[copilot-finds] Test: Task and CompletableTask base classes have zero direct unit tests

[github-actions]

Problem

The Task<T> and CompletableTask<T> classes in packages/durabletask-js/src/task/task.ts and packages/durabletask-js/src/task/completable-task.ts are the foundational building blocks of the entire SDK. Every orchestration, activity, entity operation, timer, and sub-orchestration produces a Task or CompletableTask. Despite this, they have zero direct unit tests.

Currently they are only tested indirectly through when-all-task.spec.ts, when-any-task.spec.ts, and orchestration_executor.spec.ts. None of these files exercise the base class behavior in isolation.

Missing Coverage

Task<T> (base class)

• getResult() throws "Task is not complete" when called before completion
• getResult() re-throws the stored TaskFailedError when the task has failed
• getException() throws "Task did not fail" on a non-failed task
• isFailed correctly reflects state (false before failure, true after)
• isComplete correctly reflects state

CompletableTask<T>

• complete() throws "Task is already completed" on double-completion
• fail() throws "Task is already completed" on double-failure
• complete() after fail() throws (and vice versa)
• complete() notifies parent via onChildCompleted()
• fail() sets the exception as a TaskFailedError and notifies parent
• fail() without explicit details creates a default TaskFailureDetails

Root Cause

These classes were written early in the project's lifecycle and their behavior was considered simple enough to test only through integration with composite tasks and the executor. As the SDK has grown, this gap has become more significant.

Impact

• Severity: Medium — the classes are simple but foundational. A regression here silently breaks everything above.
• Risk: Without isolated tests, changes to the base class behavior (e.g., adding new states, modifying completion semantics) could introduce subtle bugs that composite tests don't catch.