…6327 SERVER-6368 SERVER-4237
to check if its getting too large, the result array must be copied later as a result a few tests needed minor alterations
…normal projections Prep for SERVER-6175
Benchmark shows ~15-20% improvement in aggregation time
Workaround for SERVER-5752. We don't want to restart the whole aggregation on every migrate since it may never finish that way.
Changed the pre-processor guard in the test to use the exact inverse of the condition used in verifyFailed to call abort() when assertion fails.
With multithreaded replication, the call site for retrying if Cloner misses documents was changed. We now much grab a write lock before calling shouldRetry(). I also changed the logging to always log when a retry situation is present. These situations can result if Cloner misses a document because it was updated, grew, and moved backward on disk.
Depending on third_party/mongo_boost instead of third_party/boost/boost_XXX enables the use of the system version of boost libraries.