E.138. Release 8.3.1

E.138.1. Migration to Version 8.3.1

A dump/restore is not required for those running 8.3.X. However, you might need to REINDEX indexes on textual columns after updating, if you are affected by the Windows locale issue described below.

E.138.2. Changes

• Fix character string comparison for Windows locales that consider different character combinations as equal (Tom)

  This fix applies only on Windows and only when using UTF-8 database encoding. The same fix was made for all other cases over two years ago, but Windows with UTF-8 uses a separate code path that was not updated. If you are using a locale that considers some non-identical strings as equal, you may need to REINDEX to fix existing indexes on textual columns.

• Repair corner-case bugs in VACUUM FULL (Tom)

  A potential deadlock between concurrent VACUUM FULL operations on different system catalogs was introduced in 8.2. This has now been corrected. 8.3 made this worse because the deadlock could occur within a critical code section, making it a PANIC rather than just ERROR condition.

  Also, a VACUUM FULL that failed partway through vacuuming a system catalog could result in cache corruption in concurrent database sessions.

  Another VACUUM FULL bug introduced in 8.3 could result in a crash or out-of-memory report when dealing with pages containing no live tuples.

• Fix misbehavior of foreign key checks involving character or bit columns (Tom)

  If the referencing column were of a different but compatible type (for instance varchar), the constraint was enforced incorrectly.

• Avoid needless deadlock failures in no-op foreign-key checks (Stephan Szabo, Tom)

• Fix possible core dump when re-planning a prepared query (Tom)

  This bug affected only protocol-level prepare operations, not SQL PREPARE, and so tended to be seen only with JDBC, DBI, and other client-side drivers that use prepared statements heavily.

• Fix possible failure when re-planning a query that calls an SPI-using function (Tom)

• Fix failure in row-wise comparisons involving columns of different datatypes (Tom)

• Fix longstanding LISTEN/NOTIFY race condition (Tom)

  In rare cases a session that had just executed a LISTEN might not get a notification, even though one would be expected because the concurrent transaction executing NOTIFY was observed to commit later.

  A side effect of the fix is that a transaction that has executed a not-yet-committed LISTEN command will not see any row in pg_listener for the LISTEN, should it choose to look; formerly it would have. This behavior was never documented one way or the other, but it is possible that some applications depend on the old behavior.

• Disallow LISTEN and UNLISTEN within a prepared transaction (Tom)

  This was formerly allowed but trying to do it had various unpleasant consequences, notably that the originating backend could not exit as long as an UNLISTEN remained uncommitted.

• Disallow dropping a temporary table within a prepared transaction (Heikki)

  This was correctly disallowed by 8.1, but the check was inadvertently broken in 8.2 and 8.3.

• Fix rare crash when an error occurs during a query using a hash index (Heikki)

• Fix incorrect comparison of tsquery values (Teodor)

• Fix incorrect behavior of LIKE with non-ASCII characters in single-byte encodings (Rolf Jentsch)

• Disable xmlvalidate (Tom)

  This function should have been removed before 8.3 release, but was inadvertently left in the source code. It poses a small security risk since unprivileged users could use it to read the first few characters of any file accessible to the server.

• Fix memory leaks in certain usages of set-returning functions (Neil)

• Make encode(bytea, 'escape') convert all high-bit-set byte values into \nnn octal escape sequences (Tom)

  This is necessary to avoid encoding problems when the database encoding is multi-byte. This change could pose compatibility issues for applications that are expecting specific results from encode.

• Fix input of datetime values for February 29 in years BC (Tom)

  The former coding was mistaken about which years were leap years.

• Fix "unrecognized node type" error in some variants of ALTER OWNER (Tom)

• Avoid tablespace permissions errors in CREATE TABLE LIKE INCLUDING INDEXES (Tom)

• Ensure pg_stat_activity.waiting flag is cleared when a lock wait is aborted (Tom)

• Fix handling of process permissions on Windows Vista (Dave, Magnus)

  In particular, this fix allows starting the server as the Administrator user.

• Update time zone data files to tzdata release 2008a (in particular, recent Chile changes); adjust timezone abbreviation VET (Venezuela) to mean UTC-4:30, not UTC-4:00 (Tom)

• Fix ecpg problems with arrays (Michael)

• Fix pg_ctl to correctly extract the postmaster's port number from command-line options (Itagaki Takahiro, Tom)

  Previously, pg_ctl start -w could try to contact the postmaster on the wrong port, leading to bogus reports of startup failure.

• Use -fwrapv to defend against possible misoptimization in recent gcc versions (Tom)

  This is known to be necessary when building PostgreSQL with gcc 4.3 or later.

• Enable building contrib/uuid-ossp with MSVC (Hiroshi Saito)