Still too many bugs/instability

[jda]

>In fact this very wonder me Jon.
>And raise question if you have enough db.flush() calls.
>
>I think many developers have test self that right after db.flush()
>A db is on disk in good state.
>
>So if you have really enough flushes() then only possible pattern is:
>
>     you do write of some record, and during this write,
>     crash happens.
>
>     And then we need this crash log.
>

I call db.flush immediately after almost every write (the exception 
being when I am importing lots of records at once, when I call after 
every 50 or so).

The corruption I'm seeing doesn't come after a crash. They precede a 
crash. I personally have found that after I use a database long 
enough Verify starts to show problems, even when there have been no 
crashes. Unfortunately it's rarely possible to reconstruct the exact 
chain of events. But corruption does happen with routine use and no 
crashes.

I'll collect databases...

Jon