This has been one of the differences between upstream merger and image testing. After seeing a disaster related to unity8 and the inability to revert back to a usable device after running an MP, I've implemented a solution to reflash the test device after each run. The tests then assume that the device is clean when the generic-mediumtests-runner-maguro or generic-mediumtests-runner-mako starts.
The two concerns I have with this are devices falling over due to the flash and creating a backlog due to the extra time needed. I'll be watching this closely today. The flash is done with http://10.97.0.26:8080/job/touch-flash-device/, which takes the slave $NODE_NAME from the runner job and dispatches the correct touch-flash-$slave job. The touch-flash-* jobs have a higher priority than the generic-mediumtests-runner-* jobs to work out the scheduling. The flash job is also triggered before the testing actually starts to make sure it's in the queue as soon as possible. Francis -- Francis Ginther Canonical - Ubuntu Engineering - Quality Engineer -- Mailing list: https://launchpad.net/~canonical-ci-engineering Post to : [email protected] Unsubscribe : https://launchpad.net/~canonical-ci-engineering More help : https://help.launchpad.net/ListHelp
