mirror of
https://github.com/nodejs/node.git
synced 2025-05-03 09:52:21 +00:00
7f54dd4f07
When the test fails, it usually ends up failing on the assertion that
`wc` did not find as many bytes as it expected. That may not be
helpful if it is caused by a failure earlier in the command
pipeline ( cat | grep | wc ). Move the byte check to
`process.on('exit')` so that other error handlers that report the
existence of errors run first.
Refs: https://github.com/nodejs/node/issues/25988
PR-URL: https://github.com/nodejs/node/pull/37632
Reviewed-By: Gireesh Punathil <gpunathi@in.ibm.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
74 lines
2.3 KiB
JavaScript
74 lines
2.3 KiB
JavaScript
'use strict';
|
|
const common = require('../common');
|
|
const assert = require('assert');
|
|
const path = require('path');
|
|
const fs = require('fs');
|
|
const os = require('os');
|
|
const spawn = require('child_process').spawn;
|
|
const tmpdir = require('../common/tmpdir');
|
|
|
|
let cat, grep, wc;
|
|
|
|
const KB = 1024;
|
|
const MB = KB * KB;
|
|
|
|
|
|
// Make sure process chaining allows desired data flow:
|
|
// check cat <file> | grep 'x' | wc -c === 1MB
|
|
// This helps to make sure no data is lost between pipes.
|
|
|
|
{
|
|
tmpdir.refresh();
|
|
const file = path.resolve(tmpdir.path, 'data.txt');
|
|
const buf = Buffer.alloc(MB).fill('x');
|
|
|
|
// Most OS commands that deal with data, attach special
|
|
// meanings to new line - for example, line buffering.
|
|
// So cut the buffer into lines at some points, forcing
|
|
// data flow to be split in the stream.
|
|
for (let i = 1; i < KB; i++)
|
|
buf.write(os.EOL, i * KB);
|
|
fs.writeFileSync(file, buf.toString());
|
|
|
|
cat = spawn('cat', [file]);
|
|
grep = spawn('grep', ['x'], { stdio: [cat.stdout, 'pipe', 'pipe'] });
|
|
wc = spawn('wc', ['-c'], { stdio: [grep.stdout, 'pipe', 'pipe'] });
|
|
|
|
// Extra checks: We never try to start reading data ourselves.
|
|
cat.stdout._handle.readStart = common.mustNotCall();
|
|
grep.stdout._handle.readStart = common.mustNotCall();
|
|
|
|
// Keep an array of error codes and assert on them during process exit. This
|
|
// is because stdio can still be open when a child process exits, and we don't
|
|
// want to lose information about what caused the error.
|
|
const errors = [];
|
|
process.on('exit', () => {
|
|
assert.deepStrictEqual(errors, []);
|
|
});
|
|
|
|
[cat, grep, wc].forEach((child, index) => {
|
|
const errorHandler = (thing, type) => {
|
|
// Don't want to assert here, as we might miss error code info.
|
|
console.error(`unexpected ${type} from child #${index}:\n${thing}`);
|
|
};
|
|
|
|
child.stderr.on('data', (d) => { errorHandler(d, 'data'); });
|
|
child.on('error', (err) => { errorHandler(err, 'error'); });
|
|
child.on('exit', common.mustCall((code) => {
|
|
if (code !== 0) {
|
|
errors.push(`child ${index} exited with code ${code}`);
|
|
}
|
|
}));
|
|
});
|
|
|
|
let wcBuf = '';
|
|
wc.stdout.on('data', common.mustCall((data) => {
|
|
wcBuf += data;
|
|
}));
|
|
|
|
process.on('exit', () => {
|
|
// Grep always adds one extra byte at the end.
|
|
assert.strictEqual(wcBuf.trim(), (MB + 1).toString());
|
|
});
|
|
}
|