Coda File System

Re: 'exec' on 'ci' fails

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Tue, 6 Mar 2001 15:53:24 -0500
On Tue, Mar 06, 2001 at 02:58:28PM -0500, Abhinav Khushraj wrote:
> Hi All
>      Would really appreciate if some one could throw some light on
> this.
>      I am trying to incorporate incremental Coda. I am forking
> a process and calling 'exec' on 'ci', 'rcs' and other RCS commands
> from within fso_cfscalls0.cc and venuscb.cc. The ci is being used for
> checking in files on the local file system and not on /coda

Is this similar to the project work for the Mobile Computing class of
prof. Helal at UFL?

Maybe you should recheck the assignment, because as far as I remember
the "incremental hoarding" only made use of RCS on the servers.

>      The exec does not return as expected. When I call ci  for the
> first time it creates the initial version just fine. However, for
> subsequent ci's it behaves randomly and checks in only once in a
> while. This behaviour is not seen, however when I am using gdb to

I don't believe the client is necessarily reusing the same containerfile
when fetching an updated copy of the file. It might do so in most cases,
but when objects are thrown out of the case their slot is taken by some
completely unrelated fso. You could also be seeing problems when venus
is keeping track of pending operations in the CML because of shadow
copies that are created for not yet reintegrated store operations.

The way to debug this is to log the exact "ci" command that you are
executing everytime it it executed, and maybe dump the state of the FSO
around the same time. However, I'm pretty sure that you are doing
something which isn't asked for in the project description, maybe you
should discuss your chosen approach with one of the TA's.

Jan
Received on 2001-03-06 15:53:27