Problem
Tooling drift across the repo: no lint/format/coverage config, two sources of truth for mathjs, and the smoke suite depends on a global install.
Evidence
- No lint, format, or coverage config in the repo (an eslint-disable comment hints at ad-hoc use only).
- Two math.js sources must be kept in sync by hand: the vendored bundle (
vendor/math.js) vs the npm mathjs the tests use.
test/smoke.js requires Playwright from a global NODE_PATH — friction for fresh clones.
- Root-level
audit2.js / audit_screenshots.js are undocumented one-off scripts with hardcoded paths, not in CI or the README file table.
- No test filter flag, so running one test means running the whole suite.
Expected
- A minimal lint/format config (or an explicit decision to have none, documented).
- A single mathjs source (or a sync check in CI).
- Smoke/audit tooling documented or removed.
Impact
Each of these is small, but together they make the toolchain the least trustworthy part of an otherwise disciplined repo.
Problem
Tooling drift across the repo: no lint/format/coverage config, two sources of truth for mathjs, and the smoke suite depends on a global install.
Evidence
vendor/math.js) vs the npmmathjsthe tests use.test/smoke.jsrequires Playwright from a globalNODE_PATH— friction for fresh clones.audit2.js/audit_screenshots.jsare undocumented one-off scripts with hardcoded paths, not in CI or the README file table.Expected
Impact
Each of these is small, but together they make the toolchain the least trustworthy part of an otherwise disciplined repo.