800011 - [B2G] Need to be able to load entry points of a main app

Status: RESOLVED FIXED

(Remote Protocol :: Marionette, defect)

Description

[Zac C (:zac)]

The  'launch' method ( https://github.com/jonallengriffin/gaia/blob/bug796200/tests/marionette/gaiatest/gaia_test.py#L55 ) in the GaiaApps B2G Marionette testing framework does not allow you to load apps at their entry points  (see manifest: https://github.com/mozilla-b2g/gaia/blob/master/apps/communications/manifest.webapp )

jgriffin confirmed that the method needs to be updated to handle this.

Error handling of invalid app names should also be improved.

Comment 1

[Stephen Donner [:stephend] Not actively reading bugmail]

This is blocking Zac from writing about 4 Gaia smoketests right now, so appreciate traction on this -- thanks!

Comment 2

[Jonathan Griffin (:jgriffin)]

Marionette's Gaia app launching takes place here:  https://github.com/mozilla-b2g/gaia/blob/master/tests/marionette/gaiatest/gaia_apps.js#L22

The contacts app is a little different than most as it is an aspect of the communications app, rather than a separate distinct app.  The code I linked above needs to be aware of the kind of manifest used by it:  https://github.com/mozilla-b2g/gaia/blob/master/apps/communications/manifest.webapp

I.e., I imagine it needs to look for entry points in the app and not just the top-level 'name'.

Comment 3

[Jonathan Griffin (:jgriffin)]

See also this code which shows how gaia handles this internally:

https://github.com/mozilla-b2g/gaia/blob/master/apps/system/js/window_manager.js#L1017

Comment 4

[Jonathan Griffin (:jgriffin)]

I've just been told that one of the Gaia devs already has a fix for this, so we may not need to do anything here.

Comment 5

[David Burns :automatedtester]

:jgriffin is there a bug for reference? will hold off on this one then. Just for now.

Comment 6

[Jonathan Griffin (:jgriffin)]

(In reply to David Burns :automatedtester from comment #5)
> :jgriffin is there a bug for reference? will hold off on this one then. Just
> for now.

Not that I know of.  I think this will be handled via a pull request to Gaia; when it happens, I'll cc: it here.

Comment 7

[James Lal [:lightsofapollo]]

albertoq and I are working on this for JS tests. We should have a patch up later today, I am not sure what bug to file it under yet but I will reference it here.

Comment 8

[James Lal [:lightsofapollo]]

We have a PR up that unifies the atoms between the JS tests & python as well as fixing the dist bug. (So we should be able to publish the gaiatest package now).

Comment 9

[James Lal [:lightsofapollo]]

... and that PR also contains a fix (and test) that allows launching of apps that use the entry points.

Comment 10

[Zac C (:zac)]

How will this work where there could be two apps with the same entry point name?

Comment 11

[James Lal [:lightsofapollo]]

I believe (jgriffin correct me if I am wrong) that any apps named the same thing will cause some serious problems when trying to use the gaia_apps.js to launch them.

That is unlikely to be a problem for gaia specific tests but if we where to load the system with third-party apps, conflicts would be likely.

The only way to solve that would be to accept the origin+entrypoint over the name of the application.

Comment 12

[James Lal [:lightsofapollo]]

Zac C not sure how you consume this but the new version of gaiatest (0.2) just landed in gaia with these changes. Can you please try them out and let us know if/where you have problems?

Comment 13

[Zac C (:zac)]

Will give it a go, thanks James.