E.95. Release 9.1.5

E.95.1. Migration to Version 9.1.5

A dump/restore is not required for those running 9.1.X.

However, if you are upgrading from a version earlier than 9.1.4, see Section E.96.

E.95.2. Changes

• Prevent access to external files/URLs via XML entity references (Noah Misch, Tom Lane)

  xml_parse() would attempt to fetch external files or URLs as needed to resolve DTD and entity references in an XML value, thus allowing unprivileged database users to attempt to fetch data with the privileges of the database server. While the external data wouldn't get returned directly to the user, portions of it could be exposed in error messages if the data didn't parse as valid XML; and in any case the mere ability to check existence of a file might be useful to an attacker. (CVE-2012-3489)

• Prevent access to external files/URLs via contrib/xml2's xslt_process() (Peter Eisentraut)

  libxslt offers the ability to read and write both files and URLs through stylesheet commands, thus allowing unprivileged database users to both read and write data with the privileges of the database server. Disable that through proper use of libxslt's security options. (CVE-2012-3488)

  Also, remove xslt_process()'s ability to fetch documents and stylesheets from external files/URLs. While this was a documented "feature", it was long regarded as a bad idea. The fix for CVE-2012-3489 broke that capability, and rather than expend effort on trying to fix it, we're just going to summarily remove it.

• Prevent too-early recycling of btree index pages (Noah Misch)

  When we allowed read-only transactions to skip assigning XIDs, we introduced the possibility that a deleted btree page could be recycled while a read-only transaction was still in flight to it. This would result in incorrect index search results. The probability of such an error occurring in the field seems very low because of the timing requirements, but nonetheless it should be fixed.

• Fix crash-safety bug with newly-created-or-reset sequences (Tom Lane)

  If ALTER SEQUENCE was executed on a freshly created or reset sequence, and then precisely one nextval() call was made on it, and then the server crashed, WAL replay would restore the sequence to a state in which it appeared that no nextval() had been done, thus allowing the first sequence value to be returned again by the next nextval() call. In particular this could manifest for serial columns, since creation of a serial column's sequence includes an ALTER SEQUENCE OWNED BY step.

• Fix race condition in enum-type value comparisons (Robert Haas, Tom Lane)

  Comparisons could fail when encountering an enum value added since the current query started.

• Fix txid_current() to report the correct epoch when not in hot standby (Heikki Linnakangas)

  This fixes a regression introduced in the previous minor release.

• Prevent selection of unsuitable replication connections as the synchronous standby (Fujii Masao)

  The master might improperly choose pseudo-servers such as pg_receivexlog or pg_basebackup as the synchronous standby, and then wait indefinitely for them.

• Fix bug in startup of Hot Standby when a master transaction has many subtransactions (Andres Freund)

  This mistake led to failures reported as "out-of-order XID insertion in KnownAssignedXids".

• Ensure the backup_label file is fsync'd after pg_start_backup() (Dave Kerr)

• Fix timeout handling in walsender processes (Tom Lane)

  WAL sender background processes neglected to establish a SIGALRM handler, meaning they would wait forever in some corner cases where a timeout ought to happen.

• Wake walsenders after each background flush by walwriter (Andres Freund, Simon Riggs)

  This greatly reduces replication delay when the workload contains only asynchronously-committed transactions.

• Fix LISTEN/NOTIFY to cope better with I/O problems, such as out of disk space (Tom Lane)

  After a write failure, all subsequent attempts to send more NOTIFY messages would fail with messages like "Could not read from file "pg_notify/nnnn" at offset nnnnn: Success".

• Only allow autovacuum to be auto-canceled by a directly blocked process (Tom Lane)

  The original coding could allow inconsistent behavior in some cases; in particular, an autovacuum could get canceled after less than deadlock_timeout grace period.

• Improve logging of autovacuum cancels (Robert Haas)

• Fix log collector so that log_truncate_on_rotation works during the very first log rotation after server start (Tom Lane)

• Fix WITH attached to a nested set operation (UNION/INTERSECT/EXCEPT) (Tom Lane)

• Ensure that a whole-row reference to a subquery doesn't include any extra GROUP BY or ORDER BY columns (Tom Lane)

• Fix dependencies generated during ALTER TABLE ... ADD CONSTRAINT USING INDEX (Tom Lane)

  This command left behind a redundant pg_depend entry for the index, which could confuse later operations, notably ALTER TABLE ... ALTER COLUMN TYPE on one of the indexed columns.

• Fix REASSIGN OWNED to work on extensions (Alvaro Herrera)

• Disallow copying whole-row references in CHECK constraints and index definitions during CREATE TABLE (Tom Lane)

  This situation can arise in CREATE TABLE with LIKE or INHERITS. The copied whole-row variable was incorrectly labeled with the row type of the original table not the new one. Rejecting the case seems reasonable for LIKE, since the row types might well diverge later. For INHERITS we should ideally allow it, with an implicit coercion to the parent table's row type; but that will require more work than seems safe to back-patch.

• Fix memory leak in ARRAY(SELECT ...) subqueries (Heikki Linnakangas, Tom Lane)

• Fix planner to pass correct collation to operator selectivity estimators (Tom Lane)

  This was not previously required by any core selectivity estimation function, but third-party code might need it.

• Fix extraction of common prefixes from regular expressions (Tom Lane)

  The code could get confused by quantified parenthesized subexpressions, such as ^(foo)?bar. This would lead to incorrect index optimization of searches for such patterns.

• Fix bugs with parsing signed hh:mm and hh:mm:ss fields in interval constants (Amit Kapila, Tom Lane)

• Fix pg_dump to better handle views containing partial GROUP BY lists (Tom Lane)

  A view that lists only a primary key column in GROUP BY, but uses other table columns as if they were grouped, gets marked as depending on the primary key. Improper handling of such primary key dependencies in pg_dump resulted in poorly-ordered dumps, which at best would be inefficient to restore and at worst could result in outright failure of a parallel pg_restore run.

• In PL/Perl, avoid setting UTF8 flag when in SQL_ASCII encoding (Alex Hunsaker, Kyotaro Horiguchi, Alvaro Herrera)

• Use Postgres' encoding conversion functions, not Python's, when converting a Python Unicode string to the server encoding in PL/Python (Jan Urbanski)

  This avoids some corner-case problems, notably that Python doesn't support all the encodings Postgres does. A notable functional change is that if the server encoding is SQL_ASCII, you will get the UTF-8 representation of the string; formerly, any non-ASCII characters in the string would result in an error.

• Fix mapping of PostgreSQL encodings to Python encodings in PL/Python (Jan Urbanski)

• Report errors properly in contrib/xml2's xslt_process() (Tom Lane)

• Update time zone data files to tzdata release 2012e for DST law changes in Morocco and Tokelau