Looking Glass host application erroring out when fullscreening game or changing resolution

[Celmor]

Required information

Host CPU: Intel i7-6700K
Host GPU: Intel HD Graphics 530
Guest GPU: GTX 980 Ti
Host Kernel version: 4.16.4-1-MANJARO
Host QEMU version: 2.11.1
Guest Windows version: Windows 10 RS3/1709

Changing ingame resolution dramatically (e.g. 720p->1440p), starting game with different ingame resolution than desktop, alt-tabbing out of game or into game can cause Host error and prompt to exit, refer to the corresponding log file for console output and see for its file name name for a short sumarry of what I did. Either way the issue can't be reproduced reliably.
For example:

• Changing ingame resolution causes DXGI's DuplicateOutput to produce error 80070005 and Clients video freezes but spice keeps working
• switching Monitor input to guest GPU to see its desktop
• closing looking-glass-host (or in my case creating dump file via task manager first)
• starting looking-glass-host
• alt-tabbing back to game
• switching monitor input to host
• noticing looking-glass-host errored again
• switching monitor input to guest GPU to see its desktop
• starting looking-glass-host
• alt-tabbing back to game
• switching monitor input to host
• this time no error

I couldn't isolate what causes looking-glass-host/capturer to error and fullscreening a game is a gamble for if LG will correspond badly or continue working. Changing just desktop resolution doesn't cause this issue. Desktop resolution and refresh rate is usually lower than ingame's setting then when switching to Game it often causes this issue but not exclusively.
Games I tested with are GTA5 and Rainbow Six Siege.

[Celmor]

Logs/Dumps hosted in this Cloud Drive folder:
https://mega.nz/#F!4p0hWJZD!rzjzSmtLs20Oiqb6bDrS8w

[gnif]

I have been trying to replicate this using GTA5 and I can't generate a fault. Can you please give at least one exact resolution switch (including refresh) that causes the crash.

Edit: Don't worry, just had a look at the files provided, you included dumps which should help track the issue down, I will get back to you when I know more.

[Celmor]

Took me a few attemps but now it's reproducable consistently (created about 10 more dumps in a row, put them in the cloud drive folder, all with d2018-05-13 time stamp).

• Restarted client and host application
• Doesn't matter if plugged in monitor output is switched to passed through GPU or host.
• Having client window be smaller than guest resolution seems to make it fault more easily, could be just placebo though.
• After faults mouse input through client (spice) didn't work anymore

1. I have set desktop resolution at 1152x864 @24Hz
2. I have set ingame resolution at 2560x1440p or 2554x1400 when not fullscreened @24Hz (tried DirectX 10.1 and 11)
3. Started a macro to repeatedly press alt+Enter
4. For a while everything worked as expected
5. After I un-fullscreened client window faults started happening (though I can't be sure this triggered it)
6. Each time I made a dump, copied host application output, switched back to game
7. After 1 or 2 times of pressing alt+Enter (fullscreening game, un-fullscreening game) host application faulted again
8. rinse and repeat

Now rebooting VM and reducing other factors, gonna reinstall Windows later as well (newer version) maybe this will make this issue go away, if not I removed more factors at least.

Edit: It seems having non-matching resolution ingame and for desktop just increases chance of a fault and after the first fault another one is more likely to happen shortly after till I reboot guest. Currently having updated graphics driver, 1080p@60Hz ingame and on desktop I still get the fault occasionally.
Next I would try to set-up my VM again as soon as I've got some spare time.
Sry for not being able to present an accurate reproducable scenario but it seems I'm alone with this issue on my setup anyway.

[gnif]

Right now I don't have a setup I can open these dumps in, but when I find some time I will get onto this one. Thanks for the detailed information.

[Celmor]

Reproduced issue on a new installation of Windows 10 with Redstone 4 update (1803). Dump and log are uploaded to same cloud directory but tagged accordingly.

[Celmor]

My current workaround is to have the host application restart automatically via a Powershell loop when exit code is 0 (which it does when the capturer exits), exit code is 1 if killed via task manager.

Do { echo "lookin glass started"; $ps = Start-Process D:\tools\looking-glass-host.exe -Wait -PassThru } While ($ps.ExitCode -eq 0)

[gnif]

This is likely fixed as of b23c780

[gnif]

Closing as this has seen no updates, assuming fixed

[8BallBomBom]

I had the changing resolution crash recently.
Seems if i change from 1440x900@75hz down to 1366x768@60hz then the host app crashes.
Can't seem to get the host app actually running again until the resolutions been changed back.

[gnif]

Is this on the latest master?

[8BallBomBom]

Yeah sadly :/
Was going to post about it but saw this one just closing.
How would i go about error log collecting?

[gnif]

I am sorry to hear that, however the issue you describe is not the same as the one that Celmor was having here. Please create a new issue for this.

As for information, please provide the output of the host application, to obtain it you need to run it with the -f switch. If it's a full crash and not a capture error I will need to try to replicate the problem on my end to debug.

[8BallBomBom]

No worries, will do that now :)