[cvsnt] Re: cvs commit hangs when Visual SlickEdit views a file

[Oliver Koltermann]

tmh at nodomain.org (Tony Hoyle) writes:

> >> Checking in xyz.txt;
> >> R:/Repo/xyz.txt,v  <-- xyz.txt
> >> new revision: 1.8; previous revision: 1.7
> >> done
>
> It looks like you're using a network share in local mode, unless I'm reading
> you incorrectly.

Thanks for the comment, Tony, but in this case the output is
misleading. We're working in pserver mode without repository prefix
because that feature seemd to cause trouble long time ago when we
first tested CVSNT.

Maybe I should mention that there is a network share access set-up on
my workstation because I run a local ViewCVS-Testserver. But this
has no drive letter assigned (it's not R:...), I'm using UNC for
this. And none of the developers with the problems on discussion here
have this access method.

I'll try to give an overview of our installation - maybe someone sees
any problems with it:

Server is a Windows 2000 Server System also used for terminal
services. It's running CVSNT build 77a with three repositorys resting
on local drive R:. (The drive letter is setten up like this because of
the terminal server funktionality). The only installed protocol dll is
pserver (intranet-use only). Authentication goes over the passwd file
without impersonation. Because of ancient Novell 4 the user accounts
arn't even known by the server.

The clients are working with Windows NT4, 2000 and XP, with CVSNT
build 72 and WinCVS 1.3.12.1 (beta 12, some still with beta 10). They
use a CVSROOT like ":pserver:user at internal.dns.name:R:/Repo/A". The
developers having problems have the editor Visual SlickEdit open all
the time. I could repeat the hang one time with this tool open, but it
never happened again from then on. It was my hope to find the problem
there but it could be unrelated.

What can you all suggest? I could try to run a batch-commit one a
developers machine to repoduce the problem with -t option but I had no
luck when trying this on my workstation even with hardest
VSlick-usage. Or is there any possibility to generate trace output to
a logfile on the server?

I'm feeling really helpless with this. Any comment would give me new
hope...

Thanks,
  OK.