-
The previous patch is incorrect. It calls shutdown() without the 'how' parameter. My initial tests passed SHUT_RDWR as the 'how' parameter. Further testing has shown that SHUT_WR works best. Additional research revealed that a shutdown() should always be followed by a close(). So, the new patch looks like this:
diff -Naur axTLS.orig/httpd/axhttpd.c axTLS/httpd/axhttpd.c
---...
2009-10-14 15:17:43 UTC in axTLS Embedded SSL
-
The function shutdown() should be used to close the cn->networkdesc instead of SOCKET_CLOSE(). If the axhttpd launches a CGI app. and the CGI app. forks a daemon, the cn->networkdesc is still not closed after the call to SOCKET_CLOSE(). So, the browser never receives a FIN from the server and the browser continues to look like it is waiting for more data. The attached patch shows the simple...
2009-10-14 00:14:56 UTC in axTLS Embedded SSL
-
The function shutdown() should be used to close the cn->networkdesc instead of SOCKET_CLOSE(). If the axhttpd launches a CGI app. and the CGI app. forks a daemon, the cn->networkdesc is still not closed after the call to SOCKET_CLOSE(). So, the browser never receives a FIN from the server and the browser continues to look like it is waiting for more data. The attached patch shows the simple...
2009-10-13 20:13:40 UTC in axTLS Embedded SSL
-
Further testing showed that the first patch I uploaded did not use a big enough buffer. I deleted that patch and uploaded a new one with a 16k+1 buffer size. The larger buffer size should allow for the the largest chunk of data read from the socket.
2009-02-20 18:47:05 UTC in axTLS Embedded SSL
-
File Added: maxbuflength.patch.
2009-02-20 18:42:30 UTC in axTLS Embedded SSL
-
Using FF3 or IE7 to upload a file using ssl will never complete. The reason is that while in ssl mode, the axhttpd server does not copy all of the file data off the socket into the buffer.
It reads a large chunk off the socket, then copies only up to 1023 bytes into the buffer. The rest of the data gets discarded.
I created a patch to get around this. The patch was made against axTLS...
2009-02-19 20:58:10 UTC in axTLS Embedded SSL
