attended tests are not fully automated

[Robert Collins]

There appear to be two sorts of attended tests:
 - observation required
 - interaction required

The latter should probably be easily solvable using the accessiblity layer as dogtail etc have done. However the observation tests are trickier. One possible way would be to capture them as screencasts; we could look at them rapidly after the test run, and we'd have a detailed capture to use to show other folk exactly what it looks like).

This is separate to bug 439769 which is simply about separating out things for the moment to allow automating what we can cheaply.

[Mirco Müller (macslow)]

Robert, I don't know much about dogtail. How would this best be hooked up in notify-osd makefiles?

[Ara Pulido (ara)]

Dogtail is equivalent to LDTP (the framework we use in Mago), so you could try to use the mago tests we already have.

[Ted Gould (ted)]

I think it would be easier to stick with Mago, but the problem here is not that as much as the separation between fully automated and inspectable tests. What they're automated with is not of issue.

[Robert Collins (lifeless)]

On Thu, 2009-10-01 at 15:19 +0000, Ted Gould wrote:
>
> I think it would be easier to stick with Mago, but the problem here is
> not that as much as the separation between fully automated and
> inspectable tests. What they're automated with is not of issue.

I filed a separate bug for that :) We can separate and then gradually
move bugs from one to another.

@Mirco, I'm not sure - I haven't used Mago, it is possibly Ara could do
one to show you, or I can get across it and convert one to show you.

Part of doing this will be determining what layer that test should be at
(and possibly splitting it: some unit, some functional).

-Rob

[Robert Collins (lifeless)]

Mirco, I see some work towards this. Sadly we actually need an environment variable I think, so that 'distcheck' which runs 'make check' will work.

[Mirco Müller (macslow)]

The only interactive test is the one with the fallback-dialog, which needs to be skipped for automated test-runs. I suggest adding a check for env-var "UNATTENDED" to indicate any interactive tests should be skipped. I can do that as part of the SRU work after UDS.

[Robert Collins (lifeless)]

I've added this.

[Karl Lattimer (karl-qdh)]

@Mirco, Robert; has this bug been fixed? Comment #5, #6 and #7 seem to indicate it *might* be.

[Mirco Müller (macslow)]

No, that's not yet implemented.