Autopilot planning for trusty

Registered by Nicholas Skaggs on 2013-11-06

Let's talk about autopilot 1.5, and plans for autopilot during trusty.

AGENDA:
overview of tool
show off new 1.4 features
spreading autopilot love to others (how can we increase useage?)
autopilot feature requests
q and a

Blueprint information

Status:
Not started
Approver:
Jono Bacon
Priority:
Undefined
Drafter:
Nicholas Skaggs
Direction:
Approved
Assignee:
Nicholas Skaggs
Definition:
New
Series goal:
Accepted for trusty
Implementation:
Unknown
Milestone target:
milestone icon ubuntu-14.04

Related branches

Sprints

Whiteboard

Yay! Autopilot.

    New features in 1.4.

    Performance improvements specially selecting objects.

    Python 3 compliance.

    Most of the tests should just work, but we need to try them.

    Give advice or help porting.

    We would like all the tests to work equally well on python2 and python3.

    Work in the trusty cycle.

    Make a list of the most common failure causes.

    Reliable testing for autopilot

    Work to reduce spurious test failures

    Collaboration with CI tracking failures necessary: we need someone to raise a bug every time a test is rerun and then it passes, we need a way to track these bugs

What do you want from autopilot in T?

    vis tool love

    search

    notification of changes

    virtual framebuffer

    Investigate why apps don't launch with the inherited DISPLAY env var. Corey may know more...

    <dobey> if "xvfb-run -a autopilot" doesn't work, then it's probably a bug in autopilot not passing the environment to the child app, correctly

    <dobey> so should be pretty easy to fix

    pitti: you want to use autopilot-sandbox-run for that, this gets all the details right; you can even run it with --xephyr to see what it's doing (xvfb by default) example: http://bazaar.launchpad.net/~ubuntu-desktop/shotwell/ubuntu/view/head:/debian/tests/import

    Orientation support.

    -autopilot packages cause clutter

    can we drop having to package tests?

    CI team is considering using deb8 to test everything

== Notes ==

Can we use the CI infrastrucyure to run AP test suites under python3, and record success rate?

How do we organise the python 3 porting effort?

    let thomi do everything? :-P

    split up work to individual developers

    crowdsource it amongst small group

(?)

Work Items

Work items:
[fginther] Get a CI run with all the autopilot test suites with python3 to be able to compare results with the current way of running and be able to compare results: TODO
[fginther] Need the ability to select autopilot test runner based on the -autopilot package's Depends: (python3 or not): DONE
[xnox] Modify phablet-test-run to support python3-autopilot (sans pending landing): DONE
full widget tree property search in vis (pitti interested, but low-prio): TODO
[pitti] Investigate what needs to happen before we can simulate rotation changes (simulated sensor backend landed in platform-api): DONE

Dependency tree

* Blueprints in grey have been implemented.

This blueprint contains Public information 
Everyone can see this information.