Kernel Errors 348 and 351

[Erich Geiersberger]

Andrew,

disregard the reproduction scheme in my last mail.
I just found that thread named "V4RB Windows VarChar Error 351", january 2004

At 15:27 Uhr +0100 01.02.2004, Erich Geiersberger wrote:
>We have made Client-Server test during the last weeks with V4MD and 
>we also had problems with this message. While I am not sure if it 
>works in the same way with V4RB, here are some ideas:
>
>Error 351 occured quite often, when a client using read locks wrote 
>to a record and tried to read from the same table immediately 
>afterwards.
>
>This might also be related to the fact that our database gets 
>flushed after each write.
>
>The server debug window showed an error 351 "VarChar Field 
>corrupted" immediately followed by a "error 0, no error". After a 
>first occurance of that error, it happened more often, but went away 
>after a server restart.
>
>I have checked my database a couple of times and found NO errors in 
>it when that happened.
>
>After setting most of our cursors to no read locks, it occured a lot 
>less often.
>
>I am not absoultely sure what condition makes it happen, but I am 
>quite sure that the message is wrong. It seemed to happen when we 
>tried to read from a freshly edited record while it got flushed to 
>disk.

and here is some more:

At 20:42 Uhr +0100 02.02.2004, Erich Geiersberger wrote:
>I remember it worked like that:
>- set kreadWrite = 2
>- connect two clients
>- let the both make the same (order by) selection on a table
>- load records with getcolumn()
>- let one of them write to one of the records with flush()
>- immediately after the write, reload the selection
>
>
>Erich