Publish 100,000 messages of 2k to a queue (using store-async=false), then stop the broker try to restart. Get an error midway through: "Db::get: Cannot allocate memory".
Running the bdb db_stat utility on the message db file results in that program hanging.
The same thing happens when running the db_dump utility.
Fixed in r1817 of the bdbstore. Solution was to batch up recovery of messages into multiple transactions of a fxied number of ops each.
qpidc-0.2.656926-1.el5, qpidd-0.2.656926-1.el5, and rhm-0.2.2058-1.el5 have been pushed to the staging repo for testing