Expected Behavior

• Abandon: Only allow when execution is STOPPED; otherwise log the current status and return a generic error exit code.
• Restart: Only allow when execution is FAILED or STOPPED; otherwise log the current status and return a generic error exit code.
• Resolve the TODOs in CommandLineJobOperator by performing explicit precondition checks at the CLI layer without relying on deprecated exceptions.

Current Behavior

• abandon(jobExecutionId): Delegates to JobOperator#abandon without enforcing STOPPED at the CLI level. A TODO mentions throwing JobExecutionNotStoppedException, but that exception is deprecated.
• restart(jobExecutionId): Contains a TODO to check/log when the job execution “did not fail,” but does not enforce valid restart states at the CLI level.

Context

• Motivation: Align CLI precondition checks with Spring Batch semantics and the inline TODOs, prevent invalid operations earlier, and improve observability with clear error logs that include the current status.
• Alternative considered: Throwing JobExecutionNotStoppedException, but it is deprecated; the CLI should use exit codes and logging instead.
• Compatibility: No API change; behavior becomes explicit and predictable. Invalid operations return JVM_EXITCODE_GENERIC_ERROR rather than relying on downstream exceptions.

Proposed Change

• In spring-batch-core/src/main/java/org/springframework/batch/core/launch/support/CommandLineJobOperator.java:
  • restart(long jobExecutionId): Enforce FAILED or STOPPED; otherwise log current status and return generic error.
  • abandon(long jobExecutionId): Enforce STOPPED; otherwise log current status and return generic error.

Previous Code (before change)

• restart(long jobExecutionId): Only TODO present; no state check; directly delegates to jobOperator.restart(jobExecution).

// TODO should check and log error if the job execution did not fail
JobExecution restartedExecution = this.jobOperator.restart(jobExecution);
return this.exitCodeMapper.intValue(restartedExecution.getExitStatus().getExitCode());

• abandon(long jobExecutionId): Only TODO present; no state check; directly delegates to jobOperator.abandon(jobExecution).

// TODO should throw JobExecutionNotStoppedException if the job execution is
// not stopped
JobExecution abandonedExecution = this.jobOperator.abandon(jobExecution);
return this.exitCodeMapper.intValue(abandonedExecution.getExitStatus().getExitCode());

A corresponding pull request has been carefully prepared to demonstrate this improvement.
I would appreciate it if the team could review and provide feedback when possible.