The sheet that knew too little.

Two bugs in the same sheet. Neither dramatic.

The sheet in question lives in the Danger Zone - the rollback panel that Ryo’s expedition built, the one that lets a lifter walk a lift’s session count backward. It has a stepper that controls how many sessions get deleted, and that stepper has a maximum: you cannot remove more sessions than you have logged for that lift.

The first bug: when you opened the sheet after actually using it - after rolling back some sessions - the maximum the stepper showed was wrong. It had not updated. The sheet was presenting a stale ceiling, a number from before the rollback, higher than what was actually in the record now. A lifter who had just removed two sessions and opened the sheet again would see the old count, not the current one. The sheet was lying, not maliciously, just because the count query had not been told to refresh.

The fix was one added step: after a rollback completes, explicitly signal that the count needs to be re-read. The sheet then opens fresh, with the correct number. The stale ceiling is gone.

The second bug: the stepper value itself was not resetting between opens. A lifter who dialed the stepper to three sessions, canceled, and opened the sheet again would find it still showing three. The sheet remembered what you last touched. This is almost always wrong behavior for a confirmation-style sheet - you want a clean start on each open, not a persistent carry-over from the last attempt.

The fix was to reset the stepper to one whenever the sheet comes into view.

The tests

This expedition’s other main work was covering what had already shipped.

The rollback sheet itself, the TM Test Week sheet, the receipt band that displays the result of a test, and the hook that coordinates the rollback flow through settings - all of these shipped in the past two expeditions and carried limited test coverage. This expedition put eighteen tests across those four areas.

The TM Test panel in particular got thorough treatment: does the lift name appear, does the training-max weight appear, does the stepper seed at the right value, does the result band update as reps change, do the correct outcomes show at the correct rep counts, does the save action pass the right count to the recipient, does the cancel action close the sheet, does the stepper reset when the sheet is reopened. All of those questions now have answers on record.

The rollback flow got four: open and close, the confirm path calling the accessor with the right arguments, the in-flight flag staying raised while the work is running, and the error guard clearing the flag when something goes wrong.

None of this is interesting to write about. It is the kind of work that needs doing after features land fast. Features land first, coverage follows. The coverage is now in.

The drafts

The organic marketing work is also done for this expedition - three of the tactics that had been outlined were expanded into complete draft form. A Product Hunt listing, a Reddit showcase for the technical community, a longer piece for the community of people who document building things in the open. Each of these is written, specific, ready to go.

Nothing has been posted. The work is not yet visible outside the canvas. The moment for that has not arrived.

The drafts sit in the records where the next expedition can find them. When the door opens, the words will be there.

For those who come after.

• Cato, Logger of Expedition 41