PostgreSQL 12 Open Items
From PostgreSQL wiki
Jump to navigationJump to searchOpen Issues
NOTE: Please add new open items to the bottom of the list.
- Should we change jsonpath to avoid using standard syntax for not-quite-standard features?
- Additional discussion here
- After recovery starting when backup_label exists, but not recovery.signal shuts down, the restarted recovery doesn't enter an archive recovery mode and may fail.
- Interaction of recovery target options with recovery.signal/standby.signal does not match documentation and includes some surprising behaviors.
Older Bugs
Live issues
- ATPostAlterTypeCleanup causes child indexes to be recreated with wrong relfilenode
- Crash/data corruption is fixed by 02c359eeda50a71c951371c9d3e920ff8f514008
- There's more to be done here, but it's not clear whether additional work is small enough to be in-scope for v11 or v12
- parallel queries failed ERROR: invalid name syntax CONTEXT: parallel worker
- BUG #15746: cache lookup failed for function in plpgsql block
- This was already fixed in HEAD/v12 by a part of 04fe805a1734eccd8dcdd34c8cc0ddcb62c7240c
- Issue is whether it's worth the risk to back-patch unproven code
- Change resowner cleanup order for Windows?
- B-Tree Page can become unrecyclable due to PageDeleteXid wraparound
- This was fixed in the new GiST page deletion by 6655a7299d835dea9e8e0ba69cc5284611b96f29, but the same issue remains in B-tree.
- Undesirable PANIC on snapshot file open failure
- v10: memory leak in reorderbuffer having large changes
Fixed issues
- getting ERROR "relation 16401 has no triggers" with partition foreign key alter
- Fixed in: 5562272a4229cfa57354aa203cffd36b4e7f70cb
- pg_upgrade from 10-or-earlier: TRAP: FailedAssertion(»!(metad->btm_version >= 3)
- Fixed in: d004147eb3ece6b5981dbdd3d918ffc3f23fc505
- Bypass processing of wraparound autovacuums not marked as aggressive
- Problem exists since the point where aggressive vacuums have been introduced, v12 has only added extra logs to look after the impossible case of wraparound autovacuums not aggressive.
- Fixed in: 2aa6e331ead7f3ad080561495ad4bd3bc7cd8913
- An insert destined at partition created after a column has been dropped from the parent table fails
- Fixed in: 6b0208ebc436b33bd80ce264299b4b1b8d59b68a
- RelationData.rd_partcheck should get its own memory context
- Fixed in: 5f1433ac5e7f943b29ef01266b6b8fc915e6b917
- Walsender process crashing when executing SHOW ALL
- Fixed in: c34677fdaa73f089d557554a9cd479b9bd5b5143
- Possible to store invalid SCRAM-SHA-256 Passwords
- Fixed in: ccae190b916f27fbe4079ee4664d34cd1be47b79
- BUG #15781: subselect on foreign table (postgres_fdw) can crash
- Fixed in: 8cad5adb9c0be82e9f40d51b02a542439f47de9e
- selecting from partition directly can't use constraint exclusion
- Fixed in: e03ff739695cb731956763355e8e0f38c6905008
- REINDEX INDEX on an index of pg_class can fail
- Fixed in: f912d7dec29341d55315fccef8dc3fdfd068c6e3
- Partition pruning is broken for stable WHERE conditions
- Fixed in: 6630ccad7a25cad32e2d1a6833fb971602cb67fe and predecessor commits
- Statistical aggregate functions are not working with partitionwise aggregate
- Fixed in: 2657283256f1cab53d09d2c7db1ce9b7065193a0
- Race conditions with checkpointer and shutdown
- Fixed in a1a789eb5ac894b4ca4b7742f2dc2d9602116e46
- Back-patched to v10; back-patching further is unattractive from both risk and work-required standpoints
- Statistics with inheritance cause ANALYZE failures
- Change semantics of log_statement_sample_rate?
- REL_12_STABLE crashing with assertion failure in ExtractReplicaIdentity
- The report mentions REL_12_STABLE, but it's a pre-existing issue just made visible by new checks. Nonetheless, it's not clear that there are real consequences beyond the assertion failure, so no change in prior branches.
- Fixed in f63a5ead9d04467e1c1847bd5e3d87c4dca6cd35
Nothing to do
- toast_tuple_target reloption doesn't work as expected
- The consensus would be to increase the upper boundary of toast_tuple_target, but this means potentially breaking a category of dumps.
Non-bugs
Resolved Issues
resolved before 12rc1
- ECPG DECLARE STATEMENT definitional and quality-of-implementation concerns
- Reverted by 96b6c82c9dd4a6a91c7e54bf42d36da111959ec6
- ecpglib major version change
- Reverted by fbd1cb2fd9796a08a198d929774280ecce933057
- Leakproofness of texteq()/textne()
- Nondeterministic collations break text_pattern_ops
resolved before 12beta4
- EvalPlanQual crash due to wrong slot tupdesc
- Commit ad0bda5d24ea2bcc72b5e50020e3c79bab10836b Author: Andres Freund
- Fixed in 27cc7cd2bc8a5e8efc8279bc5d2a8ae42fd8ad33
- BUG #15977: Inconsistent behavior in chained transactions
- behavior of AND CHAIN on implicit transactions is not consistent
- Commit 280a408b48d5ee42969f981bceb9e9426c3a344c Author: Peter Eisentraut
- fix ready, consisting in disabling chaining when in an implicit transaction
- Fixed in 862ef372d6b23629f93d4afc123ddd7d172501ac
- Debate INFO messages in ATTACH PARTITION and SET NOT NULL
- Proposed patch is in thread Change ereport level for QueuePartitionConstraintValidation
- Fixed in db438318997b75f4b40c61258da56384039fa43f
- Images in documentation don't work in VPATH
- Fixed in 842ac79407a136
- default_table_access_method is not in sample config file
- postgresql.auto.conf file with duplicate entries not handled properly
- Code fix committed at f1bf619acdff15b88b5729f8de6df4eed609b3a0
- Doc fix committed at 45aaaa42fefad6e2f164647e373346a5a4123dad
resolved before 12beta3
- Hash join explain can fail with "bogus varno: 65000"
- Commit 5f32b29c18195299e90c1fb6c8945e9a46d772d2 Author: Andres Freund
- Fixed in: 2abd7ae9b20bcd810d4f19d28aefb97048813825
- Wrong-slot-type assertion failure
- Commit 15d8f83128e15de97de61430d0b9569f5ebecc26 Author: Andres Freund
- Fixed in: af3deff3f2ac79585481181cb198b04c67486c09
- virtual tuple table slot does not have system attributes
- Commit 277cb789836b5ddf81aabb80c2058268c70e2f36 Author: Andres Freund
- Fixed in: ecbdd009344d3a00733e4382f50137b5e0248ce8
- Adversarial/pathological case for new nbtree split point choice logic
- Commit fab2502433870d98271ba8751f3794e2ed44140a Author: Peter Geoghegan
- Fixed in: e3899ffd8beafdaaa037b503163a9f572e9fc729
- Add vacuum_index_cleanup for toast relations?
- Commit: a96c41feec6b6616eb9d5baee9a9e08c20533c38, Author: Masahiko Sawada, Owner: Robert Haas
- Fixed in: ce59b75d449d9152667ce3e9eab33ef2872bfd98
- check_recovery_target_lsn() does a PG_CATCH without a throw
- Commit 2dedf4d9a899b36d1a8ed29be5efbd1b31a8fe85 Author: Peter Eisentraut
- Fixed in: 21f428ebde39339487c271a830fed135d6032d73
- psql error messages are now too verbose
- Commit cc8d41511721d25d557fc02a46c053c0a602fed0 Author: Peter Eisentraut
- Fixed in: 3f3542621f131379e32e9283d40853cb6d03a97f
- BulkInsertStates and copy.c with partitioned tables
- Fixed in: b8ef33b2d14519da0d038e9d63f51a6714ef764a
- Crash in mcv_get_match_bitmap with sqlsmith
- Fixed in: e8b6ae2130e3a95bb776708a9a7c9cb21fe8ac87
- GiST Page can become unrecyclable due to PageDeleteXid wraparound
- Commit 7df159a620b760e289f1795b13542ed1b3e13b87 Author: Heikki Linnakangas
- Fixed in: 6655a7299d835dea9e8e0ba69cc5284611b96f29
- REINDEX CONCURRENTLY causes ALTER TABLE to fail
- Fixed in: 28bbf7a81b3a30504cc7dfdbd76b410d1f127b8e
- Progress reporting of REINDEX, CREATE TABLE and CLUSTER ignore each other
- A progress may be started while another one is already in progress. Hence, if progress gets stopped the previously-started state is removed, causing all follow-up updates to not happen.
- Progress updates happening in a code path shared between those three commands may clobber a previous state present.
- Fixed in: da47e43dc32e3c5916396f0cbcfa974b371e4875
resolved before 12beta2
- CREATE TABLE .. PARTITION OF doesn't respect default_tablespace
- Fixed in: a36c84c3e4a9bee6baa7
- with oids option not removed in pg_dumpall
- pg_dump throwing "column number -1 is out of range 0..36" on HEAD
- Fixed in: 54487d1560619a0027e0651d1b8d715ca8fc388c
- ddl.sgml still says foreign keys can't point to partitioned tables
- Commit f56f8f8da6afd8523b4d5284e02a20ed2b33ef8d Author: Alvaro Herrera
- Fixed in: f73293aba4d43e48707e361b2b1ef1465fef46e0
- psql completion bugs with access methods
- Fixed in: 0240a00fbd4fd14f577edf8d36a032237fd0b9cb
- COPY into a partitioned table breaks its indexes
- Fixed in: 56b3b3838284f53c83556592e60688522155f57f
- be-gssapi-common.h not in correct location
- Fixed in: 35b2d4bc0eb5d61a2a294ccb6b2e4abdad307604
- pg_basebackup failure after setting default_table_access_method option
- Fixed in: fff2a7d7bd09db38e1bafc1303c29b10a9805dc0
- Inconsistency between table am callback and table function names
- many commits, Author: Andres Freund
- Fixed in: 73b8c3bd2889fed986044e15aefd0911f96ccdd3
- Documents don't warn about using too many partitions
- Fixed in: e788e849addd56007a0e75f3b5514f294a0f3bca
- Multivariate MCV stats can leak data to unprivileged users
- Fixed by 6cbfb784c3c91146148a76d50cda6f69ae6a79fb et seq
- \connect uses the same IP as the existing connection, docs no longer match behavior, etc.
- Commit 6e5f8d4 Author: Fabien Coelho
- Fixed by 313f56ce2d1b9dfd3483e4f39611baa27852835a
resolved before 12beta1
- Segfault when restoring -Fd dump on current HEAD
- Fixed in 7fcdb5e0021
- pg_dump is broken for partition tablespaces
- Fixed in commits: 87259588d0ab0b8e742e30596afa7ae25caadb18
- and 3b23552ad8bb
- PANIC: cannot abort transaction XXX, it was already committed
- One issue fixed in: 41f5e04aec6cf63ba8392adf70e9289e9c3706d6
- Another issue fixed in: f7feb020c3d8d5aff24204af28359b99ee65bf8f
- remove \cset from pgbench
- Fixed in: 25ee70511ec2ccbef0ad3fe64875a4d552cdcd50
- pg_partition_root crashes when using top-most parent in input
- Fixed in: 2ab6d28d233af17987ea323e3235b2bda89b4f2e
- Server crash in transformPartitionRangeBounds
- Fixed in: cdde886d36b5a4d7ad9e1d02596f7fa1c8c129e3
- Misleading errors with column references in default expressions and partition bounds
- Fixed in: ecfed4a12247cf4659eee6b6ea27405e35fe57f8
- Planner's partitionwise-join code crashes under GEQO
- Fixed in: 7ad6498fd5a654de6e743814c36cf619a3b5ddb6
- Inadequate index locking causes Assert failure
- Fixed in: 9c703c169a872d144f2f79d2fb211c82587adfa7
- Inlining of nested CTEs with recursive terms
- Fixed in: 9476131278c7bfc435ad9a21fc8e981272ac0dd2
- Indexes part of a partition tree cannot be run with REINDEX CONCURRENTLY
- Fixed in: ef6f30fe77af69a8c775cca82bf993b10c9889ee
- Keep track of checksum failures in shared object, last failure time and pg_stat_checkums view
- Commit: 6b9e875f7286d8535bff7955e5aa3602e188e436, Author: Magnus Hagander, Owner: Magnus Hagander
- Fixed in: 77bd49adba4711b4497e7e39a5ec3a9812cbd52a
- Partition tuple routing code does not always call EndForeignInsert
- Commit: 3f2393edefa5ef2b6970a5a2fa2c7e9c55cc10cf, Author: David Rowley, Amit Langote and Álvaro Herrera: Owner: Álvaro Herrera
- Fixed in: 3a45321a491711b556d2cf8f6904ab989b9d0b08
- COLLATE: Hash partition vs UPDATE
- Fixed in: 4b40e44f07c727c7a82b291d3b60098dd99f3f64
- Consider invalid indexes for REINDEX INDEX CONCURRENTLY?
- Fixed in: a6dcf9df4d91ff0db23579f9114079abe6f3e2bf
- ExecForceStoreMinimalTuple leaks memory like there's no tomorrow
- Commit: 4da597edf1bae0cf0453b5ed6fc4347b6334dfe1, Author: Andres Freund, Ashutosh Bapat, Owner: Andres Freund
- Fixed in 88e6ad3054ddd5aa0dee12e5def2c335fe92a414
- GSSAPI encryption missing protocol documentation
- Commit: b0b39f72b9904bcb80f97b35837ccff1578aa4b8, Author: Robbie Harwood, Owner: Stephen Frost
- Fixed in eb882a1b717589777e068dc6671830226f3aa7f0
- New regression tests for GSSAPI encryption are unstable
- Commit: b0b39f72b9904bcb80f97b35837ccff1578aa4b8, Author: Robbie Harwood, Owner: Stephen Frost
- Fixed in eb882a1b717589777e068dc6671830226f3aa7f0
- clean up docs for v12
- Pathological performance when inserting NULL values into unique index
- Commit: dd299df8189bd00fbe54b72c64f43b6af2ffeccd, Author: Peter Geoghegan, Owner: Peter Geoghegan
- Fixed in 9b10926263d831fac5758f1493c929a49b55669b
- Cleanup/remove/update references to OID column
- Commit: 578b229718e8f15fa779e20f086c4b6bb3776106, Author: Andres Freund, Owner: Andres Freund
- Fixed in f6b39171f3d65155b9390c2c69bc5b3469f923a8
- topminnow triggered assertion failure with vacuum_index_cleanup
- compiler warning in pgcrypto imath.c
- Commit: 48e24ba6b7fd3bfd156b51e8d768fd48df0d288b, Author: Noah Misch, Owner: Noah Misch
- Fixed in 726cc4242a2f766c8280a72ef7c8418965d139c8
- Add TRUNCATE option to vacuum command as well as reloption
- Commit: 119dcfad988d5b5d9f52b256087869997670aa36, Author: Tsunakawa Takayuki, Owner: Fujii Masao
- Fixed in b84dbc8eb80b43e554891c459a19969d9fbdefe5
- Improvements in no-fsm-for-small-rels patch suggested by Andres Freund
- Commit: b0eaa4c51bbff3e3c600b11e5d104d6feb9ca77f, Author: John Naylor, Amit Kapila, Owner: Amit Kapila
- Fixed in 7db0cde6b58eef2ba0c70437324cbc7622230320
- Leakage of predicate locks
- Heap lock levels for REINDEX INDEX CONCURRENTLY not quite right?
- Commit: 5dc92b844e680c54a7ecd68de0ba53c949c3d605, Author: Michael Paquier, Owner: Peter Eisentraut
- Fixed in add85ead4ab969c1e31d64f4476c2335856f4aa9
- Inconsistent error message wording for REINDEX CONCURRENTLY
- Commit: 5dc92b844e680c54a7ecd68de0ba53c949c3d605, Author: Michael Paquier, Owner: Peter Eisentraut
- Fixed in 508300e2e141a9fd87758ce01374c5b0597986fd
- Wrong answers from queries using a GIST index
- Commit: 9155580fd5fc2a0cbb23376dfca7cd21f59c2c7b, Author: Anastasia Lubennikova, Andrey V. Lepikhov, Owner: Heikki Linnakangas
- Fixed in 22251686f07f70527aecda22ab5402986884f6f5
- Circular dependency between libpgcommon and libpgfeutils
- Should effective_io_concurrency + 10 be used for an index's page deletion table scans, or a new GUC
- Decision: leave the "+ 10" in for v12, and revisit in a later release when we have a better idea what kind of GUCs we want to control prefetching here and elsewhere
- Passing CopyMultiInsertInfo structure to CopyMultiInsertInfoNextFreeSlot()
- Commit 86b85044e823a304d2a265abc030254d39efe7df Author: David Rowley, Andres Freund
- Decision: it's ok to have the unused parameter
- Assertion failure when using logging_collector on Windows
- Commit 57431a911d3a650451d198846ad3194900666152 Author: Peter Eisentraut
- Fixed for 12beta1 by reverting, in 833451552925d0175e1e15128e411ddef9a36996, the necessary changes are too big for v12
Won't Fix
- Memory leak associated with dubious RelationData.rd_pdcxt handling
- Commit: 898e5e3290a72d288923260143930fb32036c00c, Author: Robert Haas, Owner: Robert Haas
- A proposed patch exists, but some feel that cure is worse than the disease, so we may end up doing nothing for v12
Important Dates
Current schedule:
- feature freeze: April 7, 2019
- beta1: May 23, 2019
- beta2: June 20, 2019
- beta3: Aug. 8, 2019
- beta4: Sep. 12, 2019
- rc1: Sep. 26, 2019
- GA: Oct. 3, 2019