Fix onMarkerHit not firing in interiors (#5018)

[QueryOfficial]

Summary

Fixed onMarkerHit not firing in interiors when a geometric collision was recorded but the event was silently cancelled due to an interior mismatch. Interior (and dimension for pickups) checks were moved from callbacks into collision state management; hit detection is now refreshed after setElementInterior and createMarker.

Affected events:

• Server: onMarkerHit, onMarkerLeave, onPickupHit, onPickupLeave
• Client: onClientMarkerHit, onClientMarkerLeave, onClientPickupHit, onClientPickupLeave

Motivation

Fixes #5018

CColManager::HandleHitDetectionResult recorded collisions based on geometry only, while CMarker::Callback_OnCollision blocked the event entirely when interiors did not match. This mismatch created a stale collision state: even after the player later entered the correct interior, CollisionExists was already true, so onMarkerHit never fired again.

The same root cause also affected pickup and client-side marker/pickup events.

Test plan

Original repro (server console):

srun m = createMarker(me.position + Vector3(0,0,0.5), 'cylinder', 1.1, 255, 255, 255)
addEventHandler('onMarkerHit', m, function(he,md) outputChatBox('HIT '..tostring(he)..' md='..tostring(md)) end)
setElementInterior(m, 14)
setElementInterior(me, 14)

• onMarkerHit should fire.

Additional scenarios:

• Player passes through the marker area in interior 0 → event does not fire; then setElementInterior(me, 14) → onMarkerHit fires.
• Player is inside the marker and setElementInterior(me, 0) is called → onMarkerLeave fires.
• With a dimension mismatch, onMarkerHit still fires with matchingDimension = false.
• Same interior stale-state scenario verified for pickups via onPickupHit.
• Client-side onClientMarkerHit tested with the local player using the same scenarios.
• When interiors match, isElementWithinMarker returns true and onMarkerHit fires consistently.

Checklist

• Your code should follow the coding guidelines.
• Smaller pull requests are easier to review. If your pull request is beefy, your pull request should be reviewable commit-by-commit.

[Lpsd]

Do you know what caused this (a specific commit/PR)? Or has it always been broken?

[HeresHavi]

Do you know what caused this (a specific commit/PR)? Or has it always been broken?

November 23, 2008

• Commit: de636610
• SetInterior only changed the interior value and did not refresh collision detection.

August 4, 2009

• Commit: f2a99c7c
• Added interior checks to server-side marker and pickup callbacks.
• Collisions were still recorded before these checks, creating the stale collision state.

March 18, 2021

• PR Add on(Client)ElementInteriorChange #2058 / commit: 01059438
• Added interior-change events, but no collision reevaluation.

July 31, 2025

• PR Fix #3310: Improved client-side markers & Fix logic #3324 / commit: 7108a8d2
• Added the equivalent interior check to client-side marker callbacks.

Conclusion

The missing collision refresh is longstanding, but the stale server-side collision behavior appears to originate from f2a99c7c in 2009. The client-side equivalent was introduced separately in 2025.

[QueryOfficial]

Do you know what caused this (a specific commit/PR)? Or has it always been broken?

Agreed with this timeline. The stale collision state was introduced when interior checks were added to callbacks (f2a99c7, 2009) without updating CColManager to skip recording collisions when interiors don't match. The missing refresh on SetElementInterior (de63661, 2008) made it easier to trigger. This PR addresses both: interior filtering is now part of collision state management, and hit detection is refreshed on interior changes.