kore-rpc ending normally says user-interupted

[ehildenb]

Because we're sending SIGINT instead of SIGTERM to the kore-rpc, we're getting a weird message even on normal exit:

Could we instead, either:

• Go back to SIGTERM, or
• Add an explicit shutdown message to the RPC server that we call?

Related: #298
Related: runtimeverification/haskell-backend#3557

[tothtamas28]

The proper fix is to undo #298 , but for that the server needs to handle the SIGTERM signal gracefully.

[goodlyrottenapple]

This is a known issue in the Haskell runtime system which seems unadressed to this day. There appear to be a couple of workarounds like this one: IntersectMBO/cardano-node#3641

However, can't tell whether this would actually prevent the error above from being thrown.

[jberthold]

The proper fix is to undo #298 , but for that the server needs to handle the SIGTERM signal gracefully.

I think we should keep SIGINT, as only this will clean up all resources in the correct way. Translating SIGTERM to SIGINT internally as in the PR which @goodlyrottenapple found may be an option if for some reason we should not keep SIGINT, but I would not see why.

we're getting a weird message even on normal exit:

If the concern is just the appearance of an error in the logs, we can probably catch UserInterrupt and print something more comforting instead.

[tothtamas28]

we can probably catch UserInterrupt and print something more comforting instead

If there's no simple way to process both SIGINT and SIGTERM on the backend, I'm fine with this workaround.

[ehildenb]

I believe this is addressed at this point? I haven't seen the error message in logs for a while, and I am indeed comforted by that.