valkyrjaio / valkyrja-ts / 30325727451 / 1

100%

26.x: 100%

push

github

[Application] Drop the interpreter path from argv before building the CLI input (#92)

# Description

Fixes the CLI entry point reading the wrong argv slots, so **every
command name was wrong in real
use**.

`Cli.getInput()` passed `process.argv` straight to
`InputFactory.fromGlobals()`. Node's
`process.argv` leads with the **interpreter path** — `[execPath, script,
...args]` — whereas the
factory (a faithful port of the PHP reference) expects the vector to
start with the caller, as
PHP's `$argv` does. Everything therefore landed one slot late:

```
node bin/valkyrja app:version
```

| | before | after |
|---|---|---|
| caller | `/usr/local/bin/node` | `bin/valkyrja` |
| command name | `bin/valkyrja` | `app:version` |
| arguments | `['app:version']` | `[]` |

The fix is `process.argv.slice(1)` at the entry point — deliberately
there rather than in the
factory, since Node is the only runtime that leads with an interpreter
path; every other port's
vector already starts with the caller.

## Why the existing tests missed it

`Cli.test.ts` set a synthetic, PHP-shaped `process.argv = ['cli',
'command', …]` — a vector Node
never produces — so the entry point's off-by-one was invisible while the
assertions passed. The
message-mapping fidelity tests added in #91 don't cover it either: they
call
`InputFactory.fromGlobals()` directly with their own array, so they
exercise the factory (which is
correct) and never the entry point (which was not).

Every vector in `Cli.test.ts` is now shaped as Node really produces it,
and three of its cases fail
against the old code — confirming the regression coverage is real.

This came out of the Java message-mapping fidelity work
(valkyrjaio/valkyrja-java#55); the per-language argv conventions are
documented in
valkyrjaio/architecture#68.

## Types of changes

- [ ] Improvement _(non-breaking change which improves code)_
- [x] Bug fix _(non-breaking change which fixes an issue)_
- [ ] New feature _(non-breaking change which ad... (continued)

1587 of 1587 branches covered (100.0%)

Branch coverage included in aggregate %.

3473 of 3473 relevant lines covered (100.0%)

29.04 hits per line