February 2002 - Wine-Devel - list.winehq.org

RE: Some more MSVC fixes
by Patrik Stridvall Feb. 6, 2002

Feb. 6, 2002

> On Wed, 6 Feb 2002, Patrik Stridvall wrote: > > > Well. If you do as I did and just add just create > > a Visual Studio Project (.dsp) for each DLL using the > > MS headers is the default. I had to add "..\..\include" > > to the include path to get it to use the Wine header. > > Sorry Patrik if my questions are sometimes naive, but I don't have a > Windows partition, let alone MSVC. So, here goes the question: > > MSVC must have a command line version of the compiler. Yes, MS Visual Studio always calls the command line version to compile. It is called "cl.exe". > Wouldn't it be > easier to use our Makefile system under cygwin to compile > Wine with MSVC? Possible. However I don't know exactly cygwin works. I try to avoid Windows as much as possible. I know how MSVS project files work though so I tryed with them first. Personally I think that the ultimate solution would be to automatically generate the .dsp files from Makefile.in or otherwise. There is a huge advantage of using the Microsoft format since Visual Studio can read them and you can for example just point and click on the files of each DLL to open them inside Visual Studio. The .dsp format is not very difficult to generate. It is just some sort of glorified Makefile. > > As to how good it works with MS header: > > Well, the problem is that if you don't add the path the > > Wine header files like debugtools.h doesn't > > exists so it doesn't compile very well and if > > you do the Wine headers take priority. > > That's why all our private headers must reside in > include/wine. That needs > to be fixed, eventually. Hmm. Well, I gave the MS headers one more try and discovered a few not so nice things. If you have a Wine specific include like below in a file: #include "heap.h" and this file heap.h in turn has a #include "winbase.h" it will include the Wine version of the file instead of the MS version. This is probably because with "" files in the same directory has priority IIRC. This is solvable by doing #include <winbase.h> but this requires changes in almost all .h files but it is just a simple search and replace. Hmm. I wonder if doing this will cause any bad effect with GNU C or other compilers? > > Of course you could add the MS header for inclusion one > > more time after the Wine headers, but where they > > are under Windows are installation dependent and not > > very portable. :-( > > Yeah, but that could be determined by configure. True. > > Additionally windef.h must be included first. > > > > #include "winbase.h" > > #include "windef.h" > > > > doesn't work. You have to do > > Of course. I was thinking would be nice to have a > > #include "wine/prologue.h" > <whatever includes you have> > #include "wine/epilogue.h" > > > #include "windef.h" > > #include "winbase.h" > > > > This is a minor issue though. > > Which should be fixed, I think. Yes. It will investigate some more. > > I also so got some other strange error with PCONTEXT > > in winnt.h which in didn't understand. > > > > In short it was a terrible mess so I primarily tested > > it with the Wine headers. > > It would be sooo nice if we could compile with other headers > as well. Just > compiling would be great, running the resulting executable would be > fantastic! Yes. :-)

4 3

RE: Some more MSVC fixes
by Patrik Stridvall Feb. 6, 2002

Feb. 6, 2002

> > Personally I think that the ultimate solution would be to > > automatically generate the .dsp files from Makefile.in or otherwise. > > May not be easy but you are pretty good with perl so if anyone can > pull it off it's you :-) I don't think it is that difficult even for somebody less comfortable with Perl. The problem is getting all the details right and that will take some time, but allmost all details are quite easy, I think. > Once you have all the .dsp files, make a big .dsw file in the > top-level directory that references all the other .dsp files. That way > you can double-click on the .dsw, and then pick which dll you want to > build, build them all at once, etc. I guess you already know > about them > (and they are really simple to generate). Yes, I have it that way right now. But I only have a .dsp for a of few DLLs since doing it manually is quite boring. It would be quite nice to have all non core DLL compiling and working and with winetest ported. Then you could just run winetest through the debugger and be able to set breakpoint and step through the code in Visual Studio in order to get all to tests to run properly. :-) > > Hmm. Well, I gave the MS headers one more try and discovered > > a few not so nice things. > > > > If you have a Wine specific include like below in a file: > > #include "heap.h" > > and this file heap.h in turn has a > > #include "winbase.h" > > it will include the Wine version of the file instead of the > > MS version. This is probably because with "" files in the > > same directory has priority IIRC. > > No, that's because we use '-I .', which is a questionable practice. > Unfortunately I believe there is something that makes it hard to avoid > in Wine's case. I cannot remember what though. No, I don't think so. First of all -I. means current directory that is the directory of the current .c file. I was refering to that the #include "winbase.h" in file heap.h in wine/include perfered the winbase.h in wine/include rather than the MS variant. This is propably because #include "" with MSC always chooses the include from the same dir first rather the Microsoft dir that was EARLIER in the -I path (or rather the MSC equivalent). In the .c file #include "winbase.h" correctly included the MS variant. I haven't tested if GNU C have the same behavior but it might. > > > > Of course you could add the MS header for inclusion one > > > > more time after the Wine headers, but where they > > > > are under Windows are installation dependent and not > > > > very portable. :-( > > > > > > Yeah, but that could be determined by configure. > > > > True. > > But you would have to run configure on Windows, no? Since configure > is a shell script this means you need something like cygwin to compile > the source :-/ Or did you mean to run configure on Unix to > generate just > the right Makefiles for use on Windows? I didn't nessisarily mean configure but rather a Perl script that generates the .dsp files. Just install Perl and double click on configure.pl and all the .dsp was generated. Then double click on wine.dsw and you are ready to go.

1 0

compiling .obj with wine
by Ricardo Feb. 6, 2002

Feb. 6, 2002

hi, Is it possible to compile/linkedit .obj, not the source, with wine libs? thanks Ricardo.

2 1

RE: Some more MSVC fixes
by Patrik Stridvall Feb. 6, 2002

Feb. 6, 2002

> On Wed, 6 Feb 2002, Patrik Stridvall wrote: > [...] > > I also so got some other strange error with PCONTEXT > > in winnt.h which in didn't understand. > > You must define _X86_ before including 'winnt.h'. > Otherwise it won't > known on which architecture it's being compile and it won't define > CONTEXT. Yes, I realized that slighty after sending the mail. _M_IX86 is defined by the complier but _X86_ is defined by windows.h iif _M_IX86 is defined. > And you must also include 'windef.h'. Otherwise CONST is > not getting > defined and again, errors ensue. > > #define _X86_ > #include <windef.h> > #include <winnt.h> > > > Defining _X86_ is done in 'windows.h' which means that it is > virtually impossible to use these headers without including the big > beast itself. Yes. I have noticed that. However it is not that bad. #include "windef.h" #include "winnt.h" #include "stdarg.h" #include "winbase.h" Seems to do the trick. Another thing. It seems that all files will need to include "config.h" anyway because of issues like MSC doesn't support inline only __inline so a #define inline __inline is needed for anything reasonable to work anyway. Adding #include "config.h" at the top of each .c files is not that bad and things like #define _X86_ from windows.h could be there as well.

1 0

RE: Some more MSVC fixes
by Patrik Stridvall Feb. 6, 2002

Feb. 6, 2002

> On Tue, 5 Feb 2002, Patrik Stridvall wrote: > > > I have finally tried to compile Wine with MSVC. > > Eventhough I had to do some hack on the whole > > it compiled quite well. > > Great. On a slightly related note, is it possible that we compile Wine > with the MS headers? What would it take to do that? Well. If you do as I did and just add just create a Visual Studio Project (.dsp) for each DLL using the MS headers is the default. I had to add "..\..\include" to the include path to get it to use the Wine header. As to how good it works with MS header: Well, the problem is that if you don't add the path the Wine header files like debugtools.h doesn't exists so it doesn't compile very well and if you do the Wine headers take priority. Of course you could add the MS header for inclusion one more time after the Wine headers, but where they are under Windows are installation dependent and not very portable. :-( Additionally windef.h must be included first. #include "winbase.h" #include "windef.h" doesn't work. You have to do #include "windef.h" #include "winbase.h" This is a minor issue though. I also so got some other strange error with PCONTEXT in winnt.h which in didn't understand. In short it was a terrible mess so I primarily tested it with the Wine headers. I have only tested a few DLL since doing .dsp files for each them was some work. I might do a script to automatically generate .dsp files for MSVC some day. > If we > could, we would > have one more degree of freedom: > > Platform Compilers BinFmt Headers > -------- --------- ------ ------- > Linux gcc ELF Wine > *BSD MSVC PE MS > Solaris Borland ReactOS(?) > Cygwin > > Esentially, most of the above possible cobinations should > work (except, of > course Windows-base compilers and ELF). Well. I think GNU C chokes on the MS headers so that won't work either. > If we get this level of portability, would be nice to set an automatic > compile farm which tries all the possible combinations, and reports > problems early, so they get fixed. Yes, that would probably be a good idea.

3 2

Conferences
by Keith Matthews Feb. 6, 2002

Feb. 6, 2002

Speaking of conferences, is anyone planning on going to FOSDEM (16th & 17th Feb, Brussels) ? -- Keith Matthews Frequentous Consultants - Linux Services, Oracle development & database administration

2 1

RE: (FWD) Announcing wineconf 2002
by Patrik Stridvall Feb. 6, 2002

Feb. 6, 2002

> For those coming internationally, just hope that Microsoft > isn't waiting > at the airport with a bunch of DCMA lawyers to arrest you :) > > Mind you, M$ wouldn't do that because they don't like to > admit that Linux, > Wine, or anything else actually exists. I don't think Wine does anything DMCA related right now. However it will be much fun when we will get to the DRM parts of Windows. Dialog from the future: "Oh, no, this is not a circumvention device, we are just trying to implement copy protection to protect the intrests of RIAA and MPAA." "But what about the function FooIsAuthorized?" "What about it? It is just a stub that nobody for some reason have gotten around to implement. :-)" BOOL WINAPI FooIsAuthorized(...) { FIXME("(...): stub\n"); return TRUE; /* Information wants to be free */ }

1 0

RE: Header files not present on Windows
by Patrik Stridvall Feb. 5, 2002

Feb. 5, 2002

> Patrik Stridvall <ps(a)leissner.se> writes: > > > Hmm. I'm note sure I like making unistd.h > > an optional header but then few part of > > Wine should include it anyway so perhaps > > it doesn't matter. > > The question is whether you can build without it. If the needed > definitions are provided by other includes then it's fine to make it > optional; but if the code doesn't compile without it then there's no > point in having a check. Yes and in many cases it did work without unistd.h, but maybe that is because it really wasn't needed at all. In some case it might have worked because some of the functions in unistd.h is defined else where in Windows and in some case we might get it to work by including needed files. Of course in such cases we probably should fix Wine and include that other header exclusively instead of unistd.h. Anyway, I just experimented a little trying to get Wine to compile under Windows. I'm not really sure exacly what to do. Eventually I guess that the non-core DLL:s never should call any non Windows functions at all and the core DLL:s are not really useful to compile and run under Windows anyway.

1 0

Re: Header files not present on Windows
by Alexandre Julliard Feb. 5, 2002

Feb. 5, 2002

Patrik Stridvall <ps(a)leissner.se> writes: > Hmm. I'm note sure I like making unistd.h > an optional header but then few part of > Wine should include it anyway so perhaps > it doesn't matter. The question is whether you can build without it. If the needed definitions are provided by other includes then it's fine to make it optional; but if the code doesn't compile without it then there's no point in having a check. -- Alexandre Julliard julliard(a)winehq.com

1 0

TSXxxx functions: die!
by Dimitrie O. Paun Feb. 5, 2002

Feb. 5, 2002

Once again, Since this list is (again!) so quiet, I feel obligated to stir the spirits a little. So here it goes: we should get rid of the TSXxxx functions. Why? -- we don't keep to standard API, which is against the Wine's spirit -- they introduce yet another (useless) layer -- we reaquire the lock like crazy for no good reason -- most modern X systems are already thread safe -- we integrate Wine's locking with glic's, so we should be OK -- they are ugly How? s/TSX/X/g rm -rf tsx + misc Makefile cleanups But I hear people screaming: how about portability? What about platforms where we don't integrate with libc's locking, or when we don't have a thread safe X? Well, we can create a wrapper driver which simply aquires the lock, calls into the regular driver, then releases the lock. In fact, I am not convinced we shouldn't deal this way with the current driver, to simplify the code. It's not like we're missing a lot on performance... -- Dimi.

1 0