Run cleanup_task on SIGINT

[SomaticIT]

Describe The Bug

The cleanup_task does not run when flow is interrupted by Ctrl+C.

This is problematic when using cargo watch (or similar).

Exemple:
Imagine we are developing a web server with actix-web. We need a redis database to store some data.
We want to setup the environment before runing cargo watch and clean up the environment when we interupt the watch.

To Reproduce

1/ Setup Makefile.toml

[tasks.dev]
run_task = { name = ["setup ", "dev-watch"], fork = true, cleanup_task = "cleanup" }

[tasks.dev-watch]
command = "cargo"
args = ["watch", "-x", "run"]

[tasks.setup]
script = "echo setup env"

[tasks.cleanup]
script = "echo cleanup env"

2/ Run cargo make dev
3/ Ctrl+C

The task cleanup is never executed.

Error Stack

None.

[sagiegurari]

@SomaticIT sorry for late reply.
i totally get your use case, but i feel its not something standard for scripts to do. usually the signal would kill the process.
maybe you need to cleanup any previous leftovers before the execution?

[SomaticIT]

@sagiegurari, thank you for your reply.

I understand your mean.

However, it's not uncommon to see scripts that react to the SIGINT signal to clean their ressources.

• eg: telepresence clean remote ressources on k8s and close the connection
• eg2: actix-web clean all workers to avoid zombies

To clarify a good use case for this feature, it would be useful to start some containers before running the dev server and stopping/cleaning these containers when killing the server.

Maybe to avoid ambiguities, we could add an option (eg : cleanup_on_sigint: true or on_sigint: "task_name").
What do you think?

[sagiegurari]

there is some unstable feature for singal handling which kills child process and might be a good start for reusing for this issue.
would you like to investigate a solution with my help and publish a PR? or prefer for me to look at it?
heads up, it won't be trivial.