Antalya 25.8: fix some failing tests

[zvonand]

• 04340_datalake_schema_deep_recursion
• 04409_explain_actions_secret_args
• 04410_explain_actions_secret_args_secondary

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
• Aarch64 tests
• All with ASAN
• All with TSAN
• All with MSAN
• All with UBSAN
• All with Coverage
• 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)

[github-actions]

Workflow [PR], commit [1db2126]

[zvonand]

@blau-ai

[blau-ai]

CI triage — 3 failing checks, 0 PR-caused (1 infra, 1 flaky/pre-existing, 1 dependent gate)

This PR touches only 5 stateless test files (04299, 04340, 04409, 04410) and zero C++/build code, so no failure can be attributed to it by construction. Details per check below.

Check	Verdict
Build (amd_release)	❌ Infra — job terminated (SIGTERM), not a compile error
Stateless tests (amd_binary, ParallelReplicas, s3 storage, parallel)	❌ Flaky / pre-existing product bug in an unrelated test
PR	⚠️ Aggregate gate — red only because of the two above

---

1. Build (amd_release) — infra, safe to re-run

The job ran ~2h05m and was killed with exit code [-15] (SIGTERM) during result/artifact packaging, not a compilation failure:

19:50:48 ...KeeperConverter.cpp.o          <- object files being packaged
19:50:51 WARNING: Failed to scan file ./ci/tmp/result_build_amd_release.json.gz, unknown encoding
19:50:52 Current job failed ...
          ... description='Job got terminated with an error, exit code [-15]'

No error:/FAILED: lines anywhere in the compile log. Exit -15 is the runner terminating the process (timeout/resource kill). A test-only PR cannot break the build.
Next step: re-run the job — no change needed.

2. Stateless tests (amd_binary, ParallelReplicas, s3 storage, parallel) — flaky, unrelated

Failed: 1, Passed: 7124, Skipped: 732, Broken: 2. The single failure is 01710_projection_additional_filters, which is not part of this PR:

01710_projection_additional_filters:   [ FAIL ] 1.04 sec.
Reason: result differs with reference:
 0	0	0
-3
+23

The failing statement is select count() from atf_p settings additional_table_filters = {'atf_p': 'x <= 2'} on a 10-row table — expected 3, got 23 (a count larger than the table itself). The test explicitly enables parallel_replicas_support_projection=1, and the failing run randomized --parallel_replicas_local_plan 0. This is a latent parallel-replicas + additional_table_filters correctness bug surfaced by a specific random-settings combination.

Evidence it is not PR-caused:

• Not in the PR diff (PR only edits 04299/04340/04409/04410).
• The same job passed on the base branch antalya-25.8 in its latest run (run 29721926034, both ParallelReplicas … parallel and … sequential = success), so it is not a hard regression on the branch — it's intermittent under randomized settings.
• The (broken) 03441_deltalake_* entries are pre-marked BROKEN, not counted as failures.

Next step: re-run the job (should pass on a different settings roll). Independently, the additional_table_filters + parallel-replicas wrong-result looks like a real product bug worth its own issue — but it is out of scope for this test-fix PR.

3. PR — dependent gate

The umbrella PR status is red purely because the two checks above are red. It should clear once they pass.

---

Health of the actual change

The four tests this PR fixes all pass wherever they run:

04299_constraint_subst_correlated_subquery_root   6× [ OK ]
04340_datalake_schema_deep_recursion              5× [ OK ], 1× [ SKIPPED ]  (no-msan/no-fasttest as intended)
04409_explain_actions_secret_args                 5× [ OK ], 1× [ SKIPPED ]
04410_explain_actions_secret_args_secondary       5× [ OK ], 1× [ SKIPPED ]

So the PR's own changes are working. No fix is needed on this PR — just re-run Build (amd_release) and the ParallelReplicas … parallel stateless job.

_{🤖 Generated with Claude Code}