Bug 478421

You''re talking about the Panel in the host OS disappearing when you enter Full Screen mode, not the guest OS, right? If so, that's expected. Full Screen Mode generally hides the panel. That's the main difference between the Full Screen the Maximized modes.

(In reply to Nate Graham from comment #1)
> You''re talking about the Panel in the host OS disappearing when you enter
> Full Screen mode, not the guest OS, right? If so, that's expected. Full
> Screen Mode generally hides the panel. That's the main difference between
> the Full Screen the Maximized modes.

No, the panel in the guest Fedora Rawhide Plasma 5.90.0 disappeared when entering full screen mode as shown in the recording I attached.

Can you check whether plasmashell process crashes?

(In reply to Vlad Zahorodnii from comment #3)
> Can you check whether plasmashell process crashes?

plasmashell didn't crash when this problem happened. I think that because the vertical resolution increased but the visible area Plasma was shown in didn't change, the panel was below the bottom of what was shown in the guest in full screen mode. This involved the spice-vdagent errors and how GNOME Boxes was using spice-vdagent. Thanks.

The video shows what I have reported in words in https://bugs.kde.org/show_bug.cgi?id=477985, and where progress is being made in getting it addressed.

The only difference is that the underlying kernel graphics driver here (most likely) is qxl, while for me it is vmwgfx. Both set the hotplug_mode_update DRM property.

IMHO, this issue here can be resolved as a duplicate of https://bugs.kde.org/show_bug.cgi?id=477985

*** This bug has been marked as a duplicate of bug 477985 ***

(In reply to Stefan Hoffmeister from comment #5)
> The video shows what I have reported in words in
> https://bugs.kde.org/show_bug.cgi?id=477985, and where progress is being
> made in getting it addressed.
> 
> The only difference is that the underlying kernel graphics driver here (most
> likely) is qxl, while for me it is vmwgfx. Both set the hotplug_mode_update
> DRM property.
> 
> IMHO, this issue here can be resolved as a duplicate of
> https://bugs.kde.org/show_bug.cgi?id=477985

The 6.7-rc5 kernel graphics driver was virtio_gpu in the GNOME Boxes VMs I used. Did you see the spice-vdagent errors I reported? I don't see them mentioned in https://bugs.kde.org/show_bug.cgi?id=477985

I see you raising two points in your bug report:
a) "The panel disappeared" inside the VM, i.e. with KDE
b) "spice ... trying to connect to mutter by dbus"

Point a) seems to be the major point, and, as a similarly affected user of KDE Plasma 6, I can confirm that resizing of a virtual screen will simply do that at the moment to KDE. https://bugs.kde.org/show_bug.cgi?id=477985 tracks that, and there is code (there) which "works for me" and which almost certainly will also work for you - best is for you to try the code and provide feedback :) https://wiki.archlinux.org/title/QEMU/Guest_graphics_acceleration seems to be ... exciting ...

Point b) would rather be a spice problem, not a KDE problem, I guess? Would the spice / GNOME Boxes developers be interested in that? I am not familiar with those virtualization components, sorry.

The spice-vdagent errors happened at the same time I entered full screen mode and the guest's panel disappeared. After I stopped spice-vdagentd with sudo systemctl stop spice-vdagentd.service spice-vdagentd.socket, the guest's panel was shown in full screen mode and the spice-vdagentd errors weren't shown. I took this to mean that the panel disappearing is due the same problem with how GNOME Boxes and spice-vdagent were interacting with Plasma 5.90.0. If you didn't see the spice-vdagent errors and stopping spice-vdagentd didn't allow the panel to be shown in full screen mode in VMware for you, then the causes of the problems we reported might not be the same. The problem I reported didn't happen using virt-manager, so there might be specific problems with each of GNOME Boxes and VMware. You can get guest GPU acceleration using the virgl mesa driver in GNOME Boxes by enabling the 3D acceleration switch in the VM's properties.

I guess https://invent.kde.org/plasma/kwin/-/merge_requests/4790 is the code you mentioned. Does that patch apply cleanly to kwin-5.90.0-2.fc40? Thanks.

I built kwin with the patch at https://invent.kde.org/plasma/kwin/-/merge_requests/4790 added to the current rawhide kwin repo with fedpkg --release rawhide mockbuild. I copied the kwin rpms into a GNOME Boxes QEMU/KVM VM using the image Fedora-KDE-Live-x86_64-Rawhide-20231214.n.0.iso with nautilus. I logged out and switched to VT2 in the VM. I updated to the kwin rpms I built. I restarted sddm. The guest's panel still disappeared when I entered full screen mode. Since the patch fixed the problem at https://bugs.kde.org/show_bug.cgi?id=477985, the problem I reported might be different.

I have installed gnome-boxes on Fedora 39 now, created Rawhide virtual machines for GNOME and KDE. 

Somehow, my setup does seem to work "nicely" in general (very janky), but I could see by running drm_info that on KDE kwin would never update to the resolution changes.

The alternative patch you may have seen on the MR mentioned on https://bugs.kde.org/show_bug.cgi?id=477985 should work for you, as virtio-gpu does not ever set the hotplug_mode_update DRM property.

I selected Code > Download Patches at https://invent.kde.org/plasma/kwin/-/merge_requests/4790 but the patch I downloaded was the same as the one I downloaded yesterday. Where do I get the alternative patch?

I built kwin 5.90.0 with the patch at https://invent.kde.org/plasma/kwin/-/merge_requests/4799 added. I updated the kwin rpms in GNOME Boxes VM using Fedora-KDE-Live-x86_64-Rawhide-20231216.n.0.iso as in Comment 10. The guest's panel was still shown in full screen mode after doing that, and Plasma covered the full screen without the black bands at the top and bottom that I reported. Thanks.