Re: [Madwifi-devel] regression tests
Status: Beta
Brought to you by:
otaku
From: Brian E. <eat...@gm...> - 2006-06-14 14:39:07
|
On 6/13/06, Michael Renzmann <ma...@no...> wrote: > I think I can manage to provide internet access to the testbed as well > as locations for implementing small indoor and outdoor setup(s), once we > have a better idea of what will be necessary. I think we could put together a decent test plan for station and AP mode with the following hardware: 1 laptop 2 wireless cards 1 wired NIC for remote administration That depends on the cards being physically positioned so that they can communicate properly. I use a laptop with one integrated wireless card and one card plugged into a cardbus slot, and those are able to see each others packets without problems. Having the wireless cards in the same machine reduces the hardware requirements and simplifies the test tooling. If we use separate machines then we will need extra code to synchronize the systems. If this test system is on the internet... well, that's a little worrisome. Shared test systems tend to be poorly maintained, and poorly maintained internet exposed systems tend to get broken into. Recovery from broken drivers could be a problem. We should expect regression testing to find bugs. If we are lucky, the bugs will cause a kernel panic, a reboot, and the system will recover. If we are unlucky, the reboot won't happen. I'm not sure how best to deal with this problem. Maybe a serial modem connection? Maybe the laptop sits in somebody's office, and we drop them a line saying "could you please reboot the victim machine, it's hosed again"? - Brian |