RFR 8004925: java/net/Socks/SocksV4Test.java failing on all platforms

[Chris Hegarty]

On 12 Dec 2012, at 19:01, Dmitry Samersoff <dmitry.samersoff at oracle.com> wrote:

> On 2012-12-12 22:29, Chris Hegarty wrote:
>> On 12/12/2012 18:15, Dmitry Samersoff wrote:
>>> Chris,
>>> 
>>> According to rfc2606 TLD .invalid is reserved for cases like this one,
>> 
>> Yes, I came across this, but there is nothing to stop an internal DNS
>> server from resolving .invalid domains. Anyway, may
>> "doesnot.exist.invalid" would be sufficient.
> 
> You can't prevent internal DNS from resolving anything without doing
> some heavy tricks, so I guess doesnot.exist.invalid and error message
> that clear states that DNS setup violates rfc2606 is sufficient.

Agreed. I'll add a debugging message with the results of a lookup of said name, to help diagnose future failures.

-Chris.

> 
> -Dmitry
> 
> 
>> 
>> -Chris.
>> 
>> 
>>> 
>>> So, it seems to me
>>> 
>>>     domainame.invalid
>>> 
>>> is the best approach.
>>> 
>>> -Dmitry
>>> 
>>> 
>>> On 2012-12-12 20:15, Chris Hegarty wrote:
>>>> 
>>>> On 12/12/2012 14:14, Alan Bateman wrote:
>>>>> ....
>>>>>> -Chris.
>>>>> Would it be better if the test SocksServer had a list of knows that it
>>>>> always rejects? That might speed up the test too as it would avoid is
>>>>> trying to resolve host names or connect to hosts that don't exist.
>>>> 
>>>> The UHE is thrown from the client socket connect(). The Server in this
>>>> case doesn't ever receive the destination address or host name. It is
>>>> simply replying to the initial/opening SOCKS handshake.
>>>> 
>>>> The updated host name is still brittle ( if a .t TLD is ever registered!
>>>> ). I don't have a better alternative.
>>>> 
>>>> -Chris.
>>>> 
>>>>> 
>>>>> -Alan
>>>>> 
>>> 
>>> 
> 
> 
> -- 
> Dmitry Samersoff
> Oracle Java development team, Saint Petersburg, Russia
> * Give Rabbit time, and he'll always get the answer