New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 761416 link

Starred by 2 users

Issue metadata

Status: WontFix
Owner:
Left Chrome team
Closed: Jan 2018
Cc:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 2
Type: Bug



Sign in to add a comment

Chrome on android consistently killed when backgrounded

Project Member Reported by boliu@chromium.org, Sep 1 2017

Issue description

Context is I was trying to verify my fix in crbug.com/756497, but can't because chrome seem to be killed consistently on being backgrounded:

repro on galaxy j1 device
* open chrome with one tab displaying https://webglsamples.org/aquarium/aquarium.html make sure can at least see some things swirling around in the page
* press home
* open chrome again and see browser is restarting

Browser seem to survive 61.0.3162.3, but consistently getting killed on near trunk 62.0.3200.4. Pretty large range
 

Comment 1 by boliu@chromium.org, Sep 1 2017

I patch this in if I need to look at what the state of bindings are:
https://chromium-review.googlesource.com/647772
Status: Assigned (was: Untriaged)
I verified this regressed between 62.0.3187.3 and 62.0.3188.0. This range contains Bo's changes to renderer importance as the app goes from foreground to background.

I suspect the order of killing processes has changed between these versions causing the system to kill the browser process before the renderer.
I verified that with Bo's change renderer and browser have the same oom_score_adj (which is good). So my current hypothesis for this behaviour is that in the past we would momentarily drop to WAIVE binding, get the renderer killed, and that enabled browser process to stay alive in background. Now when we drop to background, it seems the system may be choosing to kill the browser first.

On Samsung, it's also likely they have their own logic, I am seeing
I/IntelligentSystemControl(  607): Killing 19310:com.chrome.beta/u0a113 (adj 2): kill heavy process(1)

The IntelligentSystemControl is not something I was able to see in our Android tree.
Labels: -Pri-3 LowMemory Pri-2
Status: WontFix (was: Assigned)
I've filed a separate bug to fix the issue where browser is being killed before renderer in the background. Going to close this.

Sign in to add a comment