Structural comparison in pg-upgrade-test (both twin databases)

[jnasbyupgrade]

Supersedes #35, which is now stale: it was branched before pg-upgrade-test's twin-database redesign (the schema matrix axis was removed and replaced with a single job that prepares two databases - count_nulls_upgrade_none (schema none) and count_nulls_upgrade_quoted (schema Quoted) - ahead of one binary pg_upgrade call, then loops bin/test_existing's update/run-suite over both). #35 referenced matrix.schema and a single count_nulls_upgrade database name that no longer exist in that form.

This PR adapts #35's actual intent to the new shape:

• Adds an EXISTING_DB argument back to bin/compare_fresh_vs_update (it doesn't exist on phase6-compare-and-report's current version - both branches independently added this script, and only Run bin/compare_fresh_vs_update against the real pg_upgraded database too #35's version had this parameter). When given, the script diffs that already-populated database against a fresh install instead of creating+updating its own scratch database, and leaves it untouched on exit (only scratch databases it created itself are dropped).
• Adds a step after the existing "Run the suite against the pg_upgraded database" step in pg-upgrade-test that loops the comparison over both twin databases, each against a fresh install in its own schema:
  • bin/compare_fresh_vs_update "" 0.9.6 count_nulls_upgrade_none
  • bin/compare_fresh_vs_update Quoted 0.9.6 count_nulls_upgrade_quoted

This catches a divergence class the fixed pgTAP suite doesn't: an object left subtly different (body, comment, ACL) by surviving a real binary catalog migration, as opposed to only an in-place ALTER EXTENSION UPDATE (which extension-update-test already covers with the same tool).

Verification

This container's PG12/PG17 clusters are shared persistent dev infra, not disposable, so a real pg_upgrade between them wasn't run. Instead, verified the mechanics against plain (non-upgraded) databases with the same names/schemas the CI job produces:

• bin/test_existing prepare-old for both count_nulls_upgrade_none (schema "") and count_nulls_upgrade_quoted (schema Quoted)
• bin/test_existing update + run-suite for both - both pass
• bin/compare_fresh_vs_update "" 0.9.6 count_nulls_upgrade_none and the Quoted equivalent - both report IDENTICAL and exit 0
• Negative check: COMMENT ON FUNCTION ... on one of the existing databases, re-ran the comparison - it correctly printed a real diff and exited nonzero, then reverted and confirmed it passes clean again
• Confirmed the EXISTING_DB databases survive the comparison script (only its own scratch compare_fresh_vs_update_* databases are dropped)

The real binary pg_upgrade leg itself is left for CI's ephemeral containers.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 18e75e19-7b5d-4b28-b1bd-deb5b262f3e3

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.