This introduces the notion of a DistributedQueryHandler that uses a "provider" to read/write requests and results to and from the master. The full flow is exercised via integration tests, and unit tests for each component.
It is intended to foster discussion around this client side interface, as well as provide a base to build from.
I'd like to make sure we use expressions of sizeof to relate buffer
sizes to memcpy and memset. This should make modifying the code less
error prone.
Conflicts:
osquery/tables/system/darwin/nvram.cpp
If osqueryd was killed and another process was started with osqueryd's
old pid before a new osqueryd could start, osqueryd would encounter a
bug where osqueryd would never start.
This executes an osquery query to the processes table to make sure that
the name of the process is "osqueryd". Of course, you could perhaps
denial of service osqueryd this way, but that would require root
filesystem access (assuming that the last version of osqueryd was
ran as root). Thoughts?
