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

Issue 712849 link

Starred by 1 user

Issue metadata

Status: Untriaged
Owner: ----
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 2
Type: Bug



Sign in to add a comment

Add mobile system health benchmark to exercise CCTs

Project Member Reported by dskiba@chromium.org, Apr 18 2017

Issue description

We've recently discovered and fixed an issue where Chrome was leaking activities ( issue 710612 ).

While the leak was not specific to Chrome Custom Tabs, I found it while playing with CCTs (and it's way harder to repro it other way). Besides, CCTs are important part of Chrome, and are used by many 1st party apps.

So I'm proposing adding a system health benchmark that would exercise CCTs. We need to cover at least two cases:

1. Opening / closing CCT in the current task (the default behavior, seen e.g. in gmail app)

2. Opening / closing CCT in a separate task (the way AGSA does it)

This would not only cover the found leak, but would also help spotting other classes of leaks (e.g. in the native code).
 
Components: Speed>Benchmarks
I'm generally positive about the idea of covering CCT. As you say, they are a very commonly used feature and we have zero coverage today.
My two questions are:
1) Does CCT require an internal checkout, or does it work with ChromePublic?
2) How long (in terms of test runtime) do you think a user story would take to get representative data (i.e. catch the leak we had?)

Comment 3 by yus...@chromium.org, Apr 19 2017

1)It works with ChromePublic
2)Catching the leak would have involved creating two back to back activities in Chrome (easiest if these are CCT). So launching two intents and monitoring the result.

Sign in to add a comment