docs: add CLAUDE.md documenting the local pre-push gate - #186
Conversation
Names the two local checks CI runs: pre-commit run --all-files (black pinned to 24.2.0, bin/compile.py, bin/check_lints.py, hygiene hooks) and the docker-compose pg_regress suite. Every command verified against the repo's own .pre-commit-config.yaml and .github/workflows/*.yaml.
…DE.md PG_VERSION never selects a Postgres version -- dockerfiles/docker-compose.yml only reads SUPABASE_VERSION for its build arg, verified via docker compose config with both vars set. The rmi-first staleness fix and --rm flag now match the repo's own README/new-lint skill convention. Pre-commit builds its own isolated black env from the pinned rev, so the venv-install instruction was actively wrong; added the real prerequisite (python3.12 on PATH) instead. Also names bin/compile.py's tracked output (splinter.sql) and completes the hygiene-hook list. Every documented command re-run for real (pre-commit run --all-files: 8/8 hooks passed; the docker test suite: 28/28 tests passed) after this edit.
|
@claude review |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
LGTM — this adds a new CLAUDE.md documentation file with no code changes.
What was reviewed: verified the described pre-commit hooks (black 24.2.0/python3.12, compile.py, check_lints.py, hygiene hooks) against .pre-commit-config.yaml; verified the PG_VERSION vs SUPABASE_VERSION claim against .github/workflows/test.yml and dockerfiles/docker-compose.yml; confirmed docker-compose.yml only reads SUPABASE_VERSION for the build arg, matching the described no-op behavior. All factual claims in the new file check out.
Extended reasoning...
Overview
The diff adds a single new file, CLAUDE.md, containing guidance for Claude Code sessions working in this repo: an overview of what splinter does, and two subsections describing the local pre-commit gate and the docker-based pg_regress test suite. No source code, SQL lints, scripts, or CI workflows are modified.
Security risks
None. This is a documentation-only change with no executable code, no secrets, and no changes to build or CI configuration.
Level of scrutiny
Low. This warrants a light-touch factual check: does the documentation accurately describe the repository's actual tooling and CI behavior? I cross-checked every specific claim (hook versions, script names, env var behavior, the "docker rmi -f is required not optional" caveat) against .pre-commit-config.yaml, .github/workflows/pre-commit_hooks.yaml, .github/workflows/test.yml, and dockerfiles/docker-compose.yml, and all matched.
Other factors
No CODEOWNERS restriction found on CLAUDE.md, no prior review history to weigh, and no findings were reported by the bug hunt. This is exactly the kind of small, self-contained, mechanical change (docs addition, verifiably accurate) that merits approval without requiring further human review.
1 participant
Summary
Adds a
CLAUDE.mddocumenting what this repo is and its local pre-push gate:pre-commit run --all-filesand the docker-compose pg_regress suite. Every command in the file was run for real against this repo, not just read from the workflow YAML.(bot-generated information collapsed below)
Details
splinterhad noCLAUDE.mdat all before this PR.PG_VERSIONnever selects a Postgres version, sincedockerfiles/docker-compose.ymlonly readsSUPABASE_VERSIONfor its build arg (confirmed viadocker compose configwith both variables set). The file now documents the working form, matching this repo's own README and.claude/skills/new-lint/SKILL.md.pre-commitalready manages that itself; the hygiene-hook list was missingcheck-added-large-files.pre-commit run --all-files(8/8 hooks passed) and the docker-compose pg_regress suite (28/28 tests passed).Resolves https://linear.app/supabase/issue/INDATA-1481