Hack for io.jenkins.configuration-as-code:test-harness

[jglick]

Same as #261 but with whitespace cleaned up. Apparently needed for bom though I do not know the details.

If we at least get an incremental deployment from this, #261 and #275 could be closed.

[jglick]

Need to get this updated, and then #299 verified in bom.

[imonteroperez]

@jglick it seems there are some IT failures: https://ci.jenkins.io/job/jenkinsci-libraries/job/plugin-compat-tester/job/PR-276/9/. I recently fixed them on master branch: https://ci.jenkins.io/job/jenkinsci-libraries/job/plugin-compat-tester/job/master/lastSuccessfulBuild/

[jetersen]

https://ci.jenkins.io/blue/organizations/jenkins/jenkinsci-libraries%2Fplugin-compat-tester/detail/PR-276/10/pipeline/167#step-168-log-863

[2021-09-28T10:36:42.697Z] [ERROR] Failed to execute goal on project artifact-manager-s3: Could not resolve dependencies for project io.jenkins.plugins:artifact-manager-s3:hpi:1.6: Could not find artifact io.jenkins.configuration-as-code:test-harness:jar:1.20 in repo.jenkins-ci.org (https://repo.jenkins-ci.org/public/) -> [Help 1]

[jglick]

Still

[ERROR] Failed to execute goal on project artifact-manager-s3: Could not resolve dependencies for project io.jenkins.plugins:artifact-manager-s3:hpi:1.15: Could not find artifact io.jenkins.configuration-as-code:test-harness:jar:1.20 in repo.jenkins-ci.org (https://repo.jenkins-ci.org/public/)

[jglick]

Hmm so artifact-manager-s3 just depends on JCasC in test scope but does not bother with its test harness:

ConfigurationAsCode.get().configure(ConfigAsCodeTest.class.getResource("configuration-as-code.yml").toString());
// assertions about ċonfig…

Apparently this patch breaks that mode.

[timja]

Hmm so artifact-manager-s3 just depends on JCasC in test scope but does not bother with its test harness:

ConfigurationAsCode.get().configure(ConfigAsCodeTest.class.getResource("configuration-as-code.yml").toString());
// assertions about ċonfig…

Apparently this patch breaks that mode.

weird, could either patch that here or just switch it to use the test harness, nothing in bom hits this case at least

[jglick]

Trying to reproduce in a more development-friendly mode than the Docker IT, so far without success (running tests normally):

mvn -f plugins-compat-tester-cli -Dexec.vmArgs= '-Dexec.args=${exec.vmArgs} -classpath %classpath ${exec.mainClass} ${exec.appArgs}' '-Dexec.appArgs=-overridenPlugins io.jenkins:configuration-as-code=1.20 -includePlugins artifact-manager-s3 -reportFile /tmp/pct/report.xml -workDirectory /tmp/pct' -Dexec.mainClass=org.jenkins.tools.test.PluginCompatTesterCli -Dexec.executable=$(which java) exec:exec

This does

     <dependency> 
       <groupId>io.jenkins</groupId>  
       <artifactId>configuration-as-code</artifactId>  
       <scope>test</scope>  
+      <version>1.20</version> 
     </dependency>  

as expected. However it also does this, which is clearly wrong:

+    <dependency> 
+      <groupId>io.jenkins</groupId>  
+      <artifactId>configuration-as-code</artifactId>  
+      <version>1.20</version> 
+    </dependency>  

And this (twice!):

+    <dependency> 
+      <groupId>io.jenkins.configuration-as-code</groupId>  
+      <artifactId>test-harness</artifactId>  
+      <version>1.53</version>  
+      <scope>test</scope> 
+    </dependency>  

PCT is such a train wreck…

[jglick]

Even without my patch, PCT is wrong:

$ mvn dependency:tree|fgrep configuration-as-code
[WARNING] 'dependencies.dependency.(groupId:artifactId:type:classifier)' must be unique: io.jenkins:configuration-as-code:jar -> duplicate declaration of version 1.20 @ line 324, column 17
[WARNING] 'dependencies.dependency.(groupId:artifactId:type:classifier)' must be unique: io.jenkins:configuration-as-code:jar -> version 1.20 vs 1.53 @ line 413, column 17
[INFO] +- io.jenkins:configuration-as-code:jar:1.53:test

since besides adding the version as expected, it also

+    <dependency> 
+      <groupId>io.jenkins</groupId>  
+      <artifactId>configuration-as-code</artifactId>  
+      <version>1.20</version> 
+    </dependency>  
+    <dependency> 
+      <groupId>io.jenkins</groupId>  
+      <artifactId>configuration-as-code</artifactId>  
+      <version>1.53</version>  
+      <scope>test</scope> 
+    </dependency>  

[jglick]

Will try merging with master to pick up #295 etc.

[timja]

It's already merged

[jglick]

Oh right, I get thrown off guard when other people push to my PRs…

[jglick]

Even without my patch, PCT is wrong

Still true after merging master. Somehow it did not matter fur purposes of this IT.

[jglick]

@timja do you even remember at this point what fails in bom without this patch?

[timja]

@timja do you even remember at this point what fails in bom without this patch?

no this was a long time ago

[jglick]

Rather than merging in #341, it may be cleaner to file a separate draft PR combining the two just to get an incremental deployment.

[timja]

Rather than merging in #341, it may be cleaner to file a separate draft PR combining the two just to get an incremental deployment.

Possibly, does this version actually work though or is there stuff from master that needs backing out in this PR?

[jglick]

is there stuff from master that needs backing out

Uh…not sure. Should be tracked in jenkinsci/bom#657.

[timja]

I’ll try it out tomorrow unless you happen to have time :)

[timja]

Need to get this updated, and then #299 verified in bom.

causes bom to fail in injected test which we normally ignore failures in

bisect shows that's what is causing the failure atm in jenkinsci/bom#863

[jglick]

Presumed obsolete after jenkinsci/bom#1336.