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

Issue 731775 link

Starred by 1 user
Status: Fixed
Owner:
█ roc...@chromium.org
please use my google.com address
Closed: Jun 2017
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 1
Type: Bug



Sign in to add a comment

DataPipeTest.SendConsumerAndCloseProducer is flaky

Project Member Reported by roc...@chromium.org, Jun 9 2017 
It occasionally hangs due to the consumer process not detecting data on the data pipe.

This flake is extremely rare and thus goes unnoticed on the CQ, but I can repro locally. It seems to go back several months so isn't caused by recent changes.

I'm running into it with slightly higher frequency after some WIP local chnages, so this needs to be addressed now.
Project Member

Comment 1 by bugdroid1@chromium.org, Jun 9 2017 

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8dc1fcc39716152ef830099390396c0ed3d3e35c

commit 8dc1fcc39716152ef830099390396c0ed3d3e35c
Author: Ken Rockot <rockot@chromium.org>
Date: Fri Jun 09 21:56:08 2017

Fix NodeController race in outgoing message queue

It's possible, albeit extremely unlikely, for a node to be introduced to
a peer in the small window between when it first tries to send an event
message and when it decides to stick the message in the queue for that
peer. If this happens, the message will never get flushed to the peer.

BUG= 731775 

Change-Id: If185139b900e90819ff989ac3622271dc1837eb5
Reviewed-on: https://chromium-review.googlesource.com/530045
Reviewed-by: Jay Civelli <jcivelli@chromium.org>
Commit-Queue: Ken Rockot <rockot@chromium.org>
Cr-Commit-Position: refs/heads/master@{#478429}
[modify] https://crrev.com/8dc1fcc39716152ef830099390396c0ed3d3e35c/mojo/edk/system/node_controller.cc

Comment 2 by roc...@chromium.org, Jun 15 2017

Status: Fixed (was: Assigned)
►

Sign in to add a comment