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

Issue 705120 link

Starred by 2 users

Issue metadata

Status: Duplicate
Merged: issue 709117
Owner:
Closed: Apr 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Windows , Mac
Pri: 2
Type: Bug-Regression



Sign in to add a comment

domAutomationController event being killed by ServiceManager

Project Member Reported by wylieb@google.com, Mar 24 2017

Issue description

Chrome Version       : 57.0.2987.98 (64-bit)
URLs (if applicable) : chrome://chrome-signin/?access_point=0&reason=0
Other browsers tested:
  Add OK or FAIL, along with the version, after other browsers where you
have tested this issue:
     Safari: n/a
    Firefox: n/a
         IE: n/a

What steps will reproduce the problem?
(1) Automated login via 
https://cs.chromium.org/chromium/src/chrome/browser/ui/webui/signin/login_ui_test_utils.cc?l=174
(2) Email field is filled in.
(3) Password is filled in and sign in is clicked.

What is the expected result?
The user is signed in.

What happens instead?
The test hangs with the log:
[73914:775:0324/150629.083324:WARNING:CONSOLE(237)] "<webview>: The load has aborted with error -20: ERR_BLOCKED_BY_CLIENT.", source: extensions::webViewEvents (237)

Please provide any additional information below. Attach a screenshot if
possible.

 

Comment 1 by wylieb@chromium.org, Mar 24 2017

Components: Internals>Mojo
Labels: -Type-Bug -Pri-3 OS-Linux OS-Mac OS-Windows Pri-2 Type-Bug-Regression

Comment 2 by wylieb@chromium.org, Mar 24 2017

Cc: s...@chromium.org

Comment 3 by s...@chromium.org, Mar 24 2017

Summary: domAutomationController event being killed by ServiceManager (was: InterfaceProviderSpec causing automated login to fail)
Hey Mojo folks!

We saw a regression in our sync specific tests that are not part of the main try bots, can view our waterfall at https://uberchromegw.corp.google.com/i/internal.client.kitchensync/waterfall . We're trying to figure out what went wrong. Tests started failing on Mar 23rd with the following log error:

[16935:16975:0323/201041.803973:ERROR:service_manager.cc(480)] InterfaceProviderSpec prevented connection from: content_utility to: content_browser

This happens as we try to run some javascript, https://cs.chromium.org/chromium/src/chrome/browser/ui/webui/signin/login_ui_test_utils.cc?dr=CSs&l=267 , to close the sign in confirmation dialog, and then signal back using the domAutomationController. But no message ever comes back, our tests hang/fail, and our suspicion is that this prevention the ServiceManager is doing is killing over event.

The last successful Linux build started Mar 22nd 19:32:24 GMT, and the first failure started at Mar 23rd 19:32:24 GMT.

Now, if this is the wrong component, my apologies, hopefully this is at least close. I've glanced over 
 https://cs.chromium.org/chromium/src/services/service_manager/README.md but still lacking understanding of what's going on here. All help is appreciated!
Labels: TE-NeedsTriageHelp
Added TE-NeedsTriageHelp as it looks out of scope from TE end.

Comment 5 by yzshen@chromium.org, Mar 27 2017

Cc: roc...@chromium.org
Hi, Ken.

Are you aware of any utility-related changes around that time? Thanks!

Comment 6 by s...@chromium.org, Apr 3 2017

Friendly ping?
Cc: wylieb@chromium.org
Cc: -s...@chromium.org ben@chromium.org
Owner: s...@chromium.org
Status: Assigned (was: Unconfirmed)
Sorry, just got to look at this now:

1 - the error log is an innocuous red herring (and known issue) caused by memory profiling client code running in utility process setup, trying to reach an interface which the browser does not expose to utility processes. this has nothing to do with domAutomationController or your tests.

2 - AFAICT the service manager does not kill or in any way control the lifetime of anything relevant to domAutomationController, so this bug description is probably inaccurate

It would help to do a bisect and figure out what CL actually broke this. The timing does seem to line up with r458954, which could mean that some related interface connection has been broken or has somehow become racy. Not sure who should take a look next (I'm OOO) but giving to skym@ for further triage/investigation. 

Comment 9 by s...@chromium.org, Apr 7 2017

Mergedinto: 709117
Status: Duplicate (was: Assigned)
Thanks for the reply rockot@, really appreciate it, sorry to bug you while OOO. I think my summary in #3 was actually quite inaccurate, after more digging it doesn't seem like we're actually reaching the domAutomationController at all.

I should have updated this bug, but I created new  issue 709117  instead with a more clear description of the current symptoms.

Currently trying to run the bisect tool but I seem to be doing something wrong.

Sign in to add a comment