5 | | * Turn 139 is exactly the turn where this unit has finished gathering and decides to dropoff the resources. |
6 | | * As the hash values match before turn 139, the position of the gathering archer for those turns must be identical too. The unit didn't move until it had finished chopping wood. |
7 | | * I assume the rejoined client decides to dropoff earlier or later. |
| 7 | * Turn 138: In the replay (which matches the host), the unit has 10 wood (finished gathering). Stands at the tree without formation for one turn and then starts to walk to the dropsite. |
| 8 | * As the hash values match before turn 139, the position of the gathering archer for those turns must be identical too. |
| 9 | * I assume the rejoined client decides to dropoff one turn earlier. |
| 10 | |
| 11 | '''Why 'the serializationtest doesn't catch it:'' |
| 12 | * The serializationtest simulates only one turn of a rejoined client. But the OOS occurs multiple turns after the rejoin. Thus it is logical that serializationtest doesn't catch it. |
| 13 | * We could mofiy the serializationtest to initialize the secondary simulation on turn 124 and then continue the simulation after that, instead of reseting it. Should trigger the OOS error. |
| 14 | |