BOUNDED_TO_ALLOCA (scbeg, scend, set_cookie);
+ cookie_handle_set_cookie (wget_cookie_jar, u->host, u->port,
+ u->path, set_cookie);
}
}
- response_free (resp);
-
- if (keep_alive)
- /* The server has promised that it will not close the connection
- when we're done. This means that we can register it. */
- register_persistent (conn->host, conn->port, sock, using_ssl);
- if ((statcode == HTTP_STATUS_UNAUTHORIZED)
- && authenticate_h)
- {
- /* Authorization is required. */
- xfree_null (type);
- type = NULL;
- free_hstat (hs);
- CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
- might be more bytes in the body. */
- if (auth_tried_already)
- {
- /* If we have tried it already, then there is not point
- retrying it. */
- failed:
- logputs (LOG_NOTQUIET, _("Authorization failed.\n"));
- xfree (authenticate_h);
- return AUTHFAILED;
- }
- else if (!known_authentication_scheme_p (authenticate_h))
- {
- xfree (authenticate_h);
- logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
- return AUTHFAILED;
- }
- else if (BEGINS_WITH (authenticate_h, "Basic"))
- {
- /* The authentication scheme is basic, the one we try by
- default, and it failed. There's no sense in trying
- again. */
- goto failed;
- }
- else
- {
- auth_tried_already = 1;
- goto again;
- }
- }
- /* We do not need this anymore. */
- if (authenticate_h)
+ if (resp_header_copy (resp, "Content-Range", hdrval, sizeof (hdrval)))
{
- xfree (authenticate_h);
- authenticate_h = NULL;
+ wgint first_byte_pos, last_byte_pos, entity_length;
+ if (parse_content_range (hdrval, &first_byte_pos, &last_byte_pos,
+ &entity_length))
+ contrange = first_byte_pos;
}
+ resp_free (resp);
/* 20x responses are counted among successful by default. */
if (H_20X (statcode))
{
logprintf (LOG_VERBOSE,
_("Location: %s%s\n"),
- hs->newloc ? hs->newloc : _("unspecified"),
+ hs->newloc ? escnonprint_uri (hs->newloc) : _("unspecified"),
hs->newloc ? _(" [following]") : "");
- CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
- might be more bytes in the body. */
+ if (keep_alive && !head_only && skip_short_body (sock, contlen))
+ CLOSE_FINISH (sock);
+ else
+ CLOSE_INVALIDATE (sock);
xfree_null (type);
return NEWLOCATION;
}
text/html file. If some case-insensitive variation on ".htm[l]" isn't
already the file's suffix, tack on ".html". */
{
- char* last_period_in_local_filename = strrchr(*hs->local_file, '.');
+ char *last_period_in_local_filename = strrchr (*hs->local_file, '.');
if (last_period_in_local_filename == NULL
|| !(0 == strcasecmp (last_period_in_local_filename, ".htm")
|| 0 == strcasecmp (last_period_in_local_filename, ".html")))
{
- size_t local_filename_len = strlen(*hs->local_file);
-
- *hs->local_file = xrealloc(*hs->local_file,
- local_filename_len + sizeof(".html"));
+ int local_filename_len = strlen (*hs->local_file);
+ /* Resize the local file, allowing for ".html" preceded by
+ optional ".NUMBER". */
+ *hs->local_file = xrealloc (*hs->local_file,
+ local_filename_len + 24 + sizeof (".html"));
strcpy(*hs->local_file + local_filename_len, ".html");
-
+ /* If clobbering is not allowed and the file, as named,
+ exists, tack on ".NUMBER.html" instead. */
+ if (!ALLOW_CLOBBER)
+ {
+ int ext_num = 1;
+ do
+ sprintf (*hs->local_file + local_filename_len,
+ ".%d.html", ext_num++);
+ while (file_exists_p (*hs->local_file));
+ }
*dt |= ADDED_HTML_EXTENSION;
}
}
- if (contrange == -1)
+ if (statcode == HTTP_STATUS_RANGE_NOT_SATISFIABLE)
{
- /* We did not get a content-range header. This means that the
- server did not honor our `Range' request. Normally, this
- means we should reset hs->restval and continue normally. */
-
- /* However, if `-c' is used, we need to be a bit more careful:
-
- 1. If `-c' is specified and the file already existed when
- Wget was started, it would be a bad idea for us to start
- downloading it from scratch, effectively truncating it. I
- believe this cannot happen unless `-c' was specified.
-
- 2. If `-c' is used on a file that is already fully
- downloaded, we're requesting bytes after the end of file,
- which can result in server not honoring `Range'. If this is
- the case, `Content-Length' will be equal to the length of the
- file. */
- if (opt.always_rest)
- {
- /* Check for condition #2. */
- if (hs->restval > 0 /* restart was requested. */
- && contlen != -1 /* we got content-length. */
- && hs->restval >= contlen /* file fully downloaded
- or has shrunk. */
- )
- {
- logputs (LOG_VERBOSE, _("\
+ /* If `-c' is in use and the file has been fully downloaded (or
+ the remote file has shrunk), Wget effectively requests bytes
+ after the end of file and the server response with 416. */
+ logputs (LOG_VERBOSE, _("\
\n The file is already fully retrieved; nothing to do.\n\n"));
- /* In case the caller inspects. */
- hs->len = contlen;
- hs->res = 0;
- /* Mark as successfully retrieved. */
- *dt |= RETROKF;
- xfree_null (type);
- CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
- might be more bytes in the body. */
- return RETRUNNEEDED;
- }
-
- /* Check for condition #1. */
- if (hs->no_truncate)
- {
- logprintf (LOG_NOTQUIET,
- _("\
-\n\
-Continued download failed on this file, which conflicts with `-c'.\n\
-Refusing to truncate existing file `%s'.\n\n"), *hs->local_file);
- xfree_null (type);
- CLOSE_INVALIDATE (sock);
- return CONTNOTSUPPORTED;
- }
-
- /* Fallthrough */
- }
-
- hs->restval = 0;
+ /* In case the caller inspects. */
+ hs->len = contlen;
+ hs->res = 0;
+ /* Mark as successfully retrieved. */
+ *dt |= RETROKF;
+ xfree_null (type);
+ CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
+ might be more bytes in the body. */
+ return RETRUNNEEDED;
}
- else if (contrange != hs->restval ||
- (H_PARTIAL (statcode) && contrange == -1))
+ if ((contrange != 0 && contrange != hs->restval)
+ || (H_PARTIAL (statcode) && !contrange))
{
- /* This means the whole request was somehow misunderstood by the
- server. Bail out. */
+ /* The Range request was somehow misunderstood by the server.
+ Bail out. */
xfree_null (type);
CLOSE_INVALIDATE (sock);
return RANGEERR;
}
-
- if (hs->restval)
- {
- if (contlen != -1)
- contlen += contrange;
- else
- contrange = -1; /* If conent-length was not sent,
- content-range will be ignored. */
- }
- hs->contlen = contlen;
+ hs->contlen = contlen + contrange;
if (opt.verbose)
{
- if ((*dt & RETROKF) && !opt.server_response)
+ if (*dt & RETROKF)
{
/* No need to print this output if the body won't be
downloaded at all, or if the original server response is
logputs (LOG_VERBOSE, _("Length: "));
if (contlen != -1)
{
- logputs (LOG_VERBOSE, legible (contlen));
- if (contrange != -1)
- logprintf (LOG_VERBOSE, _(" (%s to go)"),
- legible (contlen - contrange));
+ logputs (LOG_VERBOSE, number_to_static_string (contlen + contrange));
+ if (contlen + contrange >= 1024)
+ logprintf (LOG_VERBOSE, " (%s)",
+ human_readable (contlen + contrange));
+ if (contrange)
+ {
+ if (contlen >= 1024)
+ logprintf (LOG_VERBOSE, _(", %s (%s) remaining"),
+ number_to_static_string (contlen),
+ human_readable (contlen));
+ else
+ logprintf (LOG_VERBOSE, _(", %s remaining"),
+ number_to_static_string (contlen));
+ }
}
else
logputs (LOG_VERBOSE,
opt.ignore_length ? _("ignored") : _("unspecified"));
if (type)
- logprintf (LOG_VERBOSE, " [%s]\n", type);
+ logprintf (LOG_VERBOSE, " [%s]\n", escnonprint (type));
else
logputs (LOG_VERBOSE, "\n");
}
type = NULL; /* We don't need it any more. */
/* Return if we have no intention of further downloading. */
- if (!(*dt & RETROKF) || (*dt & HEAD_ONLY))
+ if (!(*dt & RETROKF) || head_only)
{
/* In case the caller cares to look... */
- hs->len = 0L;
+ hs->len = 0;
hs->res = 0;
xfree_null (type);
- CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
- might be more bytes in the body. */
+ /* Pre-1.10 Wget used CLOSE_INVALIDATE here. Now we trust the
+ servers not to send body in response to a HEAD request. If
+ you encounter such a server (more likely a broken CGI), use
+ `--no-http-keep-alive'. */
+ CLOSE_FINISH (sock);
return RETRFINISHED;
}
/* Open the local file. */
- if (!opt.dfp)
+ if (!output_stream)
{
mkalldirs (*hs->local_file);
if (opt.backups)
rotate_backups (*hs->local_file);
- fp = fopen (*hs->local_file, hs->restval ? "ab" : "wb");
+ if (hs->restval)
+ fp = fopen (*hs->local_file, "ab");
+ else if (ALLOW_CLOBBER)
+ fp = fopen (*hs->local_file, "wb");
+ else
+ {
+ fp = fopen_excl (*hs->local_file, true);
+ if (!fp && errno == EEXIST)
+ {
+ /* We cannot just invent a new name and use it (which is
+ what functions like unique_create typically do)
+ because we told the user we'd use this name.
+ Instead, return and retry the download. */
+ logprintf (LOG_NOTQUIET,
+ _("%s has sprung into existence.\n"),
+ *hs->local_file);
+ CLOSE_INVALIDATE (sock);
+ return FOPEN_EXCL_ERR;
+ }
+ }
if (!fp)
{
logprintf (LOG_NOTQUIET, "%s: %s\n", *hs->local_file, strerror (errno));
- CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there
- might be more bytes in the body. */
+ CLOSE_INVALIDATE (sock);
return FOPENERR;
}
}
- else /* opt.dfp */
- {
- extern int global_download_count;
- fp = opt.dfp;
- /* To ensure that repeated "from scratch" downloads work for -O
- files, we rewind the file pointer, unless restval is
- non-zero. (This works only when -O is used on regular files,
- but it's still a valuable feature.)
-
- However, this loses when more than one URL is specified on
- the command line the second rewinds eradicates the contents
- of the first download. Thus we disable the above trick for
- all the downloads except the very first one.
-
- #### A possible solution to this would be to remember the
- file position in the output document and to seek to that
- position, instead of rewinding.
-
- We don't truncate stdout, since that breaks
- "wget -O - [...] >> foo".
- */
- if (!hs->restval && global_download_count == 0 && opt.dfp != stdout)
- {
- /* This will silently fail for streams that don't correspond
- to regular files, but that's OK. */
- rewind (fp);
- /* ftruncate is needed because opt.dfp is opened in append
- mode if opt.always_rest is set. */
- ftruncate (fileno (fp), 0);
- clearerr (fp);
- }
- }
+ else
+ fp = output_stream;
- /* #### This confuses the code that checks for file size. There
- should be some overhead information. */
+ /* #### This confuses the timestamping code that checks for file
+ size. Maybe we should save some additional information? */
if (opt.save_headers)
fwrite (head, 1, strlen (head), fp);
- /* Get the contents of the document. */
- hs->res = fd_read_body (sock, fp, &hs->len, hs->restval,
- (contlen != -1 ? contlen : 0),
- keep_alive, &hs->dltime);
+ /* Now we no longer need to store the response header. */
+ xfree (head);
+
+ /* Download the request body. */
+ flags = 0;
+ if (keep_alive)
+ flags |= rb_read_exactly;
+ if (hs->restval > 0 && contrange == 0)
+ /* If the server ignored our range request, instruct fd_read_body
+ to skip the first RESTVAL bytes of body. */
+ flags |= rb_skip_startpos;
+ hs->len = hs->restval;
+ hs->rd_size = 0;
+ hs->res = fd_read_body (sock, fp, contlen != -1 ? contlen : 0,
+ hs->restval, &hs->rd_size, &hs->len, &hs->dltime,
+ flags);
if (hs->res >= 0)
CLOSE_FINISH (sock);
error here. Checking the result of fwrite() is not enough --
errors could go unnoticed! */
int flush_res;
- if (!opt.dfp)
+ if (!output_stream)
flush_res = fclose (fp);
else
flush_res = fflush (fp);
int *dt, struct url *proxy)
{
int count;
- int use_ts, got_head = 0; /* time-stamping info */
+ bool use_ts, got_head = false;/* time-stamping info */
char *filename_plus_orig_suffix;
char *local_filename = NULL;
- char *tms, *locf, *tmrate;
+ char *tms, *locf;
+ const char *tmrate;
uerr_t err;
time_t tml = -1, tmr = -1; /* local and remote time-stamps */
- long local_size = 0; /* the size of the local file */
+ wgint local_size = 0; /* the size of the local file */
size_t filename_len;
struct http_stat hstat; /* HTTP status */
- struct stat st;
+ struct_stat st;
char *dummy = NULL;
/* This used to be done in main(), but it's a better idea to do it
if (opt.cookies_input && !cookies_loaded_p)
{
cookie_jar_load (wget_cookie_jar, opt.cookies_input);
- cookies_loaded_p = 1;
+ cookies_loaded_p = true;
}
}
*newloc = NULL;
- /* Warn on (likely bogus) wildcard usage in HTTP. Don't use
- has_wildcards_p because it would also warn on `?', and we know that
- shows up in CGI paths a *lot*. */
- if (strchr (u->url, '*'))
+ /* Warn on (likely bogus) wildcard usage in HTTP. */
+ if (opt.ftp_glob && has_wildcards_p (u->path))
logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n"));
+ xzero (hstat);
+
/* Determine the local filename. */
if (local_file && *local_file)
hstat.local_file = local_file;
- else if (local_file)
+ else if (local_file && !opt.output_document)
{
*local_file = url_file_name (u);
hstat.local_file = local_file;
{
dummy = url_file_name (u);
hstat.local_file = &dummy;
+ /* be honest about where we will save the file */
+ if (local_file && opt.output_document)
+ *local_file = HYPHENP (opt.output_document) ? NULL : xstrdup (opt.output_document);
}
if (!opt.output_document)
/* If opt.noclobber is turned on and file already exists, do not
retrieve the file */
logprintf (LOG_VERBOSE, _("\
-File `%s' already there, will not retrieve.\n"), *hstat.local_file);
+File `%s' already there; not retrieving.\n\n"), *hstat.local_file);
/* If the file is there, we suppose it's retrieved OK. */
*dt |= RETROKF;
return RETROK;
}
- use_ts = 0;
+ use_ts = false;
if (opt.timestamping)
{
- int local_dot_orig_file_exists = 0;
+ bool local_dot_orig_file_exists = false;
if (opt.backup_converted)
/* If -K is specified, we'll act on the assumption that it was specified
point I profiled Wget, and found that a measurable and
non-negligible amount of time was lost calling sprintf()
in url.c. Replacing sprintf with inline calls to
- strcpy() and long_to_string() made a difference.
+ strcpy() and number_to_string() made a difference.
--hniksic */
memcpy (filename_plus_orig_suffix, *hstat.local_file, filename_len);
memcpy (filename_plus_orig_suffix + filename_len,
the server has is the same version we already have, allowing us to
skip a download. */
{
- use_ts = 1;
+ use_ts = true;
tml = st.st_mtime;
#ifdef WINDOWS
/* Modification time granularity is 2 seconds for Windows, so
tml++;
#endif
local_size = st.st_size;
- got_head = 0;
+ got_head = false;
}
}
/* Reset the counter. */
count = 0;
- *dt = 0 | ACCEPTRANGES;
+ *dt = 0;
/* THE loop */
do
{
/* Print fetch message, if opt.verbose. */
if (opt.verbose)
{
- char *hurl = url_string (u, 1);
- char tmp[15];
+ char *hurl = url_string (u, true);
+ char tmp[256];
strcpy (tmp, " ");
if (count > 1)
sprintf (tmp, _("(try:%2d)"), count);
logprintf (LOG_VERBOSE, "--%s-- %s\n %s => `%s'\n",
tms, hurl, tmp, locf);
#ifdef WINDOWS
- ws_changetitle (hurl, 1);
+ ws_changetitle (hurl);
#endif
xfree (hurl);
}
*dt |= HEAD_ONLY;
else
*dt &= ~HEAD_ONLY;
- /* Assume no restarting. */
- hstat.restval = 0L;
+
/* Decide whether or not to restart. */
- if (((count > 1 && (*dt & ACCEPTRANGES)) || opt.always_rest)
- /* #### this calls access() and then stat(); could be optimized. */
- && file_exists_p (locf))
- if (stat (locf, &st) == 0 && S_ISREG (st.st_mode))
- hstat.restval = st.st_size;
-
- /* In `-c' is used and the file is existing and non-empty,
- refuse to truncate it if the server doesn't support continued
- downloads. */
- hstat.no_truncate = 0;
- if (opt.always_rest && hstat.restval)
- hstat.no_truncate = 1;
+ if (opt.always_rest
+ && stat (locf, &st) == 0
+ && S_ISREG (st.st_mode))
+ /* When -c is used, continue from on-disk size. (Can't use
+ hstat.len even if count>1 because we don't want a failed
+ first attempt to clobber existing data.) */
+ hstat.restval = st.st_size;
+ else if (count > 1)
+ /* otherwise, continue where the previous try left off */
+ hstat.restval = hstat.len;
+ else
+ hstat.restval = 0;
/* Decide whether to send the no-cache directive. We send it in
two cases:
*hstat.local_file to tack on ".html". */
if (!opt.output_document)
locf = *hstat.local_file;
- else
- locf = opt.output_document;
/* Time? */
tms = time_str (NULL);
{
case HERR: case HEOF: case CONSOCKERR: case CONCLOSED:
case CONERROR: case READERR: case WRITEFAILED:
- case RANGEERR:
+ case RANGEERR: case FOPEN_EXCL_ERR:
/* Non-fatal errors continue executing the loop, which will
bring them to "while" statement at the end, to judge
whether the number of tries was exceeded. */
free_hstat (&hstat);
printwhat (count, opt.ntry);
+ if (err == FOPEN_EXCL_ERR)
+ {
+ /* Re-determine the file name. */
+ if (local_file && *local_file)
+ {
+ xfree (*local_file);
+ *local_file = url_file_name (u);
+ hstat.local_file = local_file;
+ }
+ else
+ {
+ xfree (dummy);
+ dummy = url_file_name (u);
+ hstat.local_file = &dummy;
+ }
+ /* be honest about where we will save the file */
+ if (local_file && opt.output_document)
+ *local_file = HYPHENP (opt.output_document) ? NULL : xstrdup (opt.output_document);
+ if (!opt.output_document)
+ locf = *hstat.local_file;
+ else
+ locf = opt.output_document;
+ }
continue;
- break;
case HOSTERR: case CONIMPOSSIBLE: case PROXERR: case AUTHFAILED:
- case SSLERRCTXCREATE: case CONTNOTSUPPORTED:
+ case SSLINITFAILED: case CONTNOTSUPPORTED:
/* Fatal errors just return from the function. */
free_hstat (&hstat);
xfree_null (dummy);
return err;
- break;
case FWRITEERR: case FOPENERR:
/* Another fatal error. */
logputs (LOG_VERBOSE, "\n");
free_hstat (&hstat);
xfree_null (dummy);
return err;
- break;
case CONSSLERR:
/* Another fatal error. */
- logputs (LOG_VERBOSE, "\n");
logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n"));
free_hstat (&hstat);
xfree_null (dummy);
return err;
- break;
case NEWLOCATION:
/* Return the new location to the caller. */
if (!hstat.newloc)
free_hstat (&hstat);
xfree_null (dummy);
return NEWLOCATION;
- break;
case RETRUNNEEDED:
/* The file was already fully retrieved. */
free_hstat (&hstat);
xfree_null (dummy);
return RETROK;
- break;
case RETRFINISHED:
/* Deal with you later. */
break;
if (!opt.verbose)
{
/* #### Ugly ugly ugly! */
- char *hurl = url_string (u, 1);
+ char *hurl = url_string (u, true);
logprintf (LOG_NONVERBOSE, "%s:\n", hurl);
xfree (hurl);
}
logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"),
- tms, hstat.statcode, hstat.error);
+ tms, hstat.statcode, escnonprint (hstat.error));
logputs (LOG_VERBOSE, "\n");
free_hstat (&hstat);
xfree_null (dummy);
/* The time-stamping section. */
if (use_ts)
{
- got_head = 1;
+ got_head = true;
*dt &= ~HEAD_ONLY;
- use_ts = 0; /* no more time-stamping */
+ use_ts = false; /* no more time-stamping */
count = 0; /* the retrieve count for HEAD is
reset */
if (hstat.remote_time && tmr != (time_t) (-1))
}
else if (tml >= tmr)
logprintf (LOG_VERBOSE, _("\
-The sizes do not match (local %ld) -- retrieving.\n"), local_size);
+The sizes do not match (local %s) -- retrieving.\n"),
+ number_to_static_string (local_size));
else
logputs (LOG_VERBOSE,
_("Remote file is newer, retrieving.\n"));
const char *fl = NULL;
if (opt.output_document)
{
- if (opt.od_known_regular)
+ if (output_stream_regular)
fl = opt.output_document;
}
else
if (opt.spider)
{
- logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode, hstat.error);
+ logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode,
+ escnonprint (hstat.error));
xfree_null (dummy);
return RETROK;
}
- tmrate = retr_rate (hstat.len - hstat.restval, hstat.dltime, 0);
+ tmrate = retr_rate (hstat.rd_size, hstat.dltime);
+ total_download_time += hstat.dltime;
if (hstat.len == hstat.contlen)
{
if (*dt & RETROKF)
{
logprintf (LOG_VERBOSE,
- _("%s (%s) - `%s' saved [%ld/%ld]\n\n"),
- tms, tmrate, locf, hstat.len, hstat.contlen);
+ _("%s (%s) - `%s' saved [%s/%s]\n\n"),
+ tms, tmrate, locf,
+ number_to_static_string (hstat.len),
+ number_to_static_string (hstat.contlen));
logprintf (LOG_NONVERBOSE,
- "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
- tms, u->url, hstat.len, hstat.contlen, locf, count);
+ "%s URL:%s [%s/%s] -> \"%s\" [%d]\n",
+ tms, u->url,
+ number_to_static_string (hstat.len),
+ number_to_static_string (hstat.contlen),
+ locf, count);
}
++opt.numurls;
total_downloaded_bytes += hstat.len;
if (*dt & RETROKF)
{
logprintf (LOG_VERBOSE,
- _("%s (%s) - `%s' saved [%ld]\n\n"),
- tms, tmrate, locf, hstat.len);
+ _("%s (%s) - `%s' saved [%s]\n\n"),
+ tms, tmrate, locf,
+ number_to_static_string (hstat.len));
logprintf (LOG_NONVERBOSE,
- "%s URL:%s [%ld] -> \"%s\" [%d]\n",
- tms, u->url, hstat.len, locf, count);
+ "%s URL:%s [%s] -> \"%s\" [%d]\n",
+ tms, u->url, number_to_static_string (hstat.len),
+ locf, count);
}
++opt.numurls;
total_downloaded_bytes += hstat.len;
connection too soon */
{
logprintf (LOG_VERBOSE,
- _("%s (%s) - Connection closed at byte %ld. "),
- tms, tmrate, hstat.len);
+ _("%s (%s) - Connection closed at byte %s. "),
+ tms, tmrate, number_to_static_string (hstat.len));
printwhat (count, opt.ntry);
free_hstat (&hstat);
continue;
else if (!opt.kill_longer) /* meaning we got more than expected */
{
logprintf (LOG_VERBOSE,
- _("%s (%s) - `%s' saved [%ld/%ld])\n\n"),
- tms, tmrate, locf, hstat.len, hstat.contlen);
+ _("%s (%s) - `%s' saved [%s/%s]\n\n"),
+ tms, tmrate, locf,
+ number_to_static_string (hstat.len),
+ number_to_static_string (hstat.contlen));
logprintf (LOG_NONVERBOSE,
- "%s URL:%s [%ld/%ld] -> \"%s\" [%d]\n",
- tms, u->url, hstat.len, hstat.contlen, locf, count);
+ "%s URL:%s [%s/%s] -> \"%s\" [%d]\n",
+ tms, u->url,
+ number_to_static_string (hstat.len),
+ number_to_static_string (hstat.contlen),
+ locf, count);
++opt.numurls;
total_downloaded_bytes += hstat.len;
else /* the same, but not accepted */
{
logprintf (LOG_VERBOSE,
- _("%s (%s) - Connection closed at byte %ld/%ld. "),
- tms, tmrate, hstat.len, hstat.contlen);
+ _("%s (%s) - Connection closed at byte %s/%s. "),
+ tms, tmrate,
+ number_to_static_string (hstat.len),
+ number_to_static_string (hstat.contlen));
printwhat (count, opt.ntry);
free_hstat (&hstat);
continue;
if (hstat.contlen == -1)
{
logprintf (LOG_VERBOSE,
- _("%s (%s) - Read error at byte %ld (%s)."),
- tms, tmrate, hstat.len, strerror (errno));
+ _("%s (%s) - Read error at byte %s (%s)."),
+ tms, tmrate, number_to_static_string (hstat.len),
+ strerror (errno));
printwhat (count, opt.ntry);
free_hstat (&hstat);
continue;
else /* hstat.res == -1 and contlen is given */
{
logprintf (LOG_VERBOSE,
- _("%s (%s) - Read error at byte %ld/%ld (%s). "),
- tms, tmrate, hstat.len, hstat.contlen,
+ _("%s (%s) - Read error at byte %s/%s (%s). "),
+ tms, tmrate,
+ number_to_static_string (hstat.len),
+ number_to_static_string (hstat.contlen),
strerror (errno));
printwhat (count, opt.ntry);
free_hstat (&hstat);
}
}
/* not reached */
- break;
}
while (!opt.ntry || (count < opt.ntry));
return TRYLIMEXC;
In extended regexp parlance, the function returns 1 if P matches
"^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (which strptime
can return) is considered a failure and 0 is returned. */
-static int
+static bool
check_end (const char *p)
{
if (!p)
- return 0;
+ return false;
while (ISSPACE (*p))
++p;
if (!*p
|| (p[0] == 'G' && p[1] == 'M' && p[2] == 'T')
|| ((p[0] == '+' || p[0] == '-') && ISDIGIT (p[1])))
- return 1;
+ return true;
else
- return 0;
+ return false;
}
/* Convert the textual specification of time in TIME_STRING to the
number of seconds since the Epoch.
- TIME_STRING can be in any of the three formats RFC2068 allows the
- HTTP servers to emit -- RFC1123-date, RFC850-date or asctime-date.
+ TIME_STRING can be in any of the three formats RFC2616 allows the
+ HTTP servers to emit -- RFC1123-date, RFC850-date or asctime-date,
+ as well as the time format used in the Set-Cookie header.
Timezones are ignored, and should be GMT.
Return the computed time_t representation, or -1 if the conversion
implementations I've tested. */
static const char *time_formats[] = {
- "%a, %d %b %Y %T", /* RFC1123: Thu, 29 Jan 1998 22:12:57 */
- "%A, %d-%b-%y %T", /* RFC850: Thursday, 29-Jan-98 22:12:57 */
- "%a, %d-%b-%Y %T", /* pseudo-RFC850: Thu, 29-Jan-1998 22:12:57
- (google.com uses this for their cookies.) */
- "%a %b %d %T %Y" /* asctime: Thu Jan 29 22:12:57 1998 */
+ "%a, %d %b %Y %T", /* rfc1123: Thu, 29 Jan 1998 22:12:57 */
+ "%A, %d-%b-%y %T", /* rfc850: Thursday, 29-Jan-98 22:12:57 */
+ "%a %b %d %T %Y", /* asctime: Thu Jan 29 22:12:57 1998 */
+ "%a, %d-%b-%Y %T" /* cookies: Thu, 29-Jan-1998 22:12:57
+ (used in Set-Cookie, defined in the
+ Netscape cookie specification.) */
};
-
int i;
- struct tm t;
-
- /* According to Roger Beeman, we need to initialize tm_isdst, since
- strptime won't do it. */
- t.tm_isdst = 0;
-
- /* Note that under foreign locales Solaris strptime() fails to
- recognize English dates, which renders this function useless. We
- solve this by being careful not to affect LC_TIME when
- initializing locale.
-
- Another solution would be to temporarily set locale to C, invoke
- strptime(), and restore it back. This is slow and dirty,
- however, and locale support other than LC_MESSAGES can mess other
- things, so I rather chose to stick with just setting LC_MESSAGES.
-
- GNU strptime does not have this problem because it recognizes
- both international and local dates. */
for (i = 0; i < countof (time_formats); i++)
- if (check_end (strptime (time_string, time_formats[i], &t)))
- return mktime_from_utc (&t);
+ {
+ struct tm t;
+
+ /* Some versions of strptime use the existing contents of struct
+ tm to recalculate the date according to format. Zero it out
+ to prevent garbage from the stack influencing strptime. */
+ xzero (t);
+
+ /* Solaris strptime fails to recognize English month names in
+ non-English locales, which we work around by not setting the
+ LC_TIME category. Another way would be to temporarily set
+ locale to C before invoking strptime, but that's slow and
+ messy. GNU strptime does not have this problem because it
+ recognizes English month names along with the local ones. */
+
+ if (check_end (strptime (time_string, time_formats[i], &t)))
+ return mktime_from_utc (&t);
+ }
/* All formats have failed. */
return -1;
}
\f
-/* Authorization support: We support two authorization schemes:
+/* Authorization support: We support three authorization schemes:
* `Basic' scheme, consisting of base64-ing USER:PASSWORD string;
* `Digest' scheme, added by Junio Hamano <junio@twinsun.com>,
consisting of answering to the server's challenge with the proper
- MD5 digests. */
-
-/* How many bytes it will take to store LEN bytes in base64. */
-#define BASE64_LENGTH(len) (4 * (((len) + 2) / 3))
-
-/* Encode the string S of length LENGTH to base64 format and place it
- to STORE. STORE will be 0-terminated, and must point to a writable
- buffer of at least 1+BASE64_LENGTH(length) bytes. */
-static void
-base64_encode (const char *s, char *store, int length)
-{
- /* Conversion table. */
- static char tbl[64] = {
- 'A','B','C','D','E','F','G','H',
- 'I','J','K','L','M','N','O','P',
- 'Q','R','S','T','U','V','W','X',
- 'Y','Z','a','b','c','d','e','f',
- 'g','h','i','j','k','l','m','n',
- 'o','p','q','r','s','t','u','v',
- 'w','x','y','z','0','1','2','3',
- '4','5','6','7','8','9','+','/'
- };
- int i;
- unsigned char *p = (unsigned char *)store;
+ MD5 digests.
- /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
- for (i = 0; i < length; i += 3)
- {
- *p++ = tbl[s[0] >> 2];
- *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
- *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
- *p++ = tbl[s[2] & 0x3f];
- s += 3;
- }
- /* Pad the result if necessary... */
- if (i == length + 1)
- *(p - 1) = '=';
- else if (i == length + 2)
- *(p - 1) = *(p - 2) = '=';
- /* ...and zero-terminate it. */
- *p = '\0';
-}
+ * `NTLM' ("NT Lan Manager") scheme, based on code written by Daniel
+ Stenberg for libcurl. Like digest, NTLM is based on a
+ challenge-response mechanism, but unlike digest, it is non-standard
+ (authenticates TCP connections rather than requests), undocumented
+ and Microsoft-specific. */
/* Create the authentication header contents for the `Basic' scheme.
- This is done by encoding the string `USER:PASS' in base64 and
- prepending `HEADER: Basic ' to it. */
+ This is done by encoding the string "USER:PASS" to base64 and
+ prepending the string "Basic " in front of it. */
+
static char *
-basic_authentication_encode (const char *user, const char *passwd,
- const char *header)
+basic_authentication_encode (const char *user, const char *passwd)
{
- char *t1, *t2, *res;
+ char *t1, *t2;
int len1 = strlen (user) + 1 + strlen (passwd);
- int len2 = BASE64_LENGTH (len1);
t1 = (char *)alloca (len1 + 1);
sprintf (t1, "%s:%s", user, passwd);
- t2 = (char *)alloca (1 + len2);
- base64_encode (t1, t2, len1);
- res = (char *)xmalloc (len2 + 11 + strlen (header));
- sprintf (res, "%s: Basic %s\r\n", header, t2);
- return res;
+ t2 = (char *)alloca (BASE64_LENGTH (len1) + 1);
+ base64_encode (t1, len1, t2);
+
+ return concat_strings ("Basic ", t2, (char *) 0);
}
#define SKIP_WS(x) do { \
++(x); \
} while (0)
-#ifdef USE_DIGEST
+#ifdef ENABLE_DIGEST
/* Parse HTTP `WWW-Authenticate:' header. AU points to the beginning
of a field in such a header. If the field is the one specified by
ATTR_NAME ("realm", "opaque", and "nonce" are used by the current
static int
extract_header_attr (const char *au, const char *attr_name, char **ret)
{
- const char *cp, *ep;
-
- ep = cp = au;
+ const char *ep;
+ const char *cp = au;
if (strncmp (cp, attr_name, strlen (attr_name)) == 0)
{
gen_md5_finish (ctx, hash);
dump_hash (response_digest, hash);
- res = (char*) xmalloc (strlen (user)
- + strlen (user)
- + strlen (realm)
- + strlen (nonce)
- + strlen (path)
- + 2 * MD5_HASHLEN /*strlen (response_digest)*/
- + (opaque ? strlen (opaque) : 0)
- + 128);
- sprintf (res, "Authorization: Digest \
+ res = xmalloc (strlen (user)
+ + strlen (user)
+ + strlen (realm)
+ + strlen (nonce)
+ + strlen (path)
+ + 2 * MD5_HASHLEN /*strlen (response_digest)*/
+ + (opaque ? strlen (opaque) : 0)
+ + 128);
+ sprintf (res, "Digest \
username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"%s\", response=\"%s\"",
user, realm, nonce, path, response_digest);
if (opaque)
strcat (p, opaque);
strcat (p, "\"");
}
- strcat (res, "\r\n");
}
return res;
}
-#endif /* USE_DIGEST */
-
-
-#define BEGINS_WITH(line, string_constant) \
- (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \
- && (ISSPACE (line[sizeof (string_constant) - 1]) \
- || !line[sizeof (string_constant) - 1]))
-
-static int
-known_authentication_scheme_p (const char *au)
+#endif /* ENABLE_DIGEST */
+
+/* Computing the size of a string literal must take into account that
+ value returned by sizeof includes the terminating \0. */
+#define STRSIZE(literal) (sizeof (literal) - 1)
+
+/* Whether chars in [b, e) begin with the literal string provided as
+ first argument and are followed by whitespace or terminating \0.
+ The comparison is case-insensitive. */
+#define STARTS(literal, b, e) \
+ ((e) - (b) >= STRSIZE (literal) \
+ && 0 == strncasecmp (b, literal, STRSIZE (literal)) \
+ && ((e) - (b) == STRSIZE (literal) \
+ || ISSPACE (b[STRSIZE (literal)])))
+
+static bool
+known_authentication_scheme_p (const char *hdrbeg, const char *hdrend)
{
- return BEGINS_WITH (au, "Basic")
- || BEGINS_WITH (au, "Digest")
- || BEGINS_WITH (au, "NTLM");
+ return STARTS ("Basic", hdrbeg, hdrend)
+#ifdef ENABLE_DIGEST
+ || STARTS ("Digest", hdrbeg, hdrend)
+#endif
+#ifdef ENABLE_NTLM
+ || STARTS ("NTLM", hdrbeg, hdrend)
+#endif
+ ;
}
-#undef BEGINS_WITH
+#undef STARTS
/* Create the HTTP authorization request header. When the
`WWW-Authenticate' response header is seen, according to the
static char *
create_authorization_line (const char *au, const char *user,
const char *passwd, const char *method,
- const char *path)
+ const char *path, bool *finished)
{
- char *wwwauth = NULL;
-
- if (!strncasecmp (au, "Basic", 5))
- wwwauth = basic_authentication_encode (user, passwd, "Authorization");
- if (!strncasecmp (au, "NTLM", 4))
- wwwauth = basic_authentication_encode (user, passwd, "Authorization");
-#ifdef USE_DIGEST
- else if (!strncasecmp (au, "Digest", 6))
- wwwauth = digest_authentication_encode (au, user, passwd, method, path);
-#endif /* USE_DIGEST */
- return wwwauth;
+ /* We are called only with known schemes, so we can dispatch on the
+ first letter. */
+ switch (TOUPPER (*au))
+ {
+ case 'B': /* Basic */
+ *finished = true;
+ return basic_authentication_encode (user, passwd);
+#ifdef ENABLE_DIGEST
+ case 'D': /* Digest */
+ *finished = true;
+ return digest_authentication_encode (au, user, passwd, method, path);
+#endif
+#ifdef ENABLE_NTLM
+ case 'N': /* NTLM */
+ if (!ntlm_input (&pconn.ntlm, au))
+ {
+ *finished = true;
+ return NULL;
+ }
+ return ntlm_output (&pconn.ntlm, user, passwd, finished);
+#endif
+ default:
+ /* We shouldn't get here -- this function should be only called
+ with values approved by known_authentication_scheme_p. */
+ abort ();
+ }
}
\f
+void
+save_cookies (void)
+{
+ if (wget_cookie_jar)
+ cookie_jar_save (wget_cookie_jar, opt.cookies_output);
+}
+
void
http_cleanup (void)
{
+ xfree_null (pconn.host);
+ if (wget_cookie_jar)
+ cookie_jar_delete (wget_cookie_jar);
}
projects / wget / blobdiff