flaky test.parallel/test-diagnostics-channel-net

[kvakil]

Test

test-diagnostics-channel-net

Platform

Linux x64

Console output

00:32:11 not ok 653 parallel/test-diagnostics-channel-net
00:32:11 ---
00:32:11 duration_ms: 0.269
00:32:11 severity: fail
00:32:11 exitcode: 1
00:32:11 stack: |-
00:32:11 Mismatched <anonymous> function calls. Expected exactly 1, actual 0.
00:32:11 ...

Build links

• https://ci.nodejs.org/job/node-test-commit-custom-suites-freestyle/26384/
• https://ci.nodejs.org/job/node-test-commit-custom-suites-freestyle/26354/
• https://ci.nodejs.org/job/node-test-commit-custom-suites-freestyle/26384/
• Around 7% of failures in CI Reliability 2022-08-05 reliability#337

Additional information

1. Potentially only happens in worker test suite?
2. Could not reproduce locally with:

for i in {1..10000}; do
 ./node --abort-on-uncaught-exception \
 ./tools/run-worker.js test/parallel/test-diagnostics-channel-net.js \
 || echo "fail $i" &
done |& grep "fail "

3. Not clear to me which callback is not getting invoked.

[kvakil]

@theanarkh any thoughts? It's weird also that the stack trace does not include any indication what's not being called. :\

edit: never mind, I think I found it: #44144

[theanarkh]

Thanks. I have no ideas now, i will take a look later. But i think it is not related to gc because this case will keep the channel object alive. It is different from #42170. cc @Qard.
.

[theanarkh]

@Qard Hi, what do you think about this case. Is it the wrong way to use diagnostics_channel?

[Qard]

Already approved the fix in #44144.

[theanarkh]

But there's a lot of usage in the code, and if there's a problem, it all needs to be fixed ?

[theanarkh]

Running the code as follows will trigger this error always(./node --expose-gc test/parallel/test-diagnostics-channel-net.js). The callback of onGC will be triggered before timeout.

'use strict';
const common = require('../common');
const assert = require('assert');
const net = require('net');
const dc = require('diagnostics_channel');
const onGC = require('../common/ongc');
const { writeFileSync } = require('fs');

const netClientSocketChannel = dc.channel('net.client.socket');
const netServerSocketChannel = dc.channel('net.server.socket');

onGC(netClientSocketChannel, { 
 ongc: () => {
 writeFileSync(1, "gc netClientSocketChannel");
 }
});

onGC(netServerSocketChannel, {
 ongc: () => {
 writeFileSync(1, "gc netServerSocketChannel");
 }
});

const isNetSocket = (socket) => socket instanceof net.Socket;

netServerSocketChannel.subscribe(common.mustCall(({ socket }) => {
 assert.strictEqual(isNetSocket(socket), true);
}));

netServerSocketChannel.subscribe(common.mustCall(({ socket }) => {
 assert.strictEqual(isNetSocket(socket), true);
}));
gc();
setTimeout(() => {
 console.log('timeout emit');
 const server = net.createServer(common.mustCall((socket) => {
 socket.destroy();
 server.close();
 }));
 
 server.listen(() => {
 const { port } = server.address();
 net.connect(port);
 });
}, 10000);

it is related to gc indeed. I will send a PR to fix this.Thanks !