Antalya 26.6 Backport of #109051 - Respect use_client_time_zone for string datetime literals parsed on the server - #2208

Workflow [PR], commit [c0da01a]

@blau-ai

CI triage for #2208

Verdict: 7 red checks — 1 is a PR-related process gate (DCO, needs a sign-off), the other 6 are all pre-existing / flaky / infra and unrelated to this change. The PR itself looks healthy: the new test it adds, 04401_async_insert_use_client_time_zone, passes in both stateless jobs that ran it. The diff only touches client-side code (Client.cpp, ClientBase.cpp/.h) that seeds session_timezone from the client TZ — none of the failing suites exercise that path.

1. DCO — ❌ PR-related (action required, not a code bug)

"There is one commit incorrectly signed off… c3468da — Merge pull request ClickHouse#109051 … Sign-off not found"

Commit c3468da has no Signed-off-by line (the merge commit c0da01a is correctly skipped). This is the only failure caused by this PR, and it's purely a sign-off requirement — nothing wrong with the code.

Fix (needs a human — I'm not permitted to force-push / rewrite this branch):

• Rebase + sign-off: git rebase HEAD~2 --signoff then git push --force-with-lease origin backports/antalya-26.6/109051, or
• Remediation commit: add a commit whose message signs off on the prior commit(s), per the cncf/dco2 instructions.

2. PR (aggregate gate) — ⚪ not independent

This is the umbrella status; it's red only because it aggregates the checks below. It clears once they do.

3. Regression release iceberg_2 — ✅ pre-existing on base

89/207 features failed. The same suite fails identically on the base branch antalya-26.6 (latest MasterCI run 31866656987: RegressionTestsRelease / Iceberg (2) / iceberg_2 = failure). Iceberg table reads have nothing to do with client-side DateTime literal parsing. Not caused by this PR.

4. Regression release s3_export_part — ✅ pre-existing on base

9/17 features failed; likewise failing on base branch MasterCI (31866656987: RegressionTestsRelease / S3Export (part) / s3_export_part = failure). Unrelated to the diff. Not caused by this PR.

5. Regression release swarms — 🟡 flaky

1 of 1521 scenarios failed (1520 ok). A single-scenario blip in an otherwise-green suite, on functionality unrelated to client timezone. Flaky — safe to re-run.

6. Stateless tests (amd_debug, distributed plan, s3 storage, parallel) — 🟡 flaky/infra

03279_join_choose_build_table_auto_statistics: [ FAIL ] 181.05 sec
03217_filtering_in_system_tables:               [ FAIL ] 180.30 sec
02377_majority_insert_quorum_zookeeper_long:    [ FAIL ]  60.32 sec
04070_url_base_setting:                         [ BROKEN ] 600 sec

The ~180 s / 600 s durations are runner timeouts under a slow debug build; 04070_url_base_setting (external URL) and the quorum/statistics tests are unrelated to this change. Our new test 04401 = [ OK ] here.

7. Stateless tests (amd_debug, parallel) — 🟡 flaky/infra

00178_quantile_ddsketch:                    [ FAIL ] 600 sec (timeout)
03634_autopr_output_bytes_estimation:       [ FAIL ] 600 sec (timeout)
03441_deltalake_clickhouse_public_datasets: [ BROKEN ] (external datasets)
03441_deltalake_clickhouse_virtual_columns: [ BROKEN ] (external datasets)
04070_url_base_setting:                      [ BROKEN ] 600 sec

All are 600 s hard-timeouts or tests depending on external data (deltalake…public_datasets, url_base_setting) — classic debug-build flakiness/infra, none touching the client TZ path. 04401 = [ OK ] here too.

Recommendation

Only DCO needs an action from a maintainer: add a sign-off to c3468da (rebase --signoff + --force-with-lease, or a remediation commit). Once that's green, re-run the two amd_debug stateless jobs and the swarms regression to clear the flakes. The iceberg_2 and s3_export_part regression failures are pre-existing on antalya-26.6 and shouldn't block this backport — they need a separate fix on the base branch.

_{Automated triage — I can't build/run ClickHouse in this container, so classifications are from CI logs, base-branch comparison, and static reading of the diff.}