Log Message:
Update databases/apache-cassandra to 1.2.19, the last version released
on the 1.x branch. Set up Makefile.common to be used by newer versions.

1.2.19
 * Don't index tombstones
 * Don't allow compacted sstables to be marked as compacting
 * Track expired tombstones
 * Validate empty cell names from counter updates
 * Improve PasswordAuthenticator default super user setup
 * Remove duplicates from StorageService.getJoiningNodes
 * Clone token map outside of hot gossip loops
 * Add stop method to EmbeddedCassandraService
 * Support connecting to ipv6 jmx with nodetool
 * Set gc_grace_seconds to seven days for system schema tables
 * SimpleSeedProvider no longer caches seeds forever
 * Set correct stream ID on responses when non-Exception Throwables
   are thrown while handling native protocol messages
 * Fix row size miscalculation in LazilyCompactedRow
 * Fix race in background compaction check
 * Don't clear out range tombstones during compaction

1.2.18
 * Support Thrift tables clustering columns on CqlPagingInputFormat
 * Fix compilation with java 6 broke by CASSANDRA-7147

1.2.17
 * cqlsh: Fix CompositeType columns in DESCRIBE TABLE output
 * Expose global ColumnFamily metrics
 * Handle possible integer overflow in FastByteArrayOutputStream
 * cqlsh: 'ascii' values weren't formatted as text
 * cqlsh: ignore .cassandra permission errors
 * reduce failure detector initial value to 2s
 * Fix problem truncating on a node that was previously in a dead state
 * Don't insert tombstones that hide indexed values into 2i
 * Track metrics at a keyspace level
 * Add replace_address_first_boot flag to only replace if not bootstrapped
 * Enable keepalive for native protocol
 * Check internal addresses for seeds
 * Fix potential / by 0 in HHOM page size calculation
 * Fix handling of empty counter replication mutations
 * Use LOCAL_ONE for non-superuser auth queries
 * Set keepalive on MessagingService connections
 * Add Cloudstack snitch
 * Update system.peers correctly when relocating tokens
 * Add Google Compute Engine snitch
 * Fix nodetool display with vnodes
 * Fix schema concurrency exceptions
 * Fix BatchlogManager#deleteBatch() use of millisecond timsestamps
 * Continue assassinating even if the endpoint vanishes
 * Schedule schema pulls on change
 * Non-droppable verbs shouldn't be dropped from OTC
 * Shutdown batchlog executor in SS#drain()
 * Fix batchlog to account for CF truncation records
 * Fix CQLSH parsing of functions and BLOB literals
 * Require nodetool rebuild_index to specify index names
 * Ensure that batchlog and hint timeouts do not produce hints
 * Don't shut MessagingService down when replacing a node
 * Always clean up references in SerializingCache
 * fix npe when doing -Dcassandra.fd_initial_value_ms
 * Preserves CQL metadata when updating table from thrift
 * fix time conversion to milliseconds in SimpleCondition.await
 * remove duplicate query for local tokens
 * raise streaming phi convict threshold level
 * reduce garbage creation in calculatePendingRanges
 * exit CQLSH with error status code if script fails
 * Fix bug with some IN queries missig results
 * Fix availability validation for LOCAL_ONE CL
 * Hint streaming can cause decommission to fail
 * RepairTask didn't send a correct message on IllegalArgumentException
 * Fix race in Thrift server that can exhaust the pool of worker threads