Use a breakpoint to stop at the problem block
In Savebot: The Last Garden, use a breakpoint on the block where you expect the robot to make the wrong decision. Run that robot and inspect the active block and its live condition when the debugger stops.
Full Steam release unless a source specifies the demo. Evidence checked . Guide updated .
Steps
- Choose the robot with the repeatable problem and open its program. Identify the earliest decision where actual and intended behavior diverge.
- Set a breakpoint on that block and run a case that should reach it. Use the highlighted running line or the stop itself to verify you are observing the same robot.
- Check the live condition before editing anything. Compare the value and resulting branch with your expected action. Re-run the same case after one small change.
How to tell it worked
The debugger stops where you expected and you can name the condition or instruction that led to the problem.
If you are still stuck
If the breakpoint never fires, move it earlier to see whether the robot takes another branch or never enters the loop. Avoid adding multiple breakpoints until you know what input reaches that part of the program.
Select a location that distinguishes two explanations
Start with an expected decision and one failing run. For example, if the Collector reaches its delivery area empty, decide whether you first need to know if it selected collection or whether it managed to pick anything up. A breakpoint at the earlier relevant decision can show if its ordinary work was interrupted by a higher-priority condition. A later breakpoint on the delivery step only helps once you know the robot reached that part of its program.
Select the specific robot that failed and find the block or line governing the choice. Note the starting state you want to reproduce. Set a breakpoint and reproduce that state. If another robot makes progress elsewhere, leave the test on the original one. The live value belongs to the selected robot and is useful only if it is the one whose action you are trying to explain.
Read the condition and decide where to look next
When the debugger stops, read the marked location and the value you can observe. Predict which action should follow on that input. If both relevant conditions are true, use their order to explain the priority before moving another block. In the official Collector sample, low battery has priority over full cargo. To inspect the full-cargo path itself, wait for a run with enough battery and a full load.
If the breakpoint never fires, place the next test one step earlier. An earlier stop in the loop can show which route the program takes instead. If even that breakpoint never fires, check the selected robot and whether the relevant routine is running. If the earlier decision runs and points away from your breakpoint, inspect the condition that sent it elsewhere. It may be correctly avoiding the path you originally chose.
Preserve one test after the edit
Make the smallest correction that explains what you saw. Run the same starting case and stop at the same decision. Compare the observation against what you expected to change. Then try a neighboring state that should keep working, such as cargo with space remaining. This helps catch a fix that makes one case behave while reversing another.
Resume ordinary running once the trace is understood and look for the actual end of the task. A breakpoint explains what happened in the program. The pickup, delivery or shelter is what establishes whether the robot achieved the job. If the action was selected correctly and the result is still missing, use the route or target as the next thing to observe.