Open jzhou77 opened 3 years ago
For trivial problems: 1) no trace line produced for a while; 2) rerun passes, we can solve them.
TestHarness
can also look at /proc
file system.
This is part of the retry logic I was remembering. I haven't looked carefully to see if it does anything similar to what we want:
We observed that running in a shared environment (e.g., docker and AWS spot instances), the correctness runs of Joshua results are often
timeout
errors. These errors can become very noisy if the CPU resources on the test machine is heavily contended. So it would be very useful we can filter them out.One idea I have is for
TestHarness
to check the progress of simulation runs. Rerun the simulation fortimeout
errors.