Problem with tempfiles

[Jens T Thielemann <je...@gmail.com>]

Hi!

I have a problem with Subversion in that its atomic updates of files
conflicts with my virus killer (this is exactly the same problem as reported
in http://svn.haxx.se/users/archive-2006-08/0847.shtml). This makes
subversion usually fail in any action.

While arguments can be made that this really is a virus killer problem, this
does not really help me out. As I understand the problem, the virus killer
will temporarily look at the file, and if Subversion tries to move the file
simultaneously (due to its mechanism of providing atomic file updates), the
move will fail. There is thus an element of a race condition here.

Would it be possible to enhance Subversion to retry such move operations?
That is, if the first move fails, wait 0.5 seconds and retry? This would
probably enhance compatibility with virus killer and also indexing software,
and make the Subversion client much more usable in a viruskilling
environment.

Are there any work underway attempting to resolve this issue?

I am reproducing the error under TortoiseSVN 1.4.8, Build 12137 - 32 Bit ,
2008/02/16 11:09:38.

Best regards,

Jens Thielemann

(I am aware that there is probably some virus killers out there that don't
cause these conflicts, but I'm running subversion on a company machine, so
the virus killer vendor is not selectable...)

Re: Problem with tempfiles

[Andy Levy <an...@gmail.com>]

On Thu, May 22, 2008 at 3:08 AM, Jens T Thielemann <je...@gmail.com> wrote:
> Hi!
>
> I have a problem with Subversion in that its atomic updates of files
> conflicts with my virus killer (this is exactly the same problem as reported
> in http://svn.haxx.se/users/archive-2006-08/0847.shtml). This makes
> subversion usually fail in any action.
>
> While arguments can be made that this really is a virus killer problem, this
> does not really help me out. As I understand the problem, the virus killer
> will temporarily look at the file, and if Subversion tries to move the file
> simultaneously (due to its mechanism of providing atomic file updates), the
> move will fail. There is thus an element of a race condition here.
>
> Would it be possible to enhance Subversion to retry such move operations?
> That is, if the first move fails, wait 0.5 seconds and retry? This would
> probably enhance compatibility with virus killer and also indexing software,
> and make the Subversion client much more usable in a viruskilling
> environment.
>
> Are there any work underway attempting to resolve this issue?

I think SVN already does attempt to retry. The problem is that some
virus scanners hold onto the files longer than that (which is
excessive). The fact that most scanners don't seem to cause issues
should be a strong indicator that the problem is not really with SVN.

Tell your virus scanner to ignore the .svn directories. I understand
that it's a company machine, but company policy should *enable* you to
do your job, not *prevent* you from doing it.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org