RFC XRender add support for dibsections in more color depths

[Roderick Colenbrander]

Hi,

For some weeks I have been working on moving more 2D rendering to
XRender. XRender has three advantages for Wine. First of all it allows
us to perform more rendering operations using X which we previously
did using a combination of software rendering and back-forth copying
between the Xserver. Second XRender offers depth conversion (this
gives us proper alpha support in Wine and allows us to bypass DIB
color conversion in some cases!). Third XRender brings us more
hardware acceleration and that way more performance.

The patch attached to this thread adds support for dibsections in
additional color depths. On Xservers running at 24-bit this patch
offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
for proper alpha support) Big parts of the Wine X11 Driver make
assumptions about having only two color depths around (1-bit and
screen_depth e.g. 24-bit). I would like to request that people apply
the current patch to latest git (in combination with the Office2007
patch which I posted to wine-patches). Please report any 2D rendering
issues you see and also mention possible Wine crashes due to X errors
like BadMatch. Try as much programs as possible.

Thanks in advance,
Roderick Colenbrander

[Roderick Colenbrander]

I just forgot to mention that next to adding proper Alpha support this
patch can dramatically improve performance in some programs.
Especially in some 2D games which suffer from DIB depth conversion
this patch helps a lot. E.g. C&C Tiberian Sun is a lot faster using
this patch WITHOUT any opengl (for Tiberian Sun xrender is better than
opengl). This is on Nvidia drivers which accelerate XRender well.
Geforce6/7 users make sure InitialPixmapPlacement is set to 2 for
optimal performance (this is the default value for Geforce8+ cards).

Roderick

On Thu, Jul 23, 2009 at 10:51 PM, Roderick
Colenbrander wrote:
> Hi,
>
> For some weeks I have been working on moving more 2D rendering to
> XRender. XRender has three advantages for Wine. First of all it allows
> us to perform more rendering operations using X which we previously
> did using a combination of software rendering and back-forth copying
> between the Xserver. Second XRender offers depth conversion (this
> gives us proper alpha support in Wine and allows us to bypass DIB
> color conversion in some cases!). Third XRender brings us more
> hardware acceleration and that way more performance.
>
> The patch attached to this thread adds support for dibsections in
> additional color depths. On Xservers running at 24-bit this patch
> offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> for proper alpha support) Big parts of the Wine X11 Driver make
> assumptions about having only two color depths around (1-bit and
> screen_depth e.g. 24-bit). I would like to request that people apply
> the current patch to latest git (in combination with the Office2007
> patch which I posted to wine-patches). Please report any 2D rendering
> issues you see and also mention possible Wine crashes due to X errors
> like BadMatch. Try as much programs as possible.
>
> Thanks in advance,
> Roderick Colenbrander

[Austin English]

On Thu, Jul 23, 2009 at 3:51 PM, Roderick
Colenbrander wrote:
> Hi,
>
> For some weeks I have been working on moving more 2D rendering to
> XRender. XRender has three advantages for Wine. First of all it allows
> us to perform more rendering operations using X which we previously
> did using a combination of software rendering and back-forth copying
> between the Xserver. Second XRender offers depth conversion (this
> gives us proper alpha support in Wine and allows us to bypass DIB
> color conversion in some cases!). Third XRender brings us more
> hardware acceleration and that way more performance.
>
> The patch attached to this thread adds support for dibsections in
> additional color depths. On Xservers running at 24-bit this patch
> offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> for proper alpha support) Big parts of the Wine X11 Driver make
> assumptions about having only two color depths around (1-bit and
> screen_depth e.g. 24-bit). I would like to request that people apply
> the current patch to latest git (in combination with the Office2007
> patch which I posted to wine-patches). Please report any 2D rendering
> issues you see and also mention possible Wine crashes due to X errors
> like BadMatch. Try as much programs as possible.
>
> Thanks in advance,
> Roderick Colenbrander

http://austinenglish.com/logs/appinstall-xrender/summary.txt
Launching gimp-splash.png-1.0 (RGB, 1 layer) 275x400 – GIMP failed. Test failed.

That test launches the gimp 2.6 trying to open it's splash image.

Terminal output:

fixme:bitblt:X11DRV_ClientSideDIBCopy potential optimization:
client-side color-index mode DIB copy
fixme:bitblt:X11DRV_ClientSideDIBCopy potential optimization:
client-side color-index mode DIB copy
fixme:bitblt:X11DRV_ClientSideDIBCopy potential optimization:
client-side color-index mode DIB copy
fixme:bitblt:X11DRV_ClientSideDIBCopy potential optimization:
client-side color-index mode DIB copy
fixme:bitblt:X11DRV_ClientSideDIBCopy potential optimization:
client-side color-index mode DIB copy
fixme:bitblt:X11DRV_ClientSideDIBCopy potential optimization:
client-side color-index mode DIB copy
fixme:bitblt:X11DRV_ClientSideDIBCopy potential optimization:
client-side color-index mode DIB copy
X Error of failed request:  BadMatch (invalid parameter attributes)
  Major opcode of failed request:  56 (X_ChangeGC)
  Serial number of failed request:  19348
  Current serial number in output stream:  19352

(file-png.exe:36): LibGimpBase-WARNING **: file-png.exe: gimp_wire_read(): error

(script-fu.exe:32): LibGimpBase-WARNING **: script-fu.exe:
gimp_wire_read(): error

[Anders Jonsson]

Roderick Colenbrander wrote:
> Hi,
>
> For some weeks I have been working on moving more 2D rendering to
> XRender. XRender has three advantages for Wine. First of all it allows
> us to perform more rendering operations using X which we previously
> did using a combination of software rendering and back-forth copying
> between the Xserver. Second XRender offers depth conversion (this
> gives us proper alpha support in Wine and allows us to bypass DIB
> color conversion in some cases!). Third XRender brings us more
> hardware acceleration and that way more performance.
>
> The patch attached to this thread adds support for dibsections in
> additional color depths. On Xservers running at 24-bit this patch
> offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> for proper alpha support) Big parts of the Wine X11 Driver make
> assumptions about having only two color depths around (1-bit and
> screen_depth e.g. 24-bit). I would like to request that people apply
> the current patch to latest git (in combination with the Office2007
> patch which I posted to wine-patches). Please report any 2D rendering
> issues you see and also mention possible Wine crashes due to X errors
> like BadMatch. Try as much programs as possible.
>
> Thanks in advance,
> Roderick Colenbrander

Hi,

tested the patch, and it makes some improvement on bug 10408:
http://bugs.winehq.org/show_bug.cgi?id=10408#c24

The drawback is that some text that is supposed to be white now is
colored. See comparison:
http://bugs2.winehq.org/attachment.cgi?id=22572

[Roderick Colenbrander]

I guess some of the font or dib rendering code is making some depth
assumptions somewhere. Is there an english version or demo version of
the app? I could take a look at it then.

Roderick

On Fri, Jul 24, 2009 at 11:06 AM, Anders
Jonsson wrote:
> Roderick Colenbrander wrote:
> > Hi,
> >
> > For some weeks I have been working on moving more 2D rendering to
> > XRender. XRender has three advantages for Wine. First of all it allows
> > us to perform more rendering operations using X which we previously
> > did using a combination of software rendering and back-forth copying
> > between the Xserver. Second XRender offers depth conversion (this
> > gives us proper alpha support in Wine and allows us to bypass DIB
> > color conversion in some cases!). Third XRender brings us more
> > hardware acceleration and that way more performance.
> >
> > The patch attached to this thread adds support for dibsections in
> > additional color depths. On Xservers running at 24-bit this patch
> > offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> > for proper alpha support) Big parts of the Wine X11 Driver make
> > assumptions about having only two color depths around (1-bit and
> > screen_depth e.g. 24-bit). I would like to request that people apply
> > the current patch to latest git (in combination with the Office2007
> > patch which I posted to wine-patches). Please report any 2D rendering
> > issues you see and also mention possible Wine crashes due to X errors
> > like BadMatch. Try as much programs as possible.
> >
> > Thanks in advance,
> > Roderick Colenbrander
>
> Hi,
>
> tested the patch, and it makes some improvement on bug 10408:
> http://bugs.winehq.org/show_bug.cgi?id=10408#c24
>
> The drawback is that some text that is supposed to be white now is
> colored. See comparison:
> http://bugs2.winehq.org/attachment.cgi?id=22572

[Anders Jonsson]

Roderick Colenbrander wrote:
> I guess some of the font or dib rendering code is making some depth
> assumptions somewhere. Is there an english version or demo version of
> the app? I could take a look at it then.
>
> Roderick
>
> On Fri, Jul 24, 2009 at 11:06 AM, Anders
> Jonsson wrote:
> > Roderick Colenbrander wrote:
> > > Hi,
> > >
> > > For some weeks I have been working on moving more 2D rendering to
> > > XRender. XRender has three advantages for Wine. First of all it allows
> > > us to perform more rendering operations using X which we previously
> > > did using a combination of software rendering and back-forth copying
> > > between the Xserver. Second XRender offers depth conversion (this
> > > gives us proper alpha support in Wine and allows us to bypass DIB
> > > color conversion in some cases!). Third XRender brings us more
> > > hardware acceleration and that way more performance.
> > >
> > > The patch attached to this thread adds support for dibsections in
> > > additional color depths. On Xservers running at 24-bit this patch
> > > offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> > > for proper alpha support) Big parts of the Wine X11 Driver make
> > > assumptions about having only two color depths around (1-bit and
> > > screen_depth e.g. 24-bit). I would like to request that people apply
> > > the current patch to latest git (in combination with the Office2007
> > > patch which I posted to wine-patches). Please report any 2D rendering
> > > issues you see and also mention possible Wine crashes due to X errors
> > > like BadMatch. Try as much programs as possible.
> > >
> > > Thanks in advance,
> > > Roderick Colenbrander
> >
> > Hi,
> >
> > tested the patch, and it makes some improvement on bug 10408:
> > http://bugs.winehq.org/show_bug.cgi?id=10408#c24
> >
> > The drawback is that some text that is supposed to be white now is
> > colored. See comparison:
> > http://bugs2.winehq.org/attachment.cgi?id=22572

Free demo:  http://www16.big.or.jp/~zun/data/soft/youmu_tr011.lzh
http://www16.big.or.jp/%7Ezun/data/soft/youmu_tr011.lzh
Run th07.exe to start the game.
Controls:  z (select) ,x (cancel) and arrow keys.

Choose "music room" from main menu to see the screen from the comparison
screenshot. No English version exists.

[Jesse Allen]

On Thu, Jul 23, 2009 at 2:51 PM, Roderick
Colenbrander wrote:
> Hi,
>
> For some weeks I have been working on moving more 2D rendering to
> XRender. XRender has three advantages for Wine. First of all it allows
> us to perform more rendering operations using X which we previously
> did using a combination of software rendering and back-forth copying
> between the Xserver. Second XRender offers depth conversion (this
> gives us proper alpha support in Wine and allows us to bypass DIB
> color conversion in some cases!). Third XRender brings us more
> hardware acceleration and that way more performance.
>
> The patch attached to this thread adds support for dibsections in
> additional color depths. On Xservers running at 24-bit this patch
> offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> for proper alpha support) Big parts of the Wine X11 Driver make
> assumptions about having only two color depths around (1-bit and
> screen_depth e.g. 24-bit). I would like to request that people apply
> the current patch to latest git (in combination with the Office2007
> patch which I posted to wine-patches). Please report any 2D rendering
> issues you see and also mention possible Wine crashes due to X errors
> like BadMatch. Try as much programs as possible.
>
> Thanks in advance,
> Roderick Colenbrander

Very nice Roderick,  it seems to be working on my system with
StarEdit.exe where it is faster.  It seems like it could be a good
solution.

Jesse

[Roderick Colenbrander]

Hi all,

This is an updated version of my patch. It should fix the XChangeGC
crash but the text color issue in that japanese game is still around.
I hope to receive some more feed back.

Roderick

On Sat, Jul 25, 2009 at 11:20 PM, Jesse Allen wrote:
> On Thu, Jul 23, 2009 at 2:51 PM, Roderick
> Colenbrander wrote:
> > Hi,
> >
> > For some weeks I have been working on moving more 2D rendering to
> > XRender. XRender has three advantages for Wine. First of all it allows
> > us to perform more rendering operations using X which we previously
> > did using a combination of software rendering and back-forth copying
> > between the Xserver. Second XRender offers depth conversion (this
> > gives us proper alpha support in Wine and allows us to bypass DIB
> > color conversion in some cases!). Third XRender brings us more
> > hardware acceleration and that way more performance.
> >
> > The patch attached to this thread adds support for dibsections in
> > additional color depths. On Xservers running at 24-bit this patch
> > offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> > for proper alpha support) Big parts of the Wine X11 Driver make
> > assumptions about having only two color depths around (1-bit and
> > screen_depth e.g. 24-bit). I would like to request that people apply
> > the current patch to latest git (in combination with the Office2007
> > patch which I posted to wine-patches). Please report any 2D rendering
> > issues you see and also mention possible Wine crashes due to X errors
> > like BadMatch. Try as much programs as possible.
> >
> > Thanks in advance,
> > Roderick Colenbrander
>
> Very nice Roderick,  it seems to be working on my system with
> StarEdit.exe where it is faster.  It seems like it could be a good
> solution.
>
> Jesse

[Matej Spindler]

Roderick Colenbrander wrote:
> Hi all,
>
> This is an updated version of my patch. It should fix the XChangeGC
> crash but the text color issue in that japanese game is still around.
> I hope to receive some more feed back.
>
> Roderick

Hi Roderick

With this patch Room Arranger crashes with:
> X Error of failed request:  BadMatch (invalid parameter attributes)
>   Major opcode of failed request:  56 (X_ChangeGC)
>   Serial number of failed request:  18560
>   Current serial number in output stream:  18919

Program can be downloaded from: http://www.roomarranger.com/

Matej

[Roderick Colenbrander]

Hi,

This version should hopefully fix your issues. I hope it also fixes
other gc issues .. I'm not sure if the patch is correct yet though.

Roderick

On Mon, Jul 27, 2009 at 11:08 AM, Matej
Spindler wrote:
> Roderick Colenbrander wrote:
> > Hi all,
> >
> > This is an updated version of my patch. It should fix the XChangeGC
> > crash but the text color issue in that japanese game is still around.
> > I hope to receive some more feed back.
> >
> > Roderick
>
> Hi Roderick
>
> With this patch Room Arranger crashes with:
> > X Error of failed request:  BadMatch (invalid parameter attributes)
> >  Major opcode of failed request:  56 (X_ChangeGC)
> >  Serial number of failed request:  18560
> >  Current serial number in output stream:  18919
>
> Program can be downloaded from: http://www.roomarranger.com/
>
> Matej

[Roderick Colenbrander]

Hi all,

This is an updated version of the patch which also addresses the font
color issue in that japanese game. I hope as much people can test it
so that I can fix other potential bugs (hopefully there are none
left).

Roderick

On Tue, Jul 28, 2009 at 12:25 AM, Roderick
Colenbrander wrote:
> Hi,
>
> This version should hopefully fix your issues. I hope it also fixes
> other gc issues .. I'm not sure if the patch is correct yet though.
>
> Roderick
>
> On Mon, Jul 27, 2009 at 11:08 AM, Matej
> Spindler wrote:
> > Roderick Colenbrander wrote:
> > > Hi all,
> > >
> > > This is an updated version of my patch. It should fix the XChangeGC
> > > crash but the text color issue in that japanese game is still around.
> > > I hope to receive some more feed back.
> > >
> > > Roderick
> >
> > Hi Roderick
> >
> > With this patch Room Arranger crashes with:
> > > X Error of failed request:  BadMatch (invalid parameter attributes)
> > >  Major opcode of failed request:  56 (X_ChangeGC)
> > >  Serial number of failed request:  18560
> > >  Current serial number in output stream:  18919
> >
> > Program can be downloaded from: http://www.roomarranger.com/
> >
> > Matej

[Reece Dunn]

2009/7/29 Roderick Colenbrander :
> Hi all,
>
> This is an updated version of the patch which also addresses the font
> color issue in that japanese game. I hope as much people can test it
> so that I can fix other potential bugs (hopefully there are none
> left).

I get:

"""
reece@rowan:/media/hdd1/pkg/github/wine$ git apply
~/Desktop/0001-Add-support-for-dibsections-in-more-formats-when-XRe.patch
/home/reece/Desktop/0001-Add-support-for-dibsections-in-more-formats-when-XRe.patch:158:
trailing whitespace.
        if( dib->dsBm.bmBitsPixel == wxr_formats_template[i].depth &&
/home/reece/Desktop/0001-Add-support-for-dibsections-in-more-formats-when-XRe.patch:300:
trailing whitespace.

warning: 2 lines add whitespace errors.
"""

when applying the patch.

Do I need to get any additional packages in addition to the ones
mentioned at http://wiki.winehq.org/Recommended_Packages, and that are
installed via Dan's helper script for Ubuntu?

When testing this, do I need anything over and above the NVidia binary drivers?

- Reece

[Roderick Colenbrander]

Hi Reece,

The only thing required is a git version of Wine and XRender libraries
on your system. After that any app can be tested. On decent drivers
like Nvidia you'll notice heavily improved 2D performance (e.g. useful
for theming). When using a Geforce6/7 try to set 'nvidia-setings -a
InitialPixmapPlacement=2' which makes sure all pixmaps are placed in
video memory.

Roderick

On Wed, Jul 29, 2009 at 10:09 PM, Reece Dunn wrote:
> 2009/7/29 Roderick Colenbrander :
> > Hi all,
> >
> > This is an updated version of the patch which also addresses the font
> > color issue in that japanese game. I hope as much people can test it
> > so that I can fix other potential bugs (hopefully there are none
> > left).
>
> I get:
>
> """
> reece@rowan:/media/hdd1/pkg/github/wine$ git apply
> ~/Desktop/0001-Add-support-for-dibsections-in-more-formats-when-XRe.patch
> /home/reece/Desktop/0001-Add-support-for-dibsections-in-more-formats-when-XRe.patch:158:
> trailing whitespace.
>        if( dib->dsBm.bmBitsPixel == wxr_formats_template[i].depth &&
> /home/reece/Desktop/0001-Add-support-for-dibsections-in-more-formats-when-XRe.patch:300:
> trailing whitespace.
>
> warning: 2 lines add whitespace errors.
> """
>
> when applying the patch.
>
> Do I need to get any additional packages in addition to the ones
> mentioned at http://wiki.winehq.org/Recommended_Packages, and that are
> installed via Dan's helper script for Ubuntu?
>
> When testing this, do I need anything over and above the NVidia binary drivers?
> - Reece

[Austin English]

On Wed, Jul 29, 2009 at 3:18 PM, Roderick
Colenbrander wrote:
> Hi Reece,
>
> The only thing required is a git version of Wine and XRender libraries
> on your system. After that any app can be tested. On decent drivers
> like Nvidia you'll notice heavily improved 2D performance (e.g. useful
> for theming). When using a Geforce6/7 try to set 'nvidia-setings -a
> InitialPixmapPlacement=2' which makes sure all pixmaps are placed in
> video memory.

Still fails the gimp test in appinstall. Everything else passes.

It's just this patch on top of git, right? No other patches?

[Austin English]

On Thu, Jul 23, 2009 at 3:51 PM, Roderick
Colenbrander wrote:
> Hi,
>
> For some weeks I have been working on moving more 2D rendering to
> XRender. XRender has three advantages for Wine. First of all it allows
> us to perform more rendering operations using X which we previously
> did using a combination of software rendering and back-forth copying
> between the Xserver. Second XRender offers depth conversion (this
> gives us proper alpha support in Wine and allows us to bypass DIB
> color conversion in some cases!). Third XRender brings us more
> hardware acceleration and that way more performance.
>
> The patch attached to this thread adds support for dibsections in
> additional color depths. On Xservers running at 24-bit this patch
> offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> for proper alpha support) Big parts of the Wine X11 Driver make
> assumptions about having only two color depths around (1-bit and
> screen_depth e.g. 24-bit). I would like to request that people apply
> the current patch to latest git (in combination with the Office2007
> patch which I posted to wine-patches). Please report any 2D rendering
> issues you see and also mention possible Wine crashes due to X errors
> like BadMatch. Try as much programs as possible.
>
> Thanks in advance,
> Roderick Colenbrander

Still fails the gimp test in appinstall. It's just this patch on top
of git, nothing else right?

[Roderick Colenbrander]

On Fri, Jul 31, 2009 at 9:25 AM, Austin English wrote:
> On Thu, Jul 23, 2009 at 3:51 PM, Roderick
> Colenbrander wrote:
> > Hi,
> >
> > For some weeks I have been working on moving more 2D rendering to
> > XRender. XRender has three advantages for Wine. First of all it allows
> > us to perform more rendering operations using X which we previously
> > did using a combination of software rendering and back-forth copying
> > between the Xserver. Second XRender offers depth conversion (this
> > gives us proper alpha support in Wine and allows us to bypass DIB
> > color conversion in some cases!). Third XRender brings us more
> > hardware acceleration and that way more performance.
> >
> > The patch attached to this thread adds support for dibsections in
> > additional color depths. On Xservers running at 24-bit this patch
> > offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> > for proper alpha support) Big parts of the Wine X11 Driver make
> > assumptions about having only two color depths around (1-bit and
> > screen_depth e.g. 24-bit). I would like to request that people apply
> > the current patch to latest git (in combination with the Office2007
> > patch which I posted to wine-patches). Please report any 2D rendering
> > issues you see and also mention possible Wine crashes due to X errors
> > like BadMatch. Try as much programs as possible.
> >
> > Thanks in advance,
> > Roderick Colenbrander
>
> Still fails the gimp test in appinstall. It's just this patch on top
> of git, nothing else right?
>
> --
> -Austin

Yeah it is just this patch on top of git. At what point does gimp fail
for you? I tried to open some images including the png gimp splash
screen and I'm seeing no issues. To be certain that you are using the
right patch I have reattached it (this time with the white space
trailing issues fixed). Just to be sure what display drivers are you
using and what videocard?

Roderick

[Austin English]

On Fri, Jul 31, 2009 at 3:51 AM, Roderick
Colenbrander wrote:
> On Fri, Jul 31, 2009 at 9:25 AM, Austin English wrote:
> > On Thu, Jul 23, 2009 at 3:51 PM, Roderick
> > Colenbrander wrote:
> > > Hi,
> > >
> > > For some weeks I have been working on moving more 2D rendering to
> > > XRender. XRender has three advantages for Wine. First of all it allows
> > > us to perform more rendering operations using X which we previously
> > > did using a combination of software rendering and back-forth copying
> > > between the Xserver. Second XRender offers depth conversion (this
> > > gives us proper alpha support in Wine and allows us to bypass DIB
> > > color conversion in some cases!). Third XRender brings us more
> > > hardware acceleration and that way more performance.
> > >
> > > The patch attached to this thread adds support for dibsections in
> > > additional color depths. On Xservers running at 24-bit this patch
> > > offers us dibsections in 1-bit/15-bit/16-bit/24-bit/32-bit. (32-bit is
> > > for proper alpha support) Big parts of the Wine X11 Driver make
> > > assumptions about having only two color depths around (1-bit and
> > > screen_depth e.g. 24-bit). I would like to request that people apply
> > > the current patch to latest git (in combination with the Office2007
> > > patch which I posted to wine-patches). Please report any 2D rendering
> > > issues you see and also mention possible Wine crashes due to X errors
> > > like BadMatch. Try as much programs as possible.
> > >
> > > Thanks in advance,
> > > Roderick Colenbrander
> >
> > Still fails the gimp test in appinstall. It's just this patch on top
> > of git, nothing else right?
> >
> > --
> > -Austin
>
> Yeah it is just this patch on top of git. At what point does gimp fail
> for you? I tried to open some images including the png gimp splash
> screen and I'm seeing no issues. To be certain that you are using the
> right patch I have reattached it (this time with the white space
> trailing issues fixed). Just to be sure what display drivers are you
> using and what videocard?

That one worked...perhaps I had some cruft in my tree :-/.

This line is new in the output:
(gimp-2.6.exe:52): Gimp-Display-CRITICAL **:
gimp_display_shell_draw_get_scaled_image_size_for_scale: assertion
`GIMP_IS_IMAGE (shell->display->image)' failed

but doesn't seem to be critical.

Card is a 9800 GeForce GTX, 190.18 drivers.

Running the suite again, it seems to work fine. Photoshop had that
failure we discussed on IRC, but it's a subtle race condition
unrelated to your patch. I'm working on that now.