May 2011 - Wine-Devel - list.winehq.org

Re: [1/5] wined3d: Don't check the FBO status if FIXMEs are off
by Henri Verbeet May 25, 2011

May 25, 2011

On 24 May 2011 22:37, Stefan Dösinger <stefan(a)codeweavers.com> wrote: > I put the fixme checks outside context_check_fbo_status because the ENTER_GL() / LEAVE_GL() > locking is more expensive than the FBO validation itself. Moving the locking into the function I'm not sure that's really worth it. If the GL lock is really such a concern, you should probably change the calls in context.c to already be under the GL lock anyway. You can easily change context_apply_draw_buffers() for example to require locking by the caller.

2 1

Re: usp10/tests: Make enumFontProc() static.
by Marvin May 25, 2011

May 25, 2011

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at http://testbot.winehq.org/JobDetails.pl?Key=11170 Your paranoid android. === WNT4WSSP6 (32 bit usp10) === Timeout

1 0

Re: [PATCH 1/2] ddraw: COM cleanup for IDirectDrawSurface3 iface
by Michael Stefaniuc May 25, 2011

May 25, 2011

Hello Ričardas! Ričardas Barkauskas wrote: > My first try for COM cleanup. Please donėt be too harsh :) > If this is too large for one patch, please suggest how to split it. > > This is not automatic clean up so I don't really expect it to be good on > first try. Heh! I have attached the raw diff generated by my coccinelle script. Looking at the interdiff between the two patches it looks like you caught everything. And the BADBADBAD() (*) from the generated patch are replaced with unsafe_impl_from_IDirectDrawSurface3(), very good. If you want to do more COM cleanup I can give you the raw generated patches to save you some time. bye michael P.S.: (*) My script validates that impl_from_IFace() is called only from methods of that IFace. All other impl_from_IFace() calls are replaced with BADBADBAD().

1 1

Re: cmd: Don't read past end of thisVar in WCMD_expand_envvar (valgrind)
by Marvin May 25, 2011

May 25, 2011

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at http://testbot.winehq.org/JobDetails.pl?Key=11166 Your paranoid android. === WNT4WSSP6 (32 bit) === batch.c:223: Test failed: unexpected char 0x54 position 0 in line 86 (got 'This Line Is Never Reached. If this runs, it is a bug.', wanted '65536') batch.c:223: Test failed: unexpected char 0x31 position 0 in line 87 (got '1', wanted 'errorlevel just right, good') batch.c:223: Test failed: unexpected char 0x65 position 0 in line 88 (got 'errorlevel too low, bad', wanted 'abc0def') batch.c:223: Test failed: unexpected char 0x54 position 0 in line 89 (got 'This Line Is Never Reached. If this runs, it is a bug.', wanted 'errorlevel zero, good') batch.c:223: Test failed: unexpected char 0x61 position 0 in line 90 (got 'abc1def', wanted '7 should be 7') batch.c:223: Test failed: unexpected char 0x65 position 0 in line 91 (got 'errorlevel nonzero, bad', wanted '7 should still be 7') batch.c:223: Test failed: unexpected char 0x37 position 0 in line 92 (got '7 should be 7', wanted '3 should be 3 now') batch.c:223: Test failed: unexpected char 0x54 position 0 in line 93 (got 'This Line Is Never Reached. If this runs, it is a bug.', wanted '-----------Testing GOTO-----------') batch.c:223: Test failed: unexpected char 0x37 position 0 in line 94 (got '7 should still be 7', wanted 'goto with no leading space worked') batch.c:223: Test failed: unexpected char 0x30 position 0 in line 95 (got '0 should be 3 now', wanted 'goto with a leading space worked') batch.c:223: Test failed: unexpected char 0x2d position 0 in line 96 (got '-----------Testing GOTO-----------', wanted 'goto with a leading tab worked') batch.c:223: Test failed: unexpected char 0x67 position 0 in line 97 (got 'goto with no leading space worked', wanted 'goto with a following space worked') batch.c:223: Test failed: unexpected char 0x67 position 0 in line 98 (got 'goto with a leading space worked', wanted '-----------Done, jumping to EOF-----------') batch.c:235: Test failed: too long output, got additional goto with a leading tab worked === W2KPROSP4 (32 bit) === batch.c:223: Test failed: unexpected char 0x30 position 0 in line 92 (got '0 should be 3 now', wanted '3 should be 3 now') === WXPPROSP3 (32 bit) === batch.c:223: Test failed: unexpected char 0x30 position 0 in line 92 (got '0 should be 3 now', wanted '3 should be 3 now') === W2K3R2SESP2 (32 bit) === batch.c:223: Test failed: unexpected char 0x30 position 0 in line 92 (got '0 should be 3 now', wanted '3 should be 3 now') === WVISTAADM (32 bit) === batch.c:223: Test failed: unexpected char 0x30 position 0 in line 92 (got '0 should be 3 now', wanted '3 should be 3 now') === W2K8SE (32 bit) === batch.c:223: Test failed: unexpected char 0x30 position 0 in line 92 (got '0 should be 3 now', wanted '3 should be 3 now') === W7PRO (32 bit) === batch.c:223: Test failed: unexpected char 0x30 position 0 in line 92 (got '0 should be 3 now', wanted '3 should be 3 now') === W7PROX64 (32 bit) === batch.c:223: Test failed: unexpected char 0x30 position 0 in line 92 (got '0 should be 3 now', wanted '3 should be 3 now') === W7PROX64 (64 bit) === batch.c:223: Test failed: unexpected char 0x30 position 0 in line 92 (got '0 should be 3 now', wanted '3 should be 3 now')

2 1

windows are incompletely drawn: bisected to commit 7864ade5a8306c0078e16ae6d7e40bdece29395b
by Wolfgang Walter May 24, 2011

May 24, 2011

An application we use shows incompletely painted windows with 1.3.19. I bisected it to commit 7864ade5a8306c0078e16ae6d7e40bdece29395b is the first bad commit commit 7864ade5a8306c0078e16ae6d7e40bdece29395b Author: Dmitry Timoshkov <dmitry(a)codeweavers.com> Date: Tue Sep 14 14:24:11 2010 +0900 winex11.drv: Avoid copying invalid window bits. I reverted it in 1.3.19 and that fixed indeed the problem. The wrongly painted windows are maximized windows within a MDI window: if there are serveral such windows opened and one switches between them parts of the newly active window is not painted and the content of the old one is still displayed. Regards, -- Wolfgang Walter Studentenwerk München Anstalt des öffentlichen Rechts

2 1

Re: [2/5] d3dx9: Generate effect instances from materials for mesh loading.
by Matteo Bruni May 24, 2011

May 24, 2011

2011/5/24 Dylan Smith <dylan.ah.smith(a)gmail.com>: > --- > dlls/d3dx9_36/mesh.c | 117 +++++++++++++++++++++++++++++++++++++++++++++++-- > 1 files changed, 112 insertions(+), 5 deletions(-) > > diff --git a/dlls/d3dx9_36/mesh.c b/dlls/d3dx9_36/mesh.c > index fa32797..20b3262 100644 > --- a/dlls/d3dx9_36/mesh.c > +++ b/dlls/d3dx9_36/mesh.c ... > + strcpy(defaults->pParamName, diffuse_paramname); > + defaults->pValue = defaults->pParamName + sizeof(diffuse_paramname); > + defaults->Type = D3DXEDT_FLOATS; > + defaults->NumBytes = sizeof(D3DCOLORVALUE); > + memcpy(defaults->pValue, &material_ptr->MatD3D.Diffuse, sizeof(D3DCOLORVALUE)); > + (defaults + 1)->pParamName = (LPSTR)((D3DCOLORVALUE*)defaults->pValue + 1); > + defaults++; > + > + strcpy(defaults->pParamName, power_paramname); > + defaults->pValue = defaults->pParamName + sizeof(power_paramname); > + defaults->Type = D3DXEDT_FLOATS; > + defaults->NumBytes = sizeof(FLOAT); > + *(FLOAT*)defaults->pValue = material_ptr->MatD3D.Power; > + (defaults + 1)->pParamName = (LPSTR)((FLOAT*)defaults->pValue + 1); > + defaults++; ... A nitpick maybe, but I think you could avoid to repeat the same code for each parameter by means of a loop driven by a table. It may be worth a try if it doesn't become more cumbersome than the current approach.

1 0

Re: [2/2]usp10/test: add Tibetan shaping test
by Marvin May 24, 2011

May 24, 2011

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at http://testbot.winehq.org/JobDetails.pl?Key=11159 Your paranoid android. === WNT4WSSP6 (32 bit usp10) === Timeout

1 0

Re: usp10/test: make sure to only trace on _test_shape_ok number of glyphs if font is not valid
by Marvin May 24, 2011

May 24, 2011

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at http://testbot.winehq.org/JobDetails.pl?Key=11158 Your paranoid android. === WNT4WSSP6 (32 bit usp10) === Timeout

1 0

Re: [PATCH 4/5] wined3d: Mask out writes to unused render targets.
by Stefan Dösinger May 24, 2011

May 24, 2011

On Tuesday 24 May 2011 21:18:53 Henri Verbeet wrote: > @@ -1999,19 +1999,22 @@ static void context_apply_draw_buffers(struct > wined3d_context *context, UINT rt_ if > (wined3d_settings.offscreen_rendering_mode == ORM_FBO) { > const struct wined3d_gl_info *gl_info = context->gl_info; > - unsigned int i; > + unsigned int i = 0; > > - for (i = 0; i < rt_count; ++i) > + while (rt_mask) > { > if (rts[i] && rts[i]->resource.format->id != > WINED3DFMT_NULL) context->draw_buffers[i] = GL_COLOR_ATTACHMENT0 + i; > else > context->draw_buffers[i] = GL_NONE; > + > + rt_mask >>= 1; > + ++i; > } Shouldn't you also check if the lowest bit of rt_mask? Otherwise you'll miss situations where oC0 is written, oC1 isn't and oC2 is written. context->draw_buffer_dirty also has to be dirtified somewhere when the pixel shader is changed, their rt masks are different and a target is bound to one of the changed indices.

1 0

Re: [PATCH] wined3d: Mark wined3d_pixel_format with hidden visibility
by Henri Verbeet May 24, 2011

May 24, 2011

On 13 May 2011 18:27, Marcus Meissner <meissner(a)suse.de> wrote: > @@ -1273,7 +1273,7 @@ struct wined3d_pixel_format > BOOL doubleBuffer; > int auxBuffers; > int numSamples; > -} wined3d_pixel_format; > +} wined3d_pixel_format DECLSPEC_HIDDEN; > No, that's wrong. It's not supposed to be a variable at all, it's a leftover from the typedef.

1 0