June 2009 - Wine-devel - list.winehq.org

Re: [3/6] WineD3D: Handle SM 3.0 varyings in ARB
by Allan Tong 17 Jun '09

17 Jun '09

On Wed, Jun 17, 2009 at 1:37 PM, Stefan Dösinger <stefan(a)codeweavers.com> wrote: > + if(strcmp(rel_reg, "**aL_emul**") == 0) > + { > + DWORD idx = ctx->aL + reg->idx; > + if(idx < 10) Seems like this should be MAX_REG_INPUT instead of 10. > + { > + strcpy(register_name, ctx->ps_input[reg->idx]); Should this be "ctx->ps_input[idx]"? > + } > + else > + { > + ERR("Pixel shader input register out of bounds: %u\n", idx); > + sprintf(register_name, "out_of_bounds_%u", idx); > + } > + } - Allan

2 1

Re: [3/6] d3dx9: Implement ID3DXFont_GetDesc
by Roderick Colenbrander 17 Jun '09

17 Jun '09

Hi Tony, + WideCharToMultiByte(CP_ACP, 0, This->desc.FaceName, -1, desc->FaceName, sizeof(desc->FaceName) / sizeof(CHAR), NULL, NULL); sizeof(desc->FaceName) / sizeof(CHAR) won't give you the length of the string (remember that sizeof is evaluated at compile time, so the value is a constant) Regards, Roderick

2 2

Re: [3/6] WineD3D: Handle SM 3.0 varyings in ARB
by Henri Verbeet 17 Jun '09

17 Jun '09

> @@ -1465,9 +1471,14 @@ struct IWineD3DDeviceImpl > const struct fragment_pipeline *frag_pipe; > const struct blit_shader *blitter; > > + struct ps_signature *pshader_signatures; > + unsigned int num_pshader_signatures; > + unsigned int pshader_signatures_size; > + DWORD ps_sig_number; > + > unsigned int max_ffp_textures, max_ffp_texture_stages; > DWORD d3d_vshader_constantF, d3d_pshader_constantF; /* Advertised d3d caps, not GL ones */ > - DWORD vs_clipping; > + DWORD vs_clipping, need_ps_signatures; > > WORD view_ident : 1; /* true iff view matrix is identity */ > WORD untransformed : 1; > @@ -2730,6 +2741,7 @@ typedef struct IWineD3DPixelShaderImpl { > BOOL color0_mov; > DWORD color0_reg; > > + DWORD input_signature_idx; Please keep this private to the ARB backend. (I.e., store these in struct shader_arb_priv.) Also, it looks like you're trying to write a binary search tree. Would you like to use wine_rb_tree instead?

2 4

Re: [6/6] d3dx9: Add tests for basic ID3DXFont functions
by Roderick Colenbrander 17 Jun '09

17 Jun '09

Hi Tony, + hr = D3DXCreateFontA(device, 12, 0, FW_DONTCARE, 0, FALSE, DEFAULT_CHARSET, OUT_DEFAULT_PRECIS, DEFAULT_QUALITY, DEFAULT_PITCH, "Arial", &font); A bunch of font tests all use Arial. On Windows this font is always around, what happens on Wine when this font isn't around? Perhaps the function fails on fonts which aren't around, so this could potentially result in a lot of test failures on Wine. Not sure what should be done if this is the case though (perhaps check at run time if the font is around or perhaps use a font which is always around as a fallback). Regards, Roderick

2 1

Re: [6/6] WineD3D: Emulate clipplanes in ARB if the NV extensions are not available
by Henri Verbeet 17 Jun '09

17 Jun '09

2009/6/17 Stefan Dösinger <stefan(a)codeweavers.com>: > + //ERR("Using texcoord %u for clip\n", ps->clipplane_emulation); C99 features generally aren't allowed inside Wine code. > + if(!device->vs_clipping) > + { > + /* See if we can use fragment.texcoord[7] for clipplane emulation > + * > + * Don't do this if it is not supported, or fragment.texcoord[7] is used > + */ > + if(This->baseShader.reg_maps.shader_version.major < 3) > + { > + for(i = GL_LIMITS(texture_stages); i > 0; i--) > + { > + if(!This->baseShader.reg_maps.texcoord[i - 1]) > + { > + This->clipplane_emulation = i; > + break; > + } > + } > + } > + else > + { > + for(i = GL_LIMITS(texture_stages); i > 0; i--) > + { > + if(!This->baseShader.reg_maps.input_registers & (1 << (i - 1))) > + { > + This->clipplane_emulation = i; > + break; > + } > + } > + } > + } This is specific to ARB as well. > + char clipplane_emulation; Using a char for this is pretty useless. If you want to save memory or cachelines, use a bitfield. The compiler will just pad the structure for alignment, and if you add other elements below it you'll have a hole, like e.g. after vpos_uniform in the same structure.

2 2

Re: [4/6] d3dx9: Implement ID3DXFont_GetDC
by Stefan Dösinger 17 Jun '09

17 Jun '09

Am Wednesday 17 June 2009 18:29:57 schrieb Tony Wasserka: > @@ -300,6 +301,12 @@ HRESULT WINAPI D3DXCreateFontIndirectW(LPDIRECT3DDEVICE9 device, CONST D3DXFONT_ > object->device=device; > object->desc=*desc; > > + object->hdc = CreateCompatibleDC(NULL); Device contexts are scarce resources, please avoid holding one all the time. Instead, create it in GetDC and destroy it in ReleaseDC. Besides this, after ReleaseDC the DC is supposed to be invalid - the only way to do this is to destroy it. (Yes, I know that's broken in wined3d too. Needs fixing there as well) You'll also have to attach the DC to the font bitmap, or whatever you have, to have it draw to the correct device. I think the current code means that the DC will draw to the screen directly, which is maybe not what you want. If the font has a d3d surface underneath, you should use d3dsurface9::GetDC

2 1

Re: [4/6] WineD3D: Don't run out of varyings because of clipplanes
by Henri Verbeet 17 Jun '09

17 Jun '09

2009/6/17 Stefan Dösinger <stefan(a)codeweavers.com>: > If I have to choose between a compile failure or disabling some clipplanes > I'll throw the clipplanes under the bus. > > If the shader is using relative addressing, look at the highest directly > addressed constant register. We know that the constant array must be at least > highest_idx+1 elements large, otherwise the shader will fail for sure. If we > see this problem coming, only enable as many clipplanes as we can fit into the > hardware limit. > > Something similar should be done in GLSL, but GLSL can only support all > clipplanes or none at all. I don't think we want this for GLSL. ARB is your own responsibility, but personally I think compile errors are much nicer to debug than the kind of graphics glitches missing clipplanes will create.

2 1

Re: Request for Patchreview
by Juan Lang 17 Jun '09

17 Jun '09

Hi Andre, > I fixed the strange problems in try8 ( > http://www.winehq.org/pipermail/wine-patches/2009-June/074340.html ) and it > still dont get commit. > can you please do a review for this try? (Just say it, if it gets disturbing > you.) This one looks fine to me. And don't worry, you're not disturbing me. --Juan

1 0

Re: [PATCH 2/2] wined3d: improved ATI Radeon HD 4xxx detection
by Stefan Dösinger 17 Jun '09

17 Jun '09

The patches seem ok to me, although I'd probably merge them into one patch(there's no point in adding unused defines)

3 2

Re: oleaut32/tests: Add some tests for [lcid] parameters.
by Paul Vriens 17 Jun '09

17 Jun '09

Huw Davies wrote: > --- > dlls/oleaut32/tests/tmarshal.c | 53 +++++++++++++++++++++++++++++++- > dlls/oleaut32/tests/tmarshal.idl | 6 +++ > dlls/oleaut32/tests/tmarshal_dispids.h | 1 + > 3 files changed, 59 insertions(+), 1 deletions(-) > > > ------------------------------------------------------------------------ > > In an effort to get a green NT4, I ran your new tests on NT4: tmarshal.c:1376: Test failed: ITypeInfo_Invoke failed with error 0x80010105 tmarshal.c:1377: Test failed: got 0 tmarshal.c:1378: Test failed: got 0 On W2K3: tmarshal.c:1376: Test failed: ITypeInfo_Invoke failed with error 0x80010105 tmarshal.c:1377: Test failed: got 0 tmarshal.c:1378: Test failed: got 0 and a maybe unrelated on W2K3 (wasn't here before): tmarshal.c:605: Test failed: Struct parameter passed by pointer corrupted On W98: tmarshal.c:1376: Test failed: ITypeInfo_Invoke failed with error 0x80010105 tmarshal.c:1377: Test failed: got 0 tmarshal.c:1378: Test failed: got 0 Am I doing something wrong here? -- Cheers, Paul.

2 3