TCPSocketTest.SPWNoAdvance failing on builder chromeos-amd64-generic-rel-vm-tests |
|||||
Issue description
Test output:
[ RUN ] TCPSocketTest.SPWNoAdvance
../../net/socket/tcp_socket_unittest.cc:209: Failure
Expected equality of these values:
expect_rtt_notification_count
Which is: 3
watcher_ptr->rtt_notification_count()
Which is: 0
Stack trace:
#0 0x565eae796dfc testing::internal::UnitTestImpl::CurrentOsStackTraceExceptTop()
#1 0x565eae796746 testing::internal::AssertHelper::operator=()
#2 0x565eadca440c net::(anonymous namespace)::TCPSocketTest::TestSPWNotifications()
[ FAILED ] TCPSocketTest.SPWNoAdvance (49 ms)
,
Aug 14
Note: https://chromium.googlesource.com/chromiumos/docs/+/master/cros_vm.md#run-a-chrome-gtest-binary-in-the-vm The symptoms are consistent with tcp_info failing for some reason on this VM.
,
Aug 16
achuith: I'm not having much luck reproducing the issue locally either. Is there a way to schedule a build on this builder so that I can run an instrumented binary?
,
Aug 16
Ben may have some ideas on how to do this..
,
Aug 20
You can't schedule jobs on the fyi bot, but you can on the trybot. If you apply this patch (https://chromium-review.googlesource.com/c/chromium/src/+/1181653) and select the "chromeos-amd64-generic-rel" trybot in gerrit, it'll run net_unittests w/ the same config.
,
Aug 20
#5: Sweet! Thanks
,
Aug 21
,
Sep 19
Tentatively lowering the priority to P2 since no action has occurred for a while on this bug. Feel free to raise it back up again if you disagree.
,
Oct 25
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/3773b61598fc9e6e3a947ae74cb0cd2f45a6dbfc commit 3773b61598fc9e6e3a947ae74cb0cd2f45a6dbfc Author: Asanka Herath <asanka@chromium.org> Date: Thu Oct 25 19:46:22 2018 [net] TCP_INFO.tcpi_rtt could legitimately be zero on some kernels. Previously the return value of GetTransportRtt() didn't make it possible to distinguish between an RTT of zero from the case where an RTT could not be determined. However it is possible for this field in tcp_info to be legitimately zero due to the value being rounded down. The Linux kernel attempts to account for that by setting a floor of 1 for the RTT estimation, but that adjustment itself can get rounded off when responding to a getsockopt(TCP_INFO) call. This patch attempts to correct for this by setting a floor of 1µs for a valid RTT value. R=mattm@chromium.org Bug: 874197 , 875104 Change-Id: I3673e13301b62347b5bc2935cda4a7313447ce9c Reviewed-on: https://chromium-review.googlesource.com/c/1273808 Commit-Queue: Asanka Herath <asanka@chromium.org> Reviewed-by: Matt Menke <mmenke@chromium.org> Cr-Commit-Position: refs/heads/master@{#602833} [modify] https://crrev.com/3773b61598fc9e6e3a947ae74cb0cd2f45a6dbfc/net/nqe/network_quality_estimator_unittest.cc [modify] https://crrev.com/3773b61598fc9e6e3a947ae74cb0cd2f45a6dbfc/net/socket/tcp_socket_posix.cc [modify] https://crrev.com/3773b61598fc9e6e3a947ae74cb0cd2f45a6dbfc/net/socket/tcp_socket_unittest.cc
,
Oct 25
|
|||||
►
Sign in to add a comment |
|||||
Comment 1 by asanka@chromium.org
, Aug 14