Bug Strange errors possiblty related to the particle system

We have some strange errors which seem to be linked to the particle systems.
We have about 200 gameObjects with emitters, I hope that it's not that which is causing the problem, because in the end we'll need much more than that.

Here are the errors we have:

Invalid memory pointer was detected in ThreadsafeLinearAllocator::Deallocate!

Internal: JobTempAlloc has allocations that are more than 4 frames old - this is not allowed and likely a leak

Assertion failed: Assertion failed on expression: 'p >= m_Blocks[blockIndex].ptr && p < m_Blocks[blockIndex].ptr + m_BlockSize'

Assertion failed on expression: 'all(sheetIndexNormalized >= 0.0f) && all(sheetIndexNormalized <= 1.0f)'

In order to produce the first 3 errors, we just have to add a particle system on a one of our prefab, and set its emission rate to 200.
The first 2 happen a lot more often than the third one.
We can also let it to 10, and activate the cast and receive shadows features.
In both cases the first 3 errors do happen.

The last one happen only (along with the first error) when we use a texture sheet animation.

These errors happen in the editor or in standalone.
The editor and the standalone version ultimately crashes when there are too many of these errors.
The errors are accumulating much faster in standalone than in the editor.


We are using Unity 2017.4.25f1, and the problem is still there with Unity 2017.4.40f1.
We cannot go for Unity 2018 or later because of the fact that the game is widely released and the networking code is quite old and would probably not survive going for a more recent Unity version.

 

I doubt that a bug report would help.
We cannot reproduce the problem on an empty project, and we cannot provide the project (anyway, thousands of assets and more than 150k line of code wouldn't really help).
In addition, as I said we cannot switch to Unity 2018 or newer, so we're not hoping for a fix from Unity (if it is a Unity bug, which isn't proven at this point).

But you could help us by providing some insights about these 2 assertions:

Assertion failed on expression: 'p >= m_Blocks[blockIndex].ptr && p < m_Blocks[blockIndex].ptr + m_BlockSize'

Assertion failed on expression: 'all(sheetIndexNormalized >= 0.0f) && all(sheetIndexNormalized <= 1.0f)'

If you could tell us in what situation these can happen it may help us isolate the problem.
With some luck we can disable some part of the particle system (or something else) and it will prevent the problem from coming back.
I'm pretty sure that the second one is about the 'Texture Sheet Animation' feature of the particle system, but that would help us to have a bit more information about the situations in which this assertion (and the other) can be triggered.

 

Thanks for your reply.

The particle system with the sheet animation is not to blame unfortunately.
I have removed the particle systems and replaced them with default ones, and the error still happen from time to time (a lot less often, presumably because the default emission rate is very low).

We also observed that the behaviour is quite different between a fast computer and a slow one (knowing that the scene is slow with both computers).
It looks to me as being a multi-threading issue in the particle system, which could probably explain the memory corruption and the fact that computers with different clocks and core numbers don't produce the same results.

 

We have finally found out the root of the problem.

Each particle system we have is on a child of a parent gameObject.
This parent also has another child, which has a light on it.
The light has its halo activated.

The problem was there: the halo.
If we remove the halo, then it works perfectly fine for hours, when it was spawning errors about every second and crashed in less than 30s.
I do not know if it is linked to the particle system or if it is the halo which is causing havoc in the memory and the particle system was only an innocent victim.

Anyway, I wanted you to know about the halo which seem to be having some trouble.