[Linaro-validation] HDMI capture tech & Mozilla's project Eideticker
michael.hudson at linaro.org
Thu May 24 02:52:52 UTC 2012
Loïc Minier <loic.minier at linaro.org> writes:
> On Wed, May 23, 2012, Michael Hudson-Doyle wrote:
> Yup; looks like it
OK. So I had poke into the source code to see how it works.
A 'test' in this world is a web page that you load. JS in this page
does whatever it is you want to test, and then calls a "finish" function
from eideticker.js. For copyright reasons (I think all tests are things
like "a bunch of HTML we grabbed from cnn.com") there are no tests
distributed with eideticker and can't find any on the web, so I haven't
The way the tests work is that when you run get-metric-for-build, a web
server is started that can serve up the assets that make up the test and
also has an api to start and stop the capturing. The browser on the DUT
is instructed to load a HTML page which contains js that hits the "start
capturing" API, waits 2 seconds and loads the test page. This
(presumably) does its thing and then calls Eideticker.finish() -- which
loads another HTML page, the JS of which waits 1 second and then hits
the api to terminate the capturing. The host side code meanwhile is
polling its capturing code and, when it sees that capturing is done it
tears everything down and does some analysis on the video it got.
In sum, while I'm sure it works and provides useful results, it doesn't
seem like anything terribly sophisticated.
To put things as they might be in our world, it's like inserting
ssh power-measurement-server start-capture
ssh power-measurement-server stop-capture
after the shell commands we run to execute a test. Maybe that's enough!
>> (the reason I'm not sure if this is what Loic had in mind is that it's
>> more a Fennec/Firefox thing than a B2G thing).
> The B2G folks actually put us in touch with the QA team lead back then;
> I'm short finding the QA notes, generally the notes from chats with
> Mozilla were at:
> and I think that last call is where we had discussed QA / validation,
> but it's not too specific in the notes, mostly Linaro's side of the
Yeah, there doesn't seem to be anything directly relevant in the notes.
More information about the linaro-validation