Skip to content

development.md

Latest commit

 

History

History
72 lines (55 loc) · 2.65 KB

development.md

File metadata and controls

72 lines (55 loc) · 2.65 KB
Raw

Development

npm install
npm run check     # lint + typecheck + test + build
npm test          # vitest
npm run coverage  # vitest with a coverage report
npm run build     # vite library build into dist/

Every script calls its tool by bare name (eslint ., vitest run), which relies on npm putting node_modules/.bin at the front of PATH for the duration of the script. There is no .npmrc in the repository and no custom script-shell, so on a stock toolchain this works with no setup beyond npm install.

Windows

npm run cannot find the local binaries under Volta

If you manage Node with Volta on Windows, npm run and npx may fail to resolve any project binary:

> eslint .

'eslint' is not recognized as an internal or external command,
operable program or batch file.

The same failure hits every script that shells out to a dependency — lint, test, typecheck, build — and therefore check as well.

This is not a problem with the repository, and the following were all ruled out on a machine that reproduced it:

Checked Result
node_modules\.bin\eslint.cmd exists Yes, and runs correctly when invoked directly
PATHEXT contains .CMD Yes
node_modules\.bin present in the script PATH Yes, confirmed via npm run env
eslint invoked from the shell directly Works
Downgrading to the npm bundled with Node Still fails

The remaining difference is Volta's shim layer, which sits between the shell and node/npm/npx. Node, npm and npx all resolve to C:\Program Files\Volta\ rather than to a Node installation.

Workaround — call the shims directly, bypassing npm run:

.\node_modules\.bin\eslint.cmd .
.\node_modules\.bin\vitest.cmd run
.\node_modules\.bin\tsc.cmd --noEmit
.\node_modules\.bin\vite.cmd build

Contributors who are not on Volta, and CI, are unaffected — npm run check works normally there.

The stack safety tests are slow

The two tests in the stack safety block build their fixture on a real filesystem: one creates 20,000 files in a single directory, the other nests 400 directories. Creating those files takes well under a second on Linux and roughly 25 times longer on Windows, where NTFS and real-time antivirus scanning charge per file created.

Both tests carry a 60 second timeout for that reason. The traversal being tested is fast on every platform — walking the 20,000 entries takes about 70 ms — so a failure here is worth reading carefully before assuming it is just slowness. If these tests dominate your local runs, excluding antivirus scanning from your temp directory is the change that helps most.