[agl-discussions] AGL image stability, stress and functionality tests proposal

[Manuel Bachmann]

Hi Jeremiah,

I totally agree with you ; it is always more reliable, for requirements
such as this, to rely on well-known upstream software which already has
contributors and users (even if we'll need to modify it to match 100% of
the specification).

I am happy to let you know that I got some results ! I did the following :
- push an upstream commit to the "wayland-fits" projects, so that it can
build with Yocto  [1] ;
- created the "wayland-fits" Yocto recipes [2] ;
 - integrated it in our local copy of the "meta-agl" and "meta-agl-demo"
repositories [3] [4] because we don't have rights necessary to send reviews
yet ; ;
- build a "agl-demo-platform-testsuite" image, and run it on a Koelsch
board.

Here is a screenshot of the test suite running on AGL demo platform/Koelsch
board :
http://www.tarnyko.net/repo/agl-demo-platform-testsuite.png (screenshot)

It is creating the white windows which you can see on the screenshot, and
moving the pointer on its own.  Short story, it works :).

[1] :
https://github.com/01org/wayland-fits/commit/af4891df0c270268b2188e666fc8b6177f96897c
[2] :
https://github.com/iotbzh/meta-agl/blob/testsuite/meta-ivi-common/recipes-graphics/wayland/wayland-fits_git.bb
[3] : https://github.com/iotbzh/meta-agl/tree/testsuite
[4] : https://github.com/iotbzh/meta-agl-demo/tree/testsuite

2015-09-19 11:05 GMT+02:00 Jeremiah Foster <jeremiah.foster at pelagicore.com>:

>
>
> On Fri, Sep 18, 2015 at 5:08 PM, Manuel Bachmann [ IoT.bzh ] <
> manuel.bachmann at iot.bzh> wrote:
>
>> Hi Jens,
>>
>> I was a member of the Tizen Common maintainers team.
>> I read your proposal with attention : especially the "Image Features"
>> one, which mentions test suites for Wayland and Weston.
>>
>> The idea to have functional tests is great. You mention manual tests, and
>> then ask about automated tests ; under Tizen, we had the "wayland-fits" (1)
>> suite for this purpose.
>> It basically consists in a virtual keyboard/mouse/touchscreen driver that
>> achieves most of the checks you are listing (such as "Window will be
>> moved", "Window will be resized"....). And with regards to Rudolf's remark,
>> it can work on QEMU, too.
>>
>> I just verified, and it does not exist in Yocto yet. So I guess I will
>> have to bitbake-ize it if we want to give it a try in AGL. I am OK to do
>> the work necessary to achieve that.
>>
>> What is your opinion on this ?
>>
>
> I tell you my opinion. :-)
>
> I think this is a great idea. Why re-write a functional test suite? What
> do you gain? Perhaps you can modify the tests somewhat but what you want is
> a certain "conformance" to an API or set of functionality which this suite
> appears to give you. If it comes into a project anywhere as a Yocto layer
> or recipe I'll certainly consider its use in a GENIVI test suite. That is
> if the GENIVI test suite doesn't get merged into the burgeoning AGL test
> suite. Have to see how specific to AGL the suite will become and what
> components AGL takes. Currently AGL is not yet building much of GENIVI's
> already existing layers and recipes which is leading to fragmentation.
> Fragmentation is a really bad idea and becomes expensive really quickly.
>
> Cheers,
>
> Jeremiah
>
>
>



-- 
Regards,

*Manuel Bachmann, Graphics Engineer www.iot.bzh <http://iot.bzh> *
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfoundation.org/pipermail/automotive-discussions/attachments/20150921/1b0ba299/attachment.html>