April 2015 - Wine-Devel - list.winehq.org

Re: [1/1] dwrite: case insensitive font search
by Nikolay Sivov April 11, 2015

April 11, 2015

On 04/10/2015 10:20 PM, Julius Schwartzenberg wrote: > Running the tests on MS Windows, it turns out that Nikolay Sivov was > correct that it should just be a case insensitive search: > https://www.winehq.org/pipermail/wine-devel/2015-April/107309.html > This allows Samsung Smart View 2.0 to get further with starting up. Hi, Julius. This looks good enough, except subject line needs to be improved a bit to follow a pattern we usually use, something like 'dwrite: Font family lookup should be case-insensitive', but that's not critical at all. Thanks for tracking this down.

1 0

Re: [PATCH 8/9] wined3d: Handle color key dirtification in wined3d_cs_exec_set_texture.
by Henri Verbeet April 10, 2015

April 10, 2015

On 9 April 2015 at 22:48, Stefan Dösinger <stefan(a)codeweavers.com> wrote: > diff --git a/dlls/wined3d/wined3d_private.h b/dlls/wined3d/wined3d_private.h > index eb5c3b0..a888552 100644 > --- a/dlls/wined3d/wined3d_private.h > +++ b/dlls/wined3d/wined3d_private.h > @@ -1068,7 +1068,10 @@ DWORD get_flexible_vertex_size(DWORD d3dvtVertexType) DECLSPEC_HIDDEN; > #define STATE_POINT_SIZE_ENABLE (STATE_FRAMEBUFFER + 1) > #define STATE_IS_POINT_SIZE_ENABLE(a) ((a) == STATE_POINT_SIZE_ENABLE) > > -#define STATE_HIGHEST (STATE_POINT_SIZE_ENABLE) > +#define STATE_COLOR_KEY (STATE_POINT_SIZE_ENABLE + 1) You should handle this in debug_d3dstate() and validate_state_table() as well.

1 0

[RFC] mshtml: Remove nsIDOMWindowUtils::SetCriticalDisplayPortForElement.
by Thomas Faber April 10, 2015

April 10, 2015

Hey Jacek, I just got an MSVC /RTCs stack pointer mismatch error after this call in mshtml's CustomDoc_Release: nsIDOMWindowUtils_CycleCollect(window_utils, NULL, 0); After some investigation I found that the vtable for this class was off by one element. The culprit seems to be SetCriticalDisplayPortForElement. The function got removed here in Gecko: http://sourceforge.net/p/wine/wine-gecko/ci/ed28b621505812d584723bfb389983b… However the latest nsiface.idl in Wine still has it: https://source.winehq.org/git/wine.git/blob/HEAD:/dlls/mshtml/nsiface.idl#l… I did my testing on mshtml 1.7.37 with Wine Gecko 2.34 but it looks like this is still present in the latest code as well. Do you agree with my analysis (i.e. the attached patch) here, or do you think I'm missing something? Thanks! And thanks for maintaining Wine Gecko... this truly opened my eyes to the complexity of that task ;) -Thomas

2 1

Re: [PATCH 2/2] wined3d: Implement color keying in the glsl fragment pipeline (v2).
by Marvin April 10, 2015

April 10, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12746 Your paranoid android. === build (build) === Patch failed to apply

1 0

Re: [PATCH 3/9] wined3d: Implement color keying in arbfp_blit.
by Matteo Bruni April 10, 2015

April 10, 2015

2015-04-09 22:48 GMT+02:00 Stefan Dösinger <stefan(a)codeweavers.com>: > This intentionally discards on the alpha value. To do a proper color > compare we have to migrate fixed function pipeline color keying and > blits at the same time because the color keying conversion changes the > texture contents and can cause the key not to match. > --- > dlls/wined3d/arb_program_shader.c | 68 +++++++++++++++++++++++++++++++++------ > dlls/wined3d/surface.c | 51 +++++++++++++++++++---------- > dlls/wined3d/swapchain.c | 2 +- > dlls/wined3d/wined3d_private.h | 7 ++-- > 4 files changed, 99 insertions(+), 29 deletions(-) > > diff --git a/dlls/wined3d/arb_program_shader.c b/dlls/wined3d/arb_program_shader.c > index 79439f5..f740123 100644 > --- a/dlls/wined3d/arb_program_shader.c > +++ b/dlls/wined3d/arb_program_shader.c > @@ -7539,7 +7549,24 @@ static GLuint arbfp_gen_plain_shader(struct arbfp_blit_priv *priv, > GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)); > > shader_addline(&buffer, "!!ARBfp1.0\n"); > - shader_addline(&buffer, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target); > + > + if (type->use_color_key) > + { > + shader_addline(&buffer, "TEMP color;\n"); > + shader_addline(&buffer, "TEMP compare;\n"); > + shader_addline(&buffer, "PARAM color_key = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY); > + shader_addline(&buffer, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target); > + shader_addline(&buffer, "SGE compare.r, color.a, color_key.a;\n"); > + shader_addline(&buffer, "SGE compare.g, -color.a, color_key.a;\n"); > + shader_addline(&buffer, "MUL compare, compare.r, -compare.g;\n"); > + shader_addline(&buffer, "KIL compare;\n"); > + shader_addline(&buffer, "MOV result.color, color;\n"); If you're checking for an exact match, that doesn't look right. You probably want something like: ... SGE compare.r, color.a, color_key.a; SGE compare.g, -color.a, -color_key.a; MUL compare, compare.r, compare.g; ... I might have misunderstood the point of the shader though, if that's the case then ignore this.

2 1

Re: [PATCH 3/4] d3drm/tests: Add QueryInterface tests for IDirect3DRMFrame{1-3}
by Stefan Dösinger April 9, 2015

April 9, 2015

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Am 2015-04-09 um 16:11 schrieb Aaryaman Vasishta: > +static void test_frame1_qi(void) > +{ > ... > + test_qi("frame_qi", (IUnknown *)frame, &IID_IDirect3DRMFrame, tests, sizeof(tests) / sizeof(*tests)); > ... > +static void test_frame2_qi(void) > +{ > ... > + test_qi("frame_qi", (IUnknown *)frame2, &IID_IDirect3DRMFrame2, tests, sizeof(tests) / sizeof(*tests)); > ... > +static void test_frame3_qi(void) > +{ > ... > + test_qi("frame_qi", (IUnknown *)frame3, &IID_IDirect3DRMFrame3, tests, sizeof(tests) / sizeof(*tests)); With this it won't be possible to find out which test failed if something fails. Please use more informative names for the test_name parameter of test_qi, e.g. "frame1_qi", "frame2_qi" and "frame3_qi". It is a bit unfortunate that you have to duplicate the big structure 3 times. You could in theory share it between the different functions, but then you'd have to lie about which interface you pass to test_qi, and I am not sure if this is nicer. If you can't find a better solution keep it the way it currently is. @Alexandre: The reason why I want to test all interface versions is our experience with ddraw. It has plenty of interface-specific behaviors, some of them depend on which version is called, and some of them depend on which version was used to create the object. (see struct ddraw_surface::version, struct ddraw::version and struct d3d_device::version, as well as various thunks). For this particular test the relevant information is that all interfaces can be queried from all object versions and that there are no refcounting differences, which could be introduced by the presumably aggregated IDirect3DRMObject2 interface. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAEBAgAGBQJVJvBaAAoJEN0/YqbEcdMwXYEP/2k4k4qdiGA2sxdv0S6mTaZW RXEPGDY1BqtDyOlw9EzwK+EmQTnE6B4zC25AM55yFAQjaIs8UyAk/TN3b0K+H9xz ZJJIgqGe2x+yIrzKqT70lGqUltYMaENHEw3oh8GdkYnIIK97bG7Y7vK7VZKIDPlh vfxcwG990di77KyPTt3eI4w6BQp4SyyEUxNC5vCy213ML+sF9XjGV+WPkV0UStcX nNCLRcszExminklJgEj4h1S6qcxNVIwJks8vCrgRx9isJfmbK8g5GrAzod3iTSES K1ccZWxGFbXupiT8Je6K5Te6b2PlHZtBVVHb5ndeYQdZJmwZIrTFbE3LjX3o2Vyt m0DvbPzAzNNJO7jREtWVSDXmg2OIaxsaIlMpH+v7VU2csqLNQZCijJ9/q3hN2qLf rYIS7i2fAmWDUoE6uFDOs+VJKDx88OljZaBTV4VFAHYwfeiqr3KXVdYZ3lciViif gjfno0N20inhG/ydjmJOn+tMipj2p1iffHDMaIlXVFhxitTj2Kk/bYp5b65T5q0C dn99clpZKU8xKg8w1I6MQv03CvMla0Z5k5d01H6YCW0YDSHtSnxfXeykxN2PeCGZ wYkTxF/KeSXhg3tLkl7StD1aRJsa5lv98FnJmQA2UWDNlcPfHcvv0Dam8ZsRohEr pPvj1H5GMcaaDGhfCOiZ =Eore -----END PGP SIGNATURE-----

1 0

Re: [PATCH] shell32: Add more knownfolders
by Marvin April 9, 2015

April 9, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12738 Your paranoid android. === w2000pro (32 bit shellpath) === The test timed out

1 0

Re: TestBot job 11764 results: [1/5] comctl32: Move TaskDialogIndirect function and its test to separate files (try 3)
by Joachim Priesner April 9, 2015

April 9, 2015

Hi all, this patch series (implementing a basic version of TaskDialog/TaskDialogIndirect) has vanished from the patches list without comment. I'd appreciate any hint on what I did wrong. Thanks Joachim Am Donnerstag, 26. Februar 2015 schrieb Marvin: > VM Status Number of test failures > build completed > w2000pro completed 0 > wxppro completed 0 > w2003std completed 0 > wvistau64 completed 0 > w2008s64 completed 0 > w7u completed 0 > w8 completed 0 > wvistau64 completed 0 > w2008s64 completed 0 > w2000pro completed 0 > wxppro completed 0 > w2003std completed 0 > wvistau64 completed 0 > w2008s64 completed 0 > w7u completed 0 > w8 completed 0 > wvistau64 completed 0 > w2008s64 completed 0 > > === build (build) === > No build failures found > > === w2000pro (32 bit taskdialog) === > No test failures found > > === wxppro (32 bit taskdialog) === > No test failures found > > === w2003std (32 bit taskdialog) === > No test failures found > > === wvistau64 (32 bit taskdialog) === > No test failures found > > === w2008s64 (32 bit taskdialog) === > No test failures found > > === w7u (32 bit taskdialog) === > No test failures found > > === w8 (32 bit taskdialog) === > No test failures found > > === wvistau64 (64 bit taskdialog) === > No test failures found > > === w2008s64 (64 bit taskdialog) === > No test failures found > > === w2000pro (32 bit misc) === > No test failures found > > === wxppro (32 bit misc) === > No test failures found > > === w2003std (32 bit misc) === > No test failures found > > === wvistau64 (32 bit misc) === > No test failures found > > === w2008s64 (32 bit misc) === > No test failures found > > === w7u (32 bit misc) === > No test failures found > > === w8 (32 bit misc) === > No test failures found > > === wvistau64 (64 bit misc) === > No test failures found > > === w2008s64 (64 bit misc) === > No test failures found

2 1

Re: dpnet: Convert Address components to use a list
by Stefan Dösinger April 9, 2015

April 9, 2015

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Am 2015-04-08 um 09:10 schrieb Alistair Leslie-Hughes: > +{ > + if(This->comp_count == This->comp_array_size) > + { > + struct component **temp; > + > + temp = heap_realloc(This->components, sizeof(struct component *) * (This->comp_array_size*2) ); Style hint: You can use sizeof(*temp) here, that way the alloc code doesn't need to be changed when you change the type of temp. This is even more helpful when the declaration of the pointer is in a different file, like in the heap_alloc call that allocates client->components. > + if(This->components) > + { > + This->components[This->comp_count] = item; > + This->comp_count++; > + return TRUE; > + } Can this ever be NULL for a legitimate reason? Note that the heap_realloc call already assumes This->components is not NULL. > @@ -540,8 +596,8 @@ HRESULT DPNET_CreateDirectPlay8Address(IClassFactory *iface, IUnknown *pUnkOuter > > client->IDirectPlay8Address_iface.lpVtbl = &DirectPlay8Address_Vtbl; > client->ref = 1; > - > - list_init(&client->components); > + client->comp_array_size = 4; > + client->components = heap_alloc( sizeof(struct component *) * client->comp_array_size ); Please add error checking here. (The NULL check in add_component doesn't help if this fails. You'll just get a half-working address object where AddComponent always fails) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAEBAgAGBQJVJkU+AAoJEN0/YqbEcdMwYRYP/0Vzb0FQDV6xBWc66VFjECxa dSyWsm34yksURXd9H7ayT+cGHQiyfq9dP5MIEL3Lrf7Npiqy7b3DV2NYH1X3foL8 NZoU0R6BSq1XUkqcEWFnXRDIHWs8f0Q1kNyRcUfsuXO4c6vAQOde7CdRtMHuLEy5 DOZjnnSbfAjoKq9Xk3JGJOs1juceoNumojpYE6wtbwDM5Q3Z8DOAdDEvv9qCw9SU Gziuis+LLN2yFDb+cONIKApy/XS/zs2imFS4z/wlKAEu72aGib/64j/d7j48jIxl C7ojDWjWGQj1CDSF6Wn0X80TV5Xjgolgw3ndyPFvqResCk4+UMeoyPaBXLLmozcu VaNtd4sGVTTAe7ivfKVaZx1K0gR8mezsFXEZ2jBfZlTOGHIRiImQ84ORZvahWr7t HYh0ZdDddzNOEbJVAlvxMrB/HLL+ODAOZJaElN6QXyWPBYFIb5AH+VwXeBccahSM xg8xkZLQIPqgDyQaI4TPDCboU5VzptTtl3uKJSy4QX+gV23IFXT9AUnPrWLas+3J qW0pX1tljdTxrVqU3mGvvE3KnUsHyEvx4sBVNWOzEagnSv4TQiabNlkyPfQmJsiH WE/Lx2WTWV9ucLfsGmCNbx+Cu52hDNj1R93hqgpSgAvvUjFeGjxTy4SQFD17ig6M ZFkE66v0L53cG4lQASGw =FqJl -----END PGP SIGNATURE-----

1 0

Re: [1/2] ws2_32/tests: Add tests about retrieving the local address of an interface bound socket
by Marvin April 8, 2015

April 8, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12730 Your paranoid android. === w2008s64 (64 bit sock) === No test summary line found

1 0

dwrite font family issue
by Julius Schwartzenberg April 8, 2015

April 8, 2015

Hi, On #winehq there was a discussion on Monday as I was trying to get Samsung Smart View 2.0 to run. (DotNOT 4.5 application requiring Windows 7/8.) I would report back on my findings. Anastasius Focht figured it was running into a problem discovering a suitable font (family). Here is a patch I came up with which lets the program continue. I'm not sure whether this is a correct fix however. What is happening is that in 'dlls/dwrite/font.c: collection_find_family' the name argument is completely uppercase. This is then compared to buffer which will contain a similar string at some point but with regular casing. Before that however I see the string passed with its original casing like this: 0009:Call KERNEL32.LCMapStringEx(00000000,00000200,0172c79c L"Arial",00000005,0172c7b4,00000005,00000000,00000000,001908b8) ret=00aaa471 0009:fixme:nls:LCMapStringEx unsupported lparam 1908b8 0009:Ret KERNEL32.LCMapStringEx() retval=00000005 ret=00aaa471 0009:trace:dwrite:dwritefontcollection_FindFamilyName (0x1966e8)->(L"ARIAL" 0x33be0c 0x33be08) I added two TRACE calls myself to print the values of name & buffer which gives: 0009:trace:dwrite:collection_find_family name=L"ARIAL" and later: 0009:trace:dwrite:collection_find_family buffer=L"Arial" The call to KERNEL32.LCMapStringEx may suggest that the casing is actually lost unintentionally. Thanks, Julius

2 1

Re: [PATCH 5/5] wined3d: Set color keys through the command stream.
by Henri Verbeet April 8, 2015

April 8, 2015

On 6 April 2015 at 00:28, Stefan Dösinger <stefan(a)codeweavers.com> wrote: > +#define WINED3D_CKEY_REMOVE 0x80000000 > + It seems unfortunate to define part of the flags in include/wine/wined3d.h and part of the flags in dlls/wined3d/cs.c. You'd probably also want to mask out private flags in wined3d_cs_emit_set_color_key(). Ultimately though, why do you need this? Most of the CS ops operate on a wined3d_state structure, which is mirrored between the CS and the device, and (in theory) functions get passed the appropriate one depending on whether they're running from the CS or not. This one is different in that it operates on a texture, and anything that would need to read the color key would then need to either wait for the texture to become idle, or run from the CS itself. Perhaps that's ok, but it's more of a fundamental design decision than this commit makes it seem.

2 3

Re: [PATCH 2/2] shell32: Add more knownfolders
by Marvin April 8, 2015

April 8, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12722 Your paranoid android. === build (build) === Patch failed to apply

1 0

Re: winecfg: improve user experience by adding all Windows versions and accompanying error message
by Jason Overland April 8, 2015

April 8, 2015

Is there anything I can do to improve this patch? Thanks, Jason On 03/26/2015 10:03 PM, Jason Overland wrote: > Hi, > This patch is a resubmission of and supersedes my previous patch "winecfg: add note about 32 bit prefixes to dialog". Thanks to everyone who provided feedback on that one. > > In this patch, winecfg's Application settings tab now includes all Windows versions in the combobox. If the user is running from a 64 bit prefix and selects a version of Windows that requires a 32 bit prefix, the selection will revert to the previous selection and a modal dialog containing an error message will appear: > > """ > This Wine prefix is built to support 64 bit applications. Because the Windows version you selected does not support 64 bit applications, this prefix cannot be configured to mimic that version of Windows. > > Please refer to "How do I create a 32 bit wineprefix on a 64 bit system?" in the "Using Wine" section of the Wine FAQ (http://wiki.winehq.org/FAQ) for instructions on how to create a 32 bit Wine prefix. > """ > > In addition, I noticed two File Open dialogs in winecfg which should be modal but are not modal. I made these modal by setting the appropriate owner window. While a separate issue, this seemed like a trivial enough change to include in this patch. >

3 3

Re: [1/3] ws2_32/tests: Add some more select() related tests
by Marvin April 8, 2015

April 8, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12715 Your paranoid android. === w7u (32 bit sock) === sock.c:7148: Test failed: bytesReturned isn't supposed to be 1 sock.c:7172: Test failed: GetOverlappedResult failed, error 0

1 0

Re: dinput: Add JoystickID to the Instance Name
by Dmitry Timoshkov April 7, 2015

April 7, 2015

Aric Stewart <aric(a)codeweavers.com> wrote: > + if (SUCCEEDED(IDirectInputDevice2_GetProperty(iface, DIPROP_JOYSTICKID, (LPDIPROPHEADER)&pd))) > + index = pd.dwData; Please find a way to avoid the cast. > + if (SUCCEEDED(IDirectInputDevice2_GetProperty(iface, DIPROP_JOYSTICKID, (LPDIPROPHEADER)&pd))) > + index = pd.dwData; Same here. > + snprintf(buffer, MAX_PATH, "Joystick %d", index); 'index' is unsigned, sizeof() instead of a hardcoded value is better. -- Dmitry.

1 0

Problem receiving UDP broadcast on wxppro and w2008s64
by Bruno Jesus April 7, 2015

April 7, 2015

Hi, Francois, I hope you can help me in this situation. The patch attached to bug 31994 [1] (compiled version at [2]) runs fine in most testbot machines (tested dozens of times) and my personal XP machine, but for some reason it doesn't work in xppro and w2008s64. The test creates 2 UDP sockets and send a broadcast from socket A to the network where B is bound, but the broadcasted packet never is received by B resulting in test timeout due to recvfrom never getting data. According to [3] all testbots share similar IP addresses like and I can't imagine a reason for them not to work: wxppro: testing broadcast for interface IP 10.42.42.134 (don't work) w2003std: testing broadcast for interface IP 10.42.42.140 (works) wvistau64: testing broadcast for interface IP 10.42.42.138 (works) w2008s64: testing broadcast for interface IP 10.42.42.216 (don't work) Any ideas? I wonder if it could be a firewall. The test also works fine in NT4. [1] https://bugs.winehq.org/attachment.cgi?id=51182 [2] https://testbot.winehq.org/GetFile.pl?JobKey=12594&StepKey=1 [3] https://testbot.winehq.org/JobDetails.pl?Key=12591 Best Regards, Bruno

2 5

Re: locale: [1/1] patch - 0x10000000 flag is NORM_STOP_ON_NULL (plus test fix)
by Marvin April 7, 2015

April 7, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12697 Your paranoid android. === w2000pro (32 bit locale) === locale.c:1776: Test failed: Without NORM_STOP_ON_NULL flag - strings `don\0a` (5) and `don\0b` (5) expected CSTR_LESS_THAN, got 2

1 0

CoInternetParseUrl and concepts explanation.
by Philippe Groarke April 6, 2015

April 6, 2015

Hello, I am currently exploring Wine code base looking for low hanging fruits I can contribute to. I think CoInternetParseUrl and other URLMon stubs seemed easy enough I want to try and implement some. I also see many URLMon stubs when using Wine for games and applications. In CoInternetParseUrl, you use a switch case depending on the parameters passed. These in turn call specific methods for each parameter, which most do the same thing, call IInternetProtocolInfo_ParseUrl with the desired parse option . This function is defined in urlmon header. It seems to call a saved This on runtime, derefencing a saved (?) virtual table calling ParseUrl. It is also implemented in the header (static FORCEINLINE) and also calls the ParseUrl. There are some STDMETHODCALLTYPE pure virtual methods inside a IInternetProtocoleInfo class. A struct virtual table IInternetProtocolInfoVtbl O_o I cannot find the actual ParseUrl implementation, though in misc.c, there are many references to CHECK_CALLED(ParseUrl) and SET_EXPECT(ParseUrl). I have no clue what is going on here to be completely honest. Finally, in misc.c there are some pCoInternetParseUrl and CoInternetParseUrl tests. But nothing that seems to implement what I need. My question is, why go through all these hoops instead of implementing CoInternetParseUrl inside internet.c. I am sure there is good reasoning behind this, I would just like to understand why. Also, where do I implement ParseUrl or IInternetProtocolInfo_ParseUrl? Thank you for any help in understanding what is going on here. Good day, Philippe

2 1

Re: [PATCH 2/4] ntdll/test: Add ntdll test function test_file_full_size_information
by Nikolay Sivov April 5, 2015

April 5, 2015

On 04.04.2015 10:03, Jianqiu Zhang wrote: > + trace("TotalAllocationUnits %08x %08x\n", fsi.TotalAllocationUnits.HighPart, fsi.TotalAllocationUnits.LowPart); > + trace("CallerAvailableAllocationUnits %08x %08x\n", fsi.CallerAvailableAllocationUnits.HighPart, fsi.CallerAvailableAllocationUnits.LowPart); > + trace("ActualAvailableAllocationUnits %08x %08x\n", fsi.ActualAvailableAllocationUnits.HighPart, fsi.ActualAvailableAllocationUnits.LowPart); > + trace("SectorsPerAllocationUnit %08x\n", fsi.SectorsPerAllocationUnit); > + trace("BytesPerSector %08x\n", fsi.BytesPerSector); > + Hi, Jianqiu. This is not exactly a test, as you're just printing a result. Same thing with patch 3/4. Also to print LONGLONG values you use wine_dbgstr_longlong() probably.

2 3

Re: [PATCH 1/2] ws2_32: Fix compile on OS X.
by Bruno Jesus April 5, 2015

April 5, 2015

On Sun, Apr 5, 2015 at 4:07 PM, Socapex <philippe.groarke(a)gmail.com> wrote: > If HAS_IPX isn't defined, compilation will throw error since > WS_IPX_* aren't declared. > --- > dlls/ws2_32/socket.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/dlls/ws2_32/socket.c b/dlls/ws2_32/socket.c > index 09c8416..bc4d40a 100644 > --- a/dlls/ws2_32/socket.c > +++ b/dlls/ws2_32/socket.c > @@ -365,6 +365,7 @@ static inline const char *debugstr_sockopt(int level, int optname) > DEBUG_SOCKLEVEL(WS_NSPROTO_IPX); > switch(optname) > { > +#ifdef HAS_IPX > DEBUG_SOCKOPT(WS_IPX_PTYPE); > DEBUG_SOCKOPT(WS_IPX_FILTERPTYPE); > DEBUG_SOCKOPT(WS_IPX_DSTYPE); > @@ -372,6 +373,7 @@ static inline const char *debugstr_sockopt(int level, int optname) > DEBUG_SOCKOPT(WS_IPX_MAXSIZE); > DEBUG_SOCKOPT(WS_IPX_ADDRESS); > DEBUG_SOCKOPT(WS_IPX_MAX_ADAPTER_NUM); > +#endif > } > break; Hi, unfortunately I don't have OSX so I only test my patches in Linux and FreeBSD, so sometimes these problems happen, sorry. I think it's better to simply remove the #ifdef around wsnwlink.h [1], this will not reduce the debug capabilities of debugstr_sockopt. Since wsnwlink.h is an internal include it will always be available, can you it, please? [1] From socket.c: 160 #ifdef HAS_IPX 161 # include "wsnwlink.h" 162 #endif Best Regards, Bruno

2 1

Re: [PATCH 2/2] ws2_32: Remove trailing white-spaces.
by Stefan Dösinger April 5, 2015

April 5, 2015

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi, Am 2015-04-05 um 21:07 schrieb Socapex: > -/* Note: EAI_NODATA is deprecated, but still > +/* Note: EAI_NODATA is deprecated, but still Generally we don't commit whitespace-only changes because they cause problems with git blame (though git blame has been improved in that regard) and merges. Stefan -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAEBAgAGBQJVIYsNAAoJEN0/YqbEcdMwLUEP/jy+kbilEFxb+W5HrbvV5x4Y GKTrrE0sBYv3RmcLEdCEmivTt5BOHerIxmoy5Li2wsH/S1FglwcdVETTBBG9v8tV 0dEgJ/p3GAB8RPQTRtvHTyr5aiUqOQg0HNW08puy/OHF3Zr3BTiFfzNLfKcVRTj5 Vvp28AwThDTqjG85ajOezBfxw/H6g4D8GtmsETYr2NYQZD9lt+C28op9K5hxSuA9 E05Af00Rxye7qb8ry715cAbBg2nO/MDygRt5iYpCX6Rd8fmRuM5O27MJLDQkq/pU GDKilQuWmiwZPEQcs3PKbOql3xditgvLrGKk4jkYAiXfTtaxIvwz9GDA6KVXynom Hv/m2BFKV1odoBYkDWlZZ42XVj1aZFKFZIknVTddDoRN/AVfTBO3cz5TPWOwr0q3 jGWOBT31ksWStdXOaQsPyUjX7ZKws2T2ZTRZTKoY0o/c/LYAXIgqz0VNwVnVaIs2 x+xlrdEpg58E1JuLBL93F0z3Pe+YQXqdv3RhWNSgzosK3VmaEC8+GzWBUR8gQMlJ UW0WRSycE9AktMGJv1+Grf/PIk8pHRN+4veL4065M5a4jaeclghx/HxqswuI8WCV Ty1qe5eHeNbD/9cEzbRSHBOV5g/V+xmvgzY5M0DJ2TTSMchmkJzC00m0khRMWhaj JMLgp7jnoWasLLA2ufZH =MEUF -----END PGP SIGNATURE-----

1 0

Re: [PATCH 1/2] ws2_32: Fix compile on OS X.
by Stefan Dösinger April 5, 2015

April 5, 2015

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi, You have to use your real name, not a nickname. We need this to be able to show where our code comes from. With best regards, Stefan Am 2015-04-05 um 21:07 schrieb Socapex: > If HAS_IPX isn't defined, compilation will throw error since > WS_IPX_* aren't declared. > --- > dlls/ws2_32/socket.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/dlls/ws2_32/socket.c b/dlls/ws2_32/socket.c > index 09c8416..bc4d40a 100644 > --- a/dlls/ws2_32/socket.c > +++ b/dlls/ws2_32/socket.c > @@ -365,6 +365,7 @@ static inline const char *debugstr_sockopt(int level, int optname) > DEBUG_SOCKLEVEL(WS_NSPROTO_IPX); > switch(optname) > { > +#ifdef HAS_IPX > DEBUG_SOCKOPT(WS_IPX_PTYPE); > DEBUG_SOCKOPT(WS_IPX_FILTERPTYPE); > DEBUG_SOCKOPT(WS_IPX_DSTYPE); > @@ -372,6 +373,7 @@ static inline const char *debugstr_sockopt(int level, int optname) > DEBUG_SOCKOPT(WS_IPX_MAXSIZE); > DEBUG_SOCKOPT(WS_IPX_ADDRESS); > DEBUG_SOCKOPT(WS_IPX_MAX_ADAPTER_NUM); > +#endif > } > break; > > -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAEBAgAGBQJVIYmzAAoJEN0/YqbEcdMwmcMP/0UWQMONOFklXHtI9ibTn5rU Y8idfpQRQEj7s4UqKT5XdHbbGsmGa24byBybhWf54z/ykTqcwvcun4xDUs0/pmCK WRWx9hlgGd4xPSGrnum9XC/tnywjyP5s8eyRUU4CeMmBOIwauj15iePd/v4W+BA+ iHTyY4rjzyoyatmZURoaZaFFj/n5cLoqyqQ+fTNeP22z+Wjl7N76wXYbsylRsatB NKe8bAdBAQZL8LNngjpCVXW0PsKaH88LIdVPZpvI0T+hH8+LYRToKbfqCXtwkwYb LuE9RCdfAcRbIQTCe4FrpfduGUtQN2Nlxu7pIOH9GghWITTjIZn/XV0VYSBVDrNc d2zZ0lmyEWPU6H5RU/dzfhkJMhKP7RzZkd3j7yDKsjBlC651ysDT6ev/HpX2Jx7O aYf70/48Qk5BhbEZblQuPrssz/p4ErXHz2UQRGodqLjJm+ATSwkSlwsAGZ8iyjn0 3JQuL3BDDSF3CAjfQBLfcRXS4tNgNd58rr5dfoLmOhkKG+LIbYIQ55104UG03xDc 5LFubLU8i8r89OBR3yeZXLIX3BdR3oxANVDoPjjbbJiAKIoej5yJOCyLb8cPbchN CufNxeiY+wrKNWtuW6UXONaEJqt2MpcY0HUV70s9qRXZ2iZoPKDFkR5mPudR7qbj ZIUAqSKGQcWwmCw+aEk+ =ShND -----END PGP SIGNATURE-----

1 0

Re: [PATCH 3/3] d3drm/tests: Split tests between different versions, initially adding frame QueryInterface tests.
by Stefan Dösinger April 5, 2015

April 5, 2015

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi, This series looks good to me, and this time I ran the tests on Windows and Wine :-) Am 2015-04-03 um 14:25 schrieb Aaryaman Vasishta: > +struct qi_test > +{ > + REFIID iid; > + REFIID refcount_iid; > + HRESULT hr; > + BOOL refcount_todo; > +}; refcount_todo is always FALSE in your tables, so you might as well remove it. I don't think it's necessary to resend the patch because of that though. You can just remove it when you add the vtable inheritance test. If you expect future tests to need refcount_todo you can also keep the field. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAEBAgAGBQJVIWyeAAoJEN0/YqbEcdMwZPgP/AoYCerTMFiXH9lYnmueThK/ UPGcBOLhdU0dT0U8lI0ZVzX7mbvtvjoVFUn/cAGgBZTUH175BK3SHbuIljJ+R954 +xQtRda5QrEu759UIkuc6v+REUtNQWzPtWLRX8EhZLhTEJaTEdsGeGUKAKRwGFmc UsMEff5r1gIIFgFDSlwSm1daAduOwkNcnzFwbxTbm7Kg/qCn50nSTfWj4Nk/sYYl kr6vzKH5EAt5nL9tvtXT7q8YDu+e2TC96B2xzoseRnS3st44nAgqjSb9E7fB05Cu Lem/zEEoGJI3mrLWpoDVDoYFFaHPifoqRgyf6EcT9ID0vUmJhM97WgBWw0+b8fFR yxJ/nsBol56Rb4xssOrXINDfF2E+7S9RMu/oxSbVbW8GV3vN4ssgZ5/Q8itmCdJd u7Kyfx/OS72xE9rrUXIQrS0EM7yFq1j6cZa/xv6xGzO0SpeIbWrB2Pnujvr5GTt/ KkPqJVoiFr8VOaZfK2SQreGTaBK4f6rDdW+jkfnQLU38uVSmDTowdO3oD6aPCy2s mgB8JmhBuDsU5RAWoA61PzTtDPmdWLhmHZJWfLC4PG9/IN0m3G3JhFUAUy3qr8UZ 543kGtR21tmPHLKXc+Bt20B0j+RY0sxu8CvVJPb3qa7VmDP+c3thnjkgng1ZJo0F CrXo+6/IKapzItlvTFIE =ssTG -----END PGP SIGNATURE-----

1 0

Re: locale: [1/1] patch - 0x10000000 flag is NORM_STOP_ON_NULL (marvins test fix)
by Marvin April 4, 2015

April 4, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12662 Your paranoid android. === w2000pro (32 bit locale) === locale.c:1776: Test failed: Without NORM_STOP_ON_NULL flag - strings `don\0a` (5) and `don\0b` (5) expected CSTR_LESS_THAN, got 2 === wvistau64 (32 bit locale) === locale.c:2010: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2014: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2018: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:2022: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2028: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2032: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === w2008s64 (32 bit locale) === locale.c:2010: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2014: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2018: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:2022: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2028: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2032: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === w7u (32 bit locale) === locale.c:2010: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2014: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2018: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:2022: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2028: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2032: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === w8 (32 bit locale) === locale.c:2010: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2014: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2018: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:2022: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2028: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2032: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === wvistau64 (64 bit locale) === locale.c:2010: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2014: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2018: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:2022: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2028: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2032: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === w2008s64 (64 bit locale) === locale.c:2010: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2014: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:2018: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:2022: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2028: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2032: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0

1 0

Re: [PATCH 1/7] advapi32: Initialize buffer length to zero in LsaLookupSids to prevent crash.
by Nikolay Sivov April 4, 2015

April 4, 2015

On 04.04.2015 17:22, Qian Hong wrote: > > (*Names)[i].Name.Buffer = NULL; > + (*Names)[i].Name.Length = 0; > + (*Names)[i].Name.MaximumLength = 0; Maybe it's better to hide this in RtlInitUnicodeStringEx(..., NULL) ?

2 1

Re: mshtml: Removed return to ensure AddRef is called on an internal interface
by Piotr Caban April 4, 2015

April 4, 2015

The patch is wrong, nsCycleCollectionISupports doesn't support AddRef/Release.

4 6

Re: locale: [1/1] patch - 0x10000000 flag is NORM_STOP_ON_NULL
by Marvin April 4, 2015

April 4, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12661 Your paranoid android. === wvistau64 (32 bit locale) === locale.c:1985: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1989: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1993: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:1997: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2003: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2007: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === w2008s64 (32 bit locale) === locale.c:1985: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1989: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1993: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:1997: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2003: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2007: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === w7u (32 bit locale) === locale.c:1985: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1989: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1993: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:1997: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2003: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2007: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === w8 (32 bit locale) === locale.c:1985: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1989: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1993: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:1997: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2003: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2007: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === wvistau64 (64 bit locale) === locale.c:1985: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1989: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1993: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:1997: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2003: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2007: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 === w2008s64 (64 bit locale) === locale.c:1985: Test failed: NORM_STOP_ON_NULL passed, though its equal to length autodetect (-1) - strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1989: Test failed: NORM_STOP_ON_NULL is not set, strings considered as null terminated - so strings `don\0at` (-1) and `don\__` (-1) expected CSTR_EQUAL, got 0 locale.c:1993: Test failed: NORM_STOP_ON_NULL passed, and lengths are defined exlicitly - so strings `don\0at` (6) and `don\__` (6) are expected CSTR_EQUAL due to flag, got 0 locale.c:1997: Test failed: NORM_STOP_ON_NULL passed, and length is defined exlicitly for one string - anyway strings `don\0at` (6) and `don\__` (-1) are expected CSTR_EQUAL due to flag, got 0 locale.c:2003: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, though its equal to length autodetect (-1) - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0 locale.c:2007: Test failed: NORM_STOP_ON_NULL | NORM_IGNORECASE passed, and lengths are defined explicitely - so strings `don\0at` (-1) and `dOn\At` (-1) expected CSTR_EQUAL, got 0

1 0

Re: [PATCH 7/7] advapi32/tests: Add more tests for LsaLookupSids and LsaLookupNames2.
by Marvin April 4, 2015

April 4, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12659 Your paranoid android. === w2003std (32 bit lsa) === lsa: unhandled exception c0000005 at 7C81BB52 === wvistau64 (64 bit lsa) === No test summary line found === w2008s64 (64 bit lsa) === No test summary line found

1 0

Re: msxml3: Explicitly list acceptable libxml2->msxml node types for unknode (try 2)
by Jacek Caban April 4, 2015

April 4, 2015

Hi Nikolay, On 01/04/15 22:41, Nikolay Sivov wrote: > + switch (This->node.node->type) > + { > + case XML_ELEMENT_NODE: > + case XML_ATTRIBUTE_NODE: > + case XML_TEXT_NODE: > + case XML_CDATA_SECTION_NODE: > + case XML_ENTITY_REF_NODE: > + case XML_ENTITY_NODE: > + case XML_PI_NODE: > + case XML_COMMENT_NODE: > + case XML_DOCUMENT_NODE: > + case XML_DOCUMENT_TYPE_NODE: > + case XML_DOCUMENT_FRAG_NODE: > + case XML_NOTATION_NODE: Most of those types have their own IXMLDOMNode implementations, so they should never be the case in here. Also, ideally we shouldn't even create unknode object for types that are not supported, so it should be fixed in create_node or its callers. Is it possible in the case that you're fixing? Jacek

2 1

Re: [PATCH 3/4] ntdll/test: Add test_file_size_information
by Marvin April 4, 2015

April 4, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12656 Your paranoid android. === w8 (32 bit file) === file.c:397: Test failed: mismatched file size for L"WindowsUpdate.log"

1 0

Re: ntdll: Return STATUS_SUCCESS from SetInformationJobObject
by Nikolay Sivov April 4, 2015

April 4, 2015

I don't think it's a good idea to always claim success. It might be better to actually store those UILIMIT flags (probably by extending set_job_limits ). Alternatively if you don't want to store those for some reason you can return success for this class only while keeping a fixme, but not for all unimplemented classes.

2 1

Re: [PATCH 2/4] ntdll/test: Add ntdll test function test_file_full_size_information
by Marvin April 4, 2015

April 4, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12655 Your paranoid android. === w8 (32 bit file) === file.c:397: Test failed: mismatched file size for L"WindowsUpdate.log"

1 0

Re: [PATCH] gdiplus: Avoid unnecessary pre-multiplied alpha conversions in GdipDrawImagePointsRect
by Vincent Povirk April 3, 2015

April 3, 2015

Looks good to me.

1 0

Re: [PATCH] comctl32: Do not resize windows when giving an updown control the same buddy.
by Anton Baskanov April 3, 2015

April 3, 2015

Hello! Have you verified this behavior on native comctl32? I've tried to send UDM_SETBUDDY with the same HWND repeatedly, and the buddy is resized every time. I've tested it on Windows 7 x64 with both version 5 and 6 of comctl32. On Wed, Apr 1, 2015 at 4:02 AM, Mark Harmstone <hellas(a)burntcomma.com> wrote: > When an updown control receives UDM_SETBUDDY with the HWND of its existing > buddy, it should do nothing. At present Wine resizes the buddy as it > would do initially, resulting in the textbox getting progressively smaller > every time UDM_SETBUDDY is sent. > > --- > dlls/comctl32/updown.c | 3 +++ > 1 file changed, 3 insertions(+) > > > > >

2 2

Re: [PATCH 4/4] d3drm/tests: Add vtable comparison tests in test_frame_qi
by Stefan Dösinger April 3, 2015

April 3, 2015

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Am 2015-04-02 um 18:36 schrieb Aaryaman Vasishta: > REFIID iid; > REFIID refcount_iid; > + REFIID vtable_iid; > HRESULT hr; > BOOL refcount_todo; > + BOOL vtable_equal; > + BOOL vtable_todo; > }; I think this is more complicated than it needs to be. All you should need is the vtable_iid field. Then, in the inner loop do something like if (IsEqualGUID(tests[i].vtable_iid, tests[j].vtable_iid)) ok(iface1 == iface2, ...); else ok(iface1 != iface2, ...); I may be missing something though. If you want to compare two Win32 BOOLs for equality you have to be careful: BOOL a = 1; BOOL b = 2; if (a) printf("this is true.\n"); if (b) printf("this is also true.\n"); if (a == b) printf("oops, this is false.\n"); For something like that you can use if (!a == !b) printf("yay, this is true!\n"); IsEqualGUID(a, b) is defined as !memcmp(a, b, sizeof(GUID)), so IsEqualGUID() == (ptr1 == ptr2) should be OK though. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAEBAgAGBQJVHakvAAoJEN0/YqbEcdMwPPYP/08CiwYXURzl7Ias7EKvZGO8 QlYjp4CSEorj44LMOMPMELbN+50NJOngxhXA0mecIhZYA9Dgd5NZtVvDWBIPPu/B tXe+0+Xw/EvzubfzMYjOX5/lOKm400CW3fPEbxwYrzdFOSmeNhclZpfD+9jjuekG Lj2XUSk/TAi1B8NlD0x0LE7P6gGb61rg7jeJNZGgfPYMsuAoB2OtekvbIvpbCZmb 3WlOXtzmqVMHjsEsw47i8pdL/DwCFFfpE0xWWdhfgCChFUG/Ul/V56ltN5ERtudO j+jKtY5oNhunIyALwIifwhYNq/bY9CdfovIOI7sNfM3bXwSCZGEk+5Oq3/dnd4rP /Ejp+Kd5xk4SDMC3kyE6VucZkyFD58plgEDNlgSXx2ND/o3Mw91PTDaGj5xT0Fy1 Enk6oqORgXQhvtUaDH5Y5rOyh+AMquWhhN5N43ymT0UW40D4b6UBNSSwuQeqKaxS d1CqkbjUcivIBVd1k8GiuXEgjaEvwMlv8kPj1FGu/GCoMeyoRy7mXN1NOapPdFVu Gz2e8xR8Bri7oglvsS25i6gjYeYvP8qhTcc8tKBU2tr6MHFN3sUkuOZwtLKJLirc NxdSzaylgVGpCFQ0cwNSjm+eY0WvBd4eLychal+cs2plMmMAkYVTLDRBC4ZReCkw KIL1cdIMHQIZSSrDu22h =tKFs -----END PGP SIGNATURE-----

2 2

Re: kernel32: Stub out GetFileMUIPath.
by Nikolay Sivov April 3, 2015

April 3, 2015

> +@ stdcall GetFileMUIPath(long ptr ptr ptr ptr ptr ptr) Second and third arguments should be 'wstr'.

1 0

Re: advapi32: Handle NULL DACL in SetSecurityInfo
by Sebastian Lackner April 3, 2015

April 3, 2015

On 02.04.2015 17:20, Piotr Caban wrote: > --- > dlls/advapi32/security.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > It doesn't match the Windows behaviour, please add at least a couple of tests. https://newtestbot.winehq.org/JobDetails.pl?Key=12622&log_204=1#k204

3 4

Re: user32/tests: Avoid side-effect in an assert.
by Marvin April 3, 2015

April 3, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12631 Your paranoid android. === wvistau64 (32 bit win) === win.c:7487: Test failed: didn't get start_event in.c:7429: Test failed: button under static window didn't get WM_LBUTTONUP

1 0

Re: [1/3] server: Allow to open a directory without any permission bits.
by Piotr Caban April 2, 2015

April 2, 2015

On 04/02/15 09:42, Sebastian Lackner wrote: > Its a bit ugly, but probably the best solution to workaround the restrictions of the unix api. ;) There's similar problem with regular files (not directories). The owner should be able to read/write permissions. Also is there any need for unix_fd in this case? Thanks, Piotr

2 1

Re: [PATCH 3/4] d3drm/tests: Split tests between different versions, initially adding frame QueryInterface tests
by Stefan Dösinger April 2, 2015

April 2, 2015

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi, There are a few more issues: Am 2015-04-02 um 18:36 schrieb Aaryaman Vasishta: > + * Copyright 2012 Andr� Hentschel > + * Copyright 2011-2014 Henri Verbeet for CodeWeavers The encoding is still messed up. Is there any code you have taken from the other file? test_qi() is from dlls/ddraw/tests/ddraw*.c, so you should keep Henri's line. Taking along all the Copyright lines if you take something from an existing file is a safe choice though. Also, you should still fix your editor settings anyways. You can add your own name there if you want. (Fwiw, I don't know why we're maintaining these lines at all. The last thing I heard was that they don't have any legal meaning anyway. Also, git keep track of authorship pretty well.) > +#define CHECK_REFCOUNT(obj,rc) \ > + { \ > + int rc_new = rc; \ > + int count = get_refcount((IUnknown *)obj); \ > + ok(count == rc_new, "Invalid refcount. Expected %d got %d\n", rc_new, count); \ > + } Do you need this? You're using this in one place as far as I can see, you can also just use the get_refcount() and ok() calls directly. > +static int get_refcount(IUnknown *object) > +{ > + IUnknown_AddRef(object); > + return IUnknown_Release(object); > +} Refcounts are ULONG. > + hr = IDirect3DRM_CreateFrame(d3drm, NULL, &frame); > + ok(hr == D3DRM_OK, "Cannot get IDirect3DRMFrame3 interface (hr = %x)\n", hr); The string is a bit misleading, I think something like "Failed to create frame" is better. Not a big issue though. The same applies to d3drm2.c and d3drm3.c -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAEBAgAGBQJVHaNRAAoJEN0/YqbEcdMwZsAP/RvHLPjullG+tQC+JOZvzGDg nm2CZanwOzmC8ZEvWpXKxhCUjTIVUu0XB7yV33BgPywgRKBUUXkqEuGN+Z/wk7Lm rqNuPx17nPHL7aKjw9DxU3QP75t7RbFxU6N3Wz3Veoi+ox4xwxwmrIxRMkFNtxhv tPREjNp9yQVO79ZlLzI1VzeMVnTHKgVn/L+s68nADUkcBFrQaH6SBhB232utwbtp DdkwYOWCK0xGweaLAkt33X2Qhp+qbdtQjIoK/UZvq190owrW2tC2ORnIFhB/7zDR +FrpGzwGuTnhVocrLYCtVtB1RW5t0QealNjhlwv9HI03NSi/J7MmcwGOIE3TIatm TCCDvp034+EMThEvTDODATbYyihbgihG/jjiaQwznu+A66N9lOvhKRCbiasTjTym X06ti98dmpq9BwMd+8sRVe+U1zi9nRmGqvrcncmxUyZ1QoFdrDWFEpRog1jJvaho 8zO9NIzwsG839bqRyZIyZ0DvteaGXD7o907Nkr7aDt4jMjUmuj0ij+nrDvif9kTJ 5WOppUgN7JN4HhWNo4y5i3jei8X2TDDI9z76Juiv1DM0cbWbr2UTHSvFRysPiqGT vCg+boC2pZd94vQKnMYbyAg+06KDneZwiNYBVQpirY3Bua7PY7TwxkBgXQfh2sNv B2fjxmZVRMB3dWylqJmg =wYMH -----END PGP SIGNATURE-----

2 1

Re: [PATCH] undefined symbols compiling for macosx >10.8
by Ken Thomases April 2, 2015

April 2, 2015

That symbol is in the 10.8 SDK. We build against that SDK all the time. It works fine. Here's the documentation for that function: <https://developer.apple.com/library/mac/documentation/Security/Reference/se…>. Notice that under Availability it says, "Available in OS X v10.8 and later." Here is the declaration from Security.framework/Headers/SecureTransport.h: ---------- OSStatus SSLGetProtocolVersionMax (SSLContextRef context, SSLProtocol *maxVersion) __OSX_AVAILABLE_STARTING(__MAC_10_8, __IPHONE_5_0); ---------- Notice the __MAC_10_8 identifier in the __OSX_AVAILABLE_STARTING() decorator. You need to fix your build environment and/or your build commands. Whatever problem you're encountering while trying to build is specific to your system. -Ken On Apr 2, 2015, at 11:58 AM, Sergey Isakov <isakov-sl(a)bk.ru> wrote: > Your proposition that SDK10.8 contains the symbol is wrong > —— > #if MAC_OS_X_VERSION_MAX_ALLOWED >= 1080 > +#ifdef SSLGetProtocolVersionMax > if(SSLGetProtocolVersionMax != NULL) { > ---- > > On 01 апр. 2015 г., at 21:36, Ken Thomases <ken(a)codeweavers.com> wrote: > >> If you were having problems building Wine without this patch, then you have a system configuration problem. Or you're configuring Wine incorrectly. >> >> -Ken >> >> On Apr 1, 2015, at 11:52 AM, Sergey Isakov <isakov-sl(a)bk.ru> wrote: >> >>> Sorry but I committed the patch make me compilation be possible. >>> >>> On 01 апр. 2015 г., at 18:34, Ken Thomases <ken(a)codeweavers.com> wrote: >>> >>>> Hi, >>>> >>>> This patch makes no sense. Wine builds just fine for OS X versions greater than 10.8. Indeed, as near as I know, it still builds just fine for 10.5 and up (the Mac driver requires 10.6 or later, but on 10.5 it just will be skipped; it won't break anything). >>>> >>>> On Apr 1, 2015, at 6:48 AM, Sergey Isakov <isakov-sl(a)bk.ru> wrote: >>>> >>>>> Undefined symbols for architecture i386: >>>>> "_kCGDisplayShowDuplicateLowResolutionModes", referenced from: >>>>> -[WineApplicationController modeMatchingMode:forDisplay:] in cocoa_app.o >>>>> ld: symbol(s) not found for architecture i386 >>>> >>>> If you're getting this error, you're not building against a 10.8+ SDK properly. The compiler is seeing the headers for a 10.8+ SDK but the linker is seeing libraries and frameworks from some pre-10.8 SDK. >>>> >>>> >>>>> Instead of checking for NULL we should check #ifdef >>>> >>>> No. The things being tested are not macros in any version of the SDK. Your #ifdef checks will never pass. You have effectively done "#if 0". >>>> >>>> Finally, changes to separate DLLs (secur32 and winemac.drv, in this case) should be in separate patches unless they are inter-dependent. >>>> >>>> -Ken >>>> >>> >> >

1 0

Re: [PATCH] vbscript: Fixed Hex(try 3)
by Marvin April 2, 2015

April 2, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12612 Your paranoid android. === w2000pro (32 bit) === No test summary line found === wxppro (32 bit) === No test summary line found === w2003std (32 bit) === No test summary line found === wvistau64 (32 bit) === No test summary line found === w2008s64 (32 bit) === No test summary line found === w7u (32 bit) === No test summary line found === w8 (32 bit) === No test summary line found === wvistau64 (64 bit) === No test summary line found === w2008s64 (64 bit) === No test summary line found

1 0

Re: [1/4] windowscodecs: Add test for PNG gAMA metadata.
by Vincent Povirk April 2, 2015

April 2, 2015

I forgot to NULL-terminate a string, please ignore. On Wed, Apr 1, 2015 at 5:24 PM, Vincent Povirk <madewokherd(a)gmail.com> wrote: >

1 0

Re: [PATCH] findstr: Added test for findstr. (try 2)
by Kaipeng Zeng April 2, 2015

April 2, 2015

Excuse me,can anyone have a look at this patch? I appreciate for any comment. Thanks！ 2015-03-24 13:55 GMT+08:00 Kaipeng Zeng <kaipeng94(a)gmail.com>: > Thanks Stefan! > > Superseded patch 110143 . > ChangeLog: > - Exclude change to configure > - Return exit code of findstr in runcmd() >

2 1

Re: [1/3] server: Implement job limit flags (try 2)
by Nikolay Sivov April 2, 2015

April 2, 2015

On 02.04.2015 10:38, Andrew Cook wrote: > Changes in try 2: > - Removed the goto in SetInformationJobObject, ExtendedLimit will need its own > wineserver call eventually anyway I don't think you'll need a separate server call for extended info case. All fields should be enabled with flags, so you don't have to invent special values to ignore some limits. > + switch (class) > + { > + > + case JobObjectExtendedLimitInformation: > + if (len != sizeof(JOBOBJECT_EXTENDED_LIMIT_INFORMATION)) > + return STATUS_INVALID_PARAMETER; > + > + basic_limit = &(((JOBOBJECT_EXTENDED_LIMIT_INFORMATION *)info)->BasicLimitInformation); > + if (basic_limit->LimitFlags & ~JOB_OBJECT_EXTENDED_LIMIT_VALID_FLAGS) > + return STATUS_INVALID_PARAMETER; > + > + SERVER_START_REQ( job_set_limits ) > + { > + req->handle = wine_server_obj_handle( handle ); > + req->limit_flags = basic_limit->LimitFlags; > + status = wine_server_call( req ); > + } > + SERVER_END_REQ; > + break; This still looks strange, if you need only flags you can use switch() to prepare this flags value depending on info class, and handle invalid arguments same way you do know, no need to duplicate server calling block. Actually since extended info contains basic info you only need a switch() to test for invalid arguments, and you can cast unconditionally to JOBOBJECT_BASIC_LIMIT_INFORMATION* to get flags.

1 0

Re: [1/3] server: Implement job limit flags
by Alexandre Julliard April 2, 2015

April 2, 2015

Andrew Cook <ariscop(a)gmail.com> writes: > + switch (class) > + { > + > + case JobObjectExtendedLimitInformation: > + if (len != sizeof(JOBOBJECT_EXTENDED_LIMIT_INFORMATION)) > + return STATUS_INVALID_PARAMETER; > + > + basic_limit = &(((JOBOBJECT_EXTENDED_LIMIT_INFORMATION *)info)->BasicLimitInformation); > + if (basic_limit->LimitFlags & ~JOB_OBJECT_EXTENDED_LIMIT_VALID_FLAGS) > + return STATUS_INVALID_PARAMETER; > + > + goto set_basic_limits; > + > + case JobObjectBasicLimitInformation: > + if (len != sizeof(JOBOBJECT_BASIC_LIMIT_INFORMATION)) > + return STATUS_INVALID_PARAMETER; > + > + basic_limit = info; > + if (basic_limit->LimitFlags & ~JOB_OBJECT_BASIC_LIMIT_VALID_FLAGS) > + return STATUS_INVALID_PARAMETER; > + > + set_basic_limits: > + SERVER_START_REQ( set_job_limits ) > + { > + req->handle = wine_server_obj_handle( handle ); > + req->limit_flags = basic_limit->LimitFlags; > + status = wine_server_call( req ); > + } > + SERVER_END_REQ; > + break; A goto target in the middle of a switch statement is not very nice, please try to find a better way. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: msctf: Always return FALSE in DllCanUnloadNow
by Sebastian Lackner April 2, 2015

April 2, 2015

On 02.04.2015 03:34, Alistair Leslie-Hughes wrote: > Hi, > Fixes https://bugs.winehq.org/show_bug.cgi?id=31579 > > Changelog: > msctf: Always return FALSE in DllCanUnloadNow > > > Best Regards > Alistair Leslie-Hughes > > > I doubt that the patch has any real disadvantage, but wouldn't it be better to fix the refcounting instead of just removing this feature? Qian Hong even attached a suggested patch on the bugreport.

3 2

Re: comctl32/tests: Fix couple of metrics tests that fail on some systems (try 3)
by Marvin April 2, 2015

April 2, 2015

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 https://testbot.winehq.org/JobDetails.pl?Key=12601 Your paranoid android. === wvistau64 (32 bit listview) === listview.c:5190: Test failed: NM_HOVER allow test: the msg sequence is not complete: expected 0087 - actual 0000 listview.c:5195: Test failed: NM_HOVER block test: the msg sequence is not complete: expected 0087 - actual 0000 === w2008s64 (32 bit listview) === listview.c:5190: Test failed: NM_HOVER allow test: the msg sequence is not complete: expected 0087 - actual 0000 listview.c:5195: Test failed: NM_HOVER block test: the msg sequence is not complete: expected 0087 - actual 0000 === w7u (32 bit listview) === listview.c:5190: Test failed: NM_HOVER allow test: the msg sequence is not complete: expected 0087 - actual 0000 listview.c:5195: Test failed: NM_HOVER block test: the msg sequence is not complete: expected 0087 - actual 0000 === w2008s64 (64 bit listview) === listview.c:5190: Test failed: NM_HOVER allow test: the msg sequence is not complete: expected 0087 - actual 0000 listview.c:5195: Test failed: NM_HOVER block test: the msg sequence is not complete: expected 0087 - actual 0000

2 1

Re: [PATCH (try2)] gdiplus: Avoid unnecessary pre-multiplied alpha conversions in GdipDrawImagePointsRect
by Vincent Povirk April 1, 2015

April 1, 2015

I think we also need resample_bitmap_pixel to be set up to handle PARGB. Nearest neighbor interpolation should work OK as is, but bilinear interpolation uses blend_colors which only works correctly for ARGB. (The implementation of blend_colors that we had before 9c579023f0750539bdf7815f8545c4620540387a should actually be correct for PARGB.) On Wed, Apr 1, 2015 at 11:38 AM, Andrew Eikum <aeikum(a)codeweavers.com> wrote: > Not all of the operations we perform in the DrawImage family need to > be post-multiplied. We can avoid doing a pre-to-post-multiplied alpha > conversion by detecting whether we need post-multiplied ahead of time. > Avoiding this conversion saves considerable time, about 3/1000s in my > particular test case. > > --- > > try2: Fix return value from color_over_fgpremult > > dlls/gdiplus/gdiplus_private.h | 20 +++++++++++++ > dlls/gdiplus/graphics.c | 66 ++++++++++++++++++++++++++++-------------- > 2 files changed, 65 insertions(+), 21 deletions(-) > > diff --git a/dlls/gdiplus/gdiplus_private.h b/dlls/gdiplus/gdiplus_private.h > index 065f5ed..57b3de3 100644 > --- a/dlls/gdiplus/gdiplus_private.h > +++ b/dlls/gdiplus/gdiplus_private.h > @@ -121,6 +121,26 @@ static inline ARGB color_over(ARGB bg, ARGB fg) > return (a<<24)|(r<<16)|(g<<8)|b; > } > > +/* fg is premult, bg and return value are not */ > +static inline ARGB color_over_fgpremult(ARGB bg, ARGB fg) > +{ > + BYTE b, g, r, a; > + BYTE bg_alpha, fg_alpha; > + > + fg_alpha = (fg>>24)&0xff; > + > + if (fg_alpha == 0) return bg; > + > + bg_alpha = (((bg>>24)&0xff) * (0xff-fg_alpha)) / 0xff; > + > + a = bg_alpha + fg_alpha; > + b = ((bg&0xff)*bg_alpha + (fg&0xff)*0xff)/a; > + g = (((bg>>8)&0xff)*bg_alpha + ((fg>>8)&0xff)*0xff)/a; > + r = (((bg>>16)&0xff)*bg_alpha + ((fg>>16)&0xff)*0xff)/a; > + > + return (a<<24)|(r<<16)|(g<<8)|b; > +} > + > extern const char *debugstr_rectf(const RectF* rc) DECLSPEC_HIDDEN; > > extern const char *debugstr_pointf(const PointF* pt) DECLSPEC_HIDDEN; > diff --git a/dlls/gdiplus/graphics.c b/dlls/gdiplus/graphics.c > index 580390f..89c108e 100644 > --- a/dlls/gdiplus/graphics.c > +++ b/dlls/gdiplus/graphics.c > @@ -354,9 +354,9 @@ static GpStatus get_clip_hrgn(GpGraphics *graphics, HRGN *hrgn) > return GdipGetRegionHRgn(graphics->clip, NULL, hrgn); > } > > -/* Draw non-premultiplied ARGB data to the given graphics object */ > +/* Draw ARGB data to the given graphics object */ > static GpStatus alpha_blend_bmp_pixels(GpGraphics *graphics, INT dst_x, INT dst_y, > - const BYTE *src, INT src_width, INT src_height, INT src_stride) > + const BYTE *src, INT src_width, INT src_height, INT src_stride, const PixelFormat fmt) > { > GpBitmap *dst_bitmap = (GpBitmap*)graphics->image; > INT x, y; > @@ -372,7 +372,10 @@ static GpStatus alpha_blend_bmp_pixels(GpGraphics *graphics, INT dst_x, INT dst_ > continue; > > GdipBitmapGetPixel(dst_bitmap, x+dst_x, y+dst_y, &dst_color); > - GdipBitmapSetPixel(dst_bitmap, x+dst_x, y+dst_y, color_over(dst_color, src_color)); > + if (fmt & PixelFormatPAlpha) > + GdipBitmapSetPixel(dst_bitmap, x+dst_x, y+dst_y, color_over_fgpremult(dst_color, src_color)); > + else > + GdipBitmapSetPixel(dst_bitmap, x+dst_x, y+dst_y, color_over(dst_color, src_color)); > } > } > > @@ -380,7 +383,7 @@ static GpStatus alpha_blend_bmp_pixels(GpGraphics *graphics, INT dst_x, INT dst_ > } > > static GpStatus alpha_blend_hdc_pixels(GpGraphics *graphics, INT dst_x, INT dst_y, > - const BYTE *src, INT src_width, INT src_height, INT src_stride) > + const BYTE *src, INT src_width, INT src_height, INT src_stride, PixelFormat fmt) > { > HDC hdc; > HBITMAP hbitmap; > @@ -404,7 +407,8 @@ static GpStatus alpha_blend_hdc_pixels(GpGraphics *graphics, INT dst_x, INT dst_ > hbitmap = CreateDIBSection(hdc, (BITMAPINFO*)&bih, DIB_RGB_COLORS, > (void**)&temp_bits, NULL, 0); > > - if (GetDeviceCaps(graphics->hdc, SHADEBLENDCAPS) == SB_NONE) > + if (GetDeviceCaps(graphics->hdc, SHADEBLENDCAPS) == SB_NONE || > + fmt & PixelFormatPAlpha) > memcpy(temp_bits, src, src_width * src_height * 4); > else > convert_32bppARGB_to_32bppPARGB(src_width, src_height, temp_bits, > @@ -420,7 +424,7 @@ static GpStatus alpha_blend_hdc_pixels(GpGraphics *graphics, INT dst_x, INT dst_ > } > > static GpStatus alpha_blend_pixels_hrgn(GpGraphics *graphics, INT dst_x, INT dst_y, > - const BYTE *src, INT src_width, INT src_height, INT src_stride, HRGN hregion) > + const BYTE *src, INT src_width, INT src_height, INT src_stride, HRGN hregion, PixelFormat fmt) > { > GpStatus stat=Ok; > > @@ -470,7 +474,7 @@ static GpStatus alpha_blend_pixels_hrgn(GpGraphics *graphics, INT dst_x, INT dst > stat = alpha_blend_bmp_pixels(graphics, rects[i].left, rects[i].top, > &src[(rects[i].left - dst_x) * 4 + (rects[i].top - dst_y) * src_stride], > rects[i].right - rects[i].left, rects[i].bottom - rects[i].top, > - src_stride); > + src_stride, fmt); > } > > GdipFree(rgndata); > @@ -503,7 +507,7 @@ static GpStatus alpha_blend_pixels_hrgn(GpGraphics *graphics, INT dst_x, INT dst > ExtSelectClipRgn(graphics->hdc, hregion, RGN_AND); > > stat = alpha_blend_hdc_pixels(graphics, dst_x, dst_y, src, src_width, > - src_height, src_stride); > + src_height, src_stride, fmt); > > RestoreDC(graphics->hdc, save); > > @@ -514,9 +518,9 @@ static GpStatus alpha_blend_pixels_hrgn(GpGraphics *graphics, INT dst_x, INT dst > } > > static GpStatus alpha_blend_pixels(GpGraphics *graphics, INT dst_x, INT dst_y, > - const BYTE *src, INT src_width, INT src_height, INT src_stride) > + const BYTE *src, INT src_width, INT src_height, INT src_stride, PixelFormat fmt) > { > - return alpha_blend_pixels_hrgn(graphics, dst_x, dst_y, src, src_width, src_height, src_stride, NULL); > + return alpha_blend_pixels_hrgn(graphics, dst_x, dst_y, src, src_width, src_height, src_stride, NULL, fmt); > } > > static ARGB blend_colors(ARGB start, ARGB end, REAL position) > @@ -656,8 +660,9 @@ static BOOL color_is_gray(ARGB color) > return (r == g) && (g == b); > } > > -static void apply_image_attributes(const GpImageAttributes *attributes, LPBYTE data, > - UINT width, UINT height, INT stride, ColorAdjustType type) > +/* returns preferred pixel format for the applied attributes */ > +static PixelFormat apply_image_attributes(const GpImageAttributes *attributes, LPBYTE data, > + UINT width, UINT height, INT stride, ColorAdjustType type, PixelFormat fmt) > { > UINT x, y; > INT i; > @@ -669,6 +674,9 @@ static void apply_image_attributes(const GpImageAttributes *attributes, LPBYTE d > BYTE min_blue, min_green, min_red; > BYTE max_blue, max_green, max_red; > > + if (!data || fmt != PixelFormat32bppARGB) > + return PixelFormat32bppARGB; > + > if (attributes->colorkeys[type].enabled) > key = &attributes->colorkeys[type]; > else > @@ -702,6 +710,9 @@ static void apply_image_attributes(const GpImageAttributes *attributes, LPBYTE d > { > const struct color_remap_table *table; > > + if (!data || fmt != PixelFormat32bppARGB) > + return PixelFormat32bppARGB; > + > if (attributes->colorremaptables[type].enabled) > table = &attributes->colorremaptables[type]; > else > @@ -731,6 +742,9 @@ static void apply_image_attributes(const GpImageAttributes *attributes, LPBYTE d > int gray_matrix[5][5]; > BOOL identity; > > + if (!data || fmt != PixelFormat32bppARGB) > + return PixelFormat32bppARGB; > + > if (attributes->colormatrices[type].enabled) > colormatrices = &attributes->colormatrices[type]; > else > @@ -769,6 +783,9 @@ static void apply_image_attributes(const GpImageAttributes *attributes, LPBYTE d > { > REAL gamma; > > + if (!data || fmt != PixelFormat32bppARGB) > + return PixelFormat32bppARGB; > + > if (attributes->gamma_enabled[type]) > gamma = attributes->gamma[type]; > else > @@ -793,6 +810,8 @@ static void apply_image_attributes(const GpImageAttributes *attributes, LPBYTE d > *src_color = (*src_color & 0xff000000) | (red << 16) | (green << 8) | blue; > } > } > + > + return fmt; > } > > /* Given a bitmap and its source rectangle, find the smallest rectangle in the > @@ -1237,7 +1256,7 @@ static GpStatus brush_fill_pixels(GpGraphics *graphics, GpBrush *brush, > if (stat == Ok) > apply_image_attributes(fill->imageattributes, fill->bitmap_bits, > bitmap->width, bitmap->height, > - src_stride, ColorAdjustTypeBitmap); > + src_stride, ColorAdjustTypeBitmap, lockeddata.PixelFormat); > > if (stat != Ok) > { > @@ -2990,15 +3009,18 @@ GpStatus WINGDIPAPI GdipDrawImagePointsRect(GpGraphics *graphics, GpImage *image > return OutOfMemory; > src_stride = sizeof(ARGB) * src_area.Width; > > - /* Read the bits we need from the source bitmap into an ARGB buffer. */ > + /* Read the bits we need from the source bitmap into a compatible buffer. */ > lockeddata.Width = src_area.Width; > lockeddata.Height = src_area.Height; > lockeddata.Stride = src_stride; > - lockeddata.PixelFormat = PixelFormat32bppARGB; > lockeddata.Scan0 = src_data; > + if (bitmap->format == PixelFormat32bppPARGB) > + lockeddata.PixelFormat = apply_image_attributes(imageAttributes, NULL, 0, 0, 0, ColorAdjustTypeBitmap, bitmap->format); > + else > + lockeddata.PixelFormat = PixelFormat32bppARGB; > > stat = GdipBitmapLockBits(bitmap, &src_area, ImageLockModeRead|ImageLockModeUserInputBuf, > - PixelFormat32bppARGB, &lockeddata); > + lockeddata.PixelFormat, &lockeddata); > > if (stat == Ok) > stat = GdipBitmapUnlockBits(bitmap, &lockeddata); > @@ -3011,7 +3033,7 @@ GpStatus WINGDIPAPI GdipDrawImagePointsRect(GpGraphics *graphics, GpImage *image > > apply_image_attributes(imageAttributes, src_data, > src_area.Width, src_area.Height, > - src_stride, ColorAdjustTypeBitmap); > + src_stride, ColorAdjustTypeBitmap, lockeddata.PixelFormat); > > if (do_resampling) > { > @@ -3059,7 +3081,8 @@ GpStatus WINGDIPAPI GdipDrawImagePointsRect(GpGraphics *graphics, GpImage *image > } > > stat = alpha_blend_pixels(graphics, dst_area.left, dst_area.top, > - dst_data, dst_area.right - dst_area.left, dst_area.bottom - dst_area.top, dst_stride); > + dst_data, dst_area.right - dst_area.left, dst_area.bottom - dst_area.top, dst_stride, > + lockeddata.PixelFormat); > > GdipFree(src_data); > > @@ -4009,7 +4032,8 @@ static GpStatus SOFTWARE_GdipFillRegion(GpGraphics *graphics, GpBrush *brush, > if (stat == Ok) > stat = alpha_blend_pixels_hrgn(graphics, gp_bound_rect.X, > gp_bound_rect.Y, (BYTE*)pixel_data, gp_bound_rect.Width, > - gp_bound_rect.Height, gp_bound_rect.Width * 4, hregion); > + gp_bound_rect.Height, gp_bound_rect.Width * 4, hregion, > + PixelFormat32bppARGB); > > GdipFree(pixel_data); > } > @@ -5811,7 +5835,7 @@ GpStatus WINGDIPAPI GdipReleaseDC(GpGraphics *graphics, HDC hdc) > /* Write the changed pixels to the real target. */ > alpha_blend_pixels(graphics, 0, 0, graphics->temp_bits, > graphics->temp_hbitmap_width, graphics->temp_hbitmap_height, > - graphics->temp_hbitmap_width * 4); > + graphics->temp_hbitmap_width * 4, PixelFormat32bppARGB); > > /* Clean up. */ > DeleteDC(graphics->temp_hdc); > @@ -6366,7 +6390,7 @@ static GpStatus SOFTWARE_GdipDrawDriverString(GpGraphics *graphics, GDIPCONST UI > > /* draw the result */ > stat = alpha_blend_pixels(graphics, min_x, min_y, pixel_data, pixel_area.Width, > - pixel_area.Height, pixel_data_stride); > + pixel_area.Height, pixel_data_stride, PixelFormat32bppARGB); > > GdipFree(pixel_data); > > -- > 2.3.5 > > >

1 1