(workflows): update rollback context types

[vaishnav-mk]

adds ctx: WorkflowStepContext and deprecated legacy stepName to WorkflowRollbackContext. Makes rollback required when rollback options are provided, narrows rollbackConfig to retries and timeout, and updates generated type snapshots plus RPC type tests

[ask-bonk]

All snapshots are consistent with the source definition. The changes are straightforward and correct:

1. stepName: string removed from WorkflowRollbackContext
2. ctx: WorkflowStepContext added to WorkflowRollbackContext
3. Generated snapshots updated consistently
4. Type tests updated to verify the new shape

The WorkflowStepContext type already contains step.name (as part of step: { name: string; count: number }) along with attempt and config, so users still have access to the step name via ctx.step.name - strictly more information than the old stepName string. The type tests cover both overloads of workflowStep.do().

This is a clean, well-scoped type change with no issues.

LGTM

github run

[petebacondarwin]

Perhaps this horse has already left the stable, but won't removing this property break users of Workflows?

[vaishnav-mk]

/bonk

[ask-bonk]

Thread Summary

PR #6807 updates the WorkflowRollbackContext type for Workflows:

• Removes stepName: string from WorkflowRollbackContext
• Adds ctx: WorkflowStepContext to WorkflowRollbackContext
• Updates all generated type snapshots and RPC type tests accordingly

Discussion:

1. ask-bonk (bot) confirmed the changes are clean and consistent. Noted that WorkflowStepContext already contains the step name via ctx.step.name (plus attempt and config), so users still have access to the same info and more.

2. ryanking13 left an approval.

3. petebacondarwin raised a concern: removing stepName is a breaking change for existing Workflows users.

4. vaishnav-mk (PR author) acknowledged it's intentional. The plan is to:

   • Remove stepName from the public types to discourage its use
   • Continue supporting it internally for a while for identified users who still rely on it
   • Steer users toward ctx instead, which offers better compatibility with other features
   • Asked Pete for his thoughts and cc'd @avenceslau

Current status: Pete's concern about the breaking change is the open question. The author has explained the rationale but hasn't yet received a follow-up response from Pete or avenceslau.

github run