Checkpoints

What a checkpoint is#

A checkpoint is an ordinary save in its own slot. That is the whole design: because it is a normal save, it inherits staging, verification, backup and recovery without a second code path that could rot.

var checkpoints = new CheckpointService(slots, options.Checkpoints, clock, diagnostics);

var created = await checkpoints.CreateAsync(saveStateSource);   // a recovery point, now
var listed  = await checkpoints.ListAsync();                    // oldest first, with labels
await checkpoints.PruneAsync();                                 // rotation, on your terms

Rotation that cannot cost the player#

The oldest checkpoint is deleted only after a newer one has been written safely. A failing checkpoint can therefore never cost the player the previous one — the write that matters is already on disk before anything is removed.

When to take one#

• before a risky transition (a chapter boundary, a long cutscene, a boss gate);
• after a large, deliberate state change that the player would not want to repeat;
• on a schedule the game controls, for players who expect automatic safety nets.

Next#

• Autosave — the automatic half of the same idea.
• Backups — the per-slot previous versions.
• Save slots — how checkpoint slots look in a listing.