Add CLAUDE.md: document why extension_drop needs schema-independence testing

[jnasbyupgrade]

Summary

• extension_drop's control file has no schema= line, making it genuinely install-schema-flexible (confirmed against actual PostgreSQL semantics: schema=, not relocatable=, is what pins install-time schema choice).
• The current test suite only proves qualification-correctness via one dedicated file (test/sql/schema.sql), not across the whole suite.
• Documents the correct pattern going forward: a TEST_SCHEMA switch mirroring TEST_LOAD_SOURCE's existing GUC-propagation mechanism, per Postgres-Extensions/pg_count_nullsCI: fix claude-code-review by removing the fork-checkout step entirely #28 (a working reference implementation already in the org).

This is documentation only — no test/build changes. Implementing the actual TEST_SCHEMA dimension is tracked as follow-up work.

Test plan

• N/A (doc-only change)

[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: 982b4fb2-08ad-43d4-9908-613e21b55c49

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]

Claude finished @jnasbyupgrade's task in 2m 55s —— View job

---

Claude finished reviewing this PR

Todo list:

• Gather context (PR diff, files changed)
• Read CLAUDE.md content and surrounding repo context (control files, existing tests)
• Verify factual claims in the PR body (schema= semantics, existing TEST_SCHEMA/TEST_LOAD_SOURCE mechanism, referenced issue)
• Provide code review feedback

---

Review

This is a doc-only addition (CLAUDE.md, +43 lines), and I verified its factual claims against the actual repo state rather than taking them at face value:

• schema= vs relocatable=: confirmed. extension_drop.control has relocatable = false and no schema = line — matches the doc's claim exactly.
• Only schema.sql proves the schema-independence property: confirmed. test/sql/schema.sql:52 explicitly removes TEST_SCHEMA_2 from search_path before asserting against it, and uses a mixed-case name (_Test_Ed/_Test_Ed_2) that would break if a reference were left unquoted. test/sql/simple.sql only uses an all-lowercase schema (_test_ed) and never removes it from search_path, so it can't catch either failure mode. The doc's diagnosis of the test suite's current asymmetry is accurate.
• Existing TEST_SCHEMA mechanism: test/deps.sql:19-20 already reads a per-file \set TEST_SCHEMA variable to CREATE SCHEMA/set search_path, matching the doc's description.

One accuracy nit worth a look before/after merge:

This is best implemented as a TEST_SCHEMA switch on the whole test suite (mirroring TEST_LOAD_SOURCE's existing GUC-propagation mechanism through test/install/load.sql / test/deps.sql)...

Neither test/install/ nor a TEST_LOAD_SOURCE-style Makefile/GUC switch actually exists in this repo — only test/deps.sql's much simpler per-file \set TEST_SCHEMA exists here today. The TEST_LOAD_SOURCE pattern this sentence points to is pgxntool's documented convention, whose real example lives in cat_tools (per pgxntool/README.asc), not in extension_tools. As written, a future implementer skimming this repo's CLAUDE.md could think test/install/load.sql already exists locally to mirror, when in fact building the TEST_SCHEMA switch here means introducing that whole pattern from scratch (there's no test/install directory at all yet). Worth a small tweak to make explicit that this is an external pattern to model, not local infrastructure.

I wasn't able to fetch Postgres-Extensions/pg_count_nulls#28 to confirm its title/content matches "TEST_SCHEMA switching in test/install" — external gh/web access wasn't permitted in this run — so that citation is unverified on my end (not necessarily wrong, just unchecked).

Everything else — the reasoning about search_path-absence and mixed-case/quoting being independently necessary, and marking this as documentation-only with implementation as tracked follow-up — is sound and consistent with the current code. No blocking issues; the one nit above is optional polish.