Recover views after error in Jenkins.load

[jglick]

I had occasionally seen mysterious errors in CloudBees CI which I finally tracked down by observing that they came after a warning about reloading $JENKINS_HOME/config.xml from disk. #8413 made this somewhat safer but not enough.

(CloudBees-internal reference)

Testing done

Without the fix, the assertions about .primaryView and .views fail. More interestingly, while browsing the dashboard (goTo("")) simply fails with a 404 (Stapler trace is pretty obvious: looking up the primary view fails since it is null), browsing some other pages produces two StackOverflowErrors from Stapler/Jelly:

…
	at org.kohsuke.stapler.jelly.JellyViewScript.run(JellyViewScript.java:98)
	at org.kohsuke.stapler.jelly.IncludeTag.doTag(IncludeTag.java:174)
	at org.apache.commons.jelly.impl.TagScript.run(TagScript.java:271)
	at org.kohsuke.stapler.jelly.JellyViewScript.run(JellyViewScript.java:98)
…

and

…
	at org.apache.commons.jelly.JellyContext.getVariableWithDefaultValue(JellyContext.java:260)
	at org.apache.commons.jelly.JellyContext.getVariableWithDefaultValue(JellyContext.java:260)
…

which seems to be caused by an unfortunate default logic in IncludeTag resulting in an infinite recursion when it=null.

Proposed changelog entries

• Restore the list of views after a failed attempt to reload global configuration, to avoid further rendering errors.

Proposed upgrade guidelines

N/A

Desired reviewers

@Vlatombe

Before the changes are marked as ready-for-merge:

Maintainer checklist

Preview Give feedback

1. There are at least two (2) approvals for the pull request and no outstanding requests for change.
2. Conversations in the pull request are over, or it is explicit that a reviewer is not blocking the change.
3. Changelog entries in the pull request title and/or Proposed changelog entries are accurate, human-readable, and in the imperative mood.
4. Proper changelog labels are set so that the changelog can be generated automatically.
5. If the change needs additional upgrade steps from users, the upgrade-guide-needed label is set and there is a Proposed upgrade guidelines section in the pull request title (see example).
6. If it would make sense to backport the change to LTS, a Jira issue must exist, be a Bug or Improvement, and be labeled as lts-candidate to be considered (see query).

[dwnusbaum]

FWIW, there was a similar issue with views in folders caused by HashiCorp Vault plugin, fixed by jenkinsci/hashicorp-vault-plugin#336 and made more robust in core by #9653, leading to the same recursion as in jenkinsci/cloudbees-folder-plugin#447.

Do you have an OldDataMonitor warning for an observed loading failure? I wonder if any other XStream robustness improvements are possible here.

[jglick]

Do you have an OldDataMonitor warning for an observed loading failure?

No, OldDataMonitor is only triggered when XStream unmarshalling completed with some recoverable issues of known categories, like missing fields. The situation here is not necessarily about XStream; I used invalid XML for the test case but I think I saw the bug in response to some transient IOException loading config.xml.

[timja]

/label ready-for-merge

---

This PR is now ready for merge, after ~24 hours, we will merge it if there's no negative feedback.

Thanks!