[Before forward-ports] Antalya 26.8: patch for SettingsChangesHistory

[zvonand]

Antalya patch for SettingsChangesHistory, landed before forward-ports.

When porting to future Antalya versions:

• generate the .tsv file containing settings values in the latest Antalya release (more details in the comments in tests/queries/0_stateless/03999_stateless_settings_history.sh) and rename it to that version, keeping the 03999 prefix — ci/jobs/scripts/find_tests.py maps a data file to its owning test by five-digit prefix;
• add a new version section for the next release in src/Core/SettingsChangesHistory.cpp;
• keep the flavour list in the test in sync with the allowlist in ClickHouseVersion.cpp.

When cherry-picking, do cherry-pick -n to avoid picking:

• the ...antalya.tsv file: it needs to be re-generated;
• SettingsChangesHistory.cpp: instead, take the Antalya-related sections from the previous Antalya release's SettingsChangesHistory.cpp and manually add them to the current file, commenting all lines with Antalya-specific settings.

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
• Unit 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)
• CAS (content-addressed storage; Antalya only)
• 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 [43d6925]

[zvonand]

@blau-ai

[blau-ai]

CI triage — 3 failing checks, 0 caused by this PR

Check	Verdict	Root cause
PromQL Compliance	❌ infra	GH auth failed - required by job
Build profile diff	❌ infra	GH auth failed - required by job
Stateless tests (amd_debug, parallel)	❌ infra/flaky	setup timeout connecting to remote object-storage/distributed-cache endpoint; no test ever ran

None of these touch the code this PR changes (ClickHouseVersion parsing, SettingsChangesHistory.cpp, and the new 03999_* stateless test). Safe to re-run.

---

1 & 2 — PromQL Compliance / Build profile diff → infra (not PR-related)

Both jobs errored out before doing any work, at the praktika runner's credential check:

File ".../ci/praktika/runner.py", line 427, in _run
    Utils.raise_with_error("GH auth failed - required by job")
RuntimeError: GH auth failed - required by job

This is a GitHub-token/credential problem on the runner, not anything in the diff. A code change cannot produce this. Next step: re-run these two jobs.

3 — Stateless tests (amd_debug, parallel) → infra setup failure (not PR-related)

The suite has exactly one sub-result, and it is the setup step, not a test:

FAIL  Start ClickHouse Server

The server itself started cleanly:

ClickHouse server replica 0 started successfully, pid [553]
Server replica 0 ready

— which, notably, is positive evidence that this PR's changes to ClickHouseVersion parsing did not break startup. The suite then aborted during stateful data preparation, on a network timeout to a remote endpoint while loading the hits_s3 table:

Code: 1000. DB::Exception: Received from localhost:9440.
  DB::Exception: Timeout: connect timed out: 65.108.242.32:6000. (POCO_EXCEPTION)
(query: INSERT INTO test.hits_s3 SELECT * FROM test.hits ...)
...
SETUP FAILURE: stateful data prep failed (exit 232)

65.108.242.32:6000 is a remote object-storage/distributed-cache endpoint used by the s3_cache storage policy during fixture load — it was unreachable/timed out. No 0_stateless test (including this PR's new 03999_stateless_settings_history.sh) ever executed. Next step: re-run the job.

Positive signals for this PR

• Server starts with the new ClickHouseVersion suffix-parsing logic (see startup log above).
• Stateless tests (amd_binary, flaky check) = OK — the flaky-check job, which specifically stress-runs new/changed tests, passed, so the PR's new 03999 test is healthy.
• All builds, unit tests (incl. the new gtest_ch_version.cpp), integration tests, AST fuzzers, and Grype scans are green.

Recommendation: re-run the three failing jobs; no code change is needed. Once the runner auth/network hiccups clear, this should go green.

_{— @blau-ai · evidence from praktika result_pr.json for 43d6925}