On 27 July 2017 at 12:55, Benjamin Auclair
<benjamin.auclair@devialet.com> wrote: The problem is in the file you've attached called device-job.yaml
> Thanks for your answer
>
>> Minimal pastes only make sense in a review for a code change.
>> Debugging needs ALL the data.
>
> In fact, I did it because it's just some lava code, you can find the whole
> code in lava_dispatcher_app folder, in shell.py file at line 240.
>
>> Without seeing the test job submission and the device-type jinja2 it's
>> impossible to tell. Have you modified any code in lava-dispatcher?
>
> I didn't modified any code in lava-dispatcher and I attach you in this email
> my device-type config, test job and full test job submission.
(which is actually a test job submission - terminology can be
important when debugging automation).
There is no deploy action defined. The test shell overlay is prepared
in the deploy action so that it can be added to the ramdisk or NFS
which is to be made available over TFTP.
You need to have a deploy step - what you have currently is that your
device.jinja2 only runs bootp, so uses whatever is pre-installed on
the device. This is a mistake (see the problems of simplistic testing
in the docs https://staging.validation.linaro.org/static/docs/v2/ )simple-admin.html#problems- with-simplistic-testing
So you will need to provide a deploy step including a kernel and a
ramdisk and then change the device.jinja2 so that it does a normal
deploy action. Also, don't use the name "device" for things, give it a
sensible name that relates to the board itself, something based on the
manufacturer, model and/or DTB name.
Do not take shortcuts in adding a new device to LAVA - do a full
deploy step and then LAVA will be able to provide the scripts which
start the execution of the test.