Log Message:
The PostgreSQL Global Development Group has released an update to all supported \ 
versions of our database system, including 9.5.4, 9.4.9, 9.3.14, 9.2.18 and \ 
9.1.23. This release fixes two security issues. It also patches a number of \ 
other bugs reported over the last three months. Users who rely on security \ 
isolation between database users should update as soon as possible. Other users \ 
should plan to update at the next convenient downtime.

Security Issues
---------------
Two security holes have been closed by this release:

CVE-2016-5423: certain nested CASE expressions can cause the server to crash.
CVE-2016-5424: database and role names with embedded special characters can \ 
allow code injection during administrative operations like pg_dumpall.
The fix for the second issue also adds an option, -reuse-previous, to psql's \ 
\connect command. pg_dumpall will also refuse to handle database and role names \ 
containing line breaks after the update. For more information on these issues \ 
and how they affect backwards-compatibility, see the Release Notes.

Bug Fixes and Improvements
--------------------------
This update also fixes a number of bugs reported in the last few months. Some of \ 
these issues affect only version 9.5, but many affect all supported versions:

Fix misbehaviors of IS NULL/IS NOT NULL with composite values
Fix three areas where INSERT ... ON CONFLICT failed to work properly with other \ 
SQL features.
Make INET and CIDR data types properly reject bad IPv6 values
Prevent crash in "point ## lseg" operator for NaN input
Avoid possible crash in pg_get_expr()
Fix several one-byte buffer over-reads in to_number()
Don't needlessly plan query if WITH NO DATA is specified
Avoid crash-unsafe state in expensive heap_update() paths
Fix hint bit update during WAL replay of row locking operations
Avoid unnecessary "could not serialize access" with FOR KEY SHARE
Avoid crash in postgres -C when the specified variable is a null string
Fix two issues with logical decoding and subtransactions
Ensure that backends see up-to-date statistics for shared catalogs
Prevent possible failure when vacuuming multixact IDs in an upgraded database
When a manual ANALYZE specifies columns, don't reset changes_since_analyze
Fix ANALYZE's overestimation of n_distinct for columns with nulls
Fix bug in b-tree mark/restore processing
Fix building of large (bigger than shared_buffers) hash indexes
Prevent infinite loop in GiST index build with NaN values
Fix possible crash during a nearest-neighbor indexscan
Fix "PANIC: failed to add BRIN tuple" error
Prevent possible crash during background worker shutdown
Many fixes for issues in parallel pg_dump and pg_restore
Make pg_basebackup accept -Z 0 as no compression
Make regression tests safe for Danish and Welsh locales