sol repos can commit forge fmt drift: no Solidity hook in the pre-commit suite rainix-sol-static gates on

[thedavidmeister]

rainix-sol-static gates Solidity with forge fmt --check, but nothing in the
pre-commit suite it generates looks at a .sol file. So a Solidity repo can
commit with every hook green and red CI on formatting alone — which is what just
happened on rainlanguage/rain.deploy#26 (commit
eb0b3f5: all 11 hooks passed, rainix / static failed on one over-long call
in a test file, one more commit and one more full CI run to fix).

The drift

.github/workflows/rainix-sol-static.yaml runs, as workflow steps:

• slither .
• forge fmt --check
• rainix-sol-single-contract

The generated .pre-commit-config.yaml in a consumer holds eleven hooks —
deadnix, denofmt, nil, nixfmt, no-consumer-prettier,
prettier-rainix, rustfmt-conditional, shellcheck, statix, taplo,
yamlfmt. None of them match \.sol$. On a pure-Solidity change every hook
reports "no files to check" and the commit succeeds.

Rust does not have this problem: rustfmt-conditional is a hook, so
cargo fmt drift is caught at commit time. Solidity has no equivalent, and it
is the same formatter-drift failure the same suite already solved once.

Ask

Add a forgefmt-conditional hook to the sol shell's git-hooks.nix, guarded
the way rustfmt-conditional is (command -v forge || exit 0, so the
flake-check sandbox and non-Solidity repos skip it), matching files: \.sol$.

That makes forge fmt drift a commit-time failure in the shell CI already
pins, rather than a CI round trip.

The same shape, not in this ask

slither, rainix-sol-single-contract, no-ignored-tests, no-submodules,
frozen-snapshots-append-only and no-custom-natspec are also CI-only with no
local hook. Formatting is the one that fires most often and is purely
mechanical, so it is worth fixing on its own; whether the rest belong in
pre-commit is a separate call — several are repo-state checks rather than
per-file ones, and slither is too slow for a commit hook.

Also worth a look while in there

rain.deploy/CLAUDE.md documents nix develop -c rainix-sol-static as the
local static command. That binary is not on the consumer devshell's PATH —
it fails exec: rainix-sol-static: not found. There is no single local command
for the sol static gate today, which is part of why the drift goes unnoticed.