I think there is a small window for a race condition:
The MemoryManager is cleaning up and releasing a databuffer that it owns. At the same time the DataBufferManager is allocating a new page. As soon as the okToFree is set the page can be released and returned back to the operating system - which will zero the contents of the memory when it is accessed. In particular the value of owner may now be null.
|How could this regression have been avoided||Resolved|