軟體下載 Select Version

• 1

  版本確認

• 2

  檔案準備中

• 3

  檔案下載

What's new in this version:

- Fix for handling of NULL values in BRIN indexes. This fix does not apply to existing BRIN indexes -- you will need to run REINDEX to fix any BRIN indexes used to search for NULL values
- Avoid leaving a corrupted database behind when DROP DATABASE is interrupted
- Several fixes for partitioned indexes
- Fix ALTER EXTENSION ... SET SCHEMA to error if the extension contains any objects outside the extension's schema
- Fix dependency tracking of access methods for tables
- Don't use partial unique indexes for uniqueness proofs in the planner
- Correctly handle sub-SELECTs in RLS policy expressions and security-barrier views when expanding rule actions
- Fix race conditions in conflict detection for SERIALIZABLE transaction isolation mode
- Fix intermittent failures when trying to update a field of a composite column that requires out-of-line TOASTing
- Fix several memory leaks that occurred during the lifespan of a query
- Accept fractional seconds in the input to the jsonpath datetime() method
- Increase token limit in pg_hba.conf and pg_ident.conf to 10,240 bytes
- An out-of-memory error from JIT will now cause a PostgreSQL FATAL error instead of a C++ exception
- Allow VACUUM to continue after detecting certain types of B-tree index corruption. While this fix allows VACUUM to continue, you still need to REINDEX to fix the broken index
- Avoid double replay of prepared transactions during crash recovery
- Ensure that checkpoint calls fsync on a newly created but empty table
- Silence "missing contrecord" errors to avoid logging inaccurate messages from pg_waldump and walsender
- Fix the [fuzzystrmatch] Soundex difference() function to correctly handle empty input
- Several fixes for intarray, including disallowing oversize input arrays in a GiST index
- Fix pg_dump to correctly handle SQL-standard function bodies (BEGIN ATOMIC) that require parse-time dependencies on unique indexes