Greetings:
Why does the procedure:
Java_java_net_PlainDatagramSocketImpl_datagramSocketCreate(...), defined
in the file: jdk/src/solaris/native/java/net/PlainDatagramSocketImpl.c,
close the given socket before returning after an error in some cases,
but not others? Shouldn't the behaviour be the same throughout the
procedure? I think this behaviour should be: throw exception, close socket,
return.
For examples consider the following:
if ((fd = JVM_Socket(domain, SOCK_DGRAM, 0)) == JVM_IO_ERR) {
NET_ThrowByNameWithLastError(env, JNU_JAVANETPKG "SocketException",
"Error creating socket");
return;
}
In this case, the socket was never opened, so there no need to
close it.
#ifdef AF_INET6
/* Disable IPV6_V6ONLY to ensure dual-socket support */
if (domain == AF_INET6) {
arg = 0;
if (setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY, (char*)&arg,
sizeof(int)) < 0) {
NET_ThrowNew(env, errno, "cannot set IPPROTO_IPV6");
close(fd); <<<< This seems OK
return;
}
}
#endif /* AF_INET6 */
#ifdef __APPLE__
arg = 65507;
if (JVM_SetSockOpt(fd, SOL_SOCKET, SO_SNDBUF,
(char *)&arg, sizeof(arg)) < 0) {
JNU_ThrowByName(env, JNU_JAVANETPKG "SocketException",
strerror(errno));
return; <<<<<<<<<< Why not close it? If we can't allocate the
desired buffer size, is it then OK, to just use what we have? It's
interesting that this would fail...
}
if (JVM_SetSockOpt(fd, SOL_SOCKET, SO_RCVBUF,
(char *)&arg, sizeof(arg)) < 0) {
JNU_ThrowByName(env, JNU_JAVANETPKG "SocketException",
strerror(errno));
return; <<<< likewise
}
#endif /* __APPLE__ */
if (setsockopt(fd, SOL_SOCKET, SO_BROADCAST, (char*) &t,
sizeof(int))<0) {
JNU_ThrowByName(env, JNU_JAVANETPKG "SocketException",
"Unable to set SO_BROADCAST");
return; <<< setsockopt() failed, we threw an exception. We
didn't close the socket.
}
--
John Zavgren
john.zavg...@oracle.com
603-821-0904
US-Burlington-MA
