X-Git-Url: http://sjero.net/git/?a=blobdiff_plain;f=src%2Fhttp.c;h=14176aac0af659d58f78e354db861f2c82d1ffbe;hb=30ac043b0a4a9a983dd1b50ce1c89ed953019292;hp=42b2dbcb89c2cf67c1a575506eb9d320c41cb305;hpb=9ae0328c3dc095848faf437f03027209ec963151;p=wget diff --git a/src/http.c b/src/http.c index 42b2dbcb..14176aac 100644 --- a/src/http.c +++ b/src/http.c @@ -1,21 +1,32 @@ /* HTTP support. - Copyright (C) 1995, 1996, 1997, 1998, 2000 Free Software Foundation, Inc. + Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001, 2002 + Free Software Foundation, Inc. -This file is part of Wget. +This file is part of GNU Wget. -This program is free software; you can redistribute it and/or modify +GNU Wget is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. -This program is distributed in the hope that it will be useful, +GNU Wget is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License -along with this program; if not, write to the Free Software -Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ +along with Wget; if not, write to the Free Software +Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + +In addition, as a special exception, the Free Software Foundation +gives permission to link the code of its release of Wget with the +OpenSSL project's "OpenSSL" library (or with modified versions of it +that use the same license as the "OpenSSL" library), and distribute +the linked executables. You must obey the GNU General Public License +in all respects for all of the code used other than "OpenSSL". If you +modify this file, you may extend this exception to your version of the +file, but you are not obligated to do so. If you do not wish to do +so, delete this exception statement from your version. */ #include @@ -43,12 +54,6 @@ Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ # endif #endif -#ifdef WINDOWS -# include -#else -# include /* for h_errno */ -#endif - #include "wget.h" #include "utils.h" #include "url.h" @@ -59,26 +64,22 @@ Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ #include "connect.h" #include "fnmatch.h" #include "netrc.h" -#if USE_DIGEST -# include "md5.h" -#endif #ifdef HAVE_SSL # include "gen_sslfunc.h" #endif /* HAVE_SSL */ #include "cookies.h" +#ifdef USE_DIGEST +# include "gen-md5.h" +#endif extern char *version_string; #ifndef errno extern int errno; #endif -#ifndef h_errno -# ifndef __CYGWIN__ -extern int h_errno; -# endif -#endif static int cookies_loaded_p; +struct cookie_jar *wget_cookie_jar; #define TEXTHTML_S "text/html" #define HTTP_ACCEPT "*/*" @@ -180,6 +181,80 @@ parse_http_status_line (const char *line, const char **reason_phrase_ptr) return statcode; } +#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 + longer, read only that much; if the file is shorter, pad it with + zeros. */ + +static int +post_file (int sock, void *ssl, const char *file_name, long promised_size) +{ + static char chunk[8192]; + long written = 0; + int write_error; + FILE *fp; + + /* Only one of SOCK and SSL may be active at the same time. */ + assert (sock > -1 || ssl != NULL); + assert (sock == -1 || ssl == NULL); + + DEBUGP (("[writing POST file %s ... ", file_name)); + + fp = fopen (file_name, "rb"); + if (!fp) + goto pad; + while (written < promised_size) + { + int towrite; + int length = fread (chunk, 1, sizeof (chunk), fp); + if (length == 0) + break; + towrite = WMIN (promised_size - written, length); +#ifdef HAVE_SSL + if (ssl) + write_error = ssl_iwrite (ssl, chunk, towrite); + else +#endif + write_error = iwrite (sock, chunk, towrite); + if (write_error < 0) + { + fclose (fp); + return -1; + } + written += towrite; + } + fclose (fp); + + pad: + if (written < promised_size) + { + /* This highly unlikely case can happen only if the file has + shrunk under us. To uphold the promise that exactly + promised_size bytes would be delivered, pad the remaining + data with zeros. #### Should we abort instead? */ + DEBUGP (("padding %ld bytes ... ", promised_size - written)); + memset (chunk, '\0', sizeof (chunk)); + while (written < promised_size) + { + int towrite = WMIN (promised_size - written, sizeof (chunk)); +#ifdef HAVE_SSL + if (ssl) + write_error = ssl_iwrite (ssl, chunk, towrite); + else +#endif + write_error = iwrite (sock, chunk, towrite); + if (write_error < 0) + return -1; + written += towrite; + } + } + assert (written == promised_size); + DEBUGP (("done]\n")); + return 0; +} + /* Functions to be used as arguments to header_process(): */ struct http_process_range_closure { @@ -204,6 +279,10 @@ http_process_range (const char *hdr, void *arg) if (!strncasecmp (hdr, "bytes", 5)) { hdr += 5; + /* "JavaWebServer/1.1.1" sends "bytes: x-y/z", contrary to the + HTTP spec. */ + if (*hdr == ':') + ++hdr; hdr += skip_lws (hdr); if (!*hdr) return 0; @@ -266,6 +345,22 @@ http_process_connection (const char *hdr, void *arg) *flag = 1; return 1; } + +/* Commit the cookie to the cookie jar. */ + +int +http_process_set_cookie (const char *hdr, void *arg) +{ + struct url *u = (struct url *)arg; + + /* The jar should have been created by now. */ + assert (wget_cookie_jar != NULL); + + cookie_jar_process_set_cookie (wget_cookie_jar, u->host, u->port, u->path, + hdr); + return 1; +} + /* Persistent connections. Currently, we cache the most recently used connection as persistent, provided that the HTTP server agrees to @@ -278,7 +373,7 @@ http_process_connection (const char *hdr, void *arg) /* Whether a persistent connection is active. */ static int pc_active_p; /* Host and port of currently active persistent connection. */ -static unsigned char pc_last_host[4]; +static struct address_list *pc_last_host_ip; static unsigned short pc_last_port; /* File descriptor of the currently active persistent connection. */ @@ -303,6 +398,11 @@ invalidate_persistent (void) #ifdef HAVE_SSL pc_active_ssl = 0; #endif /* HAVE_SSL */ + if (pc_last_host_ip != NULL) + { + address_list_release (pc_last_host_ip); + pc_last_host_ip = NULL; + } DEBUGP (("Invalidating fd %d from further reuse.\n", pc_last_fd)); } @@ -314,15 +414,15 @@ invalidate_persistent (void) If a previous connection was persistent, it is closed. */ -static void -register_persistent (const char *host, unsigned short port, int fd #ifdef HAVE_SSL - , SSL *ssl -#endif - ) +static void +register_persistent (const char *host, unsigned short port, int fd, SSL *ssl) { - int success; - +#else +static void +register_persistent (const char *host, unsigned short port, int fd) +{ +#endif if (pc_active_p) { if (pc_last_fd == fd) @@ -349,10 +449,13 @@ register_persistent (const char *host, unsigned short port, int fd } } - /* This store_hostaddress may not fail, because it has the results - in the cache. */ - success = store_hostaddress (pc_last_host, host); - assert (success); + assert (pc_last_host_ip == NULL); + + /* This lookup_host cannot fail, because it has the results in the + cache. */ + pc_last_host_ip = lookup_host (host, 1); + assert (pc_last_host_ip != NULL); + pc_last_port = port; pc_last_fd = fd; pc_active_p = 1; @@ -363,17 +466,30 @@ register_persistent (const char *host, unsigned short port, int fd DEBUGP (("Registered fd %d for persistent reuse.\n", fd)); } +#ifdef HAVE_SSL +# define SHUTDOWN_SSL(ssl) do { \ + if (ssl) \ + shutdown_ssl (ssl); \ +} while (0) +#else +# define SHUTDOWN_SSL(ssl) +#endif + /* Return non-zero if a persistent connection is available for connecting to HOST:PORT. */ -static int -persistent_available_p (const char *host, unsigned short port #ifdef HAVE_SSL - , int ssl -#endif - ) +static int +persistent_available_p (const char *host, unsigned short port, int ssl) +{ +#else +static int +persistent_available_p (const char *host, unsigned short port) { - unsigned char this_host[4]; +#endif + int success; + struct address_list *this_host_ip; + /* First, check whether a persistent connection is active at all. */ if (!pc_active_p) return 0; @@ -381,6 +497,7 @@ persistent_available_p (const char *host, unsigned short port (HOST, PORT) ordered pair. */ if (port != pc_last_port) return 0; + #ifdef HAVE_SSL /* Second, a): check if current connection is (not) ssl, too. This test is unlikely to fail because HTTP and HTTPS typicaly use @@ -390,10 +507,19 @@ persistent_available_p (const char *host, unsigned short port if (ssl != pc_active_ssl) return 0; #endif /* HAVE_SSL */ - if (!store_hostaddress (this_host, host)) + + this_host_ip = lookup_host (host, 1); + if (!this_host_ip) return 0; - if (memcmp (pc_last_host, this_host, 4)) + + /* To equate the two host names for the purposes of persistent + connections, they need to share all the IP addresses in the + list. */ + success = address_list_match_all (pc_last_host_ip, this_host_ip); + address_list_release (this_host_ip); + if (!success) return 0; + /* Third: check whether the connection is still open. This is important because most server implement a liberal (short) timeout on persistent connections. Wget can of course always reconnect @@ -406,21 +532,16 @@ persistent_available_p (const char *host, unsigned short port let's invalidate the persistent connection before returning 0. */ CLOSE (pc_last_fd); +#ifdef HAVE_SSL + SHUTDOWN_SSL (pc_last_ssl); + pc_last_ssl = NULL; +#endif invalidate_persistent (); return 0; } return 1; } -#ifdef HAVE_SSL -# define SHUTDOWN_SSL(ssl) do { \ - if (ssl) \ - shutdown_ssl (ssl); \ -} while (0) -#else -# define SHUTDOWN_SSL(ssl) -#endif - /* The idea behind these two CLOSE macros is to distinguish between two cases: one when the job we've been doing is finished, and we want to close the connection and leave, and two when something is @@ -466,16 +587,22 @@ struct http_stat long dltime; /* time of the download */ int no_truncate; /* whether truncating the file is forbidden. */ + const char *referer; /* value of the referer header. */ + char **local_file; /* local file. */ }; -/* Free the elements of hstat X. */ -#define FREEHSTAT(x) do \ -{ \ - FREE_MAYBE ((x).newloc); \ - FREE_MAYBE ((x).remote_time); \ - FREE_MAYBE ((x).error); \ - (x).newloc = (x).remote_time = (x).error = NULL; \ -} while (0) +static void +free_hstat (struct http_stat *hs) +{ + FREE_MAYBE (hs->newloc); + FREE_MAYBE (hs->remote_time); + FREE_MAYBE (hs->error); + + /* Guard against being called twice. */ + hs->newloc = NULL; + hs->remote_time = NULL; + hs->error = NULL; +} static char *create_authorization_line PARAMS ((const char *, const char *, const char *, const char *, @@ -484,7 +611,7 @@ static char *basic_authentication_encode PARAMS ((const char *, const char *, const char *)); static int known_authentication_scheme_p PARAMS ((const char *)); -time_t http_atotm PARAMS ((char *)); +time_t http_atotm PARAMS ((const char *)); #define BEGINS_WITH(line, string_constant) \ (!strncasecmp (line, string_constant, sizeof (string_constant) - 1) \ @@ -501,32 +628,30 @@ time_t http_atotm PARAMS ((char *)); response code correctly, it is not used in a sane way. The caller can do that, though. - If u->proxy is non-NULL, the URL u will be taken as a proxy URL, - and u->proxy->url will be given to the proxy server (bad naming, - I'm afraid). */ + If PROXY is non-NULL, the connection will be made to the proxy + server, and u->url will be requested. */ static uerr_t -gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) +gethttp (struct url *u, struct http_stat *hs, int *dt, struct url *proxy) { - char *request, *type, *command, *path; + char *request, *type, *command, *full_path; char *user, *passwd; - char *pragma_h, *referer, *useragent, *range, *wwwauth, *remhost; + char *pragma_h, *referer, *useragent, *range, *wwwauth; char *authenticate_h; char *proxyauth; char *all_headers; char *port_maybe; char *request_keep_alive; - int sock, hcount, num_written, all_length, remport, statcode; + int sock, hcount, all_length, statcode; + int write_error; long contlen, contrange; - struct urlinfo *ou; - uerr_t err; + struct url *conn; FILE *fp; int auth_tried_already; struct rbuf rbuf; #ifdef HAVE_SSL static SSL_CTX *ssl_ctx = NULL; SSL *ssl = NULL; -#endif /* HAVE_SSL */ - struct wget_timer *timer; +#endif char *cookies = NULL; /* Whether this connection will be kept alive after the HTTP request @@ -540,11 +665,20 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) /* 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; + + /* Headers sent when using POST. */ + char *post_content_type, *post_content_length; + long post_data_size = 0; + #ifdef HAVE_SSL /* initialize ssl_ctx on first run */ if (!ssl_ctx) { - err=init_ssl (&ssl_ctx); + uerr_t err = init_ssl (&ssl_ctx); if (err != 0) { switch (err) @@ -581,12 +715,12 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) if (!(*dt & 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 (u->local != NULL); + assert (*hs->local_file != NULL); authenticate_h = 0; auth_tried_already = 0; - inhibit_keep_alive = (!opt.http_keep_alive || u->proxy != NULL); + inhibit_keep_alive = !opt.http_keep_alive || proxy != NULL; again: /* We need to come back here when the initial attempt to retrieve @@ -596,9 +730,8 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) keep_alive = 0; http_keep_alive_1 = http_keep_alive_2 = 0; - if (opt.cookies) - cookies = build_cookies_request (u->host, u->port, u->path, - u->proto == URLHTTPS); + post_content_type = NULL; + post_content_length = NULL; /* Initialize certain elements of struct http_stat. */ hs->len = 0L; @@ -608,62 +741,39 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) hs->remote_time = NULL; hs->error = NULL; - /* Which structure to use to retrieve the original URL data. */ - if (u->proxy) - ou = u->proxy; - else - ou = u; + /* If we're using a proxy, we will be connecting to the proxy + server. */ + conn = proxy ? proxy : u; /* First: establish the connection. */ if (inhibit_keep_alive || #ifndef HAVE_SSL - !persistent_available_p (u->host, u->port) + !persistent_available_p (conn->host, conn->port) #else - !persistent_available_p (u->host, u->port, (u->proto==URLHTTPS ? 1 : 0)) + !persistent_available_p (conn->host, conn->port, + u->scheme == SCHEME_HTTPS) #endif /* HAVE_SSL */ ) { - logprintf (LOG_VERBOSE, _("Connecting to %s:%hu... "), u->host, u->port); - err = make_connection (&sock, u->host, u->port); - switch (err) - { - case HOSTERR: - logputs (LOG_VERBOSE, "\n"); - logprintf (LOG_NOTQUIET, "%s: %s.\n", u->host, herrmsg (h_errno)); - return HOSTERR; - break; - case CONSOCKERR: - logputs (LOG_VERBOSE, "\n"); - logprintf (LOG_NOTQUIET, "socket: %s\n", strerror (errno)); - return CONSOCKERR; - break; - case CONREFUSED: - logputs (LOG_VERBOSE, "\n"); - logprintf (LOG_NOTQUIET, - _("Connection to %s:%hu refused.\n"), u->host, u->port); - CLOSE (sock); - return CONREFUSED; - case CONERROR: - logputs (LOG_VERBOSE, "\n"); - logprintf (LOG_NOTQUIET, "connect: %s\n", strerror (errno)); - CLOSE (sock); - return CONERROR; - break; - case NOCONERROR: - /* Everything is fine! */ - logputs (LOG_VERBOSE, _("connected!\n")); - break; - default: - abort (); - break; - } + struct address_list *al = lookup_host (conn->host, 0); + if (!al) + return HOSTERR; + set_connection_host_name (conn->host); + sock = connect_to_many (al, conn->port, 0); + set_connection_host_name (NULL); + address_list_release (al); + + if (sock < 0) + return CONNECT_ERROR (errno); + #ifdef HAVE_SSL - if (u->proto == URLHTTPS) + if (conn->scheme == SCHEME_HTTPS) if (connect_ssl (&ssl, ssl_ctx,sock) != 0) { logputs (LOG_VERBOSE, "\n"); logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n")); + ssl_printerrors (); CLOSE (sock); return CONSSLERR; } @@ -671,7 +781,8 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) } else { - logprintf (LOG_VERBOSE, _("Reusing connection to %s:%hu.\n"), u->host, u->port); + logprintf (LOG_VERBOSE, _("Reusing connection to %s:%hu.\n"), + conn->host, conn->port); /* #### pc_last_fd should be accessed through an accessor function. */ sock = pc_last_fd; @@ -681,22 +792,25 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) DEBUGP (("Reusing fd %d.\n", sock)); } - if (u->proxy) - path = u->proxy->url; + if (*dt & HEAD_ONLY) + command = "HEAD"; + else if (opt.post_file_name || opt.post_data) + command = "POST"; else - path = u->path; - - command = (*dt & HEAD_ONLY) ? "HEAD" : "GET"; + command = "GET"; + referer = NULL; - if (ou->referer) + if (hs->referer) { - referer = (char *)alloca (9 + strlen (ou->referer) + 3); - sprintf (referer, "Referer: %s\r\n", ou->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); @@ -719,8 +833,8 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) sprintf (useragent, "Wget/%s", version_string); } /* Construct the authentication, if userid is present. */ - user = ou->user; - passwd = ou->passwd; + 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; @@ -754,13 +868,18 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) } 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, ou->path); + command, pth); + xfree (pth); } } proxyauth = NULL; - if (u->proxy) + if (proxy) { char *proxy_user, *proxy_passwd; /* For normal username and password, URL components override @@ -775,31 +894,22 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) } else { - proxy_user = u->user; - proxy_passwd = u->passwd; + proxy_user = proxy->user; + proxy_passwd = proxy->passwd; } - /* #### This is junky. Can't the proxy request, say, `Digest' - authentication? */ + /* #### 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"); } - remhost = ou->host; - remport = ou->port; /* String of the form :PORT. Used only for non-standard ports. */ port_maybe = NULL; - if (1 -#ifdef HAVE_SSL - && remport != (u->proto == URLHTTPS - ? DEFAULT_HTTPS_PORT : DEFAULT_HTTP_PORT) -#else - && remport != DEFAULT_HTTP_PORT -#endif - ) + if (u->port != scheme_default_port (u->scheme)) { - port_maybe = (char *)alloca (numdigit (remport) + 2); - sprintf (port_maybe, ":%d", remport); + port_maybe = (char *)alloca (numdigit (u->port) + 2); + sprintf (port_maybe, ":%d", u->port); } if (!inhibit_keep_alive) @@ -807,10 +917,52 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) else request_keep_alive = NULL; + if (opt.cookies) + cookies = cookie_jar_generate_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 (path) + request = (char *)alloca (strlen (command) + + strlen (full_path) + strlen (useragent) - + strlen (remhost) + + strlen (u->host) + (port_maybe ? strlen (port_maybe) : 0) + strlen (HTTP_ACCEPT) + (request_keep_alive @@ -821,16 +973,22 @@ gethttp (struct urlinfo *u, struct http_stat *hs, int *dt) + (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\r\n\ +Host: %s%s%s%s\r\n\ Accept: %s\r\n\ -%s%s%s%s%s%s%s%s\r\n", - command, path, useragent, remhost, +%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 : "", @@ -839,23 +997,53 @@ Accept: %s\r\n\ wwwauth ? wwwauth : "", proxyauth ? proxyauth : "", range ? range : "", - pragma_h, + pragma_h, + post_content_type ? post_content_type : "", + post_content_length ? post_content_length : "", opt.user_header ? opt.user_header : ""); - DEBUGP (("---request begin---\n%s---request end---\n", request)); - /* Free the temporary memory. */ + DEBUGP (("---request begin---\n%s", request)); + + /* Free the temporary memory. */ FREE_MAYBE (wwwauth); FREE_MAYBE (proxyauth); FREE_MAYBE (cookies); + xfree (full_path); /* Send the request to server. */ #ifdef HAVE_SSL - if (u->proto == URLHTTPS) - num_written = ssl_iwrite (ssl, request, strlen (request)); + if (conn->scheme == SCHEME_HTTPS) + write_error = ssl_iwrite (ssl, request, strlen (request)); else -#endif /* HAVE_SSL */ - num_written = iwrite (sock, request, strlen (request)); +#endif + write_error = iwrite (sock, request, strlen (request)); + + if (write_error >= 0) + { + if (opt.post_data) + { + DEBUGP (("[POST data: %s]\n", opt.post_data)); +#ifdef HAVE_SSL + if (conn->scheme == SCHEME_HTTPS) + write_error = ssl_iwrite (ssl, opt.post_data, post_data_size); + else +#endif + write_error = iwrite (sock, opt.post_data, post_data_size); + } + else if (opt.post_file_name) + { +#ifdef HAVE_SSL + if (conn->scheme == SCHEME_HTTPS) + write_error = post_file (-1, ssl, opt.post_file_name, + post_data_size); + else +#endif + write_error = post_file (sock, NULL, opt.post_file_name, + post_data_size); + } + } + DEBUGP (("---request end---\n")); - if (num_written < 0) + if (write_error < 0) { logprintf (LOG_VERBOSE, _("Failed writing HTTP request: %s.\n"), strerror (errno)); @@ -863,7 +1051,7 @@ Accept: %s\r\n\ return WRITEFAILED; } logprintf (LOG_VERBOSE, _("%s request sent, awaiting response... "), - u->proxy ? "Proxy" : "HTTP"); + proxy ? "Proxy" : "HTTP"); contlen = contrange = -1; type = NULL; statcode = -1; @@ -872,7 +1060,7 @@ Accept: %s\r\n\ /* Before reading anything, initialize the rbuf. */ rbuf_initialize (&rbuf, sock); #ifdef HAVE_SSL - if (u->proto == URLHTTPS) + if (conn->scheme == SCHEME_HTTPS) rbuf.ssl = ssl; else rbuf.ssl = NULL; @@ -909,7 +1097,6 @@ Accept: %s\r\n\ logputs (LOG_NOTQUIET, _("End of file while parsing headers.\n")); xfree (hdr); FREE_MAYBE (type); - FREE_MAYBE (hs->newloc); FREE_MAYBE (all_headers); CLOSE_INVALIDATE (sock); return HEOF; @@ -921,7 +1108,6 @@ Accept: %s\r\n\ strerror (errno)); xfree (hdr); FREE_MAYBE (type); - FREE_MAYBE (hs->newloc); FREE_MAYBE (all_headers); CLOSE_INVALIDATE (sock); return HERR; @@ -939,10 +1125,6 @@ Accept: %s\r\n\ all_headers[all_length] = '\0'; } - /* Print the header if requested. */ - if (opt.server_response && hcount != 1) - logprintf (LOG_VERBOSE, "\n%d %s", hcount, hdr); - /* Check for status line. */ if (hcount == 1) { @@ -973,7 +1155,12 @@ Accept: %s\r\n\ && !opt.debug #endif ) - logprintf (LOG_VERBOSE, "%d %s", statcode, error); + { + if (opt.server_response) + logprintf (LOG_VERBOSE, "\n%2d %s", hcount, hdr); + else + logprintf (LOG_VERBOSE, "%2d %s", statcode, error); + } goto done_header; } @@ -985,6 +1172,10 @@ Accept: %s\r\n\ 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, @@ -1005,7 +1196,7 @@ Accept: %s\r\n\ goto done_header; /* Try getting cookies. */ if (opt.cookies) - if (header_process (hdr, "Set-Cookie", set_cookie_header_cb, u)) + if (header_process (hdr, "Set-Cookie", http_process_set_cookie, u)) goto done_header; /* Try getting www-authentication. */ if (!authenticate_h) @@ -1068,9 +1259,9 @@ Accept: %s\r\n\ /* The server has promised that it will not close the connection when we're done. This means that we can register it. */ #ifndef HAVE_SSL - register_persistent (u->host, u->port, sock); + register_persistent (conn->host, conn->port, sock); #else - register_persistent (u->host, u->port, sock, ssl); + register_persistent (conn->host, conn->port, sock, ssl); #endif /* HAVE_SSL */ if ((statcode == HTTP_STATUS_UNAUTHORIZED) @@ -1079,7 +1270,7 @@ Accept: %s\r\n\ /* Authorization is required. */ FREE_MAYBE (type); type = NULL; - FREEHSTAT (*hs); + free_hstat (hs); CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there might be more bytes in the body. */ if (auth_tried_already) @@ -1121,10 +1312,36 @@ Accept: %s\r\n\ if (H_20X (statcode)) *dt |= RETROKF; - if (type && !strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S))) + /* Return if redirected. */ + if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES) + { + /* RFC2068 says that in case of the 300 (multiple choices) + response, the server can output a preferred URL through + `Location' header; otherwise, the request should be treated + like GET. So, if the location is set, it will be a + redirection; otherwise, just proceed normally. */ + if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc) + *dt |= RETROKF; + else + { + logprintf (LOG_VERBOSE, + _("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. */ + FREE_MAYBE (type); + FREE_MAYBE (all_headers); + return NEWLOCATION; + } + } + + /* If content-type is not given, assume text/html. This is because + of the multitude of broken CGI's that "forget" to generate the + content-type. */ + if (!type || 0 == strncasecmp (type, TEXTHTML_S, strlen (TEXTHTML_S))) *dt |= TEXTHTML; else - /* We don't assume text/html by default. */ *dt &= ~TEXTHTML; if (opt.html_extension && (*dt & TEXTHTML)) @@ -1132,16 +1349,17 @@ Accept: %s\r\n\ 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(u->local, '.'); + char* last_period_in_local_filename = strrchr(*hs->local_file, '.'); if (last_period_in_local_filename == NULL || !(strcasecmp(last_period_in_local_filename, ".htm") == EQ || strcasecmp(last_period_in_local_filename, ".html") == EQ)) { - size_t local_filename_len = strlen(u->local); + size_t local_filename_len = strlen(*hs->local_file); - u->local = xrealloc(u->local, local_filename_len + sizeof(".html")); - strcpy(u->local + local_filename_len, ".html"); + *hs->local_file = xrealloc(*hs->local_file, + local_filename_len + sizeof(".html")); + strcpy(*hs->local_file + local_filename_len, ".html"); *dt |= ADDED_HTML_EXTENSION; } @@ -1168,19 +1386,24 @@ Accept: %s\r\n\ if (opt.always_rest) { /* Check for condition #2. */ - if (hs->restval == contlen) + 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, _("\ \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; FREE_MAYBE (type); - FREE_MAYBE (hs->newloc); FREE_MAYBE (all_headers); CLOSE_INVALIDATE (sock); /* would be CLOSE_FINISH, but there might be more bytes in the body. */ - return RETRFINISHED; + return RETRUNNEEDED; } /* Check for condition #1. */ @@ -1189,8 +1412,11 @@ Accept: %s\r\n\ logprintf (LOG_NOTQUIET, _("\ \n\ - The server does not support continued download;\n\ - refusing to truncate `%s'.\n\n"), u->local); +Continued download failed on this file, which conflicts with `-c'.\n\ +Refusing to truncate existing file `%s'.\n\n"), *hs->local_file); + FREE_MAYBE (type); + FREE_MAYBE (all_headers); + CLOSE_INVALIDATE (sock); return CONTNOTSUPPORTED; } @@ -1199,14 +1425,12 @@ Accept: %s\r\n\ hs->restval = 0; } - else if (contrange != hs->restval || (H_PARTIAL (statcode) && contrange == -1)) { /* This means the whole request was somehow misunderstood by the server. Bail out. */ FREE_MAYBE (type); - FREE_MAYBE (hs->newloc); FREE_MAYBE (all_headers); CLOSE_INVALIDATE (sock); return RANGEERR; @@ -1222,29 +1446,6 @@ Accept: %s\r\n\ } hs->contlen = contlen; - /* Return if redirected. */ - if (H_REDIRECTED (statcode) || statcode == HTTP_STATUS_MULTIPLE_CHOICES) - { - /* RFC2068 says that in case of the 300 (multiple choices) - response, the server can output a preferred URL through - `Location' header; otherwise, the request should be treated - like GET. So, if the location is set, it will be a - redirection; otherwise, just proceed normally. */ - if (statcode == HTTP_STATUS_MULTIPLE_CHOICES && !hs->newloc) - *dt |= RETROKF; - else - { - logprintf (LOG_VERBOSE, - _("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. */ - FREE_MAYBE (type); - FREE_MAYBE (all_headers); - return NEWLOCATION; - } - } if (opt.verbose) { if ((*dt & RETROKF) && !opt.server_response) @@ -1288,13 +1489,13 @@ Accept: %s\r\n\ /* Open the local file. */ if (!opt.dfp) { - mkalldirs (u->local); + mkalldirs (*hs->local_file); if (opt.backups) - rotate_backups (u->local); - fp = fopen (u->local, hs->restval ? "ab" : "wb"); + rotate_backups (*hs->local_file); + fp = fopen (*hs->local_file, hs->restval ? "ab" : "wb"); if (!fp) { - logprintf (LOG_NOTQUIET, "%s: %s\n", u->local, strerror (errno)); + 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. */ FREE_MAYBE (all_headers); @@ -1317,8 +1518,12 @@ Accept: %s\r\n\ #### 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. */ - if (!hs->restval && global_download_count == 0) + 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. */ @@ -1334,13 +1539,17 @@ Accept: %s\r\n\ should be some overhead information. */ if (opt.save_headers) fwrite (all_headers, 1, all_length, fp); - timer = wtimer_new (); + /* Get the contents of the document. */ hs->res = get_contents (sock, fp, &hs->len, hs->restval, (contlen != -1 ? contlen : 0), - &rbuf, keep_alive); - hs->dltime = wtimer_elapsed (timer); - wtimer_delete (timer); + &rbuf, keep_alive, &hs->dltime); + + if (hs->res >= 0) + CLOSE_FINISH (sock); + else + CLOSE_INVALIDATE (sock); + { /* Close or flush the file. We have to be careful to check for error here. Checking the result of fwrite() is not enough -- @@ -1354,7 +1563,6 @@ Accept: %s\r\n\ hs->res = -2; } FREE_MAYBE (all_headers); - CLOSE_FINISH (sock); if (hs->res == -2) return FWRITEERR; return RETRFINISHED; @@ -1363,25 +1571,35 @@ Accept: %s\r\n\ /* The genuine HTTP loop! This is the part where the retrieval is retried, and retried, and retried, and... */ uerr_t -http_loop (struct urlinfo *u, char **newloc, int *dt) +http_loop (struct url *u, char **newloc, char **local_file, const char *referer, + int *dt, struct url *proxy) { int count; int use_ts, got_head = 0; /* time-stamping info */ char *filename_plus_orig_suffix; char *local_filename = NULL; - char *tms, *suf, *locf, *tmrate; + char *tms, *locf, *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 */ size_t filename_len; struct http_stat hstat; /* HTTP status */ struct stat st; + char *dummy = NULL; /* This used to be done in main(), but it's a better idea to do it here so that we don't go through the hoops if we're just using FTP or whatever. */ - if (opt.cookies && opt.cookies_input && !cookies_loaded_p) - load_cookies (opt.cookies_input); + if (opt.cookies) + { + if (!wget_cookie_jar) + wget_cookie_jar = cookie_jar_new (); + if (opt.cookies_input && !cookies_loaded_p) + { + cookie_jar_load (wget_cookie_jar, opt.cookies_input); + cookies_loaded_p = 1; + } + } *newloc = NULL; @@ -1392,38 +1610,44 @@ http_loop (struct urlinfo *u, char **newloc, int *dt) logputs (LOG_VERBOSE, _("Warning: wildcards not supported in HTTP.\n")); /* Determine the local filename. */ - if (!u->local) - u->local = url_filename (u->proxy ? u->proxy : u); + if (local_file && *local_file) + hstat.local_file = local_file; + else if (local_file) + { + *local_file = url_filename (u); + hstat.local_file = local_file; + } + else + { + dummy = url_filename (u); + hstat.local_file = &dummy; + } if (!opt.output_document) - locf = u->local; + locf = *hstat.local_file; else locf = opt.output_document; - /* Yuck. Multiple returns suck. We need to remember to free() the space we - xmalloc() here before EACH return. This is one reason it's better to set - flags that influence flow control and then return once at the end. */ - filename_len = strlen(u->local); - filename_plus_orig_suffix = xmalloc(filename_len + sizeof(".orig")); + hstat.referer = referer; + + filename_len = strlen (*hstat.local_file); + filename_plus_orig_suffix = alloca (filename_len + sizeof (".orig")); - if (opt.noclobber && file_exists_p (u->local)) + if (opt.noclobber && file_exists_p (*hstat.local_file)) { /* 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"), u->local); +File `%s' already there, will not retrieve.\n"), *hstat.local_file); /* If the file is there, we suppose it's retrieved OK. */ *dt |= RETROKF; /* #### Bogusness alert. */ - /* If its suffix is "html" or (yuck!) "htm", we suppose it's - text/html, a harmless lie. */ - if (((suf = suffix (u->local)) != NULL) - && (!strcmp (suf, "html") || !strcmp (suf, "htm"))) + /* If its suffix is "html" or "htm" or similar, assume text/html. */ + if (has_html_suffix_p (*hstat.local_file)) *dt |= TEXTHTML; - xfree (suf); - xfree (filename_plus_orig_suffix); /* must precede every return! */ - /* Another harmless lie: */ + + FREE_MAYBE (dummy); return RETROK; } @@ -1450,11 +1674,12 @@ File `%s' already there, will not retrieve.\n"), u->local); in url.c. Replacing sprintf with inline calls to strcpy() and long_to_string() made a difference. --hniksic */ - strcpy(filename_plus_orig_suffix, u->local); - strcpy(filename_plus_orig_suffix + filename_len, ".orig"); + memcpy (filename_plus_orig_suffix, *hstat.local_file, filename_len); + memcpy (filename_plus_orig_suffix + filename_len, + ".orig", sizeof (".orig")); /* Try to stat() the .orig file. */ - if (stat(filename_plus_orig_suffix, &st) == 0) + if (stat (filename_plus_orig_suffix, &st) == 0) { local_dot_orig_file_exists = TRUE; local_filename = filename_plus_orig_suffix; @@ -1463,8 +1688,8 @@ File `%s' already there, will not retrieve.\n"), u->local); if (!local_dot_orig_file_exists) /* Couldn't stat() .orig, so try to stat() . */ - if (stat (u->local, &st) == 0) - local_filename = u->local; + if (stat (*hstat.local_file, &st) == 0) + local_filename = *hstat.local_file; if (local_filename != NULL) /* There was a local file, so we'll check later to see if the version @@ -1473,6 +1698,11 @@ File `%s' already there, will not retrieve.\n"), u->local); { use_ts = 1; tml = st.st_mtime; +#ifdef WINDOWS + /* Modification time granularity is 2 seconds for Windows, so + increase local time by 1 second for later comparison. */ + tml++; +#endif local_size = st.st_size; got_head = 0; } @@ -1491,7 +1721,7 @@ File `%s' already there, will not retrieve.\n"), u->local); /* Print fetch message, if opt.verbose. */ if (opt.verbose) { - char *hurl = str_url (u->proxy ? u->proxy : u, 1); + char *hurl = url_string (u, 1); char tmp[15]; strcpy (tmp, " "); if (count > 1) @@ -1515,34 +1745,43 @@ File `%s' already there, will not retrieve.\n"), u->local); 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; - /* Decide whether to send the no-cache directive. */ - if (u->proxy && (count > 1 || (opt.proxy_cache == 0))) + + /* 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; + + /* Decide whether to send the no-cache directive. We send it in + two cases: + a) we're using a proxy, and we're past our first retrieval. + Some proxies are notorious for caching incomplete data, so + we require a fresh get. + b) caching is explicitly inhibited. */ + if ((proxy && count > 1) /* a */ + || !opt.allow_cache /* b */ + ) *dt |= SEND_NOCACHE; else *dt &= ~SEND_NOCACHE; - /* Try fetching the document, or at least its head. :-) */ - err = gethttp (u, &hstat, dt); + /* Try fetching the document, or at least its head. */ + err = gethttp (u, &hstat, dt, proxy); /* It's unfortunate that wget determines the local filename before finding out the Content-Type of the file. Barring a major restructuring of the code, we need to re-set locf here, since gethttp() may have xrealloc()d - u->local to tack on ".html". */ + *hstat.local_file to tack on ".html". */ if (!opt.output_document) - locf = u->local; + locf = *hstat.local_file; else locf = opt.output_document; - /* In `-c' is used, check whether the file we're writing to - exists before we've done anything. If so, we'll refuse to - truncate it if the server doesn't support continued - downloads. */ - if (opt.always_rest) - hstat.no_truncate = file_exists_p (locf); - /* Time? */ tms = time_str (NULL); /* Get the new location (with or without the redirection). */ @@ -1556,31 +1795,32 @@ File `%s' already there, will not retrieve.\n"), u->local); /* 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. */ - FREEHSTAT (hstat); + free_hstat (&hstat); printwhat (count, opt.ntry); continue; break; case HOSTERR: case CONREFUSED: case PROXERR: case AUTHFAILED: case SSLERRCTXCREATE: case CONTNOTSUPPORTED: /* Fatal errors just return from the function. */ - FREEHSTAT (hstat); - xfree (filename_plus_orig_suffix); /* must precede every return! */ + free_hstat (&hstat); + FREE_MAYBE (dummy); return err; break; case FWRITEERR: case FOPENERR: /* Another fatal error. */ logputs (LOG_VERBOSE, "\n"); logprintf (LOG_NOTQUIET, _("Cannot write to `%s' (%s).\n"), - u->local, strerror (errno)); - FREEHSTAT (hstat); + *hstat.local_file, strerror (errno)); + free_hstat (&hstat); + FREE_MAYBE (dummy); return err; break; case CONSSLERR: /* Another fatal error. */ logputs (LOG_VERBOSE, "\n"); logprintf (LOG_NOTQUIET, _("Unable to establish SSL connection.\n")); - FREEHSTAT (hstat); - xfree (filename_plus_orig_suffix); /* must precede every return! */ + free_hstat (&hstat); + FREE_MAYBE (dummy); return err; break; case NEWLOCATION: @@ -1590,13 +1830,20 @@ File `%s' already there, will not retrieve.\n"), u->local); logprintf (LOG_NOTQUIET, _("ERROR: Redirection (%d) without location.\n"), hstat.statcode); - xfree (filename_plus_orig_suffix); /* must precede every return! */ + free_hstat (&hstat); + FREE_MAYBE (dummy); return WRONGCODE; } - FREEHSTAT (hstat); - xfree (filename_plus_orig_suffix); /* must precede every return! */ + free_hstat (&hstat); + FREE_MAYBE (dummy); return NEWLOCATION; break; + case RETRUNNEEDED: + /* The file was already fully retrieved. */ + free_hstat (&hstat); + FREE_MAYBE (dummy); + return RETROK; + break; case RETRFINISHED: /* Deal with you later. */ break; @@ -1609,15 +1856,15 @@ File `%s' already there, will not retrieve.\n"), u->local); if (!opt.verbose) { /* #### Ugly ugly ugly! */ - char *hurl = str_url (u->proxy ? u->proxy : u, 1); + char *hurl = url_string (u, 1); logprintf (LOG_NONVERBOSE, "%s:\n", hurl); xfree (hurl); } logprintf (LOG_NOTQUIET, _("%s ERROR %d: %s.\n"), tms, hstat.statcode, hstat.error); logputs (LOG_VERBOSE, "\n"); - FREEHSTAT (hstat); - xfree (filename_plus_orig_suffix); /* must precede every return! */ + free_hstat (&hstat); + FREE_MAYBE (dummy); return WRONGCODE; } @@ -1660,8 +1907,8 @@ Last-modified header invalid -- time-stamp ignored.\n")); logprintf (LOG_VERBOSE, _("\ Server file no newer than local file `%s' -- not retrieving.\n\n"), local_filename); - FREEHSTAT (hstat); - xfree (filename_plus_orig_suffix); /*must precede every return!*/ + free_hstat (&hstat); + FREE_MAYBE (dummy); return RETROK; } else if (tml >= tmr) @@ -1671,7 +1918,7 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); logputs (LOG_VERBOSE, _("Remote file is newer, retrieving.\n")); } - FREEHSTAT (hstat); + free_hstat (&hstat); continue; } if ((tmr != (time_t) (-1)) @@ -1690,7 +1937,7 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); fl = opt.output_document; } else - fl = u->local; + fl = *hstat.local_file; if (fl) touch (fl, tmr); } @@ -1699,15 +1946,11 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); if (opt.spider) { logprintf (LOG_NOTQUIET, "%d %s\n\n", hstat.statcode, hstat.error); - xfree (filename_plus_orig_suffix); /* must precede every return! */ + FREE_MAYBE (dummy); return RETROK; } - /* It is now safe to free the remainder of hstat, since the - strings within it will no longer be used. */ - FREEHSTAT (hstat); - - tmrate = rate (hstat.len - hstat.restval, hstat.dltime, 0); + tmrate = retr_rate (hstat.len - hstat.restval, hstat.dltime, 0); if (hstat.len == hstat.contlen) { @@ -1729,7 +1972,8 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); else downloaded_file(FILE_DOWNLOADED_NORMALLY, locf); - xfree(filename_plus_orig_suffix); /* must precede every return! */ + free_hstat (&hstat); + FREE_MAYBE (dummy); return RETROK; } else if (hstat.res == 0) /* No read error */ @@ -1755,7 +1999,8 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); else downloaded_file(FILE_DOWNLOADED_NORMALLY, locf); - xfree (filename_plus_orig_suffix); /* must precede every return! */ + free_hstat (&hstat); + FREE_MAYBE (dummy); return RETROK; } else if (hstat.len < hstat.contlen) /* meaning we lost the @@ -1765,6 +2010,7 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); _("%s (%s) - Connection closed at byte %ld. "), tms, tmrate, hstat.len); printwhat (count, opt.ntry); + free_hstat (&hstat); continue; } else if (!opt.kill_longer) /* meaning we got more than expected */ @@ -1784,7 +2030,8 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); else downloaded_file(FILE_DOWNLOADED_NORMALLY, locf); - xfree (filename_plus_orig_suffix); /* must precede every return! */ + free_hstat (&hstat); + FREE_MAYBE (dummy); return RETROK; } else /* the same, but not accepted */ @@ -1793,6 +2040,7 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); _("%s (%s) - Connection closed at byte %ld/%ld. "), tms, tmrate, hstat.len, hstat.contlen); printwhat (count, opt.ntry); + free_hstat (&hstat); continue; } } @@ -1804,6 +2052,7 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); _("%s (%s) - Read error at byte %ld (%s)."), tms, tmrate, hstat.len, strerror (errno)); printwhat (count, opt.ntry); + free_hstat (&hstat); continue; } else /* hstat.res == -1 and contlen is given */ @@ -1813,6 +2062,7 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); tms, tmrate, hstat.len, hstat.contlen, strerror (errno)); printwhat (count, opt.ntry); + free_hstat (&hstat); continue; } } @@ -1820,12 +2070,42 @@ The sizes do not match (local %ld) -- retrieving.\n"), local_size); break; } while (!opt.ntry || (count < opt.ntry)); - xfree (filename_plus_orig_suffix); /* must precede every return! */ return TRYLIMEXC; } /* Converts struct tm to time_t, assuming the data in tm is UTC rather - than local timezone (mktime assumes the latter). + than local timezone. + + mktime is similar but assumes struct tm, also known as the + "broken-down" form of time, is in local time zone. mktime_from_utc + uses mktime to make the conversion understanding that an offset + will be introduced by the local time assumption. + + mktime_from_utc then measures the introduced offset by applying + gmtime to the initial result and applying mktime to the resulting + "broken-down" form. The difference between the two mktime results + is the measured offset which is then subtracted from the initial + mktime result to yield a calendar time which is the value returned. + + tm_isdst in struct tm is set to 0 to force mktime to introduce a + consistent offset (the non DST offset) since tm and tm+o might be + on opposite sides of a DST change. + + Some implementations of mktime return -1 for the nonexistent + localtime hour at the beginning of DST. In this event, use + mktime(tm - 1hr) + 3600. + + Schematically + mktime(tm) --> t+o + gmtime(t+o) --> tm+o + mktime(tm+o) --> t+2o + t+o - (t+2o - t+o) = t + + Note that glibc contains a function of the same purpose named + `timegm' (reverse of gmtime). But obviously, it is not universally + available, and unfortunately it is not straightforwardly + extractable for use here. Perhaps configure should detect timegm + and use it where available. Contributed by Roger Beeman , with the help of Mark Baushke and the rest of the Gurus at CISCO. @@ -1867,8 +2147,8 @@ mktime_from_utc (struct tm *t) `+X', or at the end of the string. In extended regexp parlance, the function returns 1 if P matches - "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (a valid result of - strptime()) is considered a failure and 0 is returned. */ + "^ *(GMT|[+-][0-9]|$)", 0 otherwise. P being NULL (which strptime + can return) is considered a failure and 0 is returned. */ static int check_end (const char *p) { @@ -1884,28 +2164,50 @@ check_end (const char *p) return 0; } -/* Convert TIME_STRING time to time_t. TIME_STRING can be in any of - the three formats RFC2068 allows the HTTP servers to emit -- - RFC1123-date, RFC850-date or asctime-date. Timezones are ignored, - and should be GMT. - - We use strptime() to recognize various dates, which makes it a - little bit slacker than the RFC1123/RFC850/asctime (e.g. it always - allows shortened dates and months, one-digit days, etc.). It also - allows more than one space anywhere where the specs require one SP. - The routine should probably be even more forgiving (as recommended - by RFC2068), but I do not have the time to write one. - - Return the computed time_t representation, or -1 if all the - schemes fail. - - Needless to say, what we *really* need here is something like - Marcus Hennecke's atotm(), which is forgiving, fast, to-the-point, - and does not use strptime(). atotm() is to be found in the sources - of `phttpd', a little-known HTTP server written by Peter Erikson. */ +/* 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. + Timezones are ignored, and should be GMT. + + Return the computed time_t representation, or -1 if the conversion + fails. + + This function uses strptime with various string formats for parsing + TIME_STRING. This results in a parser that is not as lenient in + interpreting TIME_STRING as I would like it to be. Being based on + strptime, it always allows shortened months, one-digit days, etc., + but due to the multitude of formats in which time can be + represented, an ideal HTTP time parser would be even more + forgiving. It should completely ignore things like week days and + concentrate only on the various forms of representing years, + months, days, hours, minutes, and seconds. For example, it would + be nice if it accepted ISO 8601 out of the box. + + I've investigated free and PD code for this purpose, but none was + usable. getdate was big and unwieldy, and had potential copyright + issues, or so I was informed. Dr. Marcus Hennecke's atotm(), + distributed with phttpd, is excellent, but we cannot use it because + it is not assigned to the FSF. So I stuck it with strptime. */ + time_t -http_atotm (char *time_string) +http_atotm (const char *time_string) { + /* NOTE: Solaris strptime man page claims that %n and %t match white + space, but that's not universally available. Instead, we simply + use ` ' to mean "skip all WS", which works under all strptime + 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 */ + }; + + int i; struct tm t; /* According to Roger Beeman, we need to initialize tm_isdst, since @@ -1913,41 +2215,23 @@ http_atotm (char *time_string) t.tm_isdst = 0; /* Note that under foreign locales Solaris strptime() fails to - recognize English dates, which renders this function useless. I - assume that other non-GNU strptime's are plagued by the same - disease. We solve this by setting only LC_MESSAGES in - i18n_initialize(), instead of LC_ALL. + recognize English dates, which renders this function useless. We + solve this by being careful not to affect LC_TIME when + initializing locale. - Another solution could be to temporarily set locale to C, invoke + 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. - Also note that none of this is necessary under GNU strptime(), - because it recognizes both international and local dates. */ - - /* NOTE: We don't use `%n' for white space, as OSF's strptime uses - it to eat all white space up to (and including) a newline, and - the function fails if there is no newline (!). - - Let's hope all strptime() implementations use ` ' to skip *all* - whitespace instead of just one (it works that way on all the - systems I've tested it on). */ - - /* RFC1123: Thu, 29 Jan 1998 22:12:57 */ - if (check_end (strptime (time_string, "%a, %d %b %Y %T", &t))) - return mktime_from_utc (&t); - /* RFC850: Thursday, 29-Jan-98 22:12:57 */ - if (check_end (strptime (time_string, "%A, %d-%b-%y %T", &t))) - return mktime_from_utc (&t); - /* pseudo-RFC850: Thu, 29-Jan-1998 22:12:57 - (google.com uses this for their cookies.)*/ - if (check_end (strptime (time_string, "%a, %d-%b-%Y %T", &t))) - return mktime_from_utc (&t); - /* asctime: Thu Jan 29 22:12:57 1998 */ - if (check_end (strptime (time_string, "%a %b %d %T %Y", &t))) - return mktime_from_utc (&t); - /* Failure. */ + GNU strptime does not have this problem because it recognizes + both international and local dates. */ + + for (i = 0; i < ARRAY_SIZE (time_formats); i++) + if (check_end (strptime (time_string, time_formats[i], &t))) + return mktime_from_utc (&t); + + /* All formats have failed. */ return -1; } @@ -2081,7 +2365,7 @@ dump_hash (unsigned char *buf, const unsigned char *hash) /* Take the line apart to find the challenge, and compose a digest authorization header. See RFC2069 section 2.1.2. */ -char * +static char * digest_authentication_encode (const char *au, const char *user, const char *passwd, const char *method, const char *path) @@ -2154,37 +2438,37 @@ digest_authentication_encode (const char *au, const char *user, /* Calculate the digest value. */ { - struct md5_ctx ctx; + ALLOCA_MD5_CONTEXT (ctx); unsigned char hash[MD5_HASHLEN]; unsigned char a1buf[MD5_HASHLEN * 2 + 1], a2buf[MD5_HASHLEN * 2 + 1]; unsigned char response_digest[MD5_HASHLEN * 2 + 1]; /* A1BUF = H(user ":" realm ":" password) */ - md5_init_ctx (&ctx); - md5_process_bytes (user, strlen (user), &ctx); - md5_process_bytes (":", 1, &ctx); - md5_process_bytes (realm, strlen (realm), &ctx); - md5_process_bytes (":", 1, &ctx); - md5_process_bytes (passwd, strlen (passwd), &ctx); - md5_finish_ctx (&ctx, hash); + gen_md5_init (ctx); + gen_md5_update ((unsigned char *)user, strlen (user), ctx); + gen_md5_update ((unsigned char *)":", 1, ctx); + gen_md5_update ((unsigned char *)realm, strlen (realm), ctx); + gen_md5_update ((unsigned char *)":", 1, ctx); + gen_md5_update ((unsigned char *)passwd, strlen (passwd), ctx); + gen_md5_finish (ctx, hash); dump_hash (a1buf, hash); /* A2BUF = H(method ":" path) */ - md5_init_ctx (&ctx); - md5_process_bytes (method, strlen (method), &ctx); - md5_process_bytes (":", 1, &ctx); - md5_process_bytes (path, strlen (path), &ctx); - md5_finish_ctx (&ctx, hash); + gen_md5_init (ctx); + gen_md5_update ((unsigned char *)method, strlen (method), ctx); + gen_md5_update ((unsigned char *)":", 1, ctx); + gen_md5_update ((unsigned char *)path, strlen (path), ctx); + gen_md5_finish (ctx, hash); dump_hash (a2buf, hash); /* RESPONSE_DIGEST = H(A1BUF ":" nonce ":" A2BUF) */ - md5_init_ctx (&ctx); - md5_process_bytes (a1buf, MD5_HASHLEN * 2, &ctx); - md5_process_bytes (":", 1, &ctx); - md5_process_bytes (nonce, strlen (nonce), &ctx); - md5_process_bytes (":", 1, &ctx); - md5_process_bytes (a2buf, MD5_HASHLEN * 2, &ctx); - md5_finish_ctx (&ctx, hash); + gen_md5_init (ctx); + gen_md5_update (a1buf, MD5_HASHLEN * 2, ctx); + gen_md5_update ((unsigned char *)":", 1, ctx); + gen_md5_update ((unsigned char *)nonce, strlen (nonce), ctx); + gen_md5_update ((unsigned char *)":", 1, ctx); + gen_md5_update (a2buf, MD5_HASHLEN * 2, ctx); + gen_md5_finish (ctx, hash); dump_hash (response_digest, hash); res = (char*) xmalloc (strlen (user) @@ -2249,3 +2533,10 @@ create_authorization_line (const char *au, const char *user, #endif /* USE_DIGEST */ return wwwauth; } + +void +http_cleanup (void) +{ + if (pc_last_host_ip) + address_list_release (pc_last_host_ip); +}