SCMSource: Check task authentication context when looking forcredentials

[cghislai]

GiteaSCMSource was sometimes using the authenticated user associated with the request.
This was causing occasional requests to be sent without authorization headers, and receiving 404 responses.
This could happen by manually triggering a single branch in a multi-branch pipeline project, or replaying a job. Triggering a job that lanches a parametrized build of a branch job also triggered the issue.

This change checks for Queue.Task instance, and fallbacks to the the System credential. This is what is done elsewhere.
The user associated with the request is this not used anymore when retrieving credentials by id.

Fixes JENKINS-62326

[justusbunsi]

Hi. I think I am running into this issue often. So I'm happy to test drive your changes. Unfortunately, this is going to be a long running thing since it's not always happen.

[cghislai]

@justusbunsi thanks. Remember you can compile the hpi and upload it to any running instance manually via the plugins page, thats how i tested it. (edit: didnt read the last few words. you should be able to trigger it by launching a multibranch pipeline job for one branch manually?)

[justusbunsi]

you should be able to trigger it by launching a multibranch pipeline job for one branch manually

Hrm. Looks like I misinterpreted a lot. 🤦
After reviewing the stacktrace in the referenced issue, it don't look similar. 😕 I always get an error on fetching collaborators. Also, I am currently running my jobs as SYSTEM (bad idea), so I'm not affected by this issue.

java.lang.NullPointerException
	at org.jenkinsci.plugin.gitea.client.impl.DefaultGiteaConnection.fetchCollaborators(DefaultGiteaConnection.java:391)
	at org.jenkinsci.plugin.gitea.GiteaSCMSource.getTrustedRevision(GiteaSCMSource.java:603)
	at org.jenkinsci.plugins.workflow.multibranch.SCMBinder.create(SCMBinder.java:104)
	at org.jenkinsci.plugins.workflow.job.WorkflowRun.run(WorkflowRun.java:309)
	at hudson.model.ResourceController.execute(ResourceController.java:100)
	at hudson.model.Executor.run(Executor.java:433)

Anyway, I'll test your changes. Reading the issue and your suggestion on reproducing the issue, it happens all the time, right?

[cghislai]

Yes that is a different issue, please file a bug report if none exists already. I was able to trigger the issue at hand here by running a branch job on a multibranch project using gitea scm, every time indeed.

[justusbunsi]

@cghislai Please sync with master branch to get a working build.