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

Firefox-1.5.x on Wine slow on javascript-heavy pages
by Dan Kegel Dec. 3, 2006

Dec. 3, 2006

I've been using Firefox-1.5.x again on wine (using the winehq packages for Ubuntu) on my dual 650MHz system as my main browser again for the last few months. It's working quite well... except that javascript-heavy pages/apps like gmail are a tad sluggish at accepting keyboard input. For some reason, yahoo.com's main page is also sluggish; they must do something fancy. Both are fine with native firefox on this system. I'm sure this won't affect anyone with a modern system, but it is kind of sad that my trusty old dual CPU system isn't up to daily tasks anymore. Sniff... ah, well, I see Fry's has amd64 4000 motherboard/cpu combos for $189, guess it's time to upgrade. At some point it'd be nice to look into why Firefox's javascript is slow under Wine, though. - Dan

2 1

Re: d3d [2]: Optional callback infrastructure for implicit texture surface destruction.
by Stefan Dösinger Dec. 3, 2006

Dec. 3, 2006

Am Sonntag 03 Dezember 2006 21:52 schrieb Markus Amsler: > /* IWineD3DTexture */ > + IWineD3DTextureImpl_Destroy, Wouldn't it maybe be better to make the destroy method part of IWineD3DBase?

2 1

Re: d3d [2]: Optional callback infrastructure for implicit texture surface destruction.
by H. Verbeet Dec. 3, 2006

Dec. 3, 2006

On 03/12/06, Markus Amsler <markus.amsler(a)oribi.org> wrote: > --- > dlls/wined3d/texture.c | 53 > +++++++++++++++++++++++--------------- > include/wine/wined3d_interface.h | 7 +++++ > 2 files changed, 39 insertions(+), 21 deletions(-) I think the callback should be non-optional.

2 1

One more go at fixing GetCursorPos()
by Vitaliy Margolen Dec. 3, 2006

Dec. 3, 2006

Here is one more shot at fixing GetCursorPos() to make it usable in dinput. This time I used a refcount of sorts. Because the problem appears only during processing of hook chain. Just to remind about what is the problem: when hook handler calls GetCursorPos() position returned is the "old" position. While hook parameters contain "new" coordinates. This is on windows. On Wine returned position queried directly from X and might be even "newer" then position in hook parameters. Attached are two tests (out of lots more that I've used to find and fix the problem). test_msg.c - tests major problem that Alexandre pointed out about last fix attempt. Application in question does not process messages while calling GetCursorPos(). It still works properly with proposed patch. di_mouse_2th.c - 2 threads showing that there are no extra thread local position involved (if there is such a thing). Cached cursor position is global to all threads. (Last numbers should stay at dt(0 0) at all times). Inside hook handler d(x y) should show relative mouse move delta. If anyone has any objections, comment, suggestions, please do send them. Vitaliy Margolen. #define COBJMACROS #define _WIN32_WINNT 0x0501 #include <windows.h> #include <windowsx.h> #include <stdio.h> static HANDLE event1, event2; static HINSTANCE g_hInstance; static POINT pt_global; LRESULT CALLBACK hook_proc( int code, WPARAM wparam, LPARAM lparam ) { MSLLHOOKSTRUCT *hook = (MSLLHOOKSTRUCT *)lparam; POINT pt; GetCursorPos(&pt); if (code == HC_ACTION) { fprintf(stderr, "Hook: code=%d w=%x [(%ld %ld) %lx %lx %lx] cur(%ld %ld) d(%ld %ld) ", code, wparam, hook->pt.x, hook->pt.y, hook->mouseData, hook->flags, hook->dwExtraInfo, pt.x, pt.y, hook->pt.x - pt.x, hook->pt.y - pt.y); pt_global = pt; SetEvent(event1); WaitForSingleObject(event2, INFINITE); } return CallNextHookEx( 0, code, wparam, lparam ); } static DWORD CALLBACK thread_proc(LPVOID arg) { while (WaitForSingleObject(event1, INFINITE) == WAIT_OBJECT_0) { POINT pt; GetCursorPos(&pt); fprintf(stderr, "Th: (%ld %ld) dt(%ld %ld)\n", pt.x, pt.y, pt.x - pt_global.x, pt.y - pt_global.y); SetEvent(event2); } return 0; } static DWORD wnd_proc(void) { MSG msg; while (GetMessage(&msg, NULL, 0, 0)) { TranslateMessage(&msg); DispatchMessage(&msg); } return msg.wParam; } #define SEND(x,y) \ i.mi.time = GetCurrentTime(); \ i.mi.dx = x; i.mi.dy = y; \ SendInput(1, &i, sizeof(INPUT)); void send_input(void) { INPUT i; POINT ptorg, pt; GetCursorPos(&ptorg); fprintf(stderr, "Cur pos: %ld %ld\n", ptorg.x, ptorg.y); i.type = INPUT_MOUSE; i.mi.dwFlags = MOUSEEVENTF_MOVE; i.mi.dwExtraInfo = 0; i.mi.mouseData = 0; SEND(-4, -4) GetCursorPos(&pt); fprintf(stderr, "%ld %ld\n", pt.x, pt.y); SetCursorPos(ptorg.x, ptorg.y); Sleep(250); fprintf(stderr, "***DONE***\n"); } void test1(HWND hwnd) { HANDLE thread; HHOOK hook; MSG msg; event1 = CreateEvent(NULL, FALSE, FALSE, NULL); event2 = CreateEvent(NULL, FALSE, FALSE, NULL); thread = CreateThread(NULL, 0, thread_proc, NULL, 0, NULL); Sleep(250); hook = SetWindowsHookExA(WH_MOUSE_LL, hook_proc, g_hInstance, 0); Sleep(250); // send_input(); wnd_proc(); while (PeekMessage( &msg, 0, 0, 0, PM_REMOVE )) DispatchMessage( &msg ); UnhookWindowsHookEx(hook); CloseHandle(event1); CloseHandle(event2); CloseHandle(thread); Sleep(250); } LRESULT CALLBACK WndProc(HWND hWnd, UINT message, WPARAM wParam, LPARAM lParam) { switch (message) { case WM_DESTROY: PostQuitMessage(0); return 0; } return DefWindowProc(hWnd, message, wParam, lParam); } int WINAPI WinMain(HINSTANCE hInstance, HINSTANCE hPrevInstance, LPSTR lpCmdLine, int nCmdShow) { WNDCLASSEX wcex; HWND hwnd; g_hInstance = hInstance; wcex.cbSize = sizeof(WNDCLASSEX); wcex.style = CS_HREDRAW | CS_VREDRAW; wcex.lpfnWndProc = (WNDPROC)WndProc; wcex.cbClsExtra = 0; wcex.cbWndExtra = 0; wcex.hInstance = hInstance; wcex.hIcon = NULL; wcex.hCursor = LoadCursor(NULL, IDC_ARROW); wcex.hbrBackground = (HBRUSH)(COLOR_WINDOW+1); wcex.lpszMenuName = NULL; wcex.lpszClassName = "MainWindow"; wcex.hIconSm = NULL; RegisterClassEx(&wcex); hwnd = CreateWindow("MainWindow", "Title", WS_OVERLAPPEDWINDOW, 10, 10, 200, 200, NULL, NULL, NULL, NULL); ShowWindow(hwnd, SW_SHOW); test1(hwnd); DestroyWindow(hwnd); return 0; } #include <windows.h> #include <stdio.h> int WINAPI WinMain(HINSTANCE hInstance, HINSTANCE hPrevInstance, LPSTR lpCmdLine, int nCmdShow) { LONG started = GetTickCount(); POINT pt, pt_org; HWND hwnd; MSG msg; hwnd = CreateWindow("static", "Title", WS_OVERLAPPEDWINDOW | WS_VISIBLE, 10, 10, 200, 200, NULL, NULL, NULL, NULL); GetCursorPos(&pt_org); while (PeekMessage( &msg, 0, 0, 0, PM_REMOVE )) DispatchMessage( &msg ); while (started + 10000 > GetTickCount()) { GetCursorPos(&pt); if ((pt.x - pt_org.x) || (pt.y - pt_org.y)) printf("(%ld %ld)\n", pt.x - pt_org.x, pt.y - pt_org.y); pt_org = pt; } DestroyWindow(hwnd); return 0; }

2 2

winedump: Cast-qual warnings fix (2 of 2)
by Andrew Talbot Dec. 3, 2006

Dec. 3, 2006

Sorry, I made a mess of titling this patch, twice. Please find and use the version whose subject line and changelog entry refer to "winedump", not "winedbg". Thanks, -- Andy.

1 0

[PATCH] overflow checking for MultiByteToWideChar/WideCharToMultiByte
by Marcus Meissner Dec. 3, 2006

Dec. 3, 2006

Hi, This is more a FYI patch and GCC 4.0+ specific. This enhancement detects if the size of a known buffer is different than the specified size. Thie tests/sock.c adjustement is necessary due to stdlib.h inclusion. Ciao, Marcus --- dlls/kernel32/locale.c | 2 ++ dlls/ws2_32/tests/sock.c | 2 +- include/winnls.h | 45 +++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 48 insertions(+), 1 deletions(-) d0e7840335b6bbe4990b1c043dcff9b298c7a5a8 diff --git a/dlls/kernel32/locale.c b/dlls/kernel32/locale.c index bc7db75..1d4ceaf 100644 --- a/dlls/kernel32/locale.c +++ b/dlls/kernel32/locale.c @@ -1757,6 +1757,7 @@ BOOL WINAPI EnumSystemCodePagesW( CODEPA * is passed, and ERROR_NO_UNICODE_TRANSLATION if no translation is * possible for src. */ +#undef MultiByteToWideChar INT WINAPI MultiByteToWideChar( UINT page, DWORD flags, LPCSTR src, INT srclen, LPWSTR dst, INT dstlen ) { @@ -1850,6 +1851,7 @@ INT WINAPI MultiByteToWideChar( UINT pag * and dstlen != 0, and ERROR_INVALID_PARAMETER, if an invalid * parameter was given. */ +#undef WideCharToMultiByte INT WINAPI WideCharToMultiByte( UINT page, DWORD flags, LPCWSTR src, INT srclen, LPSTR dst, INT dstlen, LPCSTR defchar, BOOL *used ) { diff --git a/dlls/ws2_32/tests/sock.c b/dlls/ws2_32/tests/sock.c index c8b9cc8..f5ed0c1 100644 --- a/dlls/ws2_32/tests/sock.c +++ b/dlls/ws2_32/tests/sock.c @@ -22,8 +22,8 @@ #include <stdarg.h> #include <windef.h> -#include <winbase.h> #include <winsock2.h> +#include <winbase.h> #include <mswsock.h> #include "wine/test.h" #include <winnt.h> diff --git a/include/winnls.h b/include/winnls.h index 8888171..c13c0bf 100644 --- a/include/winnls.h +++ b/include/winnls.h @@ -20,6 +20,10 @@ #define __WINE_WINNLS_H #ifndef NONLS +#ifndef RC_INVOKED +# include <stdlib.h> +#endif + #ifdef __cplusplus extern "C" { #endif @@ -768,6 +772,26 @@ INT WINAPI LCMapStringW(LCID,DWO #define LCMapString WINELIB_NAME_AW(LCMapString) LCID WINAPI LocaleNameToLCID(LPCWSTR,DWORD); INT WINAPI MultiByteToWideChar(UINT,DWORD,LPCSTR,INT,LPWSTR,INT); + +static inline INT +WINAPI MultiByteToWideChar_ichk( + UINT cp,DWORD flags, + LPCSTR src,INT srclen,INT srcbuflen, + LPWSTR dst,INT dstlen,INT dstbuflen +) { + if ((srclen != -1) && (srcbuflen != -1) && (srcbuflen < srclen)) + abort(); + if ((dstlen != -1) && (dstbuflen != -1) && (dstbuflen/2 < dstlen)) + abort(); + /* later add assertions regarding conversion lengths */ + return MultiByteToWideChar(cp,flags,src,srclen,dst,dstlen); +} +#define MultiByteToWideChar(cp,flags,src,srclen,dst,dstlen) \ + MultiByteToWideChar_ichk(cp,flags, \ + src,srclen,__builtin_object_size(src,0), \ + dst,dstlen,__builtin_object_size(dst,0) \ + ) + INT WINAPI SetCalendarInfoA(LCID,CALID,CALTYPE,LPCSTR); INT WINAPI SetCalendarInfoW(LCID,CALID,CALTYPE,LPCWSTR); #define SetCalendarInfo WINELIB_NAME_AW(SetCalendarInfo) @@ -778,6 +802,27 @@ BOOL WINAPI SetThreadLocale(LCID) BOOL WINAPI SetUserGeoID(GEOID); INT WINAPI WideCharToMultiByte(UINT,DWORD,LPCWSTR,INT,LPSTR,INT,LPCSTR,LPBOOL); +static inline INT +WINAPI WideCharToMultiByte_ichk( + UINT cp,DWORD flags, + LPCWSTR src,INT srclen,INT srcbuflen, + LPSTR dst,INT dstlen,INT dstbuflen, + LPCSTR xstr,LPBOOL b +) { + if ((srclen != -1) && (srcbuflen != -1) && (srcbuflen/2 < srclen)) + abort(); + if ((dstlen != -1) && (dstbuflen != -1) && (dstbuflen < dstlen)) + abort(); + /* later add assertions regarding conversion lengths */ + return WideCharToMultiByte(cp,flags,src,srclen,dst,dstlen,xstr,b); +} +#define WideCharToMultiByte(cp,flags,src,srclen,dst,dstlen,defchar,b) \ + WideCharToMultiByte_ichk(cp,flags, \ + src,srclen,__builtin_object_size(src,0), \ + dst,dstlen,__builtin_object_size(dst,0), \ + defchar,b \ + ) + #ifdef __cplusplus } #endif -- 1.2.4

1 0

Re: [PATCH] check length in path canonicalize
by Dmitry Timoshkov Dec. 3, 2006

Dec. 3, 2006

"Marcus Meissner" <marcus(a)jet.franken.de> wrote: > - MultiByteToWideChar(CP_ACP,0,lpszPath,-1,szPath,MAX_PATH); > + INT ret; > + > + ret = MultiByteToWideChar(CP_ACP,0,lpszPath,-1,NULL,0); > + if (ret > MAX_PATH) { > + FIXME("Too long path!\n"); > + return FALSE; > + } > + ret = MultiByteToWideChar(CP_ACP,0,lpszPath,-1,szPath,sizeof(szPath)/sizeof(WCHAR)); It should be sufficient to check return value of MultiByteToWideChar to detect an overflow, there is no need to call MultiByteToWideChar twice for that. -- Dmitry.

1 0

XEMBED Implementation
by Cloudor Pu Dec. 2, 2006

Dec. 2, 2006

Hi, I find discussion about XEMBED implementation in this mailing list and that was 3 years ago. Is there now any usable implementation or examples for wine XEMBED plug? What I am doing is to embed IE into a GtkSocket. I wrote a standalone program compiled using winelib that can CreateWindow and embed IE. Can I CreateWindow using the XEMBED Id if there is already XEMBED implementation or can I just CreateWindow as a child of a GtkPlug which is a gtk-embeddee? Any advices or suggestions would be highly appreciated. Thank you! Regards, cloudor

3 3

Wined3d State management again
by Stefan Dösinger Dec. 2, 2006

Dec. 2, 2006

Hi, I've done some more work on my state management, and I uploaded an updated patch to http://stud4.tuwien.ac.at/~e0526822/wined3d-statemgmt-2.tar.bz2 . Ok, so what's updated: * Pixel shaders: Should work now * GLSL shaders: Working too. Henris shader backend stuff is integrated in a messy way. * Offscreen rendering: Works with back buffer and pbuffer rendering, fbos not yet. * Stateblocks: Brute force integration by just scheduling all modified states for a reset. No display list yet * Blitting, render target unlocking: Enabled again, sets new gl states and schedules a reapplication of the modified states. Not a nice solution yet What is still missing: * Nv register combiners. I haven't found a way yet to build a texture stage -> texture unit mapping which I like. All the ways I could think of could potentially cause lots of unneccessary state changes. One nice thing I have noticed is that this changes give a huge performance gain in Half-Life 2(dxlevel 80, and 90). Dxlevel 80 is accelerated to approximately 150-200% of the former performance, and especially slow parts are faster now. This makes a really huge playability difference :-) There are some regressions too, especially related to texture loading(3Dmark2000), and some glsl crashes. The good part about this change is that I can send it in in very small patches, making regression testing much easier, contrary to the huge ddraw rewrite. Where to go from here? As long as nobody screams I'd start sending patches now. Its easier to catch regressions while sending small patches instead of hunting them on the blob I have here(and my messy patch history in my git tree). I'm also afraid that I get too far away from the main wined3d code(See the shader backend integration). We can discuss the code based on the small changes I'll send. Stefan

1 1

Re: msi 2: Remove redundant NULL checks before x_Release
by Mike McCormack Dec. 1, 2006

Dec. 1, 2006

James Hawkins wrote: > Changelog: > * Remove redundant NULL checks before x_Release. obj->Release() will read the vtable pointer from the object, and that will crash if obj is NULL, so these checks are needed. The following code will crash: IUnknown *obj = NULL; IUnknown_Release(obj); You can see this looking at the definition of the IUnknown_Release macro: include/unknwn.h:#define IUnknown_Release(p) (p)->lpVtbl->Release(p) Mike

1 0

Video problems with FC6
by gslink Dec. 1, 2006

Dec. 1, 2006

There is currently a bug in FC6 that affects Wine. If, with any Nvidia driver from any source, a program attempts to use accelerated video and the attached monitor uses digital feed the program will often crash the user session. This has been reported and is not a problem in Wine but a good many games that used to run under Wine now crash. Other cards were not tested but may be affected. It is hoped RH will fix this rather quickly.

2 1

Re: [PATCH 1/9] dinput: Move acquired flag to the base device class.
by Peter Oberndorfer Dec. 1, 2006

Dec. 1, 2006

On Friday 01 December 2006 07:23, Vitaliy Margolen wrote: > Sorry ignore my previous patches. I didn't send them in the proper order. > > --- > dlls/dinput/device.c | 28 ++++++++++++++++++++++++++++ > dlls/dinput/device_private.h | 3 +++ > dlls/dinput/joystick_linux.c | 23 ++++++++--------------- > dlls/dinput/keyboard.c | 36 ++++++++++++++++-------------------- > dlls/dinput/mouse.c | 23 ++++++++--------------- > 5 files changed, 63 insertions(+), 50 deletions(-) > Hi, some questions/notices about this patch: Previously when the device was not acquired inside JoystickAImpl_Unacquire DIERR_NOTACQUIRED was returned. now you call IDirectInputDevice2AImpl_Unacquire, which returns DI_NOEFFECT when not acquired. but DI_NOEFFECT does not have high bit set so it won't trigger the FAILED macro. Do you know if the Acquire/Unacquire functions return value the same for all device types? In JoystickAImpl_Acquire you don't use your new IDirectInputDevice2AImpl_Acquire function but set This->base.acquired = 1; Is there a reason for this? In SysKeyboardAImpl_Unacquire you seem to call IDirectInputDevice2AImpl_Acquire? Greetings Peter

2 1

Re: user32: Factorize graphics driver's WindowCreate and ShowWindow in user32.
by Dmitry Timoshkov Dec. 1, 2006

Dec. 1, 2006

"Pierre d'Herbemont" <pdherbemont(a)free.fr> wrote: > At the same time ShowWindow gets factorized because WindowCreate and > ShowWindow share both the use of WINPOS_MinMaximize, so it is easier to > factorize both functions in a single patch. In fact ShowWindow is > removed entirely from the drivers as it is driver independant. > > To properly factorize WindowCreate we have to export 3 new functions > from the graphics driver which are SetIconicState, SyncWindowStyle, > SystrayDockWindow. I'd suggest to leave ShowWindow in the driver's interface. You already have created 3 new entry points to replace its partial functionality, but still miss ShowWindow functionality for at least SW_RESTORE and SW_MAXIMIZE. Also did you run 'make test' after this patch? -- Dmitry.

2 3

Re: oleaut32: add VarBstrCmp binary comparison for LCID==0
by Dmitry Timoshkov Dec. 1, 2006

Dec. 1, 2006

"Charles Blacklock" <charles(a)diagnos.co.uk> wrote: > + if (lcid == 0) > + { > + ret = memcmp(pbstrLeft, pbstrRight, min(SysStringByteLen(pbstrLeft), SysStringByteLen(pbstrRight))); > + if (ret < 0) > + return VARCMP_LT; > + if (ret > 0) > + return VARCMP_GT; > + if (SysStringByteLen(pbstrLeft) < SysStringByteLen(pbstrRight)) > + return VARCMP_LT; > + if (SysStringByteLen(pbstrLeft) > SysStringByteLen(pbstrRight)) > + return VARCMP_GT; > + return VARCMP_EQ; > + } It would be much more effective to call SysStringByteLen only once for each string instead of calling it again and again after a failing comparison. -- Dmitry.

2 2

How to embed wine into swt?
by Cloudor Pu Dec. 1, 2006

Dec. 1, 2006

Hi, all I am developing an swt app only for gtk. I want to embed IE into this app. I can create a SWT.EMBEDDED composite and get its embededHandle (an integer). Then, how to do next? Thanks in advace! regards, cloudor

3 3