TileMap collisions are not removed/changed when set_tile is used (#4829)

[totebo]

To replicate:

1. Create a tilemap with a factory.
2. Use set_tile to change a collision tile to a different tile.

Minimal example: SetTileCollisions.zip (31.0 KB)

Possibly related:

Removing tiles from tilemap updates collision shape but does not affect objects at rest (DEF-3810) (SOLVED)

Encountered similar issue (kinematic and dynamic collision objects interacting with static tilemap). OS: Windows 10. Defold: 1.2.146 Reproduce: Allow kinematic/dynamic collision objects to reach steady-state on collision tile. (e.g. player or object landing on tile) set_tile() sets the tile to one not included in the collision mask. kinematic/dynamic object remains at rest. Initiating movement on the kinematic object recognizes updated collision mask, but only when the contact normal vector…

[britzl]

Hmm, strange, I thought we had fixed this. @JCash what are your thoughts on this?

[Mathias_Westerdahl]

No real idea I’m afraid, we’ll just have to take a look at it again.

[totebo]

New game, same issue. Any revelations in regards to this?

[Mathias_Westerdahl]

Sorry, no. We haven’t had time to look into this yet.

[britzl]

It feels like this should be fairly easy to solve.

• tilemap.set_tile() defined here: https://github.com/defold/defold/blob/dev/engine/gamesys/src/gamesys/scripts/script_tilemap.cpp#L207
• calls SetTileGridTile(): https://github.com/defold/defold/blob/dev/engine/gamesys/src/gamesys/components/comp_tilegrid.cpp#L225
• SetTileGridTile() sets the tilegrid to dirty and UpdateRegion() will be called: https://github.com/defold/defold/blob/dev/engine/gamesys/src/gamesys/components/comp_tilegrid.cpp#L269
• UpdateRegion() will clear the dirty flag
• And the collision object component checks the dirty flag as well: https://github.com/defold/defold/blob/dev/engine/gamesys/src/gamesys/components/comp_collision_object.cpp#L901

So the question is if the dirty flag gets cleared before the collision object component reads it, and thus never updated the collision object shape?

Is this a reasonable analysis @JCash?

[Mathias_Westerdahl]

It sounds very plausible, yes.

[britzl]

Created https://github.com/defold/defold/issues/4829