July 2006 - Wine-Devel - list.winehq.org

cmp instruction in GLSL - problems
by Ivan Gyurdiev July 11, 2006

July 11, 2006

This patch is causing trouble for me. The current GLSL cmp implementation is wrong - here's 4 reasons why, only the first of which is mentioned in the source: - it ignores destination write mask - it ignores source swizzle - it ignores other source modifiers. - it works incorrectly src0 = 0 So, I replaced it with something that's supposed to fix all the issues above - based on per-component lessThan operation followed by a mix(), which blends the two things we're selecting between, but the blend factor is always 0 or 1 exactly, so it's always a select operation, not a blend. Patch improved: Hair, TransparentShadowMapping Patch broke: 3Dc, PhongLighting, ASCII, SelfShadowBump I still think it's correct - tested behavior separately, and it seems to work... but it breaks other demos. Any suggestions? All the demos above can be found on http://www.humus.ca/index.php?page=3D. They require UsingGLSL to be enabled in the registry, and a GLSL-capable card. ========= Note that all of the above demos used replicate swizzle in at least one argument, which the previous implementation did not take into account. A lot of them did this: cmp r1.a, r1.a, 0, 1 mad r0, r0, r1.a, something_else So, that looks like z-buffer test. Note that with the previous implementation, the comparison line would ignore the write mask and write to all 4 channels, which should cause a result that's too bright in all channels. I think that's exactly what happened in a number of these demos, PhongLighting in particular. If I use the new code, and simply swap the two arguments in the cmp test, the demo will work for the most part, a lot more is shown than black screen, and the color balance looks much closer to what's shown in the picture online. ...but.. I shouldn't have to swap the two arguments in cmp, they are correct as is, imho, and the previous implementation worked by accident. --- dlls/wined3d/glsl_shader.c | 11 +++++------ 1 files changed, 5 insertions(+), 6 deletions(-) diff --git a/dlls/wined3d/glsl_shader.c b/dlls/wined3d/glsl_shader.c index 223a8e0..23f944c 100644 --- a/dlls/wined3d/glsl_shader.c +++ b/dlls/wined3d/glsl_shader.c @@ -987,20 +987,19 @@ void shader_glsl_compare(SHADER_OPCODE_A /** Process CMP instruction in GLSL (dst = src0.x > 0.0 ? src1.x : src2.x), per channel */ void shader_glsl_cmp(SHADER_OPCODE_ARG* arg) { + char tmpLine[256]; char dst_str[100], src0_str[100], src1_str[100], src2_str[100]; char dst_reg[50], src0_reg[50], src1_reg[50], src2_reg[50]; char dst_mask[6], src0_mask[6], src1_mask[6], src2_mask[6]; - + shader_glsl_add_param(arg, arg->dst, 0, FALSE, dst_reg, dst_mask, dst_str); shader_glsl_add_param(arg, arg->src[0], arg->src_addr[0], TRUE, src0_reg, src0_mask, src0_str); shader_glsl_add_param(arg, arg->src[1], arg->src_addr[1], TRUE, src1_reg, src1_mask, src1_str); shader_glsl_add_param(arg, arg->src[2], arg->src_addr[2], TRUE, src2_reg, src2_mask, src2_str); - /* FIXME: This isn't correct - doesn't take the dst's swizzle into account. */ - shader_addline(arg->buffer, "%s.x = (%s.x > 0.0) ? %s.x : %s.x;\n", dst_reg, src0_reg, src1_reg, src2_reg); - shader_addline(arg->buffer, "%s.y = (%s.y > 0.0) ? %s.y : %s.y;\n", dst_reg, src0_reg, src1_reg, src2_reg); - shader_addline(arg->buffer, "%s.z = (%s.z > 0.0) ? %s.z : %s.z;\n", dst_reg, src0_reg, src1_reg, src2_reg); - shader_addline(arg->buffer, "%s.w = (%s.w > 0.0) ? %s.w : %s.w;\n", dst_reg, src0_reg, src1_reg, src2_reg); + shader_glsl_add_dst(arg->dst, dst_reg, dst_mask, tmpLine); + shader_addline(arg->buffer, "%smix(vec4(%s), vec4(%s), vec4(lessThan(vec4(%s), vec4(0.0)))))%s;\n", + tmpLine, src1_str, src2_str, src0_str, dst_mask); } /** Process the CND opcode in GLSL (dst = (src0 < 0.5) ? src1 : src2) */ -- 1.4.0

1 2

Re: Link or rename wcmd.exe to cmd.exe?
by Neal Gompa July 10, 2006

July 10, 2006

Forwarded because accidently did not add wine-devel to it. Also, noting that this is about Wine, not ReactOS. But both still apply. ---------- Forwarded message ---------- From: Neal Gompa <ngompa13(a)gmail.com> Date: Jul 10, 2006 5:59 PM Subject: Re: Link or rename wcmd.exe to cmd.exe? To: Thomas Kho <tkho(a)ucla.edu> Most applications I know of will check not for cmd.exe, but for command.com. I believe Windows uses a stub command.com to open up cmd.com with more limited capabilities. I believe we need both cmd.exe and command.com on ReactOS, in order to ensure compatibility. However, the state of wcmd.exemakes me think that it is not quite ready for regular usage. As such, I think you should just add a stub command.com that forwards everything to wcmd.exe, and nearly all Windows apps that have commandline dependencies (MikTeX comes in mind) use command.com, not cmd.exe. Just my 2¢. On 7/6/06, Thomas Kho <tkho(a)ucla.edu> wrote: > > On 7/1/06, Alexandre Julliard <julliard(a)winehq.org> wrote: > > "Thomas Kho" <tkho(a)ucla.edu> writes: > > > > > I'm currently looking into what would be required for something like a > > > symlink to wcmd.exe, but at the same time I'm wondering the historical > > > > reasons for choosing 'wcmd' and if it should be changed to just 'cmd'. > > > > Yes, I think it should be renamed to cmd.exe, at this point it's > > supposed to be a complete replacement. > > It's a ~500k patch to rename programs/wcmd to programs/cmd so I want > to get this as right as possible prior to submitting a patch. How much > of wcmd's identity should remain? That is--should it be a wholesale > s/wcmd/cmd or should wcmd keep the name wcmd with just the executable > name changed? > > Thomas Kho > > >

1 0

Re: Speed up GetTextExtentExPointW (Take 3)
by Huw Davies July 10, 2006

July 10, 2006

On Mon, Jul 10, 2006 at 01:24:53PM -0700, Dan Hipschman wrote: > > Hi, > > This is a fix for http://bugs.winehq.org/show_bug.cgi?id=5569 It speeds up > the loading of the EULA for the trial version of Acrobat by about 50x. Since > the last attempt at this fix I improved performance in the case where both > lpnFit and alpDx are NULL. I also added a few more tests to cover each of > the four cases where lpnFit and alpDx are NULL/non-NULL. The tests pass in > both Wine and XP Pro. If you don't accept this, please let me know what I can > do to improve it. Thanks. There are GetTextExtentPoint entry points in both winex11.drv and wineps.drv. Both of these will need to be updated to GetTextExtentExPoint. For one other niggle see below. > diff --git a/dlls/gdi/freetype.c b/dlls/gdi/freetype.c > index 9f52e60..2020d9d 100644 > --- a/dlls/gdi/freetype.c > +++ b/dlls/gdi/freetype.c > @@ -3849,32 +3849,43 @@ BOOL WineEngGetCharABCWidthsI(GdiFont fo ... > > size->cx = 0; > WineEngGetTextMetrics(font, &tm); > size->cy = tm.tmHeight; > > - for(idx = 0; idx < count; idx++) { > - get_glyph_index_linked(font, wstr[idx], &linked_font, &glyph_index); > + for (i = 0; i < count; ++i) > + { > + get_glyph_index_linked(font, wstr[i], &linked_font, &glyph_index); Please don't add in this sort of change, it just makes the patch even larger. Huw.

2 1

New Bugzilla Season begins
by Vijay Kiran Kamuju July 10, 2006

July 10, 2006

Hi all, I would say lets start, checking out all the old bugs. And close all the bugs which have been fixed by latest 0.9.17 The bug count is near to 1910 or more. Last time when we did this exercise was when 0.9 was to be released. Say lets start the season again. And close all the bugs that have been fixed and revive those that are neglected. Thanks, Vijay

1 0

riched20 - patch ping
by Krzysztof Foltman July 10, 2006

July 10, 2006

These could get in unchanged, perhaps? http://www.winehq.org/pipermail/wine-patches/2006-June/028096.html http://www.winehq.org/pipermail/wine-patches/2006-June/027940.html http://www.winehq.org/pipermail/wine-patches/2006-June/027978.html This is the one I want comments for before I split it into smaller parts (and update to take Thomas Kho's patch into account): http://www.winehq.org/pipermail/wine-patches/2006-June/028159.html Krzysztof

1 0

widl/parser.y: Correct use of free()?
by Andrew Talbot July 10, 2006

July 10, 2006

Hi, I am trying to constify some of the "name" strings in this file. But I am running into problems caused by what seems to be an unusual use of the free() function. To my novice eye, it appears that, in functions such as get_type(), the author is trying to free strings that were not given space via malloc(). Is free() being used correctly in this file? Can I remove these calls to free()? Thanks, -- Andy.

2 2

Re: msi: Implement the InstallServices action
by Robert Shearman July 10, 2006

July 10, 2006

James Hawkins wrote: > +static UINT ITERATE_InstallService(MSIRECORD *rec, LPVOID param) > +{ > ... > + service = CreateServiceW(hscm, name, disp, SERVICE_START | > GENERIC_ALL, > + serv_type, start_type, err_control, > file->TargetPath, > + load_order, NULL, depends, serv_name, pass); > + if (!service) > + ERR("Failed to create service %s: %ld\n", debugstr_w(disp), > GetLastError()); > + > + if (!StartServiceW(service, 0, NULL)) > + ERR("Failed to start service %s: %ld\n", debugstr_w(disp), > GetLastError()); Is this function really supposed to start the service as well? Shouldn't this be done by the StartServices action instead (http://msdn.microsoft.com/library/default.asp?url=/library/en-us/msi/setup/… -- Rob Shearman

2 1

Re: user[2/5]: implement A version of SPI_GETDESKWALLPAPER
by Dmitry Timoshkov July 10, 2006

July 10, 2006

"Andrew Ziem" <ahziem1(a)mailbolt.com> wrote: > + case SPI_GETDESKWALLPAPER: /* 115 */ > + { > + WCHAR buffer[MAX_PATH]; > + ret = SystemParametersInfoW( SPI_GETDESKWALLPAPER, uiParam, buffer, fuWinIni ); > + if (!WideCharToMultiByte(CP_ACP, 0, buffer, -1, pvParam, uiParam, NULL, NULL)) > + ret = FALSE; > + break; > + } You need to check first whether the SystemParametersInfoW call succeeded or not, otherwise it will end up translating garbage in WideCharToMultiByte. -- Dmitry.

1 0

lotus notes and wine
by Dan Kegel July 10, 2006

July 10, 2006

http://linux.slashdot.org/linux/06/07/10/121246.shtml says that IBM has Lotus Notes running in Java (on Eclipse!). Two readers commented that they've used the code, and it's uncomfortably bloated. I bet there's still a window of opportunity for Wine here. James Hawkins is fixing the wine bugs that keep Lotus Notes 6.5.3 demo from installing on Wine; if that's done before the Java version of Notes is released (in two weeks), and Notes on Wine becomes easy to set up and run, it's possible we'll get a wad of good publicity out of it...

1 0

Re: [WINED3D 1] Place immediate constants in the same array as global constants.
by H. Verbeet July 10, 2006

July 10, 2006

On 10/07/06, Ivan Gyurdiev <ivg231(a)gmail.com> wrote: > Also note that there's diffrent lifetime rules for immediate constants > in d3d8 and d3d9, and as far as I know we implement one of them wrong. > Here I stick with the current behavior, which I think is d3d9-compliant. Yes. In d3d8 constants set with the def instructions persist between shaders.

2 2