Bump cat_tools floor to 0.3.0, wire CI to the git-source override

[jnasbyupgrade]

Redesigned alongside PR #10's redesign, splitting cleanly: #10 is the
install-mechanism override (CAT_TOOLS_GIT_REF/CAT_TOOLS_SKIP_INSTALL,
defaults empty/unchanged); this PR is the actual version decision.

What this PR does

• Bumps META.in.json/META.json's declared cat_tools floor from 0.2.1
  to 0.3.0 -- extension_drop.sql calls
  cat_tools.routine__parse_arg_types_text(), which only exists starting at
  cat_tools 0.3.0. PGXN's published listing is stuck at a stale 2017 0.2.1
  lacking it, so a plain pgxn install extension_drop now fails cleanly at
  dependency-resolution time instead of installing successfully and breaking
  later, obscurely, at CREATE EXTENSION time.
• Sets CAT_TOOLS_GIT_REF: 0.3.0 at .github/workflows/ci.yml's workflow
  level (not per-job), so every job that builds the extension -- including
  jobs in other stacked PRs that already read this value via
  make print-CAT_TOOLS_GIT_REF (e.g. the pg_tle job) -- picks it up
  automatically with no further changes needed there.
• Updates RELEASE.md's release-blocking guardrail: the real signal for
  "is a release actually cuttable" is now whether ci.yml's
  CAT_TOOLS_GIT_REF is set to a non-empty value, not whether the Makefile
  merely declares the variable -- it always does now, as a normally-empty
  opt-in override, so its mere existence isn't a useful signal anymore.

Verified locally (PG17)

• CAT_TOOLS_GIT_REF=0.3.0 make install (the real path CI exercises):
  installs cat_tools 0.3.0 from git, then installs extension_drop itself,
  end to end.
• Confirmed environment-variable precedence works as needed: an exported
  CAT_TOOLS_GIT_REF correctly overrides the Makefile's ?= (empty)
  default -- verified via make -n before and after exporting it.

[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: 4890b0ad-aa24-415f-b7d5-a6436a912195

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 3m 6s —— View job

---

Claude's review

Todo list:

• Gather context (diff, changed files, CLAUDE.md)
• Understand the request (code review per /code-review)
• Review changed files: .github/workflows/ci.yml, META.in.json, META.json, RELEASE.md
• Post review feedback

---

Note on scope: the actual diff (git diff origin/fix-cat-tools-install...HEAD) doesn't touch Makefile — CAT_TOOLS_GIT_REF already existed there (default empty) from the prior PR. This PR's one commit (b9a135c) instead (1) bumps the cat_tools floor in META.in.json/META.json from 0.2.1 → 0.3.0, (2) sets CAT_TOOLS_GIT_REF: 0.3.0 as a workflow-level env: in ci.yml, and (3) adds a release guardrail to RELEASE.md. The PR description's "What changed" section (which says Makefile's CAT_TOOLS_GIT_REF moved from master to 0.3.0) doesn't match this — worth updating the description so reviewers aren't looking for a Makefile change that isn't there.

Findings

1. Stale cat_tools>=0.2.1 pin left in the Makefile's pgxn install fallback — Makefile:40

This PR's purpose is bumping the cat_tools floor to 0.3.0 (per META.in.json/META.json), but the else branch of the cat_tools target still runs:

pgxn install 'cat_tools>=0.2.1' --sudo;

This branch is exactly the path RELEASE.md's new guardrail section describes reverting to once CAT_TOOLS_GIT_REF is unset again. If someone removes the CI override later without also noticing this literal string, pgxn install would happily accept a cat_tools that doesn't have routine__parse_arg_types_text (needed by sql/extension_drop.sql:38), silently reintroducing the exact bug RELEASE.md documents (PRs #6/#7). The PR's own body even quotes the intended end state as pgxn install 'cat_tools>=0.3.0' --sudo, so this looks like an oversight rather than intentional.

Fix this →

2. README.md:7 still advertises the old floor

Requires [cat_tools](https://pgxn.org/dist/cat_tools/) >= 0.2.1.

Same issue as above, user-facing this time — should read >= 0.3.0 to match META.in.json/META.json.

Fix this →

What looked right

• ci.yml's workflow-level env: placement is correct YAML and applies to all jobs (lint, test matrix, all-checks-passed), matching the stated intent of not needing a per-job copy.
• RELEASE.md's new "Critical" section is clear, and correctly points at checking ci.yml's actual value rather than the Makefile's variable existence — good given CAT_TOOLS_GIT_REF is now unconditionally defined (empty by default) in the Makefile.
• META.in.json/META.json stay in sync with each other, and the X_comment update accurately reflects why 0.3.0 is now required (per the routine__parse_arg_types_text history already documented in RELEASE.md).
  · Branch: pin-cat-tools-0.3.0-tag