lp:~mwhudson/lava-dispatcher/bug-900990

Branch merges

Propose for merging

No branches dependent on this one.

Merged into lp:lava-dispatcher at revision 173

Paul Larson (community): Approve on 2011-12-07

Diff: 75 lines (+20/-18)

1 file modified

lava_dispatcher/client/master.py (+20/-18)

api

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Bug #900990: android deploy does not fail if untar fails

High

Fix Released

Link a bug report

Related blueprints

178. By Michael Hudson-Doyle on 2011-12-07

thanks paul

177. By Michael Hudson-Doyle on 2011-12-07

this is torture

176. By Michael Hudson-Doyle on 2011-12-07

grr tar does not autodecompress from stdin

175. By Michael Hudson-Doyle on 2011-12-07

stoopid

174. By Michael Hudson-Doyle on 2011-12-06

factor downloading a tarball to the board into a separate function, that fails if something goes wrong

173. By Michael Hudson-Doyle on 2011-12-06

make some error messages more explicit

172. By Michael Hudson-Doyle on 2011-12-06

use the context's config rather than reading it again in _install_lava_test

171. By Zygmunt Krynicki on 2011-12-01

Automatically override configuration inside virtualenv

170. By Michael Hudson-Doyle on 2011-11-30

silly mistake

169. By Michael Hudson-Doyle on 2011-11-30

Major refactoring of how the dispatcher deploys images and communicates with
the device being tested.

 * One major change is the commands are executed in 'sessions', for example:

       with client.tester_session() as session:
           session.run('ls')

   This will ensure the system is booted into the test partition and run()
   wraps up the "run a shell command, optionally looking for a particular
   response and wait for the next prompt" functionality we use in a number
   of places.

 * Another major change is to move the details of deployment to client methods.
   Nothing outside of the LavaMasterImageClient implementation of thse methods
   assumes the existence of a master image now.

 * Finally, add an experimental client subclass that builds an image and tests
   it for qemu.

The QEMU support is probably a bit fragile and specific to beagle currently,
but that should be easy to change. The support for testing on master image
based boards should work as before.