Intel MacOS memory leak due to wgpu bug

[Aultus-defora]

Bevy version

0.10.0 or 0.10.1 or main by simply clonning the repo

[Optional] Relevant system information

MacOS Ventura 13.4.1 on Fusion Drive and Intel Mac

What you did

I run example of fox animation after clonning the repo, and also noticed same problem in a game I'm making, even in older versions using older bevy versions.

What went wrong

The Activity Moniter shows that memory usage growth over time, seemingly indefinitly. I am not even sure if it is a bevy bug or some dependency broke on MacOS specifically, as running earlier versions of bevy did not help to avoid the issue
P. S. after letting fox animation example run for ~30 minutes made consumed 5.1 GB of RAM
P. P. S. WASM build IS NOT affected.

[mockersf]

on main, on macos 13.3 with a apple m1 max, memory stays between 420 and 435MB over 10 minutes for me

[Aultus-defora]

Well I once let my project run and it ate a 1.1G of RAM. I have Fusion Drive and Intel if it makes any difference

[Aultus-defora]

Tried on macbook with Monterey 12.6.7 and 1.70 rustc abd 0.10 bevt and there is no memory leak. Seems fusion drive intel iMacs in particular are affected

[adals]

this is what i got if that help
my system is mac m1

[Aultus-defora]

Do you also have Ventura 13.4.1 (22F82) OS version?

[adals]

Do you also have Ventura 13.4.1 (22F82) OS version?

yes same

[Aultus-defora]

Updating to 13.4.1 (c) didn't help

[Aultus-defora]

Apparently the camera bundle is causing this. If I wanted to help debug it, what should I do?

[Aultus-defora]

So #6417 fixed it for most plantforms but not macs (with some macOS versions?) it seems like
P. S. this issue started after MacOS update so might be entierly different

[Aultus-defora]

Updating to Sonoma 14.0 did not fix the issue

[Aultus-defora]

leaks does not actually find this issue. It does not think whatever is growing the memory is an actual leak

[Aultus-defora]

So while running boids example of wgpu, similar memory growth occurs. Turns out this is not actually a bevy bug after all

[james7132]

So while running boids example of wgpu, similar memory growth occurs. Turns out this is not actually a bevy bug after all

Is there an associated issue filed on wgpu's end?

[Aultus-defora]

Is there an associated issue filed on wgpu's end?

I think there is an issue about 0.18-> 0.19 causing memory leaks; I am not sure if it is the same problem, considering even 0.14 of wgpu DOES have this problem with boids example. v0.13 does not.

[alice-i-cecile]

Reproduced on 0.14.0-rc3, but still only on Intel Mac.

Marking as blocked, since this is fundamentally a wgpu bug.