July 2001 - Wine-devel - list.winehq.org

RE: Documentation fixes
by Patrik Stridvall 13 Jul '01

13 Jul '01

> Patrik Stridvall <ps(a)leissner.se> writes: > > > Like: > > > > > /************************************************************* > ********** > > * strcpy (MSVCRT.@) > > */ > > #if 0 /* Suppress warnings */ > > extern char *strcpy(char *dest, const char *src); > > #endif > > > > Perhaps we could do the same with atom functions and lstrcmpi? > > The #if 0 #endif isn't even needed for them I believe. > > That's quite ugly IMO. The #if 0 #endif or having to have a documentation comment at all? Anyway the #if 0 is not needed I thought it might be needed in some cases but it is not. I mistakingly believed it might be type conflicts but it can't be since any exported function must of course use the C library types otherwise you can't export them directly. If you are refering to the documentation comment, I can't really see why. All other functions have it and adding additional documentation might be a good idea in some cases as well. Furthermore the extern declaration makes it possible for the C compiler to check that the prototype is the same as the prototype in the included files. I spotted several error when doing the patch because of this. As hinted above, I have implemented a working solution that does what I have described above. See separate mail. > As long as we require comments, we might as > well put everything explicitly in the comment, something like: > > /* ##spec## msvcrt "@ cdecl strcpy(ptr str) strcpy" */ > > (we can probably come up with a better syntax, but you get the idea). > Then we can simply grep for '##spec##' and extract everything into the > spec file. We can do the same for forwards etc. (and in fact we could > do it for normal entry points as well). For forwards, stubs and other special cases, perhaps. I don't like it very much but it is acceptable. But we need not decide that now. I will think more about it and come back later with some kind of proposal that might or might not match yours. I'm not sure, we will see. For normal entry point, however, I respectfully disagree. One of the major reasons of having automatically generated .spec files is to avoid having to store redundant information. The needed information is already in the C function declaration and storing in a comment as well brings us back to the current situation. Sure it makes that generator less complicatated but it really isn't that important as long as it is reliable and not ambigeous. PS. For the record: I consider winapi_extract a prototype not something that nessary will finally use in "production". We might of course use it, but I would rather rewrite it so it becomes simpler and more reliable. Perhaps it shouldn't be in Perl but in C instead? Oh, well, that is a minor issue compared to how the information should be stored.

1 0

To strip or not to strip
by Jeremy White 13 Jul '01

13 Jul '01

We've always built the CodeWeavers Wine RPM unstripped, in the theory that it made supporting Wine newbies easier. However, AFAICT from watching cemw, we never get into a situation where a newbie goes into winedbg; more often, they shift first to the CVS tip, and from there maybe to winedbg. So, it seems like shipping debug symbols may just be a waste of bandwidth. (Note, we build with all full logging enabled, and we'll never change that). However, before we go ahead and yank the debug info out of our RPM, I thought I'd post a note, and let someone tell me if I'm wrong. Thanks, Jer

1 0

Re: Documentation fixes
by Alexandre Julliard 12 Jul '01

12 Jul '01

Patrik Stridvall <ps(a)leissner.se> writes: > Like: > > /*********************************************************************** > * strcpy (MSVCRT.@) > */ > #if 0 /* Suppress warnings */ > extern char *strcpy(char *dest, const char *src); > #endif > > Perhaps we could do the same with atom functions and lstrcmpi? > The #if 0 #endif isn't even needed for them I believe. That's quite ugly IMO. As long as we require comments, we might as well put everything explicitly in the comment, something like: /* ##spec## msvcrt "@ cdecl strcpy(ptr str) strcpy" */ (we can probably come up with a better syntax, but you get the idea). Then we can simply grep for '##spec##' and extract everything into the spec file. We can do the same for forwards etc. (and in fact we could do it for normal entry points as well). -- Alexandre Julliard julliard(a)winehq.com

2 1

RE: Documentation fixes
by Patrik Stridvall 12 Jul '01

12 Jul '01

> > The problem is that if I say that the memory directory doesn't > > "belong" to dlls/user/user.spec winapi_{check,extract,fixup} will > > not find implementations for all functions, which will cause > > problems. > > > > For example how should winapi_extract know that I should > generate the line > > 471 pascal16 lstrcmpi(str str) lstrcmpiA > > when it generates the spec files? > > Well, it's a more general problem, namely that the spec files can > reference functions that are not implemented directly in the dll. > > You also have for instance: > > @ cdecl strcpy(ptr str) strcpy > > in MSVCRT, and strcpy is clearly not implemented there since it's in > the C library. So you cannot assume that only the functions contained > in the dll can be used in the spec files, it is legal for the spec > file to reference external functions. True. I didn't even thought of that particular problem. Functions like strcpy is currently suppressed by having it listed in tools/winapi_check/nativeapi.dat, but that is a little kludgy. [Thinking...] Well you could simply have an function documentation comment without a function implementation in some file that belongs to MSVCRT. Like: /*********************************************************************** * strcpy (MSVCRT.@) */ #if 0 /* Suppress warnings */ extern char *strcpy(char *dest, const char *src); #endif Perhaps we could do the same with atom functions and lstrcmpi? The #if 0 #endif isn't even needed for them I believe. > > I think that we need to move the solution of this problem one step > > up the "ladder" and discuss the more general problem: > > How should we store the information needed to generate > > the spec files in the source, especially concerning > > the forwards and stubs? > > You'll need a file somewhere that specifies all the entries that > cannot be determined automatically from the source. You simply cannot > build a 100% correct spec file based only on the function comments. Yes. I have some ideas, but no really good one. My currently best suggestion is having a .spec.in file that approximately looks like the .spec files but supports configure like substitutions. foo.spec.in: ---8<--- name foo type win32 init @init@ rsrc foo.res import baz.dll @debug_channels@ 101 forward Bar BAZ.Bar 102 stub BarFoo @fixed@ @ forward Foo BAZ.Foo @ stub FooBar @unfixed@ ---8<--- And simply let winapi_extract or whatever simple substitute @init@, @debug_channels@, @fixed@ and @unfixed@ with what can be extracted from the source. > And you definitely do not want to need to parse KERNEL source files > (like the ones in memory/) to build the USER spec file. Of course. The question was whether the memory directory belongs to USER or not. I have now accepted that it does not.

1 0

Re: Documentation fixes
by Alexandre Julliard 12 Jul '01

12 Jul '01

Patrik Stridvall <ps(a)leissner.se> writes: > > No, USER does not use the memory directory. It links directly to > > libkernel32.so because it needs some internal functions in there, > > Like the atom functions that are implemented in memory/atom.c I guess > you mean. Actually no, the atom functions are similar to lstrcmpi, they are exported from kernel so we don't need to link to libkernel32.so to get them (and they shouldn't have USER comments either). The real problem is the internal kernel functions called by user, things like THUNK_Alloc or LOCAL_Alloc. > The problem is that if I say that the memory directory doesn't > "belong" to dlls/user/user.spec winapi_{check,extract,fixup} will > not find implementations for all functions, which will cause > problems. > > For example how should winapi_extract know that I should generate the line > 471 pascal16 lstrcmpi(str str) lstrcmpiA > when it generates the spec files? Well, it's a more general problem, namely that the spec files can reference functions that are not implemented directly in the dll. You also have for instance: @ cdecl strcpy(ptr str) strcpy in MSVCRT, and strcpy is clearly not implemented there since it's in the C library. So you cannot assume that only the functions contained in the dll can be used in the spec files, it is legal for the spec file to reference external functions. > I think that we need to move the solution of this problem one step > up the "ladder" and discuss the more general problem: > How should we store the information needed to generate > the spec files in the source, especially concerning > the forwards and stubs? You'll need a file somewhere that specifies all the entries that cannot be determined automatically from the source. You simply cannot build a 100% correct spec file based only on the function comments. And you definitely do not want to need to parse KERNEL source files (like the ones in memory/) to build the USER spec file. -- Alexandre Julliard julliard(a)winehq.com

1 0

RE: Documentation fixes
by Patrik Stridvall 12 Jul '01

12 Jul '01

> Patrik Stridvall <ps(a)leissner.se> writes: > > > PS. And yes, the added lstrcmpi (USER.471) in memory/string.c > > should really be there. Both KERNEL and USER uses the memory > > directory. > > No, USER does not use the memory directory. It links directly to > libkernel32.so because it needs some internal functions in there, Like the atom functions that are implemented in memory/atom.c I guess you mean. > but > that's a different problem. lstrcmpi is a purely KERNEL function, that > happens to be imported by USER (and by other dlls too), That might be so, but it is also _exported_ by USER: grep lstrcmpi dlls/user/user.spec 471 pascal16 lstrcmpi(str str) lstrcmpiA and lstrcmpiA is not implemented anyware else The same goes for the atom functions: grep Atom dlls/user/user.spec 268 pascal16 GlobalAddAtom(str) GlobalAddAtomA 269 pascal16 GlobalDeleteAtom(word) GlobalDeleteAtom 270 pascal16 GlobalFindAtom(str) GlobalFindAtomA 271 pascal16 GlobalGetAtomName(word ptr s_word) GlobalGetAtomNameA > so USER must > not be mentioned in the comment IMO. The problem is that if I say that the memory directory doesn't "belong" to dlls/user/user.spec winapi_{check,extract,fixup} will not find implementations for all functions, which will cause problems. For example how should winapi_extract know that I should generate the line 471 pascal16 lstrcmpi(str str) lstrcmpiA when it generates the spec files? Of course the automatic generation of spec files is not 100% complete most importantly handling of forwards and stubs is not supported. I think that we need to move the solution of this problem one step up the "ladder" and discuss the more general problem: How should we store the information needed to generate the spec files in the source, especially concerning the forwards and stubs? I will address this issue in a separate mail, but you are welcome to add comments specific to this problem in this thread, I'm not sure I feel to writing it today. I have been postponing the writing of mail for quite some time and I'm currenly beginning to get rather tired. PS. Why not as an interim solution simply add support for forward in the Win16 spec files so 471 pascal16 lstrcmpi(str str) lstrcmpiA becomes 471 forward lstrcmpi(str str) USER32.lstrcmpiA which winebuild until actual need simply interprets as 471 pascal16 lstrcmpi(str str) lstrcmpiA Hmm. Perhaps we should have forward16 and forward to separate the pascal16 and pascal cases? Anyway this gives winapi_{check,fixup} enough information to handle the situation where the memory directory doesn't "belong" to dlls/user/user.spec.

1 0

Re: New problems with the debugger
by Uwe Bonnes 12 Jul '01

12 Jul '01

>>>>> "Alexandre" == Alexandre Julliard <julliard(a)winehq.com> writes: Alexandre> Uwe Bonnes <bon(a)elektron.ikp.physik.tu-darmstadt.de> writes: >> WineDbg starting... file_set_error: No such process Can't attach >> process 8068d98: error 317 Alexandre> I don't see this here. Could you please run wineserver under Alexandre> gdb, put a breakpoint on the file_set_error message Alexandre> (server/file.c line 334), and send me the backtrace at this Alexandre> point? Thanks. Breakpoint 1, file_set_error () at file.c:334 334 default: perror("file_set_error"); set_error( ERROR_UNKNOWN /* FIXME */ ); break; (gdb) bt #0 file_set_error () at file.c:334 #1 0x804b1e5 in get_thread_context (thread=0x97fa210, flags=65537, context=0xbffff350) at context_i386.c:142 #2 0x804b567 in get_thread_single_step (thread=0x97fa210) at context_i386.c:481 #3 0x8051d2e in continue_thread (thread=0x97fa210) at ptrace.c:177 #4 0x8050e9f in resume_process (process=0x8068d98) at process.c:518 #5 0x804bfc6 in debugger_attach (process=0x8068d98, debugger=0x97fa778) at debugger.c:420 #6 0x804c278 in req_debug_process (req=0xbffff4d8) at debugger.c:540 #7 0x805669f in read_request (thread=0x97fa778) at request.c:168 #8 0x805922b in thread_poll_event (obj=0x97fa778, event=1) at thread.c:157 #9 0x8057642 in select_loop () at select.c:262 #10 0x804e5af in main (argc=1, argv=0xbffff6d4) at main.c:108 #11 0x40178baf in __libc_start_main () from /lib/libc.so.6 Perhaps to notice: I run 2.4.6 on a SMP machine, Suse 7.1 Hope this helps Bye -- Uwe Bonnes bon(a)elektron.ikp.physik.tu-darmstadt.de Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------

1 0

Re: Documentation fixes
by Alexandre Julliard 12 Jul '01

12 Jul '01

Patrik Stridvall <ps(a)leissner.se> writes: > PS. And yes, the added lstrcmpi (USER.471) in memory/string.c > should really be there. Both KERNEL and USER uses the memory > directory. No, USER does not use the memory directory. It links directly to libkernel32.so because it needs some internal functions in there, but that's a different problem. lstrcmpi is a purely KERNEL function, that happens to be imported by USER (and by other dlls too), so USER must not be mentioned in the comment IMO. -- Alexandre Julliard julliard(a)winehq.com

1 0

New problems with the debugger
by Uwe Bonnes 12 Jul '01

12 Jul '01

Hallo, since the update yesterday evening ( and a fresh update recently), on a crash the debugger has problems like this one: 08067330:Call kernel32.UnhandledExceptionFilter(404e4b00) ret=7800f2b8 fixme:pthread_kill_other_threads_np FIXME:pthread_rwlock_rdlock FIXME:pthread_rwlock_unlock Could not stat /home/bon/ikda/vol2 (No such file or directory), ignoring drive Y: FIXME:pthread_rwlock_rdlock FIXME:pthread_rwlock_unlock Could not stat /home/bon/ikda/sys (No such file or directory), ignoring drive Z: WineDbg starting... file_set_error: No such process Can't attach process 8068d98: error 317 Beendet Is this known? Bye -- Uwe Bonnes bon(a)elektron.ikp.physik.tu-darmstadt.de Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------

1 0

Re: The [Windows] section
by Alexandre Julliard 12 Jul '01

12 Jul '01

"Bill Medland" <medbi01(a)accpac.com> writes: > The answer, for those interested > > In Windows 95 a load of configuration information moved from win.ini (esp > the [windows] section) to the registry. Now (when the edits are added to > the cvs) information should come from the [windows] section of win.ini if > the winelook is WIN31 and from the registry in general otherwise, with the > Tweak.Layout overriding both. Actually I have hacked your patch a bit when committing it (hope you don't mind ;-), with the idea of phasing out the Tweak.Layout stuff. So now the values come from the registry for Win95, and from win.ini (plus some hard-coded values) for Win3.1. -- Alexandre Julliard julliard(a)winehq.com

1 0