Petr Pisar's privoxy-3.0.12-ipv6-3.diff.

[privoxy.git] / urlmatch.c

diff --git a/urlmatch.c b/urlmatch.c
index 98cecf2..fa11f20 100644 (file)
--- a/urlmatch.c
+++ b/urlmatch.c
@@ -1,4 +1,4 @@
-const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.43 2008/05/04 13:30:55 fabiankeil Exp $";
+const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.47 2009/03/02 19:18:10 fabiankeil Exp $";
 /*********************************************************************
  *
  * File        :  $Source: /cvsroot/ijbswa/current/urlmatch.c,v $
@@ -6,8 +6,8 @@ const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.43 2008/05/04 13:30:55 fabianke
  * Purpose     :  Declares functions to match URLs against URL
  *                patterns.
  *
- * Copyright   :  Written by and Copyright (C) 2001-2003, 2006-2008 the SourceForge
- *                Privoxy team. http://www.privoxy.org/
+ * Copyright   :  Written by and Copyright (C) 2001-2009
+ *                the Privoxy team. http://www.privoxy.org/
  *
  *                Based on the Internet Junkbuster originally written
  *                by and Copyright (C) 1997 Anonymous Coders and
@@ -33,6 +33,20 @@ const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.43 2008/05/04 13:30:55 fabianke
  *
  * Revisions   :
  *    $Log: urlmatch.c,v $
+ *    Revision 1.47  2009/03/02 19:18:10  fabiankeil
+ *    Streamline parse_http_request()'s prototype. As
+ *    cparser pointed out it doesn't actually use csp.
+ *
+ *    Revision 1.46  2009/02/11 19:31:32  fabiankeil
+ *    Reject request lines that end with neither HTTP/1.0 nor HTTP/1.1.
+ *
+ *    Revision 1.45  2008/06/21 21:19:18  fabiankeil
+ *    Silence bogus compiler warning.
+ *
+ *    Revision 1.44  2008/05/04 16:18:32  fabiankeil
+ *    Provide parse_http_url() with a third parameter to specify
+ *    whether or not URLs without protocol are acceptable.
+ *
  *    Revision 1.43  2008/05/04 13:30:55  fabiankeil
  *    Streamline parse_http_url()'s prototype.
  *
@@ -526,8 +540,40 @@ jb_err parse_http_url(const char *url, struct http_request *http, int require_pr
          host = buf;
       }
 
+      /* Move after hostname before port number */
+      if (*host == '[')
+      {
+         /* Numeric IPv6 address delimited by brackets */
+         host++;
+         port = strchr(host, ']');
+
+         if (port == NULL)
+         {
+            /* Missing closing bracket */
+            freez(buf);
+            return JB_ERR_PARSE;
+         }
+
+         *port++='\0';
+         
+         if (*port == '\0')
+         {
+            port = NULL;
+         }
+         else if (*port != ':')
+         {
+            /* Garbage after closing bracket */
+            freez(buf);
+            return JB_ERR_PARSE;
+         }
+      }
+      else
+      {
+         /* Plain non-escaped hostname */ 
+         port = strchr(host, ':');
+      }
+
       /* check if url contains port */
-      port = strchr(host, ':');
       if (port != NULL)
       {
          /* Contains port */
@@ -623,7 +669,6 @@ static int unknown_method(const char *method)
  * Parameters  :
  *          1  :  req = HTTP request line to break down
  *          2  :  http = pointer to the http structure to hold elements
- *          3  :  csp = Current client state (buffers, headers, etc...)
  *
  * Returns     :  JB_ERR_OK on success
  *                JB_ERR_MEMORY on out of memory
@@ -631,9 +676,7 @@ static int unknown_method(const char *method)
  *                                  or >100 domains deep.
  *
  *********************************************************************/
-jb_err parse_http_request(const char *req,
-                          struct http_request *http,
-                          const struct client_state *csp)
+jb_err parse_http_request(const char *req, struct http_request *http)
 {
    char *buf;
    char *v[10]; /* XXX: Why 10? We should only need three. */
@@ -671,6 +714,14 @@ jb_err parse_http_request(const char *req,
       return JB_ERR_PARSE;
    }
 
+   if (strcmpic(v[2], "HTTP/1.1") && strcmpic(v[2], "HTTP/1.0"))
+   {
+      log_error(LOG_LEVEL_ERROR, "The only supported HTTP "
+         "versions are 1.0 and 1.1. This rules out: %s", v[2]);
+      freez(buf);
+      return JB_ERR_PARSE;
+   }
+
    http->ssl = !strcmpic(v[0], "CONNECT");
 
    err = parse_http_url(v[1], http, !http->ssl);
@@ -726,7 +777,7 @@ static jb_err compile_pattern(const char *pattern, enum regex_anchoring anchorin
 {
    int errcode;
    char rebuf[BUFFER_SIZE];
-   const char *fmt;
+   const char *fmt = NULL;
 
    assert(pattern);
    assert(strlen(pattern) < sizeof(rebuf) - 2);
@@ -825,7 +876,31 @@ static jb_err compile_url_pattern(struct url_spec *url, char *buf)
       *p = '\0';
    }
 
-   p = strchr(buf, ':');
+   /* XXX: IPv6 numeric hostname contains colons, thus we need to delimit the
+    * hostname before real port separator. Because brackets are used in
+    * hostname matching on lower layer, we can't use it. I decided to use
+    * angle brackets '<' '>' instead. */
+   if (buf[0] == '<' && NULL != (p = strchr(buf + 1, '>')))
+   {
+      *p++ = '\0';
+      buf++;
+
+      if (*p == '\0')
+      {
+         /* Only IPv6 address without port number */
+         p = NULL;
+      }
+      else if (*p != ':')
+      {
+         /* Garbage after address delimiter */
+         return JB_ERR_PARSE;
+      }
+   }
+   else
+   {
+      p = strchr(buf, ':');
+   }
+
    if (NULL != p)
    {
       *p++ = '\0';
@@ -1430,4 +1505,6 @@ int match_portlist(const char *portlist, int port)
   Local Variables:
   tab-width: 3
   end:
+
+  vim:softtabstop=3 shiftwidth=3
 */