test: mark test-net-write-fully-async-buffer as flaky

[aduh95]

Refs: #47428

[lpinca]

Did we try to investigate? It is not flaky on my machines.

[lpinca]

On macOS, I actually get a 1 ‰ failures when running:

python3 tools/test.py -J --repeat=1000 test/parallel/test-net-write-fully-async-buffer.js

16 ‰ when running

python3 tools/test.py -J --repeat=1000 test/parallel/test-net-write-fully-async-hex-string.js

[lpinca]

Something is fishy here. I get no failures if I remove the common module.

[aduh95]

Something is fishy here. I get no failures if I remove the common module.

#52964 (comment) also reported that removing ../common import would remove the flakiness of a different test (i.e. the test no longer times out)

[lpinca]

Bisecting point to 1d29d81 as the first bad commit.

[lpinca]

I think there is a recent bug in V8 or Node.js behind the flakiness of some tests. This is one of those. I think we should not land this.

[richardlau]

Something is fishy here. I get no failures if I remove the common module.

#52964 (comment) also reported that removing ../common import would remove the flakiness of a different test (i.e. the test no longer times out)

Maybe something to do with the exit hooks common installs?

[lpinca]

@richardlau I tried commenting almost everything in common and failures still occur. Note that, as per git bisect no failures occur from 07f481c which is the parent commit of 1d29d81.

[richardlau]

I meant that it's possible that the V8 update has caused something to have changed during shutdown. common registers Javascript code to be run prior to Node.js exiting. If you've commented out the registering of those hooks then it's probably something else.

[lpinca]

If you are referring to process.on('exit', handler) then yes I've commented them.

[lpinca]

The behavior described in #52964 (comment) also occurs for this test. It correctly finishes, the server emits the 'close' event, but the process does not exit.

[lpinca]

Adding the --jitless flag fixes the issue.