PostgreSQL 17 Open Items

Open Issues

NOTE: Please place new open items at the end of the list.

NOTE: If known, please list the Owner of the open item.

• 040_pg_createsubscriber.pl test is slow and (still) unstable
  • Commit: d44032d
  • Owner: Peter Eisentraut

• Assertion failure with summarize_wal enabled during pg_createsubscriber
  • Commit: 174c480508ac
  • Owner: Robert Haas

• Incremental backup from a streaming replication standby fails
  • Commit: 174c480508ac
  • Owner: Robert Haas

• pg_upgrade --transaction-size can cause the backend to use an order of magnitude more RAM
  • Commit: 959b38d770b
  • Owner: Tom Lane

Older bugs affecting stable branches

Live issues

• long-standing data loss bug in initial sync of logical replication
  • Affects 11+ (probably)

• relfrozenxid may disagree with row XIDs after 1ccc1e05ae
  • Commit: 1ccc1e05ae
  • We have no test case to confirm the hypothesis that 1ccc1e05ae made things worse.
  • It seems likely that there are or have been bugs in this area for a while, but it is not clear that anything is worse in 17.
  • Owner: Robert Haas

• altering a column's collation leaves an invalid foreign key
  • the pattern exists as earlier as pg 12.

• Inconsistency with EXTRACT(field from INTERVAL)
  • these issues are as old as the hills, so why is this an open item for 17?

• MERGE with UNION ALL bug

• pg_init_privs still has bugs
  • Commit: 534287403
  • Dangling-ACLs issue resolved at: 35dd40d34
  • pg_dump still has issues with replicating the ownership of extension member objects, especially if the ownership was changed after CREATE EXTENSION. That's far from new though, and it's unlikely to get resolved for v17.

• JIT crashes with LLVM 14 on ARM
  • Not entirely clear whether this is our bug or LLVM's

Fixed issues

• Intermittent failures in 040_standby_failover_slots_sync test
  • Possible solution in this thread: Race condition in FetchTableStates
  • Affects 15+ (probably)
  • Fixed at: aa79bde72565

• Underscore in positional parameters?
  • Commit: faff8f8e47f
  • Affects 16+
  • Fixed at: 98b4f53d15

Non-bugs

• lower(s COLLATE pg_c_utf8), though IMMUTABLE, will change behavior in v18
  • Commit: f69319f

Resolved Issues

resolved before 17beta3

• DefineIndex() switches search path, then reverts mid-operation
  • Commit: 2af07e2
  • Owner: Jeff Davis
  • Fixed at: 4b74ebf726, 8e28778ce3, b4da732fd6, a15b0edb5d

• search_path cache: missing invalidation
  • Commit: f26c236
  • Owner: Jeff Davis
  • Fixed at: b3bd18294ee471c5a6f1db69be57f08c1e00f1a6, d3e076549b99d1130053223adb9c1fa909d75dc0

• pg_createsubscriber fails on a dbname containing a space
  • Commit: d44032d
  • Owner: Peter Eisentraut
  • Fixed at: b3f5ccebd79d9c7b73f4e04611cdf31fdf87420b, 917754557cc0002bb042341720a7ce18fe5b0a09

• SQL/JSON: don't use explicit casts for RETURNING types with typmod
  • Commit: 6185c9737c
  • Owner: Amit Langote
  • Fixed at: c2d93c3802b205d135d1ae1d7ac167d74e08a274, 716bd12d22c53d1943d41309f2dd061ec601dd5e

• Apparent bug in WAL summarizer process (hung state)
  • Commit: 174c480508ac
  • Owner: Robert Haas
  • Fixed at: 065583cf460f980a182498941ac52810f709a897

• SQL/JSON: disallow incompatible values in ON ERROR / EMPTY
  • Commit: 6185c9737c
  • Owner: Amit Langote
  • Fixed at: 55e56c84da99fe7becda2194563f48bb3083c2d1

• SQL/JSON: make context_item implicit cast errors respect ON ERROR or have no cast altogether or simply document existing cast behavior?
  • Commit: 6185c9737c
  • Owner: Amit Langote
  • Fixed at: 473a352fb393519f22cd4d31839ad3d74b1aeea1

• pg_combinebackup --clone doesn't work
  • Commit: dc212340058b
  • Owner: Tomas Vondra
  • Fixed at: e99e840b82756bc6858222d97453639cef929b53

• New standby_slot_names GUC in PG 17
  • Commit: bf279ddd1c28
  • Note: discussion is about renaming this GUC.
  • Owner: Amit Kapila
  • Fixed at: 0f934b0739

• pg_createsubscriber: drop pre-existing subscriptions from the converted node
  • Commit: d44032d
  • Owner: Amit Kapila
  • Fixed at: 622cb84d69be91931568bee180cae7c484a7f026

• PQcancelCreate oom_error mutates wrong connection
  • Commit: 61461a3
  • Owner: Alvaro Herrera
  • Fixed at: 3497c87

• comparison to wrong var in GetRunningTransactionData()
  • Commit: e85662d
  • Owner: Alexander Korotkov
  • Fixed at: 6c1af54

• /D on small subset of cmd.exe invocations
  • Commit: 9886744 + b83747a
  • Owner: Michael Paquier
  • Fixed at: 74b8e6a69802

• 64-bit pg_notify page numbers truncated to 32-bit
  • Commit: 4ed8f09
  • Owner: Alexander Korotkov
  • Fixed at: 165ea79a6077 and 3b279d89cb5c

• Subscription and slot information retrieval inefficiency in pg_upgrade
  • Commit: 9a17be1e244a
  • Owner: Nathan Bossart
  • Fixed at: 364509a2e7

• Inadvertent changes in connection negotiation fallback order
  • Commit: 05fd30c0e7
  • Owner: Heikki Linnakangas
  • Fixed at: c95d2159c1

resolved before 17beta2

• SQL/JSON: ON EMPTY default not respected
  • Commit: 6185c9737c
  • Owner: Amit Langote
  • Fixed at: 03ec203164119f11f0eab4c83c97a8527e2b108d

• ResourceOwner bug with pg_logical_slot_get_changes
  • Commit: b8bff07daa
  • Owner: Heikki Linnakangas

• Segmentation fault in tests modules with TIDStore
  • Original Commit: 30e144287a72
  • Fixed at: 18404ea60141a2e2eaf58a5ebbd2b99f7a0cd442

• Failures in constraints regression test, "read only 0 of 8192 bytes"
  • Commit: af0e7deb4a
  • Owner: Heikki Linnakangas

• suspicious valgrind reports about radixtree/tidstore on arm64
  • Commit: 30e144287a72
  • Owner: Masahiko Sawada
  • Fixed at: fd49e8f32

resolved before 17beta1

• "UNION removes almost all rows (not duplicates)"
  • Commit: 66c0185a3d14
  • Fixed at: c37267162

• not null constraints break dump/restore
  • Commit: b0e96f3119
  • Fixed by revert 6f8bb7c1e9

• DROP OWNED BY fails to clean out pg_init_privs grants
  • Fixed at: 534287403914cc9760db98f7320ac4e92f5d416e

• Assert in expandTupleDesc() fails on row mismatch with additional SRF
  • Fixed at: e0df80828abc0c01fcde123389032370b98f2015

• pg_upgrade with --link mode failing upgrade of publishers
  • Original Commit: 29d0a77fa6606f9c01ba17311fc452dabd3f793d
  • Fixed at: f17529b710977c6ae89a1bc141cd1fb587953cc3

• Rethink automated wait events when it comes to backpatching
  • Original Commit: fa88928470b5
  • Fixed at: f98dbdeb51d1

• Make injection points more easily concurrent-safe
  • Original Commit: d86d20f0ba79
  • Fixed at: f587338dec87

• Revisit implementation choices of ReorderBuffer eviction algorithm
  • Original Commit: 5bec1d6bc5e3, b840508644, bcb14f4abc
  • Fixed at: efb8acc0d0, 810f64a015

• Incorrect handling of IS [NOT] NULL quals on inheritance parents
  • Original Commit: b262ad440edecda0b1aba81d967ab560a83acb8a
  • Fixed at: 3af7040985b6df504a72cd307aad5d69ac5f5384

• Synced slot:change could be missing after promotion
  • Commit: ddd5f4f54a
  • Fixed at: 2ec005b4e2

• unnecessary differences between btree/brin in parallel CREATE INDEX
  • Commit: b43757171470
  • Owner: Tomas Vondra

• Problems with deparsed SQL/JSON query function
  • Commit: 6185c9737cf
  • Fixed at: c0fc0751862

• Improve some error messages for JSON_TABLE() by mentioning column name
  • Commit: de3600452b61d
  • Fixed at: b4fad46b6bc8

• pg_combinebackup does not detect missing files
  • Commit: dc212340058b
  • The behavior is intentional but we have improved the documentation in 2e2d4604d94318a304d350aee478bda91a521339
  • Owner: Robert Haas

• aminsertcleanup fails to release buffers during reindex
  • Commit: c1ec02be1d
  • Owner: Tomas Vondra

• fix tablespace handling in pg_combinebackup
  • Commit: dc212340058b
  • Fixed at: cd64dc42d1e1b03e57e6ba3d316e4f9dec52a78d

• dump/restores not able to handle AM for partitioned tables
  • Commit: 374c7a229042
  • Fixed at: f46bee346c3b

• Rework handling of failover option in transaction block for ALTER SUBSCRIPTION
  • Commit: 776621a5e479
  • Fixed at: b29cbd3da4e3

• Rework tests of test_json_parser
  • Commit: 3311ea86edc7
  • Fixed at: daf554dbea b8a7bfa333 42fa4b6601 ba3e6e2bca cc893b8237

• backtrace_on_internal_error lacks extensibility
  • Commit: a740b213d4b
  • Owner: Peter Eisentraut
  • Reverted: 592a2283721f7143999364ef487f2b4993f5161d

• checksums vs. incremental backup
  • Commit: dc212340058b
  • Owner: Robert Haas
  • For now, we'll do this: 205db0114e03496f1d6febf276374900b6314e67

• Check that ALPN was used with direct SSL in libpq
  • Commit: 91044ae4ba
  • Owner: Heikki Linnakangas

• Crash on merging inherited compression attributes
  • Commit: 4d969b2f8
  • Owner: Peter Eisentraut, Ashutosh Bapat
  • Fixed at: 8f0a97dfffc5b5f48f609fb894befcdf761dd576

• Issues with identity columns and partitioning
  • Commit: 699586315704
  • Owner: Peter Eisentraut
  • Fixed at: 509199587df73f06eda898ae13284292f4ae573a

• Register ALPN protocol id with IANA
  • Thread on IETF mailing list
  • apparently, request has been approved
  • Commit: 91044ae4ba
  • Owner: Heikki Linnakangas

• Race condition with local injection point detach
  • Commit: f587338dec87
  • Owner: Michael Paquier
  • Fixed at: 33181b48fd0e and 267d41dc4f41

• Temporal PKs allow duplicates with empty ranges
  • Commit: 46a0cd4ce
  • Owner: Peter Eisentraut
  • Fixed by revert 8aee330af55d8a759b2b73f5a771d9d34a7b887f

• explain format json, unit for serialize and memory are different
  • Commit: 5de890e3610d, 06286709ee06
  • Owner: Alvaro Herrera, Tom Lane
  • Fixed at: 0de37b51065bc5b5848d65a9bb6f932ef392374f

• Rename sslnegotiation "requiredirect" option to "directonly"
  • Commit: d39a49c1e4
  • Owner: Heikki Linnakangas
  • Latest patch

• Incorrect Assert in heap_end/rescan for BHS
  • Commit: 04e72ed617be
  • Owners: Melanie Plageman, Tomas Vondra
  • Fixed at: a3e6c6f929912f928fa405909d17bcbf0c1b03ee

• State of pg_createsubscriber
  • Commit: d44032d01463
  • This item is too general. Add specific items for specific issues.
  • Some things have already been fixed (e.g. test instability).

Important Dates

Current schedule:

• GA: XX
• RC 1: XX
• Beta 3: XX
• Beta 2: June 27, 2024
• Beta 1: May 23, 2024
• Feature Freeze: April 8, 2024 0:00 AoE

See also

• Release Management Team
• PostgreSQL 16 Open Items