Signal handlers not registered for redis-cli

[rabbitfang]

When running redis-cli as PID 1 (e.g. within a container), the process can only be killed with SIGKILL. This is because PID 1 ignores all signals not explicitly handled by the process, and redis-cli does not generally register its own signal handler (except when using --intrinsic-latency). This prevents the user from aborting long-running commands such as --scan and --stat.

Steps to reproduce

1. Run docker run -d redis:4 redis-cli -h $HOST --stat where $HOST is an active redis instance
2. Run docker ps to verify redis-cli is running.
3. Run docker kill -s INT $ID (where $ID is the container ID of the redis-cli container) to send the SIGINT signal to the redis-cli process.
4. Run docker ps

Expected result

• The redis-cli container is no longer running.

Actual result

• The redis-cli container is still running. If you attach to the container, you will see that the container did not react to the signal at all (it is still scanning).

Versions

Tested and verified on redis-cli version 4.0.8 and 2.8.23.

Proposed solution

• Setup a signal handler like in server.c: https://github.com/antirez/redis/blob/ffde73c57d0aa5cd58869c13cb1b76d6c27ef5b7/src/server.c#L1832-L1834