Lesson 07 · Three tasks and an early finisher - Build a Green-Thread Scheduler

Three tasks and an early finisher

Round-robin should stay fair even when tasks have different lifespans. Today a task that finishes early simply drops out of the rotation, and the ones still running close ranks - the trace proves nobody skips a turn and nobody lingers.

The goal

Run three tasks of unequal length and assert a finished task leaves the rotation cleanly.

Start here - the target

TO DO

Scenario: A short task drops out mid-rotation

Giventasks 1, 2, 3 spawned in order, where task 2 finishes on its first step and tasks 1 and 3 each yield once then finish on their second step

WhenRun is called

Thenthe run trace is exactly [1, 2, 3, 1, 3]

Andtask 2 appears only once because it never yielded back into the queue

Background

Real programs mix long-lived and short-lived green threads freely, and the scheduler must not stumble when one finishes ahead of the others. It does not, because Done just means “do not re-enqueue” - the finished task quietly vanishes and the queue carries on with whoever is left.

Follow the queue: it starts [1, 2, 3]. Task 1 yields, giving [2, 3, 1]. Task 2 finishes and is dropped, giving [3, 1]. Task 3 yields, giving [1, 3]. Task 1 finishes, giving [3]. Task 3 finishes. The trace is [1, 2, 3, 1, 3] - task 2 appears exactly once, and the survivors keep alternating without missing a beat. This is the property that lets you spawn and retire green threads at will.

Make it work

// task 2 is a one-shot; tasks 1 and 3 yield once
oneShot := func() Status { return Done }
// reuse twoStep() from the previous lesson for tasks 1 and 3
// spawn order: twoStep, oneShot, twoStep

CheckpointDONE

A finished task drops out of the rotation while the rest keep taking turns. Commit and stop here.