Update pgxntool to 2.3.0 (release tag)

[jnasbyupgrade]

Updates the vendored pgxntool subtree from 2.2.0 to the 2.3.0 release, which fixes make test always exiting 0 even when every test fails (issues #35/#79) and widens the DATA wildcard to install historical one-dash version scripts like sql/object_reference--0.1.0.sql (issue #48) -- both already worked around elsewhere in this repo's own history.

• make test now correctly exits non-zero on a real regression: confirmed locally against this repo's own current, pre-existing cat_tools-0.3.0-deprecation-driven test failure -- make test now exits 2, where it previously always exited 0 with no failure signal at all.
• make install now installs sql/object_reference--0.1.0.sql without any local Makefile workaround, since the widened DATA wildcard picks up one-dash historical version files directly.
• No changes outside pgxntool/ were needed; update-setup-files.sh's reconciliation found .gitignore/test/deps.sql unchanged from the current pgxntool release.

The still-open PR #18 currently carries a local DATA += sql/object_reference--0.1.0.sql line and an installcheck: install ordering fix for these same two issues -- both become droppable once this merges and #18 rebases onto it.

Co-Authored-By: Claude noreply@anthropic.com

[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: edb2b289-4a61-489d-87d9-da4ad7dfd7a6

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 29s —— View job

---

Reviewing PR #31: Update pgxntool to 2.3.0

• Gather context (PR description, diff vs origin/master, repo Makefile)
• Review pgxntool/base.mk functional changes
• Review vendored docs (CLAUDE.md, HISTORY.asc, README.asc/html)
• Check whether repo-local workarounds mentioned in the PR body are affected
• Post review

---

Summary

This is a clean vendor-only update (all 5 changed files are under pgxntool/, diff matches PR description: +148/-37). The two upstream fixes are real and match what's described:

• base.mk: test: now does exit 1 after cat-ing regression.diffs on failure, instead of silently exiting 0 (fixes PR #2 content not carried into #17: README.md and META.json/META.in.json metadata fixes missing #35/#79).
• base.mk: DATA now uses $(sort $(EXTENSION__CURRENT_VERSION__FILES) $(wildcard sql/*--*.sql)), widening the wildcard from two-dash-only (*--*--*.sql) to also match one-dash historical files like sql/object_reference--0.1.0.sql, with $(sort …) added to dedupe the now-overlapping current-version file (fixes #48).
• base.mk: adds installcheck: install upstream and verify-results/results now depend on $(TEST_DEPS) instead of test, preserving fail-fast semantics for make test while keeping make results/verify-results able to inspect diffs after a failure.
• PGTLE_VERSION renamed to PGXNTOOL_PGTLE_VERSION to avoid silent collision with a CI job's own PGTLE_VERSION env var (issue #78) — no usages of the old name remain in this repo (checked, none found outside pgxntool/).

Finding: root Makefile's local installcheck: install workaround is now redundant

Makefile:8-13 (top-level, not pgxntool/) already carries a local workaround added in #9 ("Update pgxntool to 2.2.0"):

# pgxntool's check-stale-expected target (added in pgxntool 2.2.0) depends on
# installcheck but is listed before install in TEST_DEPS, and Make evaluates a
# target's prerequisites in file-parse order across stanzas -- so plain
# `make test` ran installcheck before install ever happened. Force installcheck
# to require install locally until that's fixed upstream.
installcheck: install

pgxntool 2.3.0 (this PR) adds the exact same edge upstream in pgxntool/base.mk (installcheck: install, with its own explanatory comment). Declaring the same prerequisite twice is harmless to Make, but it's now dead weight, and the local comment's "until that's fixed upstream" is stale — it is fixed upstream, in this very PR. Worth removing Makefile:8-13 here (or in a quick follow-up) while this is fresh, the same way PR #18 plans to drop its own copy of this workaround plus its DATA += sql/object_reference--0.1.0.sql line once it rebases onto this.

Fix this →

Other notes

• Docs-only diffs in CLAUDE.md, HISTORY.asc, README.asc, README.html are consistent with the upstream 2.3.0 release notes — nothing repo-specific to flag there.
• DATA wildcard widening looks correct: sql/*--*.sql still matches two-dash upgrade scripts (a--b.sql) as well as one-dash full-install scripts (a.sql), and $(sort …) correctly resolves the resulting overlap with EXTENSION__CURRENT_VERSION__FILES that the PR description flags as otherwise breaking make install.
• PR body's claims (exit code 2 on the pre-existing cat_tools failure, no changes needed outside pgxntool/) are consistent with what's in the diff; I did not re-run make test myself in this sandbox (no live Postgres instance available here) but the logic checks out by inspection.

Overall: no blocking issues. One optional cleanup suggested above (redundant installcheck: install in the root Makefile).