GNOME Bugzilla – Bug 726882
Boxes cannot virtualize any GNOME-based OS
Last modified: 2016-03-31 13:22:07 UTC
I'm running Boxes 3.10.2 on Fedora 20. I've attempted to run several different GNOME systems in Boxes: Fedora 20, openSUSE 13.1, and Tanglu 1. Attempts to boot these systems inevitably end with the fail whale ("something has gone wrong and the system cannot recover"). I have noticed that the Fedora libvirt/qemu packages are updated frequently, so I often assume that there is just some temporary problem, but I've been trying again intermittently several times during the past year (on Fedora 18, 19, and now 20), and at this point I'm pretty sure that Boxes on Fedora has never been able to run any GNOME system, at least none of the ones I've tried. The only GNOME system I've tried that doesn't hit the fail whale is a QEMU image from GNOME Continuous, which stops at a kernel panic instead. I guess this is a bug rather than some known limitation, since I know GNOME Continuous images are supposed to work in Boxes, and that implies that other GNOME systems ought to as well.
Without any details, I'd blame QXL drivers. It breaks every other release. Please do file a bug on that. There is really nothing Boxes can do about this, except maybe to stop using QXL but then we don't get the cool features like autoresolution etc won't work unfortunately. :(
OK. I'm having a bit of trouble determining where to file a bug: would that be QEMU? Spice? X? (Or is there any other information I could provide that would be helpful?) (In reply to comment #1) > Without any details, I'd blame QXL drivers. It breaks every other release. > Please do file a bug on that. There is really nothing Boxes can do about this, > except maybe to stop using QXL but then we don't get the cool features like > autoresolution etc won't work unfortunately. :( That would be unfortunate, but it sounds loads better than no virtualization at all. Maybe that'd be best until things settle down.
(In reply to comment #2) > OK. I'm having a bit of trouble determining where to file a bug: would that be > QEMU? Spice? X? It could be a spice-gtk bug if not a QXL drivers (part of Spice project and in their bugzilla) so I'll start there. > (Or is there any other information I could provide that would be helpful?) I guess you'll want to provide the output of `virsh dumpxml YOUR_DOMAIN', where you can find the name of domain using 'virsh list --all'. > (In reply to comment #1) > > Without any details, I'd blame QXL drivers. It breaks every other release. > > Please do file a bug on that. There is really nothing Boxes can do about this, > > except maybe to stop using QXL but then we don't get the cool features like > > autoresolution etc won't work unfortunately. :( > > That would be unfortunate, but it sounds loads better than no virtualization at > all. Its working fine for me and other people too AFAIK so I wouldn't declare the situation so bad. Besides its not just GNOME that people use on their Linux VMs. > Maybe that'd be best until things settle down. I seriously doubt disabling QXL will push the developers to do better testing of their changes before pushing/merging.
Thanks, the Spice bug is https://bugs.freedesktop.org/show_bug.cgi?id=76521 (In reply to comment #3) > Its working fine for me and other people too AFAIK so I wouldn't declare the > situation so bad. Besides its not just GNOME that people use on their Linux > VMs. Hm, that's good to know.
(In reply to comment #1) > Without any details, I'd blame QXL drivers. It breaks every other release. > Please do file a bug on that. If you start randomly pushing the blame with a comment starting with 'Without any details', it would be much appreciated if you could make an effort to gather some of these details to improve the bug report, and even better, try to get info to support your claim.
This bug looks a lot like https://bugzilla.gnome.org/show_bug.cgi?id=720798 . This needs to be fixed in libvirt at some point ( https://bugzilla.redhat.com/show_bug.cgi?id=870071 ) but a workaround is in place in Boxes 3.11, and unfortunately not in 3.10.
(In reply to comment #6) > This bug looks a lot like https://bugzilla.gnome.org/show_bug.cgi?id=720798 . > This needs to be fixed in libvirt at some point ( > https://bugzilla.redhat.com/show_bug.cgi?id=870071 ) but a workaround is in > place in Boxes 3.11, and unfortunately not in 3.10. Yes, Boxes 3.11.92 works great. (It cannot run GNOME Continuous, but that's surely a completely different bug.) Thanks!
(In reply to comment #5) > (In reply to comment #1) > > Without any details, I'd blame QXL drivers. It breaks every other release. > > Please do file a bug on that. > > If you start randomly pushing the blame with a comment starting with 'Without > any details', Its not "random" at all. Firstly since if the issue is inside the VM, its very unlikey its a Boxes bug. Then we have a proven track record of these issues almost always some breakage in QXL (typically drivers). > it would be much appreciated if you could make an effort to > gather some of these details to improve the bug report, and even better, try to > get info to support your claim. I had even a long discussion with one of the QXL devs (Alon) on IRC about this and I pointed him to this bug. I really don't know which details exactly I'll need gathered except for the ones I mentioned in comment#3 (which as I pointed out is also guess). Keeping also in mind that I'm also not the one who reported this bug or is able to reproduce it on his end, I think you are asking for too much of me. In future, please refrain from such attacks on me or my choice of words and focus only on helping if you can. Bugzilla is not meant for this.
(In reply to comment #8) > Then we have a proven track record of these issues > almost always some breakage in QXL (typically drivers). I don't really have a recollection of such a 'track record' of brokenness with QXL. > > I really don't know which details exactly I'll need gathered except for the ones I mentioned in comment#3 (which as I pointed out is also guess). You asked the reporter to file a bug against QXL in comment #1 before any attempt at gathering more info about the bug. This is premature, and a waste of others people's time. Please try to gather at least a bit of data/evidence before closing the bug as NOTGNOME and asking to file a bug in other places. This is also nicer to the bug reporter...
(In reply to comment #9) > (In reply to comment #8) > > Then we have a proven track record of these issues > > almost always some breakage in QXL (typically drivers). > > I don't really have a recollection of such a 'track record' of brokenness with > QXL. Then you should get your facts straight before putting up comments here that are not helpful. FWIW, Alon agreed that QXL is really bad and he wasn't surprised at all about my claims of it breaking every now and then. gnome-continunous has been broken in Boxes for months recently and in the end the issue turned out be in QXL. (In reply to comment #9) > (In reply to comment #8) > > Then we have a proven track record of these issues > > almost always some breakage in QXL (typically drivers). > > I don't really have a recollection of such a 'track record' of brokenness with > QXL. > > > > > I really don't know which details exactly I'll need gathered except for the ones I mentioned in comment#3 (which as I pointed out is also guess). > > You asked the reporter to file a bug against QXL in comment #1 before any > attempt at gathering more info about the bug. This is premature, and a waste of > others people's time. I had enough facts already to conclude that this is very likely a Boxes bug. I can't be bothered gathering facts for other projects that are known to break all the time. They have their own bz for a reason. If you have time for that, please react more immediately to bug reports and help in solving them ASAP. No need to tell me how to do my job! Also, why do I have to explain all this to you? Did someone appointed you the police of bugzilla? Your behavior is extremely annoying sometimes and makes one question whether your contributions are worth dealing with you.
(In reply to comment #10) > I had enough facts already to conclude that this is very likely a Boxes bug. I > can't be bothered gathering facts for other projects that are known to break > all the time. I'm asking you to gather info to try to determine where the bug is coming from rather than make some guesses based on no evidence, and then wasting other people's time. Especially as in this case, I researched *Boxes* bugs in order to explain what the bug is. Or since I was fairly sure in the beginning this was this libvirt bug, should I have closed the qxl bug as NOTOURBUG, and sent the bug reporter to another trip to another bugzilla, where other people would have wasted time triaging it, ... ?
*** This bug has been marked as a duplicate of bug 720798 ***