August 2010 - Wine-devel - list.winehq.org

WineConf 2010
by Francois Gouget 25 Aug '10

25 Aug '10

WineConf 2010 is now all set: Where: Hotel Ibis Bastille in Paris When: November, Saturday 20th ~9am to Sunday 21st ~5pm You should book your hotel room directly from the hotel; the sooner the better. Also if you have not booked your plane tickets yet, this is your reminder. All the details should be available on the WineConf 2010 wiki page: http://wiki.winehq.org/WineConf2010 Again, note that most further communication will happen on the wineconf(a)winehq.org mailing list. So subscribe there to be sure not to miss the latest news. This is also the place to go if you have questions, etc. I hope to see you all in Paris! -- Francois Gouget <fgouget(a)free.fr> http://fgouget.free.fr/ RFC 2549: ftp://ftp.isi.edu/in-notes/rfc2549.txt IP over Avian Carriers with Quality of Service

1 0

Re: [PATCH 2/7] d3dx9_36/tests: Added HLSL test suite
by Henri Verbeet 25 Aug '10

25 Aug '10

On 25 August 2010 09:58, Travis Athougies <iammisc(a)gmail.com> wrote: > +static IDirect3DPixelShader9* compile_pixel_shader(IDirect3DDevice9* device, const char *shader, const char *profile,· > +»······»·······»·······»·······»·······»······· ID3DXConstantTable **constants) You still have basic formatting errors like trailing spaces, mixing tabs and spaces, inconsistent "*" placement, etc. > + hres = IDirect3DDevice9_CreateVertexShader(device_ptr, (DWORD*) ID3DXBuffer_GetBufferPointer(compiled), &vshader_passthru); Here's another unnecessary cast. For the record, I'm not going to point out every single flaw in every single patch. If I find an issue in a patch, you should take that as a hint that perhaps you didn't review your patches careful enough before sending them, and go over them again yourself. Anyway, we've decided to implement the shader assembler and compiler in the d3dcompiler dll. I think it makes sense to move any compiler/assembler tests there as well. Note that I'm going to request that d3dcompiler follows the same general style as used in dxgi/d3d10core/d3d10, this is different from the style used in most of d3dx9. Something else. In my opinion an internship should imply some level of guidance/training, not simply "cheap labor". In particular, while ideally you would have caught the issues mentioned in this and previous reviews yourself, whoever has been mentoring you certainly should have, these are basic C / Wine development issues. The process is supposed to look somewhat like this: 1. Carefully write the patch. 2. Carefully review the patch. 3. If you find any issues, carefully fix them and go back to 2. 4. Run the test suite. 5. If it finds any issues, carefully fix them and go back to 2. 6. Send the patch to your mentor for review. 7. If he finds any issues, carefully fix them and go back to 2. 8. If it's still the same day as on 1, consider waiting a day, so you can give the patch a fresh look in the next step. 9. Repeat steps 2 through 5 before sending the patch to wine-patches. 10. Send the patch to wine-patches. 11. If it finds any issues, carefully fix them and go back to 2. Replace "patch" with "patchset" if you're doing more complicated things. Ideally step 6 would produce something like a "Signed-off-by" or "Reviewed-by" tag. If you're at step 11 for the third or fourth time, and there are still problems with the patch, there are structural failures at 1, 2 and 6. Failures at 1 and 2 are somewhat expected for an internship, depending on what kind of standards a specific organization sets, but that's exactly what step 6 is supposed to catch. I.e., when I find an issue in a patch that means at the very least that you didn't catch it when writing the patch, didn't catch it when reviewing the patch the first time, your mentor didn't catch it, and you didn't catch it during review before sending the patch. The things wine-patches reviews are supposed to catch are non-trivial interactions with other parts of the code, not trivial whitespace errors.

1 0

Re: mscoree: improve GetCORSytemDirectory
by testbot＠testbot.winehq.org 25 Aug '10

25 Aug '10

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=4758 Your paranoid android. === WINEBUILD (build) === Patch failed

1 0

Re: d3dx9: Implement D3DXDeclaratorFromFVF. (try 2)
by Henri Verbeet 25 Aug '10

25 Aug '10

On 24 August 2010 05:05, Misha Koshelev <misha680(a)gmail.com> wrote: > + elements[idx].Stream = 0xFF; > + elements[idx].Offset = D3DXGetDeclVertexSize(elements, 0); > + elements[idx].Stream = 0; That's ugly. > + memcpy(Declaration, elements, size * sizeof(D3DVERTEXELEMENT9)); That's unnecessary.

3 4

"Spanish - Modern Sort_Spain.1252" locale on wine
by GOUJON Alexandre 25 Aug '10

25 Aug '10

Hi everyone, I'm trying to fix msvcrt:locale on wine and it appears "Spanish - Modern Sort_Spain.1252" locale is missing. According to msdn (http://msdn.microsoft.com/en-us/library/39cwe7zf%28v=VS.71%29.aspx) "esn" and "spanish-modern" are equivalent which is false [quoted from msvcrt/tests/locale.c] ret = setlocale(LC_ALL, "esn"); ok(ret != NULL || broken (ret == NULL), "ret == NULL\n"); if(ret) ok(!strcmp(ret, "Spanish_Spain.1252") || broken(!strcmp(ret, "Spanish - Modern Sort_Spain.1252")), "ret = %s\n", ret); and ret = setlocale(LC_ALL, "spanish-modern"); ok(ret != NULL || broken (ret == NULL), "ret == NULL\n"); if(ret) ok(!strcmp(ret, "Spanish - Modern Sort_Spain.1252") || broken(!strcmp(ret, "Spanish_Spain.1252")), "ret = %s\n", ret); If I modify dlls/kernerl32/nls/esn.nls LOCALE_SENGLANGUAGE "Spanish" to LOCALE_SENGLANGUAGE "Spanish - Modern Sort" then it'll fix the second test and break the first one. So I'm wondering what to do. Create another nls file based on esn with a different LOCALE_SENGLANGUAGE ? (and change _country_synonyms[] in locale.c) Hard code this (i.e. if .. == "spanish-modern" .. ) ? Other advice ? Thanks in advance

1 0

Re: ddraw [try 3]: SetCooperativeLevel with Full screen + exclusive + normal is authorized
by Henri Verbeet 25 Aug '10

25 Aug '10

On 25 August 2010 08:48, David Adam <david.adam.cnrs(a)gmail.com> wrote: > + if(cooplevel == (DDSCL_NORMAL | DDSCL_EXCLUSIVE) ) So what happens if you set e.g. "DDSCL_NORMAL | DDSCL_EXCLUSIVE | DDSCL_FPUPRESERVE"? You also still have trailing spaces. The way this would conceptually make sense to me is if you always need either NORMAL or EXCLUSIVE, with NORMAL | EXCLUSIVE -> EXCLUSIVE and EXCLUSIVE requiring FULLSCREEN. (Well, except for SETFOCUSWINDOW, which seems to live in a world of its own.) I.e., something along these lines: if (coop & SETFOCUSWINDOW) { ... return; } if (coop & EXCLUSIVE) { if (!(coop & FULLSCREEN)) fail(); ... } else if (coop & NORMAL) { ... } else fail();

1 0

wined3d and visual studio: The remaining uglyness
by Stefan Dösinger 25 Aug '10

25 Aug '10

Hi, I've spent some time to get wined3d to compile with visual studio on Windows. I sent some code fixes needed for that, but a few hacks are remaining. I'm sending them here for discussion. First, why did I look at Visual Studio: *) mingw on windows a bit messy. Cross Compiling all the time even more so *) msvc is the native toolkit, it opens up many other tools. I was interested in VsPerfKit specifically. Basically I set up a Visual Studio process to compile libport, libwine, wined3d, d3d8 and d3d9 out of tree, referencing the code files via relative paths(../../wine/dlls/wined3d/*.c). This was done manually, there's no integration with Wine's native build system. That's hack #1. Unfortunately I can't attach the project right now, it's on a different computer at home. I think we have a script to generate msvc projects for the tests. Maybe the right way is to extend this to the DLLs and programs. I wrote a hand-written config.h file defining a few HAVE_XXX_H constants. That is hack #2 and relates to Alexandre's concern about the libport isnan and isinf implementations. More specific code issues: *) MIDL uses cdecl as the default calling convention for exported functions. Patch 6 in the archive fixes wined3d.idl to work correctly with midl, but this change causes a segfault in widl. *) Msvc complains that -unsigned_int is still unsigned, and the integer overflow just happens to do what you want. The easiest way is to write 0 - unsigned_int, but I'm not sure about it. *) NaN and Inf generation *) msvc doesn't like the predeclaration of **environ. It causes a unknown symbol error at link time.

1 0

Re: ddraw: SetCooperativeLevel with Full screen + exclusive + normal is authorized
by Stefan Dösinger 25 Aug '10

25 Aug '10

Hi, Your CreateSurface tests are missing in this patch(assuming that paulo lesgaz is your alter ego) There's a comment above the SetCooperativeLevel implementation that needs adjustments. > if(cooplevel & DDSCL_NORMAL) > { > - /* Can't coexist with fullscreen or exclusive */ > - if(cooplevel & (DDSCL_FULLSCREEN | DDSCL_EXCLUSIVE) ) > + /* Can't coexist with exclusive only*/ > + if(cooplevel = DDSCL_NORMAL | DDSCL_EXCLUSIVE ) This check doesn't do what you intend it to do. An application may pass NORMAL | EXCLUSIVE | FPUSETUP for instance Also, your test shows that NORMAL | EXCLUSIVE | FULLSCREEN means fullscreen mode. The CreateSurface test passes because the FULLSCREEN and EXCLUSIVE flags are added, but the window is still set up for windowed rendering. In this flag combination you want to enter the fullscreen setup codepath.

1 0

Re: ddraw [try 3]: SetCooperativeLevel with Full screen + exclusive + normal is authorized
by testbot＠testbot.winehq.org 25 Aug '10

25 Aug '10

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=4741 Your paranoid android. === W98SE (32 bit ddrawmodes) === No test summary line found === W7PROX64 (64 bit ddrawmodes) === Timeout

1 0

Re: [3/3] d3dcompiler: Move D3DAssemble tests from d3dx9_36.
by testbot＠testbot.winehq.org 24 Aug '10

24 Aug '10

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=4740 Your paranoid android. === WINEBUILD (build) === Reconfig failed

1 0