Random err:winediag:nulldrv_CreateWindow errors related to windows and explorer

[Bruno Jesus]

I believe that since code freeze started (or maybe 1.7.55) a random
behavior was introduced related to these 2 messages:

err:winediag:nulldrv_CreateWindow Application tried to create a
window, but no driver could be loaded.
err:winediag:nulldrv_CreateWindow The explorer process failed to start.

Sometimes together there is also:
err:ole:apartment_createwindowifneeded CreateWindow failed with error 0

I'm not seeing this alone. at November 23 there was a mention to it in
bug https://bugs.winehq.org/show_bug.cgi?id=39384 and also in December
1 there was a mention at bug
https://bugs.winehq.org/show_bug.cgi?id=39057

The issue seems to manifest more when the computer is heavily loaded (cpu/ram).

I believe a regression test to be very hard since the issue is random, sorry.

Is anyone else affected by this? I'm still affected with today's git.

Best wishes,
Bruno

[Sebastian Lackner]

Hi Bruno,

I think this issue already exists since a long time - for example after
running the user32/winstation tests.
I personally have not noticed that the probability has increased, but like
with all random issues, even a
small delay / speedup can make it more or less likely to trigger such a
random issue. If you can
reproduce it reliable enough, it would be nice if you could try to do a
bisect (by running the corresponding
app/tests several times). The "bad" commit could be completely unrelated
though.

BTW: After Qian pointed out the problem with explorer.exe process leaks
some weeks ago, I've started
preparing patches for winstation and desktop cleanup. They also fix some
race-conditions, but in their
current form I would consider them too experimental to be merged. Also,
there are still a few unfixed
problems left.

Regards,
Sebastian

2015-12-01 20:20 GMT+01:00 Bruno Jesus :
> I believe that since code freeze started (or maybe 1.7.55) a random
> behavior was introduced related to these 2 messages:
>
> err:winediag:nulldrv_CreateWindow Application tried to create a
> window, but no driver could be loaded.
> err:winediag:nulldrv_CreateWindow The explorer process failed to start.
>
> Sometimes together there is also:
> err:ole:apartment_createwindowifneeded CreateWindow failed with error 0
>
> I'm not seeing this alone. at November 23 there was a mention to it in
> bug https://bugs.winehq.org/show_bug.cgi?id=39384 and also in December
> 1 there was a mention at bug
> https://bugs.winehq.org/show_bug.cgi?id=39057
>
> The issue seems to manifest more when the computer is heavily loaded
> (cpu/ram).
>
> I believe a regression test to be very hard since the issue is random,
> sorry.
>
> Is anyone else affected by this? I'm still affected with today's git.
>
> Best wishes,
> Bruno

[Bruno Jesus]

On Wed, Dec 2, 2015 at 9:54 PM, Sebastian Lackner  wrote:
> Hi Bruno,

Lali-ho!
> I think this issue already exists since a long time - for example after
> running the user32/winstation tests.
> I personally have not noticed that the probability has increased, but like
> with all random issues, even a
> small delay / speedup can make it more or less likely to trigger such a
> random issue. If you can
> reproduce it reliable enough, it would be nice if you could try to do a
> bisect (by running the corresponding
> app/tests several times). The "bad" commit could be completely unrelated
> though.

Thanks for the reply. I don't remember ever seeing those messages, I'm
trying to make it easier to reproduce here. What I have always
randomly experienced is other error (which also have a bug I can't
remember) and it is something like wine_kernel_boot: timeout, or
wine_kernel_init or something like that.
> BTW: After Qian pointed out the problem with explorer.exe process leaks some
> weeks ago, I've started
> preparing patches for winstation and desktop cleanup. They also fix some
> race-conditions, but in their
> current form I would consider them too experimental to be merged. Also,
> there are still a few unfixed
> problems left.

I'll wait for your patches them, after restarting the computer the
error seems to be happening less so I believe it is related to the
amount of memory used.

Bruno

[Austin English]

On Wed, Dec 2, 2015 at 8:32 PM, Bruno Jesus  wrote:
> On Wed, Dec 2, 2015 at 9:54 PM, Sebastian Lackner  wrote:
> > Hi Bruno,
>
> Lali-ho!
> > I think this issue already exists since a long time - for example after
> > running the user32/winstation tests.
> > I personally have not noticed that the probability has increased, but like
> > with all random issues, even a
> > small delay / speedup can make it more or less likely to trigger such a
> > random issue. If you can
> > reproduce it reliable enough, it would be nice if you could try to do a
> > bisect (by running the corresponding
> > app/tests several times). The "bad" commit could be completely unrelated
> > though.
>
> Thanks for the reply. I don't remember ever seeing those messages, I'm
> trying to make it easier to reproduce here. What I have always
> randomly experienced is other error (which also have a bug I can't
> remember) and it is something like wine_kernel_boot: timeout, or
> wine_kernel_init or something like that.

https://bugs.winehq.org/show_bug.cgi?id=15492

I was only able to reproduce in a VM on an (older) machine without
virtualization extensions IIRC (definitely slow in any case ;)).

[Alexandre Julliard]

Bruno Jesus  writes:
> On Wed, Dec 2, 2015 at 9:54 PM, Sebastian Lackner  wrote:
> > Hi Bruno,
>
> Lali-ho!
> > I think this issue already exists since a long time - for example after
> > running the user32/winstation tests.
> > I personally have not noticed that the probability has increased, but like
> > with all random issues, even a
> > small delay / speedup can make it more or less likely to trigger such a
> > random issue. If you can
> > reproduce it reliable enough, it would be nice if you could try to do a
> > bisect (by running the corresponding
> > app/tests several times). The "bad" commit could be completely unrelated
> > though.
>
> Thanks for the reply. I don't remember ever seeing those messages, I'm
> trying to make it easier to reproduce here.

I have never seen those messages either, and I run the tests a lot, so
I'm worried that it may be something new. If you could manage to
reproduce it, that would be great.

[Sebastian Lackner]

On 03.12.2015 03:32, Bruno Jesus wrote:
> I'll wait for your patches them, after restarting the computer the
> error seems to be happening less so I believe it is related to the
> amount of memory used.

Could you please test if you're still able to see this error in 1.8-rc3?
Two race-conditions have been fixed by my patches which were accepted today.

Regards,
Sebastian

[Qian Hong]

On Sat, Dec 5, 2015 at 1:34 AM, Sebastian Lackner  wrote:
> Could you please test if you're still able to see this error in 1.8-rc3?
> Two race-conditions have been fixed by my patches which were accepted today.

Today I found I can reproduce similar issue with lua for Windows and
nodejs for WIndows on Wine Staging 1.8-rc3-719-g3a955a5

err:winediag:nulldrv_CreateWindow Application tried to create a
window, but no driver could be loaded.
err:winediag:nulldrv_CreateWindow The explorer process failed to start.

Steps to reproduce:

The Lua way:

1. Install MSYS2 on Wine Staging: https://msys2.github.io/
(see https://github.com/fracting/wine-fracting/wiki/MSYS2-on-Wine for more info)

2. Start a MINGW32 shell on Wine:
$ cd ~/.wine/drive_c/msys32
$ wine start mingw32_shell.bat

3. Install lua for Windows inside MINGW32 shell using pacman:
$ pacman -Su mingw-w64-i686-lua

4. Start lua inside MINGW32 shell on Wine:
$ lua

The nodejs way is similar, just install mingw-w64-i686-nodejs

Environment:
Kernel version:
4.2.0-16-generic #19-Ubuntu SMP i686 GNU/Linux
$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 15.10
Release: 15.10
Codename: wily

I have native IE8 installed by winetricks in the same prefix, I'll
retest without native IE later.

Could anyone double confirm if you can reproduce? Thanks!

[Qian Hong]

On Tue, Dec 8, 2015 at 4:14 PM, Qian Hong  wrote:
> I have native IE8 installed by winetricks in the same prefix, I'll
> retest without native IE later.

Update: I can confirm again without native IE.

[Qian Hong]

Update:

I created a bug for it:

https://bugs.wine-staging.com/show_bug.cgi?id=626
Infinite "err:win:GetDesktopWindow failed to create desktop window"
hangs Linux system when type in lua shell / nodejs shell in MSYS2