TestBot IPv6 enabled

[Greg Geldorp]

Hi,

All TestBot VMs (with the exception of the Win9x ones) now have IPv6 connectivity. I was surprised that there was even a IPv6 stack available for NT4, it was produced by Microsoft Research.
The VMs are (and always have been) behind a IPv4 NAT router. That same router now acts as a IPv6 router/firewall, with full outbound access but blocked inbound access.
The website is also IPv6 capable, just waiting for the AAAA record to be added to the nameserver.

Greg.

[Austin Lund]

On 14 November 2010 07:40, Greg Geldorp  wrote:
> Hi,
>
> All TestBot VMs (with the exception of the Win9x ones) now have IPv6 connectivity. I was surprised that there was even a IPv6 stack available for NT4, it was produced by Microsoft Research.
> The VMs are (and always have been) behind a IPv4 NAT router. That same router now acts as a IPv6 router/firewall, with full outbound access but blocked inbound access.
> The website is also IPv6 capable, just waiting for the AAAA record to be added to the nameserver.
>
> Greg.

Are these test failures:

http://test.winehq.org/data/b8d0c0dcc70dc06674c9b0f42230357c736359f9/xp_wtb-...

related to this change?  Is there an issue on the testbot with IPV6
and security certificates?

[Greg Geldorp]

> From: Austin Lund
> On 14 November 2010 07:40, Greg Geldorp  wrote:
> > Hi,
> >
> > All TestBot VMs (with the exception of the Win9x ones) now have IPv6
> > connectivity. I was surprised that there was even a IPv6 stack available
> > for NT4, it was produced by Microsoft Research.
> > The VMs are (and always have been) behind a IPv4 NAT router. That same
> > router now acts as a IPv6 router/firewall, with full outbound access
> > but blocked inbound access.
> > The website is also IPv6 capable, just waiting for the AAAA record to be
> > added to the nameserver.
>
> Are these test failures:
>
> http://test.winehq.org/data/b8d0c0dcc70dc06674c9b0f42230357c736359f9/xp_wtb-...
>
> related to this change?

Yes, I suspect they are. These failures started appearing right after the IPv6 change.
> Is there an issue on the testbot with IPV6 and security certificates?

Sorry, haven't had the time to investigate yet (only looked up the error code which is
ERROR_INTERNET_SEC_CERT_CN_INVALID).

Greg.

[Juan Lang]

> Sorry, haven't had the time to investigate yet (only looked up the error code which is
> ERROR_INTERNET_SEC_CERT_CN_INVALID).

That would imply that the certificate checking is failing.  Please get
a +crypt,+chain log of the failing tests and attach to a bug for me to
have a look.
--Juan

[Greg Geldorp]

> From: Juan Lang [mailto:[e-mail]]
> > Sorry, haven't had the time to investigate yet (only looked up the
> > error code which is ERROR_INTERNET_SEC_CERT_CN_INVALID).
>
> That would imply that the certificate checking is failing.  Please get
> a +crypt,+chain log of the failing tests and attach to a bug for me to
> have a look.

That's going to be a bit of a problem. The tests fail on Windows, haven't
found a way yet to convince the Windows DLLs to produce trace logs...

Greg.

[Juan Lang]

> That's going to be a bit of a problem. The tests fail on Windows, haven't
> found a way yet to convince the Windows DLLs to produce trace logs...

Ha, right you are.  Hm.  Well, a wireshark capture of the connection
might help, to verify that the web server is in fact attaching the
correct certificate to the connection in the IPv6 case.
--Juan

[Greg Geldorp]

> From: Juan Lang
> Ha, right you are.  Hm.  Well, a wireshark capture of the connection
> might help, to verify that the web server is in fact attaching the
> correct certificate to the connection in the IPv6 case.

The URL in question is actually https://testbot.winehq.org/. Browsing
there using Firefox from one of the test VMs showed that the server was
indeed using the wrong certificate on a IPv6 connection. I've had a
stern talk with it and it promised to do better on the next winetest run.

Greg.