I've noticed that CI has started having intermittent failures with the h2spec.test.py test. I reproduced it locally, running the test 20 times, and noticing that it would fail consistently within a few runs. I did a git bisect and here's the results:
$ git bisect run /var/tmp/build_and_run_h2spec_test.sh
running /var/tmp/build_and_run_h2spec_test.sh
building: f80ed7354 (HEAD) Do not delete the continuation twice (#7862)
tests failed for: f80ed7354 (HEAD) Do not delete the continuation twice (#7862)
Bisecting: 5 revisions left to test after this (roughly 3 steps)
[a0c267cdc88586679ccba4ec3e394f386d52f806] Add URI Signing cdnistd Claim Implementation (#7822)
running /var/tmp/build_and_run_h2spec_test.sh
building: a0c267cdc (HEAD) Add URI Signing cdnistd Claim Implementation (#7822)
tests failed for: a0c267cdc (HEAD) Add URI Signing cdnistd Claim Implementation (#7822)
Bisecting: 2 revisions left to test after this (roughly 2 steps)
[19dfdd4753232d0b77ca555f7ef5f5ba3d2ccae1] Clean up lua plugin doc for overridable configurations (#7844)
running /var/tmp/build_and_run_h2spec_test.sh
building: 19dfdd475 (HEAD) Clean up lua plugin doc for overridable configurations (#7844)
tests failed for: 19dfdd475 (HEAD) Clean up lua plugin doc for overridable configurations (#7844)
Bisecting: 0 revisions left to test after this (roughly 1 step)
[23174f43ba543c9f56916e5d6aadb7a22f349355] Save and propagate epoll network error (#7809)
running /var/tmp/build_and_run_h2spec_test.sh
building: 23174f43b (HEAD) Save and propagate epoll network error (#7809)
tests failed for: 23174f43b (HEAD) Save and propagate epoll network error (#7809)
Bisecting: 0 revisions left to test after this (roughly 0 steps)
[fbdbb5b733c389e537d777938298219be43e619e] Add method to write an IpAddr value to a sockaddr. (#7821)
running /var/tmp/build_and_run_h2spec_test.sh
building: fbdbb5b73 (HEAD) Add method to write an IpAddr value to a sockaddr. (#7821)
tests passed for: fbdbb5b73 (HEAD) Add method to write an IpAddr value to a sockaddr. (#7821)
23174f43ba543c9f56916e5d6aadb7a22f349355 is the first bad commit
commit 23174f43ba543c9f56916e5d6aadb7a22f349355
Author: Susan Hinrichs <shinrich@verizonmedia.com>
Date: Mon May 17 19:23:48 2021 -0500
Save and propagate epoll network error (#7809)
iocore/net/NetEvent.h | 17 +++++++++++++++++
iocore/net/UnixNet.cc | 19 ++++++++-----------
iocore/net/UnixNetVConnection.cc | 12 ++++++++++++
3 files changed, 37 insertions(+), 11 deletions(-)
bisect run success
Those results seem to indicate that this PR has made the h2spec test to pass inconsistently:
#7809
For those trying to reproduce these results:
- Docker image:
controller.trafficserver.org/ats/centos:8
./configure --prefix /var/tmp/build --with-openssl=/opt/openssl-quic --enable-experimental-plugins --enable-example-plugins --enable-werror --enable-debug --enable-wccp./autest.sh --sandbox /tmp/sb_master --ats-bin /var/tmp/build/bin/ --clean=none -f h2spec
I've noticed that CI has started having intermittent failures with the h2spec.test.py test. I reproduced it locally, running the test 20 times, and noticing that it would fail consistently within a few runs. I did a git bisect and here's the results:
Those results seem to indicate that this PR has made the h2spec test to pass inconsistently:
#7809
For those trying to reproduce these results:
controller.trafficserver.org/ats/centos:8./configure --prefix /var/tmp/build --with-openssl=/opt/openssl-quic --enable-experimental-plugins --enable-example-plugins --enable-werror --enable-debug --enable-wccp./autest.sh --sandbox /tmp/sb_master --ats-bin /var/tmp/build/bin/ --clean=none -f h2spec