Wei Liu

> Thanks for the confidence, @rbradford! > > Looks like it still didn't pick up the change. I'll try to close/reopen, maybe it'll retrigger then. > > Thanks I restarted...

It seems that downloading the image takes a long time and causes the stage to be terminated.

Have you considered using `panic=abort`? Something to consider if you don't care about unwinding.

> @liuw Unfortunately the panic! handling does the correct thing. It's when a `Result::Err` is returned that the threads do not exit. Indeed. I misread. Cloud Hypervisor itself does not...

Hmm... Do you really need the root allocator to handle IRQ, PIO or PCI segment? They seem to be completely different from the memory address space.

> > @rbradford > > > What was the specific problem you were tying to solve? > > > > > > We want to pass a dtb to Cloud...

> The `test_windows_snapshot_restore` test now seems to be timing out after either the refactoring for enabling unified binary or (more likely in my opinion) the upgrade of the Azure VM...

> @liuw Do you think that the 5.15 change could also be affecting the hyperv KVM enlightenments? I cannot say for sure.

I ran this test locally for five times. It passed without any issues. My host is a D16s_v5. It is not as powerful as the ones in the CI but...

I rebooted my dev VM to D64s_v5. The test stopped working. And then I rebooted again into D16s_v5 again, it stopped working there too. This could be host specific. Unfortunately...