[Bug 30584] New: Epsitec games which need DirectX 3+ give black screen after winetricks ddr=gdi

http://bugs.winehq.org/show_bug.cgi?id=30584

Bug #: 30584
           Summary: Epsitec games which need DirectX 3+ give black screen
                    after winetricks ddr=gdi
           Product: Wine
           Version: 1.5.3
          Platform: x86
               URL: http://www.ceebot.com/blupi/load-e.php
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]
    Classification: Unclassified

Created attachment 40018
  --> http://bugs.winehq.org/attachment.cgi?id=40018
Crash log for Speedy Blupi demo near Wine 1.3.34

Several games by Epsitec which feature their Blupi character don't render
correctly if run after "winetricks ddr=gdi". This is a regression.

The games with this problem are Blupi at Home, Speedy Blupi and Planet
Blupi. Demos can be downloaded from http://www.ceebot.com/blupi/load-e.php

I tested Speedy Blupi, http://www.ceebot.com/download/blupi/speedy-e.exe

With the default opengl renderer the game works correctly. However after
"winetricks ddr=gdi", with or without virtual desktop mode the game just shows
a blank/black screen. The program is running, it just doesn't display anything.
(You can click to quit if you know where to click.)

I did a regression test, but unfortunately the result wasn't conclusive.

1.3.33 GOOD
 1.3.34 get a crash on running the program
 1.3.35 BAD, black screen

I bisected between 1.3.33 and 1.3.34 to find which commit caused the crash:

0f10ac93580d5a055f0b3f6e7c3d77738135f031 is the first bad commit
commit 0f10ac93580d5a055f0b3f6e7c3d77738135f031
Author: Henri Verbeet 
Date:   Mon Nov 28 21:30:57 2011 +0100

ddraw: Create the wined3d swapchain when setting the cooperative level.

:040000 040000 f8a8ea10d91985680fe1e29e2fb64aaf8bc59e2d
ab3f595c09537f3f18bd03bbe708c584f62e3952 M    dlls

I then bisected again between there and 1.3.35 (doing git bisect bad on black
screen/no crash), which gave this result:

c43aa1ada82e30168a373d584ad7ef62a5e04677 is the first bad commit
commit c43aa1ada82e30168a373d584ad7ef62a5e04677
Author: Henri Verbeet 
Date:   Sun Dec 4 15:50:14 2011 +0100

wined3d: Only unload the surface when we're using GL for rendering in
updateSurfaceDesc().

:040000 040000 2710baa1ce9dd8e207956e7802056a06ef874831
f2e62de0ba87f19f2705e53352affce2c88fd001 M    dlls

(Presumably commit c43aa1ada82e30168a373d584ad7ef62a5e04677 *isn't* bad, git
just said that because I had to reverse conditions for the second bisecting.)

I've attached console output when the program crashed, but that probably isn't
too relevant since it doesn't crash any more.

I guess the next step would be bisecting between commits
0f10ac93580d5a055f0b3f6e7c3d77738135f031 and
c43aa1ada82e30168a373d584ad7ef62a5e04677, but at each step reverting commit
0f10ac93580d5a055f0b3f6e7c3d77738135f031???