Box2D Forums

What is the expected behavior for a collision, after failing the contact filter's ShouldCollide()? Examining the source, it looks like the collision would be retested every step. However, in practice:

* 2 simple rectangles are overlapping - 1 dynamic & 1 static. There is no contact created because ShouldCollide() is false. Neither has vertexes within the other - picture a cross.
* I then change some internal rules so that ShouldCollide() will return a collision between them. However, it seems ShouldCollide() is not getting called again for the 2 fixtures, even when the dynamic object is moving. (There are no joints involved) It only seems to reported again if a vertex of 1 object crosses in/out of the other object.

Is this expected? If so, how can I force a reexamination of existing collisions for a fixture?

Thanks!

OK it doesn't seem to be related to vertexes - it seems to be pretty random actually, sometimes it reprocesses the collision and sometimes not.

The broad-phase uses enlarged AABBs to reduce the number of updates. Once you filter out a collision, you can't get the collision to be solid again until the objects move substantially. However, you can call b2Fixture::SetFilterData to trigger re-filtering.

From the manual:

Yes SetFilterData calls contact->FlagForFiltering on all contacts - but there is no contact for this collision so nothing to re-filter. Would I have to move the object off-screen then back the next step, to override this? Or can I remove it from the broad-phase list somehow? I'm not averse to changing the code I'm just not sure where to look.

Also, I notice you removed fixture.Refilter() - but there is another reason to refilter contacts other than changing filter data: custom code in the contact filter. In this case I have a one-way platform, and solidity is based on a combination of position and player input.

I've gotten this working correctly, by setting the position to -999, -999, then immediately setting it back to the correct value. However, it's not exactly elegant and I'd prefer not to lose existing contacts as there could be side effects from that. If there's a better solution, please let me know!

Thanks

Right. You want to force the broad-phase proxy to be added to the move buffer. There is not an API for this because no one has ever requested it before.

Thanks, ok here's my quick and dirty solution if anyone else needs it, it seems to work well. (Although, it seems the setpos(-999, -999) then setpos back worked just as well, and did not trigger any extra contact events as I feared.)

May have this quick solution implemented.

Still get errors on:


Errors:

'b2FixtureProxy' was not declared in this scope

'proxy' was not declared in this scope

'm_proxies' was not declared in this scope

I'm for the first time trying to change the source code and learn so any help as to where/how these would be declared would be much appreciated.