Specify Out-of-Memory in Autograder Debug Output

[cohnt]

Right now, if the docker instance runs out of memory, it just kills the program that was taking up that memory. The only output is that the program was killed (with no other information) -- I was only able to diagnose that this was the problem when I increased the memory limit in the Gradescope interface (and it ran successfully). It would be helpful if the system could print some sort of warning or error message when this occurs.

[adam-brucker]

Gradescope has a roadmap with ideas located here: https://www.gradescope.com/roadmap. I would post this there as users can vote on ideas.

[cohnt]

Just posted there.

[mattrussell2]

If you're working in Python and use the subprocess module you can use resource.setrlimit(resource.RLIMIT_DATA, (kill_limit, kill_limit)) as a preexec_fn, with kill_limit as the max usable virtual memory (in bytes). This will gracefully kill the running process if it exceeds the memory limits, and not crash the autograder completely.