orioledb / orioledb

90%

LAST BUILD ON BRANCH main

branch: main

CHANGE BRANCH

x

Reset

Sync Branches

• main
• ImreSamu-makefile_skip_install
• add-constraint-using-index
• add-pg-tests
• add_orioledb_zstd_version_v2
• alter-column-set-storage
• alter-table-subcommands-support
• alter_storage_missing_notice
• array_scan_step
• beta12
• beta13
• beta14
• blacksmith
• btree_seq_scan_fix
• build_toast_table_during_bridging
• bump-patches17
• bump-versions-beta14
• caching
• check_on_arm64
• checkworld_fixes
• checkworld_fixes_with_bridging
• ci_timeout
• cleanup-15
• coveralls
• cppcheck_fixes
• cur_recovery_xid_state_refactor
• debug_undo_failure
• debug_undo_failure_3
• define-in-ambuild
• deserialize_fix
• disable-range-tid-scan
• docker/check_improvements_2024nov
• docker_check_v3
• docker_orioledb_basics_test
• dockerfile_use_valgrind
• dockerhub-fix
• dont-fsync-temp-index-creation
• draft-beta13
• drop_tablespace_dangling_symlink
• enable_brin_regress
• enable_inherit_regress
• enable_insert_regress
• evict-crossdb-descr
• evict-unlogged
• fast_default_fix
• fastpath_attr_index_fix
• fetch_key_fix
• fix-485
• fix-485-2
• fix-662
• fix-bridge-index-replication
• fix-index-row-size-error
• fix-invalid-blkno-use
• fix-logical-ctid-test
• fix-orioledb-tbl-bin-structure
• fix-parallel-rebuild
• fix-plain-insert
• fix-python-compat-namedtempfile
• fix-recovery
• fix-relnode-undo-overflow
• fix-zero-flush-after
• fix578_large_values_v3
• fix_alter_table
• fix_alter_table_and_create_index
• fix_assign_oxid
• fix_evict_pk
• fix_generated_regress
• fix_headers
• fix_int_overflow
• fix_memory_leak
• fix_merge_while_io
• fix_privileges
• fix_rewind_undo_overflow
• fix_sql_header
• fix_sync_commit
• fix_tmp_evict
• fix_wrong_fit_startegy_for_replace_with_vacated
• fix_wrong_root_invalidation
• fixed-stuck-on-eviction
• fixed_truncated_bridge
• fixes_for_pg_17_5
• fk_fix
• foreign_key_test_enable
• get_commithash_from_describe
• get_rid_of_orioledb_synchronize
• github_checks_optimize
• github_ereutt_pageinspect
• github_issue_311
• github_issue_521
• github_issue_527
• github_issue_561
• github_issue_571_logical_xid
• github_issue_580
• github_issue_584
• handle_mark_restore
• improve-docs
• improve_structure_mdx
• indices_get_extended_memory_increase
• invalidate-seqscan-ptr
• isolation_tests_fixes
• issue-547
• issue-587
• issue-587-rebased-3
• issue-587-rebased-4
• issue-587-rebased-norefactor
• issue-643
• issue497_recovery_index
• issue502_stuck_worker
• issue_255_heap_toast_copy
• issue_487_seqscan_backwards
• issue_508_temp_table_cleanup
• issue_515_toast_cleanup
• issue_524_crash
• issue_535_analyze_detoast
• issue_569_crash_on_database_drop
• issue_579_create_database_template
• issue_588_advance_logical_slot
• issue_722_memory_leak
• iterator_toast_comments
• keep_coverage_on_fail
• kiwicopple-patch-1
• lcov_fix
• license-update
• limit-rewind
• local-page-pool
• logical-apply-xid-origin-advance
• logical-fix-ctid
• logical-subxid-guc
• logrep-bridge-fix
• logrep-bridge-fix-squashed
• macos_typedefs
• main_17_7
• main_17_8
• main_17_9
• makefile_skip_install
• memset_undo_stacks
• nickb/adjust_makefile
• nickb/call_check_for_interrupts_in_seqscan
• nickb/fix_worker_crash_pgstat
• nickb/further_logical_replication_tests
• nickb/origin_tracking
• nickb/recovery_stop
• nickb/rewind_docs
• nickb/rewind_optional_restart
• nickb/rewind_restart
• nickb/rewind_restart_cleanup
• nickb/workaround_streaming
• null_ptrs_after_free
• or/eraliest-typo
• origin/bump-patches17-previous
• oriole_systrees_visibility_on_logical_decoding
• oriole_systrees_visibility_on_logical_decoding-3
• oriole_systrees_visibility_on_logical_decoding-4
• oriole_systrees_visibility_on_logical_decoding-5
• orioledb_pronunciation
• page-io-improvements
• page-pool-ops
• page-version-squashed
• page-versions
• pagequeueing_reorg_fix2
• parallel-scan-condition
• parallel-sortmemory-size
• patches17_rebased_PG17_7
• perf-test
• pg-database-size-rebased
• pg_test_fixes_without_exclude
• pgtag_with_hash
• pgtags_check_in_python
• proc_utility_hook_chain
• rebuild-parallel-limit
• recovery_fixes
• recovery_target_time_support
• refactor-wal-decode
• relation-sizes
• relation-sizes-squashed
• remote_apply_accelerate
• remove-generated-sql
• remove-not-needed-assignment
• remove-outdated-ddl-test
• remove-parallel-warning
• rename_data_version
• revert-593-pg_test_fixes_without_exclude
• revert-701-remove-not-needed-assignment
• revert-714-proc_utility_hook_chain
• revise_parallel_seq_scan
• rewind-tests-improve
• rewind_tests
• rewind_tests_improve
• rewind_tests_squashed
• rewind_worker_rebased_4
• rightlink_race
• rll_delete_early
• rr_conflict_fix
• run_pg_tests
• sanitize_fix
• seq_scan_fixes
• seq_scan_on_merge
• set-expression-support
• speedup-valgrind
• split-big-undo-stack-item
• stabilize_rightlink_spec
• stamp-beta14
• subscription-xacts-test
• support-replica-identity-7
• support-replica-identity-8
• temp_patchset_bump
• test_recovery_index
• track_cumulative_stats
• trigger-perf-test
• truncate-compression-bug
• turn-off-writeback-diag
• typcache_recovery_invalidate
• undo-log-size-2
• undo_fix
• undolog_minor_updates
• undometa_docs
• update_copyrights
• update_dockerfiles_2025_debian_etc
• use-strlcat
• use_pg_hash_tmp
• versioning-comments
• wal-version-naming
• wal-versions
• wal_parser_common
• wal_record_refactor
• walk_page_fixes
• xidless-commit-replorigin-advance
• yaroher/perf-test

Build # 23842752611

Build Type

push

github

Committed by akorotkov

Commit Message

Relax undo cleanup assertion to account for recovery retain locations

During recovery, apply_xids_record() replays transaction records from
the previous checkpoint's xid file.  Transactions that were in-progress
during that checkpoint have their retain locations in the
[checkpointRetainStartLocation, checkpointRetainEndLocation] range.
These get loaded into minProcRetainLocation, which can push it below
cleanedLocation (initialized from the control file's lastUndoLocation).

This caused the Assert(oldCleanedNum <= newCleanedNum) in
update_min_undo_locations() to fire during end-of-recovery checkpoint,
because minRetainLocation (used as the new cleaned location) could be
in a lower undo file than the old cleaned location.

The undo files in the checkpoint retain range were persisted during the
previous checkpoint and were never cleaned, so it is safe for the
cleaned location to move back into that range.  Relax the assertion to
allow newCleanedNum < oldCleanedNum as long as newCleanedNum is within
the old checkpoint retain range (>= oldCheckpointStartNum).

Assisted-by: Claude Opus 4.6

Coverage Stats

45705 of 50622 relevant lines covered (90.29%)

77891956.59 hits per line