- * Revisions :
- * $Log: gateway.c,v $
- * Revision 1.32 2008/10/16 16:27:22 fabiankeil
- * Fix compiler warning.
- *
- * Revision 1.31 2008/10/16 07:31:11 fabiankeil
- * - Factor socket_is_still_usable() out of get_reusable_connection().
- * - If poll() isn't available, show a warning and assume the socket
- * is still usable.
- *
- * Revision 1.30 2008/10/13 17:31:03 fabiankeil
- * If a remembered connection is no longer usable and
- * has been marked closed, don't bother checking if the
- * destination matches.
- *
- * Revision 1.29 2008/10/11 16:59:41 fabiankeil
- * Add missing dots for two log messages.
- *
- * Revision 1.28 2008/10/09 18:21:41 fabiankeil
- * Flush work-in-progress changes to keep outgoing connections
- * alive where possible. Incomplete and mostly #ifdef'd out.
- *
- * Revision 1.27 2008/09/27 15:05:51 fabiankeil
- * Return only once in forwarded_connect().
- *
- * Revision 1.26 2008/08/18 17:42:06 fabiankeil
- * Fix typo in macro name.
- *
- * Revision 1.25 2008/02/07 18:09:46 fabiankeil
- * In socks5_connect:
- * - make the buffers quite a bit smaller.
- * - properly report "socks5 server unreachable" failures.
- * - let strncpy() use the whole buffer. Using a length of 0xffu wasn't actually
- * wrong, but requires too much thinking as it doesn't depend on the buffer size.
- * - log a message if the socks5 server sends more data than expected.
- * - add some assertions and comments.
- *
- * Revision 1.24 2008/02/04 14:56:29 fabiankeil
- * - Fix a compiler warning.
- * - Stop assuming that htonl(INADDR_NONE) equals INADDR_NONE.
- *
- * Revision 1.23 2008/02/04 13:11:35 fabiankeil
- * Remember the cause of the SOCKS5 error for the CGI message.
- *
- * Revision 1.22 2008/02/03 13:46:15 fabiankeil
- * Add SOCKS5 support. Patch #1862863 by Eric M. Hopper with minor changes.
- *
- * Revision 1.21 2007/07/28 12:30:03 fabiankeil
- * Modified patch from Song Weijia (#1762559) to
- * fix socks requests on big-endian platforms.
- *
- * Revision 1.20 2007/05/14 10:23:48 fabiankeil
- * - Use strlcpy() instead of strcpy().
- * - Use the same buffer for socks requests and socks responses.
- * - Fix bogus warning about web_server_addr being used uninitialized.
- *
- * Revision 1.19 2007/01/25 14:09:45 fabiankeil
- * - Save errors in socks4_connect() to csp->error_message.
- * - Silence some gcc43 warnings, hopefully the right way.
- *
- * Revision 1.18 2006/07/18 14:48:46 david__schmidt
- * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
- * with what was really the latest development (the v_3_0_branch branch)
- *
- * Revision 1.16 2002/05/12 21:36:29 jongfoster
- * Correcting function comments
- *
- * Revision 1.15 2002/03/26 22:29:54 swa
- * we have a new homepage!
- *
- * Revision 1.14 2002/03/24 13:25:43 swa
- * name change related issues
- *
- * Revision 1.13 2002/03/13 00:29:59 jongfoster
- * Killing warnings
- *
- * Revision 1.12 2002/03/09 20:03:52 jongfoster
- * - Making various functions return int rather than size_t.
- * (Undoing a recent change). Since size_t is unsigned on
- * Windows, functions like read_socket that return -1 on
- * error cannot return a size_t.
- *
- * THIS WAS A MAJOR BUG - it caused frequent, unpredictable
- * crashes, and also frequently caused JB to jump to 100%
- * CPU and stay there. (Because it thought it had just
- * read ((unsigned)-1) == 4Gb of data...)
- *
- * - The signature of write_socket has changed, it now simply
- * returns success=0/failure=nonzero.
- *
- * - Trying to get rid of a few warnings --with-debug on
- * Windows, I've introduced a new type "jb_socket". This is
- * used for the socket file descriptors. On Windows, this
- * is SOCKET (a typedef for unsigned). Everywhere else, it's
- * an int. The error value can't be -1 any more, so it's
- * now JB_INVALID_SOCKET (which is -1 on UNIX, and in
- * Windows it maps to the #define INVALID_SOCKET.)
- *
- * - The signature of bind_port has changed.
- *
- * Revision 1.11 2002/03/08 17:46:04 jongfoster
- * Fixing int/size_t warnings
- *
- * Revision 1.10 2002/03/07 03:50:19 oes
- * - Improved handling of failed DNS lookups
- * - Fixed compiler warnings
- *
- * Revision 1.9 2001/10/25 03:40:48 david__schmidt
- * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
- * threads to call select() simultaneously. So, it's time to do a real, live,
- * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
- * (native). Both versions will work, but using __OS2__ offers multi-threading.
- *
- * Revision 1.8 2001/09/13 20:10:12 jongfoster
- * Fixing missing #include under Windows
- *
- * Revision 1.7 2001/09/12 17:58:26 steudten
- *
- * add #include <string.h>
- *
- * Revision 1.6 2001/09/10 10:41:16 oes
- * Added #include in.h
- *
- * Revision 1.5 2001/07/29 18:47:57 jongfoster
- * Adding missing #include project.h
- *
- * Revision 1.4 2001/07/24 12:47:06 oes
- * Applied BeOS support update by Eugenia
- *
- * Revision 1.3 2001/06/09 10:55:28 jongfoster
- * Changing BUFSIZ ==> BUFFER_SIZE
- *
- * Revision 1.2 2001/06/07 23:11:38 jongfoster
- * Removing gateways[] list - no longer used.
- * Replacing function pointer in struct gateway with a directly
- * called function forwarded_connect(), which can do the common
- * task of deciding whether to connect to the web server or HTTP
- * proxy.
- * Replacing struct gateway with struct forward_spec
- * Fixing bug with SOCKS4A and HTTP proxy server in combination.
- * It was a bug which led to the connection being made to the web
- * server rather than the HTTP proxy, and also a buffer overrun.
- *
- * Revision 1.1.1.1 2001/05/15 13:58:54 oes
- * Initial import of version 2.9.3 source tree
- *
- *