I was looking at extensions today and noticed that private bytes was significantly larger than WS private for both of them.
Beyond Corp: 17 MB versus 34 MB
Password Alsert: 13.5 versus 28.5 MB
Is there a separate bug for investigating and reducing this discrepancy to avoid memory-shock?
The test failure is caused by my patch, but the root cause is unrelated. See https://bugs.chromium.org/p/chromium/issues/detail?id=771805 for more details.
A slight change I want to make, regardless:
1) Only queue memory-infra dumps if there isn't an outstanding request.
"""
../../chrome/browser/task_manager/task_manager_browsertest_util.cc:136: Failure
Failed
Timed out.
Waiting for exactly 1 matches of wildcard pattern "Subframe: http://d.com/" && [Memory Footprint >= 1000]
Currently there are 0 matches.
Current Task Manager Model is:
> Browser [Memory Footprint == 208584704]
> GPU Process [Memory Footprint == 52879360]
> Tab: Cross-site iframe factory [Memory Footprint == 87560192]
> Tab: Cross-site iframe factory [Memory Footprint == 87560192]
> Subframe: http://b.com/ [Memory Footprint == 82747392]
> Subframe: http://b.com/ [Memory Footprint == 82747392]
> Subframe: http://c.com/ [Memory Footprint == -1]
> Subframe: http://d.com/ [Memory Footprint == -1]
> Subframe: http://e.com/ [Memory Footprint == -1]
"""
The error *looks* comparable. But I cannot reproduce it locally anymore. Investigating further to see what's happening. It's possible we're seeing a timeout caused by other (?) reasons. The test does take 12s to run on my local, fast machine.
Comment 1 by erikc...@chromium.org
, Aug 30 2017