Bug: `AI_Smart_Encore` can lead to incorrect type matchup checks

[UnderscorePoY]

File : engine/battle/ai/scoring.asm (linked to Draft Pull Request #784)

The raw call to CheckTypeMatchup might have some oversights.

• Contrarily to the general comment of this routine, since wEnemyMoveStruct is populated with the last player move, this will be used properly in CheckTypeMatchup (putting it in a is indeed not how to use the matchup routine, but it is irrelevant in that case).

• However, if the AI flow can lead to this clause being called with a hBattleTurn set to 0, the type matchup will consider the player types instead of the enemy types.

• AI layers which can cause this oversight if called before AI_Smart_Encore (without hBattleTurn being set back to 1 unvoluntarily) :

  • AI_Smart_Conversion2
  • TODO : there could be some other very tricky situations like the last turn using Metronome->Conversion2 if the AI switching conserves hBattleTurn, etc.

• AI layers which unvoluntarily avoid this oversight (by setting hBattleTurn to 1) :

  • AI_Types (SuperNerd, Juggler, ExecutiveM/F, PokefanM/F).
  • AI_Status reaching the .typeimmunity jump label.
  • AI_Smart_LeechHit.
  • AI_Smart_LockOn when the player isn't locked on.
  • AI_Smart_Mimic when player used a move & enemy HP > 50%.
  • AI_Smart_PriorityHit when the enemy is slower or speedtied and the player is not using Fly or Dig.
  • AI_Smart_HiddenPower.

Example (not encoutered in vanilla) :

• Player : Ho-Oh with Mud-Slap.
• Enemy : SMART_AI Raichu with Conversion2, Encore in that order (any other moves before or after).
• Turn 1 : Raichu uses another move. Ho-Oh uses Mud-Slap.
• Turn 2 : Raichu will wrongly consider types and will use Ho-Oh's after the AI_Smart_Conversion2 layer. It will see Ho-Oh as immune to Mud-Slap and consider this matchup its. Thus Encore is 72% greatly encouraged (-2 score) despite Mud-Slap being super-effective against Raichu.

[FredrIQ]

I'm fairly sure that the AI in general assumes that SetEnemyTurn has been prepared, meaning that the problem here is Conversion2, not Encore?

[rawr51919]

Wondering if the documentation from https://github.com/pret/pokecrystal/pull/656 fixed this originally

[rawr51919]

Documented in #936, awaiting on a fix