"Jamie McClymont" <jamie@kwiius.com> writes:

> Hello
>
> If anyone else has struggled with sepectrum-vm on AMD graphics, where
> the "child virtio-pci (virtio-wl)" process dies without a descriptive
> error, I believe the issue is fixed here:
> https://chromium-review.googlesource.com/c/chromiumos/platform/crosvm/+/2570453/2/seccomp/x86_64/video_device.policy
>
> It doesn't cleanly apply (I assume the seccomp policies have been shuffled between files), but --disable-sandbox is there as a functioning workaround :)

Thanks for the tip. :) I haven't tested with AMD graphics -- not sure if
anybody else will have.

> Does anyone have a branch where the chromium-os upstream-info is more
> up to date? I can have a go myself, but didn't want to duplicate
> effort :)

I don't think so.  Somebody was interested in updating to a more recent
Nixpkgs a few weeks ago but I don't think they got anywhere.  I was
planning on doing it once I got some stuff I have locally cleaned up and
committed.  If you wanted to have a go that'd be great, and I'll happily
provide any help you need. :)

In particular, if you were to run into any failures due to versions of
dependencies being too old or something, we might need to sync with
upstream Nixpkgs first.  If that's the case I can prioritise getting
that done unless you'd like to try that as well -- I'd expect it to be
easier than updating chromiumOSPackages because we don't heavily modify
Nixpkgs, just add to it.

> Thanks
> - Jamie McClymont
>
> --
>
> P.S. an introduction: Hi everyone!
>
> I have just started a new job where VM isolation is necessary, and
> have been hoping for a way to get integrated window management across
> VM boundaries - it looks like this is clearly the ideal approach, and
> I'm already a NixOS user, so am keen to contribute :)

Nice to make your acquaintance!  I'm Alyssa, and I'm the main person
behind Spectrum at the moment.  I hope you stick around. :)