Fixes Dancer, adds Revelation Dance interactions with Z-Move, Roost and typeless mons

[PhallenTree]

Adds Tests for Revelation Dance.

The following Revelation Dance interactions were added:

• If a Pokémon is Typeless, Revelation Dance will have no type (in which case Revelation Dance will not receive STAB). If a Pokémon has lost its primary Flying type due to having used Roost earlier this turn and uses Revelation Dance due to Dancer, Revelation Dance's type will match its current type: its secondary type if it has any, or else be Normal-type (regardless of type additions by Forest's Curse or Trick-or-Treat).
• Typeless Revelation Dance does not activate Color Change and is blocked by Wonder Guard. Conversion 2 will fail if the last move used by the target was typeless Revelation Dance.
• Regardless of the user's types, Revelation Dance is always treated as Normal-type for Z-Move mechanics, and can be upgraded to Breakneck Blitz by a Normalium Z.

Also fixes Dancer-called moves not updating their type according to the new user of the move.

Feature(s) this PR does NOT handle:

When merging to upcoming, will be necessary to modify the checks for CheckDynamicMoveType.

Discord contact info

PhallenTree

[AlexOn1ine]

The CI fails.

[PhallenTree]

I've run out of ideas on how to fix this. Fails on CI, passes on manual build

[AsparagusEduardo]

I've run out of ideas on how to fix this. Fails on CI, passes on manual build

Are you able to make it fail locally by running at 4 threads? make -j4 check?

[PhallenTree]

Are you able to make it fail locally by running at 4 threads? make -j4 check?

Yes, while running locally at 4 threads the test fails

[AsparagusEduardo]

Yes, while running locally at 4 threads the test fails

What does the replay looks like? (you can open the elf file on mGBA)

[AsparagusEduardo]

Yes, while running locally at 4 threads the test fails

What does the replay looks like? (you can open the elf file on mGBA)

Wait, that's too many tests. Let's see if it fails just with make -j4 check TESTS="Revelation Dance".
EDIT: should be make -j4 check TESTS="Dancer"

[AsparagusEduardo]

Ok, finally made it fail locally with just Dancer tests: make -j3 check TESTS="Dancer"

[PhallenTree]

The Dancer test that fails seems to only fail if handled by thread 0. Maybe a problem with the test runner?

I tried moving the Dancer test to above the previous one and it passes after doing so (with 4 threads).
Should I just do that and add a comment to let others know that test fails if executed on thread 0?

[AsparagusEduardo]

In some cases, it seems that the battler sent to AbilityBattleEffects is changed. I'm not sure why it's happening.

But @AlexOn1ine identified that MOVEEND_DANCER's loop should use gBattlersCount instead of MAX_BATTLERS_COUNT, which fixes the issue. Can you apply this change?

-                   for (battler = 0; battler < MAX_BATTLERS_COUNT; battler++)
+                   for (battler = 0; battler < gBattlersCount; battler++)
                    {
                        if (GetBattlerAbility(battler) == ABILITY_DANCER && !gSpecialStatuses[battler].dancerUsedMove)
                        {
                            if (turnOnHitmarker)
                                gHitMarker |= HITMARKER_ATTACKSTRING_PRINTED;
                            if (!nextDancer || (gBattleMons[battler].speed < gBattleMons[nextDancer & 0x3].speed))
                                nextDancer = battler | 0x4;
                        }
                    }
                    if (nextDancer && AbilityBattleEffects(ABILITYEFFECT_MOVE_END_OTHER, nextDancer & 0x3, 0, 0, 0))
                        effect = TRUE;

[PhallenTree]

It seems to me that it still had the speeds of the battlers from the previous test ("Dancer triggers from slowest to fastest") hence it using battler = 3 (which is the slowest in that test).

But now with the fix, it should pick the right battler.