/* HTTP support.
- Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001, 2002
- Free Software Foundation, Inc.
+ Copyright (C) 2003 Free Software Foundation, Inc.
This file is part of GNU Wget.
#include "url.h"
#include "host.h"
#include "retr.h"
-#include "headers.h"
#include "connect.h"
#include "netrc.h"
#ifdef HAVE_SSL
extern char *version_string;
extern LARGE_INT total_downloaded_bytes;
+extern FILE *output_stream;
+extern int 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;
#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
-/* Parse the HTTP status line, which is of format:
+enum rp {
+ rel_none, rel_name, rel_value, rel_both
+};
- HTTP-Version SP Status-Code SP Reason-Phrase
+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 ()
+{
+ 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.
+
+ 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=%ld-", 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)
+ 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->hcount)
+ {
+ req->hcapacity <<= 1;
+ req->headers = xrealloc (req->headers,
+ req->hcapacity * sizeof (struct request_header));
+ }
+ 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);
+}
+
+#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. */
- The function returns the status-code, or -1 if the status line is
- malformed. The pointer to reason-phrase is returned in RP. */
static int
-parse_http_status_line (const char *line, const char **reason_phrase_ptr)
+request_send (const struct request *req, int fd)
{
- /* (the variables must not be named `major' and `minor', because
- that breaks compilation with SunOS4 cc.) */
- int mjr, mnr, statcode;
- const char *p;
+ char *request_string, *p;
+ int i, size, write_error;
- *reason_phrase_ptr = NULL;
+ /* Count the request size. */
+ size = 0;
- /* The standard format of HTTP-Version is: `HTTP/X.Y', where X is
- major version, and Y is minor version. */
- if (strncmp (line, "HTTP/", 5) != 0)
- return -1;
- line += 5;
+ /* METHOD " " ARG " " "HTTP/1.0" "\r\n" */
+ size += strlen (req->method) + 1 + strlen (req->arg) + 1 + 8 + 2;
- /* Calculate major HTTP version. */
- p = line;
- for (mjr = 0; ISDIGIT (*line); line++)
- mjr = 10 * mjr + (*line - '0');
- if (*line != '.' || p == line)
- return -1;
- ++line;
+ 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;
+ }
- /* Calculate minor HTTP version. */
- p = line;
- for (mnr = 0; ISDIGIT (*line); line++)
- mnr = 10 * mnr + (*line - '0');
- if (*line != ' ' || p == line)
- return -1;
- /* Wget will accept only 1.0 and higher HTTP-versions. The value of
- minor version can be safely ignored. */
- if (mjr < 1)
- return -1;
- ++line;
+ /* "\r\n\0" */
+ size += 3;
- /* Calculate status code. */
- if (!(ISDIGIT (*line) && ISDIGIT (line[1]) && ISDIGIT (line[2])))
- return -1;
- statcode = 100 * (*line - '0') + 10 * (line[1] - '0') + (line[2] - '0');
+ 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;
- /* Set up the reason phrase pointer. */
- line += 3;
- /* RFC2068 requires SPC here, but we allow the string to finish
- here, in case no reason-phrase is present. */
- if (*line != ' ')
+ for (i = 0; i < req->hcount; i++)
{
- if (!*line)
- *reason_phrase_ptr = line;
- else
- return -1;
+ struct request_header *hdr = &req->headers[i];
+ APPEND (p, hdr->name);
+ *p++ = ':', *p++ = ' ';
+ APPEND (p, hdr->value);
+ *p++ = '\r', *p++ = '\n';
}
- else
- *reason_phrase_ptr = line + 1;
- return statcode;
+ *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);
}
-\f
-#define WMIN(x, y) ((x) > (y) ? (y) : (x))
/* 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
int length = fread (chunk, 1, sizeof (chunk), fp);
if (length == 0)
break;
- towrite = WMIN (promised_size - written, length);
+ towrite = MIN (promised_size - written, length);
write_error = fd_write (sock, chunk, towrite, -1);
if (write_error < 0)
{
}
\f
static const char *
-next_header (const char *h)
+head_terminator (const char *hunk, int oldlen, int peeklen)
{
- const char *end = NULL;
- const char *p = h;
- do
+ const char *start, *end;
+
+ /* If at first peek, verify whether HUNK starts with "HTTP". If
+ not, this is a HTTP/0.9 request and we must bail out without
+ reading anything. */
+ if (oldlen == 0 && 0 != memcmp (hunk, "HTTP", MIN (peeklen, 4)))
+ return hunk;
+
+ if (oldlen < 4)
+ start = hunk;
+ else
+ start = hunk + oldlen - 4;
+ end = hunk + oldlen + peeklen;
+
+ for (; start < end - 1; start++)
+ if (*start == '\n')
+ {
+ if (start < end - 2
+ && start[1] == '\r'
+ && start[2] == '\n')
+ return start + 3;
+ if (start[1] == '\n')
+ return start + 2;
+ }
+ return NULL;
+}
+
+/* Read the HTTP request head from FD and return it. The error
+ conditions are the same as with fd_read_hunk.
+
+ To support HTTP/0.9 responses, this function tries to make sure
+ that the data begins with "HTTP". If this is not the case, no data
+ is read and an empty request is returned, so that the remaining
+ data can be treated as body. */
+
+static char *
+fd_read_http_head (int fd)
+{
+ return fd_read_hunk (fd, head_terminator, 512);
+}
+
+struct response {
+ /* The response data. */
+ const char *data;
+
+ /* The array of pointers that indicate where each header starts.
+ 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
+ response_header_*. */
+
+static struct response *
+response_new (const char *head)
+{
+ const char *hdr;
+ int count, size;
+
+ struct response *resp = xnew0 (struct response);
+ resp->data = head;
+
+ if (*head == '\0')
+ {
+ /* Empty head means that we're dealing with a headerless
+ (HTTP/0.9) response. In that case, don't set HEADERS at
+ all. */
+ return resp;
+ }
+
+ /* Split HEAD into header lines, so that response_header_* functions
+ don't need to do this over and over again. */
+
+ size = count = 0;
+ hdr = head;
+ while (1)
+ {
+ DO_REALLOC (resp->headers, size, count + 1, const char *);
+ resp->headers[count++] = hdr;
+
+ /* Break upon encountering an empty line. */
+ if (!hdr[0] || (hdr[0] == '\r' && hdr[1] == '\n') || hdr[0] == '\n')
+ break;
+
+ /* Find the end of HDR, including continuations. */
+ do
+ {
+ const char *end = strchr (hdr, '\n');
+ if (end)
+ hdr = end + 1;
+ else
+ hdr += strlen (hdr);
+ }
+ while (*hdr == ' ' || *hdr == '\t');
+ }
+ DO_REALLOC (resp->headers, size, count + 1, const char *);
+ resp->headers[count++] = NULL;
+
+ return resp;
+}
+
+/* Locate the header named NAME in the request data. If found, set
+ *BEGPTR to its starting, and *ENDPTR to its ending position, and
+ return 1. Otherwise return 0.
+
+ This function is used as a building block for response_header_copy
+ and response_header_strdup. */
+
+static int
+response_header_bounds (const struct response *resp, const char *name,
+ const char **begptr, const char **endptr)
+{
+ int i;
+ const char **headers = resp->headers;
+ int name_len;
+
+ if (!headers || !headers[1])
+ return 0;
+
+ name_len = strlen (name);
+
+ for (i = 1; headers[i + 1]; i++)
{
- p = strchr (p, '\n');
- if (!p)
- return end;
- end = ++p;
+ const char *b = headers[i];
+ const char *e = headers[i + 1];
+ if (e - b > name_len
+ && b[name_len] == ':'
+ && 0 == strncasecmp (b, name, name_len))
+ {
+ b += name_len + 1;
+ while (b < e && ISSPACE (*b))
+ ++b;
+ while (b < e && ISSPACE (e[-1]))
+ --e;
+ *begptr = b;
+ *endptr = e;
+ return 1;
+ }
}
- while (*p == ' ' || *p == '\t');
+ return 0;
+}
- return end;
+/* Copy the response header named NAME to buffer BUF, no longer than
+ BUFSIZE (BUFSIZE includes the terminating 0). If the header
+ exists, 1 is returned, otherwise 0. If there should be no limit on
+ the size of the header, use response_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 int
+response_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 (bufsize)
+ {
+ int len = MIN (e - b, bufsize - 1);
+ memcpy (buf, b, len);
+ buf[len] = '\0';
+ }
+ return 1;
}
-\f
-/* Functions to be used as arguments to header_process(): */
+/* Return the value of header named NAME in RESP, allocated with
+ malloc. If such a header does not exist in RESP, return NULL. */
-struct http_process_range_closure {
- long first_byte_pos;
- long last_byte_pos;
- long entity_length;
-};
+static char *
+response_header_strdup (const struct response *resp, const char *name)
+{
+ const char *b, *e;
+ if (!response_header_bounds (resp, name, &b, &e))
+ return NULL;
+ return strdupdelim (b, e);
+}
+
+/* Parse the HTTP status line, which is of format:
+
+ HTTP-Version SP Status-Code SP Reason-Phrase
+
+ The function returns the status-code, or -1 if the status line
+ appears malformed. The pointer to "reason-phrase" message is
+ returned in *MESSAGE. */
+
+static int
+response_status (const struct response *resp, char **message)
+{
+ int status;
+ const char *p, *end;
+
+ if (!resp->headers)
+ {
+ /* For a HTTP/0.9 response, assume status 200. */
+ if (message)
+ *message = xstrdup (_("No headers, assuming HTTP/0.9"));
+ return 200;
+ }
+
+ p = resp->headers[0];
+ end = resp->headers[1];
+
+ if (!end)
+ return -1;
+
+ /* "HTTP" */
+ if (end - p < 4 || 0 != strncmp (p, "HTTP", 4))
+ return -1;
+ p += 4;
+
+ /* Match the HTTP version. This is optional because Gnutella
+ servers have been reported to not specify HTTP version. */
+ if (p < end && *p == '/')
+ {
+ ++p;
+ while (p < end && ISDIGIT (*p))
+ ++p;
+ if (p < end && *p == '.')
+ ++p;
+ while (p < end && ISDIGIT (*p))
+ ++p;
+ }
+
+ while (p < end && ISSPACE (*p))
+ ++p;
+ if (end - p < 3 || !ISDIGIT (p[0]) || !ISDIGIT (p[1]) || !ISDIGIT (p[2]))
+ return -1;
+
+ status = 100 * (p[0] - '0') + 10 * (p[1] - '0') + (p[2] - '0');
+ p += 3;
+
+ if (message)
+ {
+ while (p < end && ISSPACE (*p))
+ ++p;
+ while (p < end && ISSPACE (end[-1]))
+ --end;
+ *message = strdupdelim (p, end);
+ }
+
+ return status;
+}
+
+/* Release the resources used by RESP. */
+
+static void
+response_free (struct response *resp)
+{
+ xfree_null (resp->headers);
+ xfree (resp);
+}
+
+/* Print [b, e) to the log, omitting the trailing CRLF. */
+
+static void
+print_server_response_1 (const char *prefix, 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%s\n", prefix, ln);
+}
+
+/* Print the server response, line by line, omitting the trailing CR
+ characters, prefixed with PREFIX. */
+
+static void
+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 (prefix, resp->headers[i], resp->headers[i + 1]);
+}
/* Parse the `Content-Range' header and extract the information it
contains. Returns 1 if successful, -1 otherwise. */
static int
-http_process_range (const char *hdr, void *arg)
+parse_content_range (const char *hdr, long *first_byte_ptr,
+ long *last_byte_ptr, long *entity_length_ptr)
{
- struct http_process_range_closure *closure
- = (struct http_process_range_closure *)arg;
long num;
- /* Certain versions of Nutscape proxy server send out
- `Content-Length' without "bytes" specifier, which is a breach of
- RFC2068 (as well as the HTTP/1.1 draft which was current at the
- time). But hell, I must support it... */
+ /* Ancient versions of Netscape proxy server, presumably predating
+ rfc2068, sent out `Content-Range' without the "bytes"
+ specifier. */
if (!strncasecmp (hdr, "bytes", 5))
{
hdr += 5;
HTTP spec. */
if (*hdr == ':')
++hdr;
- hdr += skip_lws (hdr);
+ while (ISSPACE (*hdr))
+ ++hdr;
if (!*hdr)
return 0;
}
num = 10 * num + (*hdr - '0');
if (*hdr != '-' || !ISDIGIT (*(hdr + 1)))
return 0;
- closure->first_byte_pos = num;
+ *first_byte_ptr = num;
++hdr;
for (num = 0; ISDIGIT (*hdr); hdr++)
num = 10 * num + (*hdr - '0');
if (*hdr != '/' || !ISDIGIT (*(hdr + 1)))
return 0;
- closure->last_byte_pos = num;
+ *last_byte_ptr = num;
++hdr;
for (num = 0; ISDIGIT (*hdr); hdr++)
num = 10 * num + (*hdr - '0');
- closure->entity_length = num;
- return 1;
-}
-
-/* Place 1 to ARG if the HDR contains the word "none", 0 otherwise.
- Used for `Accept-Ranges'. */
-static int
-http_process_none (const char *hdr, void *arg)
-{
- int *where = (int *)arg;
-
- if (strstr (hdr, "none"))
- *where = 1;
- else
- *where = 0;
- return 1;
-}
-
-/* Place the malloc-ed copy of HDR hdr, to the first `;' to ARG. */
-static int
-http_process_type (const char *hdr, void *arg)
-{
- char **result = (char **)arg;
- /* Locate P on `;' or the terminating zero, whichever comes first. */
- const char *p = strchr (hdr, ';');
- if (!p)
- p = hdr + strlen (hdr);
- while (p > hdr && ISSPACE (*(p - 1)))
- --p;
- *result = strdupdelim (hdr, p);
- return 1;
-}
-
-/* Check whether the `Connection' header is set to "keep-alive". */
-static int
-http_process_connection (const char *hdr, void *arg)
-{
- int *flag = (int *)arg;
- if (!strcasecmp (hdr, "Keep-Alive"))
- *flag = 1;
+ *entity_length_ptr = num;
return 1;
}
-/* Commit the cookie to the cookie jar. */
+/* 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 error
+ responses whose bodies don't need to be displayed or logged, but
+ which need to be read anyway. */
-int
-http_process_set_cookie (const char *hdr, void *arg)
+static void
+skip_short_body (int fd, long contlen)
{
- struct url *u = (struct url *)arg;
-
- /* 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, hdr);
- return 1;
+ /* Skipping the body doesn't make sense if the content length is
+ unknown because, in that case, persistent connections cannot be
+ used. (#### This is not the case with HTTP/1.1 where they can
+ still be used with the magic of the "chunked" transfer!) */
+ if (contlen == -1)
+ return;
+ DEBUGP (("Skipping %ld bytes of body data... ", contlen));
+
+ while (contlen > 0)
+ {
+ char dlbuf[512];
+ int ret = fd_read (fd, dlbuf, MIN (contlen, sizeof (dlbuf)), -1);
+ if (ret <= 0)
+ return;
+ contlen -= ret;
+ }
+ DEBUGP (("done.\n"));
}
-
\f
/* Persistent connections. Currently, we cache the most recently used
connection as persistent, provided that the HTTP server agrees to
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
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. */
+ long 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. */
};
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 char *basic_authentication_encode PARAMS ((const char *, const char *));
static int known_authentication_scheme_p PARAMS ((const char *));
time_t http_atotm PARAMS ((const char *));
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, hcount, statcode;
+ int statcode;
int write_error;
long contlen, contrange;
struct url *conn;
FILE *fp;
- int auth_tried_already;
+
+ int sock = -1;
+ int flags;
+
+ /* Whether authorization has been already tried. */
+ int auth_tried_already = 0;
+
+ /* Whether our connection to the remote host is through SSL. */
int using_ssl = 0;
- char *cookies = NULL;
char *head;
- const char *hdr_beg, *hdr_end;
+ struct response *resp;
+ char hdrval[256];
+ char *message;
/* Whether this connection will be kept alive after the HTTP request
is done. */
int keep_alive;
- /* Flags that detect the two ways of specifying HTTP keep-alive
- response. */
- int http_keep_alive_1, http_keep_alive_2;
-
/* Whether keep-alive should be inhibited. */
- int inhibit_keep_alive;
-
- /* 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;
+ int inhibit_keep_alive = !opt.http_keep_alive;
/* Headers sent when using POST. */
- char *post_content_type, *post_content_length;
long post_data_size = 0;
- int host_lookup_failed;
+ int host_lookup_failed = 0;
#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. */
+ switch (ssl_init ())
+ {
+ 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;
+ }
}
#endif /* HAVE_SSL */
know the local filename so we can save to it. */
assert (*hs->local_file != NULL);
- authenticate_h = 0;
auth_tried_already = 0;
- inhibit_keep_alive = !opt.http_keep_alive || proxy != NULL;
-
- 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;
- http_keep_alive_1 = http_keep_alive_2 = 0;
-
- post_content_type = NULL;
- post_content_length = NULL;
-
/* Initialize certain elements of struct http_stat. */
hs->len = 0L;
hs->contlen = -1;
hs->remote_time = NULL;
hs->error = NULL;
- /* If we're using a proxy, we will be connecting to the proxy
- server. */
- conn = proxy ? proxy : u;
-
- host_lookup_failed = 0;
-
- /* First: establish the connection. */
- if (inhibit_keep_alive
- || !persistent_available_p (conn->host, conn->port,
-#ifdef HAVE_SSL
- u->scheme == SCHEME_HTTPS
-#else
- 0
-#endif
- , &host_lookup_failed))
- {
- /* In its current implementation, persistent_available_p will
- 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;
-
- sock = connect_to_host (conn->host, conn->port);
- if (sock == E_HOST)
- return HOSTERR;
- else if (sock < 0)
- return (retryable_socket_connect_error (errno)
- ? CONERROR : CONIMPOSSIBLE);
-
-#ifdef HAVE_SSL
- if (conn->scheme == SCHEME_HTTPS)
- {
- if (!ssl_connect (sock))
- {
- logputs (LOG_VERBOSE, "\n");
- logprintf (LOG_NOTQUIET,
- _("Unable to establish SSL connection.\n"));
- fd_close (sock);
- return CONSSLERR;
- }
- using_ssl = 1;
- }
-#endif /* HAVE_SSL */
- }
- else
- {
- logprintf (LOG_VERBOSE, _("Reusing existing connection to %s:%d.\n"),
- pconn.host, pconn.port);
- sock = pconn.socket;
- using_ssl = pconn.ssl;
- DEBUGP (("Reusing fd %d.\n", sock));
- }
-
- 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);
- }
- }
+ conn = u;
proxyauth = NULL;
if (proxy)
{
char *proxy_user, *proxy_passwd;
/* For normal username and password, URL components override
- command-line/wgetrc parameters. With proxy authentication,
- it's the reverse, because proxy URLs are normally the
- "permanent" ones, so command-line args should take
- precedence. */
+ command-line/wgetrc parameters. With proxy
+ authentication, it's the reverse, because proxy URLs are
+ normally the "permanent" ones, so command-line args
+ should take precedence. */
if (opt.proxy_user && opt.proxy_passwd)
{
proxy_user = opt.proxy_user;
proxy_passwd = proxy->passwd;
}
/* #### This does not appear right. Can't the proxy request,
- say, `Digest' authentication? */
+ say, `Digest' authentication? */
if (proxy_user && proxy_passwd)
- proxyauth = basic_authentication_encode (proxy_user, proxy_passwd,
- "Proxy-Authorization");
+ proxyauth = basic_authentication_encode (proxy_user, proxy_passwd);
+
+ /* If we're using a proxy, we will be connecting to the proxy
+ server. */
+ conn = proxy;
}
- /* String of the form :PORT. Used only for non-standard ports. */
- port_maybe = NULL;
- if (u->port != scheme_default_port (u->scheme))
+ /* Prepare the request to send. */
+
+ req = request_new ();
+ {
+ const char *meth = "GET";
+ if (*dt & 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". */
+ request_set_method (req, meth,
+ proxy ? xstrdup (u->url) : url_full_path (u));
+ }
+
+ 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=%ld-", hs->restval), rel_value);
+ if (opt.useragent)
+ request_set_header (req, "User-Agent", opt.useragent, rel_none);
+ else
+ request_set_header (req, "User-Agent",
+ aprintf ("Wget/%s", version_string), rel_value);
+ 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;
+ passwd = passwd ? passwd : opt.http_passwd;
+
+ if (user && passwd)
{
- port_maybe = (char *)alloca (numdigit (u->port) + 2);
- sprintf (port_maybe, ":%d", u->port);
+ /* 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);
}
+ {
+ /* 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 = 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_keep_alive = "Connection: Keep-Alive\r\n";
- else
- request_keep_alive = NULL;
+ request_set_header (req, "Connection", "Keep-Alive", rel_none);
if (opt.cookies)
- cookies = cookie_header (wget_cookie_jar, u->host, u->port, u->path,
+ request_set_header (req, "Cookie",
+ cookie_header (wget_cookie_jar,
+ u->host, u->port, u->path,
#ifdef HAVE_SSL
- u->scheme == SCHEME_HTTPS
+ u->scheme == SCHEME_HTTPS
#else
- 0
+ 0
#endif
- );
+ ),
+ rel_value);
if (opt.post_data || opt.post_file_name)
{
- post_content_type = "Content-Type: application/x-www-form-urlencoded\r\n";
+ 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 = 0;
}
}
- post_content_length = xmalloc (16 + numdigit (post_data_size) + 2 + 1);
- sprintf (post_content_length,
- "Content-Length: %ld\r\n", post_data_size);
+ request_set_header (req, "Content-Length",
+ aprintf ("%ld", post_data_size), rel_value);
}
- 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);
+ /* 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]);
+ }
+
+ 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 = 0;
+
+ /* Establish the connection. */
+
+ if (!inhibit_keep_alive)
+ {
+ /* Look for a persistent connection to target host, unless a
+ proxy is used. The exception is when SSL is in use, in which
+ case the proxy is nothing but a passthrough to the target
+ host, registered as a connection to the latter. */
+ struct url *relevant = conn;
+#ifdef HAVE_SSL
+ if (u->scheme == SCHEME_HTTPS)
+ relevant = u;
+#endif
+
+ if (persistent_available_p (relevant->host, relevant->port,
+#ifdef HAVE_SSL
+ relevant->scheme == SCHEME_HTTPS,
+#else
+ 0,
+#endif
+ &host_lookup_failed))
+ {
+ sock = pconn.socket;
+ using_ssl = pconn.ssl;
+ logprintf (LOG_VERBOSE, _("Reusing existing connection to %s:%d.\n"),
+ pconn.host, pconn.port);
+ DEBUGP (("Reusing fd %d.\n", sock));
+ }
+ }
+
+ if (sock < 0)
+ {
+ /* In its current implementation, persistent_available_p will
+ 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;
+
+ sock = connect_to_host (conn->host, conn->port);
+ if (sock == E_HOST)
+ return HOSTERR;
+ else if (sock < 0)
+ 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. */
+ struct request *connreq = request_new ();
+ request_set_method (connreq, "CONNECT",
+ aprintf ("%s:%d", u->host, u->port));
+ 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;
+ }
+
+ write_error = request_send (connreq, sock);
+ request_free (connreq);
+ if (write_error < 0)
+ {
+ logprintf (LOG_VERBOSE, _("Failed writing to proxy: %s.\n"),
+ strerror (errno));
+ CLOSE_INVALIDATE (sock);
+ return WRITEFAILED;
+ }
+
+ head = fd_read_http_head (sock);
+ if (!head)
+ {
+ logprintf (LOG_VERBOSE, _("Failed reading proxy response: %s\n"),
+ strerror (errno));
+ CLOSE_INVALIDATE (sock);
+ return HERR;
+ }
+ message = NULL;
+ if (!*head)
+ {
+ xfree (head);
+ goto failed_tunnel;
+ }
+ DEBUGP (("proxy responded with: [%s]\n", head));
+
+ resp = response_new (head);
+ statcode = response_status (resp, &message);
+ response_free (resp);
+ if (statcode != 200)
+ {
+ failed_tunnel:
+ logprintf (LOG_NOTQUIET, _("Proxy tunneling failed: %s"),
+ message ? message : "?");
+ xfree_null (message);
+ return CONSSLERR;
+ }
+ xfree (message);
+
+ /* SOCK is now *really* connected to u->host, so update CONN
+ to reflect this. That way register_persistent will
+ register SOCK as being connected to u->host:u->port. */
+ conn = u;
+ }
+
+ if (conn->scheme == SCHEME_HTTPS)
+ {
+ if (!ssl_connect (sock))
+ {
+ fd_close (sock);
+ return CONSSLERR;
+ }
+ using_ssl = 1;
+ }
+#endif /* HAVE_SSL */
+ }
/* 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;
+ contlen = -1;
+ contrange = 0;
type = NULL;
statcode = -1;
*dt &= ~RETROKF;
- DEBUGP (("\n---response begin---\n"));
-
- head = fd_read_head (sock);
+ head = fd_read_http_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));
- /* Loop through the headers and process them. */
+ resp = response_new (head);
- hcount = 0;
- for (hdr_beg = head;
- (hdr_end = next_header (hdr_beg));
- hdr_beg = hdr_end)
+ /* Check for status line. */
+ message = NULL;
+ statcode = response_status (resp, &message);
+ if (!opt.server_response)
+ logprintf (LOG_VERBOSE, "%2d %s\n", statcode, message ? message : "");
+ else
{
- char *hdr = strdupdelim (hdr_beg, hdr_end);
- {
- char *tmp = hdr + strlen (hdr);
- if (tmp > hdr && tmp[-1] == '\n')
- *--tmp = '\0';
- if (tmp > hdr && tmp[-1] == '\r')
- *--tmp = '\0';
- }
- ++hcount;
-
- /* Check for status line. */
- if (hcount == 1)
- {
- const char *error;
- /* Parse the first line of server response. */
- statcode = parse_http_status_line (hdr, &error);
- hs->statcode = statcode;
- /* Store the descriptive response. */
- if (statcode == -1) /* malformed response */
- {
- /* A common reason for "malformed response" error is the
- case when no data was actually received. Handle this
- special case. */
- if (!*hdr)
- hs->error = xstrdup (_("No data received"));
- else
- hs->error = xstrdup (_("Malformed status line"));
- xfree (hdr);
- break;
- }
- else if (!*error)
- hs->error = xstrdup (_("(no description)"));
- else
- hs->error = xstrdup (error);
-
- if ((statcode != -1)
-#ifdef ENABLE_DEBUG
- && !opt.debug
-#endif
- )
- {
- if (opt.server_response)
- logprintf (LOG_VERBOSE, "\n%2d %s", hcount, hdr);
- else
- logprintf (LOG_VERBOSE, "%2d %s", statcode, error);
- }
-
- goto done_header;
- }
-
- /* Exit on empty header. */
- if (!*hdr)
- {
- xfree (hdr);
- break;
- }
-
- /* Print the header if requested. */
- if (opt.server_response && hcount != 1)
- logprintf (LOG_VERBOSE, "\n%2d %s", hcount, hdr);
-
- /* Try getting content-length. */
- if (contlen == -1 && !opt.ignore_length)
- if (header_process (hdr, "Content-Length", header_extract_number,
- &contlen))
- goto done_header;
- /* Try getting content-type. */
- if (!type)
- if (header_process (hdr, "Content-Type", http_process_type, &type))
- goto done_header;
- /* Try getting location. */
- if (!hs->newloc)
- if (header_process (hdr, "Location", header_strdup, &hs->newloc))
- goto done_header;
- /* Try getting last-modified. */
- if (!hs->remote_time)
- if (header_process (hdr, "Last-Modified", header_strdup,
- &hs->remote_time))
- goto done_header;
- /* Try getting cookies. */
- if (opt.cookies)
- if (header_process (hdr, "Set-Cookie", http_process_set_cookie, u))
- goto done_header;
- /* Try getting www-authentication. */
- if (!authenticate_h)
- if (header_process (hdr, "WWW-Authenticate", header_strdup,
- &authenticate_h))
- goto done_header;
- /* Check for accept-ranges header. If it contains the word
- `none', disable the ranges. */
- if (*dt & ACCEPTRANGES)
- {
- int nonep;
- if (header_process (hdr, "Accept-Ranges", http_process_none, &nonep))
- {
- if (nonep)
- *dt &= ~ACCEPTRANGES;
- goto done_header;
- }
- }
- /* Try getting content-range. */
- if (contrange == -1)
- {
- struct http_process_range_closure closure;
- if (header_process (hdr, "Content-Range", http_process_range, &closure))
- {
- contrange = closure.first_byte_pos;
- goto done_header;
- }
- }
- /* Check for keep-alive related responses. */
- if (!inhibit_keep_alive)
- {
- /* Check for the `Keep-Alive' header. */
- if (!http_keep_alive_1)
- {
- if (header_process (hdr, "Keep-Alive", header_exists,
- &http_keep_alive_1))
- goto done_header;
- }
- /* Check for `Connection: Keep-Alive'. */
- if (!http_keep_alive_2)
- {
- if (header_process (hdr, "Connection", http_process_connection,
- &http_keep_alive_2))
- goto done_header;
- }
- }
- done_header:
- xfree (hdr);
+ logprintf (LOG_VERBOSE, "\n");
+ print_server_response (resp, " ");
}
- DEBUGP (("---response end---\n"));
- logputs (LOG_VERBOSE, "\n");
+ if (response_header_copy (resp, "Content-Length", hdrval, sizeof (hdrval)))
+ contlen = strtol (hdrval, NULL, 10);
- if (contlen != -1
- && (http_keep_alive_1 || http_keep_alive_2))
+ /* Check for keep-alive related responses. */
+ if (!inhibit_keep_alive && contlen != -1)
{
- assert (inhibit_keep_alive == 0);
- keep_alive = 1;
+ if (response_header_copy (resp, "Keep-Alive", NULL, 0))
+ keep_alive = 1;
+ else if (response_header_copy (resp, "Connection", hdrval,
+ sizeof (hdrval)))
+ {
+ if (0 == strcasecmp (hdrval, "Keep-Alive"))
+ keep_alive = 1;
+ }
}
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)
+ if (statcode == HTTP_STATUS_UNAUTHORIZED)
{
/* 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)
+ skip_short_body (sock, contlen);
+ CLOSE_FINISH (sock);
+ if (auth_tried_already || !(user && passwd))
{
/* 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"))
+ else
{
- /* The authentication scheme is basic, the one we try by
- default, and it failed. There's no sense in trying
- again. */
- goto failed;
+ char *www_authenticate = response_header_strdup (resp,
+ "WWW-Authenticate");
+ /* If the authentication scheme is unknown or if it's the
+ "Basic" authentication (which we try by default), there's
+ no sense in retrying. */
+ if (!www_authenticate
+ || !known_authentication_scheme_p (www_authenticate)
+ || BEGINS_WITH (www_authenticate, "Basic"))
+ {
+ xfree_null (www_authenticate);
+ logputs (LOG_NOTQUIET, _("Unknown authentication scheme.\n"));
+ }
+ else
+ {
+ char *pth;
+ auth_tried_already = 1;
+ pth = url_full_path (u);
+ request_set_header (req, "Authorization",
+ create_authorization_line (www_authenticate,
+ user, passwd,
+ request_method (req),
+ pth),
+ rel_value);
+ xfree (pth);
+ xfree (www_authenticate);
+ goto retry_with_auth;
+ }
}
- else
+ request_free (req);
+ return AUTHFAILED;
+ }
+ request_free (req);
+
+ hs->statcode = statcode;
+ if (statcode == -1)
+ hs->error = xstrdup (_("Malformed status line"));
+ else if (!*message)
+ hs->error = xstrdup (_("(no description)"));
+ else
+ hs->error = xstrdup (message);
+
+ type = response_header_strdup (resp, "Content-Type");
+ if (type)
+ {
+ char *tmp = strchr (type, ';');
+ if (tmp)
{
- auth_tried_already = 1;
- goto again;
+ while (tmp > type && ISSPACE (tmp[-1]))
+ --tmp;
+ *tmp = '\0';
}
}
- /* We do not need this anymore. */
- if (authenticate_h)
+ hs->newloc = response_header_strdup (resp, "Location");
+ hs->remote_time = response_header_strdup (resp, "Last-Modified");
+ {
+ char *set_cookie = response_header_strdup (resp, "Set-Cookie");
+ if (set_cookie)
+ {
+ /* 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);
+ }
+ }
+ if (response_header_copy (resp, "Content-Range", hdrval, sizeof (hdrval)))
{
- xfree (authenticate_h);
- authenticate_h = NULL;
+ 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;
}
+ response_free (resp);
/* 20x responses are counted among successful by default. */
if (H_20X (statcode))
_("Location: %s%s\n"),
hs->newloc ? 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)
+ skip_short_body (sock, contlen);
+ CLOSE_FINISH (sock);
xfree_null (type);
return NEWLOCATION;
}
}
}
- 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, legible (contlen + contrange));
+ if (contrange)
+ logprintf (LOG_VERBOSE, _(" (%s to go)"), legible (contlen));
}
else
logputs (LOG_VERBOSE,
hs->len = 0L;
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)
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);
+ /* 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);
if (strchr (u->url, '*'))
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;
}
/* Reset the counter. */
count = 0;
- *dt = 0 | ACCEPTRANGES;
+ *dt = 0;
/* THE loop */
do
{
if (opt.verbose)
{
char *hurl = url_string (u, 1);
- char tmp[15];
+ char tmp[256];
strcpy (tmp, " ");
if (count > 1)
sprintf (tmp, _("(try:%2d)"), count);
*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;
+ hstat.restval = 0;
+ if (count > 1)
+ hstat.restval = hstat.len; /* continue where we left off */
+ else if (opt.always_rest
+ && stat (locf, &st) == 0
+ && S_ISREG (st.st_mode))
+ hstat.restval = st.st_size;
/* Decide whether to send the no-cache directive. We send it in
two cases:
const char *fl = NULL;
if (opt.output_document)
{
- if (opt.od_known_regular)
+ if (output_stream_regular)
fl = opt.output_document;
}
else
return RETROK;
}
- tmrate = retr_rate (hstat.len - hstat.restval, hstat.dltime, 0);
+ tmrate = retr_rate (hstat.rd_size, hstat.dltime, 0);
if (hstat.len == hstat.contlen)
{
This is done by encoding the string `USER:PASS' in base64 and
prepending `HEADER: Basic ' to 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;
int len1 = strlen (user) + 1 + strlen (passwd);
t1 = (char *)alloca (len1 + 1);
sprintf (t1, "%s:%s", user, passwd);
- t2 = (char *)alloca (1 + len2);
+
+ t2 = (char *)alloca (len2 + 1);
base64_encode (t1, t2, len1);
- res = (char *)xmalloc (len2 + 11 + strlen (header));
- sprintf (res, "%s: Basic %s\r\n", header, t2);
+
+ res = (char *)xmalloc (6 + len2 + 1);
+ sprintf (res, "Basic %s", t2);
return res;
}
+#define SKIP_WS(x) do { \
+ while (ISSPACE (*(x))) \
+ ++(x); \
+} while (0)
+
#ifdef USE_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
cp += strlen (attr_name);
if (!*cp)
return -1;
- cp += skip_lws (cp);
+ SKIP_WS (cp);
if (*cp != '=')
return -1;
if (!*++cp)
return -1;
- cp += skip_lws (cp);
+ SKIP_WS (cp);
if (*cp != '\"')
return -1;
if (!*++cp)
{
int i;
- au += skip_lws (au);
+ SKIP_WS (au);
for (i = 0; i < countof (options); i++)
{
int skip = extract_header_attr (au, options[i].name,
au++;
if (*au && *++au)
{
- au += skip_lws (au);
+ SKIP_WS (au);
if (*au == '\"')
{
au++;
+ 2 * MD5_HASHLEN /*strlen (response_digest)*/
+ (opaque ? strlen (opaque) : 0)
+ 128);
- sprintf (res, "Authorization: Digest \
+ 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;
}
const char *passwd, const char *method,
const char *path)
{
- 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");
+ if (0 == strncasecmp (au, "Basic", 5))
+ return basic_authentication_encode (user, passwd);
#ifdef USE_DIGEST
- else if (!strncasecmp (au, "Digest", 6))
- wwwauth = digest_authentication_encode (au, user, passwd, method, path);
+ if (0 == strncasecmp (au, "Digest", 6))
+ return digest_authentication_encode (au, user, passwd, method, path);
#endif /* USE_DIGEST */
- return wwwauth;
+ return NULL;
}
\f
void