dgou at mac.com
Tue Apr 28 09:25:05 PDT 2009
On or about Tuesday, April 28, 2009, at 12:05PM, holger krekel indited:
>maybe i am missing something, but i am not talking about
>isolating each test in a fresh subprocess, rather a mode
>where you run multiple tests in one process and restart
>the process in case one test crashed it. This way
>you would look for device drivers N+1 times with N
>being the number of crashes of your tests (and/or the
>number of timeouts etc).
Ah, I misunderstood. Sounds like a nice feature.
Our experience has been that when a test "misbehaves" we're deeply sunk, the device gets into a state where we are down into the windows driver stack, and the process can't be killed, etc. Sometimes the device bricks. So far, knock wood, we haven't had any problems with our Python/C++/driver-stack code. Given the number of tests and the duration (anywhere from a few minutes to a few weeks), we've pretty well shaken out any memory management issues, which is always a bug-a-boo for C-based extensions. The degree to which Python manages to "just work" here has been awesome, and its one of the reasons we're able to use Python at all. Not sure how much we'd gain by having the subprocess capability...
More information about the testing-in-python