October 2003 - Wine-Devel - list.winehq.org

Re: DINPUT device unacquire (mouse) fix
by Andreas Mohr Oct. 29, 2003

Oct. 29, 2003

Hi, On Tue, Oct 28, 2003 at 02:53:44AM +0200, Hannu Valtonen wrote: > License: LGPL > > I stumbled upon this when looking at fixing mouse support for Crimson > lands. According to MSDN DirectInputDevice Unacquire is allowed even for > devices which haven't been acquired yet. > > Changelog: > - SysMouseAImpl_Unacquire should return DI_NOEFFECT if the device is not > in an acquired state. Ah, thanks!! Maybe this is why some game (I believe it was UT) broke... Darn, why didn't I discover that when I stumbled on that place years ago? But don't you think keyboard and joystick might need the same code?? (keyboard has it as well, but with completely different code --> mess) This could probably be cleaned up. Andreas Mohr

2 2

WinE-Bug? InternetCrackUrl incorrectly sets dwUrlPathLength
by Jan Sporbeck Oct. 29, 2003

Oct. 29, 2003

Hi, During some application testing, we discovered that WinE does not correctly mimic the WinINet function InternetCrackUrl. The real WinAPI sets the dwUrlPathLength member of the URL_COMPONENTS structure to additionally hold the extra info of the given URL if the extra info is not requested as a separate component. Although this behaviour is not well documented by the Win SDK, it is for instance implemented on WinXP, Win2000 and WinME and therefore might be considered to be intended by Microsoft. The following lines of code demonstrate the problem: LPCTSTR lpszURL = _T("http://www.myserver.com/myscript.php?arg=1"); URL_COMPONENTS url; // call 1: crack out url path and extra info ::ZeroMemory(&url, sizeof(url)); url.dwStructSize = sizeof(url); url.dwHostNameLength = 1; url.dwUrlPathLength = 1; url.dwExtraInfoLength = 1; ::InternetCrackUrl(lpszURL, 0, 0, &url); // dwUrlPathLength is 13, holding "/myscript.php" // call 2: only crack out url path ::ZeroMemory(&url, sizeof(url)); url.dwStructSize = sizeof(url); url.dwHostNameLength = 1; url.dwUrlPathLength = 1; ::InternetCrackUrl(lpszURL, 0, 0, &url); // now, dwUrlPathLength should be 19, holding "/myscript.php?arg=1", // while WinE still sets it to 13 OK, it is of course no great deal to write code that correctly runs on real Win systems and on WinE (2-3 lines did it). But programers who rely on this behaviour will for example wonder why their applications don't transmit any HTTP GET parameters when they call HttpOpenRequest with lpszObjectName set to the URL path component returned by InternetCrackUrl... This applies to Wine 20031016, haven't found any recent report on this though. /jan.

1 0

Re: spec files - syntax and generation
by Jerry Jenkins Oct. 29, 2003

Oct. 29, 2003

Martin Tröster wrote: > How do I find out whether a function uses CDECL instead of STDCALL? I found the information (link lost unfortunately) that entries in the def file like testfunc1(a)0 are called via STDCALL, whereas in case of testfunc2 without any @ specifying the space is CDECL. Is this guess right? If not, how else do I find this out? The most safe one I have known is to disassemble it and look over the output. If the function have no parameters, you can declare it as either CDECL or STDCALL. Suppose it have some parameters and its type is CDECL, you can find: 1. "lret" in the function 2. it will be called in a way like pushl something1 pushl something2 call address of testfunc1(a)0 addl $8, %%esp and for a STDCALL, there are: 1. something that is similar to "lret $8" in the function 2. it will be called in a way like pushl something1 pushl something2 call address of testfunc1(a)0 > > Thanks again for any help! > > Cheers, > Martin

2 1

Re: spec files - syntax and generation
by Shachar Shemesh Oct. 29, 2003

Oct. 29, 2003

Martin Tröster wrote: >Hi, > >- Stdcall vs. cdecl >How do I find out whether a function uses CDECL instead of STDCALL? I found the information (link lost unfortunately) that entries in the def file like testfunc1(a)0 are called via STDCALL, whereas in case of testfunc2 without any @ specifying the space is CDECL. Is this guess right? If not, how else do I find this out? > > > STDCALL means PASCAL call format. This means two things, mostly. First - it means arguments are pushed on the stack left to right instead of right to left (no way you can pass printf style variable length arguments). The second is that it is the functions' responsibility, rather than the caller, to remove the arguments from the stack. With the C calling convention, if the caller places more arguments on the stack than the function expects, nothing bad will happen. The same caller (that presumably knows how much arguments there are on the stack) releases the stack. As the arguments are pushed right to left, the first argument is always in the same position in relation to the stack pointer when calling the function. This means that if the function expects three arguments, and the caller sends 4, noone will ever notice (assuming the first three are understood correctly, of course). With the Pascal calling convention, such a case is disastrous. As the arguments are cleared by the called function, if it thinks there is a different number or arguments on the stack than there really is, the function will return to the caller with the stack pointer in the wrong place. The program is most likely to crash the moment any further stack activity takes place. In order to avoid this problem, STDCALL functions mangle into the function's name a number indicating the number of bytes the arguments for the function take on the stack. This gives a certain assurance that a mismatch between caller and callee notion of what the arguments are will be translated into a linker failure, rather than inexplicable, seemingly random, runtime crashes. Warning: the following paragraph is pure guesswork. The long explanation was just a way to say that your idea that symbols that have the form of "name(a)number" are STDCALL symbols, while symobls that are pure name are CDECL, seems the only way to understand things. >Thanks again for any help! > >Cheers, >Martin > > Shachar -- Shachar Shemesh Open Source integration consultant Home page & resume - http://www.shemesh.biz/

2 1

Re: spec files - syntax and generation
by Alexandre Julliard Oct. 29, 2003

Oct. 29, 2003

"Martin Tröster" <troymcclure(a)web.de> writes: > - Stack size and parameters in spec file: > The def files are, as far as I know, very similar to the spec files: > they contain the function name (either mangled or not mangled), as > well as (most of the time) the stack space - format: > functionname(a)stackspace. Is there anything wrong to translate an > entry for "char* testfunc(char* str, int var1, long var2, long* > ptr)" called "testfunc(a)16" into "@stdcall testfunc(ptr ptr ptr > ptr)", without caring for the type of parameters? More explicitely > asked, what use do I have from using str or long instead of ptr in > the spec file? All it does is allow you to get more readable output from the relay traces. If you don't care about that, using 'ptr' everywhere works just fine. > - Stdcall vs. cdecl > How do I find out whether a function uses CDECL instead of STDCALL? > I found the information (link lost unfortunately) that entries in > the def file like testfunc1(a)0 are called via STDCALL, whereas in > case of testfunc2 without any @ specifying the space is CDECL. Is > this guess right? If not, how else do I find this out? Yes this is correct, the stdcall functions are the ones decorated with @xxx. -- Alexandre Julliard julliard(a)winehq.com

1 0

Wine coredumps with Linux kernel 2.6.0-test9
by David D. Hagood Oct. 28, 2003

Oct. 28, 2003

I'm playing with 2.6.0-test9 under RH9.0, and Wine just coredumps when I run it. This is after a full rebuild of Wine, with a CVS pull of a couple of days ago. Since I've seen some other apps behaving weirdly this may be a compatibility issue with 2.6, but I thought I'd see if anybody else has seen this, and has any suggestions. I have updated glibc to the latest in Rawhide.

3 4

Re: PATCH: server/named_pipe.c (RESENT)
by Alexandre Julliard Oct. 28, 2003

Oct. 28, 2003

Gerald Pfeifer <gerald(a)pfeifer.com> writes: > was there any problem with the following patch, or did it just fall > through the cracks? ;-) > > Gerald > > ChangeLog: > Remove unused variable pipe_client_fd_ops. Well, either the client side should really use pipe_client_fd_ops, or pipe_server_fd_ops should be renamed to show that it is used by both client and server. I'm not sure which one was intended, could the named pipe experts please check that? -- Alexandre Julliard julliard(a)winehq.com

1 0

Support for TLS in CrossOver
by Mike Hearn Oct. 28, 2003

Oct. 28, 2003

Hi Alexandre, I noticed the ChangeLog for the new CXOffice release (congrats on that btw!) said it now works with all glibc 2.3 variants. Does that mean the TLS related named pipe breakage we saw earlier is now fixed? If so, do you know when the fix will be in WineHQ CVS? thanks -mike

3 3

WineHQ: PATCH to fix IPersistFile::Load() and associated functions
by Subhobroto Sinha Oct. 28, 2003

Oct. 28, 2003

Hello all I finally produced the "official patch" after verifying that everything is OK. This patch has been inlined as well as attached as a ZIP file, so suit yourself. All reports regaring this patch is to be sent to <subhobrotosinha at yahoo.com> with a proper subject prefixed by "IPersistFile patch" in the subject header. Just don't simply hit the 'Reply' button ! This patch fixes: (1) Stream_LoadString: Conflict of unicode strings being taken as LPSTR and AGAIN being converted to Unicode due to wrong value of SCF_UNICODE .(I have really NOT changed the original code, but redirected this call to a fixed function which I named wineStream_LoadString. This should remain so until SCF_UNICODE value is corrected) (2) Stream_LoadLocation: Previously the GetPath() failed. It works great now. For the uninitiated, here's what has to be done: Please make sure that you have the 20031016 source (latest tarball) (1)Copy this .diff to the root folder of the Wine source (2)Run 'patch' on this diff file (3)It should patch "dlls/shell32/shelllink.c". Change to that directory (4)Type 'make' to rebuild 'shell32.dll.so' (5)'su' out and overwrite/symlink the current 'shell32.dll.so' over the installed one Now IPersistFile::Load() should work as intended. Try out McCormack's 'winemenubuilder' or my WineLib 'linkresolve' (NOT my C++ version) If the diff is not proper, etc, please alert me at the mentioned address PS: As I am not a member of 'wine-patches',can the moderator be kind enough... :=) I really want these functions to get fixed. That's all. Regards Subhobroto Sinha __________________________________ Do you Yahoo!? Exclusive Video Premiere - Britney Spears http://launch.yahoo.com/promos/britneyspears/ --- dlls/shell32/shelllink.c.original 2003-10-22 17:56:31.000000000 +0530 +++ dlls/shell32/shelllink.c 2003-10-26 16:30:57.000000000 +0530 @@ -401,9 +401,49 @@ return S_OK; } +static HRESULT wineStream_LoadString( IStream* stm, LPWSTR *pstr ) +{ + DWORD count=0; + USHORT len; + LPWSTR str=NULL; + HRESULT r; + /* + From Win98 upwards unicode==TRUE.(Atleast for shortcuts..) + However, the value of SCF_UNICODE does NOT seem to be 0x1000. + Until then, we assume it true. + Anybody differing mail to <subhobrotosinha at yahoo.com>. + */ + TRACE("%p\n", stm); + + r = IStream_Read(stm, &len, sizeof(len), &count); + if(FAILED(r)||(count != sizeof(len))) return E_FAIL; + + len *= sizeof(WCHAR); + + TRACE("reading %d\n", len); + str = HeapAlloc(GetProcessHeap(), 0, len+sizeof(WCHAR)); + if(!str) return E_OUTOFMEMORY; + + count = 0; + r = IStream_Read(stm, str, len, &count); + if( FAILED(r)||( count!=len) ) + { + HeapFree( GetProcessHeap(), 0, str ); + return E_FAIL; + } + + str[count/2]=0; + *pstr = str; + TRACE("read %s\n", debugstr_w(str)); + return S_OK; +} static HRESULT Stream_LoadString( IStream* stm, BOOL unicode, LPWSTR *pstr ) { +/*This is the original function short circuited to the above till we find out correct 'SCF_UNICODE'. +Now it's assumed that the data's unicode. (This assumption failes iff the OS was <=Win95*/ +return (wineStream_LoadString(stm,pstr)); +#if 0 DWORD count; USHORT len; LPVOID temp; @@ -453,6 +493,7 @@ *pstr = str; return S_OK; +#endif } static HRESULT Stream_LoadLocation( IStream* stm ) @@ -533,6 +574,14 @@ if( FAILED( r ) ) return r; } + ////////////////////////////////////////////////////////////////////////////////////// + char szTemp[MAX_PATH]={0}; + SHGetPathFromIDListA(This->pPidl,szTemp); + This->sPath=HeapAlloc( GetProcessHeap(), 0,(strlen(szTemp)+1)*sizeof(WCHAR)); + MultiByteToWideChar(CP_ACP,0,szTemp,(strlen(szTemp)+1),This->sPath,(strlen(szTemp)+1)); + //TRACE("%s\n",szTemp); + ////////////////////////////////////////////////////////////////////////////////////// + This->wHotKey = hdr.wHotKey; This->iIcoNdx = hdr.nIcon; FileTimeToSystemTime (&hdr.Time1, &This->time1);

3 3

Winesetuptk 0.7 released
by Ivan Leo Murray-Smith Oct. 28, 2003

Oct. 28, 2003

Hi lists, Winesetuptk 0.7 is at the wine sourceforge page http://sourceforge.net/project/showfiles.php?group_id=6241 This new version contains an updated winedefault.reg, you don't need to run regedit winedefault.reg any more, and it's updated to configurate all the latest and greatest features of wine. This release was possible thanks to the help and contributions of Vincent Béron and Alex Pasadyn. Without them, winesetuptk 0.7 wouldn't exist. Download and enjoy!

1 0