Andy Seaborne (Commented) (JIRA) wrote:
> Progress.  At least one cause of this (or at least a problem which needed 
> fixing) identified.  Files could be open multiple times and so have kernel or 
> JVM state that was inconsistent - one might be flushed on close resulting in 
> potnetially bad information later.  Highly dependent on the exact order 
> between threads that occurs.

Great news, thanks Andy.

> Running on an SSD makes the test catch things earlier (based on number of 
> transactions needed before problems show up) and make transactional TDB run 
> much faster anyway.

Good to know. :-)

Thanks,
Paolo

Reply via email to