Fix: Handle bot users gracefully when metadata is unavailable

[isidromontero]

Problem Description

Repository scans fail completely with a FATAL error when a Pull Request is authored by a bot user (such as GitHub Copilot) whose user metadata is not accessible via the GitHub API.

Error Details

When scanning a repository with PRs from bot users, the scan aborts with:

java.io.FileNotFoundException: https://api.github.com/users/Copilot {"message":"Not Found","documentation_url":"https://docs.github.com/rest/users/users#get-a-user","status":"404"}
    at org.kohsuke.github.GitHubResponse.wrapException(GitHubResponse.java:101)
    ...

Impact

• Entire repository scan fails with FATAL error status
• Orphaned item cleanup strategy does not execute because the scan never completes successfully
• This leads to hundreds of orphaned PR branches accumulating and consuming disk space
• Real-world case: 1,757 PR branches consuming 360GB+ of disk space

Root Cause

In GitHubSCMSource.java, the CacheUpdatingIterable.observe() method (line 2596) throws a WrappedException when user metadata cannot be fetched:

} catch (FileNotFoundException e) {
    request.listener()
            .getLogger()
            .format(
                    "%n  Could not find user %s for pull request %d.%n",
                    user == null ? "null" : user.getLogin(), number);
    throw new WrappedException(e);  // FATAL - aborts entire scan
}

This behavior is problematic because:

1. Bot users like GitHub Copilot don't have accessible user metadata via /users/Copilot endpoint (returns 404)
2. This is expected behavior, not an exceptional error that should abort the scan
3. The scan should be resilient to missing metadata and continue processing other PRs

Changes Made

This PR implements tolerant error handling that uses default values when user metadata is unavailable:

Key Changes

1. Graceful degradation for user metadata:

   • Try to fetch user.getName() and user.getEmail()
   • On FileNotFoundException (404), use default values: login as name, login@users.noreply.github.com as email
   • Log warning for visibility without aborting scan

2. Remove unnecessary try-catch:

   • Removed redundant IOException catch for PR metadata methods (getTitle(), getBody(), getHtmlUrl()) which don't throw checked exceptions

3. Maintain backward compatibility:

   • All existing successful paths remain unchanged
   • Only affects error handling for unavailable metadata

Code Example

// Try to get user metadata, but be tolerant to API failures
String userName = login;
String userEmail = login + "@users.noreply.github.com";
try {
    userName = user.getName();
    if (userName == null || userName.isEmpty()) {
        userName = login;
    }
    userEmail = user.getEmail();
    if (userEmail == null || userEmail.isEmpty()) {
        userEmail = login + "@users.noreply.github.com";
    }
} catch (FileNotFoundException e) {
    // User metadata not accessible (e.g., bot users like Copilot)
    // Log warning but continue with default values
    request.listener()
            .getLogger()
            .format(
                    "%n  Could not find user metadata for %s (PR #%d). Using default values.%n",
                    login, number);
} catch (IOException e) {
    // Other IO errors, use default values but log
    request.listener()
            .getLogger()
            .format(
                    "%n  IO error fetching user metadata for %s (PR #%d): %s. Using default values.%n",
                    login, number, e.getMessage());
}

Benefits

• ✅ Repository scans complete successfully even with bot-authored PRs
• ✅ Orphaned item cleanup strategy executes as configured
• ✅ Bot PRs are indexed with sensible default values
• ✅ Warnings are logged for visibility without blocking the entire scan
• ✅ Graceful degradation instead of catastrophic failure

Testing

Tested locally by:

1. Building custom plugin with these changes
2. Installing on Jenkins instance with 1,757 PR branches (including Copilot-authored PRs)
3. Confirming scans complete successfully with warnings instead of fatal errors
4. Verifying orphaned item strategy executes properly

Checklist

• Code compiles successfully (mvn clean package -DskipTests)
• Changes are backward compatible
• Error messages are informative
• Real-world testing completed
• Commit message follows conventions

---

Note: This fix addresses a critical production issue where bot-authored PRs prevent repository indexing and orphaned item cleanup, leading to significant disk space consumption.

[tcassaert]

This is also impacting us. It leads to not triggering builds because we rely on the branch indexing.