Attempting to SSH into the robot breaks code deployment

[nathanwalters]

We've been running into a nasty memory leak on our robot, so I tried to SSH into our robot to try to set the command line arguments of the user program to do a heap dump when we run out of memory (we use Java). SSH connected, and the rio prompted my for a password. I first tried the default blank password; that didn't work. I then set a password for the admin user and then tried to login with that password; that also failed. Thinking I might have typed the password wrong, I typed it again (very carefully!). I was now greeted with a nice "permission denied". Well, there's no obvious way to SSH into the robot. Oh well.

Fast-forward a few minutes to when I try to redeploy new code to the rio. The Eclipse console shows the error

SSH_MSG_DISCONNECT: 2 Too many authentication failures for admin

I tried again. Same thing. I rebooted the rio. Same thing. We are now locked out of deploying code to our rio until we reflash it.

So, question time:

1. Is it possible to SSH into the rio? Or does NI/FIRST not like us mucking around with its internals?

2. Is there a way to configure the rio to do a heap dump when we run out of memory?

3. Is there any way to regain our ability to deploy without reflashing the rio?

4. Who at NI possibly thought this was a good idea?

Thanks for the help.