Upgrading from 15.1 to 15.6 gives you 1.2 years worth of fixes (308 of them)

List of changes:

• ⇑ Upgrade to 15.2 released on 2023-02-09 - docs

  • libpq can leak memory contents after GSSAPI transport encryption initiation fails (Jacob Champion)

    A modified server, or an unauthenticated man-in-the-middle, can send a not-zero-terminated error message during setup of GSSAPI (Kerberos) transport encryption. libpq will then copy that string, as well as following bytes in application memory up to the next zero byte, to its error report. Depending on what the calling application does with the error report, this could result in disclosure of application memory contents. There is also a small probability of a crash due to reading beyond the end of memory. Fix by properly zero-terminating the server message. (CVE-2022-41862)

  • Fix calculation of which GENERATED columns need to be updated in child tables during an UPDATE on a partitioned table or inheritance tree (Amit Langote, Tom Lane)

    This fixes failure to update GENERATED columns that do not exist in the parent table, or that have different dependencies than are in the parent column's generation expression.

  • Fix possible failure of MERGE to compute GENERATED columns (Dean Rasheed)

    When the first row-level action of the MERGE was an UPDATE, any subsequent INSERT actions would fail to compute GENERATED columns that were deemed unnecessary to compute for the UPDATE action (due to not depending on any of the UPDATE target columns).

  • Fix MERGE's check for unreachable WHEN clauses (Dean Rasheed)

    A WHEN clause following an unconditional WHEN clause should be rejected as unreachable, but this case was not always detected.

  • Fix MERGE's rule-detection test (Dean Rasheed)

    MERGE is not supported on tables with rules; but it also failed on tables that once had rules but no longer do.

  • In MERGE, don't count a DO NOTHING action as a processed tuple (Álvaro Herrera)

    This makes the code's behavior match the documentation.

  • Allow a WITH RECURSIVE ... CYCLE CTE to access its output column (Tom Lane)

    A reference to the SET column from within the CTE would fail with “cache lookup failed for type 0”.

  • Fix handling of pending inserts when doing a bulk insertion to a foreign table (Etsuro Fujita)

    In some cases pending insertions were not flushed to the FDW soon enough, leading to logical inconsistencies, for example BEFORE ROW triggers not seeing rows they should be able to see.

  • Allow REPLICA IDENTITY to be set on an index that's not (yet) valid (Tom Lane)

    When pg_dump dumps a partitioned index that's marked REPLICA IDENTITY, it generates a command sequence that applies REPLICA IDENTITY before the partitioned index has been marked valid, causing restore to fail. There seems no very good reason to prohibit doing it in that order, so allow it. The marking will have no effect anyway until the index becomes valid.

  • Fix handling of DEFAULT markers in rules that perform an INSERT from a multi-row VALUES list (Dean Rasheed)

    In some cases a DEFAULT marker would not get replaced with the proper default-value expression, leading to an “unrecognized node type” error.

  • Reject uses of undefined variables in jsonpath existence checks (Alexander Korotkov, David G. Johnston)

    While jsonpath match operators threw an error for an undefined variable in the path pattern, the existence operators silently treated it as a match.

  • Fix jsonb subscripting to cope with toasted subscript values (Tom Lane, David G. Johnston)

    Using a text value fetched directly from a table as a jsonb subscript was likely to fail. Fetches would usually not find any matching element. Assignments could store the value with a garbage key, although keys long enough to cause that problem are probably rare in the field.

  • Fix edge-case data corruption in parallel hash joins (Dmitry Astapov)

    If the final chunk of a large tuple being written out to a temporary file was exactly 32760 bytes, it would be corrupted due to a fencepost bug. The query would typically fail later with corrupted-data symptoms.

  • Honor non-default settings of checkpoint_completion_target (Bharath Rupireddy)

    Internal state was not updated after a change in checkpoint_completion_target, possibly resulting in performing checkpoint I/O faster or slower than desired, especially if that setting was changed on-the-fly.

  • Log the correct ending timestamp in recovery_target_xid mode (Tom Lane)

    When ending recovery based on the recovery_target_xid setting with recovery_target_inclusive = off, we printed an incorrect timestamp (always 2000-01-01) in the “recovery stopping before ... transaction” log message.

  • Improve error reporting for some buffered file read failures (Peter Eisentraut)

    Correctly report a short read, giving the numbers of bytes desired and actually read, instead of reporting an irrelevant error code. Most places got this right already, but some recently-written replication logic did not.

  • Remove arbitrary limit on number of elements in int2vector and oidvector (Tom Lane)

    The input functions for these types previously rejected more than 100 elements. With the introduction of the logical replication column list feature, it's necessary to accept int2vectors having up to 1600 columns, otherwise long column lists cause logical-replication failures.

  • In extended query protocol, avoid an immediate commit after ANALYZE if we're running a pipeline (Tom Lane)

    If there's not been an explicit BEGIN TRANSACTION, ANALYZE would take it on itself to commit, which should not happen within a pipelined series of commands.

  • Reject cancel request packets having the wrong length (Andrey Borodin)

    The server would process a cancel request even if its length word was too small. This led to reading beyond the end of the allocated buffer. In theory that could cause a segfault, but it seems quite unlikely to happen in practice, since the buffer would have to be very close to the end of memory. The more likely outcome was a bogus log message about wrong backend PID or cancel code. Complain about the wrong length, instead.

  • Fix planner preprocessing oversights for window function run-condition expressions (Richard Guo, David Rowley)

    This could lead to planner errors such as “WindowFunc not found in subplan target lists”.

  • Fix possible dangling-pointer access during execution of window function run-condition expressions (David Rowley)

    In practice, because the run-condition optimization is only applied to certain window functions that happen to all return int8, this only manifested as a problem on 32-bit builds.

  • Add recursion and looping defenses in subquery pullup (Tom Lane)

    A contrived query can result in deep recursion and unreasonable amounts of time spent trying to flatten subqueries. A proper fix for that seems unduly invasive for a back-patch, but we can at least add stack depth checks and an interrupt check to allow the query to be cancelled.

  • Fix planner issues when combining Memoize nodes with partitionwise joins or parameterized nestloops (Richard Guo)

    These errors could lead to not using Memoize in contexts where it would be useful, or possibly to wrong query plans.

  • Fix partitionwise-join code to tolerate failure to produce a plan for each partition (Tom Lane)

    This could result in “could not devise a query plan for the given query” errors.

  • Limit the amount of cleanup work done by get_actual_variable_range (Simon Riggs)

    Planner runs occurring just after deletion of a large number of tuples appearing at the end of an index could expend significant amounts of work setting the “killed” bits for those index entries. Limit the amount of work done in any one query by giving up on this process after examining 100 heap pages. All the cleanup will still happen eventually, but without so large a performance hiccup.

  • Prevent the statistics machinery from getting confused when a relation's relkind changes (Andres Freund)

    Converting a table to a view could lead to crashes or assertion failures.

  • Fix under-parenthesized display of AT TIME ZONE constructs (Tom Lane)

    This could result in dump/restore failures for rules or views in which an argument of AT TIME ZONE is itself an expression.

  • Prevent clobbering of cached parsetrees for utility statements in SQL functions (Tom Lane, Daniel Gustafsson)

    If a SQL-language function executes the same utility command more than once within a single calling query, it could crash or report strange errors such as “unrecognized node type”.

  • Ensure that execution of full-text-search queries can be cancelled while they are performing phrase matches (Tom Lane)

  • Fix memory leak in hashing strings with nondeterministic collations (Jeff Davis)

  • Fix deadlock between DROP DATABASE and logical replication worker process (Hou Zhijie)

    This was caused by an ill-advised choice to block interrupts while creating a logical replication slot in the worker. In version 15 that could lead to an undetected deadlock. In version 14, no deadlock has been observed, but it's still a bad idea to block interrupts while waiting for network I/O.

  • Clean up the libpq connection object after a failed replication connection attempt (Andres Freund)

    The previous coding leaked the connection object. In background code paths that's pretty harmless because the calling process will give up and exit. But in commands such as CREATE SUBSCRIPTION, such a failure resulted in a small session-lifespan memory leak.

  • In hot-standby servers, reduce processing effort for tracking XIDs known to be active on the primary (Simon Riggs, Michail Nikolaev)

    Insufficiently-aggressive cleanup of the KnownAssignedXids array could lead to poor performance, particularly when max_connections is set to a large value on the standby.

  • Ignore invalidated logical-replication slots while determining oldest catalog xmin (Sirisha Chamarthi)

    A replication slot could prevent cleanup of dead tuples in the system catalogs even after it becomes invalidated due to exceeding max_slot_wal_keep_size. Thus, failure of a replication consumer could lead to indefinitely-large catalog bloat.

  • In logical decoding, notify the remote node when a transaction is detected to have crashed (Hou Zhijie)

    After a server restart, we'll re-stream the changes for transactions occurring shortly before the restart. Some of these transactions probably never completed; when we realize that one didn't we throw away the relevant decoding state locally, but we neglected to tell the subscriber about it. That led to the subscriber keeping useless streaming files until it's next restarted.

  • Fix uninitialized-memory usage in logical decoding (Masahiko Sawada)

    In certain cases, resumption of logical decoding could try to re-use XID data that had already been freed, leading to unpredictable behavior.

  • Acquire spinlock while updating shared state during logical decoding context creation (Masahiko Sawada)

    We neglected to acquire the appropriate lock while updating data about two-phase transactions, potentially allowing other processes to see inconsistent data.

  • Fix pgoutput replication plug-in to not send columns not listed in a table's replication column list (Hou Zhijie)

    UPDATE and DELETE events did not pay attention to the configured column list, thus sending more data than expected. This did not cause a problem when the receiver is our built-in logical replication code, but it might confuse other receivers, and in any case it wasted network bandwidth.

  • Avoid rare “failed to acquire cleanup lock” panic during WAL replay of hash-index page split operations (Robert Haas)

  • Advance a heap page's LSN when setting its all-visible bit during WAL replay (Jeff Davis)

    Failure to do this left the page possibly different on standby servers than the primary, and violated some other expectations about when the LSN changes. This seems only a theoretical hazard so far as PostgreSQL itself is concerned, but it could upset third-party tools.

  • Fix int64_div_fast_to_numeric() to work for a wider range of inputs (Dean Rasheed)

    This function misbehaved with some values of its second argument. No such usages exist in core PostgreSQL, but it's clearly a hazard for external modules, so repair.

  • Fix latent buffer-overrun problem in WaitEventSet logic (Thomas Munro)

    The epoll-based and kqueue-based implementations could ask the kernel for too many events if the size of their internal buffer was different from the size of the caller's output buffer. That case is not known to occur in released PostgreSQL versions, but this error is a hazard for external modules and future bug fixes.

  • Avoid nominally-undefined behavior when accessing shared memory in 32-bit builds (Andres Freund)

    clang's undefined-behavior sanitizer complained about use of a pointer that was less aligned than it should be. It's very unlikely that this would cause a problem in non-debug builds, but it's worth fixing for testing purposes.

  • Fix assertion failure in BRIN minmax-multi opclasses (Tomas Vondra)

    The assertion was overly strict, so this mistake was harmless in non-assert builds.

  • Remove faulty assertion in useless-RESULT-RTE optimization logic (Tom Lane)

  • Fix copy-and-paste errors in cache-lookup-failure messages for ACL checks (Justin Pryzby)

    In principle these errors should never be reached. But if they are, some of them reported the wrong type of object.

  • Fix possible corruption of very large tablespace map files in pg_basebackup (Antonin Houska)

  • Avoid harmless warning from pg_dump in --if-exists mode (Tom Lane)

    If the public schema has a non-default owner then use of pg_dump's --if-exists option resulted in a warning message “warning: could not find where to insert IF EXISTS in statement "-- *not* dropping schema, since initdb creates it"”. The dump output was okay, though.

  • Fix psql's \sf and \ef commands to handle SQL-language functions that have SQL-standard function bodies (Tom Lane)

    These commands misidentified the start of the function body when it used new-style syntax.

  • Fix tab completion of ALTER FUNCTION/PROCEDURE/ROUTINE ... SET SCHEMA (Dean Rasheed)

  • Update contrib/pageinspect to mark its disk-accessing functions as PARALLEL RESTRICTED (Tom Lane)

    This avoids possible failure if one of these functions is used to examine a temporary table, since a session's temporary tables are not accessible from parallel workers.

  • Fix contrib/seg to not crash or print garbage if an input number has more than 127 digits (Tom Lane)

  • Fix build on Microsoft Visual Studio 2013 (Tom Lane)

    A previous patch supposed that all platforms of interest have snprintf(), but MSVC 2013 isn't quite there yet. Revert to using sprintf() on that platform.

  • Fix compile failure in building PL/Perl with MSVC when using Strawberry Perl (Andrew Dunstan)

  • Fix mismatch of PL/Perl built with MSVC versus a Perl library built with gcc (Andrew Dunstan)

    Such combinations could previously fail with “loadable library and perl binaries are mismatched” errors.

  • Suppress compiler warnings from Perl's header files (Andres Freund)

    Our preferred compiler options provoke warnings about constructs appearing in recent versions of Perl's header files. When using gcc, we can suppress these warnings with a pragma.

  • Fix pg_waldump to build on compilers that don't discard unused static-inline functions (Tom Lane)

  • Update time zone data files to tzdata release 2022g for DST law changes in Greenland and Mexico, plus historical corrections for northern Canada, Colombia, and Singapore.

    Notably, a new timezone America/Ciudad_Juarez has been split off from America/Ojinaga.

• ⇑ Upgrade to 15.3 released on 2023-05-11 - docs

  • Prevent CREATE SCHEMA from defeating changes in search_path (Alexander Lakhin)

    Within a CREATE SCHEMA command, objects in the prevailing search_path, as well as those in the newly-created schema, would be visible even within a called function or script that attempted to set a secure search_path. This could allow any user having permission to create a schema to hijack the privileges of a security definer function or extension script.

    The PostgreSQL Project thanks Alexander Lakhin for reporting this problem. (CVE-2023-2454)

  • Enforce row-level security policies correctly after inlining a set-returning function (Stephen Frost, Tom Lane)

    If a set-returning SQL-language function refers to a table having row-level security policies, and it can be inlined into a calling query, those RLS policies would not get enforced properly in some cases involving re-using a cached plan under a different role. This could allow a user to see or modify rows that should have been invisible.

    The PostgreSQL Project thanks Wolfgang Walther for reporting this problem. (CVE-2023-2455)

  • Fix potential corruption of the template (source) database after CREATE DATABASE with the STRATEGY WAL_LOG option (Nathan Bossart, Ryo Matsumura)

    Improper buffer handling created a risk that any later modification of the template's pg_class catalog would be lost.

  • Fix memory leakage and unnecessary disk reads during CREATE DATABASE with the STRATEGY WAL_LOG option (Andres Freund)

  • Avoid crash when the new schema name is omitted in CREATE SCHEMA (Michael Paquier)

    The SQL standard allows writing CREATE SCHEMA AUTHORIZATION owner_name, with the schema name defaulting to owner_name. However some code paths expected the schema name to be present and would fail.

  • Fix various planner failures with MERGE commands (Tom Lane)

    Planning could fail with errors like “variable not found in subplan target list” or “PlaceHolderVar found where not expected”.

  • Fix the row count reported by MERGE for some corner cases (Dean Rasheed)

    The row count reported in the command tag counted rows that actually hadn't been modified due to a BEFORE ROW trigger returning NULL. This is inconsistent with what happens in plain UPDATE or DELETE, so change it to not count such rows. Also, avoid counting a row twice when MERGE moves it into a different partition of a partitioned table.

  • Fix MERGE problems with concurrent updates (Dean Rasheed, Álvaro Herrera)

    Some cases misbehaved if a row to be updated or deleted by MERGE had just been updated by a concurrent transaction. This could lead to a crash, or the wrong merge action being executed, or no action at all.

  • Add support for decompiling MERGE commands (Álvaro Herrera)

    This was overlooked when MERGE was added, but it's essential support for MERGE in new-style SQL functions.

  • Fix enabling/disabling of foreign-key triggers in partitioned tables (Tom Lane)

    ALTER TABLE ... ENABLE/DISABLE TRIGGER failed if applied to a partitioned table's foreign-key enforcement triggers, because it tried to locate the clone triggers for the partitions by name, and they do not have the same name. Locate them by parent-trigger OID instead.

  • Disallow altering composite types that are stored in indexes (Tom Lane)

    ALTER TYPE disallows non-binary-compatible modifications of composite types if they are stored in any table columns. (Perhaps that will be allowed someday, but it hasn't happened yet; the locking implications of rewriting many tables are daunting.) We overlooked the possibility that an index might contain a composite type that doesn't also appear in its table.

  • Disallow system columns as elements of foreign keys (Tom Lane)

    Since the removal of OID as a system column, there is no plausible use-case for this, and various bits of code no longer support it. Disallow it rather than trying to fix all the cases.

  • Ensure that COPY TO from an RLS-enabled parent table does not copy any rows from child tables (Antonin Houska)

    The documentation is quite clear that COPY TO copies rows from only the named table, not any inheritance children it may have. However, if row-level security was enabled on the table then this stopped being true.

  • Avoid possible crash when array_position() or array_positions() is passed an empty array (Tom Lane)

  • Fix possible out-of-bounds fetch in to_char() (Tom Lane)

    With bad luck this could have resulted in a server crash.

  • Avoid buffer overread in translate() function (Daniil Anisimov)

    When using the deletion feature, the function might fetch the byte just after the input string, creating a small risk of crash.

  • Adjust text-search-related character classification logic to correctly detect whether the prevailing locale is C (Jeff Davis)

    This code got confused if the database's default collation uses ICU.

  • Avoid possible crash on empty input for type interval (Tom Lane)

  • Re-allow exponential notation in ISO-8601 interval fields (Tom Lane)

    Interval input like P0.1e10D isn't officially sanctioned by ISO-8601, but we accepted it for a long time before version 15, so re-allow it.

  • Fix error cursor setting for parse errors in JSON string literals (Tom Lane)

    Most cases in which a syntax error is detected in a string literal within a JSON value failed to set the error cursor appropriately. This led at least to an unhelpful error message (pointing to the token before the string, rather than the actual trouble spot), and could even result in a crash in v14 and later.

  • Fix data corruption due to vacuum_defer_cleanup_age being larger than the current 64-bit xid (Andres Freund)

    In v14 and later with non-default settings of vacuum_defer_cleanup_age, it was possible to compute a very large vacuum cleanup horizon xid, leading to vacuum removing rows that are still live. v12 and v13 have a lesser form of the same problem affecting only GiST indexes, which could lead to index pages getting recycled too early.

  • Fix parser's failure to detect some cases of improperly-nested aggregates (Tom Lane)

    This oversight could lead to executor failures for queries that should have been rejected as invalid.

  • Fix data structure corruption during parsing of serial SEQUENCE NAME options (David Rowley)

    This can lead to trouble if an event trigger captures the corrupted parse tree.

  • Correctly update plan nodes' parallel-safety markings when moving initplans from one node to another (Tom Lane)

    This planner oversight could lead to “subplan was not initialized” errors at runtime.

  • Avoid failure with PlaceHolderVars in extended-statistics code (Tom Lane)

    Use of dependency-type extended statistics could fail with “PlaceHolderVar found where not expected”.

  • Fix incorrect tests for whether a qual clause applied to a subquery can be transformed into a window aggregate “run condition” within the subquery (David Rowley)

    A SubPlan within such a clause would cause assertion failures or incorrect answers, as would some other unusual cases.

  • Disable the inverse-transition optimization for window aggregates when the call contains sub-SELECTs (David Rowley)

    This optimization requires that the aggregate's argument expressions have repeatable results, which might not hold for a sub-SELECT.

  • Fix oversights in execution of nested ARRAY[] constructs (Alexander Lakhin, Tom Lane)

    Correctly detect overflow of the total space needed for the result array, avoiding a possible crash due to undersized output allocation. Also ensure that any trailing padding space in the result array is zeroed; while leaving garbage there is harmless for most purposes, it can result in odd behavior later.

  • Prevent crash when updating a field within an array-of-domain-over-composite-type column (Dmitry Dolgov)

  • Fix partition pruning logic for partitioning on boolean columns (David Rowley)

    Pruning with a condition like boolcol IS NOT TRUE was done incorrectly, leading to possibly not returning rows in which boolcol is NULL. Also, the rather unlikely case of partitioning on NOT boolcol was handled incorrectly.

  • Fix race condition in per-batch cleanup during parallel hash join (Thomas Munro, Melanie Plageman)

    A crash was possible given unlucky timing and parallel_leader_participation = off (which is not the default).

  • Recalculate GENERATED columns after an EvalPlanQual check (Tom Lane)

    In READ COMMITTED isolation mode, the effects of a row update might need to get reapplied to a newer version of the row than the query found originally. If so, we need to recompute any GENERATED columns, in case they depend on columns that were changed by the concurrent update.

  • Fix memory leak in Memoize plan execution (David Rowley)

  • Fix buffer refcount leak when using batched inserts for a foreign table included in a partitioned tree (Alexander Pyhalov)

  • Restore support for sub-millisecond vacuum_cost_delay settings (Thomas Munro)

  • Don't balance vacuum cost delay when a table has a per-relation vacuum_cost_delay setting of zero (Masahiko Sawada)

    Delay balancing is supposed to be disabled whenever autovacuum is processing a table with a per-relation vacuum_cost_delay setting, but this was done only for positive settings, not zero.

  • Fix corner-case crashes when columns have been added to the end of a view (Tom Lane)

  • Repair rare failure of MULTIEXPR_SUBLINK subplans in partitioned updates (Andres Freund, Tom Lane)

    Use of the syntax INSERT ... ON CONFLICT DO UPDATE SET (c1, ...) = (SELECT ...) with a partitioned target table could result in failure if any child table is dissimilar from the parent (for example, different physical column order). This typically manifested as failure of consistency checks in the executor; but a crash or incorrect data updates are also possible.

  • Fix handling of DEFAULT markers within a multi-row INSERT ... VALUES query on a view that has a DO ALSO INSERT ... SELECT rule (Dean Rasheed)

    Such cases typically failed with “unrecognized node type” errors or assertion failures.

  • Support references to OLD and NEW within subqueries in rule actions (Dean Rasheed, Tom Lane)

    Such references are really lateral references, but the server could crash if the subquery wasn't explicitly marked with LATERAL. Arrange to do that implicitly when necessary.

  • When decompiling a rule or SQL function body containing INSERT/UPDATE/DELETE within WITH, take care to print the correct alias for the target table (Tom Lane)

  • Fix glitches in SERIALIZABLE READ ONLY optimization (Thomas Munro)

    Transactions already marked as “doomed” confused the safe-snapshot optimization for SERIALIZABLE READ ONLY transactions. The optimization was unnecessarily skipped in some cases. In other cases an assertion failure occurred (but there was no problem in non-assert builds).

  • Avoid leaking cache callback slots in the pgoutput logical decoding plugin (Shi Yu)

    Multiple cycles of starting up and shutting down the plugin within a single session would eventually lead to an “out of relcache_callback_list slots” error.

  • Avoid unnecessary calls to custom validators for index operator class options (Alexander Korotkov)

    This change fixes some cases where an unexpected error was thrown.

  • Avoid useless work while scanning a multi-column BRIN index with multiple scan keys (Tomas Vondra)

    The existing code effectively considered only the last scan key while deciding whether a range matched, thus usually scanning more of the index than it needed to.

  • Fix netmask handling in BRIN inet_minmax_multi_ops opclass (Tomas Vondra)

    This error triggered an assertion failure in assert-enabled builds, but is mostly harmless in production builds.

  • Fix dereference of dangling pointer during buffering build of a GiST index (Alexander Lakhin)

    This error seems to usually be harmless in production builds, as the fetched value is noncritical; but in principle it could cause a server crash.

  • Ignore dropped columns and generated columns during logical replication of an update or delete action (Onder Kalaci, Shi Yu)

    Replication with the REPLICA IDENTITY FULL option failed if the table contained such columns.

  • Correct the name of the wait event for SLRU buffer I/O for commit timestamps (Alexander Lakhin)

    This wait event is named CommitTsBuffer according to the documentation, but the code had it as CommitTSBuffer. Change the code to match the documentation, as that way is more consistent with the naming of related wait events.

  • Re-activate reporting of wait event SLRUFlushSync (Thomas Munro)

    Reporting of this type of wait was accidentally removed in code refactoring.

  • Avoid possible underflow when calculating how many WAL segments to keep (Kyotaro Horiguchi)

    This could result in not honoring wal_keep_size accurately.

  • Disable startup progress reporting overhead in standby mode (Bharath Rupireddy)

    In standby mode, we don't actually report progress of recovery, but we were doing work to track it anyway.

  • Support RSA-PSS certificates with SCRAM-SHA-256 channel binding (Jacob Champion, Heikki Linnakangas)

    This feature requires building with OpenSSL 1.1.1 or newer. Both the server and libpq are affected.

  • Avoid race condition with process ID tracking on Windows (Thomas Munro)

    The operating system could recycle a PID before the postmaster observed that that child process was gone. This could lead to tracking more than one child with the same PID, resulting in confusion.

  • Fix list_copy_head() to work correctly on an empty List (David Rowley)

    This case is not known to be reached by any core PostgreSQL code, but extensions might rely on it working.

  • Add missing cases to SPI_result_code_string() (Dean Rasheed)

  • Fix erroneous Valgrind markings in AllocSetRealloc() (Karina Litskevich)

    In the unusual case where the size of a large (>8kB) palloc chunk is decreased, a Valgrind-aware build would mismark the defined-ness state of the memory released from the chunk, possibly causing incorrect results during Valgrind testing.

  • Fix assertion failure for MERGE into a partitioned table with row-level security enabled (Dean Rasheed)

  • Avoid assertion failure when decoding a transactional logical replication message (Tomas Vondra)

  • Avoid locale sensitivity when processing regular expression escapes (Jeff Davis)

    A backslash followed by a non-ASCII character could sometimes cause an assertion failure, depending on the prevailing locale.

  • Avoid trying to write an empty WAL record in log_newpage_range() when the last few pages in the specified range are empty (Matthias van de Meent)

    It is not entirely clear whether this case is reachable in released branches, but if it is then an assertion failure could occur.

  • Fix session-lifespan memory leakage in plpgsql DO blocks that use cast expressions (Ajit Awekar, Tom Lane)

  • Tighten array dimensionality checks when converting Perl list structures to multi-dimensional SQL arrays (Tom Lane)

    plperl could misbehave when the nesting of sub-lists is inconsistent so that the data does not represent a rectangular array of values. Such cases now produce errors, but previously they could result in a crash or garbage output.

  • Tighten array dimensionality checks when converting Python list structures to multi-dimensional SQL arrays (Tom Lane)

    plpython could misbehave when dealing with empty sub-lists, or when the nesting of sub-lists is inconsistent so that the data does not represent a rectangular array of values. The former should result in an empty output array, and the latter in an error. But some cases resulted in a crash, and others in unexpected output.

  • Fix unwinding of exception stack in plpython (Xing Guo)

    Some rare failure cases could return without cleaning up the PG_TRY exception stack, risking a crash if another error was raised before the next stack level was unwound.

  • Fix inconsistent GSS-encryption error handling in libpq's PQconnectPoll() (Michael Paquier)

    With gssencmode set to require, the connection was not marked dead after a GSS initialization failure. Make it fail immediately, as the equivalent case for TLS encryption has long done.

  • Fix possible data corruption in ecpg programs built with the -C ORACLE option (Kyotaro Horiguchi)

    When ecpg_get_data() is called with varcharsize set to zero, it could write a terminating zero character into the last byte of the preceding field, truncating the data in that field.

  • Fix pg_dump so that partitioned tables that are hash-partitioned on an enum-type column can be restored successfully (Tom Lane)

    Since the hash codes for enum values depend on the OIDs assigned to the enum, they are typically different after a dump and restore, meaning that rows often need to go into a different partition than they were in originally. Users can work around that by specifying the --load-via-partition-root option; but since there is very little chance of success without that, teach pg_dump to apply it automatically to such tables.

    Also, fix pg_restore to not try to TRUNCATE target tables before restoring into them when --load-via-partition-root mode is used. This avoids a hazard of deadlocks and lost data.

  • Correctly detect non-seekable files on Windows (Juan José Santamaría Flecha, Michael Paquier, Daniel Watzinger)

    This bug led to misbehavior when pg_dump writes to a pipe or pg_restore reads from one.

  • In pgbench's “prepared” mode, prepare all the commands in a pipeline before starting the pipeline (Álvaro Herrera)

    This avoids a failure when a pgbench script tries to start a serializable transaction inside a pipeline.

  • In contrib/amcheck's heap checking code, deal correctly with tuples having zero xmin or xmax (Robert Haas)

  • In contrib/amcheck, deal sanely with xids that appear to be before epoch zero (Andres Freund)

    In cases of corruption we might see a wrapped-around 32-bit xid that appears to be before the first xid epoch. Promoting such a value to 64-bit form produced a value far in the future, resulting in wrong reports. Return FirstNormalFullTransactionId in such cases so that things work reasonably sanely.

  • In contrib/basebackup_to_shell, properly detect failure to open a pipe (Robert Haas)

  • In contrib/hstore_plpython, avoid crashing if the Python value to be transformed isn't a mapping (Dmitry Dolgov, Tom Lane)

    This should give an error, but Python 3 changed some APIs in a way that caused the check to misbehave, allowing a crash to ensue.

  • Require the siglen option of a GiST index on an ltree column, if specified, to be a multiple of 4 (Alexander Korotkov)

    Other values result in misaligned accesses to index content, which is harmless on Intel-compatible hardware but can cause a crash on some other architectures.

  • In contrib/pageinspect, add defenses against incorrect input for the gist_page_items() function (Dmitry Koval)

  • Fix misbehavior in contrib/pg_trgm with an unsatisfiable regular expression (Tom Lane)

    A regex such as $foo is legal but unsatisfiable; the regex compiler recognizes that and produces an empty NFA graph. Attempting to optimize such a graph into a pg_trgm GIN or GiST index qualification resulted in accessing off the end of a work array, possibly leading to crashes.

  • Fix handling of escape sequences in contrib/postgres_fdw's application_name parameter (Kyotaro Horiguchi, Michael Paquier)

    The code to expand these could fail if executed in a background process, as for example during auto-analyze of a foreign table.

  • In contrib/pg_walinspect, limit memory usage of pg_get_wal_records_info() (Bharath Rupireddy)

  • Use the --strip-unneeded option when stripping static libraries with GNU-compatible strip (Tom Lane)

    Previously, make install-strip used the -x option in this case. This change avoids misbehavior of llvm-strip, and gives slightly smaller output as well.

  • Stop recommending auto-download of DTD files for building the documentation, and indeed disable it (Aleksander Alekseev, Peter Eisentraut, Tom Lane)

    It appears no longer possible to build the SGML documentation without a local installation of the DocBook DTD files. Formerly xsltproc could download those files on-the-fly from sourceforge.net; but sourceforge.net now permits only HTTPS access, and no common version of xsltproc supports that. Hence, remove the bits of our documentation suggesting that that's possible or useful, and instead add xsltproc's --nonet option to the build recipes.

  • When running TAP tests in PGXS builds, use a saner location for the temporary portlock directory (Peter Eisentraut)

    Place it under tmp_check in the build directory. With the previous coding, a PGXS build would try to place it in the installation directory, which is not necessarily writable.

  • Update time zone data files to tzdata release 2023c for DST law changes in Egypt, Greenland, Morocco, and Palestine.

    When observing Moscow time, Europe/Kirov and Europe/Volgograd now use the abbreviations MSK/MSD instead of numeric abbreviations, for consistency with other timezones observing Moscow time. Also, America/Yellowknife is no longer distinct from America/Edmonton; this affects some pre-1948 timestamps in that area.

• ⇑ Upgrade to 15.4 released on 2023-08-10 - docs

  • Disallow substituting a schema or owner name into an extension script if the name contains a quote, backslash, or dollar sign (Noah Misch)

    This restriction guards against SQL-injection hazards for trusted extensions.

    The PostgreSQL Project thanks Micah Gate, Valerie Woolard, Tim Carey-Smith, and Christoph Berg for reporting this problem. (CVE-2023-39417)

  • Fix MERGE to enforce row security policies properly (Dean Rasheed)

    When MERGE performs an UPDATE action, it should enforce any UPDATE or SELECT RLS policies defined on the target table, to be consistent with the way that a plain UPDATE with a WHERE clause works. Instead it was enforcing INSERT RLS policies for both INSERT and UPDATE actions.

    In addition, when MERGE performs a DO NOTHING action, it applied the target table's DELETE RLS policies to existing rows, even though those rows are not being deleted. While it's not a security problem, this could result in unwanted errors.

    The PostgreSQL Project thanks Dean Rasheed for reporting this problem. (CVE-2023-39418)

  • Fix confusion between empty (no rows) ranges and all-NULL ranges in BRIN indexes, as well as incorrect merging of all-NULL summaries (Tomas Vondra)

    Each of these oversights could result in forgetting that a BRIN index range contains any NULL values, potentially allowing subsequent queries that should return NULL values to miss doing so.

    This fix will not in itself correct faulty BRIN entries. It's recommended to REINDEX any BRIN indexes that may be used to search for nulls.

  • Avoid leaving a corrupted database behind when DROP DATABASE is interrupted (Andres Freund)

    If DROP DATABASE was interrupted after it had already begun taking irreversible steps, the target database remained accessible (because the removal of its pg_database row would roll back), but it would have corrupt contents. Fix by marking the database as inaccessible before we begin to perform irreversible operations. A failure after that will leave the database still partially present, but nothing can be done with it except to issue another DROP DATABASE.

  • Ensure that partitioned indexes are correctly marked as valid or not at creation (Michael Paquier)

    If a new partitioned index matches an existing but invalid index on one of the partitions, the partitioned index could end up being marked valid prematurely. This could lead to misbehavior or assertion failures in subsequent queries on the partitioned table.

  • Ignore invalid child indexes when matching partitioned indexes to child indexes during ALTER TABLE ATTACH PARTITION (Michael Paquier)

    Such an index will now be ignored, and a new child index created instead.

  • Fix possible failure when marking a partitioned index valid after all of its partitions have been attached (Michael Paquier)

    The update of the index's pg_index entry could use stale data for other columns. One reported symptom is an “attempted to update invisible tuple” error.

  • Fix ALTER EXTENSION SET SCHEMA to complain if the extension contains any objects outside the extension's schema (Michael Paquier, Heikki Linnakangas)

    Erroring out if the extension contains objects in multiple schemas was always intended; but the check was mis-coded so that it would fail to detect some cases, leading to surprising behavior.

  • Fix tracking of tables' access method dependencies (Michael Paquier)

    ALTER TABLE ... SET ACCESS METHOD failed to update relevant pg_depend entries when changing a table's access method. When using non-built-in access methods, this creates a risk that an access method could be dropped even though tables still depend on it. This fix corrects the logic in ALTER TABLE, but it will not adjust any already-missing pg_depend entries.

  • Don't use partial unique indexes for uniqueness proofs in the planner (David Rowley)

    This could give rise to incorrect plans, since the presumed uniqueness of rows read from a table might not hold if the index in question isn't used to scan the table.

  • Don't Memoize lateral joins with volatile join conditions (Richard Guo)

    Applying Memoize to a sub-plan that contains volatile filter conditions is likely to lead to wrong answers. The check to avoid doing this missed some cases that can arise when using LATERAL.

  • Avoid producing incorrect plans for foreign joins with pseudoconstant join clauses (Etsuro Fujita)

    The planner currently lacks support for attaching pseudoconstant join clauses to a pushed-down remote join, so disable generation of remote joins in such cases. (A better solution will require ABI-breaking changes of planner data structures, so it will have to wait for a future major release.)

  • Correctly handle sub-SELECTs in RLS policy expressions and security-barrier views when expanding rule actions (Tom Lane)

  • Fix race conditions in conflict detection for SERIALIZABLE isolation mode (Thomas Munro)

    Conflicts could be missed when using bitmap heap scans, when using GIN indexes, and when examining an initially-empty btree index. All these cases could lead to serializability failures due to improperly allowing conflicting transactions to commit.

  • Fix misbehavior of EvalPlanQual checks with inherited or partitioned target tables (Tom Lane)

    This oversight could lead to update or delete actions in READ COMMITTED isolation mode getting performed when they should have been skipped because of a conflicting concurrent update.

  • Fix hash join with an inner-side hash key that contains Params coming from an outer nested loop (Tom Lane)

    When rescanning the join after the values of such Params have changed, we must rebuild the hash table, but neglected to do so. This could result in missing join output rows.

  • Fix intermittent failures when trying to update a field of a composite column (Tom Lane)

    If the overall value of the composite column is wide enough to require out-of-line toasting, then an unluckily-timed cache flush could cause errors or server crashes.

  • Prevent query-lifespan memory leaks in some UPDATE queries with triggers (Tomas Vondra)

  • Prevent query-lifespan memory leaks when an Incremental Sort plan node is rescanned (James Coleman, Laurenz Albe, Tom Lane)

  • Accept fractional seconds in the input to jsonpath's datetime() method (Tom Lane)

  • Prevent stack-overflow crashes with very complex text search patterns (Tom Lane)

  • Allow tokens up to 10240 bytes long in pg_hba.conf and pg_ident.conf (Tom Lane)

    The previous limit of 256 bytes has been found insufficient for some use-cases.

  • Ensure that all existing placeholders are checked for matches when an extension declares its GUC prefix to be reserved (Karina Litskevich, Ekaterina Sokolova)

    Faulty loop logic could cause some entries to be skipped.

  • Fix mishandling of C++ out-of-memory conditions (Heikki Linnakangas)

    If JIT is in use, running out of memory in a C++ new call would lead to a PostgreSQL FATAL error, instead of the expected C++ exception.

  • Fix rare null-pointer crash in plancache.c (Tom Lane)

  • Avoid leaking a stats entry for a subscription when it is dropped (Masahiko Sawada)

  • Avoid losing track of possibly-useful shared memory segments when a page free results in coalescing ranges of free space (Dongming Liu)

    Ensure that the segment is moved into the appropriate “bin” for its new amount of free space, so that it will be found by subsequent searches.

  • Allow VACUUM to continue after detecting certain types of b-tree index corruption (Peter Geoghegan)

    If an invalid sibling-page link is detected, log the issue and press on, rather than throwing an error as before. Nothing short of REINDEX will fix the broken index, but preventing VACUUM from completing until that is done risks making matters far worse.

  • Ensure that WrapLimitsVacuumLock is released after VACUUM detects invalid data in pg_database.datfrozenxid or pg_database.datminmxid (Andres Freund)

    Failure to release this lock could lead to a deadlock later, although the lock would be cleaned up if the session exits or encounters some other error.

  • Avoid double replay of prepared transactions during crash recovery (suyu.cmj, Michael Paquier)

    After a crash partway through a checkpoint with some two-phase transaction state data already flushed to disk by this checkpoint, crash recovery could attempt to replay the prepared transaction(s) twice, leading to a fatal error such as “lock is already held” in the startup process.

  • Ensure that a newly created, but still empty table is fsync'ed at the next checkpoint (Heikki Linnakangas)

    Without this, if there is an operating system crash causing the empty file to disappear, subsequent operations on the table might fail with “could not open file” errors.

  • Ensure that creation of the init fork of an unlogged index is WAL-logged (Heikki Linnakangas)

    While an unlogged index's main data fork is not WAL-logged, its init fork should be, to ensure that we have a consistent state to restore the index to after a crash. This step was missed if the init fork contains no data, which is a case not used by any standard index AM; but perhaps some extension behaves that way.

  • Silence bogus “missing contrecord” errors (Thomas Munro)

    Treat this case as plain end-of-WAL to avoid logging inaccurate complaints from pg_waldump and walsender.

  • Fix overly strict assertion in jsonpath code (David Rowley)

    This assertion failed if a query applied the .type() operator to a like_regex result. There was no bug in non-assert builds.

  • Avoid assertion failure when processing an empty statement via the extended query protocol in an already-aborted transaction (Tom Lane)

  • Avoid assertion failure when the stats_fetch_consistency setting is changed intra-transaction (Kyotaro Horiguchi)

  • Fix contrib/fuzzystrmatch's Soundex difference() function to handle empty input sanely (Alexander Lakhin, Tom Lane)

    An input string containing no alphabetic characters resulted in unpredictable output.

  • Tighten whitespace checks in contrib/hstore input (Evan Jones)

    In some cases, characters would be falsely recognized as whitespace and hence discarded.

  • Disallow oversize input arrays with contrib/intarray's gist__int_ops index opclass (Ankit Kumar Pandey, Alexander Lakhin)

    Previously this code would report a NOTICE but press on anyway, creating an invalid index entry that presents a risk of crashes when the index is read.

  • Avoid useless double decompression of GiST index entries in contrib/intarray (Konstantin Knizhnik, Matthias van de Meent, Tom Lane)

  • Fix contrib/pageinspect's gist_page_items() function to work when there are included index columns (Alexander Lakhin, Michael Paquier)

    Previously, if the index has included columns, gist_page_items() would fail to display those values on index leaf pages, or crash outright on non-leaf pages.

  • In psql, ignore the PSQL_WATCH_PAGER environment variable when stdin/stdout are not a terminal (Tom Lane)

    This corresponds to the treatment of PSQL_PAGER in commands besides \watch.

  • Fix pg_dump to correctly handle new-style SQL-language functions whose bodies require parse-time dependencies on unique indexes (Tom Lane)

    Such cases can arise from GROUP BY and ON CONFLICT clauses, for example. The function must then be postponed until after the unique index in the dump output, but pg_dump did not do that and instead printed a warning about “could not resolve dependency loop”.

  • Improve pg_dump's display of details about dependency-loop problems (Tom Lane)

  • Avoid crash in pgbench with an empty pipeline and prepared mode (Álvaro Herrera)

  • Ensure that pg_index.indisreplident is kept up-to-date in relation cache entries (Shruthi Gowda)

    This value could be stale in some cases. There is no core code that relies on the relation cache's copy, so this is only a latent bug as far as Postgres itself is concerned; but there may be extensions for which it is a live bug.

  • Fix make_etags script to work with non-Exuberant ctags (Masahiko Sawada)

• ⇑ Upgrade to 15.5 released on 2023-11-09 - docs

  • Fix handling of unknown-type arguments in DISTINCT "any" aggregate functions (Tom Lane)

    This error led to a text-type value being interpreted as an unknown-type value (that is, a zero-terminated string) at runtime. This could result in disclosure of server memory following the text value.

    The PostgreSQL Project thanks Jingzhou Fu for reporting this problem. (CVE-2023-5868)

  • Detect integer overflow while computing new array dimensions (Tom Lane)

    When assigning new elements to array subscripts that are outside the current array bounds, an undetected integer overflow could occur in edge cases. Memory stomps that are potentially exploitable for arbitrary code execution are possible, and so is disclosure of server memory.

    The PostgreSQL Project thanks Pedro Gallegos for reporting this problem. (CVE-2023-5869)

  • Prevent the pg_signal_backend role from signalling background workers and autovacuum processes (Noah Misch, Jelte Fennema-Nio)

    The documentation says that pg_signal_backend cannot issue signals to superuser-owned processes. It was able to signal these background processes, though, because they advertise a role OID of zero. Treat that as indicating superuser ownership. The security implications of cancelling one of these process types are fairly small so far as the core code goes (we'll just start another one), but extensions might add background workers that are more vulnerable.

    Also ensure that the is_superuser parameter is set correctly in such processes. No specific security consequences are known for that oversight, but it might be significant for some extensions.

    The PostgreSQL Project thanks Hemanth Sandrana and Mahendrakar Srinivasarao for reporting this problem. (CVE-2023-5870)

  • Fix misbehavior during recursive page split in GiST index build (Heikki Linnakangas)

    Fix a case where the location of a page downlink was incorrectly tracked, and introduce some logic to allow recovering from such situations rather than silently doing the wrong thing. This error could result in incorrect answers from subsequent index searches. It may be advisable to reindex all GiST indexes after installing this update.

  • Prevent de-duplication of btree index entries for interval columns (Noah Misch)

    There are interval values that are distinguishable but compare equal, for example 24:00:00 and 1 day. This breaks assumptions made by btree de-duplication, so interval columns need to be excluded from de-duplication. This oversight can cause incorrect results from index-only scans. Moreover, after updating amcheck will report an error for almost all such indexes. Users should reindex any btree indexes on interval columns.

  • Process date values more sanely in BRIN datetime_minmax_multi_ops indexes (Tomas Vondra)

    The distance calculation for dates was backward, causing poor decisions about which entries to merge. The index still produces correct results, but is much less efficient than it should be. Reindexing BRIN minmax_multi indexes on date columns is advisable.

  • Process large timestamp and timestamptz values more sanely in BRIN datetime_minmax_multi_ops indexes (Tomas Vondra)

    Infinities were mistakenly treated as having distance zero rather than a large distance from other values, causing poor decisions about which entries to merge. Also, finite-but-very-large values (near the endpoints of the representable timestamp range) could result in internal overflows, again causing poor decisions. The index still produces correct results, but is much less efficient than it should be. Reindexing BRIN minmax_multi indexes on timestamp and timestamptz columns is advisable if the column contains, or has contained, infinities or large finite values.

  • Avoid calculation overflows in BRIN interval_minmax_multi_ops indexes with extreme interval values (Tomas Vondra)

    This bug might have caused unexpected failures while trying to insert large interval values into such an index.

  • Fix partition step generation and runtime partition pruning for hash-partitioned tables with multiple partition keys (David Rowley)

    Some cases involving an IS NULL condition on one of the partition keys could result in a crash.

  • Fix inconsistent rechecking of concurrently-updated rows during MERGE (Dean Rasheed)

    In READ COMMITTED mode, an update that finds that its target row was just updated by a concurrent transaction will recheck the query's WHERE conditions on the updated row. MERGE failed to ensure that the proper rows of other joined tables were used during this recheck, possibly resulting in incorrect decisions about whether the newly-updated row should be updated again by MERGE.

  • Correctly identify the target table in an inherited UPDATE/DELETE/MERGE even when the parent table is excluded by constraints (Amit Langote, Tom Lane)

    If the initially-named table is excluded by constraints, but not all its inheritance descendants are, the first non-excluded descendant was identified as the primary target table. This would lead to firing statement-level triggers associated with that table, rather than the initially-named table as should happen. In v16, the same oversight could also lead to “invalid perminfoindex 0 in RTE with relid NNNN” errors.

  • Fix edge case in btree mark/restore processing of ScalarArrayOpExpr clauses (Peter Geoghegan)

    When restoring an indexscan to a previously marked position, the code could miss required setup steps if the scan had advanced exactly to the end of the matches for a ScalarArrayOpExpr (that is, an indexcol = ANY(ARRAY[])) clause. This could result in missing some rows that should have been fetched.

  • Fix intra-query memory leak in Memoize execution (Orlov Aleksej, David Rowley)

  • Fix intra-query memory leak when a set-returning function repeatedly returns zero rows (Tom Lane)

  • Don't crash if cursor_to_xmlschema() is applied to a non-data-returning Portal (Boyu Yang)

  • Throw the intended error if pgrowlocks() is applied to a partitioned table (David Rowley)

    Previously, a not-on-point complaint “only heap AM is supported” would be raised.

  • Handle invalid indexes more cleanly in assorted SQL functions (Noah Misch)

    Report an error if pgstatindex(), pgstatginindex(), pgstathashindex(), or pgstattuple() is applied to an invalid index. If brin_desummarize_range(), brin_summarize_new_values(), brin_summarize_range(), or gin_clean_pending_list() is applied to an invalid index, do nothing except to report a debug-level message. Formerly these functions attempted to process the index, and might fail in strange ways depending on what the failed CREATE INDEX had left behind.

  • Fix pg_stat_reset_single_table_counters() to do the right thing for a shared catalog (Masahiro Ikeda)

    Previously the reset would be ineffective.

  • Avoid premature memory allocation failure with long inputs to to_tsvector() (Tom Lane)

  • Fix over-allocation of the constructed tsvector in tsvectorrecv() (Denis Erokhin)

    If the incoming vector includes position data, the binary receive function left wasted space (roughly equal to the size of the position data) in the finished tsvector. In extreme cases this could lead to “maximum total lexeme length exceeded” failures for vectors that were under the length limit when emitted. In any case it could lead to wasted space on-disk.

  • Fix incorrect coding in gtsvector_picksplit() (Alexander Lakhin)

    This could lead to poor page-split decisions in GiST indexes on tsvector columns.

  • Improve checks for corrupt PGLZ compressed data (Flavien Guedez)

  • In COPY FROM, fail cleanly when an unsupported encoding conversion is needed (Tom Lane)

    Recent refactoring accidentally removed the intended error check for this, such that it ended in “cache lookup failed for function 0” instead of a useful error message.

  • Avoid crash in EXPLAIN if a parameter marked to be displayed by EXPLAIN has a NULL boot-time value (Xing Guo, Aleksander Alekseev, Tom Lane)

    No built-in parameter fits this description, but an extension could define such a parameter.

  • Ensure we have a snapshot while dropping ON COMMIT DROP temp tables (Tom Lane)

    This prevents possible misbehavior if any catalog entries for the temp tables have fields wide enough to require toasting (such as a very complex CHECK condition).

  • Avoid improper response to shutdown signals in child processes just forked by system() (Nathan Bossart)

    This fix avoids a race condition in which a child process that has been forked off by system(), but hasn't yet exec'd the intended child program, might receive and act on a signal intended for the parent server process. That would lead to duplicate cleanup actions being performed, which will not end well.

  • Cope with torn reads of pg_control in frontend programs (Thomas Munro)

    On some file systems, reading pg_control may not be an atomic action when the server concurrently writes that file. This is detectable via a bad CRC. Retry a few times to see if the file becomes valid before we report error.

  • Avoid torn reads of pg_control in relevant SQL functions (Thomas Munro)

    Acquire the appropriate lock before reading pg_control, to ensure we get a consistent view of that file.

  • Avoid integer overflow when computing size of backend activity string array (Jakub Wartak)

    On 64-bit machines we will allow values of track_activity_query_size large enough to cause 32-bit overflow when multiplied by the allowed number of connections. The code actually allocating the per-backend local array was careless about this though, and allocated the array incorrectly.

  • Fix briefly showing inconsistent progress statistics for ANALYZE on inherited tables (Heikki Linnakangas)

    The block-level counters should be reset to zero at the same time we update the current-relation field.

  • Fix the background writer to report any WAL writes it makes to the statistics counters (Nazir Bilal Yavuz)

  • Fix confusion about forced-flush behavior in pgstat_report_wal() (Ryoga Yoshida, Michael Paquier)

    This could result in some statistics about WAL I/O being forgotten in a shutdown.

  • Track the dependencies of cached CALL statements, and re-plan them when needed (Tom Lane)

    DDL commands, such as replacement of a function that has been inlined into a CALL argument, can create the need to re-plan a CALL that has been cached by PL/pgSQL. That was not happening, leading to misbehavior or strange errors such as “cache lookup failed”.

  • Avoid a possible pfree-a-NULL-pointer crash after an error in OpenSSL connection setup (Sergey Shinderuk)

  • Track nesting depth correctly when inspecting RECORD-type Vars from outer query levels (Richard Guo)

    This oversight could lead to assertion failures, core dumps, or “bogus varno” errors.

  • Track hash function and negator function dependencies of ScalarArrayOpExpr plan nodes (David Rowley)

    In most cases this oversight was harmless, since these functions would be unlikely to disappear while the node's original operator remains present.

  • Fix error-handling bug in RECORD type cache management (Thomas Munro)

    An out-of-memory error occurring at just the wrong point could leave behind inconsistent state that would lead to an infinite loop.

  • Fix assertion failure when logical decoding is retried in the same session after an error (Hou Zhijie)

  • Treat out-of-memory failures as fatal while reading WAL (Michael Paquier)

    Previously this would be treated as a bogus-data condition, leading to the conclusion that we'd reached the end of WAL, which is incorrect and could lead to inconsistent WAL replay.

  • Fix possible recovery failure due to trying to allocate memory based on a bogus WAL record length field (Thomas Munro, Michael Paquier)

  • Fix race condition in database dropping that could lead to the autovacuum launcher getting stuck (Andres Freund, Will Mortensen, Jacob Speidel)

    The race could lead to a statistics entry for the removed database remaining present, confusing the launcher's selection of which database to process.

  • Fix datatype size confusion in logical tape management (Ranier Vilela)

    Integer overflow was possible on platforms where long is wider than int, although it would take a multiple-terabyte temporary file to cause a problem.

  • Avoid unintended close of syslogger process's stdin (Heikki Linnakangas)

  • Avoid doing plan cache revalidation of utility statements that do not receive interesting processing during parse analysis (Tom Lane)

    Aside from saving a few cycles, this prevents failure after a cache invalidation for statements that must not set a snapshot, such as SET TRANSACTION ISOLATION LEVEL.

  • Keep by-reference attmissingval values in a long-lived context while they are being used (Andrew Dunstan)

    This avoids possible use of dangling pointers when a tuple slot outlives the tuple descriptor with which its value was constructed.

  • Recalculate the effective value of search_path after ALTER ROLE (Jeff Davis)

    This ensures that after renaming a role, the meaning of the special string $user is re-determined.

  • Fix “could not duplicate handle” error occurring on Windows when min_dynamic_shared_memory is set above zero (Thomas Munro)

  • Fix order of operations in GenericXLogFinish (Jeff Davis)

    This code violated the conditions required for crash safety by writing WAL before marking changed buffers dirty. No core code uses this function, but extensions do (contrib/bloom does, for example).

  • Remove incorrect assertion in PL/Python exception handling (Alexander Lakhin)

  • Fix assertion failure in pg_dump when it's asked to dump the pg_catalog schema (Peter Eisentraut)

  • Fix pg_restore so that selective restores will include both table-level and column-level ACLs for selected tables (Euler Taveira, Tom Lane)

    Formerly, only the table-level ACL would get restored if both types were present.

  • Add logic to pg_upgrade to check for use of abstime, reltime, and tinterval data types (Álvaro Herrera)

    These obsolete data types were removed in PostgreSQL version 12, so check to make sure they aren't present in an older database before claiming it can be upgraded.

  • Avoid generating invalid temporary slot names in pg_basebackup (Jelte Fennema)

    This has only been seen to occur when the server connection runs through pgbouncer.

  • Avoid false “too many client connections” errors in pgbench on Windows (Noah Misch)

  • In contrib/amcheck, do not report interrupted page deletion as corruption (Noah Misch)

    This fix prevents false-positive reports of “the first child of leftmost target page is not leftmost of its level”, “block NNNN is not leftmost” or “left link/right link pair in index XXXX not in agreement”. They appeared if amcheck ran after an unfinished btree index page deletion and before VACUUM had cleaned things up.

  • Fix failure of contrib/btree_gin indexes on interval columns, when an indexscan using the < or <= operator is performed (Dean Rasheed)

    Such an indexscan failed to return all the entries it should.

  • Add support for LLVM 16 and 17 (Thomas Munro, Dmitry Dolgov)

  • Suppress assorted build-time warnings on recent macOS (Tom Lane)

    Xcode 15 (released with macOS Sonoma) changed the linker's behavior in a way that causes many duplicate-library warnings while building PostgreSQL. These were harmless, but they're annoying so avoid citing the same libraries twice. Also remove use of the -multiply_defined suppress linker switch, which apparently has been a no-op for a long time, and is now actively complained of.

  • When building contrib/unaccent's rules file, fall back to using python if --with-python was not given and make variable PYTHON was not set (Japin Li)

  • Remove PHOT (Phoenix Islands Time) from the default timezone abbreviations list (Tom Lane)

    Presence of this abbreviation in the default list can cause failures on recent Debian and Ubuntu releases, as they no longer install the underlying tzdb entry by default. Since this is a made-up abbreviation for a zone with a total human population of about two dozen, it seems unlikely that anyone will miss it. If someone does, they can put it back via a custom abbreviations file.

• ⇑ Upgrade to 15.6 released on 2024-02-08 - docs

  • Tighten security restrictions within REFRESH MATERIALIZED VIEW CONCURRENTLY (Heikki Linnakangas)

    One step of a concurrent refresh command was run under weak security restrictions. If a materialized view's owner could persuade a superuser or other high-privileged user to perform a concurrent refresh on that view, the view's owner could control code executed with the privileges of the user running REFRESH. Fix things so that all user-determined code is run as the view's owner, as expected.

    The only known exploit for this error does not work in PostgreSQL 16.0 and later, so it may be that v16 is not vulnerable in practice.

    The PostgreSQL Project thanks Pedro Gallegos for reporting this problem. (CVE-2024-0985)

  • Fix memory leak when performing JIT inlining (Andres Freund, Daniel Gustafsson)

    There have been multiple reports of backend processes suffering out-of-memory conditions after sufficiently many JIT compilations. This fix should resolve that.

  • When dequeueing from an LWLock, avoid needing to search the list of waiting processes (Andres Freund)

    This fixes O(N^2) behavior when the list of waiters is long. In some use-cases this results in substantial throughput improvements.

  • Avoid generating incorrect partitioned-join plans (Richard Guo)

    Some uncommon situations involving lateral references could create incorrect plans. Affected queries could produce wrong answers, or odd failures such as “variable not found in subplan target list”, or executor crashes.

  • Fix incorrect wrapping of subquery output expressions in PlaceHolderVars (Tom Lane)

    This fixes incorrect results when a subquery is underneath an outer join and has an output column that laterally references something outside the outer join's scope. The output column might not appear as NULL when it should do so due to the action of the outer join.

  • Fix misprocessing of window function run conditions (Richard Guo)

    This oversight could lead to “WindowFunc not found in subplan target lists” errors.

  • Skip inappropriate actions when MERGE causes a cross-partition update (Dean Rasheed)

    When executing a MERGE UPDATE action on a partitioned table, if the UPDATE is turned into a DELETE and INSERT due to changing a partition key column, skip firing AFTER UPDATE ROW triggers, as well as other post-update actions such as RLS checks. These actions would typically fail, which is why a regular UPDATE doesn't do them in such cases; MERGE shouldn't either.

  • Cope with BEFORE ROW DELETE triggers in cross-partition MERGE updates (Dean Rasheed)

    If such a trigger attempted to prevent the update by returning NULL, MERGE would suffer an error or assertion failure.

  • Prevent access to a no-longer-pinned buffer in BEFORE ROW UPDATE triggers (Alexander Lakhin, Tom Lane)

    If the tuple being updated had just been updated and moved to another page by another session, there was a narrow window where we would attempt to fetch data from the new tuple version without any pin on its buffer. In principle this could result in garbage data appearing in non-updated columns of the proposed new tuple. The odds of problems in practice seem rather low, however.

  • Avoid requesting an oversize shared-memory area in parallel hash join (Thomas Munro, Andrei Lepikhov, Alexander Korotkov)

    The limiting value was too large, allowing “invalid DSA memory alloc request size” errors to occur with sufficiently large expected hash table sizes.

  • Avoid assertion failures in heap_update() and heap_delete() when a tuple to be updated by a foreign-key enforcement trigger fails the extra visibility crosscheck (Alexander Lakhin)

    This error had no impact in non-assert builds.

  • Fix possible failure during ALTER TABLE ADD COLUMN on a complex inheritance tree (Tender Wang)

    If a grandchild table would inherit the new column via multiple intermediate parents, the command failed with “tuple already updated by self”.

  • Fix problems with duplicate token names in ALTER TEXT SEARCH CONFIGURATION ... MAPPING commands (Tender Wang, Michael Paquier)

  • Properly lock the associated table during DROP STATISTICS (Tomas Vondra)

    Failure to acquire the lock could result in “tuple concurrently deleted” errors if the DROP executes concurrently with ANALYZE.

  • Fix function volatility checking for GENERATED and DEFAULT expressions (Tom Lane)

    These places could fail to detect insertion of a volatile function default-argument expression, or decide that a polymorphic function is volatile although it is actually immutable on the datatype of interest. This could lead to improperly rejecting or accepting a GENERATED clause, or to mistakenly applying the constant-default-value optimization in ALTER TABLE ADD COLUMN.

  • Detect that a new catalog cache entry became stale while detoasting its fields (Tom Lane)

    We expand any out-of-line fields in a catalog tuple before inserting it into the catalog caches. That involves database access which might cause invalidation of catalog cache entries — but the new entry isn't in the cache yet, so we would miss noticing that it should get invalidated. The result is a race condition in which an already-stale cache entry could get made, and then persist indefinitely. This would lead to hard-to-predict misbehavior. Fix by rechecking the tuple's visibility after detoasting.

  • Fix edge-case integer overflow detection bug on some platforms (Dean Rasheed)

    Computing 0 - INT64_MIN should result in an overflow error, and did on most platforms. However, platforms with neither integer overflow builtins nor 128-bit integers would fail to spot the overflow, instead returning INT64_MIN.

  • Detect Julian-date overflow when adding or subtracting an interval to/from a timestamp (Tom Lane)

    Some cases that should cause an out-of-range error produced an incorrect result instead.

  • Add more checks for overflow in interval_mul() and interval_div() (Dean Rasheed)

    Some cases that should cause an out-of-range error produced an incorrect result instead.

  • Ensure cached statistics are discarded after a change to stats_fetch_consistency (Shinya Kato)

    In some code paths, it was possible for stale statistics to be returned.

  • Make the pg_file_settings view check validity of unapplied values for settings with backend or superuser-backend context (Tom Lane)

    Invalid values were not noted in the view as intended. This escaped detection because there are very few settings in these groups.

  • Match collation too when matching an existing index to a new partitioned index (Peter Eisentraut)

    Previously we could accept an index that has a different collation from the corresponding element of the partition key, possibly leading to misbehavior.

  • Avoid failure if a child index is dropped concurrently with REINDEX INDEX on a partitioned index (Fei Changhong)

  • Fix insufficient locking when cleaning up an incomplete split of a GIN index's internal page (Fei Changhong, Heikki Linnakangas)

    The code tried to do this with shared rather than exclusive lock on the buffer. This could lead to index corruption if two processes attempted the cleanup concurrently.

  • Avoid premature release of buffer pin in GIN index insertion (Tom Lane)

    If an index root page split occurs concurrently with our own insertion, the code could fail with “buffer NNNN is not owned by resource owner”.

  • Avoid failure with partitioned SP-GiST indexes (Tom Lane)

    Trying to use an index of this kind could lead to “No such file or directory” errors.

  • Fix ownership change reporting for large objects (Tom Lane)

    A no-op ALTER LARGE OBJECT OWNER command (that is, one selecting the existing owner) passed the wrong class ID to the PostAlterHook, probably confusing any extension using that hook.

  • Fix reporting of I/O timing data in EXPLAIN (BUFFERS) (Michael Paquier)

    The numbers labeled as “shared/local” actually refer only to shared buffers, so change that label to “shared”.

  • Ensure durability of CREATE DATABASE (Noah Misch)

    If an operating system crash occurred during or shortly after CREATE DATABASE, recovery could fail, or subsequent connections to the new database could fail. If a base backup was taken in that window, similar problems could be observed when trying to use the backup. The symptom would be that the database directory, PG_VERSION file, or pg_filenode.map file was missing or empty.

  • Add more LOG messages when starting and ending recovery from a backup (Andres Freund)

    This change provides additional information in the postmaster log that may be useful for diagnosing recovery problems.

  • Prevent standby servers from incorrectly processing dead index tuples during subtransactions (Fei Changhong)

    The startedInRecovery flag was not correctly set for a subtransaction. This affects only processing of dead index tuples. It could allow a query in a subtransaction to ignore index entries that it should return (if they are already dead on the primary server, but not dead to the standby transaction), or to prematurely mark index entries as dead that are not yet dead on the primary. It is not clear that the latter case has any serious consequences, but it's not the intended behavior.

  • Fix integer overflow hazard in checking whether a record will fit into the WAL decoding buffer (Thomas Munro)

    This bug appears to be only latent except when running a 32-bit PostgreSQL build on a 64-bit platform.

  • Fix deadlock between a logical replication apply worker, its tablesync worker, and a session process trying to alter the subscription (Shlok Kyal)

    One edge of the deadlock loop did not involve a lock wait, so the deadlock went undetected and would persist until manual intervention.

  • Ensure that column default values are correctly transmitted by the pgoutput logical replication plugin (Nikhil Benesch)

    ALTER TABLE ADD COLUMN with a constant default value for the new column avoids rewriting existing tuples, instead expecting that reading code will insert the correct default into a tuple that lacks that column. If replication was subsequently initiated on the table, pgoutput would transmit NULL instead of the correct default for such a column, causing incorrect replication on the subscriber.

  • Fix failure of logical replication's initial sync for a table with no columns (Vignesh C)

    This case generated an improperly-formatted COPY command.

  • Prevent examining system catalogs with the wrong snapshot during logical decoding (Fei Changhong)

    If decoding begins partway through a transaction that modifies system catalogs, the decoder may not recognize that, causing it to fail to treat that transaction as in-progress for catalog lookups. This fix deals with the case that a top-level transaction is already marked as containing catalog changes, but its subtransaction(s) are not.

  • Return the correct status code when a new client disconnects without responding to the server's password challenge (Liu Lang, Tom Lane)

    In some cases we'd treat this as a loggable error, which was not the intention and tends to create log spam, since common clients like psql frequently do this. It may also confuse extensions that use ClientAuthentication_hook.

  • Fix incompatibility with OpenSSL 3.2 (Tristan Partin, Bo Andreson)

    Use the BIO “app_data” field for our private storage, instead of assuming it's okay to use the “data” field. This mistake didn't cause problems before, but with 3.2 it leads to crashes and complaints about double frees.

  • Be more wary about OpenSSL not setting errno on error (Tom Lane)

    If errno isn't set, assume the cause of the reported failure is read EOF. This fixes rare cases of strange error reports like “could not accept SSL connection: Success”.

  • Fix file descriptor leakage when a foreign data wrapper's ForeignAsyncRequest function fails (Heikki Linnakangas)

  • Report ENOMEM errors from file-related system calls as ERRCODE_OUT_OF_MEMORY, not ERRCODE_INTERNAL_ERROR (Alexander Kuzmenkov)

  • In PL/pgSQL, support SQL commands that are CREATE FUNCTION/CREATE PROCEDURE with SQL-standard bodies (Tom Lane)

    Previously, such cases failed with parsing errors due to the semicolon(s) appearing in the function body.

  • Fix libpq's handling of errors in pipelines (Álvaro Herrera)

    The pipeline state could get out of sync if an error is returned for reasons other than a query problem (for example, if the connection is lost). Potentially this would lead to a busy-loop in the calling application.

  • Make libpq's PQsendFlushRequest() function flush the client output buffer under the same rules as other PQsend functions (Jelte Fennema-Nio)

    In pipeline mode, it may still be necessary to call PQflush() as well; but this change removes some inconsistency.

  • Avoid race condition when libpq initializes OpenSSL support concurrently in two different threads (Willi Mann, Michael Paquier)

  • Fix timing-dependent failure in GSSAPI data transmission (Tom Lane)

    When using GSSAPI encryption in non-blocking mode, libpq sometimes failed with “GSSAPI caller failed to retransmit all data needing to be retried”.

  • In pg_dump, don't dump RLS policies or security labels for extension member objects (Tom Lane, Jacob Champion)

    Previously, commands would be included in the dump to set these properties, which is really incorrect since they should be considered as internal affairs of the extension. Moreover, the restoring user might not have adequate privilege to set them, and indeed the dumping user might not have enough privilege to dump them (since dumping RLS policies requires acquiring lock on their table).

  • In pg_dump, don't dump an extended statistics object if its underlying table isn't being dumped (Rian McGuire, Tom Lane)

    This conforms to the behavior for other dependent objects such as indexes.

  • Make it an error for a pgbench script to end with an open pipeline (Anthonin Bonnefoy)

    Previously, pgbench would behave oddly if a \startpipeline command lacked a matching \endpipeline. This seems like a scripting mistake rather than a case that pgbench needs to handle nicely, so throw an error.

  • In contrib/bloom, fix overly tight assertion about false_positive_rate (Alexander Lakhin)

  • Fix crash in contrib/intarray if an array with an element equal to INT_MAX is inserted into a gist__int_ops index (Alexander Lakhin, Tom Lane)

  • Report a better error when contrib/pageinspect's hash_bitmap_info() function is applied to a partitioned hash index (Alexander Lakhin, Michael Paquier)

  • Report a better error when contrib/pgstattuple's pgstathashindex() function is applied to a partitioned hash index (Alexander Lakhin)

  • On Windows, suppress autorun options when launching subprocesses in pg_ctl and pg_regress (Kyotaro Horiguchi)

    When launching a child process via cmd.exe, pass the /D flag to prevent executing any autorun commands specified in the registry. This avoids possibly-surprising side effects.

  • Move is_valid_ascii() from mb/pg_wchar.h to utils/ascii.h (Jubilee Young)

    This change avoids the need to include <simd.h> in pg_wchar.h, which was causing problems for some third-party code.

  • Fix compilation failures with libxml2 version 2.12.0 and later (Tom Lane)

  • Fix compilation failure of WAL_DEBUG code on Windows (Bharath Rupireddy)

  • Suppress compiler warnings from Python's header files (Peter Eisentraut, Tom Lane)

    Our preferred compiler options provoke warnings about constructs appearing in recent versions of Python's header files. When using gcc, we can suppress these warnings with a pragma.

  • Avoid deprecation warning when compiling with LLVM 18 (Thomas Munro)

  • Update time zone data files to tzdata release 2024a for DST law changes in Greenland, Kazakhstan, and Palestine, plus corrections for the Antarctic stations Casey and Vostok. Also historical corrections for Vietnam, Toronto, and Miquelon.

• source repo
• author info
• meta-info