Antalya 26.6: Do not read empty namespaces from catalog

[zvonand]

Changelog category (leave one):

• Bug Fix (user-visible misbehavior in an official stable release)

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Fix infinite recursion with empty namespace in Iceberg catalog (#1762 by @ianton-ru).

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)
• Parquet (1.5h)
• RBAC (1.5h)
• SSL Server (1h)
• S3 (2h)
• S3 Export (2h)
• Swarms (30m)
• Tiered Storage (2h)

Cherry-picked from #1762.

---

Documentation entry for user-facing changes

Solved #1382

[github-actions]

Workflow [PR], commit [a7dbf46]

[ianton-ru]

Failed tests looks unrelated, but I think need to investigate to stabilize tests.
Looks like all failed stateless tests failed by timeout. May be deadlock somewhere in code.

[zvonand]

@blau-ai

[blau-ai]

CI triage for #2124

Verdict: none of the failures are caused by this PR. All red checks are pre-existing baseline failures on antalya-26.6 or flaky/infra timeouts. This PR is a single defensive guard in RestCatalog.cpp and is effectively a no-op for every passing code path — so it can't be responsible for the failures below.

Why the change is safe

The diff adds one branch to getNamespacesRecursive:

if (current_namespace == base_namespace) { LOG_WARNING(...); continue; }

The loop already chasserts current_namespace.starts_with(base_namespace), and in normal operation current_namespace is always a strict child of base_namespace. So this branch only fires on the pathological empty-subnamespace case it's meant to fix; for every well-formed catalog it's never taken → no behavior change.

Per-failure breakdown

Check	Classification	Evidence
Regression aarch64 iceberg_1 / iceberg_2	Pre-existing (branch baseline)	Same suites fail on #2154 and #2156, which don't touch this code. The Iceberg regression baseline is currently red on antalya-26.6.
Regression aarch64 s3_export_partition	Pre-existing / infra	Fails identically on #2154 and #2156. Unrelated to Iceberg catalog code.
Regression aarch64 swarms	Infra/setup	1322/1520 scenarios failed; job died in ~5 min on #2154/#2156. Whole-suite environment failure, not a test regression.
Stateless (amd_debug, parallel) — Failed 5, Broken 3	Flaky / timeout / infra	Failures are 900s timeouts on the slow debug runner: 03634_autopr_output_bytes_estimation, 04033_tpc_ds_q62, 04033_tpc_ds_q69, 00171_grouping_aggregated_transform_bug; 04070_url_base_setting BROKEN (900s). 03441_deltalake_* BROKEN (need external public datasets). None exercise Iceberg REST catalog.
Stateless (arm_binary, parallel) — Failed 1	Flaky / timeout	03634_autopr_input_bytes_estimation FAIL at 600.14s = hard timeout. On #2156 this arm job passed while amd failed, and vice-versa on #2154 → flaky, not PR-caused.
PR	Rollup	Aggregate gate; red only because of the above.

(Note: all Build jobs show skipping = CI cache hit / reused artifacts, not failures.)

Suggested next steps

• Stateless amd_debug / arm_binary: safe to re-run — these are runner-timeout/flaky failures. The autopr *_bytes_estimation and tpc_ds tests are chronically slow on debug builds.
• Regression iceberg/s3_export/swarms: pre-existing red on antalya-26.6; not actionable from this PR. Worth confirming the iceberg_1/iceberg_2 failing-scenario set matches the current branch baseline (open the report.html artifacts and compare against Antalya 26.6: Iceberg: support external paths in tables #2154/Antalya 26.6: Cache vended credentials for REST catalogs #2156) — if it matches, this PR is clean to merge behind those known-flaky suites.

No code fix is needed for this PR. Happy to dig into any specific regression report if you want a scenario-level diff against the baseline.