[Bug 15636] New: Colin McRae Rally 2.0 draws no background during race

http://bugs.winehq.org/show_bug.cgi?id=15636

Summary: Colin McRae Rally 2.0  draws no background during race
           Product: Wine
           Version: 1.1.6
          Platform: PC
               URL: http://cdn.gamezone.com/pub/gamezone/16/0/86/cmr2pcdemo.
                    zip
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Created an attachment (id=16677)
 --> (http://bugs.winehq.org/attachment.cgi?id=16677)
screenshot. The black area in the lower right comes from the snapshot tool:
This area was not redrawn.

During race, Colin McRae Rally 2.0 draws neither road nor other
background (see attached picture), i.e. the game is unplayable.

This affects both the commercial and the demo version, available from
http://downloads.gamezone.com/demosfiles/t1582.htm
http://cdn.gamezone.com/pub/gamezone/16/0/86/cmr2pcdemo.zip

This may be a driver issue. I'm using Ubuntu 8.04 with the Intel XOrg
video driver.

This could as well be a "Wine's DirectX6 not good enough" issue.
The in-game graphics selection menu displays no other "choice" than
"DirectDraw HAL", i.e. only DirectX6 is recognized.

Indeed, during install, a requester says "your machine is not ... for
DirectX7". The DirectX7a installer is distributed with the game.
So this could be a DirectX7 detection issue.

Alas, the game's configuration (Configuration/*.rcf) is a binary file, so I
cannot change the graphics driver.

Default trace output is attached. It contains a lot of
fixme:d3d_surface:surface_convert_format Cannot find a conversion function from
format WINED3DFMT_DXT5 to WINED3DFMT_A4R4G4B4
with XxYx16bit or
fixme:d3d_surface:surface_convert_format Cannot find a conversion function from
format WINED3DFMT_DXT5 to WINED3DFMT_A8R8G8B8
with a XxYx32bit resolution setting.
Could that be related to the missing background?

[Bug 15636] Colin McRae Rally 2.0 draws no background during race

http://bugs.winehq.org/show_bug.cgi?id=15636

--- Comment #1 from Jörg Höhle   2008-10-16 10:10:55 ---
Created an attachment (id=16678)
 --> (http://bugs.winehq.org/attachment.cgi?id=16678)
normal trace output

no particular component was traced

[Bug 15636] Colin McRae Rally 2.0 draws no background during race

http://bugs.winehq.org/show_bug.cgi?id=15636

--- Comment #2 from Jörg Höhle   2008-10-16 10:56:46 ---
CMR2 is quite sensitive to some graphics settings:
- With deactivated pixel shader in winecfg, the initial (intro/menu) screen
remains black.
  (Vertex shader makes no difference).
- Within the game, setting the car details from low to high causes the car not
be drawn (only its shadow is visible), as Daniel Szollosi-Nagy reported in
AppDB.

[Bug 15636] Colin McRae Rally 2.0 draws no background during race

http://bugs.winehq.org/show_bug.cgi?id=15636

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |download

[Bug 15636] Colin McRae Rally 2.0 draws no background during race

http://bugs.winehq.org/show_bug.cgi?id=15636

--- Comment #3 from Lei Zhang   2008-10-21 17:07:09 ---
What video card / driver version are you using? Can you try on another computer
with a nvidia / ati card and see if that makes a difference?

[Bug 15636] Colin McRae Rally 2.0 draws no background during race

http://bugs.winehq.org/show_bug.cgi?id=15636

--- Comment #4 from Jörg Höhle   2008-10-26 03:36:34 ---
The hardware is
Intel Corporation Mobile 915GM/GMS/910GML Express Graphics Controller
X.Org X Server 1.4.0.90
Release Date: 5 September 2007 (prerelease, used by Ubuntu 8.04)
I'm using the Intel Xorg driver on the laptop.
(II) Loading /usr/lib/xorg/modules/drivers//intel_drv.so
(II) Module intel: vendor="X.Org Foundation"
        compiled for 1.4.0.90, module version = 2.2.1
        Module class: X.Org Video Driver
        ABI class: X.Org Video Driver, version 2.0

Alas, I have no other computer to try on.

I've observed that versions of Wine around 1.0..1.1.2 manage to display *some*
background (i.e. grey rectangles) and will do regression testing. Also, with
these versions, the cars are rendered correctly.

[Bug 15636] Colin McRae Rally 2.0 draws no background during race

http://bugs.winehq.org/show_bug.cgi?id=15636

--- Comment #5 from Jörg Höhle   2008-11-14 04:41:40 ---
> Can you try on another computer

I still have no other HW, but I had hoped that an upgrade of the software from
Ubuntu Hardy 8.04 to Ubuntu Intrepid 8.10 might make a difference.
It does not (tested with 1.1.6), although XOrg & mesa is now much newer:

X.Org X Server 1.5.2
Release Date: 10 October 2008

[Bug 15636] Colin McRae Rally 2.0 draws no background during race

http://bugs.winehq.org/show_bug.cgi?id=15636

--- Comment #6 from Jörg Höhle   2008-12-10 12:52:12 ---
> I've observed that versions of Wine around 1.0..1.1.2 manage
> to display *some* background (i.e. grey rectangles)

More precisely, the change between 1.1.2 and 1.1.3
14b24058d69d73ebf6b70bc36c8aa62993351079 is first bad commit
commit 14b24058d69d73ebf6b70bc36c8aa62993351079
Author: Stefan Dösinger 
Date:   Mon Jul 28 11:41:02 2008 -0500
    wined3d: GL_ARB_fragment_program ffp implementation.
broke the rendering of grey building rectangles.

Very old versions of wine, e.g. 0.9.57, behave much like 1.1.3-1.1.10: no
background is drawn at all. Sadly, I cannot identify what commit enabled the
greyscale rectangles, as I too often get crashes during regression testing. It
happened somewhere between 0.9.57 and 1.0-rc1.

I thought that the missing background might be some hardware driver issue. So I
tried and switched to software rendering instead, using a recent enough version
of XOrg/Mesa, as found in Ubuntu Intrepid (previous tests were with Ubuntu
Hardy).

LIBGL_ALWAYS_SOFTWARE=yes changes nothing to rendering (except it's dog slow)
LIBGL_ALWAYS_INDIRECT=yes crashes the Xorg server (version 1.5.2, Ubuntu
Intrepid)

Do the identical results with LIBGL_ALWAYS_SOFTWARE eliminate the possibility
of a bug in the Intel XOrg driver?
Does *_ALWAYS_SOFTWARE act like a portable, hardware-independent display driver
so one could "prove" bugs must be somewhere else?

[Bug 15636] Colin McRae Rally 2.0 draws no background during race

http://bugs.winehq.org/show_bug.cgi?id=15636

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |regression

[Bug 15636] Colin McRae Rally 2.0 draws no background during race - missing DXT/S3TC decompression

http://bugs.winehq.org/show_bug.cgi?id=15636

Jörg Höhle  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID
            Summary|Colin McRae Rally 2.0  draws|Colin McRae Rally 2.0  draws
                   |no background during race   |no background during race -
                   |                            |missing DXT/S3TC
                   |                            |decompression

--- Comment #7 from Jörg Höhle   2009-04-20 03:35:53 ---
> This may be a driver issue.

Indeed it is: DXT/S3TC in HW is (partly) lacking in the Intel GPU.

Surprisingly,
http://homepage.hispeed.ch/rscheidegger/dri_experimental/s3tc_index.html
says: "Currently, s3tc is supported on all graphic cards which use the radeon,
r200, i830 and i915 driver".

Alas, OpenGL does not announce the S3TC capability because the specification
requires both decompression *and* compression.  From what I understood, the
i915 can only decompress in HW.  Luckily, that is enough for many applications.

Here's my ~/.drirc (or /etc/drirc) to force OpenGL to declare the S3TC
capability.  Then glxinfo reports it.  The file produces immediate effect,
there's no need to restart the X server.
<driconf>
    <device screen="0" driver="i915">
        <application name="Default">
            <option name="force_s3tc_enable" value="true" />
        </application>
    </device>
</driconf>

With that, Colin McRae 2.0 becomes playable, as everything is drawn correctly.
The fixmes about DXT vanish (tested with wine-1.1.16 or -1.1.18 with Ubuntu
Hardy and Intrepid).

Sadly, like in TORCS, the image freezes from time to time during race, exactly
like satellite TV on bad weather.  Sound still plays but motion is stopped,
then resumes after half a second.  All objects are redrawn to their new
position, as the application never stopped.  Meanwhile, the car bumped into a
fence as you lost visual control of it.  But that's another issue.

Changing resolution to invalid.

[Bug 15636] Colin McRae Rally 2.0 draws no background during race - missing DXT/S3TC decompression

http://bugs.winehq.org/show_bug.cgi?id=15636

Jeff Zaroyko  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #8 from Jeff Zaroyko   2009-04-20 04:18:44 ---
closing

[Bug 15636] Colin McRae Rally 2.0 draws no background during race - missing DXT/S3TC decompression

http://bugs.winehq.org/show_bug.cgi?id=15636

Ben Klein  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #9 from Ben Klein   2009-04-20 21:34:26 ---
(In reply to comment #7)
> > This may be a driver issue.
>
> Indeed it is: DXT/S3TC in HW is (partly) lacking in the Intel GPU.
>
> Surprisingly,
> http://homepage.hispeed.ch/rscheidegger/dri_experimental/s3tc_index.html
> says: "Currently, s3tc is supported on all graphic cards which use the radeon,
> r200, i830 and i915 driver".
>
> Alas, OpenGL does not announce the S3TC capability because the specification
> requires both decompression *and* compression.  From what I understood, the
> i915 can only decompress in HW.  Luckily, that is enough for many applications.
>
> Here's my ~/.drirc (or /etc/drirc) to force OpenGL to declare the S3TC
> capability.  Then glxinfo reports it.  The file produces immediate effect,
> there's no need to restart the X server.
>
> <driconf>
>     <device screen="0" driver="i915">
>         <application name="Default">
>             <option name="force_s3tc_enable" value="true" />
>         </application>
>     </device>
> </driconf>

Umm ... if you're forcing S3TC support on, and your hardware doesn't support
it, it shouldn't change the behaviour of your app (you should still be missing
textures).
> With that, Colin McRae 2.0 becomes playable, as everything is drawn correctly.
> The fixmes about DXT vanish (tested with wine-1.1.16 or -1.1.18 with Ubuntu
> Hardy and Intrepid).

There's something else going on here. Possibly the application is asking the
driver for S3TC, which is not supported by the driver (glxinfo | grep -i s3tc)
and as a result, asking DirectX to decompress it into ARGB, which is supported
by the drivers. Or possibly the application (or Wine) is querying the wrong GL
extension when testing for S3TC support.

I think your glxinfo would be interesting here. If they're different, once with
the driconf hack above, and once without.
> Sadly, like in TORCS, the image freezes from time to time during race, exactly
> like satellite TV on bad weather.  Sound still plays but motion is stopped,
> then resumes after half a second.  All objects are redrawn to their new
> position, as the application never stopped.  Meanwhile, the car bumped into a
> fence as you lost visual control of it.  But that's another issue.
>
> Changing resolution to invalid.

I don't think this bug is resolved. The symptoms are the same as in bug 14939
(missing DXTC/ARGB conversion). The proposed solution in driconf is hacky and,
if S3TC really is missing in the hardware, could cause breakages. It also only
works on supported intel and radeon drivers, not on nvidia (and bug 14939 is
reported with an nvidia card).

[Bug 15636] Colin McRae Rally 2.0 draws no background during race - missing DXT/S3TC decompression

http://bugs.winehq.org/show_bug.cgi?id=15636

--- Comment #10 from Jörg Höhle   2009-04-21 03:56:53 ---
> if you're forcing S3TC support on, and your hardware doesn't support
> it, it shouldn't change the behaviour of your app

AFAIK, my hardware supports decompression, but not compression.  The former is
apparently all that's needed for this particular application.  The referred URI
further reads: "you will get errors if applications try to use
online-compression (like QuakeIII does)".

BTW, somewhere I read that it could be that MS-Windows' D3D drivers would
report the S3TC capability even on my hardware.  That would make CMR2 run
there.  So the force_s3tc_enable option gets wine closer to MS-Windows'
behaviour.  It's not to be frowned upon.  Material for wine-wiki?
> I think your glxinfo would be interesting here. If they're different

As I said, they are.  A single additional capability is output, something
xyz_s3tc (from memory, I cannot check right now).  Must be
ext_texture_compression_s3tc as it's this one wine checks in
dlls/wined3d/directx.c.
> There's something else going on here.

Well, what's puzzled me is the following quote from
http://dri.freedesktop.org/wiki/S3TC
"force_s3tc_enable... This may be of use for some games which require S3TC or
don't use the ARB_texture_compression extension correctly."
So there would be some correct use of ARB_texture_compression (which wine
should thus implement) that does not depend on this option?!?  I could not find
further information on this.

The two pages about S3TC are not clear enough, IMHO.  It took me some time
after initially reading them to combine that it would benefit CMR2.
> bug 14939 is reported with an nvidia card

Yes, much to my surprise.  NVidia is probably supported by CMR2 (I will look up
the package cover).  Does the CMR2Demo (URL in top template) work for you with
NVidia?

[Bug 15636] Colin McRae Rally 2.0 draws no background during race - missing DXT/S3TC decompression

http://bugs.winehq.org/show_bug.cgi?id=15636

--- Comment #11 from Jörg Höhle   2009-04-23 07:42:24 ---
CMR2's package cover (published 2000 approx.) lists these supported HW devices:
3Dfx, Voodoo2, Voodoo Banshee, Voodoo3, Voodoo5, 3DLabs Permedia 3, ATI Rage
Pro, Rage 128, Intel i740, Matrox G100, G200, G400, NVidia Riva 128, Riva TNT,
Riva TNT2, GeForce 256, GeForce 2, S3 Savage 4, Savage 2000.
3D with 8MB RAM, 233MHz, D3DX7.0a.

I don't know whether they all support(ed) S3TC or whether CMR2 may contain
alternate code path not using DXT textures. Note that i740 is supported, an
ancestor of i915.

(Although I now have s3tc enabled in /etc/drirc, it's clear to me it's not a
completely satisfying solution.  E.g. from what Scheidegger writes, I would get
into trouble if I wanted to use QuakeIII or DoomIII.  But that's as far as one
can currently go without too much hassle.)

[Bug 15636] Colin McRae Rally 2.0 draws no background during race - missing DXT/S3TC decompression

http://bugs.winehq.org/show_bug.cgi?id=15636

--- Comment #12 from Jörg Höhle   2009-07-13 09:23:43 ---
Chris Dragan's D3Dcaps database http://zp.amsnet.pl/cdragan/d3dcaps.html
shows that among the cards supported (comment #11) by Colin Mac Rae 2.0, the
following ones do not support DXT1 .. DXT5.  Obviously, the app is able to take
different code paths, but I don't know how to trigger it (will try PCI ID).
- Voodoo3
- NVidia Riva TNT2
- ATI Rage Pro