Temporarily loses video while upgrading from gutsy to hardy

[Nick Barcet]

On a MacBookPro Santa Rosa (nvidia 8600GTM), video is lost twice while the upgrade process from gutsy -> hardy.
This is blocking the upgrade process as while the video is off, questions cannot be seen by the user (and are therefore hard to reply to).

I was able to continue by looking at the at the content of /var/log/dist-upgrade/term.log via an ssh connection and anwser the questions blindly, but this would be a show stopper for most users.

[Nick Barcet (nijaba)]

Note that after rebooting, the default session had been change to xfce while I had always been using Gnome...

[Michael Vogt (mvo)]

Thanks for your bugreport.

I can see the following sources that might have caused this problem:
* problem with the X driver somehow
* gnome-screensaver locked the screen
* gnome-powermanager went into dpms

Does any of this sounds likely to you ? Did you run the upgrade on battery?

[Bryce Harrington (bryce)]

gnome-screensaver misbehaves sometimes, the most typical X issue related to screensavers is with a 3D screensaver turning on, when the system is running on a driver that has an OpenGL bug. So I could imagine after upgrading the screensaver package, maybe one screensaver triggered a bug on the old video driver, or vice versa, which may not occur with both packages upgraded; this seems rather a stretch though. Alternatively, I could also imagine a screensaver issue to get triggered if the upgrade process dropped you from using -nvidia to -nv; but I don't think we do that.

It sort of sounds a bit more like the old ddcprobing-on-intel-laptops bugs we got sorted out for Gutsy. Of course, that entire chunk of code is now entirely gone (no ddcprobing going on), so I would be surprised if there was ddcprobing at the root of this.

Like Timo said, though, it's possible the -nvidia driver is just doing something sneaky that we can't see. If this is the case, then there may already be other reports of similar behavior.

Going into dpms should normally never be a problem, obviously. If it did, and got stuck trying to come back, then that's going to be a handled as either a video driver or kernel issue. Since the underlying system was functioning (just without video), that seems to rule out a kernel issue.

I would love to know if this issue can be recreated (I know it's next to impossible to redo an upgrade though). It would be interesting to test switching to -nv or -vesa prior to upgrade, and see if that eliminates the problem, as that would seem to pin the problem on the -nvidia proprietary driver. I would also love to know more details about when in the process it happened - what exactly was being upgraded, whether the system had been left on its own (such that the screensaver might have kicked in), if the system was still being used (and if so, what programs were in use), etc. Nick, can you give more insights into this?

Lacking that, it would be useful to at least know the graphic card in use here, driver, and so on. Nick, can you please attach your Xorg.0.log file? Then if anyone else reports the same issue, we can compare PCI ID's; possibly there are certain nvidia cards that fail on something that happens during an upgrade (such as a dpms power save event).

[Nick Barcet (nijaba)]

I am about to rerun an upgrade on the same machine in another partition
and will take a full snapshot of /var/log if I can reproduce.

[Nick Barcet (nijaba)]

First tentative to reproduce the problem from a fresh gutsy failed with a 386 cd image (meaning that there was no issue).
Now trying again with a amd64 bit version as that is what is used on the system I originally upgraded.

[Nick Barcet (nijaba)]

Second tentative to reproduce the problem from a fresh gusty failed with the amd64 image. Only thing left I can think of is the fact that I've been using an external monitor (I did check at that time that video did not get redirected to the laptop screenç. Will try this again tomorrow...

[Richard Birnie (rbirnie-deactivatedaccount)]

Any further news on this or has the problem been resolved?

[Nick Barcet (nijaba)]

Marking this bug as invalid as I have been unable to reproduce it after multiple tries and running out of idea on what could have caused it.