[cvsnt] Re: New Bugs - Upgrading from CVSNT 2.5.01 (Travis) Build 1976 to CVSNT 2.5.02 (Servalan) Build 2115

[Jakob Kruse]

"Tony Hoyle" <tony.hoyle at march-hare.com> wrote in message 
news:dj3lft$9u8$3 at paris.nodomain.org...
> Jakob Kruse wrote:
>>
>> I tried the exact same thing earlier, when the server was 2.5.02. Same 
>> output except for the last line, being the "unknown error -1" thing.
>>
> You will occasionally get the -1 error... it's just the way the timing 
> works sometimes (about 1 in 20 for me if I run it in a loop).  Win32 isn't 
> very good at handling one end of a socket closing early....

Definitely not an occasional thing. 100% on affected machines/logins.

> If your machine isn't in a domain then it could just be down to the 
> configuration of the network, or even whether there are active connections 
> already (network shares normally).

There are users both on the LAN and on the Internet. None of the Internet 
users are logged on to the CVS server in any way other that what CVS does. 
None of the Internet users access the CVS server using the username/password 
they are currently logged on to Windows with. This has all worked fine until 
2.5, and it still works fine in many places.

> If you're using kerberos (as seems likely) all machines should have a 
> common clock source also.

How can I find out if I use Kerberos? The server is out-of-the-box. Clients 
are out-of-the-box using sspi.

> In the last resort you can force NTLM but that's something that normally 
> wouldn't be needed (except for old servers pre-2.0.58, because they don't 
> handle Kerberos->NTLM fallback properly).

No old servers involved. How can I force NTLM (as a test)? Is it a server 
thing or a client thing?

/Jakob