0.3.0 -> stable

[jnasbyupgrade]

Per RELEASE.md steps 4/7: default_version moves from the just-released 0.3.0 to the permanent stable pseudo-version, so routine source edits on master regenerate sql/cat_tools--stable.sql.in instead of corrupting the frozen 0.3.0 install script.

• cat_tools.control / META.in.json: default_version -> stable.
• sql/cat_tools--0.3.0--stable.sql.in: seeded, empty -- the live update script future fixes extend.
• sql/cat_tools--stable.sql.in: gitignored, not tracked (pure derived output, no real version to anchor it).
• HISTORY.asc: new STABLE section opened for the next release.
• RELEASE.md: clarified (once, in step 4) that .sql.in install scripts are always tracked and stable is the one exception -- step 7 now just references that instead of restating it.
• CLAUDE.md: added rule 6 to "SQL file conventions" cross-referencing this exception to pgxntool's own "don't gitignore a skipped version" guidance, since stable isn't a skipped version (a one-time transient case) but a permanently-current one (would be re-diffed every commit if tracked, for no test-coverage gain).

[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: 5f367d94-1702-4e02-9c9c-13c6e001e4bf

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 49s —— View job

---

Code review

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

Reviewed the mechanical version-flip (cat_tools.control, META.in.json/META.json → stable), the new HISTORY.asc STABLE section, the new accumulator update script sql/cat_tools--0.3.0--stable.sql.in, and the generated sql/cat_tools--stable.sql.in:

• sql/cat_tools--stable.sql.in is byte-identical to sql/cat_tools--0.3.0.sql.in (confirmed via diff) — expected, since no source content has changed since the 0.3.0 release yet.
• Naming (cat_tools--<prev>--<new>.sql.in) and structure match the existing update-script convention (e.g. sql/cat_tools--0.2.3--0.3.0.sql.in).
• META.in.json/META.json were updated together with matching version fields, consistent with pgxntool's "edit META.in.json, META.json is generated from it" convention.
• HISTORY.asc's new STABLE section matches the wording prescribed in RELEASE.md's step 7.
• Per CLAUDE.md's "SQL file conventions" Add support for Postgres 12 - 17 #5, no version-specific file was hand-edited — the new stable install script and the update script are both newly generated/added, and the frozen 0.3.0 files are untouched.

Note: I wasn't able to run make/git write commands in this sandboxed session (approval-gated), so I relied on static diff review plus the extensive verification already documented in the PR description (make verify-results, update-path testing, structural diff, checksum comparison).

[github-actions]

Claude finished @jnasbyupgrade's task in 4m 5s —— View job

---

Reviewing PR #81

• Gather context (diff, changed files, CLAUDE.md rules)
• Review mechanical version-flip changes (cat_tools.control, META.in.json/META.json)
• Review HISTORY.asc new STABLE section
• Review new update script sql/cat_tools--0.3.0--stable.sql.in
• Review generated sql/cat_tools--stable.sql.in handling
• Review RELEASE.md and sql/.gitignore changes
• Post final review feedback

Code review

PR description is now out of sync with the diff (needs fixing before merge). The
description says:

sql/cat_tools--stable.sql.in is generated and tracked like any other per-version
install script.

but the current diff does the opposite — sql/.gitignore (+4 lines) adds
cat_tools--stable.sql.in to the ignore list, and the file is not tracked. This isn't a
stale leftover from a prior push either — RELEASE.md's new wording is internally
consistent with the gitignored approach (RELEASE.md:75: "...unrelated to
sql/cat_tools--stable.sql.in, which is gitignored... precisely because it's pure derived
output with no independent value..."). So the code was intentionally revised since the
description was written, and the description's opening paragraph — which, per this repo's
CLAUDE.md, becomes the squash commit message — needs to be updated to match, or the merged
commit message will describe behavior the code doesn't have.

Worth a second look: this deviates from pgxntool's own documented convention, though for
articulable reasons. pgxntool/README.asc ("Don't .gitignore a Skipped Version — rm
It Once", around line 405) explicitly says not to .gitignore an omitted version's
generated install script, and to rm it once instead — a .gitignore entry is framed as
"perpetual clutter" for a file that's normally only transiently present after a version
bump. stable doesn't fit that description cleanly (it's not a one-time transient leftover
— it's the permanently current version between releases, so unlike a normal skipped
version it would be re-regenerated and re-diffed on every single source edit if tracked,
with zero incremental test-coverage value since CI already builds+tests a fresh install
from source on every push regardless of whether the copy is committed). That reasoning is
sound, but:

• CLAUDE.md's "SQL file conventions" section (rules 0–4) doesn't mention this exception at
  all — a reader following just CLAUDE.md would expect version-specific install scripts to
  be tracked by default and could reasonably flag the new sql/.gitignore line as a
  violation of pgxntool's convention. Given pgxntool's README has an explicit callout that
  this exact topic ("CAUTION: This is especially relevant for AI coding agents...") is easy
  to get wrong without the rationale spelled out, it'd be worth adding a short note to
  CLAUDE.md's SQL file conventions cross-referencing the sql/.gitignore comment/RELEASE.md
  rationale, so this doesn't need to be re-derived (or mistakenly "fixed") later.

Everything else checks out:

• cat_tools.control / META.in.json / META.json version flips to stable are
  consistent with each other, and META.json was regenerated to match META.in.json
  (pgxntool convention: never hand-edit META.json).
• HISTORY.asc's new top STABLE section is correctly placed above the frozen 0.3.0
  entry.
• sql/cat_tools--0.3.0--stable.sql.in is a genuinely empty file (0 bytes) — correct, since
  no source changes have landed since 0.3.0 shipped, and per CLAUDE.md's terminology rule
  this is an "update script," tracked per rule 3.
• No frozen sql/cat_tools--0.3.0.sql[.in] files were touched, consistent with CLAUDE.md
  rule 5 (never hand-edit a non-current version file) and pgxntool's "never hand-edit an old
  version file" guidance.
• .gitignore's new entry is scoped precisely to cat_tools--stable.sql.in (not a broader
  glob), so it doesn't accidentally hide the 0.3.0--stable update script or any other
  tracked per-version file.

I wasn't able to run make/git write commands in this sandboxed session (approval-gated
for some git plumbing commands), so this is a static review of the diff plus the extensive
verification already documented in the PR description (make verify-results, update-path
testing, structural diff, checksum comparison).