# Which binary ran?

> Diagnose a build that differs between machines: list every install of a tool on PATH, best first, and report which one a run actually used.

Source: https://bellpull.interlace.tools/docs/recipes/which-binary-ran

Two machines, one command, two answers — and usually two installs of the same tool on one of
them. The fastest way to the finding is to ask for every match, and for the `PATH` entry each
came from:

```js title="doctor.mjs"
import { relative } from 'node:path';

import { run } from 'bellpull';
import { whichAllSync } from 'bellpull/which';

import { install, runtimeWith } from './tools.mjs';

install();
const runtime = runtimeWith('bin', 'vendor/bin');
const here = (p) => relative(runtime.cwd, p);

const hits = whichAllSync('greet', { runtime });
if (hits.length > 1) console.log(`${hits.length} installs of greet on PATH:`);
for (const [i, hit] of hits.entries()) console.log(`  ${i === 0 ? '→' : ' '} ${here(hit.path)}  (PATH entry ${here(hit.from)})`);

const result = await run('greet', [], { runtime });
console.log(`ran ${here(result.executable.path)}: ${result.stdout.trim()}`);
```

```text title="node doctor.mjs"
2 installs of greet on PATH:
  → bin/greet  (PATH entry bin)
    vendor/bin/greet  (PATH entry vendor/bin)
ran bin/greet: hello from bin
```

Put this behind a `doctor` subcommand and a bug report carries the answer instead of a request
for `echo $PATH`. The `run` result's `executable` is the one that actually ran, because
bellpull spawns the file it resolved rather than handing the name back to be resolved again.