Possible thread safety issue with interaction handling about physx HOT 5 CLOSED

Hi Benoit,
Thanks a lot for this detailed bug report. We will look into it.

from physx.

Hello Benoit,
To follow up on your report: "issue 1" looks not thread safe indeed. For "issue 2" I can not yet see the clash though. From what I see:

lostTouchReports => Sc::activateInteractions()
Sc::Scene::notifyInteractionActivated() will only get called for pairs that are not eOVERLAP nor eMARKER (for example trigger pairs)

Sc::Scene::unregisterInteractions() on the other hand only runs for eOVERLAP or eMARKER pairs

as such, it does looks safe since concurrent activation happens for different types of pairs only. I seem to be missing something. What was the object/scene setup where you saw "issue 2" trigger unsafe concurrent access?

from physx.

Hi Michael,

Indeed issue 2 was a false positive on my side, my debug code was over-conservative and detected changes between interactions of different types.

from physx.

Thanks for the confirmation. Fixing "issue 1" on our side will likely be postponed to 5.3.0. Please use your proposed workaround in the meantime.

from physx.

Thank you for looking into it, we will do that in the meantime.

from physx.