Helper tools useful for working with the SSH Agent.
Will check for existing ssh-agent processes and reuse them if possible.
The reason behind this script is that a huge number of stale ssh-agent processes can accumulate if you're simply just running eval $(ssh-agent) every time you're opening a Git Bash terminal. This script can actually reuse a lot of these no-so-stale processes without having to start a new one.
All that's necessary is to source the script file ssh-agent-tools.sh -- for example from .bashrc. The rest is automatic.
First, the script will check the environment variable SSH_AGENT_PID. If the variable is set, the script will then check whether there is an actual ssh-agent process behind the process ID. If that's the case, then the environment variable SSH_AUTH_SOCK is checked next. If that one is fine as well, then the identities found in ~/.ssh are simply added. If either of these steps failed for any reason, both of the environment variables are unset. In that case you will see a message like:
Removing stale SSH_AGENT_PID
The next step is to check for socket files in /tmp. If any of those files points to a valid process, then the SSH_AGENT_PID and SSH_AUTH_SOCK environment variables are set according to the process ID. If, during the socket file discovery, any stale socket file were found, then they will be removed and you will see one or more messages like:
Removing stale socket /tmp/ssh-vpP0nkSM33dz/agent.118
Only if all that fails, the the eval $(ssh-agent) command is actually run and the identities are added.
If adding an identity requires a passphrase, then the way the identities are added at the moment might actually not work as expected (or at all).
Also, this script has so far only been tested and used on a Windows system in the scope of Git Bash. Running it on Linux may have unexpected consequences.