Clipper locking

[Manish Agarwal]

Elvis,

Are the clients that you are using WinNT ?
It would be easier to find the problem if you could post the network
traces and samba debug traces.

Elvis Pfutzenreuter wrote:
> 
> > Does this work correctly if the server is an NT box?
> 
> Some people reported that it behaves correctly. I didn't try by myself.
> 
> > Clipper goes way back.  It was, if I recall correctly, a compiler for dBase
> > programs.  Now, it's been about 15 years since I last used dBase.  I
> > believe, however, that they used a cooperative locking scheme.  That is, if
> > process 1 set a lock, process 2 had to check the lock before using the data
> > or corruption would result.
> 
> I have made some tests with samba logging level = 3 and strace. The locking
> mechanism is working correctly i.e. when someone has a lock over a record,
> the other client waits until the first releases the lock.
> 
> The problem seems to be related with Windows client caching. Clipper says
> "Corruption detected". It detects corruption by comparing the true record
> contents and the index contents (that are a function of one or more record
> fields). Windows is keeping one of them in cache.
> 
> I have tried 'mount -o mand', SGID-bit on to order mandatory locking, and
> nothing changed. So Windows or Samba is caching something when it shouldn't.

-- 
-Manish