Haskell: shelltestrunner, treefmt --ci (fourmolu/hlint/actionlint), nix-built wrapper commands (postgrest-*)

Start with project_validator/ecosystems/nix.py around line 83, the only place the payload shows treefmt, then grep project_validator/ for pytest, sphinx-build and shelltest to find where recognised test and lint commands are listed. The nix wrapper case needs a design decision about how to treat opaque nix-provided commands, so agree on that before coding. Done means the corpus results for PostgREST and hledger change as expected.

hledger's 585 functional tests are shelltest files (*.test, unrecognised) run by every workflow; PostgREST runs fourmolu, hlint and actionlint through treefmt --ci (treefmt-nix config), and its spec/hasql/io tests, pytest, sphinx-build -b linkcheck and weeder through nix-built postgrest-* wrappers from a composite action's tools: input → code-formatting, static-analysis, ci-lint, tests-run and link-check all wrong. treefmt is a common multiplexer in Nix-managed repos (treefmt.toml / treefmt-nix).

Evidence from the corpus audit

PostgREST/postgrest / code-formatting, static-analysis, ci-lint (treefmt)

• kind: UNSUPPORTED
• actual: code-formatting FAIL Haskell: fourmolu.yaml is defined but CI does not run ormolu, fourmolu or stylish-haskell; static-analysis FAIL Haskell: no static analyzer runs; ci-lint WARN CI does not lint its GitHub Actions configuration
• expected: all three are run by treefmt --ci (fails on unformatted files) with the treefmt-nix configuration enabling fourmolu, hlint and actionlint. treefmt (treefmt.toml / treefmt-nix) is a common multiplexer in Nix-managed repos; the validator knows the word only as a Nix formatter.
• evidence: .github/workflows/check.yaml:33-34: - name: Run treefmt / run: treefmt --ci; nix/treefmt.nix:26: programs.actionlint.enable = true;; :29-30: programs.fourmolu.enable = true; / programs.hlint.enable = true;; .github/actionlint.yml (actionlint config present); project_validator/ecosystems/nix.py:83: must take it: nixfmt, alejandra and treefmt do). (only mention)
• repro: treefmt.toml with [formatter.fourmolu] command = "fourmolu" options = ["--mode","check"] and a step run: treefmt --ci
• confidence: high

PostgREST/postgrest / tests-run, link-check, static-analysis (nix wrapper commands)

• kind: UNSUPPORTED
• actual: tests-run FAIL Haskell: tests found in 1 file(s), but no CI configuration runs cabal test (or stack test); Documentation site: ... no CI runs sphinx-build; Python: tests found in 14 file(s), but no CI configuration runs pytest; link-check WARN no link check found in CI configuration; static-analysis Haskell: no static analyzer runs
• expected: these all run, through nix-built postgrest-* wrapper scripts provided by the ./.github/actions/setup-nix composite action's tools: input: spec/hasql/io tests, pytest, sphinx-build -b linkcheck (scheduled workflow named Linkcheck), weeder. The commands are opaque to the validator (their bodies are in nix/tools/*.nix), so this is a gap rather than a parser bug; a "nix-provided command" notion (or reading checkedShellScript bodies in nix/tools) would be needed.
• evidence: .github/workflows/test.yaml:103-116: run: postgrest-with-${{ matrix.pgVersion }} postgrest-test-spec ... run: postgrest-with-${{ matrix.pgVersion }} postgrest-test-io -vv; nix/tools/tests.nix:116: ${ioTestPython}/bin/pytest --ignore=test/io/test_big_schema.py --ignore=test/io/test_replica.py -v test/io; .github/workflows/linkcheck.yaml:3-5,25: schedule: / - cron: '1 2 * * 3' ... run: postgrest-docs-linkcheck; nix/tools/docs.nix:175: ${python}/bin/sphinx-build --color -b linkcheck . ../.docs-build; nix/tools/tests.nix:168: docs = "Run spec and io tests while collecting hpc coverage data. First runs weeder to detect dead code.";
• repro: unclear (needs reading nix shell-script derivations)
• confidence: medium

simonmichael/hledger / tests-run (shelltestrunner)

• kind: UNSUPPORTED
• actual: Haskell tests-run PASS via stack install --test, but the 585 functional tests are invisible: .map.json unrecognised_suffixes: {"suffix": ".test", "files": 585}
• expected: shelltestrunner (shelltest) is the standard functional-test runner for Haskell CLIs (hledger, pandoc-adjacent tools); recognising shelltest <dir> and *.test files would make the Haskell test picture complete.
• evidence: .github/workflows/ci.yml:351: COLUMNS=80 $stack exec -- shelltest --execdir -j16 hledger/test -x /_ -x ledger-compat/ledger-baseline -x ledger-compat/ledger-regress -x ledger-compat/ledger-collected; hledger/test/cli/addons.test; grep -rn -i shelltest project_validator/ → no match
• repro: n/a (gap)
• confidence: medium

Found by the regression-corpus audit (235 real repositories, one or more per ecosystem profile; corpus and expected results in corpus/). Evidence is quoted from the audited checkouts at the commits pinned in corpus/projects.toml; <checkouts>/<owner>__<repo> is such a checkout.

🤖 Generated with Claude Code

https://claude.ai/code/session_01ByTXNqDWdM7HzezrFtEvQt