Re: Incorrect, repeatable, unremovable "in the way" errors

[John Nagle]

   Actually, I was a bit incorrect here.  The problem does seem
to go away if the sandbox is replaced.

   But the problem is induced by an incorrect timestamp on
a commit from one client, and affects other clients via the
server.  It's apparently possible to check in with a timestamp
in the future from the server's perspective, which causes
problems.  A clock inconsistency check on commit would prevent
this problem.

                                John Nagle

John Nagle wrote:
>   There's a known problem with spurious "in the way" errors which survive
> even removal and re-creation of the sandbox. A good description of this
> problem was provided back in 2001:
>
> http://mail.gnu.org/archive/html/info-cvs/2001-03/msg00163.html
>
> Read through the discussion and note that the problem did NOT go away
> after creation of a new sandbox.Nor was a satisfactory resolution to the
> problem obtained. A Google search indicates about 2000 references to
> problems of this nature, so it is reasonably common. The CVS documentation
> does not address the issue directly.
>
> We see this problem too, but have some insight into what is causing it.
>
> The problem is created when a commit to a remote project is made
> when the clock on the server and client are hours apart. This can occur
> because the time zone was set incorrectly on the client. This seems to
> create an out-of-sync condition in the repository that cannot be easily
> resolved.
>
> How can this be cleaned up?
>
>                                                            John Nagle
>                                                            Team Overbot