…graders on a single machine
…es app to run this hw3
…ace period in assignment part configuration
… that properly check this
This bug has another aspect that becomes apparent when we read the output of the spawned process so slow, that the process terminates long before we read its whole output. This alone is not bad, however the condition thread.alive? was used in the while loop. This led to the premature termination of run_with_timeout.
Now load_due_date and load_grace_period have the below priority: 1)Part specific due date and grace period in autograders config file 2)Assignment specific due date and grace period in autograders config 3)Default values in EdXClient If part_name is not supplied in load_due_date or load_grace_period, they revert back to their previous version
Ignore EOFError raised by stderrout.read_nonblock, since with sufficiently big stdout, small stderrout and small buffer_size, err will reach EOF before the stdout is completely read. This would cause the graders output to be truncated.
…ize, so am defaulting to 1024 and adding test
… to replicate on OSX
…nt - was screwing up the autograder sub process results parsing
… to support the testing I was doing
…- some oddness around should_receive - am stubbing some edx-client routines - want to move to should receive