test: make inheritance search test config-independent

[tjgreen42]

What

Make the inheritance regression test independent of the server's default_text_search_config.

Why

Test 1 checks "searching on child table works" with core text search:

WHERE content @@ to_tsquery('english', 'postgresql & inheritance');

Here @@ resolves to ts_match_tq(text, tsquery) — a sequential scan that stems the text side using the database default_text_search_config, while the tsquery side is explicitly english (stems inheritance → inherit). On any server whose default config is not english — notably PostgreSQL's compiled-in boot default pg_catalog.simple — inheritance is not stemmed to inherit, the match returns 0 rows instead of 1, and the test fails. The line also never exercised the bm25 index, so it wasn't really testing pg_textsearch.

What changed

Replace it with a bm25 search using the <@> operator against the child's bm25 index (child1_bm25_idx). bm25 parses the query with the index's text_config (english), so the result no longer depends on default_text_search_config — and it actually exercises pg_textsearch.

Testing

make installcheck — verified the inheritance test passes under both default_text_search_config = 'pg_catalog.simple' and 'pg_catalog.english', and the full suite (71 tests) passes.

[claude]

LGTM — straightforward test-only change that makes the inheritance test config-independent and actually exercises the bm25 index.

Extended reasoning...

Overview

This PR modifies a single regression test (test/sql/inheritance.sql and its expected output) to replace a core text-search @@ predicate (which depends on the server's default_text_search_config) with a bm25 <@> ordering query against the child's bm25 index. Both files change in a consistent way: the SQL query and its expected output.

Security risks

None. This is purely a test file change with no impact on production code, auth, data handling, or external interfaces.

Level of scrutiny

Low. The change touches no extension C/SQL implementation code, only the regression test suite. The motivation in the PR description is technically accurate: to_tsquery('english', ...) only stems the query side under the english config, while the @@ operator stems the document side under default_text_search_config, so the test was fragile on servers with the default pg_catalog.simple config.

Other factors

The replacement query uses <@> with to_bm25query('postgresql inheritance', 'child1_bm25_idx'), which exercises pg_textsearch's own code path (an improvement over the previous test). The expected output correctly reflects the single LIMIT 1 result. Bug hunting found no issues.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!