fix: handle null FlowExecution in PipelineLogExtractor

[lidiams96]

Environment

• Jenkins Version: 2.479.3 (tested with mvn hpi:run)
• Plugin Version: latest main (post PR feat: add AWS Bedrock provider #95 merge)
• Java Version: 17

Steps to Reproduce

1. Install the plugin and configure a Bedrock (or any) provider
2. Run a pipeline that fails early or where WorkflowRun.getExecution() returns null
3. The plugin attempts to walk the flow graph via PipelineLogExtractor.getFailedStepLog()
4. FlowGraphWalker receives a null execution and throws NullPointerException

Reproduced locally with mvn hpi:run when running a simple failing pipeline — the WorkflowRun.getExecution() returned null because the execution context was not fully initialized.

Expected Behavior

The plugin should gracefully fall back to the overall build log (run.getLog(maxLines)) when no flow execution is available, consistent with its existing behavior for non-pipeline builds.

Actual Behavior

java.lang.NullPointerException: Cannot invoke "org.jenkinsci.plugins.workflow.flow.FlowExecution.getCurrentHeads()" because "exec" is null

Fix

Added a null check for FlowExecution before creating FlowGraphWalker. When execution is null, the method falls back to run.getLog(maxLines) — the same fallback used when no failed step log is found or when the build is not a pipeline.

FlowExecution execution = ((WorkflowRun) this.run).getExecution();
if (execution == null) {
    setUrl("0");
    return run.getLog(maxLines);
}

This can happen in several scenarios:

• Pipeline execution has not started yet (very early failure)
• Execution was cleaned up by garbage collection
• Local test environments (mvn hpi:run) where the pipeline context is not fully initialized

[Copilot]

Pull request overview

This PR fixes a NullPointerException that occurs when WorkflowRun.getExecution() returns null during pipeline log extraction. The fix adds a defensive null check before creating FlowGraphWalker, allowing the plugin to gracefully fall back to the overall build console log instead of crashing. This scenario can occur when pipeline execution hasn't fully initialized, has been cleaned up by garbage collection, or in local test environments.

Changes:

• Added null check for FlowExecution before creating FlowGraphWalker in PipelineLogExtractor.getFailedStepLog()
• Graceful fallback to run.getLog(maxLines) when execution is null, consistent with existing fallback behavior for non-pipeline builds

[shenxianpeng]

Thanks!