chore(clickhouse): bump hclexp pin to sha-5756e98

[orian]

Problem

The declarative ClickHouse schema tooling (hclexp) is pinned to sha-2f80e66 in posthog/clickhouse/hcl/bin/image.txt, the single place both the offline check.sh gate and the multinode convergence gate resolve. That build predates several chschema releases the HCL recreate work depends on:

• Update README.md #139 — diff -format json per-object objects/summary with attribute-level field changes, drift -format json, per-role roles in plan, and native -exclude (incl. object_types) on all three.
• Fix funnel not saving #142 / Added gif back to readme #143 — dictionary CREATE OR REPLACE reconcile and precise named-collection redaction.
• 'Start sending events to PostHog' notice won't disappear #145 — hclexp locate with -duplicates, the once-only-declaration audit that becomes the new composition's CI guard.

This is the first PR of the HCL recreate plan and unblocks the later phases (per-object comparison for decomposition, locate -duplicates as a check). It should land before anything else.

Changes

Bump the pin from sha-2f80e66 to sha-5756e98 (hclexp v0.1.1-11 to v0.1.1-16). One line, no other files.

The goldens and generated SQL are unchanged: the new build regenerates golden/ and sql/ byte-for-byte identical to what is committed. #139 is documented as a no-op for introspection and DDL generation, and #143's named-collection redaction touches no object in this composition, so nothing shifts.

How did you test this code?

All runs use the pinned Docker image (I forced the container path so a stray local hclexp on PATH could not shadow it):

1. Baseline at the old pin. Regenerated golden/ + sql/ with sha-2f80e66 first. git diff was empty, confirming the harness reproduces the committed artifacts exactly.
2. After the bump. Regenerated with sha-5756e98. golden/ and sql/ diff is empty, so the pin change alters no output.
3. check.sh green under the new pin: every role validates, every (env, role) stack diffs clean against its golden, and sql/ freshness passes.
4. hogli ci:preflight on the branch diff: 0 failures, nothing maps to a known CI failure class.

Automatic notifications

• Publish to changelog?
• Alert Sales and Marketing teams?

Docs update

No user-facing behavior changes.

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

I (Claude) executed Task 0.1 Step 2 of docs/plans/2026-07-14-hcl-recreate.md under Pawel's direction. Followed superpowers:subagent-driven-development for the session, though for a one-line pin bump the substantive work was verification, not delegation, so I did it directly.

Key checks along the way: established a clean old-pin baseline before touching anything so any golden shift would be attributable purely to the bump; forced the Docker image path because a dev-build hclexp was on PATH and would otherwise have generated the goldens instead of the pinned build. The plan flagged #143 as a possible source of golden drift that "must be reviewed, not rubber-stamped" — verified it produces no shift here rather than assuming so.

[greptile-apps]

_{Reviews (1): Last reviewed commit: "Merge remote-tracking branch 'origin/mas..." | Re-trigger Greptile}

[trunk-io]

     

Failed Test	Failure Summary	Logs
test_insert_into_stage_activity_reports_records_total_for_events_model[data_interval_end0-model0-day]	The number of records counted does not match the expected total, indicating a discrepancy in the data.	Logs ↗︎

View Full Report ↗︎ ⋅ Docs

[github-actions]

🤖 CI report

✅ Playwright — all passed

All tests passed.

View test results →

[deployment-status-posthog]

Deploy status

Environment	Status	Deployed At	Workflow
dev	✅ Deployed	2026-07-15 14:33 UTC	Run
prod-us	✅ Deployed	2026-07-15 15:01 UTC	Run
prod-eu	✅ Deployed	2026-07-15 15:01 UTC	Run