[Bug 15644] New: Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Summary: Morrowind: Waterplane/sprite flickering with ORM=fbo
           Product: Wine
           Version: 1.1.4
          Platform: PC-x86-64
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: directx-d3d
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Hi there,

this is a regression introduces between wine-1.1.3 and wine-1.1.4

It affects Morrowind version 1.2.0722, when OffscreenRenderingMode is set to
fbo. GLSL settings don't seem to have any effect on the issue.

Console output doesn't seem to be related to the problem. I get a lot of those
though on my Nvidia NV35:
fixme:d3d_shader:shader_glsl_deselect_depth_blt >>>>>>>>>>>>>>>>>
GL_INVALID_OPERATION (0x502) from glUseProgramObjectARB @ glsl_shader.c / 3450

I'm currently bisecting and these msgs also appear when the state is good, so
they're probably not really related.

How does the issue look like:
The Waterplane (pixel shader water is on) and the sprites flicker a lot when
moving through the terrain. Sprites also does include parts of the trees and
e.g. lillies in the various pools outside of Seyda Neen.

Interesting observation:
When pressing ESC, which does bring up the ingame menu (Save, Load, Exit - you
get the idea) the graphics are perfect. Always.

Posting bisect info later.

Greets,
Tobias

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

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

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

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

--- Comment #1 from Tobias Jakobi   2008-10-16 19:14:56 ---
d37c6fc0a94b3ac3b4cce58afeafbdf225107b80 is first bad commit
commit d37c6fc0a94b3ac3b4cce58afeafbdf225107b80
Author: Henri Verbeet 
Date:   Wed Sep 3 16:25:44 2008 +0200

wined3d: Depth stencil tracking depends on the value of
This->render_offscreen.

:040000 040000 3541a357c05d4e97e4f89f84ccb25b94dcfbf717
2e7eabd89f0ace15bc99bb13a3e7247ccb2bf56a M      dlls

URL:
http://source.winehq.org/git/wine.git/?a=commit;h=d37c6fc0a94b3ac3b4cce58afe...

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #2 from Austin English   2008-10-16 21:44:07 ---
A +d3d,+d3d_draw trace would help.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

H. Verbeet  changed:

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

--- Comment #3 from H. Verbeet   2008-10-17 01:47:51 ---
I don't have the game, I'm afraid. Could you figure out where/why the GL error
is generated? It might not be the cause of the problem, but it would be nice to
eliminate it, for a start. Does the isInDraw flag have any effect on the
behaviour? Please make sure that reverting the patch fixes the issue in current
git.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #4 from H. Verbeet   2008-10-17 01:50:45 ---
Created an attachment (id=16696)
 --> (http://bugs.winehq.org/attachment.cgi?id=16696)
patch

Does this patch make any difference?

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #5 from Tobias Jakobi   2008-10-17 02:52:00 ---
(In reply to comment #2)
> A +d3d,+d3d_draw trace would help.

I can do this later.

(In reply to comment #3)
> I don't have the game, I'm afraid. Could you figure out where/why the GL error
> is generated?

You mean by using a debugger and dumping the callstack, or something like that?
> It might not be the cause of the problem, but it would be nice to
> eliminate it, for a start.

The thing is that this message is not only generated with Morrowind, but also
e.g. with Max Payne 2 or F.E.A.R. I also see a lot of FBO issues there, but
nothing like this flickering problem.
> Does the isInDraw flag have any effect on the
> behaviour? Please make sure that reverting the patch fixes the issue in current
> git.

Will do! (but also later)

(In reply to comment #4)
> Created an attachment (id=16696)

--> (http://bugs.winehq.org/attachment.cgi?id=16696) [details]
> patch
>
> Does this patch make any difference?

Going to check later.

Greets,
Tobias

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #6 from H. Verbeet   2008-10-17 03:19:06 ---
(In reply to comment #5)
> (In reply to comment #3)
> > I don't have the game, I'm afraid. Could you figure out where/why the GL error
> > is generated?
>
> You mean by using a debugger and dumping the callstack, or something like that?

checkGLcall() is just a wrapper around glGetError(). That means it will dump
any GL errors generated since the last time checkGLcall() was called. So while
the error could be caused by the glUseProgramObjectARB() call, it's rather
unlikely. shader_glsl_deselect_depth_blt() is called from depth_blt() in
drawprim.c. depth_blt() is called from surface_load_ds_location() in surface.c,
but is preceeded by a checkGLcall() in both cases, so the GL call that
generates the error happens somewhere between the start and end of depth_blt().
Just adding a few more checkGLcall()'s to depth_blt() should be enough to
narrow it down to a specific call. Note that the error could also be generated
by shader_glsl_select_depth_blt() though, since that's called by depth_blt().

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Jan Buecken  changed:

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

--- Comment #7 from Jan Buecken   2008-10-17 06:22:50 ---
(In reply to comment #4)
> Created an attachment (id=16696)

--> (http://bugs.winehq.org/attachment.cgi?id=16696) [details]
> patch
>
> Does this patch make any difference?

no, not for me and git 30-45 min ago on ati

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #8 from Tobias Jakobi   2008-10-17 14:29:19 ---
OK, just tried the patch and it does indeed change something.

The waterplane and sprites are now gone completly when looking straight
forward.
One way to bring them back (again flickering) is to look at the ground.

The 'open menu' technique still works and all looks good there.

Patch was applied on top of latest git master.

Next: Attaching trace like Austin English proposed (for this I'm going to
remove the patch again).

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #9 from Tobias Jakobi   2008-10-17 15:49:17 ---
d3d, d3d_draw trace of Morrowind:
http://www.math.uni-bielefeld.de/~tjakobi/wine/morrowind_debug.log.bz2

Had to bzip2 this log, because it go rather huge. Shows again that the engine
sucks (uses too much API calls).

Note that I couldn't reproduce the flickering while the trace logging was on
(the game was really slow with it). Instead of flickering the waterplane and
sprites were never rendered at all (even not when looking at the ground).

The 'game menu' technique again worked here again.

Greets,
Tobias

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #10 from Tobias Jakobi   2008-10-17 16:59:46 ---
Bisection verification:

Reverting the commit on top of current git master produces the same behaviour
like the patch by Henri.

Checking out the tree with latest commit
d37c6fc0a94b3ac3b4cce58afeafbdf225107b80 produces:
Issue there!

Reverting the commit on top of that:
Issue gone!

So something happens between d37c6fc0a94b3ac3b4cce58afeafbdf225107b80 and git
master tip that changes behaviour of the code in such a way, that reverting the
commit does not fix the problem.

Next thing: Check isInDraw flag and the GL error, that is produced.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #11 from Tobias Jakobi   2008-10-18 08:27:26 ---
Hi there,

I have sort of isolated the problem.

It's the glBindTexture(GL_TEXTURE_2D, texture) call in depth_blt that fails.
According to the OpenGL reference there are two possible ways to trigger such a
failure:
(i) GL_INVALID_OPERATION is generated if texture was previously created with a
target that doesn't match that of target.
(ii) GL_INVALID_OPERATION is generated if glBindTexture is executed between the
execution of glBegin and the corresponding execution of glEnd.

I highly doubt that (ii) is the cause, since that would also cause some other
OpenGL calls to fail as well. Leaves us with (i).

I'm currently trying to figure out if GL_TEXTURE_2D is not always the correct
target to use with the supplied texture id.

Note that (as Henri stated) the depth_blt function gets called from
surface_load_ds_location. There are two calls there, only the second one (the
"Copying depth texture to onscreen depth buffer" one) fails. The other one
doesn't seem to cause any trouble.

Major difference: The first one is called with device->depth_blt_texture as
texture id, the second one with This->glDescription.textureName.

AFAIK we can be sure that device->depth_blt_texture has a 2D target if it has
been created inside surface_load_ds_location by the glGenTextures(1,
&device->depth_blt_texture) call. That's because the first target where a tex
id is bound to specifies the target of the texture object (that's what the
OpenGL reference tells me).

I dunno where else it could be created, but at least for now I don't see why we
can ALWAYS assume that it has target GL_TEXTURE_2D.

Can anyone comment on that?

The same applies for This->glDescription.textureName, as it may have a
different target.

That's my research so far. I'm currently trying to figure out the correct
target for the texture object passed to depth_blt and also pass this info to
depth_blt as well, so it can be correctly handled there. We'll see if that
fixes the problem.

Greets,
Tobias

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #12 from H. Verbeet   2008-10-18 09:22:44 ---
Yes, "texture" can potentially be a cube or rect texture. depth_blt_texture
should always be a GL_TEXURE_2D, although that might actually be a problem in
some cases if the card doesn't support NPOT textures (ie, might cause a
software fallback). This->glDescription.textureName depends on the type of the
surfaces used for the depth stencil, and can be either GL_TEXTURE_2D,
GL_TEXTURE_RECT, or a cube face.

The other issue that's less obvious, is that the depth blit shader needs to be
modified as well. It currently always uses texture2D to sample (in case of
GLSL), but that's incorrect if the texture is a cube or rect texture, of
course.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #13 from Tobias Jakobi   2008-10-18 09:32:11 ---
Created an attachment (id=16719)
 --> (http://bugs.winehq.org/attachment.cgi?id=16719)
[wined3d] use correct texture target in depth_blt

Hi there,

this patch fixes the GL errors mentioned in this bugreport. It does NOT fix the
flickering/disappearing issue.

It's (the patch) currently not going into wine because the issue lies much
deeper.

For the GLSL path create_glsl_blt_shader() is flawed, because it assumes a 2D
target (so it uses sample2D, texture2D, and so on) which is not always the
case.

The attached patch currently handles normal 2d, cubemap and rect targets.
According to Henri Verbeet the other targes (1d and 3d) don't appear in
combination with depthstencil stuff.

So code must also be fixed in create_glsl_blt_shader (probably creating two
more functions which handle cubemaps and rects). The same would have to be done
for the ARB shader path.

Introduces more problems:
- Selection of the correct face for cubemap textures
- Non-normalized texcoords for rect textures (problem when generating texcoords
in vertex shader)

Note: The information from above is a summary of what Henri told me on IRC :)

Greets,
Tobias

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #16719|0                           |1
        is obsolete|                            |

--- Comment #14 from Tobias Jakobi   2008-10-18 10:03:10 ---
Created an attachment (id=16720)
 --> (http://bugs.winehq.org/attachment.cgi?id=16720)
[wined3d] use correct texture target in depth_blt

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #16720|0                           |1
        is obsolete|                            |

--- Comment #15 from Tobias Jakobi   2008-10-18 10:07:29 ---
Created an attachment (id=16721)
 --> (http://bugs.winehq.org/attachment.cgi?id=16721)
[wined3d] use correct texture target in depth_blt

now with less debug code...

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #16721|0                           |1
        is obsolete|                            |

--- Comment #16 from Tobias Jakobi   2008-10-18 10:18:16 ---
Created an attachment (id=16722)
 --> (http://bugs.winehq.org/attachment.cgi?id=16722)
[wined3d] use correct texture target in depth_blt

Remove junk, now it should compile again :)

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|patch                       |

--- Comment #17 from Tobias Jakobi   2008-10-18 17:14:23 ---
@Austin: I recommend not adding the patch keyword.

We don't know yet if the GL error has something to do with the actual issue
described in this bugreport. For example Jan has the same issue but he has
neither the GL error from depth_blt popping up on the console nor does my patch
change anything for him.
So this patch might be TOTALLY unrelated to the issue.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #18 from H. Verbeet   2008-10-19 12:13:50 ---
Created an attachment (id=16748)
 --> (http://bugs.winehq.org/attachment.cgi?id=16748)
depth_blt patch

How does this work? It should fix the GL error, but chances are it won't solve
the rest of the problem.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #19 from Tobias Jakobi   2008-10-19 13:54:39 ---
(In reply to comment #18)
> How does this work? It should fix the GL error, but chances are it won't solve
> the rest of the problem.

Hi Henri,

the patch does work, however with a slight modification. You forgot to enable
ARB_texture_rectangle in the GLSL program.

const char *blt_pshader_RECT[] = {
    "#version 120\n"
    "#extension GL_ARB_texture_rectangle : enable\n"
    "uniform sampler2DRect sampler;\n"
    "void main(void)\n"
    "{\n"
    "    gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
    "}\n"
};

This is how it should look like. Without this modification I get again a lot of
GL errors from shader_glsl_deselect_depth_blt.

But guess what. After modifying the patch it works. Which means the GL errors
are gone and the scene renders without any flickering and disappearing
sprites/waterplane.

Performance also increased. Going to check what kind of difference that makes
in Max Payne 2.

@Jan: Check this out on your ATI and report if it helps.

Greets,
Tobias

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #20 from Tobias Jakobi   2008-10-19 14:18:08 ---
Note: Leave this open because it's only fixed for the GLSL shader path and not
for the ARB one.

Still flickering with ARB shaders plus I get Henri's new FIXME message :)

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #21 from Tobias Jakobi   2008-10-19 15:30:31 ---
I'm wondering where depth_blt_glsl_program_id is initialized properly. Is this
even done at all? Henri?

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #22 from H. Verbeet   2008-10-19 15:31:40 ---
Yeah, I only properly implemented RECT textures with GLSL. I wasn't sure if the
GLSL extension string was strictly required, issue 15 of the
ARB_texture_rectangle spec seems to imply it isn't. I'll clean it up and submit
a patch in a couple of days.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #23 from H. Verbeet   2008-10-19 15:34:22 ---
(In reply to comment #21)
> I'm wondering where depth_blt_glsl_program_id is initialized properly. Is this
> even done at all? Henri?

It's initialized with zeroes in shader_glsl_alloc(). (the HeapAlloc() with
HEAP_ZERO_MEMORY)

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #24 from Tobias Jakobi   2008-10-19 16:51:35 ---
Created an attachment (id=16749)
 --> (http://bugs.winehq.org/attachment.cgi?id=16749)
[wined3d] henri's patch plus my port for the arb path

Description already says it.

I took a look at henri's patch, added the ARB_tex_rect extension stuff and
ported the changes to GLSL over to the ARB path. I don't think everything is
correct there, but at least it works for me now.

So if someone feels like he wants to test this, it's available now.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #25 from Jan Buecken   2008-10-19 17:49:58 ---
Created an attachment (id=16750)
 --> (http://bugs.winehq.org/attachment.cgi?id=16750)
log with Tobias patch and GLSL

I tested Tobias version of the patch with newest git on ati with fglrx-driver
(8.542).

In all cases ORM=fbo
0) no patch: regression: I can only see the water if I stand in it and look
down, then it is flickering (I call this regression because I can see the
waterplane with 1.1.6)
1) with patch, GLSL disabeld: like 0), only fixme and one err message
2) with patch, GLSL enabled: like 1), but with a GLSL errors, for this I
attached the relevant part of the command line output.

With other words, I can't see differences with or without Tobias patch.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #26 from Jan Buecken   2008-10-19 18:09:51 ---
Sorry,
but I want to make clear that I test my the cases 1) and 2) without the patch,
too!
And the output is the same.
Hence the patch makes really no difference for me (in this game) including
outputs.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Austin English  changed:

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

--- Comment #27 from Austin English   2008-10-19 19:11:29 ---
Henri's patch seems to be a real one, so readding keyword.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #28 from Tobias Jakobi   2008-10-20 04:09:04 ---
Yeah, it least it fixes the issue for me (NV35 hardware).
We'll see about Jan.

I'm also going to check this on my Intel i915, I think it should be able to run
Morrowind.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #16749|0                           |1
        is obsolete|                            |

--- Comment #29 from Tobias Jakobi   2008-10-23 16:06:20 ---
Created an attachment (id=16829)
 --> (http://bugs.winehq.org/attachment.cgi?id=16829)
[wined3d] Henri's patch plus my port for the ARB path

Rebased against current git master, should now again apply cleanly.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #16722|0                           |1
        is obsolete|                            |

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

H. Verbeet  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #16696|0                           |1
        is obsolete|                            |
  Attachment #16748|0                           |1
        is obsolete|                            |

--- Comment #30 from H. Verbeet   2008-10-26 19:30:22 ---
Created an attachment (id=16914)
 --> (http://bugs.winehq.org/attachment.cgi?id=16914)
Another patch

Could you give this patch a try? If it works I'll submit it.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #16829|0                           |1
        is obsolete|                            |

--- Comment #31 from Tobias Jakobi   2008-10-27 05:27:20 ---
Created an attachment (id=16929)
 --> (http://bugs.winehq.org/attachment.cgi?id=16929)
[wined3d] Corrected version of Henri's patch

Fixes some pointer casting issues that lead to an early crash.
Should compile fine.

Fixes the flickering issue for both GLSL and ARB mode for me.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #32 from Jan Buecken   2008-10-28 09:39:04 ---
Created an attachment (id=16946)
 --> (http://bugs.winehq.org/attachment.cgi?id=16946)
log with tobias patches

Hi,
one new log with git, Tobias patch, and his debug patch from here
http://bugs.winehq.org/attachment.cgi?id=16903

Same bug.

New ideas?

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #33 from H. Verbeet   2008-10-28 10:08:11 ---
Could you post a +d3d8,+d3d,+d3d_shader log?

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #34 from Tobias Jakobi   2008-10-28 12:35:15 ---
Created an attachment (id=16951)
 --> (http://bugs.winehq.org/attachment.cgi?id=16951)
[wined3d] dump GLSL shader to console on error

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #35 from Jan Buecken   2008-10-28 13:38:48 ---
New log with Tobias
   [wined3d] dump GLSL shader to console on error 
and
   [wined3d] Corrected version of Henri's patch 
patches
and
   +d3d8,+d3d,+d3d_shader

Because it is more than 4 MB after bzip2, I stored it here
www.math.uni-bielefeld.de/~jbuecken/log_new_debug_patch_and_winedebug.bz2

Hope this help.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #36 from H. Verbeet   2008-10-28 15:16:09 ---
Those GL errors are caused by the debug patch. You should first check the
object type (GL_OBJECT_SUBTYPE_ARB) before trying to get the shader source,
since obj can be a program as well.

Just to be sure, is d37c6fc0a94b3ac3b4cce58afeafbdf225107b80 also the patch
that causes the regression for you?

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #37 from Tobias Jakobi   2008-10-28 16:18:27 ---
(In reply to comment #36)
> Those GL errors are caused by the debug patch. You should first check the
> object type (GL_OBJECT_SUBTYPE_ARB) before trying to get the shader source,
> since obj can be a program as well.

Hmm, don't you mean calling glGetObjectParameterARBiv with GL_OBJECT_TYPE_ARB
and checking whether it is GL_SHADER_OBJECT_ARB or GL_PROGRAM_OBJECT_ARB?

AFAIK subtype only returns if the shader is a fragment or vertex one.

I'm also not quite sure how to deal with the case when it's a program object
(and not a shader object, I think the shader ones can be dumped correctly with
my current code). Do I need to get the attached shader objects and dump them as
well?

Or is this functionality already implemented somewhere else in the code?

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #38 from H. Verbeet   2008-10-28 16:56:20 ---
(In reply to comment #37)
> (In reply to comment #36)
> > Those GL errors are caused by the debug patch. You should first check the
> > object type (GL_OBJECT_SUBTYPE_ARB) before trying to get the shader source,
> > since obj can be a program as well.
>
> Hmm, don't you mean calling glGetObjectParameterARBiv with GL_OBJECT_TYPE_ARB
> and checking whether it is GL_SHADER_OBJECT_ARB or GL_PROGRAM_OBJECT_ARB?

Probably :-)
> I'm also not quite sure how to deal with the case when it's a program object
> (and not a shader object, I think the shader ones can be dumped correctly with
> my current code). Do I need to get the attached shader objects and dump them as
> well?

You probably don't need to do anything for program objects.
> Or is this functionality already implemented somewhere else in the code?

+d3d_shader will show the source as it's being generated. Finding the source
for a program is then a matter of searching back in the log for the shader
objects that are linked into a program.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #39 from Tobias Jakobi   2008-10-30 17:35:28 ---
Just wanted to confirm that latest git master fixes the problem entirely in
GLSL mode for me. Flickering is gone and no errors are produces on the console.

It's a bit different in ARB mode, here I'm hitting this (new) fixme, which is
continously printed on the console:
fixme:d3d_shader:shader_arb_select_depth_blt Unsupported tex_type 0x4

However it's not flickering in ARB mode either.

Leaving this open because it's not working for Jan and because of the FIXME
(maybe we can remove this as well).

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #40 from Tobias Jakobi   2008-10-30 18:04:40 ---
@Henri: This line in shader_arb_select_depth_blt:
if (tex_type != tex_2d) FIXME("Unsupported tex_type %#x\n", tex_type);

Isn't that already implemented?

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #41 from H. Verbeet   2008-10-31 02:05:20 ---
Oops. Yeah.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #16951|0                           |1
        is obsolete|                            |

--- Comment #42 from Tobias Jakobi   2008-11-01 11:36:33 ---
Created an attachment (id=17030)
 --> (http://bugs.winehq.org/attachment.cgi?id=17030)
[wined3d] dump GLSL shader to console on error

Fixed what Henri told me.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #16929|0                           |1
        is obsolete|                            |

--- Comment #43 from Tobias Jakobi   2008-11-01 11:39:33 ---
(From update of attachment 16929)
Now obsolete, since code is in git master.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #44 from Tobias Jakobi   2008-11-03 13:30:56 ---
Issue now completly fixed for me with latest git master.
The remaining FIXME is gone.

A big thanks goes to Henri for looking into this! :)

Leaving open so Jan can fix his variant of the issue.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #45 from Tobias Jakobi   2009-01-29 15:53:57 ---
Bumping this for Jan. I saw current state (with wine-1.1.13) on his system
today and it's gotten worse.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #46 from Tobias Jakobi   2009-03-11 13:42:26 ---
I took a look at Jan's log from comment #35 to find out what GLSL shader object
#37 consisted of.

Looks like it consists of objects #38, #36 and #35.

35 is a fragment program and 36 is the vertex shader. 38 isn't found anywhere -
which is kinda strange (probably the problem??).

To cite the error message from object #37:
ERROR: Varying gl_TexCoord[1] read in fragment shader, but not written

That here is #35:
#version 120
#extension GL_ARB_draw_buffers : enable
uniform vec4 PC[8];
uniform sampler2D Psampler0;
uniform sampler2D Psampler3;
vec4 T0 = gl_TexCoord[0];
vec4 T1 = gl_TexCoord[1];
vec4 T2 = gl_TexCoord[2];
vec4 T3 = gl_TexCoord[3];
vec4 R0;
vec4 R1;
vec4 tmp0;
vec4 tmp1;
uniform vec4 PLC5;
uniform vec4 PLC1;
void main() {
T0.xyzw = (texture2D(Psampler0, T0.xy).xyzw);
tmp0.x = dot(T2.xyz, (2.0 * (T0.xyz - 0.5)));
tmp0.y = dot(T3.xyz, (2.0 * (T0.xyz - 0.5)));
T3.xyzw = (texture2D(Psampler3, tmp0.xy).xyzw);
T1.xyzw = (PC[2].xyzw * (2.0 * (T0.xyzw - 0.5)));
R1.xyzw = (vec4(dot(T1.xyz, (2.0 * (gl_Color.xyz - 0.5)))));
R0.xyz = (T3.xyz * PC[4].xyz);
R0.w = ((R1.w * PLC1.w) + PLC1.w);
R1.xyzw = ((2.0 * (gl_Color.xyzw - 0.5)));
R0.xyz = (mix(R0.xyz, PC[6].xyz, R1.www));
R0.w = (R0.w * R0.w);
R0.w = (R0.w + -PC[7].w);
gl_FragData[0] = R0;
float Fog = clamp(gl_FogFragCoord * gl_Fog.start + gl_Fog.end, 0.0, 1.0);
gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);
}

To me the GLSL compiler error makes no sense at all since T1 is never read
before it's overwritten with data that comes from elsewhere. So it's in fact
written, not not read?!

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #47 from Jan Buecken   2009-03-12 10:13:56 ---
For me, this bug is a combination between wine and the ati driver:
Some time ago, I test this and wine 1.0.1 worked, but
with new ati-driver (8.552 or higher, I use 8.528):
wine 1.0, 1.0.1, 1.1.1 1.1.2 crashes with unhandeld pagefault.
With 1.1.3 or higher this bug holds (and the new ati-drivers causes new errors,
the menu filckers, too, I think I post a new bug for this).

New log with
ati-driver 8.528
+d3d8,+d3d,+d3d_shader
newest git (wine version = 1.1.16-367-gae1802d)
fbo
glsl enabled

9,1 MB => 
http://www.math.uni-bielefeld.de/~jbuecken/wine-1.1.16-367-gae1802d_fbo_GLSL...

(Behavior doesn't change with backbuffer or glsl disbabled)

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #48 from Tobias Jakobi   2009-03-12 13:11:44 ---
Looks like the problem is consisting of two components.

The first thing is the handling of "optimizable code" by the ATI GLSL compiler.

What do I mean by that. Well, the problem is that one GLSL shader program (id
#35 in Jan's latest log) doesn't get compile properly, because of the (already
known) error:

Fragment shader(s) linked, vertex shader(s) linked.
WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported
WARNING: built-in varying gl_TexCoord [1] has mismatched access semantics
between the vertex and fragment shader
ERROR: Varying gl_TexCoord[1] read in fragment shader, but not written

There are three shader objects linked to the program #35. These are #36, #37
and #38. I could identify #36 as the vertex shader and #38 as the fragment
shader.

That is #38, the fragment shader:
#version 120
#extension GL_ARB_draw_buffers : enable
uniform vec4 PC[8];
uniform sampler2D Psampler0;
uniform sampler2D Psampler3;
vec4 T0 = gl_TexCoord[0];
vec4 T1 = gl_TexCoord[1];
vec4 T2 = gl_TexCoord[2];
vec4 T3 = gl_TexCoord[3];
vec4 R0;
vec4 R1;
vec4 tmp0;
vec4 tmp1;
uniform vec4 PLC5;
uniform vec4 PLC1;
void main() {
T0.xyzw = (texture2D(Psampler0, T0.xy).xyzw);
tmp0.x = dot(T2.xyz, (2.0 * (T0.xyz - 0.5)));
tmp0.y = dot(T3.xyz, (2.0 * (T0.xyz - 0.5)));
T3.xyzw = (texture2D(Psampler3, tmp0.xy).xyzw);
T1.xyzw = (PC[2].xyzw * (2.0 * (T0.xyzw - 0.5)));
R1.xyzw = (vec4(dot(T1.xyz, (2.0 * (gl_Color.xyz - 0.5)))));
R0.xyz = (T3.xyz * PC[4].xyz);
R0.w = ((R1.w * PLC1.w) + PLC1.w);
R1.xyzw = ((2.0 * (gl_Color.xyzw - 0.5)));
R0.xyz = (mix(R0.xyz, PC[6].xyz, R1.www));
R0.w = (R0.w * R0.w);
R0.w = (R0.w + -PC[7].w);
gl_FragData[0] = R0;
float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);
float fogend = gl_Fog.end * -fogstart;
float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);
gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);
}

The problematic line is "vec4 T1 = gl_TexCoord[1];"
The compiler considers this reading from gl_TexCoord[1] even though this line
could probably be optimized into "vec4 T1;" without the initialization through
data from gl_TexCoord[1]. T1 is never read before it is fully overwritten
through "T1.xyzw = (PC[2].xyzw * (2.0 * (T0.xyzw - 0.5)));"

Now taking a look at the vertex shader (#36):
#version 120
uniform vec4 VC[107];
uniform vec4 posFixup;
void order_ps_input();
vec4 R4;
vec4 R5;
vec4 R6;
vec4 R7;
vec4 R8;
vec4 R9;
attribute vec4 attrib0;
attribute vec4 attrib1;
vec4 tmp0;
vec4 tmp1;
void main() {
gl_Position.x = (dot(attrib0.xyzw, VC[2].xyzw));
gl_Position.y = (dot(attrib0.xyzw, VC[3].xyzw));
gl_Position.z = (dot(attrib0.xyzw, VC[4].xyzw));
gl_Position.w = (dot(attrib0.xyzw, VC[5].xyzw));
R5.xyzw = (vec4(dot(attrib0.xyzw, VC[4].xyzw)));
R5.xyzw = (R5.xyzw + -VC[54].xxxx);
gl_FogFragCoord = ((-R5.x * VC[54].z) + VC[54].w);
R4.xyzw = (-attrib0.xyzw + VC[6].xyzw);
R9.xyzw = (R4.xyzw);
R4.z = (VC[30].x);
R4.w = (dot(R4.xyz, R4.xyz));
R4.w = (inversesqrt(R4.w));
R4.xyz = (R4.xyz * R4.www);
R6.xyzw = (R4.xyzw);
R6.z = (VC[30].y);
R5.xyzw = (VC[30].xxzx);
R8.xyzw = (R5.yzxw * R4.zxyw);
R8.xyzw = ((-R4.yzxw * R5.zxyw) + R8.xyzw);
R8.z = (VC[30].y);
R7.xyzw = (R6.xyzw);
R7.y = (R8.x);
R8.x = (R6.y);
R9.w = (dot(R9.xyz, R9.xyz));
R9.w = (inversesqrt(R9.w));
R9.xyz = (R9.xyz * R9.www);
R9.z = (max(R9.z, -R9.z));
R5.xyzw = (VC[1].xyzw + -R9.zzzz);
R5.xyzw = (R5.xyzw * R5.xyzw);
R5.xyzw = (R5.xyzw * VC[30].yyyy);
R7.z = ((R5.z * R4.x) + R7.z);
R8.z = ((R5.z * R4.y) + R8.z);
R7.xy = (R7.xy * VC[53].xy);
R8.xy = (R8.xy * VC[53].xy);
gl_TexCoord[2].xyzw = (R7.xyzw);
gl_TexCoord[3].xyzw = (R8.xyzw);
gl_FrontColor.xyzw = ((R9.xyzz * VC[30].yyyy) + VC[30].yyyy);
gl_TexCoord[0].xy = (attrib1.xy);
order_ps_input();
gl_Position.y = gl_Position.y * posFixup.y;
gl_Position.xy += posFixup.zw * gl_Position.ww;
gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;
}

As you can see gl_TexCoord[1] is never accessed (more important: written to) in
the vshader. Just inserting a "gl_TexCoord[1] = vec4(0.0, 0.0, 0.0, 0.0);" line
after "void main() {" solves the compiler error, but sadly the flickering
remains.

However I suspect something else to be a more critical source of the
flickering, that's the second component:

One of the first errors from OpenGL is this one:
fixme:d3d:IWineD3DDeviceImpl_CreateSwapChain >>>>>>>>>>>>>>>>>
GL_INVALID_OPERATION (0x502) from glDrawBuffer(GL_BACK) @ device.c / 1838

Later the console is flooded with these kind of errors:
fixme:d3d:apply_draw_buffer >>>>>>>>>>>>>>>>> GL_INVALID_OPERATION (0x502) from
glDrawBuffers() @ context.c / 1445

Please note that not "everything" is affected by the flickering. It's the
ingame waterplane, the ingame menu and the main menu. And some part of the
vegetation (leafage of trees e.g.). The main ingame geometry is unaffected.

Looks like that the flickering is happening because no backbuffer is existant
for these parts. Which sounds logical: No backbuffer so everything is always
rendered to GL_FRONT, resulting is vsync-like flickering artifacts.

Also note that the GL_BACK problem didn't always exist. It looks like it
doesn't depend on the wine version, but on the version of the ATI driver. In
comment #32 e.g. I can't see any of these messages (like the swapchain one) in
the logs.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #49 from Tobias Jakobi   2009-03-12 13:22:10 ---
And that's the surface making all the trouble:
trace:d3d:IWineD3DDeviceImpl_CreateSurface (0x19ed58) : w(1024) h(768)
fmt(3,WINED3DFMT_X8R8G8B8) lockable(0) surf@0x1b0880, surfmem@0xe20020, 3145728
bytes

SwapChain fails here and it's also the surf that makes apply_draw_buffer break.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

--- Comment #50 from Jan Buecken   2009-03-29 20:35:07 ---
fixed with wine-1.1.18 and ati-catalyst-9.3 (=8.593)

I'll delete the external logs (links above) in the near future.

Thanks for your help!

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Tobias Jakobi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED

--- Comment #51 from Tobias Jakobi   2009-03-30 06:45:29 ---
Reported as FIXED.

[Bug 15644] Morrowind: Waterplane/sprite flickering with ORM=fbo

http://bugs.winehq.org/show_bug.cgi?id=15644

Alexandre Julliard  changed:

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

--- Comment #52 from Alexandre Julliard   2009-04-10 11:19:30 ---
Closing bugs fixed in 1.1.19.