Wine 5.x can't create new Windows (same with Proton 5)

[Patola]

Hi all, this bug is killing me, I am eager to be able to update to Wine 5 but even with 5.0rc1 it won't create graphical windows! Same packages, in a very standard installation, Ubuntu 18.04.4 with nvidia driver 440.59 in an RTX 2070 SUPER, official wine PPAs, and wine 4.21 and earlier versions work very well, proton too.

I have opened a bug report for it, https://bugs.winehq.org/show_bug.cgi?id=48647 (all sorts of details there including logs from both wine 4.21 and 5.2) but it doesn't seem to be gathering much attention. Can anyone please help?

[dimesio]

The fact that you are the only one experiencing this suggests it's a problem with your system rather than a bug in Wine.

The "no driver could be loaded" message usually means a problem with your graphics drivers. Have you tried reinstalling them? I've also seen that message sometimes when there's a Wine process left running in the background, so make sure there are none.

It's not clear from the bug report whether the output is from a clean wineprefix. If not, you need to try that.

Beyond that, if you really believe this is a bug in Wine, you should run a regression test and post the results in the the bug report. https://wiki.winehq.org/Regression_Testing

[Patola]

The fact that you are the only one experiencing this suggests it's a problem with your system rather than a bug in Wine.

I know it seems like that. But no, I've checked everything in my system, every time I test it I checked whether there are wine or proton processes running, and I have upgraded and downgraded multiple times from multiple 4.x versions (staging, devel and stable) to multiple 5.x versions (staging, devel and stable) and the results are completely consistent: no 5.x version works, and all 4.x version work.

The "no driver could be loaded" message usually means a problem with your graphics drivers. Have you tried reinstalling them? I've also seen that message sometimes when there's a Wine process left running in the background, so make sure there are none.

Yes, I have changed the driver versions a few times to see if they were to blame. No change.

It's not clear from the bug report whether the output is from a clean wineprefix. If not, you need to try that.

Well, as I said in the bug report, for every test I would "rm -rf ~/.wine" and "wineboot -u" to create a new prefix, I think that it makes quite clear that the wineprefix is recreated from scratch each time.

Beyond that, if you really believe this is a bug in Wine, you should run a regression test and post the results in the the bug report. https://wiki.winehq.org/Regression_Testing

bisecting and recompiling wine multiple times would take too long... My current personal situation does not allow me much time to do that.

[Patola]

Made a few more tests. Also tried Proton-GE-4.20, it also runs successfully, but any Proton-GE-5.x fails the same way.

Changed NVIDIA drivers from 440.59 to 435.21 and rebooted, no change, the same things happen, 4.x and derivatives work and 5.x and derivatives don't.

Tried to run the commands (rm -rf ~/.wine ; wineboot -u ; winecfg ) as root and as other newly created user, no change: 4.x works, 5.x doesn't.

[dimesio]

Tried to run the commands (rm -rf ~/.wine ; wineboot -u ; winecfg ) as root

Never run Wine as root, and doing so can be the reason your graphics driver won't load.

[Patola]

Tried to run the commands (rm -rf ~/.wine ; wineboot -u ; winecfg ) as root

Never run Wine as root, and doing so can be the reason your graphics driver won't load.

I know. What I mean is that also tried running as root as one of the late tests -- it would have been useful to see if it was e.g. a permission or device access problem. But for all the other tests I used my own regular user.

Since it is a test I have performed days later than almost all other tests, in no way it could have been the cause of the driver not loading.

[Patola]

Ok, since I am not getting any new responses, and I am being doubted even in basic things, I took the time and performed the bisect from 4.21 to 5.0-rc1, and as predicted, some intermediary builds worked OK (created a window -- again, testing with winecfg) and others have shown the error (no window ever created). This is the final veredict of the bisect:

f2e5b8070776268912e1886d4516d7ddec6969fc is the first bad commit
commit f2e5b8070776268912e1886d4516d7ddec6969fc
Author: Alexandre Julliard <[email protected]>
Date: Thu Dec 12 19:19:12 2019 +0100

kernel32: Use the Get/SetComputerName functions from kernelbase.

Signed-off-by: Alexandre Julliard <[email protected]>

:040000 040000 ba60044892e47da777fd00316aee29f5779654af 1f9766be875cf9e2e241f4a3da0a685ed617ff93 M dlls

Can you PLEASE help me now? It took a lot of effort to get that time, I'll practically not sleep today.

[Patola]

BTW, before each test I DID verify if there were any running wine processes (there weren't) and I purge ~/.wine every time. Also I completely uninstalled wine from my system and just had the one compiled from git (in its directory, not make-installed) for testing. So, I took the necessary precautions to ensure a neutral testing scenario.

[Patola]

FIXED the problem based on what I read on the diffs. The problem is, I had a /etc/resolvconf/resolv.conf.d/head file which had the contents:
nameserver 192.168.15.5
search patola.org

As I had changed my internet, this nameserver does not work anymore, and also this domain is invalid for my intranet. Removed the file and restarted the internet, and then the window appears normally.