[gvdutra]

Fixes #8934.

Use the current branch's configured upstream when pulling instead of assuming that the remote branch has the same name. The existing origin/<local-branch> fallback remains available for branches without upstream configuration.

Pulling a branch with no configured upstream and no matching branch on origin now reports a clear error instead of failing with a NullPointerException.

Tests:

• Added coverage for pulling from an upstream with a different branch name.
• Added coverage for the missing-upstream error.
• ./mvnw -pl plugins/misc/git test (234 standard tests and 90 UI tests)
• ./mvnw -pl plugins/misc/git spotless:apply spotless:check

---

• Run mvn clean install apache-rat:check to make sure basic checks pass. A more thorough check will be performed on this pull request automatically.
• The change is a single commit.
• The related issue is mentioned above.
• I hereby declare this contribution to be licensed under the Apache License Version 2.0, January 2004

[bamaer]

Thanks @gvdutra, this fixes the case from #8934. I built the PR and ran the git plugin tests: all 324 pass, and both new tests fail on main.

Should fix

• When the upstream is on a remote other than origin (for example feature tracking upstream/master), pull() still fetches only origin (UIGit.java:777). It then merges a stale upstream/master and reports "already up to date" even though upstream has new commits. On main this case failed with the NPE; with this PR, nothing happens and no error is shown.
  Suggested fix: fetch the branch's configured remote, new BranchConfig(config, branch).getRemote(), fall back to origin, and skip the fetch when the remote is . (the upstream is a local branch). With that change, a test with a second remote pulls the new commit and the existing tests still pass. A variant of testPullFromConfiguredUpstreamWithDifferentName with a second remote would cover it.