Skip to content

Antalya 26.3: Bump to v26.3.13.31-lts#1899

Open
zvonand wants to merge 69 commits into
antalya-26.3from
bump/antalya-26.3/v26.3.13.31-lts
Open

Antalya 26.3: Bump to v26.3.13.31-lts#1899
zvonand wants to merge 69 commits into
antalya-26.3from
bump/antalya-26.3/v26.3.13.31-lts

Conversation

@zvonand

@zvonand zvonand commented Jun 10, 2026

Copy link
Copy Markdown
Member

Changelog category (leave one):

  • Not for changelog (changelog entry is not required)

CI/CD Options

Exclude tests:

  • Fast test
  • Integration Tests
  • Stateless tests
  • Stateful tests
  • Performance tests
  • All with ASAN
  • All with TSAN
  • All with MSAN
  • All with UBSAN
  • All with Coverage
  • All with Aarch64
  • All Regression
  • Disable CI Cache

Regression jobs to run:

  • Fast suites (mostly <1h)
  • Aggregate Functions (2h)
  • Alter (1.5h)
  • Benchmark (30m)
  • ClickHouse Keeper (1h)
  • Iceberg (2h)
  • LDAP (1h)
  • OAuth (5m)
  • Parquet (1.5h)
  • RBAC (1.5h)
  • SSL Server (1h)
  • S3 (2h)
  • S3 Export (2h)
  • Swarms (30m)
  • Tiered Storage (2h)

robot-clickhouse and others added 30 commits April 22, 2026 12:39
…ogical error when client side data encoding is incorrect
…uery when Glue catalog reads hit ExpiredToken
… being ignored for CAST to Nullable(DateTime)
The EXPLAIN-step assertion looked for `'%Partition Min-Max%'`. That string
only exists on master/26.4+ where ClickHouse#94140 renamed `IndexType::MinMax` to
`IndexType::PartitionMinMax`. In 26.3 the step is still printed as `MinMax`,
so the test fails on the partition-pruning columns of the reference.

Loosen the pattern to `%MinMax%`, which matches both names. No behavioral
change to the test — the assertion still verifies that the partition-key
min/max index step is present in the EXPLAIN plan.

This is a 26.3-only fix; master and 26.4+ already have a matching step name.
…1932

Backport ClickHouse#101932 to 26.3: Fix materialize_skip_indexes_on_merge not suppressing text indexes
…n-pattern-on-26.3

Fix 04229_defer_partition_pruning_after_final test on 26.3
…5636

Backport ClickHouse#105636 to 26.3: Fix test_mysql_dotnet_client by anchoring to /testapp
`aspell-dict.txt` was deleted in master along with the `aspell` and
`codespell` style checks. Removing the file here so the PR can merge
cleanly without a modify/delete conflict.
Instead of deleting the existing 26.2 and 25.12 block entries and adding a
new 25.10 entry (the master pattern), change the existing entries' values
in place from `{false, true, ...}` and `{false, false, ...}` to
`{true, true, ...}`. Preserves the original PR-history comments
("Enabling apply_row_policy_after_final by default, as if was in 25.8
before ClickHouse#87303" and "New setting to control if row policies and PREWHERE
are applied after FINAL...") rather than replacing them with a single
synthetic comment in a different version block.

Same effect on the compatibility walker: both `{true, true}` entries are
no-ops, so the setting resolves to `true` for every `compatibility`
level on 26.3, matching the DECLARE default.
…g FD_SETSIZE limitation of select() (replace with poll())
…5637

Backport ClickHouse#105637 to 26.3: Make `apply_row_policy_after_final` consistent accross versions
…3035

Backport ClickHouse#103035 to 26.3: Fix cast_string_to_date_time_mode being ignored for CAST to Nullable(DateTime)
…2975

Backport ClickHouse#102975 to 26.3: Use `INCORRECT_DATA` instead of logical error when client side data encoding is incorrect
…5559

Backport ClickHouse#105559 to 26.3: Fix crash in replxx due to hitting FD_SETSIZE limitation of select() (replace with poll())
…1292

Backport ClickHouse#101292 to 26.3: Fix UBSan in MergeTreeDataPartWriterCompact::cancel
…efresh on brief keeper connection loss (ClickHouse#105589)

Co-authored-by: robot-clickhouse <robot-clickhouse@users.noreply.github.com>
…PAUSE VIEWS (ClickHouse#104562)

Co-authored-by: robot-clickhouse <robot-clickhouse@users.noreply.github.com>
…sh buffer in WriteBufferFromHDFS::finalizeImpl
…red marks to PREWHERE in QueryConditionCache
…5686

Backport ClickHouse#105686 to 26.3: Do not attribute skip-index-filtered marks to PREWHERE in QueryConditionCache
…5776

Backport ClickHouse#105776 to 26.3: Reject empty or NUL-containing MongoDB collection names
robot-clickhouse and others added 28 commits May 28, 2026 13:33
The 26.3 backport should only add `show_remote_databases_in_system_tables` from PR ClickHouse#104416. Dropping the unrelated setting keeps `02995_new_settings_history` aligned with the branch.

PR: ClickHouse#106045
The PR makes `system.databases` list remote database names unconditionally because that is local metadata and does not enumerate remote tables. Keep `test_database_glue` aligned with master by expecting the data lake catalog database to remain visible when `show_data_lake_catalogs_in_system_tables` is false.

PR: ClickHouse#106045
The 26.3 stateless test runner does not provide `CLICKHOUSE_DATABASE_2`, while the original PR converted this coverage to SQL using two generated database names. Keep the same assertions in a shell test and derive both remote database names from `CLICKHOUSE_DATABASE`, which is available on 26.3.

PR: ClickHouse#106045
…ng 0 rows with partial `SHOW DICTIONARIES` revoke
…5384

Backport ClickHouse#105384 to 26.3: Fix set skip index throwing "It's a bug!" on non-integer WHERE atoms
…6105

Backport ClickHouse#106105 to 26.3: Fix `system.dictionaries` returning 0 rows with partial `SHOW DICTIONARIES` revoke
`check_cpp.sh` fails on the 26.3 backport branch because this blank line contains trailing spaces. Remove them so the branch passes style before pushing the BackportPR fix.

CI report:
https://s3.amazonaws.com/clickhouse-test-reports/json.html?PR=106045&sha=a7b453b7045232485dd9b3aa7c8d5f942187d5e4&name_0=BackportPR
Test 04299_deltalake_invalid_utf8_credentials exercises the deltaLakeAzure
path, but Delta Lake on Azure with delta-kernel-rs (AzureKernelHelper) does
not exist in 26.3, so getKernelHelper throws NOT_IMPLEMENTED for Azure and the
query never reaches the setBuilderOption UTF-8 check that emits BAD_ARGUMENTS.
The S3 fix and its tests (04300, 04301) are valid and stay.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…4416

Backport ClickHouse#104416 to 26.3: Hide MySQL and PostgreSQL databases from system.tables by default
…6109

Backport ClickHouse#106109 to 26.3: Fix DeltaLake delta-kernel-rs FFI server crash and builder leak
…5584

Backport ClickHouse#105584 to 26.3: Honor cast_string_to_date_time_mode for hive partition values
…5894

Backport ClickHouse#105894 to 26.3: Fix `IN Array` with negative values for distributed queries
…1218

Backport ClickHouse#101218 to 26.3: Fix `NOT_FOUND_COLUMN_IN_BLOCK` error when selecting from a VIEW with a PROJECTION
…5535

Backport ClickHouse#105535 to 26.3: Fix SEGFAULT in `singleValueOrNull` deserialization for JSON type
…3268

Backport ClickHouse#103268 to 26.3: Fix data loss on HDFS writes: flush buffer in WriteBufferFromHDFS::finalizeImpl
…ckport/26.3/105584

Revert "Backport ClickHouse#105584 to 26.3: Honor cast_string_to_date_time_mode for hive partition values"
…1403

Backport ClickHouse#101403 to 26.3: Fix mutation UPDATE SET DateTime literal not rewritten with session_t…
…5996

Backport ClickHouse#105996 to 26.3: Fix possible OOM due to a too-large string literal sent within the query
Release v26.3.13.31-lts

Conflict resolution notes:
- Adopted upstream's generalized `show_remote_databases_in_system_tables`
  API (renamed from Antalya's `show_data_lake_catalogs_in_system_tables`,
  which upstream keeps as a runtime alias). Converted the remaining
  `with_datalake_catalogs` usages in `StorageSystemHybridWatermarks` and
  `StorageSystemReplicatedPartitionExports` to `with_remote_databases`.
- `InterpreterShowTablesQuery`: kept Antalya's single-block refactor, using
  the renamed `isRemoteDatabase` / `show_remote_databases_in_system_tables`.
- `ci/defs/job_configs.py`: kept the Altinity `integration-tests-runner`
  image and `CLICKHOUSE_TEST_STAT_*` env vars, plus upstream's new
  `--ulimit nofile=262144:262144`.
- Kept `aspell-dict.txt` (deleted upstream) because Antalya still runs the
  `aspell` style check, which requires the dictionary.
- Bumped submodules `delta-kernel-rs`, `mongo-c-driver`, and `replxx` to the
  upstream pins from the release tag.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown

Workflow [PR], commit [60be687]

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.