New issue
Advanced search Search tips

Issue 887545 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: Sep 24
Cc:
EstimatedDays: ----
NextAction: 2018-09-24
OS: ----
Pri: 3
Type: Bug

Blocked on:
issue 887543



Sign in to add a comment

SessionRestoreTest.MemoryPressureLoadsNotAllTabs is flaky on Linux Chromium OS ASan LSan

Project Member Reported by lukasza@chromium.org, Sep 20

Issue description

It seems that the test hits a segmentation fault, but
1) I can't repro locally
2) The callstack from the bots is unreadable (see  issue 887543 )
3) I don't want to revert r590331 which I think is desirable (for insulating browser_tests from their environment)

Therefore, I am going to just disable the flaky test on affected configurations.
Cc: chrisha@chromium.org sky@chromium.org ericrk@chromium.org
<adding people from issue 860090 where the same test was marked as flaky [but is fixed/passing now / before r590331?]>
Blockedon: 887543
Owner: lukasza@chromium.org
Status: Started (was: Untriaged)
Hmmm... actually I speculate that something in the system tries to dereference base::MemoryPressureMonitor::Get() [which will be null after r590331].  The fix for this is to use base::test::FakeMemoryPressureMonitor from the test (instead of having the test directly dispatch a memory pressure notification).  Maybe we can try this speculative fix first, before disabling the test.

WIP CL @ https://crrev.com/c/1236502
Project Member

Comment 6 by bugdroid1@chromium.org, Sep 20

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/481767eb9c2c2034f8cd53f9a699a0f6545c22f0

commit 481767eb9c2c2034f8cd53f9a699a0f6545c22f0
Author: Lukasz Anforowicz <lukasza@chromium.org>
Date: Thu Sep 20 19:51:45 2018

Speculative fix for flakiness of MemoryPressureLoadsNotAllTabs.

The test flakily crashes with an unknown callstack (the crash is not
reproducible locally).  The speculation is that the crash is related
to dereferencing a null base::MemoryPressureMonitor::Get().  If this is
indeed the case, then the crash should be fixed by using
base::test::FakeMemoryPressureMonitor in the test (rather than directly
dispatching a memory pressure notification from the test).

Bug:  887545 
Change-Id: Ib922ebe4c7f85ec028f4bb202f2357c71e20687a
Reviewed-on: https://chromium-review.googlesource.com/1236502
Reviewed-by: Chris Hamilton <chrisha@chromium.org>
Commit-Queue: Ɓukasz Anforowicz <lukasza@chromium.org>
Cr-Commit-Position: refs/heads/master@{#592913}
[modify] https://crrev.com/481767eb9c2c2034f8cd53f9a699a0f6545c22f0/chrome/browser/sessions/session_restore_browsertest.cc

NextAction: 2018-09-24
Let's check next week if the flakiness is gone.
The NextAction date has arrived: 2018-09-24
Status: Fixed (was: Started)
The flakiness dashboard link from #c5 is all green - let's mark this as fixed.

Sign in to add a comment