836313 - Switching from an app using a trusted UI to one that isn't while content is being loaded can result in that app getting killed in the background

Status: RESOLVED INVALID

(Firefox OS Graveyard :: General, defect)

Description

[Jason Smith [:jsmith]]

Break out in bug 834871 of the app process getting killed part.

When a user switches from an app that has a trusted UI open that is loading content, we occasionally see the app process get killed in the background. As a result of the app process getting killed, we end up hitting bug 834871 by caveat.

This happens pretty easily with marketplace app with persona signing. I'm noming this separately from bug 834871 as I'm pretty concerned we are able to kill the marketplace app process pretty easily in the background unexpectedly.

Comment 1

[Justin Lebar (not reading bugmail)]

Which app is getting killed, the one with the trusted UI, or the one you're switching to?

In general, what do you think is the right behavior here?  Should apps showing trusted UIs get higher priority when they're in the background, do you think?

Comment 2

[Jason Smith [:jsmith]]

(In reply to Justin Lebar [:jlebar] from comment #1)
> Which app is getting killed, the one with the trusted UI, or the one you're
> switching to?

The app with the trusted UI. Specifically in the background while a different app is open.

> 
> In general, what do you think is the right behavior here?  Should apps
> showing trusted UIs get higher priority when they're in the background, do
> you think?

Hmm...I was originally expecting that the app process wouldn't get killed in the first place. Let me get a logcat and dig into what might be going on here to get more information on why the process is getting killed in the first place.

Priority wise I'd expect us to follow the same priorities that we follow for any background app - whether there's a trusted UI there or not. Although giving higher priority to an app with a trusted UI open out of the list of background apps would carry an advantage of reducing the possibility of bug 834871 happening.

Comment 3

[Justin Lebar (not reading bugmail)]

My guess was that it was dying due to oom.  If it's not, that's another story.

You can find OOM crashes by looking through the output of dmesg.

Comment 4

[Jason Smith [:jsmith]]

Attachment: Logcat

Comment 5

[Jason Smith [:jsmith]]

Yeah, this looks like a general OOM case then (confirmed). I actually don't think we can need to do anything here, so I'll close this out then.