I am using the Parallel Test Executor plugin, so my unit tests are splitted into splits executed in parallel on different nodes. This is a simplified part of my Jenkinsfile:
{code:java}
#!/usr/bin/env groovy
stage('Test') {
def testSteps = [:]
def splits = splitTests parallelism: count(env.NUMBER_EXECUTORS as int)
for (int i = 0; i < splits.size(); i++) {
def num = i
def split = splits[num]
testSteps["Test split${num}"] = {
node('!master') {
checkout scm
writeFile file: ("test_exclusions.txt"), text: split.join("\n")
sh './gradlew test'
junit '**/build/test-results/**/TEST-*.xml'
}
}
}
parallel testSteps
}
{code}
We make heavy use of the test result history to see when the build first broke and who is responsible for that.
The issue I am facing is the fact, that the ID of the test result contains the name of the parallel execution step as the prefix, in this case "Test split2" for example. As a result, a broken test is shown in the history as "Test / Test split2 / my.package.MyTest.testStuff()".
This is nice in theory, but there is one issue: The history only shows me the history of a test case executed in this specific split. For example, it may be that *"Test / Test split2 / my.package.MyTest.testStuff()"* is shown as broken with an age of 1, although it may be broken since multiple builds, just that this specific test case happened to be put into different splits before.
To summarize: *"Test / Test split2 / my.package.MyTest.testStuff()"* and *"Test / Test split3 / my.package.MyTest.testStuff()"* have separate histories, but I need to track their histories together.
|
|
