1164078 - In test_add_photo_to_contact.py, after tapping button that closes frame, you get MarionetteException: Error sending message to frame (NS_ERROR_NOT_INITIALIZED)

Status: RESOLVED INCOMPLETE

(Remote Protocol :: Marionette, defect)

Description

[Martijn Wargers (dead)]

In the test for test_add_photo_to_contact.py , I had to add a workaround to not get failures as mentioned in bug 1161088: https://github.com/mozilla-b2g/gaia/pull/29874

However, this should not be necessary. 

On irc:
23:04 jgriffin: mwargers: you can ask ato tomorrow; I think we used to catch the frame's unload event to handle this situation, but maybe that was broken recently

From bug 1161088, comment 5:
If a frame has been closed there is a requirement that your next step be to switch frame otherwise you will get errors. This was documented in  bug 1104624.

I tried that and that didn't work.

Andreas, can you take a look at this when you have the time?

Comment 1

[Martijn Wargers (dead)]

There is a similar issue in bug 1109213, see bug 1109213, comment 15.

Comment 2

[Andreas Tolfsen ❲:ato❳]

Can you provide a test case that I can use to reproduce this?

Comment 3

[Martijn Wargers (dead)]

And I suspect in bug 1170090, we have a similar issue going on.
I think I need to look all the code that closes a window and add workarounds for it.

Comment 4

[Martijn Wargers (dead)]

(In reply to Andreas Tolfsen (:ato) from comment #2)
> Can you provide a test case that I can use to reproduce this?

See comment 0. Basically when you back out the pull request from bug 1161088: https://github.com/mozilla-b2g/gaia/pull/29874 , then you'll see the failure in test_add_photo_to_contact.py.
Or are you looking for a Gaia UI unit test or something?

Comment 5

[Martijn Wargers (dead)]

In bug 1188924, we've now made a centralized workaround for this issue.

Comment 6

[Andreas Tolfsen ❲:ato❳]

Clearing needinfo request until I have a way to reproduce this.

Comment 7

[Martijn Wargers (dead)]

Attachment: 1164078.diff

Sorry for the delay, this is basically the issue that is going on.
When the Marionette is in an iframe and that iframe gets closed, then calling Marionette methods (like self.marionette.switch_to_frame()) afterwards are causing NoSuchWindowExceptions.
It seems to me that Marionette should automatically switch to the root frame or something, when this happens.

Comment 8

[Andreas Tolfsen ❲:ato❳]

(In reply to Martijn Wargers [:mwargers] (QA) from comment #7)
> Sorry for the delay, this is basically the issue that is going on.
> When the Marionette is in an iframe and that iframe gets closed, then
> calling Marionette methods (like self.marionette.switch_to_frame())
> afterwards are causing NoSuchWindowExceptions.

Thanks for investigating this.  Switching to frame should not throw this error if the current frame disappears.

> It seems to me that Marionette should automatically switch to the root frame
> or something, when this happens.

This is the expected behaviour.  If something closes the frame you would want to know that some action has removed the frame.

The correct action to take in this case is to catch that error, or by some other measure detect that the frame has disappeared, and switch to a valid frame.

Comment 9

[Martijn Wargers (dead)]

(In reply to Andreas Tolfsen (:ato) from comment #8)
> This is the expected behaviour.  If something closes the frame you would
> want to know that some action has removed the frame.
> 
> The correct action to take in this case is to catch that error, or by some
> other measure detect that the frame has disappeared, and switch to a valid
> frame.

How do I do switch to a valid frame in this case? When I call self.marionette.switch_to_frame(), I would normally have switched to the system frame, which is always there. But in this case, I'm getting this error.
Or do I need to re-instantiate a new Marionette client in this case?

Comment 10

[Andreas Tolfsen ❲:ato❳]

(In reply to Martijn Wargers [:mwargers] (QA) from comment #9)
> (In reply to Andreas Tolfsen (:ato) from comment #8)
> > The correct action to take in this case is to catch that error, or by some
> > other measure detect that the frame has disappeared, and switch to a valid
> > frame.
> 
> How do I do switch to a valid frame in this case? When I call
> self.marionette.switch_to_frame(), I would normally have switched to the
> system frame, which is always there. But in this case, I'm getting this
> error.
> Or do I need to re-instantiate a new Marionette client in this case?

No, that is definitely a bug.  Switching to a frame should not check that the current frame context is still present.  It, and Switch To Window, are special commands that should circumvent these checks.

Comment 11

[David Burns :automatedtester]

closing b2g related bugs