NPE in ByteBufferVol which appears to corrupt header

[dan-blum]

We got this exception:

java.lang.NullPointerException at org.mapdb.volume.ByteBufferVol.getSlice(ByteBufferVol.java:42) at org.mapdb.volume.ByteBufferVol.getLong(ByteBufferVol.java:121) at org.mapdb.StoreDirect.longStackTake(StoreDirect.kt:389) at org.mapdb.StoreDirectAbstract.allocateData(StoreDirectAbstract.kt:293) at org.mapdb.StoreDirect.put(StoreDirect.kt:618) at org.mapdb.HTreeMap.valueWrap(HTreeMap.kt:1208) at org.mapdb.HTreeMap.putprotected(HTreeMap.kt:344) at org.mapdb.HTreeMap.put(HTreeMap.kt:324)

This is similar to the exception reported in #963 but not identical.

This appears to have corrupted the MapDB header as upon restart we got

java[11865]: at org.mapdb.DBMaker$Maker.make(DBMaker.kt:450) java[11865]: at org.mapdb.StoreDirect$Companion.make$default(StoreDirect.kt:56) java[11865]: at org.mapdb.StoreDirect$Companion.make(StoreDirect.kt:57) java[11865]: at org.mapdb.StoreDirect.(StoreDirect.kt:114) java[11865]: at org.mapdb.StoreDirectAbstract.fileHeaderCheck(StoreDirectAbstract.kt:113) java[11865]: Exception in thread "main" org.mapdb.DBException$DataCorruption: Header checksum broken. Sto

(This is inverted because it's from a systemctl log.)

We are using MapDB 3.0.8 and it is instantiated like so:

        mapDB = DBMaker
            .fileDB(mapDBFile)
            .closeOnJvmShutdown()
            .fileMmapEnable()
            .fileMmapPreclearDisable()
            .make();

Transactions are not enabled because we need to be able to shrink the file. However, we

1. Always call commit() after changing the map (there is only one map).
2. Always change the map inside a Guava Monitor so two threads cannot change it at the same time.