search

gazebosim / gz-transport

Transport library for component communication based on publication/subscription and service calls.
https://gazebosim.org
Apache License 2.0
29 stars 37 forks source link

:farmer: Gz-transport11-homebrew `UNIT_gz_TEST ignTest.TopicInfo` failed #510

Open Crola1702 opened 1 month ago

Crola1702 commented 1 month ago

Environment

Steps to reproduce

  1. Run a build in gz-transport11
  2. See ignTest.TopicInfo fail

Output

Reference build: https://build.osrfoundation.org/job/gz_transport-ci-ign-transport11-homebrew-amd64/59/consoleFull

Log output:

33: [ RUN      ] ignTest.TopicInfo
33: /Users/jenkins/jenkins-agent/workspace/gz_transport-ci-ign-transport11-homebrew-amd64/gz-transport/src/cmd/gz_TEST.cc:135: Failure
33: Value of: infoFound
33:   Actual: false
33: Expected: true
33: OUTPUT[No publishers on topic [/foo]
33: ] Size[30]. Expected Size=50
33: 
33: /Users/jenkins/jenkins-agent/workspace/gz_transport-ci-ign-transport11-homebrew-amd64/gz-transport/src/cmd/gz_TEST.cc:138: Failure
33: Value of: output.find("ignition.msgs.Vector3d") != std::string::npos
33:   Actual: false
33: Expected: true
33: [  FAILED  ] ignTest.TopicInfo (26598 ms)

Might be related to #509

scpeters commented 1 month ago

I suspect that this is a flaky test. It is now failing consistently on macOS since build 58, though it still passes on Ubuntu. There were no changes in the ign-transport11 branch from build 57 to build 58, and I compared the log files and did not see any significant changes in dependencies. Looking at the code for this failing test case, I see it starts a publisher and waits a few seconds, which seems like a race condition to me.