2011-04-18 This release contains a variety of fixes from 9.0.3. For information about new features in the 9.0 major release, see . A dump/restore is not required for those running 9.0.X. However, if your installation was upgraded from a previous major release by running pg_upgrade, you should take action to prevent possible data loss due to a now-fixed bug in pg_upgrade. The recommended solution is to run VACUUM FREEZE on all TOAST tables. More information is available at http://wiki.postgresql.org/wiki/20110408pg_upgrade_fix. Fix pg_upgrade's handling of TOAST tables (Bruce Momjian) The pg_class.relfrozenxid value for TOAST tables was not correctly copied into the new installation during pg_upgrade. This could later result in pg_clog files being discarded while they were still needed to validate tuples in the TOAST tables, leading to could not access status of transaction failures. This error poses a significant risk of data loss for installations that have been upgraded with pg_upgrade. This patch corrects the problem for future uses of pg_upgrade, but does not in itself cure the issue in installations that have been processed with a buggy version of pg_upgrade. Suppress incorrect PD_ALL_VISIBLE flag was incorrectly set warning (Heikki Linnakangas) VACUUM would sometimes issue this warning in cases that are actually valid. Use better SQLSTATE error codes for hot standby conflict cases (Tatsuo Ishii and Simon Riggs) All retryable conflict errors now have an error code that indicates that a retry is possible. Also, session closure due to the database being dropped on the master is now reported as ERRCODE_DATABASE_DROPPED, rather than ERRCODE_ADMIN_SHUTDOWN, so that connection poolers can handle the situation correctly. Prevent intermittent hang in interactions of startup process with bgwriter process (Simon Riggs) This affected recovery in non-hot-standby cases. Disallow including a composite type in itself (Tom Lane) This prevents scenarios wherein the server could recurse infinitely while processing the composite type. While there are some possible uses for such a structure, they don't seem compelling enough to justify the effort required to make sure it always works safely. Avoid potential deadlock during catalog cache initialization (Nikhil Sontakke) In some cases the cache loading code would acquire share lock on a system index before locking the index's catalog. This could deadlock against processes trying to acquire exclusive locks in the other, more standard order. Fix dangling-pointer problem in BEFORE ROW UPDATE trigger handling when there was a concurrent update to the target tuple (Tom Lane) This bug has been observed to result in intermittent cannot extract system attribute from virtual tuple failures while trying to do UPDATE RETURNING ctid. There is a very small probability of more serious errors, such as generating incorrect index entries for the updated tuple. Disallow DROP TABLE when there are pending deferred trigger events for the table (Tom Lane) Formerly the DROP would go through, leading to could not open relation with OID nnn errors when the triggers were eventually fired. Allow replication as a user name in pg_hba.conf (Andrew Dunstan) replication is special in the database name column, but it was mistakenly also treated as special in the user name column. Prevent crash triggered by constant-false WHERE conditions during GEQO optimization (Tom Lane) Improve planner's handling of semi-join and anti-join cases (Tom Lane) Fix handling of SELECT FOR UPDATE in a sub-SELECT (Tom Lane) This bug typically led to cannot extract system attribute from virtual tuple errors. Fix selectivity estimation for text search to account for NULLs (Jesper Krogh) Fix get_actual_variable_range() to support hypothetical indexes injected by an index adviser plugin (Gurjeet Singh) Fix PL/Python memory leak involving array slices (Daniel Popowich) Allow libpq's SSL initialization to succeed when user's home directory is unavailable (Tom Lane) If the SSL mode is such that a root certificate file is not required, there is no need to fail. This change restores the behavior to what it was in pre-9.0 releases. Fix libpq to return a useful error message for errors detected in conninfo_array_parse (Joseph Adams) A typo caused the library to return NULL, rather than the PGconn structure containing the error message, to the application. Fix ecpg preprocessor's handling of float constants (Heikki Linnakangas) Fix parallel pg_restore to handle comments on POST_DATA items correctly (Arnd Hannemann) Fix pg_restore to cope with long lines (over 1KB) in TOC files (Tom Lane) Put in more safeguards against crashing due to division-by-zero with overly enthusiastic compiler optimization (Aurelien Jarno) Support use of dlopen() in FreeBSD and OpenBSD on MIPS (Tom Lane) There was a hard-wired assumption that this system function was not available on MIPS hardware on these systems. Use a compile-time test instead, since more recent versions have it. Fix compilation failures on HP-UX (Heikki Linnakangas) Avoid crash when trying to write to the Windows console very early in process startup (Rushabh Lathia) Support building with MinGW 64 bit compiler for Windows (Andrew Dunstan) Fix version-incompatibility problem with libintl on Windows (Hiroshi Inoue) Fix usage of xcopy in Windows build scripts to work correctly under Windows 7 (Andrew Dunstan) This affects the build scripts only, not installation or usage. Fix path separator used by pg_regress on Cygwin (Andrew Dunstan) Update time zone data files to tzdata release 2011f for DST law changes in Chile, Cuba, Falkland Islands, Morocco, Samoa, and Turkey; also historical corrections for South Australia, Alaska, and Hawaii. 2011-01-31 This release contains a variety of fixes from 9.0.2. For information about new features in the 9.0 major release, see . A dump/restore is not required for those running 9.0.X. Before exiting walreceiver, ensure all the received WAL is fsync'd to disk (Heikki Linnakangas) Otherwise the standby server could replay some un-synced WAL, conceivably leading to data corruption if the system crashes just at that point. Avoid excess fsync activity in walreceiver (Heikki Linnakangas) Make ALTER TABLE revalidate uniqueness and exclusion constraints when needed (Noah Misch) This was broken in 9.0 by a change that was intended to suppress revalidation during VACUUM FULL and CLUSTER, but unintentionally affected ALTER TABLE as well. Fix EvalPlanQual for UPDATE of an inheritance tree in which the tables are not all alike (Tom Lane) Any variation in the table row types (including dropped columns present in only some child tables) would confuse the EvalPlanQual code, leading to misbehavior or even crashes. Since EvalPlanQual is only executed during concurrent updates to the same row, the problem was only seen intermittently. Avoid failures when EXPLAIN tries to display a simple-form CASE expression (Tom Lane) If the CASE's test expression was a constant, the planner could simplify the CASE into a form that confused the expression-display code, resulting in unexpected CASE WHEN clause errors. Fix assignment to an array slice that is before the existing range of subscripts (Tom Lane) If there was a gap between the newly added subscripts and the first pre-existing subscript, the code miscalculated how many entries needed to be copied from the old array's null bitmap, potentially leading to data corruption or crash. Avoid unexpected conversion overflow in planner for very distant date values (Tom Lane) The date type supports a wider range of dates than can be represented by the timestamp types, but the planner assumed it could always convert a date to timestamp with impunity. Fix PL/Python crash when an array contains null entries (Alex Hunsaker) Remove ecpg's fixed length limit for constants defining an array dimension (Michael Meskes) Fix erroneous parsing of tsquery values containing ... & !(subexpression) | ... (Tom Lane) Queries containing this combination of operators were not executed correctly. The same error existed in contrib/intarray's query_int type and contrib/ltree's ltxtquery type. Fix buffer overrun in contrib/intarray's input function for the query_int type (Apple) This bug is a security risk since the function's return address could be overwritten. Thanks to Apple Inc's security team for reporting this issue and supplying the fix. (CVE-2010-4015) Fix bug in contrib/seg's GiST picksplit algorithm (Alexander Korotkov) This could result in considerable inefficiency, though not actually incorrect answers, in a GiST index on a seg column. If you have such an index, consider REINDEXing it after installing this update. (This is identical to the bug that was fixed in contrib/cube in the previous update.) 2010-12-16 This release contains a variety of fixes from 9.0.1. For information about new features in the 9.0 major release, see . A dump/restore is not required for those running 9.0.X. Force the default wal_sync_method to be fdatasync on Linux (Tom Lane, Marti Raudsepp) The default on Linux has actually been fdatasync for many years, but recent kernel changes caused PostgreSQL to choose open_datasync instead. This choice did not result in any performance improvement, and caused outright failures on certain filesystems, notably ext4 with the data=journal mount option. Fix too many KnownAssignedXids error during Hot Standby replay (Heikki Linnakangas) Fix race condition in lock acquisition during Hot Standby (Simon Riggs) Avoid unnecessary conflicts during Hot Standby (Simon Riggs) This fixes some cases where replay was considered to conflict with standby queries (causing delay of replay or possibly cancellation of the queries), but there was no real conflict. Fix assorted bugs in WAL replay logic for GIN indexes (Tom Lane) This could result in bad buffer id: 0 failures or corruption of index contents during replication. Fix recovery from base backup when the starting checkpoint WAL record is not in the same WAL segment as its redo point (Jeff Davis) Fix corner-case bug when streaming replication is enabled immediately after creating the master database cluster (Heikki Linnakangas) Fix persistent slowdown of autovacuum workers when multiple workers remain active for a long time (Tom Lane) The effective vacuum_cost_limit for an autovacuum worker could drop to nearly zero if it processed enough tables, causing it to run extremely slowly. Fix long-term memory leak in autovacuum launcher (Alvaro Herrera) Avoid failure when trying to report an impending transaction wraparound condition from outside a transaction (Tom Lane) This oversight prevented recovery after transaction wraparound got too close, because database startup processing would fail. Add support for detecting register-stack overrun on IA64 (Tom Lane) The IA64 architecture has two hardware stacks. Full prevention of stack-overrun failures requires checking both. Add a check for stack overflow in copyObject() (Tom Lane) Certain code paths could crash due to stack overflow given a sufficiently complex query. Fix detection of page splits in temporary GiST indexes (Heikki Linnakangas) It is possible to have a concurrent page split in a temporary index, if for example there is an open cursor scanning the index when an insertion is done. GiST failed to detect this case and hence could deliver wrong results when execution of the cursor continued. Fix error checking during early connection processing (Tom Lane) The check for too many child processes was skipped in some cases, possibly leading to postmaster crash when attempting to add the new child process to fixed-size arrays. Improve efficiency of window functions (Tom Lane) Certain cases where a large number of tuples needed to be read in advance, but work_mem was large enough to allow them all to be held in memory, were unexpectedly slow. percent_rank(), cume_dist() and ntile() in particular were subject to this problem. Avoid memory leakage while ANALYZE'ing complex index expressions (Tom Lane) Ensure an index that uses a whole-row Var still depends on its table (Tom Lane) An index declared like create index i on t (foo(t.*)) would not automatically get dropped when its table was dropped. Add missing support in DROP OWNED BY for removing foreign data wrapper/server privileges belonging to a user (Heikki Linnakangas) Do not inline a SQL function with multiple OUT parameters (Tom Lane) This avoids a possible crash due to loss of information about the expected result rowtype. Fix crash when inline-ing a set-returning function whose argument list contains a reference to an inline-able user function (Tom Lane) Behave correctly if ORDER BY, LIMIT, FOR UPDATE, or WITH is attached to the VALUES part of INSERT ... VALUES (Tom Lane) Make the OFF keyword unreserved (Heikki Linnakangas) This prevents problems with using off as a variable name in PL/pgSQL. That worked before 9.0, but was now broken because PL/pgSQL now treats all core reserved words as reserved. Fix constant-folding of COALESCE() expressions (Tom Lane) The planner would sometimes attempt to evaluate sub-expressions that in fact could never be reached, possibly leading to unexpected errors. Fix could not find pathkey item to sort planner failure with comparison of whole-row Vars (Tom Lane) Fix postmaster crash when connection acceptance (accept() or one of the calls made immediately after it) fails, and the postmaster was compiled with GSSAPI support (Alexander Chernikov) Retry after receiving an invalid response packet from a RADIUS authentication server (Magnus Hagander) This fixes a low-risk potential denial of service condition. Fix missed unlink of temporary files when log_temp_files is active (Tom Lane) If an error occurred while attempting to emit the log message, the unlink was not done, resulting in accumulation of temp files. Add print functionality for InhRelation nodes (Tom Lane) This avoids a failure when debug_print_parse is enabled and certain types of query are executed. Fix incorrect calculation of distance from a point to a horizontal line segment (Tom Lane) This bug affected several different geometric distance-measurement operators. Fix incorrect calculation of transaction status in ecpg (Itagaki Takahiro) Fix errors in psql's Unicode-escape support (Tom Lane) Speed up parallel pg_restore when the archive contains many large objects (blobs) (Tom Lane) Fix PL/pgSQL's handling of simple expressions to not fail in recursion or error-recovery cases (Tom Lane) Fix PL/pgSQL's error reporting for no-such-column cases (Tom Lane) As of 9.0, it would sometimes report missing FROM-clause entry for table foo when record foo has no field bar would be more appropriate. Fix PL/Python to honor typmod (i.e., length or precision restrictions) when assigning to tuple fields (Tom Lane) This fixes a regression from 8.4. Fix PL/Python's handling of set-returning functions (Jan Urbanski) Attempts to call SPI functions within the iterator generating a set result would fail. Fix bug in contrib/cube's GiST picksplit algorithm (Alexander Korotkov) This could result in considerable inefficiency, though not actually incorrect answers, in a GiST index on a cube column. If you have such an index, consider REINDEXing it after installing this update. Don't emit identifier will be truncated notices in contrib/dblink except when creating new connections (Itagaki Takahiro) Fix potential coredump on missing public key in contrib/pgcrypto (Marti Raudsepp) Fix buffer overrun in contrib/pg_upgrade (Hernan Gonzalez) Fix memory leak in contrib/xml2's XPath query functions (Tom Lane) Update time zone data files to tzdata release 2010o for DST law changes in Fiji and Samoa; also historical corrections for Hong Kong. 2010-10-04 This release contains a variety of fixes from 9.0.0. For information about new features in the 9.0 major release, see . A dump/restore is not required for those running 9.0.X. Use a separate interpreter for each calling SQL userid in PL/Perl and PL/Tcl (Tom Lane) This change prevents security problems that can be caused by subverting Perl or Tcl code that will be executed later in the same session under another SQL user identity (for example, within a SECURITY DEFINER function). Most scripting languages offer numerous ways that that might be done, such as redefining standard functions or operators called by the target function. Without this change, any SQL user with Perl or Tcl language usage rights can do essentially anything with the SQL privileges of the target function's owner. The cost of this change is that intentional communication among Perl and Tcl functions becomes more difficult. To provide an escape hatch, PL/PerlU and PL/TclU functions continue to use only one interpreter per session. This is not considered a security issue since all such functions execute at the trust level of a database superuser already. It is likely that third-party procedural languages that claim to offer trusted execution have similar security issues. We advise contacting the authors of any PL you are depending on for security-critical purposes. Our thanks to Tim Bunce for pointing out this issue (CVE-2010-3433). Improve pg_get_expr() security fix so that the function can still be used on the output of a sub-select (Tom Lane) Fix incorrect placement of placeholder evaluation (Tom Lane) This bug could result in query outputs being non-null when they should be null, in cases where the inner side of an outer join is a sub-select with non-strict expressions in its output list. Fix join removal's handling of placeholder expressions (Tom Lane) Fix possible duplicate scans of UNION ALL member relations (Tom Lane) Prevent infinite loop in ProcessIncomingNotify() after unlistening (Jeff Davis) Prevent show_session_authorization() from crashing within autovacuum processes (Tom Lane) Re-allow input of Julian dates prior to 0001-01-01 AD (Tom Lane) Input such as 'J100000'::date worked before 8.4, but was unintentionally broken by added error-checking. Make psql recognize DISCARD ALL as a command that should not be encased in a transaction block in autocommit-off mode (Itagaki Takahiro) Update build infrastructure and documentation to reflect the source code repository's move from CVS to Git (Magnus Hagander and others) 2010-09-20 This release of PostgreSQL adds features that have been requested for years, such as easy-to-use replication, a mass permission-changing facility, and anonymous code blocks. While past major releases have been conservative in their scope, this release shows a bold new desire to provide facilities that new and existing users of PostgreSQL will embrace. This has all been done with few incompatibilities. Major enhancements include: Built-in replication based on log shipping. This advance consists of two features: Streaming Replication, allowing continuous archive (WAL) files to be streamed over a network connection to a standby server, and Hot Standby, allowing continuous archive standby servers to execute read-only queries. The net effect is to support a single master with multiple read-only slave servers. Easier database object permissions management. GRANT/REVOKE IN SCHEMA supports mass permissions changes on existing objects, while ALTER DEFAULT PRIVILEGES allows control of privileges for objects created in the future. Large objects (BLOBs) now support permissions management as well. Broadly enhanced stored procedure support. The DO statement supports ad-hoc or anonymous code blocks. Functions can now be called using named parameters. PL/pgSQL is now installed by default, and PL/Perl and PL/Python have been enhanced in several ways, including support for Python3. Full support for 64-bit Windows. More advanced reporting queries, including additional windowing options (PRECEDING and FOLLOWING) and the ability to control the order in which values are fed to aggregate functions. New trigger features, including SQL-standard-compliant per-column triggers and conditional trigger execution. Deferrable unique constraints. Mass updates to unique keys are now possible without trickery. Exclusion constraints. These provide a generalized version of unique constraints, allowing enforcement of complex conditions. New and enhanced security features, including RADIUS authentication, LDAP authentication improvements, and a new contrib module passwordcheck for testing password strength. New high-performance implementation of the LISTEN/NOTIFY feature. Pending events are now stored in a memory-based queue rather than a table. Also, a payload string can be sent with each event, rather than transmitting just an event name as before. New implementation of VACUUM FULL. This command now rewrites the entire table and indexes, rather than moving individual rows to compact space. It is substantially faster in most cases, and no longer results in index bloat. New contrib module pg_upgrade to support in-place upgrades from 8.3 or 8.4 to 9.0. Multiple performance enhancements for specific types of queries, including elimination of unnecessary joins. This helps optimize some automatically-generated queries, such as those produced by object-relational mappers (ORMs). EXPLAIN enhancements. The output is now available in JSON, XML, or YAML format, and includes buffer utilization and other data not previously available. hstore improvements, including new functions and greater data capacity. The above items are explained in more detail in the sections below. A dump/restore using pg_dump, or use of pg_upgrade, is required for those wishing to migrate data from any previous release. Version 9.0 contains a number of changes that selectively break backwards compatibility in order to support new features and code quality improvements. In particular, users who make extensive use of PL/pgSQL, Point-In-Time Recovery (PITR), or Warm Standby should test their applications because of slight user-visible changes in those areas. Observe the following incompatibilities: Remove server parameter add_missing_from, which was defaulted to off for many years (Tom Lane) Remove server parameter regex_flavor, which was defaulted to advanced for many years (Tom Lane) archive_mode now only affects archive_command; a new setting, wal_level, affects the contents of the write-ahead log (Heikki Linnakangas) log_temp_files now uses default file size units of kilobytes (Robert Haas) When querying a parent table, do not do any separate permission checks on child tables scanned as part of the query (Peter Eisentraut) The SQL standard specifies this behavior, and it is also much more convenient in practice than the former behavior of checking permissions on each child as well as the parent. bytea output now appears in hex format by default (Peter Eisentraut) The server parameter bytea_output can be used to select the traditional output format if needed for compatibility. Array input now considers only plain ASCII whitespace characters to be potentially ignorable; it will never ignore non-ASCII characters, even if they are whitespace according to some locales (Tom Lane) This avoids some corner cases where array values could be interpreted differently depending on the server's locale settings. Improve standards compliance of SIMILAR TO patterns and SQL-style substring() patterns (Tom Lane) This includes treating ? and {...} as pattern metacharacters, while they were simple literal characters before; that corresponds to new features added in SQL:2008. Also, ^ and $ are now treated as simple literal characters; formerly they were treated as metacharacters, as if the pattern were following POSIX rather than SQL rules. Also, in SQL-standard substring(), use of parentheses for nesting no longer interferes with capturing of a substring. Also, processing of bracket expressions (character classes) is now more standards-compliant. Reject negative length values in 3-parameter substring() for bit strings, per the SQL standard (Tom Lane) Make date_trunc truncate rather than round when reducing precision of fractional seconds (Tom Lane) The code always acted this way for integer-based dates/times. Now float-based dates/times behave similarly. Tighten enforcement of column name consistency during RENAME when a child table inherits the same column from multiple unrelated parents (KaiGai Kohei) No longer automatically rename indexes and index columns when the underlying table columns are renamed (Tom Lane) Administrators can still rename such indexes and columns manually. This change will require an update of the JDBC driver, and possibly other drivers, so that unique indexes are correctly recognized after a rename. CREATE OR REPLACE FUNCTION can no longer change the declared names of function parameters (Pavel Stehule) In order to avoid creating ambiguity in named-parameter calls, it is no longer allowed to change the aliases for input parameters in the declaration of an existing function (although names can still be assigned to previously unnamed parameters). You now have to DROP and recreate the function to do that. PL/pgSQL now throws an error if a variable name conflicts with a column name used in a query (Tom Lane) The former behavior was to bind ambiguous names to PL/pgSQL variables in preference to query columns, which often resulted in surprising misbehavior. Throwing an error allows easy detection of ambiguous situations. Although it's recommended that functions encountering this type of error be modified to remove the conflict, the old behavior can be restored if necessary via the configuration parameter plpgsql.variable_conflict, or via the per-function option #variable_conflict. PL/pgSQL no longer allows variable names that match certain SQL reserved words (Tom Lane) This is a consequence of aligning the PL/pgSQL parser to match the core SQL parser more closely. If necessary, variable names can be double-quoted to avoid this restriction. PL/pgSQL now requires columns of composite results to match the expected type modifier as well as base type (Pavel Stehule, Tom Lane) For example, if a column of the result type is declared as NUMERIC(30,2), it is no longer acceptable to return a NUMERIC of some other precision in that column. Previous versions neglected to check the type modifier and would thus allow result rows that didn't actually conform to the declared restrictions. PL/pgSQL now treats selection into composite fields more consistently (Tom Lane) Formerly, a statement like SELECT ... INTO rec.fld FROM ... was treated as a scalar assignment even if the record field fld was of composite type. Now it is treated as a record assignment, the same as when the INTO target is a regular variable of composite type. So the values to be assigned to the field's subfields should be written as separate columns of the SELECT list, not as a ROW(...) construct as in previous versions. If you need to do this in a way that will work in both 9.0 and previous releases, you can write something like rec.fld := ROW(...) FROM .... Remove PL/pgSQL's RENAME declaration (Tom Lane) Instead of RENAME, use ALIAS, which can now create an alias for any variable, not only dollar sign parameter names (such as $1) as before. Deprecate use of => as an operator name (Robert Haas) Future versions of PostgreSQL will probably reject this operator name entirely, in order to support the SQL-standard notation for named function parameters. For the moment, it is still allowed, but a warning is emitted when such an operator is defined. Remove support for platforms that don't have a working 64-bit integer data type (Tom Lane) It is believed all still-supported platforms have working 64-bit integer data types. Version 9.0 has an unprecedented number of new major features, and over 200 enhancements, improvements, new commands, new functions, and other changes. PostgreSQL's existing standby-server capability has been expanded both to support read-only queries on standby servers and to greatly reduce the lag between master and standby servers. For many users, this will be a useful and low-administration form of replication, either for high availability or for horizontal scalability. Allow a standby server to accept read-only queries (Simon Riggs, Heikki Linnakangas) This feature is called Hot Standby. There are new postgresql.conf and recovery.conf settings to control this feature, as well as extensive documentation. Allow write-ahead log (WAL) data to be streamed to a standby server (Fujii Masao, Heikki Linnakangas) This feature is called Streaming Replication. Previously WAL data could be sent to standby servers only in units of entire WAL files (normally 16 megabytes each). Streaming Replication eliminates this inefficiency and allows updates on the master to be propagated to standby servers with very little delay. There are new postgresql.conf and recovery.conf settings to control this feature, as well as extensive documentation. Add pg_last_xlog_receive_location() and pg_last_xlog_replay_location(), which can be used to monitor standby server WAL activity (Simon Riggs, Fujii Masao, Heikki Linnakangas) Allow per-tablespace values to be set for sequential and random page cost estimates (seq_page_cost/random_page_cost) via ALTER TABLESPACE ... SET/RESET (Robert Haas) Improve performance and reliability of EvalPlanQual rechecks in join queries (Tom Lane) UPDATE, DELETE, and SELECT FOR UPDATE/SHARE queries that involve joins will now behave much better when encountering freshly-updated rows. Improve performance of TRUNCATE when the table was created or truncated earlier in the same transaction (Tom Lane) Improve performance of finding inheritance child tables (Tom Lane) Remove unnecessary outer joins (Robert Haas) Outer joins where the inner side is unique and not referenced above the join are unnecessary and are therefore now removed. This will accelerate many automatically generated queries, such as those created by object-relational mappers (ORMs). Allow IS NOT NULL restrictions to use indexes (Tom Lane) This is particularly useful for finding MAX()/MIN() values in indexes that contain many null values. Improve the optimizer's choices about when to use materialize nodes, and when to use sorting versus hashing for DISTINCT (Tom Lane) Improve the optimizer's equivalence detection for expressions involving boolean <> operators (Tom Lane) Use the same random seed every time GEQO plans a query (Andres Freund) While the Genetic Query Optimizer (GEQO) still selects random plans, it now always selects the same random plans for identical queries, thus giving more consistent performance. You can modify geqo_seed to experiment with alternative plans. Improve GEQO plan selection (Tom Lane) This avoids the rare error failed to make a valid plan, and should also improve planning speed. Improve ANALYZE to support inheritance-tree statistics (Tom Lane) This is particularly useful for partitioned tables. However, autovacuum does not yet automatically re-analyze parent tables when child tables change. Improve autovacuum's detection of when re-analyze is necessary (Tom Lane) Improve optimizer's estimation for greater/less-than comparisons (Tom Lane) When looking up statistics for greater/less-than comparisons, if the comparison value is in the first or last histogram bucket, use an index (if available) to fetch the current actual column minimum or maximum. This greatly improves the accuracy of estimates for comparison values near the ends of the data range, particularly if the range is constantly changing due to addition of new data. Allow setting of number-of-distinct-values statistics using ALTER TABLE (Robert Haas) This allows users to override the estimated number or percentage of distinct values for a column. This statistic is normally computed by ANALYZE, but the estimate can be poor, especially on tables with very large numbers of rows. Add support for RADIUS (Remote Authentication Dial In User Service) authentication (Magnus Hagander) Allow LDAP (Lightweight Directory Access Protocol) authentication to operate in search/bind mode (Robert Fleming, Magnus Hagander) This allows the user to be looked up first, then the system uses the DN (Distinguished Name) returned for that user. Add samehost and samenet designations to pg_hba.conf (Stef Walter) These match the server's IP address and subnet address respectively. Pass trusted SSL root certificate names to the client so the client can return an appropriate client certificate (Craig Ringer) Add the ability for clients to set an application name, which is displayed in pg_stat_activity (Dave Page) This allows administrators to characterize database traffic and troubleshoot problems by source application. Add a SQLSTATE option (%e) to log_line_prefix (Guillaume Smet) This allows users to compile statistics on errors and messages by error code number. Write to the Windows event log in UTF16 encoding (Itagaki Takahiro) Now there is true multilingual support for PostgreSQL log messages on Windows. Add pg_stat_reset_shared('bgwriter') to reset the cluster-wide shared statistics for the background writer (Greg Smith) Add pg_stat_reset_single_table_counters() and pg_stat_reset_single_function_counters() to allow resetting the statistics counters for individual tables and functions (Magnus Hagander) Allow setting of configuration parameters based on database/role combinations (Alvaro Herrera) Previously only per-database and per-role settings were possible, not combinations. All role and database settings are now stored in the new pg_db_role_setting system catalog. A new psql command \drds shows these settings. The legacy system views pg_roles, pg_shadow, and pg_user do not show combination settings, and therefore no longer completely represent the configuration for a user or database. Add server parameter bonjour, which controls whether a Bonjour-enabled server advertises itself via Bonjour (Tom Lane) The default is off, meaning it does not advertise. This allows packagers to distribute Bonjour-enabled builds without worrying that individual users might not want the feature. Add server parameter enable_material, which controls the use of materialize nodes in the optimizer (Robert Haas) The default is on. When off, the optimizer will not add materialize nodes purely for performance reasons, though they will still be used when necessary for correctness. Change server parameter log_temp_files to use default file size units of kilobytes (Robert Haas) Previously this setting was interpreted in bytes if no units were specified. Log changes of parameter values when postgresql.conf is reloaded (Peter Eisentraut) This lets administrators and security staff audit changes of database settings, and is also very convenient for checking the effects of postgresql.conf edits. Properly enforce superuser permissions for custom server parameters (Tom Lane) Non-superusers can no longer issue ALTER ROLE/DATABASE SET for parameters that are not currently known to the server. This allows the server to correctly check that superuser-only parameters are only set by superusers. Previously, the SET would be allowed and then ignored at session start, making superuser-only custom parameters much less useful than they should be. Perform SELECT FOR UPDATE/SHARE processing after applying LIMIT, so the number of rows returned is always predictable (Tom Lane) Previously, changes made by concurrent transactions could cause a SELECT FOR UPDATE to unexpectedly return fewer rows than specified by its LIMIT. FOR UPDATE in combination with ORDER BY can still produce surprising results, but that can be corrected by placing FOR UPDATE in a subquery. Allow mixing of traditional and SQL-standard LIMIT/OFFSET syntax (Tom Lane) Extend the supported frame options in window functions (Hitoshi Harada) Frames can now start with CURRENT ROW, and the ROWS n PRECEDING/FOLLOWING options are now supported. Make SELECT INTO and CREATE TABLE AS return row counts to the client in their command tags (Boszormenyi Zoltan) This can save an entire round-trip to the client, allowing result counts and pagination to be calculated without an additional COUNT query. Support Unicode surrogate pairs (dual 16-bit representation) in U& strings and identifiers (Peter Eisentraut) Support Unicode escapes in E'...' strings (Marko Kreen) Speed up CREATE DATABASE by deferring flushes to disk (Andres Freund, Greg Stark) Allow comments on columns of tables, views, and composite types only, not other relation types such as indexes and TOAST tables (Tom Lane) Allow the creation of enumerated types containing no values (Bruce Momjian) Let values of columns having storage type MAIN remain on the main heap page unless the row cannot fit on a page (Kevin Grittner) Previously MAIN values were forced out to TOAST tables until the row size was less than one-quarter of the page size. Implement IF EXISTS for ALTER TABLE DROP COLUMN and ALTER TABLE DROP CONSTRAINT (Andres Freund) Allow ALTER TABLE commands that rewrite tables to skip WAL logging (Itagaki Takahiro) Such operations either produce a new copy of the table or are rolled back, so WAL archiving can be skipped, unless running in continuous archiving mode. This reduces I/O overhead and improves performance. Fix failure of ALTER TABLE table ADD COLUMN col serial when done by non-owner of table (Tom Lane) Add support for copying COMMENTS and STORAGE settings in CREATE TABLE ... LIKE commands (Itagaki Takahiro) Add a shortcut for copying all properties in CREATE TABLE ... LIKE commands (Itagaki Takahiro) Add the SQL-standard CREATE TABLE ... OF type command (Peter Eisentraut) This allows creation of a table that matches an existing composite type. Additional constraints and defaults can be specified in the command. Add deferrable unique constraints (Dean Rasheed) This allows mass updates, such as UPDATE tab SET col = col + 1, to work reliably on columns that have unique indexes or are marked as primary keys. If the constraint is specified as DEFERRABLE it will be checked at the end of the statement, rather than after each row is updated. The constraint check can also be deferred until the end of the current transaction, allowing such updates to be spread over multiple SQL commands. Add exclusion constraints (Jeff Davis) Exclusion constraints generalize uniqueness constraints by allowing arbitrary comparison operators, not just equality. They are created with the CREATE TABLE CONSTRAINT ... EXCLUDE clause. The most common use of exclusion constraints is to specify that column entries must not overlap, rather than simply not be equal. This is useful for time periods and other ranges, as well as arrays. This feature enhances checking of data integrity for many calendaring, time-management, and scientific applications. Improve uniqueness-constraint violation error messages to report the values causing the failure (Itagaki Takahiro) For example, a uniqueness constraint violation might now report Key (x)=(2) already exists. Add the ability to make mass permission changes across a whole schema using the new GRANT/REVOKE IN SCHEMA clause (Petr Jelinek) This simplifies management of object permissions and makes it easier to utilize database roles for application data security. Add ALTER DEFAULT PRIVILEGES command to control privileges of objects created later (Petr Jelinek) This greatly simplifies the assignment of object privileges in a complex database application. Default privileges can be set for tables, views, sequences, and functions. Defaults may be assigned on a per-schema basis, or database-wide. Add the ability to control large object (BLOB) permissions with GRANT/REVOKE (KaiGai Kohei) Formerly, any database user could read or modify any large object. Read and write permissions can now be granted and revoked per large object, and the ownership of large objects is tracked. Make LISTEN/NOTIFY store pending events in a memory queue, rather than in a system table (Joachim Wieland) This substantially improves performance, while retaining the existing features of transactional support and guaranteed delivery. Allow NOTIFY to pass an optional payload string to listeners (Joachim Wieland) This greatly improves the usefulness of LISTEN/NOTIFY as a general-purpose event queue system. Allow CLUSTER on all per-database system catalogs (Tom Lane) Shared catalogs still cannot be clustered. Accept COPY ... CSV FORCE QUOTE * (Itagaki Takahiro) Now * can be used as shorthand for all columns in the FORCE QUOTE clause. Add new COPY syntax that allows options to be specified inside parentheses (Robert Haas, Emmanuel Cecchet) This allows greater flexibility for future COPY options. The old syntax is still supported, but only for pre-existing options. Allow EXPLAIN to output in XML, JSON, or YAML format (Robert Haas, Greg Sabino Mullane) The new output formats are easily machine-readable, supporting the development of new tools for analysis of EXPLAIN output. Add new BUFFERS option to report query buffer usage during EXPLAIN ANALYZE (Itagaki Takahiro) This allows better query profiling for individual queries. Buffer usage is no longer reported in the output for log_statement_stats and related settings. Add hash usage information to EXPLAIN output (Robert Haas) Add new EXPLAIN syntax that allows options to be specified inside parentheses (Robert Haas) This allows greater flexibility for future EXPLAIN options. The old syntax is still supported, but only for pre-existing options. Change VACUUM FULL to rewrite the entire table and rebuild its indexes, rather than moving individual rows around to compact space (Itagaki Takahiro, Tom Lane) The previous method was usually slower and caused index bloat. Note that the new method will use more disk space transiently during VACUUM FULL; potentially as much as twice the space normally occupied by the table and its indexes. Add new VACUUM syntax that allows options to be specified inside parentheses (Itagaki Takahiro) This allows greater flexibility for future VACUUM options. The old syntax is still supported, but only for pre-existing options. Allow an index to be named automatically by omitting the index name in CREATE INDEX (Tom Lane) By default, multicolumn indexes are now named after all their columns; and index expression columns are now named based on their expressions (Tom Lane) Reindexing shared system catalogs is now fully transactional and crash-safe (Tom Lane) Formerly, reindexing a shared index was only allowed in standalone mode, and a crash during the operation could leave the index in worse condition than it was before. Add point_ops operator class for GiST (Teodor Sigaev) This feature permits GiST indexing of point columns. The index can be used for several types of queries such as point <@ polygon (point is in polygon). This should make many PostGIS queries faster. Use red-black binary trees for GIN index creation (Teodor Sigaev) Red-black trees are self-balancing. This avoids slowdowns in cases where the input is in nonrandom order. Allow bytea values to be written in hex notation (Peter Eisentraut) The server parameter bytea_output controls whether hex or traditional format is used for bytea output. Libpq's PQescapeByteaConn() function automatically uses the hex format when connected to PostgreSQL 9.0 or newer servers. However, pre-9.0 libpq versions will not correctly process hex format from newer servers. The new hex format will be directly compatible with more applications that use binary data, allowing them to store and retrieve it without extra conversion. It is also significantly faster to read and write than the traditional format. Allow server parameter extra_float_digits to be increased to 3 (Tom Lane) The previous maximum extra_float_digits setting was 2. There are cases where 3 digits are needed to dump and restore float4 values exactly. pg_dump will now use the setting of 3 when dumping from a server that allows it. Tighten input checking for int2vector values (Caleb Welton) Add prefix support in synonym dictionaries (Teodor Sigaev) Add filtering dictionaries (Teodor Sigaev) Filtering dictionaries allow tokens to be modified then passed to subsequent dictionaries. Allow underscores in email-address tokens (Teodor Sigaev) Use more standards-compliant rules for parsing URL tokens (Tom Lane) Allow function calls to supply parameter names and match them to named parameters in the function definition (Pavel Stehule) For example, if a function is defined to take parameters a and b, it can be called with func(a := 7, b := 12) or func(b := 12, a := 7). Support locale-specific regular expression processing with UTF-8 server encoding (Tom Lane) Locale-specific regular expression functionality includes case-insensitive matching and locale-specific character classes. Previously, these features worked correctly for non-ASCII characters only if the database used a single-byte server encoding (such as LATIN1). They will still misbehave in multi-byte encodings other than UTF-8. Add support for scientific notation in to_char() (EEEE specification) (Pavel Stehule, Brendan Jurd) Make to_char() honor FM (fill mode) in Y, YY, and YYY specifications (Bruce Momjian, Tom Lane) It was already honored by YYYY. Fix to_char() to output localized numeric and monetary strings in the correct encoding on Windows (Hiroshi Inoue, Itagaki Takahiro, Bruce Momjian) Correct calculations of overlaps and contains operations for polygons (Teodor Sigaev) The polygon && (overlaps) operator formerly just checked to see if the two polygons' bounding boxes overlapped. It now does a more correct check. The polygon @> and <@ (contains/contained by) operators formerly checked to see if one polygon's vertexes were all contained in the other; this can wrongly report true for some non-convex polygons. Now they check that all line segments of one polygon are contained in the other. Allow aggregate functions to use ORDER BY (Andrew Gierth) For example, this is now supported: array_agg(a ORDER BY b). This is useful with aggregates for which the order of input values is significant, and eliminates the need to use a nonstandard subquery to determine the ordering. Multi-argument aggregate functions can now use DISTINCT (Andrew Gierth) Add the string_agg() aggregate function to combine values into a single string (Pavel Stehule) Aggregate functions that are called with DISTINCT are now passed NULL values if the aggregate transition function is not marked as STRICT (Andrew Gierth) For example, agg(DISTINCT x) might pass a NULL x value to agg(). This is more consistent with the behavior in non-DISTINCT cases. Add get_bit() and set_bit() functions for bit strings, mirroring those for bytea (Leonardo F) Implement OVERLAY() (replace) for bit strings and bytea (Leonardo F) Add pg_table_size() and pg_indexes_size() to provide a more user-friendly interface to the pg_relation_size() function (Bernd Helmle) Add has_sequence_privilege() for sequence permission checking (Abhijit Menon-Sen) Update the information_schema views to conform to SQL:2008 (Peter Eisentraut) Make the information_schema views correctly display maximum octet lengths for char and varchar columns (Peter Eisentraut) Speed up information_schema privilege views (Joachim Wieland) Support execution of anonymous code blocks using the DO statement (Petr Jelinek, Joshua Tolley, Hannu Valtonen) This allows execution of server-side code without the need to create and delete a temporary function definition. Code can be executed in any language for which the user has permissions to define a function. Implement SQL-standard-compliant per-column triggers (Itagaki Takahiro) Such triggers are fired only when the specified column(s) are affected by the query, e.g. appear in an UPDATE's SET list. Add the WHEN clause to CREATE TRIGGER to allow control over whether a trigger is fired (Itagaki Takahiro) While the same type of check can always be performed inside the trigger, doing it in an external WHEN clause can have performance benefits. Add the OR REPLACE clause to CREATE LANGUAGE (Tom Lane) This is helpful to optionally install a language if it does not already exist, and is particularly helpful now that PL/pgSQL is installed by default. Install PL/pgSQL by default (Bruce Momjian) The language can still be removed from a particular database if the administrator has security or performance concerns about making it available. Improve handling of cases where PL/pgSQL variable names conflict with identifiers used in queries within a function (Tom Lane) The default behavior is now to throw an error when there is a conflict, so as to avoid surprising behaviors. This can be modified, via the configuration parameter plpgsql.variable_conflict or the per-function option #variable_conflict, to allow either the variable or the query-supplied column to be used. In any case PL/pgSQL will no longer attempt to substitute variables in places where they would not be syntactically valid. Make PL/pgSQL use the main lexer, rather than its own version (Tom Lane) This ensures accurate tracking of the main system's behavior for details such as string escaping. Some user-visible details, such as the set of keywords considered reserved in PL/pgSQL, have changed in consequence. Avoid throwing an unnecessary error for an invalid record reference (Tom Lane) An error is now thrown only if the reference is actually fetched, rather than whenever the enclosing expression is reached. For example, many people have tried to do this in triggers: if TG_OP = 'INSERT' and NEW.col1 = ... then This will now actually work as expected. Improve PL/pgSQL's ability to handle row types with dropped columns (Pavel Stehule) Allow input parameters to be assigned values within PL/pgSQL functions (Steve Prentice) Formerly, input parameters were treated as being declared CONST, so the function's code could not change their values. This restriction has been removed to simplify porting of functions from other DBMSes that do not impose the equivalent restriction. An input parameter now acts like a local variable initialized to the passed-in value. Improve error location reporting in PL/pgSQL (Tom Lane) Add count and ALL options to MOVE FORWARD/BACKWARD in PL/pgSQL (Pavel Stehule) Allow PL/pgSQL's WHERE CURRENT OF to use a cursor variable (Tom Lane) Allow PL/pgSQL's OPEN cursor FOR EXECUTE to use parameters (Pavel Stehule, Itagaki Takahiro) This is accomplished with a new USING clause. Add new PL/Perl functions: quote_literal(), quote_nullable(), quote_ident(), encode_bytea(), decode_bytea(), looks_like_number(), encode_array_literal(), encode_array_constructor() (Tim Bunce) Add server parameter plperl.on_init to specify a PL/Perl initialization function (Tim Bunce) plperl.on_plperl_init and plperl.on_plperlu_init are also available for initialization that is specific to the trusted or untrusted language respectively. Support END blocks in PL/Perl (Tim Bunce) END blocks do not currently allow database access. Allow use strict in PL/Perl (Tim Bunce) Perl strict checks can also be globally enabled with the new server parameter plperl.use_strict. Allow require in PL/Perl (Tim Bunce) This basically tests to see if the module is loaded, and if not, generates an error. It will not allow loading of modules that the administrator has not preloaded via the initialization parameters. Allow use feature in PL/Perl if Perl version 5.10 or later is used (Tim Bunce) Verify that PL/Perl return values are valid in the server encoding (Andrew Dunstan) Add Unicode support in PL/Python (Peter Eisentraut) Strings are automatically converted from/to the server encoding as necessary. Improve bytea support in PL/Python (Caleb Welton) Bytea values passed into PL/Python are now represented as binary, rather than the PostgreSQL bytea text format. Bytea values containing null bytes are now also output properly from PL/Python. Passing of boolean, integer, and float values was also improved. Support arrays as parameters and return values in PL/Python (Peter Eisentraut) Improve mapping of SQL domains to Python types (Peter Eisentraut) Add Python 3 support to PL/Python (Peter Eisentraut) The new server-side language is called plpython3u. This cannot be used in the same session with the Python 2 server-side language. Improve error location and exception reporting in PL/Python (Peter Eisentraut) Add an --analyze-only option to vacuumdb, to analyze without vacuuming (Bruce Momjian) Add support for quoting/escaping the values of psql variables as SQL strings or identifiers (Pavel Stehule, Robert Haas) For example, :'var' will produce the value of var quoted and properly escaped as a literal string, while :"var" will produce its value quoted and escaped as an identifier. Ignore a leading UTF-8-encoded Unicode byte-order marker in script files read by psql (Itagaki Takahiro) This is enabled when the client encoding is UTF-8. It improves compatibility with certain editors, mostly on Windows, that insist on inserting such markers. Fix psql --file - to properly honor --single-transaction (Bruce Momjian) Avoid overwriting of psql's command-line history when two psql sessions are run concurrently (Tom Lane) Improve psql's tab completion support (Itagaki Takahiro) Show \timing output when it is enabled, regardless of quiet mode (Peter Eisentraut) Improve display of wrapped columns in psql (Roger Leigh) This behavior is now the default. The previous formatting is available by using \pset linestyle old-ascii. Allow psql to use fancy Unicode line-drawing characters via \pset linestyle unicode (Roger Leigh) Make \d show child tables that inherit from the specified parent (Damien Clochard) \d shows only the number of child tables, while \d+ shows the names of all child tables. Show definitions of index columns in \d index_name (Khee Chin) The definition is useful for expression indexes. Show a view's defining query only in \d+, not in \d (Peter Eisentraut) Always including the query was deemed overly verbose. Make pg_dump/pg_restore --clean also remove large objects (Itagaki Takahiro) Fix pg_dump to properly dump large objects when standard_conforming_strings is enabled (Tom Lane) The previous coding could fail when dumping to an archive file and then generating script output from pg_restore. pg_restore now emits large-object data in hex format when generating script output (Tom Lane) This could cause compatibility problems if the script is then loaded into a pre-9.0 server. To work around that, restore directly to the server, instead. Allow pg_dump to dump comments attached to columns of composite types (Taro Minowa (Higepon)) Make pg_dump --verbose output the pg_dump and server versions in text output mode (Jim Cox, Tom Lane) These were already provided in custom output mode. pg_restore now complains if any command-line arguments remain after the switches and optional file name (Tom Lane) Previously, it silently ignored any such arguments. Allow pg_ctl to be used safely to start the postmaster during a system reboot (Tom Lane) Previously, pg_ctl's parent process could have been mistakenly identified as a running postmaster based on a stale postmaster lock file, resulting in a transient failure to start the database. Give pg_ctl the ability to initialize the database (by invoking initdb) (Zdenek Kotala) Add new libpq functions PQconnectdbParams() and PQconnectStartParams() (Guillaume Lelarge) These functions are similar to PQconnectdb() and PQconnectStart() except that they accept a null-terminated array of connection options, rather than requiring all options to be provided in a single string. Add libpq functions PQescapeLiteral() and PQescapeIdentifier() (Robert Haas) These functions return appropriately quoted and escaped SQL string literals and identifiers. The caller is not required to pre-allocate the string result, as is required by PQescapeStringConn(). Add support for a per-user service file (.pg_service.conf), which is checked before the site-wide service file (Peter Eisentraut) Properly report an error if the specified libpq service cannot be found (Peter Eisentraut) Add TCP keepalive settings in libpq (Tollef Fog Heen, Fujii Masao, Robert Haas) Keepalive settings were already supported on the server end of TCP connections. Avoid extra system calls to block and unblock SIGPIPE in libpq, on platforms that offer alternative methods (Jeremy Kerr) When a .pgpass-supplied password fails, mention where the password came from in the error message (Bruce Momjian) Load all SSL certificates given in the client certificate file (Tom Lane) This improves support for indirectly-signed SSL certificates. Add SQLDA (SQL Descriptor Area) support to ecpg (Boszormenyi Zoltan) Add the DESCRIBE [ OUTPUT ] statement to ecpg (Boszormenyi Zoltan) Add an ECPGtransactionStatus function to return the current transaction status (Bernd Helmle) Add the string data type in ecpg Informix-compatibility mode (Boszormenyi Zoltan) Allow ecpg to use new and old variable names without restriction (Michael Meskes) Allow ecpg to use variable names in free() (Michael Meskes) Make ecpg_dynamic_type() return zero for non-SQL3 data types (Michael Meskes) Previously it returned the negative of the data type OID. This could be confused with valid type OIDs, however. Support long long types on platforms that already have 64-bit long (Michael Meskes) Add out-of-scope cursor support in ecpg's native mode (Boszormenyi Zoltan) This allows DECLARE to use variables that are not in scope when OPEN is called. This facility already existed in ecpg's Informix-compatibility mode. Allow dynamic cursor names in ecpg (Boszormenyi Zoltan) Allow ecpg to use noise words FROM and IN in FETCH and MOVE (Boszormenyi Zoltan) Enable client thread safety by default (Bruce Momjian) The thread-safety option can be disabled with configure --disable-thread-safety. Add support for controlling the Linux out-of-memory killer (Alex Hunsaker, Tom Lane) Now that /proc/self/oom_adj allows disabling of the Linux out-of-memory (OOM) killer, it's recommendable to disable OOM kills for the postmaster. It may then be desirable to re-enable OOM kills for the postmaster's child processes. The new compile-time option LINUX_OOM_ADJ allows the killer to be reactivated for child processes. New Makefile targets world, install-world, and installcheck-world (Andrew Dunstan) These are similar to the existing all, install, and installcheck targets, but they also build the HTML documentation, build and test contrib, and test server-side languages and ecpg. Add data and documentation installation location control to PGXS Makefiles (Mark Cave-Ayland) Add Makefile rules to build the PostgreSQL documentation as a single HTML file or as a single plain-text file (Peter Eisentraut, Bruce Momjian) Support compiling on 64-bit Windows and running in 64-bit mode (Tsutomu Yamada, Magnus Hagander) This allows for large shared memory sizes on Windows. Support server builds using Visual Studio 2008 (Magnus Hagander) Distribute prebuilt documentation in a subdirectory tree, rather than as tar archive files inside the distribution tarball (Peter Eisentraut) For example, the prebuilt HTML documentation is now in doc/src/sgml/html/; the manual pages are packaged similarly. Make the server's lexer reentrant (Tom Lane) This was needed for use of the lexer by PL/pgSQL. Improve speed of memory allocation (Tom Lane, Greg Stark) User-defined constraint triggers now have entries in pg_constraint as well as pg_trigger (Tom Lane) Because of this change, pg_constraint.pgconstrname is now redundant and has been removed. Add system catalog columns pg_constraint.conindid and pg_trigger.tgconstrindid to better document the use of indexes for constraint enforcement (Tom Lane) Allow multiple conditions to be communicated to backends using a single operating system signal (Fujii Masao) This allows new features to be added without a platform-specific constraint on the number of signal conditions. Improve source code test coverage, including contrib, PL/Python, and PL/Perl (Peter Eisentraut, Andrew Dunstan) Remove the use of flat files for system table bootstrapping (Tom Lane, Alvaro Herrera) This improves performance when using many roles or databases, and eliminates some possible failure conditions. Automatically generate the initial contents of pg_attribute for bootstrapped catalogs (John Naylor) This greatly simplifies changes to these catalogs. Split the processing of INSERT/UPDATE/DELETE operations out of execMain.c (Marko Tiikkaja) Updates are now executed in a separate ModifyTable node. This change is necessary infrastructure for future improvements. Simplify translation of psql's SQL help text (Peter Eisentraut) Reduce the lengths of some file names so that all file paths in the distribution tarball are less than 100 characters (Tom Lane) Some decompression programs have problems with longer file paths. Add a new ERRCODE_INVALID_PASSWORD SQLSTATE error code (Bruce Momjian) With authors' permissions, remove the few remaining personal source code copyright notices (Bruce Momjian) The personal copyright notices were insignificant but the community occasionally had to answer questions about them. Add new documentation section about running PostgreSQL in non-durable mode to improve performance (Bruce Momjian) Restructure the HTML documentation Makefile rules to make their dependency checks work correctly, avoiding unnecessary rebuilds (Peter Eisentraut) Use DocBook XSL stylesheets for man page building, rather than Docbook2X (Peter Eisentraut) This changes the set of tools needed to build the man pages. Improve PL/Perl code structure (Tim Bunce) Improve error context reports in PL/Perl (Alexey Klyukin) Note that these requirements do not apply when building from a distribution tarball, since tarballs include the files that these programs are used to build. Require Autoconf 2.63 to build configure (Peter Eisentraut) Require Flex 2.5.31 or later to build from a CVS checkout (Tom Lane) Require Perl version 5.8 or later to build from a CVS checkout (John Naylor, Andrew Dunstan) Use a more modern API for Bonjour (Tom Lane) Bonjour support now requires OS X 10.3 or later. The older API has been deprecated by Apple. Add spinlock support for the SuperH architecture (Nobuhiro Iwamatsu) Allow non-GCC compilers to use inline functions if they support them (Kurt Harriman) Remove support for platforms that don't have a working 64-bit integer data type (Tom Lane) Restructure use of LDFLAGS to be more consistent across platforms (Tom Lane) LDFLAGS is now used for linking both executables and shared libraries, and we add on LDFLAGS_EX when linking executables, or LDFLAGS_SL when linking shared libraries. Make backend header files safe to include in C++ (Kurt Harriman, Peter Eisentraut) These changes remove keyword conflicts that previously made C++ usage difficult in backend code. However, there are still other complexities when using C++ for backend functions. extern "C" { } is still necessary in appropriate places, and memory management and error handling are still problematic. Add AggCheckCallContext() for use in detecting if a C function is being called as an aggregate (Hitoshi Harada) Change calling convention for SearchSysCache() and related functions to avoid hard-wiring the maximum number of cache keys (Robert Haas) Existing calls will still work for the moment, but can be expected to break in 9.1 or later if not converted to the new style. Require calls of fastgetattr() and heap_getattr() backend macros to provide a non-NULL fourth argument (Robert Haas) Custom typanalyze functions should no longer rely on VacAttrStats.attr to determine the type of data they will be passed (Tom Lane) This was changed to allow collection of statistics on index columns for which the storage type is different from the underlying column data type. There are new fields that tell the actual datatype being analyzed. Add parser hooks for processing ColumnRef and ParamRef nodes (Tom Lane) Add a ProcessUtility hook so loadable modules can control utility commands (Itagaki Takahiro) Add contrib/pg_upgrade to support in-place upgrades (Bruce Momjian) This avoids the requirement of dumping/reloading the database when upgrading to a new major release of PostgreSQL, thus reducing downtime by orders of magnitude. It supports upgrades to 9.0 from PostgreSQL 8.3 and 8.4. Add support for preserving relation relfilenode values during binary upgrades (Bruce Momjian) Add support for preserving pg_type and pg_enum OIDs during binary upgrades (Bruce Momjian) Move data files within tablespaces into PostgreSQL-version-specific subdirectories (Bruce Momjian) This simplifies binary upgrades. Add multithreading option (-j) to contrib/pgbench (Itagaki Takahiro) This allows multiple CPUs to be used by pgbench, reducing the risk of pgbench itself becoming the test bottleneck. Add \shell and \setshell meta commands to contrib/pgbench (Michael Paquier) New features for contrib/dict_xsyn (Sergey Karpov) The new options are matchorig, matchsynonyms, and keepsynonyms. Add full text dictionary contrib/unaccent (Teodor Sigaev) This filtering dictionary removes accents from letters, which makes full-text searches over multiple languages much easier. Add dblink_get_notify() to contrib/dblink (Marcus Kempe) This allows asynchronous notifications in dblink. Improve contrib/dblink's handling of dropped columns (Tom Lane) This affects dblink_build_sql_insert() and related functions. These functions now number columns according to logical not physical column numbers. Greatly increase contrib/hstore's data length limit, and add B-tree and hash support so GROUP BY and DISTINCT operations are possible on hstore columns (Andrew Gierth) New functions and operators were also added. These improvements make hstore a full-function key-value store embedded in PostgreSQL. Add contrib/passwordcheck to support site-specific password strength policies (Laurenz Albe) The source code of this module should be modified to implement site-specific password policies. Add contrib/pg_archivecleanup tool (Simon Riggs) This is designed to be used in the archive_cleanup_command server parameter, to remove no-longer-needed archive files. Add query text to contrib/auto_explain output (Andrew Dunstan) Add buffer access counters to contrib/pg_stat_statements (Itagaki Takahiro) Update contrib/start-scripts/linux to use /proc/self/oom_adj to disable the Linux out-of-memory (OOM) killer (Alex Hunsaker, Tom Lane)