May 2004 - Wine-Devel - list.winehq.org

Getting the API's for Windows Applications
by Jacobus Erasmus May 18, 2004

May 18, 2004

Is it possible to workout or read the API's that a specific Windows application use ? The theory is that it would be possible to estimate the amount of work needed to get a Windows app to work on Wine by calculating the number of API's that is not working or only half working and estimating the time it would take to correct these API's. By running wine-debug it is possible to get the API's used by an application. I'm hoping that there is some simpler way (although I'm not holding my breath) to read the API's used by an application directly from the .exe or .dll. I'm working on a utility that would collect this information and estimate the number of man hours it would take to get an application running on Wine. Also I'm hoping to build a utility that can collect stats on the API's being used in an application as well. The idea is to build a web page with these statistics for the applications that people want to run on a wine installation. If we can get some participation on the work and completeness of an API. It should be able to build a pretty accurate picture of the amount of work that needs to be done to get an app running under Wine. There is some other interesting side effects that I will not elaborate on until I have something in place (this is very simple but time does not permit. Somebody feels better equip to do this please say so) If not I will probably get arround to doing something this weekend. Cu jacobus(a)kewlconcepts.co.za

3 2

Re: Incoming Message
by Julliard May 18, 2004

May 18, 2004

1 0

RE: Protected message
by Dmitry May 18, 2004

May 18, 2004

1 0

Re: Re: Empty (Stubs in spec only) d3drm
by fenix＠club-internet.fr May 18, 2004

May 18, 2004

>On Tue, May 18, 2004 at 12:59:57AM +0200, Ivan Leo Murray-Smith wrote: >> This dll is needed by a game, bug report number 1202. With this empty dll, the >> game crashes at the same point where it crashes when used with the native copy, >> mainly because it crashes before any function in the dll is called. But at least >> it doesn't crash with a dll not found error. You'll have to run the autofoo >> stuff after applying this patch. > >Ivan, > >As far as I know, you did not submit also a patch to the default >configuration file to add this DLL with 'native' as default (which is a good >thing as all the people having a Windows box have this DLL - it could even >be installed if someone manages to install DirectX under Wine). > > Lionel > Hi Lionel, This dll should be easy to implement (almost geometries utilities functions as d3dx8/d3dx9 for d3d8/d3d9). It can be a begin for people who want to begin patches for wine (as for d3dx8/d3dx9) :) Regards, Raphael PS: d3dx8/d3dx9 have difficults parts on some textures and geometry loading (stream/unstream) and shaders utilities.

3 3

VirtualProtect bug?
by Robert Reif May 18, 2004

May 18, 2004

I'm trying to write a test that checks for data overruns using VirtualAlloc and VirtualProtect and I think I may have found a bug. Here is the problem: DWORD dwPageSize; BYTE * twoPages, temp; DWORD flOldProtect; twoPages = VirtualAlloc(NULL, 2 * dwPageSize, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE); res = VirtualProtect(twoPages + dwPageSize, dwPageSize, PAGE_NOACCESS, &flOldProtect); temp = twoPages[dwPageSize + 1]; // should cause exception but doesn't in wine twoPages[dwPageSize + 1] = 0; // does cause exception I would like to write a test to check the exception handling but I don't know how to do a __try and __except in C. Any ideas?

2 2

RE: Message Notify
by Julliard May 18, 2004

May 18, 2004

1 0

Re: [MSI] Added version info to MSI dll
by Filip Navara May 18, 2004

May 18, 2004

Christian Costa wrote: >+#define WINE_FILENAME_STR "amstream.dll" > This should be '#define WINE_FILENAME_STR "msi.dll"'. - Filip

2 1

Path too deep???
by Clare May 18, 2004

May 18, 2004

Hello groups, I would like to use Wine to run a Windows compiler "cc" to complie a c file "test.c" on Linux, but I have the problem. <try1> cd /a1/a2 wine -- /usr/local/bin/cc -c /b1/b2/b3/b4/b5/b6/test.c -o test.o Catastrophic error: could not open source file \b1\b2\b3\b4\b5\b6\test.c I try this in another way like this: <try2> cd /a1 wine -- /usr/local/bin/cc -c /b1/b2/b3/b4/b5/b6/test.c -o test.o This works well. Could you give me some informations why <try1> has the error? Thank you. Best Regards, Clare

2 1

RE: Text message
by Dmitry May 18, 2004

May 18, 2004

1 0

Re: Document
by Julliard May 18, 2004

May 18, 2004

1 0

New changes
by Dmitry May 18, 2004

May 18, 2004

1 0

Almost forgot...
by Robert van Herk May 18, 2004

May 18, 2004

This is the url of the website of the GP install people http://www.qsc.co.uk/source_ins.htm The idea is that you pay for another product (version controlling system) and then get the source code for GP install for free :-). Grtz, Robert

1 0

GP Install
by Robert van Herk May 18, 2004

May 18, 2004

Hi people, GP-install is a free win32 setup builder. I have been using it in Wine, and it is quite usable except for some irritating (wine) bugs. For example, it forgets that it is installed and before each run needs to be installed again... :-( I read on the GP install website that they have source available, though at a price. Would it be wise if I would mail them and ask whether for Wine they would grant us (or me) free access to their source code, just to patch up Wine a bit? After all, it would also increase their possible install base... Greetings, Robert

1 0

RE: Getting the API's for Windows Applications
by Rolf Kalbermatter May 18, 2004

May 18, 2004

Jacobus Erasmus <jacobus(a)kewlconcepts.co.za> wrote: >Is it possible to workout or read the API's that a specific Windows >application use ? > >The theory is that it would be possible to estimate the amount of work >needed to get a Windows app to work on Wine by calculating the number of >API's that is not working or only half working and estimating the time >it would take to correct these API's. This would be a very rough estimation. Some APIs are simple to implement and some are not. The ones not yet in Wine are often of the second class but that is not a general rule. >By running wine-debug it is possible to get the API's used by an >application. I'm hoping that there is some simpler way (although I'm not >holding my breath) to read the API's used by an application directly >from the .exe or .dll. It really depends. You could walk the import table of the executable both the standard and delayed imports and get quite far. dumpbin or whatever it is called can do that. However this will not account for dynamically imported APIs some applications do, to work for instance around API differences between different Windows version. Or another reason might be that a certain API is optional and the app should keep running eventhough that API is not available on the target system. I think to catch those APIs as well only API profiling would help. But it may not be the main issue here as the dynamic runtime linking is most often done to prevent the application from not starting up when the underlyiing API is not present. Rolf Kalbermatter

3 2

Hidden message
by Julliard May 18, 2004

May 18, 2004

1 0

Re: Thanks :)
by Dmitry May 18, 2004

May 18, 2004

1 0

Re: Empty (Stubs in spec only) d3drm
by Lionel Ulmer May 18, 2004

May 18, 2004

On Tue, May 18, 2004 at 12:59:57AM +0200, Ivan Leo Murray-Smith wrote: > This dll is needed by a game, bug report number 1202. With this empty dll, the > game crashes at the same point where it crashes when used with the native copy, > mainly because it crashes before any function in the dll is called. But at least > it doesn't crash with a dll not found error. You'll have to run the autofoo > stuff after applying this patch. Ivan, As far as I know, you did not submit also a patch to the default configuration file to add this DLL with 'native' as default (which is a good thing as all the people having a Windows box have this DLL - it could even be installed if someone manages to install DirectX under Wine). Lionel -- Lionel Ulmer - http://www.bbrox.org/

1 0

Re: IAutocomplete(5) Add sizing grip
by Dimitrie O. Paun May 17, 2004

May 17, 2004

On Sun, May 16, 2004 at 02:44:25PM +0200, Maxime Bellengé wrote: > Autocomplete list boxes have a sizing grip to resize them. This patch > adds support for that. I'm not sure how the sizing grip should look exactly, but it would be very nice if you could use DrawFrameControl(hdc, &rc, DFC_SCROLL, DFCS_SCROLLSIZEGRIP); to draw it. I'm sure it will make it a lot easier to theme the code in the (near) future. -- Dimi.

2 1

Patch for ReadFile with NULL hEvent
by Scott W Gifford May 17, 2004

May 17, 2004

According to MSDN, the hEvent member of an OVERLAPPED structure can be NULL: http://msdn.microsoft.com/library/default.asp?url=/library/en-us/dllproc/ba… OVERLAPPED ... Members ... hEvent Handle to an event set to the signaled state when the operation has been completed. The calling process must set this member either to zero or a valid event handle before calling any overlapped functions. A program I'm trying to run takes advantage of this, and as a result triggered a lot of these errors from Wine: err:file:GetOverlappedResult lpOverlapped->hEvent was null I wrote a patch that solves this, although it may not be the best solution. If when ReadFile is called overlapped->hEvent is NULL, the patch generates a new event with CreateEventA and sets the structure member to that event. To create a unique name, it uses the hex address of the overlapped structure along with a fairly long string, which is pretty likely to be unique. This works for the application I'm running. If this seems like a good enough solution, I can easily modify the other functions which initiate an overlapped read to do the same. ----ScottG.

2 5

StarCraft speed
by sysadmin May 17, 2004

May 17, 2004

Чистый листI have Celeron333(a)417+MatroxG200, FC1, kernel 2.6.6 latest Wine and try to speedup StarCraft. IMHO, there is bug in mga video driver or something related that do not allow wine to switch video modes from 16 to 8 bit and Wine falldown to 16->8 emulation. I seen the same in some posts at wine-devel. Week ago I try StarCraft on Athlon 2500(a)3200+ATI Radeon8500, same FC1,Wine and amazing how small speed increase. Using oprofile I track that 99% CPU spend in dib.c in X11DRV_XXXX_8 (sorry don’t remember exactly). Using printf I investigate that 99% CPU used to convert internal wine 8bit with palette 640x480 picture to 16 bit format. There is some asm magic there, I try to speedup it ( lodsb/loop was be slowest instr. in IA86 at least at Pentiun Classic ) and replace by my 8X loop unroll code but nothing changes. The question is: Does anybody look at curent directx 2d code alg ? May be StarCraft engine use some “update region” and current Wine DirectX does not track clip list exactly, so Wine waste CPU converting big 640x480 but not small regions ? And what status of dga2 , why it cannot change color depth ? /dev/mem permissions is 777. I try StarCraft as root and in 8 bit desktop and it load main menu and completely block keyboard and mouse (btw, menu animation work really fast J) so I must press reset to reboot.

1 0

A little question o suggestion...
by Urbez Santana Roma May 17, 2004

May 17, 2004

El Lunes, 8 de Marzo de 2004 23:20, wine-devel-request(a)winehq.org escribió: The file wine/server/file.c contains in line 151 to 156 (function create_file) the next lines: ----------------------------------------------------------------------------------------------------------------- mode = (attrs & FILE_ATTRIBUTE_READONLY) ? 0444 : 0666; if (len >= 4 && (!strcasecmp( name + len - 4, ".exe" ) || !strcasecmp( name + len - 4, ".com" ))) mode |= 0111; ------------------------------------------------------------------------------------------------------------------ That is a problem for a instal.lation programs or another programs that writen a Driver *.drv or a *.dll file, or another unknown extension. When created the file, the permisions not be 0555 or 0777, and the program instal.led not works, wine cannot execute them. My opinion, is that the files in a Windows Filesystem are virtually executable (the flag not exist, and all is an executable if one loads to exec) I cray that we must written the mode|=0111, whithout query the extension '.exe' '.com' '.drv' '.dll', etc... (note that the code only wants .exe and .com) I have write these modification and solutions the problem. ---------------------------------------------------------------------------------------------- mode = (attrs & FILE_ATTRIBUTE_READONLY) ? 0444 : 0666; mode |= 0111; ---------------------------------------------------------------------------------------------- -- Another problem is in file wine/dlls/kernel/ne_module.c : If the extension is not dll, exe, or com, the dll with another name cannot be executed. for example extensions with .drv .ldd .dat that are really DLL's cannot be executed, i have write a patch to solve it, but only i have added the extensions (that is not the correct solution). If u will, i can write a patch that detect if it is a NE module really without test the extension, (testing NE signature anz MZ, of the file...) ################################################ #- Urbez Santana i Roma - #- Email: urbez(a)linuxupc.upc.es #- Private Web: http://linuxupc.upc.es/~urbez/ ################################################

1 0

stdole2.tlb and stdole32.tlb
by Jonathan Wilson May 16, 2004

May 16, 2004

Is anyone working on getting sutable-to-use-in-wine versions of these? Would having these benifit OLE apps?

2 1

Re: VirtualProtect bug?
by Dan Kegel May 15, 2004

May 15, 2004

> I'm trying to write a test that checks for data overruns > using VirtualAlloc and VirtualProtect and I think > I may have found a bug. Here is the problem: > > DWORD dwPageSize; > BYTE * twoPages, temp; > DWORD flOldProtect; > twoPages = VirtualAlloc(NULL, 2 * dwPageSize, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE); > res = VirtualProtect(twoPages + dwPageSize, dwPageSize, PAGE_NOACCESS, &flOldProtect); > temp = twoPages[dwPageSize + 1]; // should cause exception but doesn't in wine > twoPages[dwPageSize + 1] = 0; // does cause exception Say, that code fragment wasn't complete, was it? Looks like dwPageSize was uninitialized... > I would like to write a test to check the exception handling > but I don't know how to do a __try and __except in C. OK, I don't know much about it either. A little googling shows: http://www.kerneltraffic.org/wine/wn20030516_170.html#8 makes it sound like __try is hard to do with gcc. You might need to punt and, when compiling with gcc, do nothing unless building in winelib, where you'd set up a handler for SIGSEGV to catch the bad access. http://www.codeproject.com/cpp/transactions.asp?df=100&forumid=11253&exp=0&… has some examples of how __try is used in visual C. (Note that you can download a copy of visual C's commandline compiler free from Microsoft.) What confuses me is, doesn't MinGW support __try? Hrm. - Dan -- My technical stuff: http://kegel.com My politics: see http://www.misleader.org for examples of why I'm for regime change

1 0

Patch to reduce warnings from comm.c
by Scott W Gifford May 15, 2004

May 15, 2004

I have two programs I run under Wine that make use of the serial ports. I get this error and fixme when I run then: err:comm:GetCommState tcgetattr or ioctl error 'Invalid argument' fixme:comm:SetupComm insize 1050 outsize 1050 unimplemented stub I fixed these two in my copy of Wine with the attached patch. First, the GetCommState errors were because I was using a pseudo-tty as the serial port, so setserial didn't work on it. I modified it to ignore errors from setserial that set errno==EINVAL. SetupComm is a function that sets the communications buffer sizes. Wine doesn't have its own communication buffers, so I filled out this function to set the kernel's xmit buffer to the requested amount if possible, and ignore the receive buffer. With these changes, my programs work properly (they worked properly before), and I don't get these complaints from Wine. -----ScottG.

1 0

Re: symbol hiding
by Dan Timis May 15, 2004

May 15, 2004

> From: Mike Hearn <mh(a)codeweavers.com> > Subject: Re: symbol hiding > Date: Sat, 15 May 2004 17:07:15 +0100 > Organization: CodeWeavers, Inc > > I think Pauls point was that if what you are doing is so secret and so > proprietary that you're afraid symbol names will leak trade secrets > you're > probably better off with some full blown code obfuscater/encryption > system. Actually, we do have an encryption system, and it is the name of the functions used internally by the encryption library that we want to hide. It is not for hiding trade secrets, but to prevent dishonest customers from accessing features without paying for them. This has been already used successfully on Mac and Windows where it is very easy to hide those names. Doing this will not make our copy protection full proof, but will slow down the inevitable hack. Hopefully, by the time someone will crack our copy protection we will have a new stronger version. Dan

2 1