I am evaluating Stardog 2.1.2 using a Java application that is currently running in Stardog 1.2.2 (I've tested both Windows 7 and CentOS with the same behavior). I have been getting random exceptions when committing to the database that indicate the DB is corrupt and needs recovery. So far the only transaction that I can get to fail consistently in the application is one that includes, potentially, several hundred queries/adds/removes before the commit. I haven't quite been able to narrow down exactly what is causing the issue but I do see the same exception in the following very simple case.
Create a new db and remove a non-existent graph:
stardog-admin db create -n testdb -s
stardog data remove http://localhost:5820/testdb -g http://my.example/ns/graph#g1
There was a mixed exception while committing the transaction, some data was not persisted and the database requires recovery.
The message matches the exception message I see in the Java code and we are removing graphs in the transaction.
That same scenario in Stardog 1.2.2 simply says "Removed 0 triples" and carries on.
Is this a regression or the new expected behavior?
Daniel