Giovanni Mascellani

This is the backtrace I can observe: ``` #0 __pthread_kill_implementation (threadid=, signo=signo@entry=0x6, no_tid=no_tid@entry=0x0) at ./nptl/pthread_kill.c:44 #1 0x00007fbc35a9debf in __pthread_kill_internal (threadid=, signo=0x6) at ./nptl/pthread_kill.c:78 #2 0x00007fbc35a49c82 in __GI_raise (sig=sig@entry=0x6) at ../sysdeps/posix/raise.c:26...

> Does this mean you can work around the issue by passing 0 as the first element in `pValues` in `vkCmdPushConstants`? Yes, thought I don't really consider it a workaround....

> Unfortunately I'm not able to replicate this running your app on M3 macOS 14.6.1 with latest MoltenVK. Now I have an M3 Max myself too, and I cannot replicate...

I finally found a setup with an HTTP proxy instead of a SOCKS proxy that works for me, so this feature request is not an immediate need for me anymore....

I am hitting what appears to be the same issue for the vkd3d CI, whose macOS runner indeed runs inside a virtual machine, therefore with a paravirtualized device. An example...

It turns out that's pretty easy to reproduce in a virtualized environment. This code is enough to trigger the crash: ```swift import Metal for gpu in MTLCopyAllDevices() { let desc...