Upgrading from 9.3.1 to 9.3.25 gives you 5.1 years worth of fixes (783 of them)

List of changes:

• ⇑ Upgrade to 9.3.2 released on 2013-12-05 - docs

  • Fix VACUUM's tests to see whether it can update relfrozenxid (Andres Freund)

    In some cases VACUUM (either manual or autovacuum) could incorrectly advance a table's relfrozenxid value, allowing tuples to escape freezing, causing those rows to become invisible once 2^31 transactions have elapsed. The probability of data loss is fairly low since multiple incorrect advancements would need to happen before actual loss occurs, but it's not zero. In 9.2.0 and later, the probability of loss is higher, and it's also possible to get “could not access status of transaction” errors as a consequence of this bug. Users upgrading from releases 9.0.4 or 8.4.8 or earlier are not affected, but all later versions contain the bug.

    The issue can be ameliorated by, after upgrading, vacuuming all tables in all databases while having vacuum_freeze_table_age set to zero. This will fix any latent corruption but will not be able to fix all pre-existing data errors. However, an installation can be presumed safe after performing this vacuuming if it has executed fewer than 2^31 update transactions in its lifetime (check this with SELECT txid_current() < 2^31).

  • Fix multiple bugs in MultiXactId freezing (Andres Freund, Álvaro Herrera)

    These bugs could lead to “could not access status of transaction” errors, or to duplicate or vanishing rows. Users upgrading from releases prior to 9.3.0 are not affected.

    The issue can be ameliorated by, after upgrading, vacuuming all tables in all databases while having vacuum_freeze_table_age set to zero. This will fix latent corruption but will not be able to fix all pre-existing data errors.

    As a separate issue, these bugs can also cause standby servers to get out of sync with the primary, thus exhibiting data errors that are not in the primary. Therefore, it's recommended that 9.3.0 and 9.3.1 standby servers be re-cloned from the primary (e.g., with a new base backup) after upgrading.

  • Fix initialization of pg_clog and pg_subtrans during hot standby startup (Andres Freund, Heikki Linnakangas)

    This bug can cause data loss on standby servers at the moment they start to accept hot-standby queries, by marking committed transactions as uncommitted. The likelihood of such corruption is small unless, at the time of standby startup, the primary server has executed many updating transactions since its last checkpoint. Symptoms include missing rows, rows that should have been deleted being still visible, and obsolete versions of updated rows being still visible alongside their newer versions.

    This bug was introduced in versions 9.3.0, 9.2.5, 9.1.10, and 9.0.14. Standby servers that have only been running earlier releases are not at risk. It's recommended that standby servers that have ever run any of the buggy releases be re-cloned from the primary (e.g., with a new base backup) after upgrading.

  • Fix multiple bugs in update chain traversal (Andres Freund, Álvaro Herrera)

    These bugs could result in incorrect behavior, such as locking or even updating the wrong row, in the presence of concurrent updates. Spurious “unable to fetch updated version of tuple” errors were also possible.

  • Fix dangling-pointer problem in fast-path locking (Tom Lane)

    This could lead to corruption of the lock data structures in shared memory, causing “lock already held” and other odd errors.

  • Fix assorted race conditions in timeout management (Tom Lane)

    These errors could result in a server process becoming unresponsive because it had blocked SIGALRM and/or SIGINT.

  • Truncate pg_multixact contents during WAL replay (Andres Freund)

    This avoids ever-increasing disk space consumption in standby servers.

  • Ensure an anti-wraparound VACUUM counts a page as scanned when it's only verified that no tuples need freezing (Sergey Burladyan, Jeff Janes)

    This bug could result in failing to advance relfrozenxid, so that the table would still be thought to need another anti-wraparound vacuum. In the worst case the database might even shut down to prevent wraparound.

  • Fix full-table-vacuum request mechanism for MultiXactIds (Andres Freund)

    This bug could result in large amounts of useless autovacuum activity.

  • Fix race condition in GIN index posting tree page deletion (Heikki Linnakangas)

    This could lead to transient wrong answers or query failures.

  • Fix “unexpected spgdoinsert() failure” error during SP-GiST index creation (Teodor Sigaev)

  • Fix assorted bugs in materialized views (Kevin Grittner, Andres Freund)

  • Re-allow duplicate table aliases if they're within aliased JOINs (Tom Lane)

    Historically PostgreSQL has accepted queries like

    SELECT ... FROM tab1 x CROSS JOIN (tab2 x CROSS JOIN tab3 y) z

    although a strict reading of the SQL standard would forbid the duplicate usage of table alias x. A misguided change in 9.3.0 caused it to reject some such cases that were formerly accepted. Restore the previous behavior.

  • Avoid flattening a subquery whose SELECT list contains a volatile function wrapped inside a sub-SELECT (Tom Lane)

    This avoids unexpected results due to extra evaluations of the volatile function.

  • Fix planner's processing of non-simple-variable subquery outputs nested within outer joins (Tom Lane)

    This error could lead to incorrect plans for queries involving multiple levels of subqueries within JOIN syntax.

  • Fix incorrect planning in cases where the same non-strict expression appears in multiple WHERE and outer JOIN equality clauses (Tom Lane)

  • Fix planner crash with whole-row reference to a subquery (Tom Lane)

  • Fix incorrect generation of optimized MIN()/MAX() plans for inheritance trees (Tom Lane)

    The planner could fail in cases where the MIN()/MAX() argument was an expression rather than a simple variable.

  • Fix premature deletion of temporary files (Andres Freund)

  • Prevent intra-transaction memory leak when printing range values (Tom Lane)

    This fix actually cures transient memory leaks in any datatype output function, but range types are the only ones known to have had a significant problem.

  • Fix memory leaks when reloading configuration files (Heikki Linnakangas, Hari Babu)

  • Prevent incorrect display of dropped columns in NOT NULL and CHECK constraint violation messages (Michael Paquier and Tom Lane)

  • Allow default arguments and named-argument notation for window functions (Tom Lane)

    Previously, these cases were likely to crash.

  • Suppress trailing whitespace on each line when pretty-printing rules and views (Tom Lane)

    9.3.0 generated such whitespace in many more cases than previous versions did. To reduce unexpected behavioral changes, suppress unnecessary whitespace in all cases.

  • Fix possible read past end of memory in rule printing (Peter Eisentraut)

  • Fix array slicing of int2vector and oidvector values (Tom Lane)

    Expressions of this kind are now implicitly promoted to regular int2 or oid arrays.

  • Return a valid JSON value when converting an empty hstore value to json (Oskari Saarenmaa)

  • Fix incorrect behaviors when using a SQL-standard, simple GMT offset timezone (Tom Lane)

    In some cases, the system would use the simple GMT offset value when it should have used the regular timezone setting that had prevailed before the simple offset was selected. This change also causes the timeofday function to honor the simple GMT offset zone.

  • Prevent possible misbehavior when logging translations of Windows error codes (Tom Lane)

  • Properly quote generated command lines in pg_ctl (Naoya Anzai and Tom Lane)

    This fix applies only to Windows.

  • Fix pg_dumpall to work when a source database sets default_transaction_read_only via ALTER DATABASE SET (Kevin Grittner)

    Previously, the generated script would fail during restore.

  • Fix pg_isready to handle its -d option properly (Fabrízio de Royes Mello and Fujii Masao)

  • Fix parsing of WAL file names in pg_receivexlog (Heikki Linnakangas)

    This error made pg_receivexlog unable to restart streaming after stopping, once at least 4 GB of WAL had been written.

  • Report out-of-disk-space failures properly in pg_upgrade (Peter Eisentraut)

  • Make ecpg search for quoted cursor names case-sensitively (Zoltán Böszörményi)

  • Fix ecpg's processing of lists of variables declared varchar (Zoltán Böszörményi)

  • Make contrib/lo defend against incorrect trigger definitions (Marc Cousin)

  • Update time zone data files to tzdata release 2013h for DST law changes in Argentina, Brazil, Jordan, Libya, Liechtenstein, Morocco, and Palestine. Also, new timezone abbreviations WIB, WIT, WITA for Indonesia.

• ⇑ Upgrade to 9.3.3 released on 2014-02-20 - docs

  • Shore up GRANT ... WITH ADMIN OPTION restrictions (Noah Misch)

    Granting a role without ADMIN OPTION is supposed to prevent the grantee from adding or removing members from the granted role, but this restriction was easily bypassed by doing SET ROLE first. The security impact is mostly that a role member can revoke the access of others, contrary to the wishes of his grantor. Unapproved role member additions are a lesser concern, since an uncooperative role member could provide most of his rights to others anyway by creating views or SECURITY DEFINER functions. (CVE-2014-0060)

  • Prevent privilege escalation via manual calls to PL validator functions (Andres Freund)

    The primary role of PL validator functions is to be called implicitly during CREATE FUNCTION, but they are also normal SQL functions that a user can call explicitly. Calling a validator on a function actually written in some other language was not checked for and could be exploited for privilege-escalation purposes. The fix involves adding a call to a privilege-checking function in each validator function. Non-core procedural languages will also need to make this change to their own validator functions, if any. (CVE-2014-0061)

  • Avoid multiple name lookups during table and index DDL (Robert Haas, Andres Freund)

    If the name lookups come to different conclusions due to concurrent activity, we might perform some parts of the DDL on a different table than other parts. At least in the case of CREATE INDEX, this can be used to cause the permissions checks to be performed against a different table than the index creation, allowing for a privilege escalation attack. (CVE-2014-0062)

  • Prevent buffer overrun with long datetime strings (Noah Misch)

    The MAXDATELEN constant was too small for the longest possible value of type interval, allowing a buffer overrun in interval_out(). Although the datetime input functions were more careful about avoiding buffer overrun, the limit was short enough to cause them to reject some valid inputs, such as input containing a very long timezone name. The ecpg library contained these vulnerabilities along with some of its own. (CVE-2014-0063)

  • Prevent buffer overrun due to integer overflow in size calculations (Noah Misch, Heikki Linnakangas)

    Several functions, mostly type input functions, calculated an allocation size without checking for overflow. If overflow did occur, a too-small buffer would be allocated and then written past. (CVE-2014-0064)

  • Prevent overruns of fixed-size buffers (Peter Eisentraut, Jozef Mlich)

    Use strlcpy() and related functions to provide a clear guarantee that fixed-size buffers are not overrun. Unlike the preceding items, it is unclear whether these cases really represent live issues, since in most cases there appear to be previous constraints on the size of the input string. Nonetheless it seems prudent to silence all Coverity warnings of this type. (CVE-2014-0065)

  • Avoid crashing if crypt() returns NULL (Honza Horak, Bruce Momjian)

    There are relatively few scenarios in which crypt() could return NULL, but contrib/chkpass would crash if it did. One practical case in which this could be an issue is if libc is configured to refuse to execute unapproved hashing algorithms (e.g., “FIPS mode”). (CVE-2014-0066)

  • Document risks of make check in the regression testing instructions (Noah Misch, Tom Lane)

    Since the temporary server started by make check uses “trust” authentication, another user on the same machine could connect to it as database superuser, and then potentially exploit the privileges of the operating-system user who started the tests. A future release will probably incorporate changes in the testing procedure to prevent this risk, but some public discussion is needed first. So for the moment, just warn people against using make check when there are untrusted users on the same machine. (CVE-2014-0067)

  • Rework tuple freezing protocol (Álvaro Herrera, Andres Freund)

    The logic for tuple freezing was unable to handle some cases involving freezing of multixact IDs, with the practical effect that shared row-level locks might be forgotten once old enough.

    Fixing this required changing the WAL record format for tuple freezing. While this is no issue for standalone servers, when using replication it means that standby servers must be upgraded to 9.3.3 or later before their masters are. An older standby will be unable to interpret freeze records generated by a newer master, and will fail with a PANIC message. (In such a case, upgrading the standby should be sufficient to let it resume execution.)

  • Create separate GUC parameters to control multixact freezing (Álvaro Herrera)

    9.3 requires multixact tuple labels to be frozen before they grow too old, in the same fashion as plain transaction ID labels have been frozen for some time. Previously, the transaction ID freezing parameters were used for multixact IDs too; but since the consumption rates of transaction IDs and multixact IDs can be quite different, this did not work very well. Introduce new settings vacuum_multixact_freeze_min_age, vacuum_multixact_freeze_table_age, and autovacuum_multixact_freeze_max_age to control when to freeze multixacts.

  • Account for remote row locks propagated by local updates (Álvaro Herrera)

    If a row was locked by transaction A, and transaction B updated it, the new version of the row created by B would be locked by A, yet visible only to B. If transaction B then again updated the row, A's lock wouldn't get checked, thus possibly allowing B to complete when it shouldn't. This case is new in 9.3 since prior versions did not have any types of row locking that would permit another transaction to update the row at all.

    This oversight could allow referential integrity checks to give false positives (for instance, allow deletes that should have been rejected). Applications using the new commands SELECT FOR KEY SHARE and SELECT FOR NO KEY UPDATE might also have suffered locking failures of this kind.

  • Prevent “forgetting” valid row locks when one of several holders of a row lock aborts (Álvaro Herrera)

    This was yet another mechanism by which a shared row lock could be lost, thus possibly allowing updates that should have been prevented by foreign-key constraints.

  • Fix incorrect logic during update chain locking (Álvaro Herrera)

    This mistake could result in spurious “could not serialize access due to concurrent update” errors in REPEATABLE READ and SERIALIZABLE transaction isolation modes.

  • Handle wraparound correctly during extension or truncation of pg_multixact/members (Andres Freund, Álvaro Herrera)

  • Fix handling of 5-digit filenames in pg_multixact/members (Álvaro Herrera)

    As of 9.3, these names can be more than 4 digits, but the directory cleanup code ignored such files.

  • Improve performance of multixact cache code (Álvaro Herrera)

  • Optimize updating a row that's already locked by the same transaction (Andres Freund, Álvaro Herrera)

    This fixes a performance regression from pre-9.3 versions when doing SELECT FOR UPDATE followed by UPDATE/DELETE.

  • During archive recovery, prefer highest timeline number when WAL segments with the same ID are present in both the archive and pg_xlog/ (Kyotaro Horiguchi)

    Previously, not-yet-archived segments could get ignored during recovery. This reverts an undesirable behavioral change in 9.3.0 back to the way things worked pre-9.3.

  • Fix possible mis-replay of WAL records when some segments of a relation aren't full size (Greg Stark, Tom Lane)

    The WAL update could be applied to the wrong page, potentially many pages past where it should have been. Aside from corrupting data, this error has been observed to result in significant “bloat” of standby servers compared to their masters, due to updates being applied far beyond where the end-of-file should have been. This failure mode does not appear to be a significant risk during crash recovery, only when initially synchronizing a standby created from a base backup taken from a quickly-changing master.

  • Fix bug in determining when recovery has reached consistency (Tomonari Katsumata, Heikki Linnakangas)

    In some cases WAL replay would mistakenly conclude that the database was already consistent at the start of replay, thus possibly allowing hot-standby queries before the database was really consistent. Other symptoms such as “PANIC: WAL contains references to invalid pages” were also possible.

  • Fix WAL logging of visibility map changes (Heikki Linnakangas)

  • Fix improper locking of btree index pages while replaying a VACUUM operation in hot-standby mode (Andres Freund, Heikki Linnakangas, Tom Lane)

    This error could result in “PANIC: WAL contains references to invalid pages” failures.

  • Ensure that insertions into non-leaf GIN index pages write a full-page WAL record when appropriate (Heikki Linnakangas)

    The previous coding risked index corruption in the event of a partial-page write during a system crash.

  • When pause_at_recovery_target and recovery_target_inclusive are both set, ensure the target record is applied before pausing, not after (Heikki Linnakangas)

  • Ensure walreceiver sends hot-standby feedback messages on time even when there is a continuous stream of data (Andres Freund, Amit Kapila)

  • Prevent timeout interrupts from taking control away from mainline code unless ImmediateInterruptOK is set (Andres Freund, Tom Lane)

    This is a serious issue for any application making use of statement timeouts, as it could cause all manner of strange failures after a timeout occurred. We have seen reports of “stuck” spinlocks, ERRORs being unexpectedly promoted to PANICs, unkillable backends, and other misbehaviors.

  • Fix race conditions during server process exit (Robert Haas)

    Ensure that signal handlers don't attempt to use the process's MyProc pointer after it's no longer valid.

  • Fix race conditions in walsender shutdown logic and walreceiver SIGHUP signal handler (Tom Lane)

  • Fix unsafe references to errno within error reporting logic (Christian Kruse)

    This would typically lead to odd behaviors such as missing or inappropriate HINT fields.

  • Fix possible crashes from using ereport() too early during server startup (Tom Lane)

    The principal case we've seen in the field is a crash if the server is started in a directory it doesn't have permission to read.

  • Clear retry flags properly in OpenSSL socket write function (Alexander Kukushkin)

    This omission could result in a server lockup after unexpected loss of an SSL-encrypted connection.

  • Fix length checking for Unicode identifiers (U&"..." syntax) containing escapes (Tom Lane)

    A spurious truncation warning would be printed for such identifiers if the escaped form of the identifier was too long, but the identifier actually didn't need truncation after de-escaping.

  • Fix parsing of Unicode literals and identifiers just before the end of a command string or function body (Tom Lane)

  • Allow keywords that are type names to be used in lists of roles (Stephen Frost)

    A previous patch allowed such keywords to be used without quoting in places such as role identifiers; but it missed cases where a list of role identifiers was permitted, such as DROP ROLE.

  • Fix parser crash for EXISTS(SELECT * FROM zero_column_table) (Tom Lane)

  • Fix possible crash due to invalid plan for nested sub-selects, such as WHERE (... x IN (SELECT ...) ...) IN (SELECT ...) (Tom Lane)

  • Fix mishandling of WHERE conditions pulled up from a LATERAL subquery (Tom Lane)

    The typical symptom of this bug was a “JOIN qualification cannot refer to other relations” error, though subtle logic errors in created plans seem possible as well.

  • Disallow LATERAL references to the target table of an UPDATE/DELETE (Tom Lane)

    While this might be allowed in some future release, it was unintentional in 9.3, and didn't work quite right anyway.

  • Fix UPDATE/DELETE of an inherited target table that has UNION ALL subqueries (Tom Lane)

    Without this fix, UNION ALL subqueries aren't correctly inserted into the update plans for inheritance child tables after the first one, typically resulting in no update happening for those child table(s).

  • Fix ANALYZE to not fail on a column that's a domain over a range type (Tom Lane)

  • Ensure that ANALYZE creates statistics for a table column even when all the values in it are “too wide” (Tom Lane)

    ANALYZE intentionally omits very wide values from its histogram and most-common-values calculations, but it neglected to do something sane in the case that all the sampled entries are too wide.

  • In ALTER TABLE ... SET TABLESPACE, allow the database's default tablespace to be used without a permissions check (Stephen Frost)

    CREATE TABLE has always allowed such usage, but ALTER TABLE didn't get the memo.

  • Fix support for extensions containing event triggers (Tom Lane)

  • Fix “cannot accept a set” error when some arms of a CASE return a set and others don't (Tom Lane)

  • Fix memory leakage in JSON functions (Craig Ringer)

  • Properly distinguish numbers from non-numbers when generating JSON output (Andrew Dunstan)

  • Fix checks for all-zero client addresses in pgstat functions (Kevin Grittner)

  • Fix possible misclassification of multibyte characters by the text search parser (Tom Lane)

    Non-ASCII characters could be misclassified when using C locale with a multibyte encoding. On Cygwin, non-C locales could fail as well.

  • Fix possible misbehavior in plainto_tsquery() (Heikki Linnakangas)

    Use memmove() not memcpy() for copying overlapping memory regions. There have been no field reports of this actually causing trouble, but it's certainly risky.

  • Fix placement of permissions checks in pg_start_backup() and pg_stop_backup() (Andres Freund, Magnus Hagander)

    The previous coding might attempt to do catalog access when it shouldn't.

  • Accept SHIFT_JIS as an encoding name for locale checking purposes (Tatsuo Ishii)

  • Fix *-qualification of named parameters in SQL-language functions (Tom Lane)

    Given a composite-type parameter named foo, $1.* worked fine, but foo.* not so much.

  • Fix misbehavior of PQhost() on Windows (Fujii Masao)

    It should return localhost if no host has been specified.

  • Improve error handling in libpq and psql for failures during COPY TO STDOUT/FROM STDIN (Tom Lane)

    In particular this fixes an infinite loop that could occur in 9.2 and up if the server connection was lost during COPY FROM STDIN. Variants of that scenario might be possible in older versions, or with other client applications.

  • Fix incorrect translation handling in some psql \d commands (Peter Eisentraut, Tom Lane)

  • Ensure pg_basebackup's background process is killed when exiting its foreground process (Magnus Hagander)

  • Fix possible incorrect printing of filenames in pg_basebackup's verbose mode (Magnus Hagander)

  • Avoid including tablespaces inside PGDATA twice in base backups (Dimitri Fontaine, Magnus Hagander)

  • Fix misaligned descriptors in ecpg (MauMau)

  • In ecpg, handle lack of a hostname in the connection parameters properly (Michael Meskes)

  • Fix performance regression in contrib/dblink connection startup (Joe Conway)

    Avoid an unnecessary round trip when client and server encodings match.

  • In contrib/isn, fix incorrect calculation of the check digit for ISMN values (Fabien Coelho)

  • Fix contrib/pgbench's progress logging to avoid overflow when the scale factor is large (Tatsuo Ishii)

  • Fix contrib/pg_stat_statement's handling of CURRENT_DATE and related constructs (Kyotaro Horiguchi)

  • Improve lost-connection error handling in contrib/postgres_fdw (Tom Lane)

  • Ensure client-code-only installation procedure works as documented (Peter Eisentraut)

  • In Mingw and Cygwin builds, install the libpq DLL in the bin directory (Andrew Dunstan)

    This duplicates what the MSVC build has long done. It should fix problems with programs like psql failing to start because they can't find the DLL.

  • Avoid using the deprecated dllwrap tool in Cygwin builds (Marco Atzeri)

  • Enable building with Visual Studio 2013 (Brar Piening)

  • Don't generate plain-text HISTORY and src/test/regress/README files anymore (Tom Lane)

    These text files duplicated the main HTML and PDF documentation formats. The trouble involved in maintaining them greatly outweighs the likely audience for plain-text format. Distribution tarballs will still contain files by these names, but they'll just be stubs directing the reader to consult the main documentation. The plain-text INSTALL file will still be maintained, as there is arguably a use-case for that.

  • Update time zone data files to tzdata release 2013i for DST law changes in Jordan and historical changes in Cuba.

    In addition, the zones Asia/Riyadh87, Asia/Riyadh88, and Asia/Riyadh89 have been removed, as they are no longer maintained by IANA, and never represented actual civil timekeeping practice.

• ⇑ Upgrade to 9.3.4 released on 2014-03-20 - docs

  • Fix WAL replay of locking an already-updated tuple (Andres Freund, Álvaro Herrera)

    This error caused updated rows to not be found by index scans, resulting in inconsistent query results depending on whether an index scan was used. Subsequent processing could result in constraint violations, since the previously updated row would not be found by later index searches, thus possibly allowing conflicting rows to be inserted. Since this error is in WAL replay, it would only manifest during crash recovery or on standby servers. The improperly-replayed case most commonly arises when a table row that is referenced by a foreign-key constraint is updated concurrently with creation of a referencing row.

  • Restore GIN metapages unconditionally to avoid torn-page risk (Heikki Linnakangas)

    Although this oversight could theoretically result in a corrupted index, it is unlikely to have caused any problems in practice, since the active part of a GIN metapage is smaller than a standard 512-byte disk sector.

  • Avoid race condition in checking transaction commit status during receipt of a NOTIFY message (Marko Tiikkaja)

    This prevents a scenario wherein a sufficiently fast client might respond to a notification before database updates made by the notifier have become visible to the recipient.

  • Allow materialized views to be referenced in UPDATE and DELETE commands (Michael Paquier)

    Previously such queries failed with a complaint about not being able to lock rows in the materialized view.

  • Allow regular-expression operators to be terminated early by query cancel requests (Tom Lane)

    This prevents scenarios wherein a pathological regular expression could lock up a server process uninterruptibly for a long time.

  • Remove incorrect code that tried to allow OVERLAPS with single-element row arguments (Joshua Yanovski)

    This code never worked correctly, and since the case is neither specified by the SQL standard nor documented, it seemed better to remove it than fix it.

  • Avoid getting more than AccessShareLock when de-parsing a rule or view (Dean Rasheed)

    This oversight resulted in pg_dump unexpectedly acquiring RowExclusiveLock locks on tables mentioned as the targets of INSERT/UPDATE/DELETE commands in rules. While usually harmless, that could interfere with concurrent transactions that tried to acquire, for example, ShareLock on those tables.

  • Improve performance of index endpoint probes during planning (Tom Lane)

    This change fixes a significant performance problem that occurred when there were many not-yet-committed rows at the end of the index, which is a common situation for indexes on sequentially-assigned values such as timestamps or sequence-generated identifiers.

  • Use non-default selectivity estimates for value IN (list) and value operator ANY (array) expressions when the righthand side is a stable expression (Tom Lane)

  • Remove the correct per-database statistics file during DROP DATABASE (Tomas Vondra)

    This fix prevents a permanent leak of statistics file space. Users who have done many DROP DATABASE commands since upgrading to PostgreSQL 9.3 may wish to check their statistics directory and delete statistics files that do not correspond to any existing database. Please note that db_0.stat should not be removed.

  • Fix walsender ping logic to avoid inappropriate disconnects under continuous load (Andres Freund, Heikki Linnakangas)

    walsender failed to send ping messages to the client if it was constantly busy sending WAL data; but it expected to see ping responses despite that, and would therefore disconnect once wal_sender_timeout elapsed.

  • Fix walsender's failure to shut down cleanly when client is pg_receivexlog (Fujii Masao)

  • Check WAL level and hot standby parameters correctly when doing crash recovery that will be followed by archive recovery (Heikki Linnakangas)

  • Fix test to see if hot standby connections can be allowed immediately after a crash (Heikki Linnakangas)

  • Add read-only data_checksums parameter to display whether page checksums are enabled (Heikki Linnakangas)

    Without this parameter, determining the state of checksum processing was difficult.

  • Prevent interrupts while reporting non-ERROR messages (Tom Lane)

    This guards against rare server-process freezeups due to recursive entry to syslog(), and perhaps other related problems.

  • Fix memory leak in PL/Perl when returning a composite result, including multiple-OUT-parameter cases (Alex Hunsaker)

  • Fix tracking of psql script line numbers during \copy from out-of-line data (Kumar Rajeev Rastogi, Amit Khandekar)

    \copy ... from incremented the script file line number for each data line, even if the data was not coming from the script file. This mistake resulted in wrong line numbers being reported for any errors occurring later in the same script file.

  • Fix contrib/postgres_fdw to handle multiple join conditions properly (Tom Lane)

    This oversight could result in sending WHERE clauses to the remote server for execution even though the clauses are not known to have the same semantics on the remote server (for example, clauses that use non-built-in operators). The query might succeed anyway, but it could also fail with errors from the remote server, or worse give silently wrong answers.

  • Prevent intermittent “could not reserve shared memory region” failures on recent Windows versions (MauMau)

  • Update time zone data files to tzdata release 2014a for DST law changes in Fiji and Turkey, plus historical changes in Israel and Ukraine.

• ⇑ Upgrade to 9.3.5 released on 2014-07-24 - docs

  • In pg_upgrade, remove pg_multixact files left behind by initdb (Bruce Momjian)

    If you used a pre-9.3.5 version of pg_upgrade to upgrade a database cluster to 9.3, it might have left behind a file $PGDATA/pg_multixact/offsets/0000 that should not be there and will eventually cause problems in VACUUM. However, in common cases this file is actually valid and must not be removed. To determine whether your installation has this problem, run this query as superuser, in any database of the cluster:

    WITH list(file) AS (SELECT * FROM pg_ls_dir('pg_multixact/offsets'))
    SELECT EXISTS (SELECT * FROM list WHERE file = '0000') AND
           NOT EXISTS (SELECT * FROM list WHERE file = '0001') AND
           NOT EXISTS (SELECT * FROM list WHERE file = 'FFFF') AND
           EXISTS (SELECT * FROM list WHERE file != '0000')
           AS file_0000_removal_required;

    If this query returns t, manually remove the file $PGDATA/pg_multixact/offsets/0000. Do nothing if the query returns f.

  • Correctly initialize padding bytes in contrib/btree_gist indexes on bit columns (Heikki Linnakangas)

    This error could result in incorrect query results due to values that should compare equal not being seen as equal. Users with GiST indexes on bit or bit varying columns should REINDEX those indexes after installing this update.

  • Protect against torn pages when deleting GIN list pages (Heikki Linnakangas)

    This fix prevents possible index corruption if a system crash occurs while the page update is being written to disk.

  • Don't clear the right-link of a GiST index page while replaying updates from WAL (Heikki Linnakangas)

    This error could lead to transiently wrong answers from GiST index scans performed in Hot Standby.

  • Fix corner-case infinite loop during insertion into an SP-GiST text index (Tom Lane)

  • Fix incorrect answers from SP-GiST index searches with -|- (range adjacency) operator (Heikki Linnakangas)

  • Fix wraparound handling for pg_multixact/members (Álvaro Herrera)

  • Truncate pg_multixact during checkpoints, not during VACUUM (Álvaro Herrera)

    This change ensures that pg_multixact segments can't be removed if they'd still be needed during WAL replay after a crash.

  • Fix possible inconsistency of all-visible flags after WAL recovery (Heikki Linnakangas)

  • Fix possibly-incorrect cache invalidation during nested calls to ReceiveSharedInvalidMessages (Andres Freund)

  • Fix race condition when updating a tuple concurrently locked by another process (Andres Freund, Álvaro Herrera)

  • Fix “could not find pathkey item to sort” planner failures with UNION ALL over subqueries reading from tables with inheritance children (Tom Lane)

  • Don't assume a subquery's output is unique if there's a set-returning function in its targetlist (David Rowley)

    This oversight could lead to misoptimization of constructs like WHERE x IN (SELECT y, generate_series(1,10) FROM t GROUP BY y).

  • Improve planner to drop constant-NULL inputs of AND/OR when possible (Tom Lane)

    This change fixes some cases where the more aggressive parameter substitution done by 9.2 and later can lead to a worse plan than older versions produced.

  • Ensure that the planner sees equivalent VARIADIC and non-VARIADIC function calls as equivalent (Tom Lane)

    This bug could for example result in failure to use expression indexes involving variadic functions. It might be necessary to re-create such indexes, and/or re-create views including variadic function calls that should match the indexes, for the fix to be effective for existing 9.3 installations.

  • Fix handling of nested JSON objects in json_populate_recordset() and friends (Michael Paquier, Tom Lane)

    A nested JSON object could result in previous fields of the parent object not being shown in the output.

  • Fix identification of input type category in to_json() and friends (Tom Lane)

    This is known to have led to inadequate quoting of money fields in the JSON result, and there may have been wrong results for other data types as well.

  • Fix failure to detoast fields in composite elements of structured types (Tom Lane)

    This corrects cases where TOAST pointers could be copied into other tables without being dereferenced. If the original data is later deleted, it would lead to errors like “missing chunk number 0 for toast value ...” when the now-dangling pointer is used.

  • Fix “record type has not been registered” failures with whole-row references to the output of Append plan nodes (Tom Lane)

  • Fix possible crash when invoking a user-defined function while rewinding a cursor (Tom Lane)

  • Fix query-lifespan memory leak while evaluating the arguments for a function in FROM (Tom Lane)

  • Fix session-lifespan memory leaks in regular-expression processing (Tom Lane, Arthur O'Dwyer, Greg Stark)

  • Fix data encoding error in hungarian.stop (Tom Lane)

  • Prevent foreign tables from being created with OIDS when default_with_oids is true (Etsuro Fujita)

  • Fix liveness checks for rows that were inserted in the current transaction and then deleted by a now-rolled-back subtransaction (Andres Freund)

    This could cause problems (at least spurious warnings, and at worst an infinite loop) if CREATE INDEX or CLUSTER were done later in the same transaction.

  • Clear pg_stat_activity.xact_start during PREPARE TRANSACTION (Andres Freund)

    After the PREPARE, the originating session is no longer in a transaction, so it should not continue to display a transaction start time.

  • Fix REASSIGN OWNED to not fail for text search objects (Álvaro Herrera)

  • Prevent pg_class.relminmxid values from going backwards during VACUUM FULL (Álvaro Herrera)

  • Reduce indentation in rule/view dumps to improve readability and avoid excessive whitespace (Greg Stark, Tom Lane)

    This change reduces the amount of indentation applied to nested constructs, including some cases that the user probably doesn't think of as nested, such as UNION lists. Previously, deeply nested constructs were printed with an amount of whitespace growing as O(N^2), which created a performance problem and even risk of out-of-memory failures. Now the indentation is reduced modulo 40, which is initially odd to look at but seems to preserve readability better than simply limiting the indentation would do. Redundant parenthesization of UNION lists has been reduced as well.

  • Fix dumping of rules/views when subsequent addition of a column has resulted in multiple input columns matching a USING specification (Tom Lane)

  • Repair view printing for some cases involving functions in FROM that return a composite type containing dropped columns (Tom Lane)

  • Block signals during postmaster startup (Tom Lane)

    This ensures that the postmaster will properly clean up after itself if, for example, it receives SIGINT while still starting up.

  • Fix client host name lookup when processing pg_hba.conf entries that specify host names instead of IP addresses (Tom Lane)

    Ensure that reverse-DNS lookup failures are reported, instead of just silently not matching such entries. Also ensure that we make only one reverse-DNS lookup attempt per connection, not one per host name entry, which is what previously happened if the lookup attempts failed.

  • Allow the root user to use postgres -C variable and postgres --describe-config (MauMau)

    The prohibition on starting the server as root does not need to extend to these operations, and relaxing it prevents failure of pg_ctl in some scenarios.

  • Secure Unix-domain sockets of temporary postmasters started during make check (Noah Misch)

    Any local user able to access the socket file could connect as the server's bootstrap superuser, then proceed to execute arbitrary code as the operating-system user running the test, as we previously noted in CVE-2014-0067. This change defends against that risk by placing the server's socket in a temporary, mode 0700 subdirectory of /tmp. The hazard remains however on platforms where Unix sockets are not supported, notably Windows, because then the temporary postmaster must accept local TCP connections.

    A useful side effect of this change is to simplify make check testing in builds that override DEFAULT_PGSOCKET_DIR. Popular non-default values like /var/run/postgresql are often not writable by the build user, requiring workarounds that will no longer be necessary.

  • Fix tablespace creation WAL replay to work on Windows (MauMau)

  • Fix detection of socket creation failures on Windows (Bruce Momjian)

  • On Windows, allow new sessions to absorb values of PGC_BACKEND parameters (such as log_connections) from the configuration file (Amit Kapila)

    Previously, if such a parameter were changed in the file post-startup, the change would have no effect.

  • Properly quote executable path names on Windows (Nikhil Deshpande)

    This oversight could cause initdb and pg_upgrade to fail on Windows, if the installation path contained both spaces and @ signs.

  • Fix linking of libpython on macOS (Tom Lane)

    The method we previously used can fail with the Python library supplied by Xcode 5.0 and later.

  • Avoid buffer bloat in libpq when the server consistently sends data faster than the client can absorb it (Shin-ichi Morita, Tom Lane)

    libpq could be coerced into enlarging its input buffer until it runs out of memory (which would be reported misleadingly as “lost synchronization with server”). Under ordinary circumstances it's quite far-fetched that data could be continuously transmitted more quickly than the recv() loop can absorb it, but this has been observed when the client is artificially slowed by scheduler constraints.

  • Ensure that LDAP lookup attempts in libpq time out as intended (Laurenz Albe)

  • Fix ecpg to do the right thing when an array of char * is the target for a FETCH statement returning more than one row, as well as some other array-handling fixes (Ashutosh Bapat)

  • Fix pg_dump to cope with a materialized view that depends on a table's primary key (Tom Lane)

    This occurs if the view's query relies on functional dependency to abbreviate a GROUP BY list. pg_dump got sufficiently confused that it dumped the materialized view as a regular view.

  • Fix parsing of pg_dumpall's -i switch (Tom Lane)

  • Fix pg_restore's processing of old-style large object comments (Tom Lane)

    A direct-to-database restore from an archive file generated by a pre-9.0 version of pg_dump would usually fail if the archive contained more than a few comments for large objects.

  • Fix pg_upgrade for cases where the new server creates a TOAST table but the old version did not (Bruce Momjian)

    This rare situation would manifest as “relation OID mismatch” errors.

  • In pg_upgrade, preserve pg_database.datminmxid and pg_class.relminmxid values from the old cluster, or insert reasonable values when upgrading from pre-9.3; also defend against unreasonable values in the core server (Bruce Momjian, Álvaro Herrera, Tom Lane)

    These changes prevent scenarios in which autovacuum might insist on scanning the entire cluster's contents immediately upon starting the new cluster, or in which tracking of unfrozen MXID values might be disabled completely.

  • Prevent contrib/auto_explain from changing the output of a user's EXPLAIN (Tom Lane)

    If auto_explain is active, it could cause an EXPLAIN (ANALYZE, TIMING OFF) command to nonetheless print timing information.

  • Fix query-lifespan memory leak in contrib/dblink (MauMau, Joe Conway)

  • In contrib/pgcrypto functions, ensure sensitive information is cleared from stack variables before returning (Marko Kreen)

  • Prevent use of already-freed memory in contrib/pgstattuple's pgstat_heap() (Noah Misch)

  • In contrib/uuid-ossp, cache the state of the OSSP UUID library across calls (Tom Lane)

    This improves the efficiency of UUID generation and reduces the amount of entropy drawn from /dev/urandom, on platforms that have that.

  • Update time zone data files to tzdata release 2014e for DST law changes in Crimea, Egypt, and Morocco.

• ⇑ Upgrade to 9.3.6 released on 2015-02-05 - docs

  • Fix buffer overruns in to_char() (Bruce Momjian)

    When to_char() processes a numeric formatting template calling for a large number of digits, PostgreSQL would read past the end of a buffer. When processing a crafted timestamp formatting template, PostgreSQL would write past the end of a buffer. Either case could crash the server. We have not ruled out the possibility of attacks that lead to privilege escalation, though they seem unlikely. (CVE-2015-0241)

  • Fix buffer overrun in replacement *printf() functions (Tom Lane)

    PostgreSQL includes a replacement implementation of printf and related functions. This code will overrun a stack buffer when formatting a floating point number (conversion specifiers e, E, f, F, g or G) with requested precision greater than about 500. This will crash the server, and we have not ruled out the possibility of attacks that lead to privilege escalation. A database user can trigger such a buffer overrun through the to_char() SQL function. While that is the only affected core PostgreSQL functionality, extension modules that use printf-family functions may be at risk as well.

    This issue primarily affects PostgreSQL on Windows. PostgreSQL uses the system implementation of these functions where adequate, which it is on other modern platforms. (CVE-2015-0242)

  • Fix buffer overruns in contrib/pgcrypto (Marko Tiikkaja, Noah Misch)

    Errors in memory size tracking within the pgcrypto module permitted stack buffer overruns and improper dependence on the contents of uninitialized memory. The buffer overrun cases can crash the server, and we have not ruled out the possibility of attacks that lead to privilege escalation. (CVE-2015-0243)

  • Fix possible loss of frontend/backend protocol synchronization after an error (Heikki Linnakangas)

    If any error occurred while the server was in the middle of reading a protocol message from the client, it could lose synchronization and incorrectly try to interpret part of the message's data as a new protocol message. An attacker able to submit crafted binary data within a command parameter might succeed in injecting his own SQL commands this way. Statement timeout and query cancellation are the most likely sources of errors triggering this scenario. Particularly vulnerable are applications that use a timeout and also submit arbitrary user-crafted data as binary query parameters. Disabling statement timeout will reduce, but not eliminate, the risk of exploit. Our thanks to Emil Lenngren for reporting this issue. (CVE-2015-0244)

  • Fix information leak via constraint-violation error messages (Stephen Frost)

    Some server error messages show the values of columns that violate a constraint, such as a unique constraint. If the user does not have SELECT privilege on all columns of the table, this could mean exposing values that the user should not be able to see. Adjust the code so that values are displayed only when they came from the SQL command or could be selected by the user. (CVE-2014-8161)

  • Lock down regression testing's temporary installations on Windows (Noah Misch)

    Use SSPI authentication to allow connections only from the OS user who launched the test suite. This closes on Windows the same vulnerability previously closed on other platforms, namely that other users might be able to connect to the test postmaster. (CVE-2014-0067)

  • Cope with the Windows locale named “Norwegian (Bokmål)” (Heikki Linnakangas)

    Non-ASCII locale names are problematic since it's not clear what encoding they should be represented in. Map the troublesome locale name to a plain-ASCII alias, “Norwegian_Norway”.

  • Avoid possible data corruption if ALTER DATABASE SET TABLESPACE is used to move a database to a new tablespace and then shortly later move it back to its original tablespace (Tom Lane)

  • Avoid corrupting tables when ANALYZE inside a transaction is rolled back (Andres Freund, Tom Lane, Michael Paquier)

    If the failing transaction had earlier removed the last index, rule, or trigger from the table, the table would be left in a corrupted state with the relevant pg_class flags not set though they should be.

  • Ensure that unlogged tables are copied correctly during CREATE DATABASE or ALTER DATABASE SET TABLESPACE (Pavan Deolasee, Andres Freund)

  • Fix incorrect processing of CreateEventTrigStmt.eventname (Petr Jelinek)

    This could result in misbehavior if CREATE EVENT TRIGGER were executed as a prepared query, or via extended query protocol.

  • Fix DROP's dependency searching to correctly handle the case where a table column is recursively visited before its table (Petr Jelinek, Tom Lane)

    This case is only known to arise when an extension creates both a datatype and a table using that datatype. The faulty code might refuse a DROP EXTENSION unless CASCADE is specified, which should not be required.

  • Fix use-of-already-freed-memory problem in EvalPlanQual processing (Tom Lane)

    In READ COMMITTED mode, queries that lock or update recently-updated rows could crash as a result of this bug.

  • Avoid possible deadlock while trying to acquire tuple locks in EvalPlanQual processing (Álvaro Herrera, Mark Kirkwood)

  • Fix failure to wait when a transaction tries to acquire a FOR NO KEY EXCLUSIVE tuple lock, while multiple other transactions currently hold FOR SHARE locks (Álvaro Herrera)

  • Fix planning of SELECT FOR UPDATE when using a partial index on a child table (Kyotaro Horiguchi)

    In READ COMMITTED mode, SELECT FOR UPDATE must also recheck the partial index's WHERE condition when rechecking a recently-updated row to see if it still satisfies the query's WHERE condition. This requirement was missed if the index belonged to an inheritance child table, so that it was possible to incorrectly return rows that no longer satisfy the query condition.

  • Fix corner case wherein SELECT FOR UPDATE could return a row twice, and possibly miss returning other rows (Tom Lane)

    In READ COMMITTED mode, a SELECT FOR UPDATE that is scanning an inheritance tree could incorrectly return a row from a prior child table instead of the one it should return from a later child table.

  • Improve performance of EXPLAIN with large range tables (Tom Lane)

  • Reject duplicate column names in the referenced-columns list of a FOREIGN KEY declaration (David Rowley)

    This restriction is per SQL standard. Previously we did not reject the case explicitly, but later on the code would fail with bizarre-looking errors.

  • Re-enable error for SELECT ... OFFSET -1 (Tom Lane)

    A negative offset value has been an error since 8.4, but an optimization added in 9.3 accidentally turned the case into a no-op. Restore the expected behavior.

  • Restore previous behavior of conversion of domains to JSON (Tom Lane)

    This change causes domains over numeric and boolean to be treated like their base types for purposes of conversion to JSON. It worked like that before 9.3.5 and 9.2.9, but was unintentionally changed while fixing a related problem.

  • Fix json_agg() to not return extra trailing right brackets in its result (Tom Lane)

  • Fix bugs in raising a numeric value to a large integral power (Tom Lane)

    The previous code could get a wrong answer, or consume excessive amounts of time and memory before realizing that the answer must overflow.

  • In numeric_recv(), truncate away any fractional digits that would be hidden according to the value's dscale field (Tom Lane)

    A numeric value's display scale (dscale) should never be less than the number of nonzero fractional digits; but apparently there's at least one broken client application that transmits binary numeric values in which that's true. This leads to strange behavior since the extra digits are taken into account by arithmetic operations even though they aren't printed. The least risky fix seems to be to truncate away such “hidden” digits on receipt, so that the value is indeed what it prints as.

  • Fix incorrect search for shortest-first regular expression matches (Tom Lane)

    Matching would often fail when the number of allowed iterations is limited by a ? quantifier or a bound expression.

  • Reject out-of-range numeric timezone specifications (Tom Lane)

    Simple numeric timezone specifications exceeding +/- 168 hours (one week) would be accepted, but could then cause null-pointer dereference crashes in certain operations. There's no use-case for such large UTC offsets, so reject them.

  • Fix bugs in tsquery @> tsquery operator (Heikki Linnakangas)

    Two different terms would be considered to match if they had the same CRC. Also, if the second operand had more terms than the first, it would be assumed not to be contained in the first; which is wrong since it might contain duplicate terms.

  • Improve ispell dictionary's defenses against bad affix files (Tom Lane)

  • Allow more than 64K phrases in a thesaurus dictionary (David Boutin)

    The previous coding could crash on an oversize dictionary, so this was deemed a back-patchable bug fix rather than a feature addition.

  • Fix namespace handling in xpath() (Ali Akbar)

    Previously, the xml value resulting from an xpath() call would not have namespace declarations if the namespace declarations were attached to an ancestor element in the input xml value, rather than to the specific element being returned. Propagate the ancestral declaration so that the result is correct when considered in isolation.

  • Ensure that whole-row variables expose nonempty column names to functions that pay attention to column names within composite arguments (Tom Lane)

    In some contexts, constructs like row_to_json(tab.*) may not produce the expected column names. This is fixed properly as of 9.4; in older branches, just ensure that we produce some nonempty name. (In some cases this will be the underlying table's column name rather than the query-assigned alias that should theoretically be visible.)

  • Fix mishandling of system columns, particularly tableoid, in FDW queries (Etsuro Fujita)

  • Fix assorted oversights in range-operator selectivity estimation (Emre Hasegeli)

    This patch fixes corner-case “unexpected operator NNNN” planner errors, and improves the selectivity estimates for some other cases.

  • Avoid doing indexed_column = ANY (array) as an index qualifier if that leads to an inferior plan (Andrew Gierth)

    In some cases, = ANY conditions applied to non-first index columns would be done as index conditions even though it would be better to use them as simple filter conditions.

  • Fix “variable not found in subplan target list” planner failure when an inline-able SQL function taking a composite argument is used in a LATERAL subselect and the composite argument is a lateral reference (Tom Lane)

  • Fix planner problems with nested append relations, such as inherited tables within UNION ALL subqueries (Tom Lane)

  • Fail cleanly when a GiST index tuple doesn't fit on a page, rather than going into infinite recursion (Andrew Gierth)

  • Exempt tables that have per-table cost_limit and/or cost_delay settings from autovacuum's global cost balancing rules (Álvaro Herrera)

    The previous behavior resulted in basically ignoring these per-table settings, which was unintended. Now, a table having such settings will be vacuumed using those settings, independently of what is going on in other autovacuum workers. This may result in heavier total I/O load than before, so such settings should be re-examined for sanity.

  • Avoid wholesale autovacuuming when autovacuum is nominally off (Tom Lane)

    Even when autovacuum is nominally off, we will still launch autovacuum worker processes to vacuum tables that are at risk of XID wraparound. However, such a worker process then proceeded to vacuum all tables in the target database, if they met the usual thresholds for autovacuuming. This is at best pretty unexpected; at worst it delays response to the wraparound threat. Fix it so that if autovacuum is turned off, workers only do anti-wraparound vacuums and not any other work.

  • During crash recovery, ensure that unlogged relations are rewritten as empty and are synced to disk before recovery is considered complete (Abhijit Menon-Sen, Andres Freund)

    This prevents scenarios in which unlogged relations might contain garbage data following database crash recovery.

  • Fix race condition between hot standby queries and replaying a full-page image (Heikki Linnakangas)

    This mistake could result in transient errors in queries being executed in hot standby.

  • Fix several cases where recovery logic improperly ignored WAL records for COMMIT/ABORT PREPARED (Heikki Linnakangas)

    The most notable oversight was that recovery_target_xid could not be used to stop at a two-phase commit.

  • Prevent latest WAL file from being archived a second time at completion of crash recovery (Fujii Masao)

  • Avoid creating unnecessary .ready marker files for timeline history files (Fujii Masao)

  • Fix possible null pointer dereference when an empty prepared statement is used and the log_statement setting is mod or ddl (Fujii Masao)

  • Change “pgstat wait timeout” warning message to be LOG level, and rephrase it to be more understandable (Tom Lane)

    This message was originally thought to be essentially a can't-happen case, but it occurs often enough on our slower buildfarm members to be a nuisance. Reduce it to LOG level, and expend a bit more effort on the wording: it now reads “using stale statistics instead of current ones because stats collector is not responding”.

  • Fix possible corruption of postmaster's list of dynamic background workers (Andres Freund)

  • Fix SPARC spinlock implementation to ensure correctness if the CPU is being run in a non-TSO coherency mode, as some non-Solaris kernels do (Andres Freund)

  • Warn if macOS's setlocale() starts an unwanted extra thread inside the postmaster (Noah Misch)

  • Fix processing of repeated dbname parameters in PQconnectdbParams() (Alex Shulgin)

    Unexpected behavior ensued if the first occurrence of dbname contained a connection string or URI to be expanded.

  • Ensure that libpq reports a suitable error message on unexpected socket EOF (Marko Tiikkaja, Tom Lane)

    Depending on kernel behavior, libpq might return an empty error string rather than something useful when the server unexpectedly closed the socket.

  • Clear any old error message during PQreset() (Heikki Linnakangas)

    If PQreset() is called repeatedly, and the connection cannot be re-established, error messages from the failed connection attempts kept accumulating in the PGconn's error string.

  • Properly handle out-of-memory conditions while parsing connection options in libpq (Alex Shulgin, Heikki Linnakangas)

  • Fix array overrun in ecpg's version of ParseDateTime() (Michael Paquier)

  • In initdb, give a clearer error message if a password file is specified but is empty (Mats Erik Andersson)

  • Fix psql's \s command to work nicely with libedit, and add pager support (Stepan Rutz, Tom Lane)

    When using libedit rather than readline, \s printed the command history in a fairly unreadable encoded format, and on recent libedit versions might fail altogether. Fix that by printing the history ourselves rather than having the library do it. A pleasant side-effect is that the pager is used if appropriate.

    This patch also fixes a bug that caused newline encoding to be applied inconsistently when saving the command history with libedit. Multiline history entries written by older psql versions will be read cleanly with this patch, but perhaps not vice versa, depending on the exact libedit versions involved.

  • Improve consistency of parsing of psql's special variables (Tom Lane)

    Allow variant spellings of on and off (such as 1/0) for ECHO_HIDDEN and ON_ERROR_ROLLBACK. Report a warning for unrecognized values for COMP_KEYWORD_CASE, ECHO, ECHO_HIDDEN, HISTCONTROL, ON_ERROR_ROLLBACK, and VERBOSITY. Recognize all values for all these variables case-insensitively; previously there was a mishmash of case-sensitive and case-insensitive behaviors.

  • Make psql's \watch command display nulls as specified by \pset null (Fujii Masao)

  • Fix psql's expanded-mode display to work consistently when using border = 3 and linestyle = ascii or unicode (Stephen Frost)

  • Fix pg_dump to handle comments on event triggers without failing (Tom Lane)

  • Allow parallel pg_dump to use --serializable-deferrable (Kevin Grittner)

  • Improve performance of pg_dump when the database contains many instances of multiple dependency paths between the same two objects (Tom Lane)

  • Fix pg_dumpall to restore its ability to dump from pre-8.1 servers (Gilles Darold)

  • Fix possible deadlock during parallel restore of a schema-only dump (Robert Haas, Tom Lane)

  • Fix core dump in pg_dump --binary-upgrade on zero-column composite type (Rushabh Lathia)

  • Fix failure to fsync tables in nondefault tablespaces during pg_upgrade (Abhijit Menon-Sen, Andres Freund)

    With an operating system crash and some bad luck, this could result in data loss during an upgrade.

  • In pg_upgrade, cope with cases where the new cluster creates a TOAST table for a table that didn't previously have one (Bruce Momjian)

    Previously this could result in failures due to OID conflicts.

  • In pg_upgrade, don't try to set autovacuum_multixact_freeze_max_age for the old cluster (Bruce Momjian)

    This could result in failure because not all 9.3.X versions have that parameter. Fortunately, we don't actually need to set it at all.

  • In pg_upgrade, preserve the transaction ID epoch (Bruce Momjian)

    This oversight did not bother PostgreSQL proper, but could confuse some external replication tools.

  • Prevent WAL files created by pg_basebackup -x/-X from being archived again when the standby is promoted (Andres Freund)

  • Fix memory leak in pg_receivexlog (Fujii Masao)

  • Fix unintended suppression of pg_receivexlog verbose messages (Fujii Masao)

  • Fix failure of contrib/auto_explain to print per-node timing information when doing EXPLAIN ANALYZE (Tom Lane)

  • Fix upgrade-from-unpackaged script for contrib/citext (Tom Lane)

  • Avoid integer overflow and buffer overrun in contrib/hstore's hstore_to_json() (Heikki Linnakangas)

  • Fix recognition of numbers in hstore_to_json_loose(), so that JSON numbers and strings are correctly distinguished (Andrew Dunstan)

  • Fix block number checking in contrib/pageinspect's get_raw_page() (Tom Lane)

    The incorrect checking logic could prevent access to some pages in non-main relation forks.

  • Fix contrib/pgcrypto's pgp_sym_decrypt() to not fail on messages whose length is 6 less than a power of 2 (Marko Tiikkaja)

  • Fix file descriptor leak in contrib/pg_test_fsync (Jeff Janes)

    This could cause failure to remove temporary files on Windows.

  • Handle unexpected query results, especially NULLs, safely in contrib/tablefunc's connectby() (Michael Paquier)

    connectby() previously crashed if it encountered a NULL key value. It now prints that row but doesn't recurse further.

  • Avoid a possible crash in contrib/xml2's xslt_process() (Mark Simonetti)

    libxslt seems to have an undocumented dependency on the order in which resources are freed; reorder our calls to avoid a crash.

  • Mark some contrib I/O functions with correct volatility properties (Tom Lane)

    The previous over-conservative marking was immaterial in normal use, but could cause optimization problems or rejection of valid index expression definitions. Since the consequences are not large, we've just adjusted the function definitions in the extension modules' scripts, without changing version numbers.

  • Numerous cleanups of warnings from Coverity static code analyzer (Andres Freund, Tatsuo Ishii, Marko Kreen, Tom Lane, Michael Paquier)

    These changes are mostly cosmetic but in some cases fix corner-case bugs, for example a crash rather than a proper error report after an out-of-memory failure. None are believed to represent security issues.

  • Fix setup of background workers in EXEC_BACKEND builds, eg Windows (Robert Haas)

  • Detect incompatible OpenLDAP versions during build (Noah Misch)

    With OpenLDAP versions 2.4.24 through 2.4.31, inclusive, PostgreSQL backends can crash at exit. Raise a warning during configure based on the compile-time OpenLDAP version number, and test the crashing scenario in the contrib/dblink regression test.

  • In non-MSVC Windows builds, ensure libpq.dll is installed with execute permissions (Noah Misch)

  • Make pg_regress remove any temporary installation it created upon successful exit (Tom Lane)

    This results in a very substantial reduction in disk space usage during make check-world, since that sequence involves creation of numerous temporary installations.

  • Support time zone abbreviations that change UTC offset from time to time (Tom Lane)

    Previously, PostgreSQL assumed that the UTC offset associated with a time zone abbreviation (such as EST) never changes in the usage of any particular locale. However this assumption fails in the real world, so introduce the ability for a zone abbreviation to represent a UTC offset that sometimes changes. Update the zone abbreviation definition files to make use of this feature in timezone locales that have changed the UTC offset of their abbreviations since 1970 (according to the IANA timezone database). In such timezones, PostgreSQL will now associate the correct UTC offset with the abbreviation depending on the given date.

  • Update time zone abbreviations lists (Tom Lane)

    Add CST (China Standard Time) to our lists. Remove references to ADT as “Arabia Daylight Time”, an abbreviation that's been out of use since 2007; therefore, claiming there is a conflict with “Atlantic Daylight Time” doesn't seem especially helpful. Fix entirely incorrect GMT offsets for CKT (Cook Islands), FJT, and FJST (Fiji); we didn't even have them on the proper side of the date line.

  • Update time zone data files to tzdata release 2015a.

    The IANA timezone database has adopted abbreviations of the form AxST/AxDT for all Australian time zones, reflecting what they believe to be current majority practice Down Under. These names do not conflict with usage elsewhere (other than ACST for Acre Summer Time, which has been in disuse since 1994). Accordingly, adopt these names into our “Default” timezone abbreviation set. The “Australia” abbreviation set now contains only CST, EAST, EST, SAST, SAT, and WST, all of which are thought to be mostly historical usage. Note that SAST has also been changed to be South Africa Standard Time in the “Default” abbreviation set.

    Also, add zone abbreviations SRET (Asia/Srednekolymsk) and XJT (Asia/Urumqi), and use WSST/WSDT for western Samoa. Also, there were DST law changes in Chile, Mexico, the Turks & Caicos Islands (America/Grand_Turk), and Fiji. There is a new zone Pacific/Bougainville for portions of Papua New Guinea. Also, numerous corrections for historical (pre-1970) time zone data.

• ⇑ Upgrade to 9.3.7 released on 2015-05-22 - docs

  • Avoid possible crash when client disconnects just before the authentication timeout expires (Benkocs Norbert Attila)

    If the timeout interrupt fired partway through the session shutdown sequence, SSL-related state would be freed twice, typically causing a crash and hence denial of service to other sessions. Experimentation shows that an unauthenticated remote attacker could trigger the bug somewhat consistently, hence treat as security issue. (CVE-2015-3165)

  • Improve detection of system-call failures (Noah Misch)

    Our replacement implementation of snprintf() failed to check for errors reported by the underlying system library calls; the main case that might be missed is out-of-memory situations. In the worst case this might lead to information exposure, due to our code assuming that a buffer had been overwritten when it hadn't been. Also, there were a few places in which security-relevant calls of other system library functions did not check for failure.

    It remains possible that some calls of the *printf() family of functions are vulnerable to information disclosure if an out-of-memory error occurs at just the wrong time. We judge the risk to not be large, but will continue analysis in this area. (CVE-2015-3166)

  • In contrib/pgcrypto, uniformly report decryption failures as “Wrong key or corrupt data” (Noah Misch)

    Previously, some cases of decryption with an incorrect key could report other error message texts. It has been shown that such variance in error reports can aid attackers in recovering keys from other systems. While it's unknown whether pgcrypto's specific behaviors are likewise exploitable, it seems better to avoid the risk by using a one-size-fits-all message. (CVE-2015-3167)

  • Protect against wraparound of multixact member IDs (Álvaro Herrera, Robert Haas, Thomas Munro)

    Under certain usage patterns, the existing defenses against this might be insufficient, allowing pg_multixact/members files to be removed too early, resulting in data loss. The fix for this includes modifying the server to fail transactions that would result in overwriting old multixact member ID data, and improving autovacuum to ensure it will act proactively to prevent multixact member ID wraparound, as it does for transaction ID wraparound.

  • Fix incorrect declaration of contrib/citext's regexp_matches() functions (Tom Lane)

    These functions should return setof text[], like the core functions they are wrappers for; but they were incorrectly declared as returning just text[]. This mistake had two results: first, if there was no match you got a scalar null result, whereas what you should get is an empty set (zero rows). Second, the g flag was effectively ignored, since you would get only one result array even if there were multiple matches.

    While the latter behavior is clearly a bug, there might be applications depending on the former behavior; therefore the function declarations will not be changed by default until PostgreSQL 9.5. In pre-9.5 branches, the old behavior exists in version 1.0 of the citext extension, while we have provided corrected declarations in version 1.1 (which is not installed by default). To adopt the fix in pre-9.5 branches, execute ALTER EXTENSION citext UPDATE TO '1.1' in each database in which citext is installed. (You can also “update” back to 1.0 if you need to undo that.) Be aware that either update direction will require dropping and recreating any views or rules that use citext's regexp_matches() functions.

  • Fix incorrect checking of deferred exclusion constraints after a HOT update (Tom Lane)

    If a new row that potentially violates a deferred exclusion constraint is HOT-updated (that is, no indexed columns change and the row can be stored back onto the same table page) later in the same transaction, the exclusion constraint would be reported as violated when the check finally occurred, even if the row(s) the new row originally conflicted with had been deleted.

  • Fix planning of star-schema-style queries (Tom Lane)

    Sometimes, efficient scanning of a large table requires that index parameters be provided from more than one other table (commonly, dimension tables whose keys are needed to index a large fact table). The planner should be able to find such plans, but an overly restrictive search heuristic prevented it.

  • Prevent improper reordering of antijoins (NOT EXISTS joins) versus other outer joins (Tom Lane)

    This oversight in the planner has been observed to cause “could not find RelOptInfo for given relids” errors, but it seems possible that sometimes an incorrect query plan might get past that consistency check and result in silently-wrong query output.

  • Fix incorrect matching of subexpressions in outer-join plan nodes (Tom Lane)

    Previously, if textually identical non-strict subexpressions were used both above and below an outer join, the planner might try to re-use the value computed below the join, which would be incorrect because the executor would force the value to NULL in case of an unmatched outer row.

  • Fix GEQO planner to cope with failure of its join order heuristic (Tom Lane)

    This oversight has been seen to lead to “failed to join all relations together” errors in queries involving LATERAL, and that might happen in other cases as well.

  • Fix possible deadlock at startup when max_prepared_transactions is too small (Heikki Linnakangas)

  • Don't archive useless preallocated WAL files after a timeline switch (Heikki Linnakangas)

  • Recursively fsync() the data directory after a crash (Abhijit Menon-Sen, Robert Haas)

    This ensures consistency if another crash occurs shortly later. (The second crash would have to be a system-level crash, not just a database crash, for there to be a problem.)

  • Fix autovacuum launcher's possible failure to shut down, if an error occurs after it receives SIGTERM (Álvaro Herrera)

  • Cope with unexpected signals in LockBufferForCleanup() (Andres Freund)

    This oversight could result in spurious errors about “multiple backends attempting to wait for pincount 1”.

  • Fix crash when doing COPY IN to a table with check constraints that contain whole-row references (Tom Lane)

    The known failure case only crashes in 9.4 and up, but there is very similar code in 9.3 and 9.2, so back-patch those branches as well.

  • Avoid waiting for WAL flush or synchronous replication during commit of a transaction that was read-only so far as the user is concerned (Andres Freund)

    Previously, a delay could occur at commit in transactions that had written WAL due to HOT page pruning, leading to undesirable effects such as sessions getting stuck at startup if all synchronous replicas are down. Sessions have also been observed to get stuck in catchup interrupt processing when using synchronous replication; this will fix that problem as well.

  • Fix crash when manipulating hash indexes on temporary tables (Heikki Linnakangas)

  • Fix possible failure during hash index bucket split, if other processes are modifying the index concurrently (Tom Lane)

  • Check for interrupts while analyzing index expressions (Jeff Janes)

    ANALYZE executes index expressions many times; if there are slow functions in such an expression, it's desirable to be able to cancel the ANALYZE before that loop finishes.

  • Ensure tableoid of a foreign table is reported correctly when a READ COMMITTED recheck occurs after locking rows in SELECT FOR UPDATE, UPDATE, or DELETE (Etsuro Fujita)

  • Add the name of the target server to object description strings for foreign-server user mappings (Álvaro Herrera)

  • Include the schema name in object identity strings for conversions (Álvaro Herrera)

  • Recommend setting include_realm to 1 when using Kerberos/GSSAPI/SSPI authentication (Stephen Frost)

    Without this, identically-named users from different realms cannot be distinguished. For the moment this is only a documentation change, but it will become the default setting in PostgreSQL 9.5.

  • Remove code for matching IPv4 pg_hba.conf entries to IPv4-in-IPv6 addresses (Tom Lane)

    This hack was added in 2003 in response to a report that some Linux kernels of the time would report IPv4 connections as having IPv4-in-IPv6 addresses. However, the logic was accidentally broken in 9.0. The lack of any field complaints since then shows that it's not needed anymore. Now we have reports that the broken code causes crashes on some systems, so let's just remove it rather than fix it. (Had we chosen to fix it, that would make for a subtle and potentially security-sensitive change in the effective meaning of IPv4 pg_hba.conf entries, which does not seem like a good thing to do in minor releases.)

  • Report WAL flush, not insert, position in IDENTIFY_SYSTEM replication command (Heikki Linnakangas)

    This avoids a possible startup failure in pg_receivexlog.

  • While shutting down service on Windows, periodically send status updates to the Service Control Manager to prevent it from killing the service too soon; and ensure that pg_ctl will wait for shutdown (Krystian Bigaj)

  • Reduce risk of network deadlock when using libpq's non-blocking mode (Heikki Linnakangas)

    When sending large volumes of data, it's important to drain the input buffer every so often, in case the server has sent enough response data to cause it to block on output. (A typical scenario is that the server is sending a stream of NOTICE messages during COPY FROM STDIN.) This worked properly in the normal blocking mode, but not so much in non-blocking mode. We've modified libpq to opportunistically drain input when it can, but a full defense against this problem requires application cooperation: the application should watch for socket read-ready as well as write-ready conditions, and be sure to call PQconsumeInput() upon read-ready.

  • In libpq, fix misparsing of empty values in URI connection strings (Thomas Fanghaenel)

  • Fix array handling in ecpg (Michael Meskes)

  • Fix psql to sanely handle URIs and conninfo strings as the first parameter to \connect (David Fetter, Andrew Dunstan, Álvaro Herrera)

    This syntax has been accepted (but undocumented) for a long time, but previously some parameters might be taken from the old connection instead of the given string, which was agreed to be undesirable.

  • Suppress incorrect complaints from psql on some platforms that it failed to write ~/.psql_history at exit (Tom Lane)

    This misbehavior was caused by a workaround for a bug in very old (pre-2006) versions of libedit. We fixed it by removing the workaround, which will cause a similar failure to appear for anyone still using such versions of libedit. Recommendation: upgrade that library, or use libreadline.

  • Fix pg_dump's rule for deciding which casts are system-provided casts that should not be dumped (Tom Lane)

  • In pg_dump, fix failure to honor -Z compression level option together with -Fd (Michael Paquier)

  • Make pg_dump consider foreign key relationships between extension configuration tables while choosing dump order (Gilles Darold, Michael Paquier, Stephen Frost)

    This oversight could result in producing dumps that fail to reload because foreign key constraints are transiently violated.

  • Avoid possible pg_dump failure when concurrent sessions are creating and dropping temporary functions (Tom Lane)

  • Fix dumping of views that are just VALUES(...) but have column aliases (Tom Lane)

  • In pg_upgrade, force timeline 1 in the new cluster (Bruce Momjian)

    This change prevents upgrade failures caused by bogus complaints about missing WAL history files.

  • In pg_upgrade, check for improperly non-connectable databases before proceeding (Bruce Momjian)

  • In pg_upgrade, quote directory paths properly in the generated delete_old_cluster script (Bruce Momjian)

  • In pg_upgrade, preserve database-level freezing info properly (Bruce Momjian)

    This oversight could cause missing-clog-file errors for tables within the postgres and template1 databases.

  • Run pg_upgrade and pg_resetxlog with restricted privileges on Windows, so that they don't fail when run by an administrator (Muhammad Asif Naeem)

  • Improve handling of readdir() failures when scanning directories in initdb and pg_basebackup (Marco Nenciarini)

  • Fix slow sorting algorithm in contrib/intarray (Tom Lane)

  • Fix compile failure on Sparc V8 machines (Rob Rowan)

  • Silence some build warnings on macOS (Tom Lane)

  • Update time zone data files to tzdata release 2015d for DST law changes in Egypt, Mongolia, and Palestine, plus historical changes in Canada and Chile. Also adopt revised zone abbreviations for the America/Adak zone (HST/HDT not HAST/HADT).

• ⇑ Upgrade to 9.3.8 released on 2015-06-04 - docs

  • Avoid failures while fsync'ing data directory during crash restart (Abhijit Menon-Sen, Tom Lane)

    In the previous minor releases we added a patch to fsync everything in the data directory after a crash. Unfortunately its response to any error condition was to fail, thereby preventing the server from starting up, even when the problem was quite harmless. An example is that an unwritable file in the data directory would prevent restart on some platforms; but it is common to make SSL certificate files unwritable by the server. Revise this behavior so that permissions failures are ignored altogether, and other types of failures are logged but do not prevent continuing.

    Also apply the same rules in initdb --sync-only. This case is less critical but it should act similarly.

  • Fix pg_get_functiondef() to show functions' LEAKPROOF property, if set (Jeevan Chalke)

  • Remove configure's check prohibiting linking to a threaded libpython on OpenBSD (Tom Lane)

    The failure this restriction was meant to prevent seems to not be a problem anymore on current OpenBSD versions.

  • Allow libpq to use TLS protocol versions beyond v1 (Noah Misch)

    For a long time, libpq was coded so that the only SSL protocol it would allow was TLS v1. Now that newer TLS versions are becoming popular, allow it to negotiate the highest commonly-supported TLS version with the server. (PostgreSQL servers were already capable of such negotiation, so no change is needed on the server side.) This is a back-patch of a change already released in 9.4.0.

• ⇑ Upgrade to 9.3.9 released on 2015-06-12 - docs

  • Fix possible failure to recover from an inconsistent database state (Robert Haas)

    Recent PostgreSQL releases introduced mechanisms to protect against multixact wraparound, but some of that code did not account for the possibility that it would need to run during crash recovery, when the database may not be in a consistent state. This could result in failure to restart after a crash, or failure to start up a secondary server. The lingering effects of a previously-fixed bug in pg_upgrade could also cause such a failure, in installations that had used pg_upgrade versions between 9.3.0 and 9.3.4.

    The pg_upgrade bug in question was that it would set oldestMultiXid to 1 in pg_control even if the true value should be higher. With the fixes introduced in this release, such a situation will result in immediate emergency autovacuuming until a correct oldestMultiXid value can be determined. If that would pose a hardship, users can avoid it by doing manual vacuuming before upgrading to this release. In detail:

    1. Check whether pg_controldata reports “Latest checkpoint's oldestMultiXid” to be 1. If not, there's nothing to do.

    2. Look in PGDATA/pg_multixact/offsets to see if there's a file named 0000. If there is, there's nothing to do.

    3. Otherwise, for each table that has pg_class.relminmxid equal to 1, VACUUM that table with both vacuum_multixact_freeze_min_age and vacuum_multixact_freeze_table_age set to zero. (You can use the vacuum cost delay parameters described in Section 19.4.4 to reduce the performance consequences for concurrent sessions.) You must use PostgreSQL 9.3.5 or later to perform this step.

  • Fix rare failure to invalidate relation cache init file (Tom Lane)

    With just the wrong timing of concurrent activity, a VACUUM FULL on a system catalog might fail to update the “init file” that's used to avoid cache-loading work for new sessions. This would result in later sessions being unable to access that catalog at all. This is a very ancient bug, but it's so hard to trigger that no reproducible case had been seen until recently.

  • Avoid deadlock between incoming sessions and CREATE/DROP DATABASE (Tom Lane)

    A new session starting in a database that is the target of a DROP DATABASE command, or is the template for a CREATE DATABASE command, could cause the command to wait for five seconds and then fail, even if the new session would have exited before that.

  • Improve planner's cost estimates for semi-joins and anti-joins with inner indexscans (Tom Lane, Tomas Vondra)

    This type of plan is quite cheap when all the join clauses are used as index scan conditions, even if the inner scan would nominally fetch many rows, because the executor will stop after obtaining one row. The planner only partially accounted for that effect, and would therefore overestimate the cost, leading it to possibly choose some other much less efficient plan type.

• ⇑ Upgrade to 9.3.10 released on 2015-10-08 - docs

  • Guard against stack overflows in json parsing (Oskari Saarenmaa)

    If an application constructs PostgreSQL json or jsonb values from arbitrary user input, the application's users can reliably crash the PostgreSQL server, causing momentary denial of service. (CVE-2015-5289)

  • Fix contrib/pgcrypto to detect and report too-short crypt() salts (Josh Kupershmidt)

    Certain invalid salt arguments crashed the server or disclosed a few bytes of server memory. We have not ruled out the viability of attacks that arrange for presence of confidential information in the disclosed bytes, but they seem unlikely. (CVE-2015-5288)

  • Fix subtransaction cleanup after a portal (cursor) belonging to an outer subtransaction fails (Tom Lane, Michael Paquier)

    A function executed in an outer-subtransaction cursor could cause an assertion failure or crash by referencing a relation created within an inner subtransaction.

  • Ensure all relations referred to by an updatable view are properly locked during an update statement (Dean Rasheed)

  • Fix insertion of relations into the relation cache “init file” (Tom Lane)

    An oversight in a patch in the most recent minor releases caused pg_trigger_tgrelid_tgname_index to be omitted from the init file. Subsequent sessions detected this, then deemed the init file to be broken and silently ignored it, resulting in a significant degradation in session startup time. In addition to fixing the bug, install some guards so that any similar future mistake will be more obvious.

  • Avoid O(N^2) behavior when inserting many tuples into a SPI query result (Neil Conway)

  • Improve LISTEN startup time when there are many unread notifications (Matt Newell)

  • Fix performance problem when a session alters large numbers of foreign key constraints (Jan Wieck, Tom Lane)

    This was seen primarily when restoring pg_dump output for databases with many thousands of tables.

  • Disable SSL renegotiation by default (Michael Paquier, Andres Freund)

    While use of SSL renegotiation is a good idea in theory, we have seen too many bugs in practice, both in the underlying OpenSSL library and in our usage of it. Renegotiation will be removed entirely in 9.5 and later. In the older branches, just change the default value of ssl_renegotiation_limit to zero (disabled).

  • Lower the minimum values of the *_freeze_max_age parameters (Andres Freund)

    This is mainly to make tests of related behavior less time-consuming, but it may also be of value for installations with limited disk space.

  • Limit the maximum value of wal_buffers to 2GB to avoid server crashes (Josh Berkus)

  • Avoid logging complaints when a parameter that can only be set at server start appears multiple times in postgresql.conf, and fix counting of line numbers after an include_dir directive (Tom Lane)

  • Fix rare internal overflow in multiplication of numeric values (Dean Rasheed)

  • Guard against hard-to-reach stack overflows involving record types, range types, json, jsonb, tsquery, ltxtquery and query_int (Noah Misch)

  • Fix handling of DOW and DOY in datetime input (Greg Stark)

    These tokens aren't meant to be used in datetime values, but previously they resulted in opaque internal error messages rather than “invalid input syntax”.

  • Add more query-cancel checks to regular expression matching (Tom Lane)

  • Add recursion depth protections to regular expression, SIMILAR TO, and LIKE matching (Tom Lane)

    Suitable search patterns and a low stack depth limit could lead to stack-overrun crashes.

  • Fix potential infinite loop in regular expression execution (Tom Lane)

    A search pattern that can apparently match a zero-length string, but actually doesn't match because of a back reference, could lead to an infinite loop.

  • In regular expression execution, correctly record match data for capturing parentheses within a quantifier even when the match is zero-length (Tom Lane)

  • Fix low-memory failures in regular expression compilation (Andreas Seltenreich)

  • Fix low-probability memory leak during regular expression execution (Tom Lane)

  • Fix rare low-memory failure in lock cleanup during transaction abort (Tom Lane)

  • Fix “unexpected out-of-memory situation during sort” errors when using tuplestores with small work_mem settings (Tom Lane)

  • Fix very-low-probability stack overrun in qsort (Tom Lane)

  • Fix “invalid memory alloc request size” failure in hash joins with large work_mem settings (Tomas Vondra, Tom Lane)

  • Fix assorted planner bugs (Tom Lane)

    These mistakes could lead to incorrect query plans that would give wrong answers, or to assertion failures in assert-enabled builds, or to odd planner errors such as “could not devise a query plan for the given query”, “could not find pathkey item to sort”, “plan should not reference subplan's variable”, or “failed to assign all NestLoopParams to plan nodes”. Thanks are due to Andreas Seltenreich and Piotr Stefaniak for fuzz testing that exposed these problems.

  • Improve planner's performance for UPDATE/DELETE on large inheritance sets (Tom Lane, Dean Rasheed)

  • Ensure standby promotion trigger files are removed at postmaster startup (Michael Paquier, Fujii Masao)

    This prevents unwanted promotion from occurring if these files appear in a database backup that is used to initialize a new standby server.

  • During postmaster shutdown, ensure that per-socket lock files are removed and listen sockets are closed before we remove the postmaster.pid file (Tom Lane)

    This avoids race-condition failures if an external script attempts to start a new postmaster as soon as pg_ctl stop returns.

  • Fix postmaster's handling of a startup-process crash during crash recovery (Tom Lane)

    If, during a crash recovery cycle, the startup process crashes without having restored database consistency, we'd try to launch a new startup process, which typically would just crash again, leading to an infinite loop.

  • Make emergency autovacuuming for multixact wraparound more robust (Andres Freund)

  • Do not print a WARNING when an autovacuum worker is already gone when we attempt to signal it, and reduce log verbosity for such signals (Tom Lane)

  • Prevent autovacuum launcher from sleeping unduly long if the server clock is moved backwards a large amount (Álvaro Herrera)

  • Ensure that cleanup of a GIN index's pending-insertions list is interruptable by cancel requests (Jeff Janes)

  • Allow all-zeroes pages in GIN indexes to be reused (Heikki Linnakangas)

    Such a page might be left behind after a crash.

  • Fix handling of all-zeroes pages in SP-GiST indexes (Heikki Linnakangas)

    VACUUM attempted to recycle such pages, but did so in a way that wasn't crash-safe.

  • Fix off-by-one error that led to otherwise-harmless warnings about “apparent wraparound” in subtrans/multixact truncation (Thomas Munro)

  • Fix misreporting of CONTINUE and MOVE statement types in PL/pgSQL's error context messages (Pavel Stehule, Tom Lane)

  • Fix PL/Perl to handle non-ASCII error message texts correctly (Alex Hunsaker)

  • Fix PL/Python crash when returning the string representation of a record result (Tom Lane)

  • Fix some places in PL/Tcl that neglected to check for failure of malloc() calls (Michael Paquier, Álvaro Herrera)

  • In contrib/isn, fix output of ISBN-13 numbers that begin with 979 (Fabien Coelho)

    EANs beginning with 979 (but not 9790) are considered ISBNs, but they must be printed in the new 13-digit format, not the 10-digit format.

  • Improve contrib/postgres_fdw's handling of collation-related decisions (Tom Lane)

    The main user-visible effect is expected to be that comparisons involving varchar columns will be sent to the remote server for execution in more cases than before.

  • Improve libpq's handling of out-of-memory conditions (Michael Paquier, Heikki Linnakangas)

  • Fix memory leaks and missing out-of-memory checks in ecpg (Michael Paquier)

  • Fix psql's code for locale-aware formatting of numeric output (Tom Lane)

    The formatting code invoked by \pset numericlocale on did the wrong thing for some uncommon cases such as numbers with an exponent but no decimal point. It could also mangle already-localized output from the money data type.

  • Prevent crash in psql's \c command when there is no current connection (Noah Misch)

  • Make pg_dump handle inherited NOT VALID check constraints correctly (Tom Lane)

  • Fix selection of default zlib compression level in pg_dump's directory output format (Andrew Dunstan)

  • Ensure that temporary files created during a pg_dump run with tar-format output are not world-readable (Michael Paquier)

  • Fix pg_dump and pg_upgrade to support cases where the postgres or template1 database is in a non-default tablespace (Marti Raudsepp, Bruce Momjian)

  • Fix pg_dump to handle object privileges sanely when dumping from a server too old to have a particular privilege type (Tom Lane)

    When dumping data types from pre-9.2 servers, and when dumping functions or procedural languages from pre-7.3 servers, pg_dump would produce GRANT/REVOKE commands that revoked the owner's grantable privileges and instead granted all privileges to PUBLIC. Since the privileges involved are just USAGE and EXECUTE, this isn't a security problem, but it's certainly a surprising representation of the older systems' behavior. Fix it to leave the default privilege state alone in these cases.

  • Fix pg_dump to dump shell types (Tom Lane)

    Shell types (that is, not-yet-fully-defined types) aren't useful for much, but nonetheless pg_dump should dump them.

  • Fix assorted minor memory leaks in pg_dump and other client-side programs (Michael Paquier)

  • Fix spinlock assembly code for PPC hardware to be compatible with AIX's native assembler (Tom Lane)

    Building with gcc didn't work if gcc had been configured to use the native assembler, which is becoming more common.

  • On AIX, test the -qlonglong compiler option rather than just assuming it's safe to use (Noah Misch)

  • On AIX, use -Wl,-brtllib link option to allow symbols to be resolved at runtime (Noah Misch)

    Perl relies on this ability in 5.8.0 and later.

  • Avoid use of inline functions when compiling with 32-bit xlc, due to compiler bugs (Noah Misch)

  • Use librt for sched_yield() when necessary, which it is on some Solaris versions (Oskari Saarenmaa)

  • Fix Windows install.bat script to handle target directory names that contain spaces (Heikki Linnakangas)

  • Make the numeric form of the PostgreSQL version number (e.g., 90405) readily available to extension Makefiles, as a variable named VERSION_NUM (Michael Paquier)

  • Update time zone data files to tzdata release 2015g for DST law changes in Cayman Islands, Fiji, Moldova, Morocco, Norfolk Island, North Korea, Turkey, and Uruguay. There is a new zone name America/Fort_Nelson for the Canadian Northern Rockies.

• ⇑ Upgrade to 9.3.11 released on 2016-02-11 - docs

  • Fix infinite loops and buffer-overrun problems in regular expressions (Tom Lane)

    Very large character ranges in bracket expressions could cause infinite loops in some cases, and memory overwrites in other cases. (CVE-2016-0773)

  • Perform an immediate shutdown if the postmaster.pid file is removed (Tom Lane)

    The postmaster now checks every minute or so that postmaster.pid is still there and still contains its own PID. If not, it performs an immediate shutdown, as though it had received SIGQUIT. The main motivation for this change is to ensure that failed buildfarm runs will get cleaned up without manual intervention; but it also serves to limit the bad effects if a DBA forcibly removes postmaster.pid and then starts a new postmaster.

  • In SERIALIZABLE transaction isolation mode, serialization anomalies could be missed due to race conditions during insertions (Kevin Grittner, Thomas Munro)

  • Fix failure to emit appropriate WAL records when doing ALTER TABLE ... SET TABLESPACE for unlogged relations (Michael Paquier, Andres Freund)

    Even though the relation's data is unlogged, the move must be logged or the relation will be inaccessible after a standby is promoted to master.

  • Fix possible misinitialization of unlogged relations at the end of crash recovery (Andres Freund, Michael Paquier)

  • Ensure walsender slots are fully re-initialized when being re-used (Magnus Hagander)

  • Fix ALTER COLUMN TYPE to reconstruct inherited check constraints properly (Tom Lane)

  • Fix REASSIGN OWNED to change ownership of composite types properly (Álvaro Herrera)

  • Fix REASSIGN OWNED and ALTER OWNER to correctly update granted-permissions lists when changing owners of data types, foreign data wrappers, or foreign servers (Bruce Momjian, Álvaro Herrera)

  • Fix REASSIGN OWNED to ignore foreign user mappings, rather than fail (Álvaro Herrera)

  • Fix possible crash after doing query rewrite for an updatable view (Stephen Frost)

  • Fix planner's handling of LATERAL references (Tom Lane)

    This fixes some corner cases that led to “failed to build any N-way joins” or “could not devise a query plan” planner failures.

  • Add more defenses against bad planner cost estimates for GIN index scans when the index's internal statistics are very out-of-date (Tom Lane)

  • Make planner cope with hypothetical GIN indexes suggested by an index advisor plug-in (Julien Rouhaud)

  • Speed up generation of unique table aliases in EXPLAIN and rule dumping, and ensure that generated aliases do not exceed NAMEDATALEN (Tom Lane)

  • Fix dumping of whole-row Vars in ROW() and VALUES() lists (Tom Lane)

  • Fix possible internal overflow in numeric division (Dean Rasheed)

  • Fix enforcement of restrictions inside parentheses within regular expression lookahead constraints (Tom Lane)

    Lookahead constraints aren't allowed to contain backrefs, and parentheses within them are always considered non-capturing, according to the manual. However, the code failed to handle these cases properly inside a parenthesized subexpression, and would give unexpected results.

  • Conversion of regular expressions to indexscan bounds could produce incorrect bounds from regexps containing lookahead constraints (Tom Lane)

  • Fix regular-expression compiler to handle loops of constraint arcs (Tom Lane)

    The code added for CVE-2007-4772 was both incomplete, in that it didn't handle loops involving more than one state, and incorrect, in that it could cause assertion failures (though there seem to be no bad consequences of that in a non-assert build). Multi-state loops would cause the compiler to run until the query was canceled or it reached the too-many-states error condition.

  • Improve memory-usage accounting in regular-expression compiler (Tom Lane)

    This causes the code to emit “regular expression is too complex” errors in some cases that previously used unreasonable amounts of time and memory.

  • Improve performance of regular-expression compiler (Tom Lane)

  • Make %h and %r escapes in log_line_prefix work for messages emitted due to log_connections (Tom Lane)

    Previously, %h/%r started to work just after a new session had emitted the “connection received” log message; now they work for that message too.

  • On Windows, ensure the shared-memory mapping handle gets closed in child processes that don't need it (Tom Lane, Amit Kapila)

    This oversight resulted in failure to recover from crashes whenever logging_collector is turned on.

  • Fix possible failure to detect socket EOF in non-blocking mode on Windows (Tom Lane)

    It's not entirely clear whether this problem can happen in pre-9.5 branches, but if it did, the symptom would be that a walsender process would wait indefinitely rather than noticing a loss of connection.

  • Avoid leaking a token handle during SSPI authentication (Christian Ullrich)

  • In psql, ensure that libreadline's idea of the screen size is updated when the terminal window size changes (Merlin Moncure)

    Previously, libreadline did not notice if the window was resized during query output, leading to strange behavior during later input of multiline queries.

  • Fix psql's \det command to interpret its pattern argument the same way as other \d commands with potentially schema-qualified patterns do (Reece Hart)

  • Avoid possible crash in psql's \c command when previous connection was via Unix socket and command specifies a new hostname and same username (Tom Lane)

  • In pg_ctl start -w, test child process status directly rather than relying on heuristics (Tom Lane, Michael Paquier)

    Previously, pg_ctl relied on an assumption that the new postmaster would always create postmaster.pid within five seconds. But that can fail on heavily-loaded systems, causing pg_ctl to report incorrectly that the postmaster failed to start.

    Except on Windows, this change also means that a pg_ctl start -w done immediately after another such command will now reliably fail, whereas previously it would report success if done within two seconds of the first command.

  • In pg_ctl start -w, don't attempt to use a wildcard listen address to connect to the postmaster (Kondo Yuta)

    On Windows, pg_ctl would fail to detect postmaster startup if listen_addresses is set to 0.0.0.0 or ::, because it would try to use that value verbatim as the address to connect to, which doesn't work. Instead assume that 127.0.0.1 or ::1, respectively, is the right thing to use.

  • In pg_ctl on Windows, check service status to decide where to send output, rather than checking if standard output is a terminal (Michael Paquier)

  • In pg_dump and pg_basebackup, adopt the GNU convention for handling tar-archive members exceeding 8GB (Tom Lane)

    The POSIX standard for tar file format does not allow archive member files to exceed 8GB, but most modern implementations of tar support an extension that fixes that. Adopt this extension so that pg_dump with -Ft no longer fails on tables with more than 8GB of data, and so that pg_basebackup can handle files larger than 8GB. In addition, fix some portability issues that could cause failures for members between 4GB and 8GB on some platforms. Potentially these problems could cause unrecoverable data loss due to unreadable backup files.

  • Fix assorted corner-case bugs in pg_dump's processing of extension member objects (Tom Lane)

  • Make pg_dump mark a view's triggers as needing to be processed after its rule, to prevent possible failure during parallel pg_restore (Tom Lane)

  • Ensure that relation option values are properly quoted in pg_dump (Kouhei Sutou, Tom Lane)

    A reloption value that isn't a simple identifier or number could lead to dump/reload failures due to syntax errors in CREATE statements issued by pg_dump. This is not an issue with any reloption currently supported by core PostgreSQL, but extensions could allow reloptions that cause the problem.

  • Avoid repeated password prompts during parallel pg_dump (Zeus Kronion)

  • Fix pg_upgrade's file-copying code to handle errors properly on Windows (Bruce Momjian)

  • Install guards in pgbench against corner-case overflow conditions during evaluation of script-specified division or modulo operators (Fabien Coelho, Michael Paquier)

  • Fix failure to localize messages emitted by pg_receivexlog and pg_recvlogical (Ioseph Kim)

  • Avoid dump/reload problems when using both plpython2 and plpython3 (Tom Lane)

    In principle, both versions of PL/Python can be used in the same database, though not in the same session (because the two versions of libpython cannot safely be used concurrently). However, pg_restore and pg_upgrade both do things that can fall foul of the same-session restriction. Work around that by changing the timing of the check.

  • Fix PL/Python regression tests to pass with Python 3.5 (Peter Eisentraut)

  • Fix premature clearing of libpq's input buffer when socket EOF is seen (Tom Lane)

    This mistake caused libpq to sometimes not report the backend's final error message before reporting “server closed the connection unexpectedly”.

  • Prevent certain PL/Java parameters from being set by non-superusers (Noah Misch)

    This change mitigates a PL/Java security bug (CVE-2016-0766), which was fixed in PL/Java by marking these parameters as superuser-only. To fix the security hazard for sites that update PostgreSQL more frequently than PL/Java, make the core code aware of them also.

  • Improve libpq's handling of out-of-memory situations (Michael Paquier, Amit Kapila, Heikki Linnakangas)

  • Fix order of arguments in ecpg-generated typedef statements (Michael Meskes)

  • Use %g not %f format in ecpg's PGTYPESnumeric_from_double() (Tom Lane)

  • Fix ecpg-supplied header files to not contain comments continued from a preprocessor directive line onto the next line (Michael Meskes)

    Such a comment is rejected by ecpg. It's not yet clear whether ecpg itself should be changed.

  • Fix hstore_to_json_loose()'s test for whether an hstore value can be converted to a JSON number (Tom Lane)

    Previously this function could be fooled by non-alphanumeric trailing characters, leading to emitting syntactically-invalid JSON.

  • Ensure that contrib/pgcrypto's crypt() function can be interrupted by query cancel (Andreas Karlsson)

  • Accept flex versions later than 2.5.x (Tom Lane, Michael Paquier)

    Now that flex 2.6.0 has been released, the version checks in our build scripts needed to be adjusted.

  • Improve reproducibility of build output by ensuring filenames are given to the linker in a fixed order (Christoph Berg)

    This avoids possible bitwise differences in the produced executable files from one build to the next.

  • Install our missing script where PGXS builds can find it (Jim Nasby)

    This allows sane behavior in a PGXS build done on a machine where build tools such as bison are missing.

  • Ensure that dynloader.h is included in the installed header files in MSVC builds (Bruce Momjian, Michael Paquier)

  • Add variant regression test expected-output file to match behavior of current libxml2 (Tom Lane)

    The fix for libxml2's CVE-2015-7499 causes it not to output error context reports in some cases where it used to do so. This seems to be a bug, but we'll probably have to live with it for some time, so work around it.

  • Update time zone data files to tzdata release 2016a for DST law changes in Cayman Islands, Metlakatla, and Trans-Baikal Territory (Zabaykalsky Krai), plus historical corrections for Pakistan.

• ⇑ Upgrade to 9.3.12 released on 2016-03-31 - docs

  • Fix incorrect handling of NULL index entries in indexed ROW() comparisons (Tom Lane)

    An index search using a row comparison such as ROW(a, b) > ROW('x', 'y') would stop upon reaching a NULL entry in the b column, ignoring the fact that there might be non-NULL b values associated with later values of a.

  • Avoid unlikely data-loss scenarios due to renaming files without adequate fsync() calls before and after (Michael Paquier, Tomas Vondra, Andres Freund)

  • Correctly handle cases where pg_subtrans is close to XID wraparound during server startup (Jeff Janes)

  • Fix corner-case crash due to trying to free localeconv() output strings more than once (Tom Lane)

  • Fix parsing of affix files for ispell dictionaries (Tom Lane)

    The code could go wrong if the affix file contained any characters whose byte length changes during case-folding, for example I in Turkish UTF8 locales.

  • Avoid use of sscanf() to parse ispell dictionary files (Artur Zakirov)

    This dodges a portability problem on FreeBSD-derived platforms (including macOS).

  • Avoid a crash on old Windows versions (before 7SP1/2008R2SP1) with an AVX2-capable CPU and a Postgres build done with Visual Studio 2013 (Christian Ullrich)

    This is a workaround for a bug in Visual Studio 2013's runtime library, which Microsoft have stated they will not fix in that version.

  • Fix psql's tab completion logic to handle multibyte characters properly (Kyotaro Horiguchi, Robert Haas)

  • Fix psql's tab completion for SECURITY LABEL (Tom Lane)

    Pressing TAB after SECURITY LABEL might cause a crash or offering of inappropriate keywords.

  • Make pg_ctl accept a wait timeout from the PGCTLTIMEOUT environment variable, if none is specified on the command line (Noah Misch)

    This eases testing of slower buildfarm members by allowing them to globally specify a longer-than-normal timeout for postmaster startup and shutdown.

  • Fix incorrect test for Windows service status in pg_ctl (Manuel Mathar)

    The previous set of minor releases attempted to fix pg_ctl to properly determine whether to send log messages to Window's Event Log, but got the test backwards.

  • Fix pgbench to correctly handle the combination of -C and -M prepared options (Tom Lane)

  • In pg_upgrade, skip creating a deletion script when the new data directory is inside the old data directory (Bruce Momjian)

    Blind application of the script in such cases would result in loss of the new data directory.

  • In PL/Perl, properly translate empty Postgres arrays into empty Perl arrays (Alex Hunsaker)

  • Make PL/Python cope with function names that aren't valid Python identifiers (Jim Nasby)

  • Fix multiple mistakes in the statistics returned by contrib/pgstattuple's pgstatindex() function (Tom Lane)

  • Remove dependency on psed in MSVC builds, since it's no longer provided by core Perl (Michael Paquier, Andrew Dunstan)

  • Update time zone data files to tzdata release 2016c for DST law changes in Azerbaijan, Chile, Haiti, Palestine, and Russia (Altai, Astrakhan, Kirov, Sakhalin, Ulyanovsk regions), plus historical corrections for Lithuania, Moldova, and Russia (Kaliningrad, Samara, Volgograd).

• ⇑ Upgrade to 9.3.13 released on 2016-05-12 - docs

  • Clear the OpenSSL error queue before OpenSSL calls, rather than assuming it's clear already; and make sure we leave it clear afterwards (Peter Geoghegan, Dave Vitek, Peter Eisentraut)

    This change prevents problems when there are multiple connections using OpenSSL within a single process and not all the code involved follows the same rules for when to clear the error queue. Failures have been reported specifically when a client application uses SSL connections in libpq concurrently with SSL connections using the PHP, Python, or Ruby wrappers for OpenSSL. It's possible for similar problems to arise within the server as well, if an extension module establishes an outgoing SSL connection.

  • Fix “failed to build any N-way joins” planner error with a full join enclosed in the right-hand side of a left join (Tom Lane)

  • Fix incorrect handling of equivalence-class tests in multilevel nestloop plans (Tom Lane)

    Given a three-or-more-way equivalence class of variables, such as X.X = Y.Y = Z.Z, it was possible for the planner to omit some of the tests needed to enforce that all the variables are actually equal, leading to join rows being output that didn't satisfy the WHERE clauses. For various reasons, erroneous plans were seldom selected in practice, so that this bug has gone undetected for a long time.

  • Fix possible misbehavior of TH, th, and Y,YYY format codes in to_timestamp() (Tom Lane)

    These could advance off the end of the input string, causing subsequent format codes to read garbage.

  • Fix dumping of rules and views in which the array argument of a value operator ANY (array) construct is a sub-SELECT (Tom Lane)

  • Make pg_regress use a startup timeout from the PGCTLTIMEOUT environment variable, if that's set (Tom Lane)

    This is for consistency with a behavior recently added to pg_ctl; it eases automated testing on slow machines.

  • Fix pg_upgrade to correctly restore extension membership for operator families containing only one operator class (Tom Lane)

    In such a case, the operator family was restored into the new database, but it was no longer marked as part of the extension. This had no immediate ill effects, but would cause later pg_dump runs to emit output that would cause (harmless) errors on restore.

  • Fix pg_upgrade to not fail when new-cluster TOAST rules differ from old (Tom Lane)

    pg_upgrade had special-case code to handle the situation where the new PostgreSQL version thinks that a table should have a TOAST table while the old version did not. That code was broken, so remove it, and instead do nothing in such cases; there seems no reason to believe that we can't get along fine without a TOAST table if that was okay according to the old version's rules.

  • Back-port 9.4-era memory-barrier code changes into 9.2 and 9.3 (Tom Lane)

    These changes were not originally needed in pre-9.4 branches, but we recently back-patched a fix that expected the barrier code to work properly. Only IA64 (when using icc), HPPA, and Alpha platforms are affected.

  • Reduce the number of SysV semaphores used by a build configured with --disable-spinlocks (Tom Lane)

  • Rename internal function strtoi() to strtoint() to avoid conflict with a NetBSD library function (Thomas Munro)

  • Fix reporting of errors from bind() and listen() system calls on Windows (Tom Lane)

  • Reduce verbosity of compiler output when building with Microsoft Visual Studio (Christian Ullrich)

  • Fix putenv() to work properly with Visual Studio 2013 (Michael Paquier)

  • Avoid possibly-unsafe use of Windows' FormatMessage() function (Christian Ullrich)

    Use the FORMAT_MESSAGE_IGNORE_INSERTS flag where appropriate. No live bug is known to exist here, but it seems like a good idea to be careful.

  • Update time zone data files to tzdata release 2016d for DST law changes in Russia and Venezuela. There are new zone names Europe/Kirov and Asia/Tomsk to reflect the fact that these regions now have different time zone histories from adjacent regions.

• ⇑ Upgrade to 9.3.14 released on 2016-08-11 - docs

  • Fix possible mis-evaluation of nested CASE-WHEN expressions (Heikki Linnakangas, Michael Paquier, Tom Lane)

    A CASE expression appearing within the test value subexpression of another CASE could become confused about whether its own test value was null or not. Also, inlining of a SQL function implementing the equality operator used by a CASE expression could result in passing the wrong test value to functions called within a CASE expression in the SQL function's body. If the test values were of different data types, a crash might result; moreover such situations could be abused to allow disclosure of portions of server memory. (CVE-2016-5423)

  • Fix client programs' handling of special characters in database and role names (Noah Misch, Nathan Bossart, Michael Paquier)

    Numerous places in vacuumdb and other client programs could become confused by database and role names containing double quotes or backslashes. Tighten up quoting rules to make that safe. Also, ensure that when a conninfo string is used as a database name parameter to these programs, it is correctly treated as such throughout.

    Fix handling of paired double quotes in psql's \connect and \password commands to match the documentation.

    Introduce a new -reuse-previous option in psql's \connect command to allow explicit control of whether to re-use connection parameters from a previous connection. (Without this, the choice is based on whether the database name looks like a conninfo string, as before.) This allows secure handling of database names containing special characters in pg_dumpall scripts.

    pg_dumpall now refuses to deal with database and role names containing carriage returns or newlines, as it seems impractical to quote those characters safely on Windows. In future we may reject such names on the server side, but that step has not been taken yet.

    These are considered security fixes because crafted object names containing special characters could have been used to execute commands with superuser privileges the next time a superuser executes pg_dumpall or other routine maintenance operations. (CVE-2016-5424)

  • Fix corner-case misbehaviors for IS NULL/IS NOT NULL applied to nested composite values (Andrew Gierth, Tom Lane)

    The SQL standard specifies that IS NULL should return TRUE for a row of all null values (thus ROW(NULL,NULL) IS NULL yields TRUE), but this is not meant to apply recursively (thus ROW(NULL, ROW(NULL,NULL)) IS NULL yields FALSE). The core executor got this right, but certain planner optimizations treated the test as recursive (thus producing TRUE in both cases), and contrib/postgres_fdw could produce remote queries that misbehaved similarly.

  • Make the inet and cidr data types properly reject IPv6 addresses with too many colon-separated fields (Tom Lane)

  • Prevent crash in close_ps() (the point ## lseg operator) for NaN input coordinates (Tom Lane)

    Make it return NULL instead of crashing.

  • Avoid possible crash in pg_get_expr() when inconsistent values are passed to it (Michael Paquier, Thomas Munro)

  • Fix several one-byte buffer over-reads in to_number() (Peter Eisentraut)

    In several cases the to_number() function would read one more character than it should from the input string. There is a small chance of a crash, if the input happens to be adjacent to the end of memory.

  • Do not run the planner on the query contained in CREATE MATERIALIZED VIEW or CREATE TABLE AS when WITH NO DATA is specified (Michael Paquier, Tom Lane)

    This avoids some unnecessary failure conditions, for example if a stable function invoked by the materialized view depends on a table that doesn't exist yet.

  • Avoid unsafe intermediate state during expensive paths through heap_update() (Masahiko Sawada, Andres Freund)

    Previously, these cases locked the target tuple (by setting its XMAX) but did not WAL-log that action, thus risking data integrity problems if the page were spilled to disk and then a database crash occurred before the tuple update could be completed.

  • Fix hint bit update during WAL replay of row locking operations (Andres Freund)

    The only known consequence of this problem is that row locks held by a prepared, but uncommitted, transaction might fail to be enforced after a crash and restart.

  • Avoid unnecessary “could not serialize access” errors when acquiring FOR KEY SHARE row locks in serializable mode (Álvaro Herrera)

  • Avoid crash in postgres -C when the specified variable has a null string value (Michael Paquier)

  • Ensure that backends see up-to-date statistics for shared catalogs (Tom Lane)

    The statistics collector failed to update the statistics file for shared catalogs after a request from a regular backend. This problem was partially masked because the autovacuum launcher regularly makes requests that did cause such updates; however, it became obvious with autovacuum disabled.

  • Avoid redundant writes of the statistics files when multiple backends request updates close together (Tom Lane, Tomas Vondra)

  • Avoid consuming a transaction ID during VACUUM (Alexander Korotkov)

    Some cases in VACUUM unnecessarily caused an XID to be assigned to the current transaction. Normally this is negligible, but if one is up against the XID wraparound limit, consuming more XIDs during anti-wraparound vacuums is a very bad thing.

  • Avoid canceling hot-standby queries during VACUUM FREEZE (Simon Riggs, Álvaro Herrera)

    VACUUM FREEZE on an otherwise-idle master server could result in unnecessary cancellations of queries on its standby servers.

  • Prevent possible failure when vacuuming multixact IDs in an installation that has been pg_upgrade'd from pre-9.3 (Andrew Gierth, Álvaro Herrera)

    The usual symptom of this bug is errors like “MultiXactId NNN has not been created yet -- apparent wraparound”.

  • When a manual ANALYZE specifies a column list, don't reset the table's changes_since_analyze counter (Tom Lane)

    If we're only analyzing some columns, we should not prevent routine auto-analyze from happening for the other columns.

  • Fix ANALYZE's overestimation of n_distinct for a unique or nearly-unique column with many null entries (Tom Lane)

    The nulls could get counted as though they were themselves distinct values, leading to serious planner misestimates in some types of queries.

  • Prevent autovacuum from starting multiple workers for the same shared catalog (Álvaro Herrera)

    Normally this isn't much of a problem because the vacuum doesn't take long anyway; but in the case of a severely bloated catalog, it could result in all but one worker uselessly waiting instead of doing useful work on other tables.

  • Prevent infinite loop in GiST index build for geometric columns containing NaN component values (Tom Lane)

  • Fix contrib/btree_gin to handle the smallest possible bigint value correctly (Peter Eisentraut)

  • Teach libpq to correctly decode server version from future servers (Peter Eisentraut)

    It's planned to switch to two-part instead of three-part server version numbers for releases after 9.6. Make sure that PQserverVersion() returns the correct value for such cases.

  • Fix ecpg's code for unsigned long long array elements (Michael Meskes)

  • In pg_dump with both -c and -C options, avoid emitting an unwanted CREATE SCHEMA public command (David Johnston, Tom Lane)

  • Improve handling of SIGTERM/control-C in parallel pg_dump and pg_restore (Tom Lane)

    Make sure that the worker processes will exit promptly, and also arrange to send query-cancel requests to the connected backends, in case they are doing something long-running such as a CREATE INDEX.

  • Fix error reporting in parallel pg_dump and pg_restore (Tom Lane)

    Previously, errors reported by pg_dump or pg_restore worker processes might never make it to the user's console, because the messages went through the master process, and there were various deadlock scenarios that would prevent the master process from passing on the messages. Instead, just print everything to stderr. In some cases this will result in duplicate messages (for instance, if all the workers report a server shutdown), but that seems better than no message.

  • Ensure that parallel pg_dump or pg_restore on Windows will shut down properly after an error (Kyotaro Horiguchi)

    Previously, it would report the error, but then just sit until manually stopped by the user.

  • Make pg_dump behave better when built without zlib support (Kyotaro Horiguchi)

    It didn't work right for parallel dumps, and emitted some rather pointless warnings in other cases.

  • Make pg_basebackup accept -Z 0 as specifying no compression (Fujii Masao)

  • Fix makefiles' rule for building AIX shared libraries to be safe for parallel make (Noah Misch)

  • Fix TAP tests and MSVC scripts to work when build directory's path name contains spaces (Michael Paquier, Kyotaro Horiguchi)

  • Be more predictable about reporting “statement timeout” versus “lock timeout” (Tom Lane)

    On heavily loaded machines, the regression tests sometimes failed due to reporting “lock timeout” even though the statement timeout should have occurred first.

  • Make regression tests safe for Danish and Welsh locales (Jeff Janes, Tom Lane)

    Change some test data that triggered the unusual sorting rules of these locales.

  • Update our copy of the timezone code to match IANA's tzcode release 2016c (Tom Lane)

    This is needed to cope with anticipated future changes in the time zone data files. It also fixes some corner-case bugs in coping with unusual time zones.

  • Update time zone data files to tzdata release 2016f for DST law changes in Kemerovo and Novosibirsk, plus historical corrections for Azerbaijan, Belarus, and Morocco.

• ⇑ Upgrade to 9.3.15 released on 2016-10-27 - docs

  • Fix WAL-logging of truncation of relation free space maps and visibility maps (Pavan Deolasee, Heikki Linnakangas)

    It was possible for these files to not be correctly restored during crash recovery, or to be written incorrectly on a standby server. Bogus entries in a free space map could lead to attempts to access pages that have been truncated away from the relation itself, typically producing errors like “could not read block XXX: read only 0 of 8192 bytes”. Checksum failures in the visibility map are also possible, if checksumming is enabled.

    Procedures for determining whether there is a problem and repairing it if so are discussed at https://wiki.postgresql.org/wiki/Free_Space_Map_Problems.

  • Fix SELECT FOR UPDATE/SHARE to correctly lock tuples that have been updated by a subsequently-aborted transaction (Álvaro Herrera)

    In 9.5 and later, the SELECT would sometimes fail to return such tuples at all. A failure has not been proven to occur in earlier releases, but might be possible with concurrent updates.

  • Fix EvalPlanQual rechecks involving CTE scans (Tom Lane)

    The recheck would always see the CTE as returning no rows, typically leading to failure to update rows that were recently updated.

  • Fix improper repetition of previous results from hashed aggregation in a subquery (Andrew Gierth)

    The test to see if we can reuse a previously-computed hash table of the aggregate state values neglected the possibility of an outer query reference appearing in an aggregate argument expression. A change in the value of such a reference should lead to recalculating the hash table, but did not.

  • Fix EXPLAIN to emit valid XML when track_io_timing is on (Markus Winand)

    Previously the XML output-format option produced syntactically invalid tags such as <I/O-Read-Time>. That is now rendered as <I-O-Read-Time>.

  • Suppress printing of zeroes for unmeasured times in EXPLAIN (Maksim Milyutin)

    Certain option combinations resulted in printing zero values for times that actually aren't ever measured in that combination. Our general policy in EXPLAIN is not to print such fields at all, so do that consistently in all cases.

  • Fix timeout length when VACUUM is waiting for exclusive table lock so that it can truncate the table (Simon Riggs)

    The timeout was meant to be 50 milliseconds, but it was actually only 50 microseconds, causing VACUUM to give up on truncation much more easily than intended. Set it to the intended value.

  • Fix bugs in merging inherited CHECK constraints while creating or altering a table (Tom Lane, Amit Langote)

    Allow identical CHECK constraints to be added to a parent and child table in either order. Prevent merging of a valid constraint from the parent table with a NOT VALID constraint on the child. Likewise, prevent merging of a NO INHERIT child constraint with an inherited constraint.

  • Remove artificial restrictions on the values accepted by numeric_in() and numeric_recv() (Tom Lane)

    We allow numeric values up to the limit of the storage format (more than 1e100000), so it seems fairly pointless that numeric_in() rejected scientific-notation exponents above 1000. Likewise, it was silly for numeric_recv() to reject more than 1000 digits in an input value.

  • Avoid very-low-probability data corruption due to testing tuple visibility without holding buffer lock (Thomas Munro, Peter Geoghegan, Tom Lane)

  • Fix file descriptor leakage when truncating a temporary relation of more than 1GB (Andres Freund)

  • Disallow starting a standalone backend with standby_mode turned on (Michael Paquier)

    This can't do anything useful, since there will be no WAL receiver process to fetch more WAL data; and it could result in misbehavior in code that wasn't designed with this situation in mind.

  • Don't try to share SSL contexts across multiple connections in libpq (Heikki Linnakangas)

    This led to assorted corner-case bugs, particularly when trying to use different SSL parameters for different connections.

  • Avoid corner-case memory leak in libpq (Tom Lane)

    The reported problem involved leaking an error report during PQreset(), but there might be related cases.

  • Make ecpg's --help and --version options work consistently with our other executables (Haribabu Kommi)

  • In pg_dump, never dump range constructor functions (Tom Lane)

    This oversight led to pg_upgrade failures with extensions containing range types, due to duplicate creation of the constructor functions.

  • In pg_xlogdump, retry opening new WAL segments when using --follow option (Magnus Hagander)

    This allows for a possible delay in the server's creation of the next segment.

  • Fix pg_xlogdump to cope with a WAL file that begins with a continuation record spanning more than one page (Pavan Deolasee)

  • Fix contrib/intarray/bench/bench.pl to print the results of the EXPLAIN it does when given the -e option (Daniel Gustafsson)

  • Update Windows time zone mapping to recognize some time zone names added in recent Windows versions (Michael Paquier)

  • Prevent failure of obsolete dynamic time zone abbreviations (Tom Lane)

    If a dynamic time zone abbreviation does not match any entry in the referenced time zone, treat it as equivalent to the time zone name. This avoids unexpected failures when IANA removes abbreviations from their time zone database, as they did in tzdata release 2016f and seem likely to do again in the future. The consequences were not limited to not recognizing the individual abbreviation; any mismatch caused the pg_timezone_abbrevs view to fail altogether.

  • Update time zone data files to tzdata release 2016h for DST law changes in Palestine and Turkey, plus historical corrections for Turkey and some regions of Russia. Switch to numeric abbreviations for some time zones in Antarctica, the former Soviet Union, and Sri Lanka.

    The IANA time zone database previously provided textual abbreviations for all time zones, sometimes making up abbreviations that have little or no currency among the local population. They are in process of reversing that policy in favor of using numeric UTC offsets in zones where there is no evidence of real-world use of an English abbreviation. At least for the time being, PostgreSQL will continue to accept such removed abbreviations for timestamp input. But they will not be shown in the pg_timezone_names view nor used for output.

    In this update, AMT is no longer shown as being in use to mean Armenia Time. Therefore, we have changed the Default abbreviation set to interpret it as Amazon Time, thus UTC-4 not UTC+4.

• ⇑ Upgrade to 9.3.16 released on 2017-02-09 - docs

  • Fix a race condition that could cause indexes built with CREATE INDEX CONCURRENTLY to be corrupt (Pavan Deolasee, Tom Lane)

    If CREATE INDEX CONCURRENTLY was used to build an index that depends on a column not previously indexed, then rows updated by transactions that ran concurrently with the CREATE INDEX command could have received incorrect index entries. If you suspect this may have happened, the most reliable solution is to rebuild affected indexes after installing this update.

  • Unconditionally WAL-log creation of the “init fork” for an unlogged table (Michael Paquier)

    Previously, this was skipped when wal_level = minimal, but actually it's necessary even in that case to ensure that the unlogged table is properly reset to empty after a crash.

  • If the stats collector dies during hot standby, restart it (Takayuki Tsunakawa)

  • Ensure that hot standby feedback works correctly when it's enabled at standby server start (Ants Aasma, Craig Ringer)

  • Check for interrupts while hot standby is waiting for a conflicting query (Simon Riggs)

  • Avoid constantly respawning the autovacuum launcher in a corner case (Amit Khandekar)

    This fix avoids problems when autovacuum is nominally off and there are some tables that require freezing, but all such tables are already being processed by autovacuum workers.

  • Fix check for when an extension member object can be dropped (Tom Lane)

    Extension upgrade scripts should be able to drop member objects, but this was disallowed for serial-column sequences, and possibly other cases.

  • Make sure ALTER TABLE preserves index tablespace assignments when rebuilding indexes (Tom Lane, Michael Paquier)

    Previously, non-default settings of default_tablespace could result in broken indexes.

  • Prevent dropping a foreign-key constraint if there are pending trigger events for the referenced relation (Tom Lane)

    This avoids “could not find trigger NNN” or “relation NNN has no triggers” errors.

  • Fix processing of OID column when a table with OIDs is associated to a parent with OIDs via ALTER TABLE ... INHERIT (Amit Langote)

    The OID column should be treated the same as regular user columns in this case, but it wasn't, leading to odd behavior in later inheritance changes.

  • Report correct object identity during ALTER TEXT SEARCH CONFIGURATION (Artur Zakirov)

    The wrong catalog OID was reported to extensions such as logical decoding.

  • Check for serializability conflicts before reporting constraint-violation failures (Thomas Munro)

    When using serializable transaction isolation, it is desirable that any error due to concurrent transactions should manifest as a serialization failure, thereby cueing the application that a retry might succeed. Unfortunately, this does not reliably happen for duplicate-key failures caused by concurrent insertions. This change ensures that such an error will be reported as a serialization error if the application explicitly checked for the presence of a conflicting key (and did not find it) earlier in the transaction.

  • Prevent multicolumn expansion of foo.* in an UPDATE source expression (Tom Lane)

    This led to “UPDATE target count mismatch --- internal error”. Now the syntax is understood as a whole-row variable, as it would be in other contexts.

  • Ensure that column typmods are determined accurately for multi-row VALUES constructs (Tom Lane)

    This fixes problems occurring when the first value in a column has a determinable typmod (e.g., length for a varchar value) but later values don't share the same limit.

  • Throw error for an unfinished Unicode surrogate pair at the end of a Unicode string (Tom Lane)

    Normally, a Unicode surrogate leading character must be followed by a Unicode surrogate trailing character, but the check for this was missed if the leading character was the last character in a Unicode string literal (U&'...') or Unicode identifier (U&"...").

  • Ensure that a purely negative text search query, such as !foo, matches empty tsvectors (Tom Dunstan)

    Such matches were found by GIN index searches, but not by sequential scans or GiST index searches.

  • Prevent crash when ts_rewrite() replaces a non-top-level subtree with an empty query (Artur Zakirov)

  • Fix performance problems in ts_rewrite() (Tom Lane)

  • Fix ts_rewrite()'s handling of nested NOT operators (Tom Lane)

  • Fix array_fill() to handle empty arrays properly (Tom Lane)

  • Fix one-byte buffer overrun in quote_literal_cstr() (Heikki Linnakangas)

    The overrun occurred only if the input consisted entirely of single quotes and/or backslashes.

  • Prevent multiple calls of pg_start_backup() and pg_stop_backup() from running concurrently (Michael Paquier)

    This avoids an assertion failure, and possibly worse things, if someone tries to run these functions in parallel.

  • Avoid discarding interval-to-interval casts that aren't really no-ops (Tom Lane)

    In some cases, a cast that should result in zeroing out low-order interval fields was mistakenly deemed to be a no-op and discarded. An example is that casting from INTERVAL MONTH to INTERVAL YEAR failed to clear the months field.

  • Ensure that cached plans are invalidated by changes in foreign-table options (Amit Langote, Etsuro Fujita, Ashutosh Bapat)

  • Fix pg_dump to dump user-defined casts and transforms that use built-in functions (Stephen Frost)

  • Fix possible pg_basebackup failure on standby server when including WAL files (Amit Kapila, Robert Haas)

  • Ensure that the Python exception objects we create for PL/Python are properly reference-counted (Rafa de la Torre, Tom Lane)

    This avoids failures if the objects are used after a Python garbage collection cycle has occurred.

  • Fix PL/Tcl to support triggers on tables that have .tupno as a column name (Tom Lane)

    This matches the (previously undocumented) behavior of PL/Tcl's spi_exec and spi_execp commands, namely that a magic .tupno column is inserted only if there isn't a real column named that.

  • Allow DOS-style line endings in ~/.pgpass files, even on Unix (Vik Fearing)

    This change simplifies use of the same password file across Unix and Windows machines.

  • Fix one-byte buffer overrun if ecpg is given a file name that ends with a dot (Takayuki Tsunakawa)

  • Fix psql's tab completion for ALTER DEFAULT PRIVILEGES (Gilles Darold, Stephen Frost)

  • In psql, treat an empty or all-blank setting of the PAGER environment variable as meaning “no pager” (Tom Lane)

    Previously, such a setting caused output intended for the pager to vanish entirely.

  • Improve contrib/dblink's reporting of low-level libpq errors, such as out-of-memory (Joe Conway)

  • Teach contrib/dblink to ignore irrelevant server options when it uses a contrib/postgres_fdw foreign server as the source of connection options (Corey Huinker)

    Previously, if the foreign server object had options that were not also libpq connection options, an error occurred.

  • On Windows, ensure that environment variable changes are propagated to DLLs built with debug options (Christian Ullrich)

  • Sync our copy of the timezone library with IANA release tzcode2016j (Tom Lane)

    This fixes various issues, most notably that timezone data installation failed if the target directory didn't support hard links.

  • Update time zone data files to tzdata release 2016j for DST law changes in northern Cyprus (adding a new zone Asia/Famagusta), Russia (adding a new zone Europe/Saratov), Tonga, and Antarctica/Casey. Historical corrections for Italy, Kazakhstan, Malta, and Palestine. Switch to preferring numeric zone abbreviations for Tonga.

• ⇑ Upgrade to 9.3.17 released on 2017-05-11 - docs

  • Restrict visibility of pg_user_mappings.umoptions, to protect passwords stored as user mapping options (Michael Paquier, Feike Steenbergen)

    The previous coding allowed the owner of a foreign server object, or anyone he has granted server USAGE permission to, to see the options for all user mappings associated with that server. This might well include passwords for other users. Adjust the view definition to match the behavior of information_schema.user_mapping_options, namely that these options are visible to the user being mapped, or if the mapping is for PUBLIC and the current user is the server owner, or if the current user is a superuser. (CVE-2017-7486)

    By itself, this patch will only fix the behavior in newly initdb'd databases. If you wish to apply this change in an existing database, follow the corrected procedure shown in the changelog entry for CVE-2017-7547, in Section E.44.

  • Prevent exposure of statistical information via leaky operators (Peter Eisentraut)

    Some selectivity estimation functions in the planner will apply user-defined operators to values obtained from pg_statistic, such as most common values and histogram entries. This occurs before table permissions are checked, so a nefarious user could exploit the behavior to obtain these values for table columns he does not have permission to read. To fix, fall back to a default estimate if the operator's implementation function is not certified leak-proof and the calling user does not have permission to read the table column whose statistics are needed. At least one of these criteria is satisfied in most cases in practice. (CVE-2017-7484)

  • Restore libpq's recognition of the PGREQUIRESSL environment variable (Daniel Gustafsson)

    Processing of this environment variable was unintentionally dropped in PostgreSQL 9.3, but its documentation remained. This creates a security hazard, since users might be relying on the environment variable to force SSL-encrypted connections, but that would no longer be guaranteed. Restore handling of the variable, but give it lower priority than PGSSLMODE, to avoid breaking configurations that work correctly with post-9.3 code. (CVE-2017-7485)

  • Fix possible corruption of “init forks” of unlogged indexes (Robert Haas, Michael Paquier)

    This could result in an unlogged index being set to an invalid state after a crash and restart. Such a problem would persist until the index was dropped and rebuilt.

  • Fix incorrect reconstruction of pg_subtrans entries when a standby server replays a prepared but uncommitted two-phase transaction (Tom Lane)

    In most cases this turned out to have no visible ill effects, but in corner cases it could result in circular references in pg_subtrans, potentially causing infinite loops in queries that examine rows modified by the two-phase transaction.

  • Ensure parsing of queries in extension scripts sees the results of immediately-preceding DDL (Julien Rouhaud, Tom Lane)

    Due to lack of a cache flush step between commands in an extension script file, non-utility queries might not see the effects of an immediately preceding catalog change, such as ALTER TABLE ... RENAME.

  • Skip tablespace privilege checks when ALTER TABLE ... ALTER COLUMN TYPE rebuilds an existing index (Noah Misch)

    The command failed if the calling user did not currently have CREATE privilege for the tablespace containing the index. That behavior seems unhelpful, so skip the check, allowing the index to be rebuilt where it is.

  • Fix ALTER TABLE ... VALIDATE CONSTRAINT to not recurse to child tables when the constraint is marked NO INHERIT (Amit Langote)

    This fix prevents unwanted “constraint does not exist” failures when no matching constraint is present in the child tables.

  • Fix VACUUM to account properly for pages that could not be scanned due to conflicting page pins (Andrew Gierth)

    This tended to lead to underestimation of the number of tuples in the table. In the worst case of a small heavily-contended table, VACUUM could incorrectly report that the table contained no tuples, leading to very bad planning choices.

  • Ensure that bulk-tuple-transfer loops within a hash join are interruptible by query cancel requests (Tom Lane, Thomas Munro)

  • Fix cursor_to_xml() to produce valid output with tableforest = false (Thomas Munro, Peter Eisentraut)

    Previously it failed to produce a wrapping <table> element.

  • Improve performance of pg_timezone_names view (Tom Lane, David Rowley)

  • Fix sloppy handling of corner-case errors from lseek() and close() (Tom Lane)

    Neither of these system calls are likely to fail in typical situations, but if they did, fd.c could get quite confused.

  • Fix incorrect check for whether postmaster is running as a Windows service (Michael Paquier)

    This could result in attempting to write to the event log when that isn't accessible, so that no logging happens at all.

  • Fix ecpg to support COMMIT PREPARED and ROLLBACK PREPARED (Masahiko Sawada)

  • Fix a double-free error when processing dollar-quoted string literals in ecpg (Michael Meskes)

  • In pg_dump, fix incorrect schema and owner marking for comments and security labels of some types of database objects (Giuseppe Broccolo, Tom Lane)

    In simple cases this caused no ill effects; but for example, a schema-selective restore might omit comments it should include, because they were not marked as belonging to the schema of their associated object.

  • Avoid emitting an invalid list file in pg_restore -l when SQL object names contain newlines (Tom Lane)

    Replace newlines by spaces, which is sufficient to make the output valid for pg_restore -L's purposes.

  • Fix pg_upgrade to transfer comments and security labels attached to “large objects” (blobs) (Stephen Frost)

    Previously, blobs were correctly transferred to the new database, but any comments or security labels attached to them were lost.

  • Improve error handling in contrib/adminpack's pg_file_write() function (Noah Misch)

    Notably, it failed to detect errors reported by fclose().

  • In contrib/dblink, avoid leaking the previous unnamed connection when establishing a new unnamed connection (Joe Conway)

  • Fix contrib/pg_trgm's extraction of trigrams from regular expressions (Tom Lane)

    In some cases it would produce a broken data structure that could never match anything, leading to GIN or GiST indexscans that use a trigram index not finding any matches to the regular expression.

  • In contrib/postgres_fdw, transmit query cancellation requests to the remote server (Michael Paquier, Etsuro Fujita)

    Previously, a local query cancellation request did not cause an already-sent remote query to terminate early. This is a back-patch of work originally done for 9.6.

  • Support OpenSSL 1.1.0 (Heikki Linnakangas, Andreas Karlsson, Tom Lane)

    This is a back-patch of work previously done in newer branches; it's needed since many platforms are adopting newer OpenSSL versions.

  • Support Tcl 8.6 in MSVC builds (Álvaro Herrera)

  • Sync our copy of the timezone library with IANA release tzcode2017b (Tom Lane)

    This fixes a bug affecting some DST transitions in January 2038.

  • Update time zone data files to tzdata release 2017b for DST law changes in Chile, Haiti, and Mongolia, plus historical corrections for Ecuador, Kazakhstan, Liberia, and Spain. Switch to numeric abbreviations for numerous time zones in South America, the Pacific and Indian oceans, and some Asian and Middle Eastern countries.

    The IANA time zone database previously provided textual abbreviations for all time zones, sometimes making up abbreviations that have little or no currency among the local population. They are in process of reversing that policy in favor of using numeric UTC offsets in zones where there is no evidence of real-world use of an English abbreviation. At least for the time being, PostgreSQL will continue to accept such removed abbreviations for timestamp input. But they will not be shown in the pg_timezone_names view nor used for output.

  • Use correct daylight-savings rules for POSIX-style time zone names in MSVC builds (David Rowley)

    The Microsoft MSVC build scripts neglected to install the posixrules file in the timezone directory tree. This resulted in the timezone code falling back to its built-in rule about what DST behavior to assume for a POSIX-style time zone name. For historical reasons that still corresponds to the DST rules the USA was using before 2007 (i.e., change on first Sunday in April and last Sunday in October). With this fix, a POSIX-style zone name will use the current and historical DST transition dates of the US/Eastern zone. If you don't want that, remove the posixrules file, or replace it with a copy of some other zone file (see Section 8.5.3). Note that due to caching, you may need to restart the server to get such changes to take effect.

• ⇑ Upgrade to 9.3.18 released on 2017-08-10 - docs

  • Further restrict visibility of pg_user_mappings.umoptions, to protect passwords stored as user mapping options (Noah Misch)

    The fix for CVE-2017-7486 was incorrect: it allowed a user to see the options in her own user mapping, even if she did not have USAGE permission on the associated foreign server. Such options might include a password that had been provided by the server owner rather than the user herself. Since information_schema.user_mapping_options does not show the options in such cases, pg_user_mappings should not either. (CVE-2017-7547)

    By itself, this patch will only fix the behavior in newly initdb'd databases. If you wish to apply this change in an existing database, you will need to do the following:

    1. Restart the postmaster after adding allow_system_table_mods = true to postgresql.conf. (In versions supporting ALTER SYSTEM, you can use that to make the configuration change, but you'll still need a restart.)

    2. In each database of the cluster, run the following commands as superuser:

       SET search_path = pg_catalog;
       CREATE OR REPLACE VIEW pg_user_mappings AS
           SELECT
               U.oid       AS umid,
               S.oid       AS srvid,
               S.srvname   AS srvname,
               U.umuser    AS umuser,
               CASE WHEN U.umuser = 0 THEN
                   'public'
               ELSE
                   A.rolname
               END AS usename,
               CASE WHEN (U.umuser <> 0 AND A.rolname = current_user
                            AND (pg_has_role(S.srvowner, 'USAGE')
                                 OR has_server_privilege(S.oid, 'USAGE')))
                           OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
                           OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
                           THEN U.umoptions
                        ELSE NULL END AS umoptions
           FROM pg_user_mapping U
                LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
               pg_foreign_server S ON (U.umserver = S.oid);

    3. Do not forget to include the template0 and template1 databases, or the vulnerability will still exist in databases you create later. To fix template0, you'll need to temporarily make it accept connections. In PostgreSQL 9.5 and later, you can use

       ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;

       and then after fixing template0, undo that with

       ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;

       In prior versions, instead use

       UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
       UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';

    4. Finally, remove the allow_system_table_mods configuration setting, and again restart the postmaster.

  • Disallow empty passwords in all password-based authentication methods (Heikki Linnakangas)

    libpq ignores empty password specifications, and does not transmit them to the server. So, if a user's password has been set to the empty string, it's impossible to log in with that password via psql or other libpq-based clients. An administrator might therefore believe that setting the password to empty is equivalent to disabling password login. However, with a modified or non-libpq-based client, logging in could be possible, depending on which authentication method is configured. In particular the most common method, md5, accepted empty passwords. Change the server to reject empty passwords in all cases. (CVE-2017-7546)

  • Fix concurrent locking of tuple update chains (Álvaro Herrera)

    If several sessions concurrently lock a tuple update chain with nonconflicting lock modes using an old snapshot, and they all succeed, it was possible for some of them to nonetheless fail (and conclude there is no live tuple version) due to a race condition. This had consequences such as foreign-key checks failing to see a tuple that definitely exists but is being updated concurrently.

  • Fix potential data corruption when freezing a tuple whose XMAX is a multixact with exactly one still-interesting member (Teodor Sigaev)

  • On Windows, retry process creation if we fail to reserve the address range for our shared memory in the new process (Tom Lane, Amit Kapila)

    This is expected to fix infrequent child-process-launch failures that are probably due to interference from antivirus products.

  • Fix low-probability corruption of shared predicate-lock hash table in Windows builds (Thomas Munro, Tom Lane)

  • Avoid logging clean closure of an SSL connection as though it were a connection reset (Michael Paquier)

  • Prevent sending SSL session tickets to clients (Tom Lane)

    This fix prevents reconnection failures with ticket-aware client-side SSL code.

  • Fix code for setting tcp_keepalives_idle on Solaris (Tom Lane)

  • Fix statistics collector to honor inquiry messages issued just after a postmaster shutdown and immediate restart (Tom Lane)

    Statistics inquiries issued within half a second of the previous postmaster shutdown were effectively ignored.

  • Ensure that the statistics collector's receive buffer size is at least 100KB (Tom Lane)

    This reduces the risk of dropped statistics data on older platforms whose default receive buffer size is less than that.

  • Fix possible creation of an invalid WAL segment when a standby is promoted just after it processes an XLOG_SWITCH WAL record (Andres Freund)

  • Fix SIGHUP and SIGUSR1 handling in walsender processes (Petr Jelinek, Andres Freund)

  • Fix unnecessarily slow restarts of walreceiver processes due to race condition in postmaster (Tom Lane)

  • Fix cases where an INSERT or UPDATE assigns to more than one element of a column that is of domain-over-array type (Tom Lane)

  • Allow window functions to be used in sub-SELECTs that are within the arguments of an aggregate function (Tom Lane)

  • Move autogenerated array types out of the way during ALTER ... RENAME (Vik Fearing)

    Previously, we would rename a conflicting autogenerated array type out of the way during CREATE; this fix extends that behavior to renaming operations.

  • Ensure that ALTER USER ... SET accepts all the syntax variants that ALTER ROLE ... SET does (Peter Eisentraut)

  • Properly update dependency info when changing a datatype I/O function's argument or return type from opaque to the correct type (Heikki Linnakangas)

    CREATE TYPE updates I/O functions declared in this long-obsolete style, but it forgot to record a dependency on the type, allowing a subsequent DROP TYPE to leave broken function definitions behind.

  • Reduce memory usage when ANALYZE processes a tsvector column (Heikki Linnakangas)

  • Fix unnecessary precision loss and sloppy rounding when multiplying or dividing money values by integers or floats (Tom Lane)

  • Tighten checks for whitespace in functions that parse identifiers, such as regprocedurein() (Tom Lane)

    Depending on the prevailing locale, these functions could misinterpret fragments of multibyte characters as whitespace.

  • Use relevant #define symbols from Perl while compiling PL/Perl (Ashutosh Sharma, Tom Lane)

    This avoids portability problems, typically manifesting as a “handshake” mismatch during library load, when working with recent Perl versions.

  • In libpq, reset GSS/SASL and SSPI authentication state properly after a failed connection attempt (Michael Paquier)

    Failure to do this meant that when falling back from SSL to non-SSL connections, a GSS/SASL failure in the SSL attempt would always cause the non-SSL attempt to fail. SSPI did not fail, but it leaked memory.

  • In psql, fix failure when COPY FROM STDIN is ended with a keyboard EOF signal and then another COPY FROM STDIN is attempted (Thomas Munro)

    This misbehavior was observed on BSD-derived platforms (including macOS), but not on most others.

  • Fix pg_dump and pg_restore to emit REFRESH MATERIALIZED VIEW commands last (Tom Lane)

    This prevents errors during dump/restore when a materialized view refers to tables owned by a different user.

  • Fix pg_dump with the --clean option to drop event triggers as expected (Tom Lane)

    It also now correctly assigns ownership of event triggers; before, they were restored as being owned by the superuser running the restore script.

  • Fix pg_dump to not emit invalid SQL for an empty operator class (Daniel Gustafsson)

  • Fix pg_dump output to stdout on Windows (Kuntal Ghosh)

    A compressed plain-text dump written to stdout would contain corrupt data due to failure to put the file descriptor into binary mode.

  • Fix pg_get_ruledef() to print correct output for the ON SELECT rule of a view whose columns have been renamed (Tom Lane)

    In some corner cases, pg_dump relies on pg_get_ruledef() to dump views, so that this error could result in dump/reload failures.

  • Fix dumping of outer joins with empty constraints, such as the result of a NATURAL LEFT JOIN with no common columns (Tom Lane)

  • Fix dumping of function expressions in the FROM clause in cases where the expression does not deparse into something that looks like a function call (Tom Lane)

  • Fix pg_basebackup output to stdout on Windows (Haribabu Kommi)

    A backup written to stdout would contain corrupt data due to failure to put the file descriptor into binary mode.

  • Fix pg_upgrade to ensure that the ending WAL record does not have wal_level = minimum (Bruce Momjian)

    This condition could prevent upgraded standby servers from reconnecting.

  • In postgres_fdw, re-establish connections to remote servers after ALTER SERVER or ALTER USER MAPPING commands (Kyotaro Horiguchi)

    This ensures that option changes affecting connection parameters will be applied promptly.

  • In postgres_fdw, allow cancellation of remote transaction control commands (Robert Haas, Rafia Sabih)

    This change allows us to quickly escape a wait for an unresponsive remote server in many more cases than previously.

  • Always use -fPIC, not -fpic, when building shared libraries with gcc (Tom Lane)

    This supports larger extension libraries on platforms where it makes a difference.

  • Fix unescaped-braces issue in our build scripts for Microsoft MSVC, to avoid a warning or error from recent Perl versions (Andrew Dunstan)

  • In MSVC builds, handle the case where the openssl library is not within a VC subdirectory (Andrew Dunstan)

  • In MSVC builds, add proper include path for libxml2 header files (Andrew Dunstan)

    This fixes a former need to move things around in standard Windows installations of libxml2.

  • In MSVC builds, recognize a Tcl library that is named tcl86.lib (Noah Misch)

• ⇑ Upgrade to 9.3.19 released on 2017-08-31 - docs

  • Show foreign tables in information_schema.table_privileges view (Peter Eisentraut)

    All other relevant information_schema views include foreign tables, but this one ignored them.

    Since this view definition is installed by initdb, merely upgrading will not fix the problem. If you need to fix this in an existing installation, you can, as a superuser, do this in psql:

    SET search_path TO information_schema;
    CREATE OR REPLACE VIEW table_privileges AS
        SELECT CAST(u_grantor.rolname AS sql_identifier) AS grantor,
               CAST(grantee.rolname AS sql_identifier) AS grantee,
               CAST(current_database() AS sql_identifier) AS table_catalog,
               CAST(nc.nspname AS sql_identifier) AS table_schema,
               CAST(c.relname AS sql_identifier) AS table_name,
               CAST(c.prtype AS character_data) AS privilege_type,
               CAST(
                 CASE WHEN
                      -- object owner always has grant options
                      pg_has_role(grantee.oid, c.relowner, 'USAGE')
                      OR c.grantable
                      THEN 'YES' ELSE 'NO' END AS yes_or_no) AS is_grantable,
               CAST(CASE WHEN c.prtype = 'SELECT' THEN 'YES' ELSE 'NO' END AS yes_or_no) AS with_hierarchy
    
        FROM (
                SELECT oid, relname, relnamespace, relkind, relowner, (aclexplode(coalesce(relacl, acldefault('r', relowner)))).* FROM pg_class
             ) AS c (oid, relname, relnamespace, relkind, relowner, grantor, grantee, prtype, grantable),
             pg_namespace nc,
             pg_authid u_grantor,
             (
               SELECT oid, rolname FROM pg_authid
               UNION ALL
               SELECT 0::oid, 'PUBLIC'
             ) AS grantee (oid, rolname)
    
        WHERE c.relnamespace = nc.oid
              AND c.relkind IN ('r', 'v', 'f')
              AND c.grantee = grantee.oid
              AND c.grantor = u_grantor.oid
              AND c.prtype IN ('INSERT', 'SELECT', 'UPDATE', 'DELETE', 'TRUNCATE', 'REFERENCES', 'TRIGGER')
              AND (pg_has_role(u_grantor.oid, 'USAGE')
                   OR pg_has_role(grantee.oid, 'USAGE')
                   OR grantee.rolname = 'PUBLIC');

    This must be repeated in each database to be fixed, including template0.

  • Clean up handling of a fatal exit (e.g., due to receipt of SIGTERM) that occurs while trying to execute a ROLLBACK of a failed transaction (Tom Lane)

    This situation could result in an assertion failure. In production builds, the exit would still occur, but it would log an unexpected message about “cannot drop active portal”.

  • Remove assertion that could trigger during a fatal exit (Tom Lane)

  • Correctly identify columns that are of a range type or domain type over a composite type or domain type being searched for (Tom Lane)

    Certain ALTER commands that change the definition of a composite type or domain type are supposed to fail if there are any stored values of that type in the database, because they lack the infrastructure needed to update or check such values. Previously, these checks could miss relevant values that are wrapped inside range types or sub-domains, possibly allowing the database to become inconsistent.

  • Fix crash in pg_restore when using parallel mode and using a list file to select a subset of items to restore (Fabrízio de Royes Mello)

  • Change ecpg's parser to allow RETURNING clauses without attached C variables (Michael Meskes)

    This allows ecpg programs to contain SQL constructs that use RETURNING internally (for example, inside a CTE) rather than using it to define values to be returned to the client.

  • Improve selection of compiler flags for PL/Perl on Windows (Tom Lane)

    This fix avoids possible crashes of PL/Perl due to inconsistent assumptions about the width of time_t values. A side-effect that may be visible to extension developers is that _USE_32BIT_TIME_T is no longer defined globally in PostgreSQL Windows builds. This is not expected to cause problems, because type time_t is not used in any PostgreSQL API definitions.

• ⇑ Upgrade to 9.3.20 released on 2017-11-09 - docs

  • Fix crash due to rowtype mismatch in json{b}_populate_recordset() (Michael Paquier, Tom Lane)

    These functions used the result rowtype specified in the FROM ... AS clause without checking that it matched the actual rowtype of the supplied tuple value. If it didn't, that would usually result in a crash, though disclosure of server memory contents seems possible as well. (CVE-2017-15098)

  • Fix sample server-start scripts to become $PGUSER before opening $PGLOG (Noah Misch)

    Previously, the postmaster log file was opened while still running as root. The database owner could therefore mount an attack against another system user by making $PGLOG be a symbolic link to some other file, which would then become corrupted by appending log messages.

    By default, these scripts are not installed anywhere. Users who have made use of them will need to manually recopy them, or apply the same changes to their modified versions. If the existing $PGLOG file is root-owned, it will need to be removed or renamed out of the way before restarting the server with the corrected script. (CVE-2017-12172)

  • Properly reject attempts to convert infinite float values to type numeric (Tom Lane, KaiGai Kohei)

    Previously the behavior was platform-dependent.

  • Fix corner-case crashes when columns have been added to the end of a view (Tom Lane)

  • Record proper dependencies when a view or rule contains FieldSelect or FieldStore expression nodes (Tom Lane)

    Lack of these dependencies could allow a column or data type DROP to go through when it ought to fail, thereby causing later uses of the view or rule to get errors. This patch does not do anything to protect existing views/rules, only ones created in the future.

  • Correctly detect hashability of range data types (Tom Lane)

    The planner mistakenly assumed that any range type could be hashed for use in hash joins or hash aggregation, but actually it must check whether the range's subtype has hash support. This does not affect any of the built-in range types, since they're all hashable anyway.

  • Fix low-probability loss of NOTIFY messages due to XID wraparound (Marko Tiikkaja, Tom Lane)

    If a session executed no queries, but merely listened for notifications, for more than 2 billion transactions, it started to miss some notifications from concurrently-committing transactions.

  • Prevent low-probability crash in processing of nested trigger firings (Tom Lane)

  • Correctly restore the umask setting when file creation fails in COPY or lo_export() (Peter Eisentraut)

  • Give a better error message for duplicate column names in ANALYZE (Nathan Bossart)

  • Fix mis-parsing of the last line in a non-newline-terminated pg_hba.conf file (Tom Lane)

  • Fix libpq to not require user's home directory to exist (Tom Lane)

    In v10, failure to find the home directory while trying to read ~/.pgpass was treated as a hard error, but it should just cause that file to not be found. Both v10 and previous release branches made the same mistake when reading ~/.pg_service.conf, though this was less obvious since that file is not sought unless a service name is specified.

  • Fix libpq to guard against integer overflow in the row count of a PGresult (Michael Paquier)

  • Fix ecpg's handling of out-of-scope cursor declarations with pointer or array variables (Michael Meskes)

  • Make ecpglib's Informix-compatibility mode ignore fractional digits in integer input strings, as expected (Gao Zengqi, Michael Meskes)

  • Sync our copy of the timezone library with IANA release tzcode2017c (Tom Lane)

    This fixes various issues; the only one likely to be user-visible is that the default DST rules for a POSIX-style zone name, if no posixrules file exists in the timezone data directory, now match current US law rather than what it was a dozen years ago.

  • Update time zone data files to tzdata release 2017c for DST law changes in Fiji, Namibia, Northern Cyprus, Sudan, Tonga, and Turks & Caicos Islands, plus historical corrections for Alaska, Apia, Burma, Calcutta, Detroit, Ireland, Namibia, and Pago Pago.

• ⇑ Upgrade to 9.3.21 released on 2018-02-08 - docs

  • Ensure that all temporary files made by pg_upgrade are non-world-readable (Tom Lane, Noah Misch)

    pg_upgrade normally restricts its temporary files to be readable and writable only by the calling user. But the temporary file containing pg_dumpall -g output would be group- or world-readable, or even writable, if the user's umask setting allows. In typical usage on multi-user machines, the umask and/or the working directory's permissions would be tight enough to prevent problems; but there may be people using pg_upgrade in scenarios where this oversight would permit disclosure of database passwords to unfriendly eyes. (CVE-2018-1053)

  • Fix vacuuming of tuples that were updated while key-share locked (Andres Freund, Álvaro Herrera)

    In some cases VACUUM would fail to remove such tuples even though they are now dead, leading to assorted data corruption scenarios.

  • Fix inadequate buffer locking in some LSN fetches (Jacob Champion, Asim Praveen, Ashwin Agrawal)

    These errors could result in misbehavior under concurrent load. The potential consequences have not been characterized fully.

  • Avoid unnecessary failure in a query on an inheritance tree that occurs concurrently with some child table being removed from the tree by ALTER TABLE NO INHERIT (Tom Lane)

  • Repair failure with correlated sub-SELECT inside VALUES inside a LATERAL subquery (Tom Lane)

  • Fix “could not devise a query plan for the given query” planner failure for some cases involving nested UNION ALL inside a lateral subquery (Tom Lane)

  • Fix has_sequence_privilege() to support WITH GRANT OPTION tests, as other privilege-testing functions do (Joe Conway)

  • In databases using UTF8 encoding, ignore any XML declaration that asserts a different encoding (Pavel Stehule, Noah Misch)

    We always store XML strings in the database encoding, so allowing libxml to act on a declaration of another encoding gave wrong results. In encodings other than UTF8, we don't promise to support non-ASCII XML data anyway, so retain the previous behavior for bug compatibility. This change affects only xpath() and related functions; other XML code paths already acted this way.

  • Provide for forward compatibility with future minor protocol versions (Robert Haas, Badrul Chowdhury)

    Up to now, PostgreSQL servers simply rejected requests to use protocol versions newer than 3.0, so that there was no functional difference between the major and minor parts of the protocol version number. Allow clients to request versions 3.x without failing, sending back a message showing that the server only understands 3.0. This makes no difference at the moment, but back-patching this change should allow speedier introduction of future minor protocol upgrades.

  • Prevent stack-overflow crashes when planning extremely deeply nested set operations (UNION/INTERSECT/EXCEPT) (Tom Lane)

  • Fix null-pointer crashes for some types of LDAP URLs appearing in pg_hba.conf (Thomas Munro)

  • Fix sample INSTR() functions in the PL/pgSQL documentation (Yugo Nagata, Tom Lane)

    These functions are stated to be Oracle® compatible, but they weren't exactly. In particular, there was a discrepancy in the interpretation of a negative third parameter: Oracle thinks that a negative value indicates the last place where the target substring can begin, whereas our functions took it as the last place where the target can end. Also, Oracle throws an error for a zero or negative fourth parameter, whereas our functions returned zero.

    The sample code has been adjusted to match Oracle's behavior more precisely. Users who have copied this code into their applications may wish to update their copies.

  • Fix pg_dump to make ACL (permissions), comment, and security label entries reliably identifiable in archive output formats (Tom Lane)

    The “tag” portion of an ACL archive entry was usually just the name of the associated object. Make it start with the object type instead, bringing ACLs into line with the convention already used for comment and security label archive entries. Also, fix the comment and security label entries for the whole database, if present, to make their tags start with DATABASE so that they also follow this convention. This prevents false matches in code that tries to identify large-object-related entries by seeing if the tag starts with LARGE OBJECT. That could have resulted in misclassifying entries as data rather than schema, with undesirable results in a schema-only or data-only dump.

    Note that this change has user-visible results in the output of pg_restore --list.

  • In ecpg, detect indicator arrays that do not have the correct length and report an error (David Rader)

  • Avoid triggering a libc assertion in contrib/hstore, due to use of memcpy() with equal source and destination pointers (Tomas Vondra)

  • Provide modern examples of how to auto-start Postgres on macOS (Tom Lane)

    The scripts in contrib/start-scripts/osx use infrastructure that's been deprecated for over a decade, and which no longer works at all in macOS releases of the last couple of years. Add a new subdirectory contrib/start-scripts/macos containing scripts that use the newer launchd infrastructure.

  • Fix incorrect selection of configuration-specific libraries for OpenSSL on Windows (Andrew Dunstan)

  • Support linking to MinGW-built versions of libperl (Noah Misch)

    This allows building PL/Perl with some common Perl distributions for Windows.

  • Fix MSVC build to test whether 32-bit libperl needs -D_USE_32BIT_TIME_T (Noah Misch)

    Available Perl distributions are inconsistent about what they expect, and lack any reliable means of reporting it, so resort to a build-time test on what the library being used actually does.

  • On Windows, install the crash dump handler earlier in postmaster startup (Takayuki Tsunakawa)

    This may allow collection of a core dump for some early-startup failures that did not produce a dump before.

  • On Windows, avoid encoding-conversion-related crashes when emitting messages very early in postmaster startup (Takayuki Tsunakawa)

  • Use our existing Motorola 68K spinlock code on OpenBSD as well as NetBSD (David Carlier)

  • Add support for spinlocks on Motorola 88K (David Carlier)

  • Update time zone data files to tzdata release 2018c for DST law changes in Brazil, Sao Tome and Principe, plus historical corrections for Bolivia, Japan, and South Sudan. The US/Pacific-New zone has been removed (it was only an alias for America/Los_Angeles anyway).

• ⇑ Upgrade to 9.3.22 released on 2018-03-01 - docs

  • Document how to configure installations and applications to guard against search-path-dependent trojan-horse attacks from other users (Noah Misch)

    Using a search_path setting that includes any schemas writable by a hostile user enables that user to capture control of queries and then run arbitrary SQL code with the permissions of the attacked user. While it is possible to write queries that are proof against such hijacking, it is notationally tedious, and it's very easy to overlook holes. Therefore, we now recommend configurations in which no untrusted schemas appear in one's search path. Relevant documentation appears in Section 5.8.6 (for database administrators and users), Section 33.1 (for application authors), Section 37.15.1 (for extension authors), and CREATE FUNCTION (for authors of SECURITY DEFINER functions). (CVE-2018-1058)

  • Avoid use of insecure search_path settings in pg_dump and other client programs (Noah Misch, Tom Lane)

    pg_dump, pg_upgrade, vacuumdb and other PostgreSQL-provided applications were themselves vulnerable to the type of hijacking described in the previous changelog entry; since these applications are commonly run by superusers, they present particularly attractive targets. To make them secure whether or not the installation as a whole has been secured, modify them to include only the pg_catalog schema in their search_path settings. Autovacuum worker processes now do the same, as well.

    In cases where user-provided functions are indirectly executed by these programs — for example, user-provided functions in index expressions — the tighter search_path may result in errors, which will need to be corrected by adjusting those user-provided functions to not assume anything about what search path they are invoked under. That has always been good practice, but now it will be necessary for correct behavior. (CVE-2018-1058)

  • Fix misbehavior of concurrent-update rechecks with CTE references appearing in subplans (Tom Lane)

    If a CTE (WITH clause reference) is used in an InitPlan or SubPlan, and the query requires a recheck due to trying to update or lock a concurrently-updated row, incorrect results could be obtained.

  • Fix planner failures with overlapping mergejoin clauses in an outer join (Tom Lane)

    These mistakes led to “left and right pathkeys do not match in mergejoin” or “outer pathkeys do not match mergeclauses” planner errors in corner cases.

  • Repair pg_upgrade's failure to preserve relfrozenxid for materialized views (Tom Lane, Andres Freund)

    This oversight could lead to data corruption in materialized views after an upgrade, manifesting as “could not access status of transaction” or “found xmin from before relfrozenxid” errors. The problem would be more likely to occur in seldom-refreshed materialized views, or ones that were maintained only with REFRESH MATERIALIZED VIEW CONCURRENTLY.

    If such corruption is observed, it can be repaired by refreshing the materialized view (without CONCURRENTLY).

  • Fix incorrect reporting of PL/Python function names in error CONTEXT stacks (Tom Lane)

    An error occurring within a nested PL/Python function call (that is, one reached via a SPI query from another PL/Python function) would result in a stack trace showing the inner function's name twice, rather than the expected results. Also, an error in a nested PL/Python DO block could result in a null pointer dereference crash on some platforms.

  • Allow contrib/auto_explain's log_min_duration setting to range up to INT_MAX, or about 24 days instead of 35 minutes (Tom Lane)

• ⇑ Upgrade to 9.3.23 released on 2018-05-10 - docs

  • Fix incorrect volatility markings on a few built-in functions (Thomas Munro, Tom Lane)

    The functions query_to_xml, cursor_to_xml, cursor_to_xmlschema, query_to_xmlschema, and query_to_xml_and_xmlschema should be marked volatile because they execute user-supplied queries that might contain volatile operations. They were not, leading to a risk of incorrect query optimization. This has been repaired for new installations by correcting the initial catalog data, but existing installations will continue to contain the incorrect markings. Practical use of these functions seems to pose little hazard, but in case of trouble, it can be fixed by manually updating these functions' pg_proc entries, for example ALTER FUNCTION pg_catalog.query_to_xml(text, boolean, boolean, text) VOLATILE. (Note that that will need to be done in each database of the installation.) Another option is to pg_upgrade the database to a version containing the corrected initial data.

  • Avoid re-using TOAST value OIDs that match dead-but-not-yet-vacuumed TOAST entries (Pavan Deolasee)

    Once the OID counter has wrapped around, it's possible to assign a TOAST value whose OID matches a previously deleted entry in the same TOAST table. If that entry were not yet vacuumed away, this resulted in “unexpected chunk number 0 (expected 1) for toast value nnnnn” errors, which would persist until the dead entry was removed by VACUUM. Fix by not selecting such OIDs when creating a new TOAST entry.

  • Change ANALYZE's algorithm for updating pg_class.reltuples (David Gould)

    Previously, pages not actually scanned by ANALYZE were assumed to retain their old tuple density. In a large table where ANALYZE samples only a small fraction of the pages, this meant that the overall tuple density estimate could not change very much, so that reltuples would change nearly proportionally to changes in the table's physical size (relpages) regardless of what was actually happening in the table. This has been observed to result in reltuples becoming so much larger than reality as to effectively shut off autovacuuming. To fix, assume that ANALYZE's sample is a statistically unbiased sample of the table (as it should be), and just extrapolate the density observed within those pages to the whole table.

  • Fix UPDATE/DELETE ... WHERE CURRENT OF to not fail when the referenced cursor uses an index-only-scan plan (Yugo Nagata, Tom Lane)

  • Fix incorrect planning of join clauses pushed into parameterized paths (Andrew Gierth, Tom Lane)

    This error could result in misclassifying a condition as a “join filter” for an outer join when it should be a plain “filter” condition, leading to incorrect join output.

  • Fix misoptimization of CHECK constraints having provably-NULL subclauses of top-level AND/OR conditions (Tom Lane, Dean Rasheed)

    This could, for example, allow constraint exclusion to exclude a child table that should not be excluded from a query.

  • Avoid failure if a query-cancel or session-termination interrupt occurs while committing a prepared transaction (Stas Kelvich)

  • Fix query-lifespan memory leakage in repeatedly executed hash joins (Tom Lane)

  • Fix overly strict sanity check in heap_prepare_freeze_tuple (Álvaro Herrera)

    This could result in incorrect “cannot freeze committed xmax” failures in databases that have been pg_upgrade'd from 9.2 or earlier.

  • Prevent dangling-pointer dereference when a C-coded before-update row trigger returns the “old” tuple (Rushabh Lathia)

  • Reduce locking during autovacuum worker scheduling (Jeff Janes)

    The previous behavior caused drastic loss of potential worker concurrency in databases with many tables.

  • Ensure client hostname is copied while copying pg_stat_activity data to local memory (Edmund Horner)

    Previously the supposedly-local snapshot contained a pointer into shared memory, allowing the client hostname column to change unexpectedly if any existing session disconnected.

  • Fix incorrect processing of multiple compound affixes in ispell dictionaries (Arthur Zakirov)

  • Fix collation-aware searches (that is, indexscans using inequality operators) in SP-GiST indexes on text columns (Tom Lane)

    Such searches would return the wrong set of rows in most non-C locales.

  • Count the number of index tuples correctly during initial build of an SP-GiST index (Tomas Vondra)

    Previously, the tuple count was reported to be the same as that of the underlying table, which is wrong if the index is partial.

  • Count the number of index tuples correctly during vacuuming of a GiST index (Andrey Borodin)

    Previously it reported the estimated number of heap tuples, which might be inaccurate, and is certainly wrong if the index is partial.

  • Allow scalarltsel and scalargtsel to be used on non-core datatypes (Tomas Vondra)

  • Reduce libpq's memory consumption when a server error is reported after a large amount of query output has been collected (Tom Lane)

    Discard the previous output before, not after, processing the error message. On some platforms, notably Linux, this can make a difference in the application's subsequent memory footprint.

  • Fix double-free crashes in ecpg (Patrick Krecker, Jeevan Ladhe)

  • Fix ecpg to handle long long int variables correctly in MSVC builds (Michael Meskes, Andrew Gierth)

  • Fix mis-quoting of values for list-valued GUC variables in dumps (Michael Paquier, Tom Lane)

    The local_preload_libraries, session_preload_libraries, shared_preload_libraries, and temp_tablespaces variables were not correctly quoted in pg_dump output. This would cause problems if settings for these variables appeared in CREATE FUNCTION ... SET or ALTER DATABASE/ROLE ... SET clauses.

  • Fix overflow handling in PL/pgSQL integer FOR loops (Tom Lane)

    The previous coding failed to detect overflow of the loop variable on some non-gcc compilers, leading to an infinite loop.

  • Adjust PL/Python regression tests to pass under Python 3.7 (Peter Eisentraut)

  • Support testing PL/Python and related modules when building with Python 3 and MSVC (Andrew Dunstan)

  • Rename internal b64_encode and b64_decode functions to avoid conflict with Solaris 11.4 built-in functions (Rainer Orth)

  • Sync our copy of the timezone library with IANA tzcode release 2018e (Tom Lane)

    This fixes the zic timezone data compiler to cope with negative daylight-savings offsets. While the PostgreSQL project will not immediately ship such timezone data, zic might be used with timezone data obtained directly from IANA, so it seems prudent to update zic now.

  • Update time zone data files to tzdata release 2018d for DST law changes in Palestine and Antarctica (Casey Station), plus historical corrections for Portugal and its colonies, as well as Enderbury, Jamaica, Turks & Caicos Islands, and Uruguay.

• ⇑ Upgrade to 9.3.24 released on 2018-08-09 - docs

  • Fix failure to reset libpq's state fully between connection attempts (Tom Lane)

    An unprivileged user of dblink or postgres_fdw could bypass the checks intended to prevent use of server-side credentials, such as a ~/.pgpass file owned by the operating-system user running the server. Servers allowing peer authentication on local connections are particularly vulnerable. Other attacks such as SQL injection into a postgres_fdw session are also possible. Attacking postgres_fdw in this way requires the ability to create a foreign server object with selected connection parameters, but any user with access to dblink could exploit the problem. In general, an attacker with the ability to select the connection parameters for a libpq-using application could cause mischief, though other plausible attack scenarios are harder to think of. Our thanks to Andrew Krasichkov for reporting this issue. (CVE-2018-10915)

  • Ensure that updates to the relfrozenxid and relminmxid values for “nailed” system catalogs are processed in a timely fashion (Andres Freund)

    Overoptimistic caching rules could prevent these updates from being seen by other sessions, leading to spurious errors and/or data corruption. The problem was significantly worse for shared catalogs, such as pg_authid, because the stale cache data could persist into new sessions as well as existing ones.

  • Fix case where a freshly-promoted standby crashes before having completed its first post-recovery checkpoint (Michael Paquier, Kyotaro Horiguchi, Pavan Deolasee, Álvaro Herrera)

    This led to a situation where the server did not think it had reached a consistent database state during subsequent WAL replay, preventing restart.

  • Avoid emitting a bogus WAL record when recycling an all-zero btree page (Amit Kapila)

    This mistake has been seen to cause assertion failures, and potentially it could result in unnecessary query cancellations on hot standby servers.

  • Improve performance of WAL replay for transactions that drop many relations (Fujii Masao)

    This change reduces the number of times that shared buffers are scanned, so that it is of most benefit when that setting is large.

  • Improve performance of lock releasing in standby server WAL replay (Thomas Munro)

  • Ensure a table's cached index list is correctly rebuilt after an index creation fails partway through (Peter Geoghegan)

    Previously, the failed index's OID could remain in the list, causing problems later in the same session.

  • Fix misoptimization of equivalence classes involving composite-type columns (Tom Lane)

    This resulted in failure to recognize that an index on a composite column could provide the sort order needed for a mergejoin on that column.

  • Fix SQL-standard FETCH FIRST syntax to allow parameters ($n), as the standard expects (Andrew Gierth)

  • Fix failure to schema-qualify some object names in getObjectDescription output (Kyotaro Horiguchi, Tom Lane)

    Names of collations, conversions, and text search objects were not schema-qualified when they should be.

  • Widen COPY FROM's current-line-number counter from 32 to 64 bits (David Rowley)

    This avoids two problems with input exceeding 4G lines: COPY FROM WITH HEADER would drop a line every 4G lines, not only the first line, and error reports could show a wrong line number.

  • Add a string freeing function to ecpg's pgtypes library, so that cross-module memory management problems can be avoided on Windows (Takayuki Tsunakawa)

    On Windows, crashes can ensue if the free call for a given chunk of memory is not made from the same DLL that malloc'ed the memory. The pgtypes library sometimes returns strings that it expects the caller to free, making it impossible to follow this rule. Add a PGTYPESchar_free() function that just wraps free, allowing applications to follow this rule.

  • Fix ecpg's support for long long variables on Windows, as well as other platforms that declare strtoll/strtoull nonstandardly or not at all (Dang Minh Huong, Tom Lane)

  • Fix misidentification of SQL statement type in PL/pgSQL, when a rule change causes a change in the semantics of a statement intra-session (Tom Lane)

    This error led to assertion failures, or in rare cases, failure to enforce the INTO STRICT option as expected.

  • Fix password prompting in client programs so that echo is properly disabled on Windows when stdin is not the terminal (Matthew Stickney)

  • Further fix mis-quoting of values for list-valued GUC variables in dumps (Tom Lane)

    The previous fix for quoting of search_path and other list-valued variables in pg_dump output turned out to misbehave for empty-string list elements, and it risked truncation of long file paths.

  • Make pg_upgrade check that the old server was shut down cleanly (Bruce Momjian)

    The previous check could be fooled by an immediate-mode shutdown.

  • Fix crash in contrib/ltree's lca() function when the input array is empty (Pierre Ducroquet)

  • Fix various error-handling code paths in which an incorrect error code might be reported (Michael Paquier, Tom Lane, Magnus Hagander)

  • Rearrange makefiles to ensure that programs link to freshly-built libraries (such as libpq.so) rather than ones that might exist in the system library directories (Tom Lane)

    This avoids problems when building on platforms that supply old copies of PostgreSQL libraries.

  • Update time zone data files to tzdata release 2018e for DST law changes in North Korea, plus historical corrections for Czechoslovakia.

    This update includes a redefinition of “daylight savings” in Ireland, as well as for some past years in Namibia and Czechoslovakia. In those jurisdictions, legally standard time is observed in summer, and daylight savings time in winter, so that the daylight savings offset is one hour behind standard time not one hour ahead. This does not affect either the actual UTC offset or the timezone abbreviations in use; the only known effect is that the is_dst column in the pg_timezone_names view will now be true in winter and false in summer in these cases.

• ⇑ Upgrade to 9.3.25 released on 2018-11-08 - docs

  • Fix corner-case failures in has_foo_privilege() family of functions (Tom Lane)

    Return NULL rather than throwing an error when an invalid object OID is provided. Some of these functions got that right already, but not all. has_column_privilege() was additionally capable of crashing on some platforms.

  • Avoid O(N^2) slowdown in regular expression match/split functions on long strings (Andrew Gierth)

  • Avoid O(N^3) slowdown in lexer for long strings of + or - characters (Andrew Gierth)

  • Fix mis-execution of SubPlans when the outer query is being scanned backwards (Andrew Gierth)

  • Fix failure of UPDATE/DELETE ... WHERE CURRENT OF ... after rewinding the referenced cursor (Tom Lane)

    A cursor that scans multiple relations (particularly an inheritance tree) could produce wrong behavior if rewound to an earlier relation.

  • Fix EvalPlanQual to handle conditionally-executed InitPlans properly (Andrew Gierth, Tom Lane)

    This resulted in hard-to-reproduce crashes or wrong answers in concurrent updates, if they contained code such as an uncorrelated sub-SELECT inside a CASE construct.

  • Fix character-class checks to not fail on Windows for Unicode characters above U+FFFF (Tom Lane, Kenji Uno)

    This bug affected full-text-search operations, as well as contrib/ltree and contrib/pg_trgm.

  • Ensure that sequences owned by a foreign table are processed by ALTER OWNER on the table (Peter Eisentraut)

    The ownership change should propagate to such sequences as well, but this was missed for foreign tables.

  • Fix over-allocation of space for array_out()'s result string (Keiichi Hirobe)

  • Fix memory leak in repeated SP-GiST index scans (Tom Lane)

    This is only known to amount to anything significant in cases where an exclusion constraint using SP-GiST receives many new index entries in a single command.

  • Avoid crash if a utility command causes infinite recursion (Tom Lane)

  • When initializing a hot standby, cope with duplicate XIDs caused by two-phase transactions on the master (Michael Paquier, Konstantin Knizhnik)

  • Randomize the random() seed in bootstrap and standalone backends, and in initdb (Noah Misch)

    The main practical effect of this change is that it avoids a scenario where initdb might mistakenly conclude that POSIX shared memory is not available, due to name collisions caused by always using the same random seed.

  • Ensure that hot standby processes use the correct WAL consistency point (Alexander Kukushkin, Michael Paquier)

    This prevents possible misbehavior just after a standby server has reached a consistent database state during WAL replay.

  • Don't run atexit callbacks when servicing SIGQUIT (Heikki Linnakangas)

  • Don't record foreign-server user mappings as members of extensions (Tom Lane)

    If CREATE USER MAPPING is executed in an extension script, an extension dependency was created for the user mapping, which is unexpected. Roles can't be extension members, so user mappings shouldn't be either.

  • Make syslogger more robust against failures in opening CSV log files (Tom Lane)

  • Fix possible inconsistency in pg_dump's sorting of dissimilar object names (Jacob Champion)

  • Ensure that pg_restore will schema-qualify the table name when emitting DISABLE/ENABLE TRIGGER commands (Tom Lane)

    This avoids failures due to the new policy of running restores with restrictive search path.

  • Fix pg_upgrade to handle event triggers in extensions correctly (Haribabu Kommi)

    pg_upgrade failed to preserve an event trigger's extension-membership status.

  • Fix pg_upgrade's cluster state check to work correctly on a standby server (Bruce Momjian)

  • Enforce type cube's dimension limit in all contrib/cube functions (Andrey Borodin)

    Previously, some cube-related functions could construct values that would be rejected by cube_in(), leading to dump/reload failures.

  • Fix contrib/unaccent's unaccent() function to use the unaccent text search dictionary that is in the same schema as the function (Tom Lane)

    Previously it tried to look up the dictionary using the search path, which could fail if the search path has a restrictive value.

  • Fix build problems on macOS 10.14 (Mojave) (Tom Lane)

    Adjust configure to add an -isysroot switch to CPPFLAGS; without this, PL/Perl and PL/Tcl fail to configure or build on macOS 10.14. The specific sysroot used can be overridden at configure time or build time by setting the PG_SYSROOT variable in the arguments of configure or make.

    It is now recommended that Perl-related extensions write $(perl_includespec) rather than -I$(perl_archlibexp)/CORE in their compiler flags. The latter continues to work on most platforms, but not recent macOS.

    Also, it should no longer be necessary to specify --with-tclconfig manually to get PL/Tcl to build on recent macOS releases.

  • Fix MSVC build and regression-test scripts to work on recent Perl versions (Andrew Dunstan)

    Perl no longer includes the current directory in its search path by default; work around that.

  • Support building on Windows with Visual Studio 2015 or Visual Studio 2017 (Michael Paquier, Haribabu Kommi)

  • Allow btree comparison functions to return INT_MIN (Tom Lane)

    Up to now, we've forbidden datatype-specific comparison functions from returning INT_MIN, which allows callers to invert the sort order just by negating the comparison result. However, this was never safe for comparison functions that directly return the result of memcmp(), strcmp(), etc, as POSIX doesn't place any such restriction on those functions. At least some recent versions of memcmp() can return INT_MIN, causing incorrect sort ordering. Hence, we've removed this restriction. Callers must now use the INVERT_COMPARE_RESULT() macro if they wish to invert the sort order.

  • Fix recursion hazard in shared-invalidation message processing (Tom Lane)

    This error could, for example, result in failure to access a system catalog or index that had just been processed by VACUUM FULL.

    This change adds a new result code for LockAcquire, which might possibly affect external callers of that function, though only very unusual usage patterns would have an issue with it. The API of LockAcquireExtended is also changed.

  • Save and restore SPI's global variables during SPI_connect() and SPI_finish() (Chapman Flack, Tom Lane)

    This prevents possible interference when one SPI-using function calls another.

  • Provide ALLOCSET_DEFAULT_SIZES and sibling macros in back branches (Tom Lane)

    These macros have existed since 9.6, but there were requests to add them to older branches to allow extensions to rely on them without branch-specific coding.

  • Avoid using potentially-under-aligned page buffers (Tom Lane)

    Invent new union types PGAlignedBlock and PGAlignedXLogBlock, and use these in place of plain char arrays, ensuring that the compiler can't place the buffer at a misaligned start address. This fixes potential core dumps on alignment-picky platforms, and may improve performance even on platforms that allow misalignment.

  • Make src/port/snprintf.c follow the C99 standard's definition of snprintf()'s result value (Tom Lane)

    On platforms where this code is used (mostly Windows), its pre-C99 behavior could lead to failure to detect buffer overrun, if the calling code assumed C99 semantics.

  • When building on i386 with the clang compiler, require -msse2 to be used (Andres Freund)

    This avoids problems with missed floating point overflow checks.

  • Fix configure's detection of the result type of strerror_r() (Tom Lane)

    The previous coding got the wrong answer when building with icc on Linux (and perhaps in other cases), leading to libpq not returning useful error messages for system-reported errors.

  • Update time zone data files to tzdata release 2018g for DST law changes in Chile, Fiji, Morocco, and Russia (Volgograd), plus historical corrections for China, Hawaii, Japan, Macau, and North Korea.

• source repo
• author info
• meta-info