I believe this is a problem in the networking code, which is causing
all the regression tests to fail on my machine. Any machine with a similar
setup will probably fail in the same manner.

The problem code looks like:

NetworkInterface.getNetworkInterfaces().getInetAddresses();

I have XP Pro with both a wired and wireless network card.

Before testing, the wireless adapter appears in the getNetworkInterfaces()
list (and ipconfig /all) *FIRST* before the wired one. When tests
like java/net/ipv6tests/TcpTest.java run, they look for the first IP
address, regardless of whether the adapter is actually on and enabled/disabled.
In my case, the wireless card was initially turned off,
and in the second case, it was completely
disabled. For some reason, an IP address is still somehow being obtained from
this adapter, and the connection is then attempted, which of course times out.
Even when I completely disabled my wireless adapter, it is still showing this as
a valid interface via getInterfaces().

In the disabled case, ipconfig /all doesn't even
mention the wireless interface. In the active case, it doesn't report any
active IPs when the wireless card is turned off.

See attachments for more info. output is the result of java/net/ipv6tests/TcpTest.java when run on this machine.`

###@###.### 2004-09-17