Background tasks in Cram tests are not terminated

[Willenbrink]

Expected Behavior

When a cram test terminates, all running background tasks (e.g. started via &) should be terminated too.

A reasonable alternative would be to wait for all tasks to terminate. As some background tasks never terminate and that behavior can be replicated by calling wait I think the first option is preferable.

Actual Behavior

Background tasks continue to run and never terminate.

Reproduction

1. Create a cram test that start some server in the background, e.g.

  $ ncat -l 2000 -k -c 'xargs -n1 echo' &

1. dune runtest
2. The server is still available and responds to e.g. ncat localhost 2000

Specifications

• Version of dune (output of dune --version): 3.18.2
• Version of ocaml (output of ocamlc --version): 5.3.0
• Operating system (distribution and version): Fedora

Workaround

Adding pkill -P $$ as the last line to a cram test kills all child tasks.

[Alizter]

This is definitely a missing feature and should be implemented. We even have our own workaround in our tests to get around this limitation. I'll need to do some research to find the best way to terminate all the children of the shell.

[Alizter]

@Willenbrink I have implemented a feature that cleans up child PIDs when the cram test exits. Could you have a look and see if it fixes your issue?

• feat(cram): cleanup subprocesses on exit #11827

[rgrinberg]

I don't think this issue is cram specific. One should be able to reproduce the same with any action. For example, with (action (bash "sleep 100 &"))