CVS - Notes Breakage

[Kevin DeKorte]

This mornings CVS of wine breaks Lotus Notes 5.0.11

This is the message I get from wine-dbg

Unhandled exception: unimplemented function user32.dll.SendIMEMessageExA 
called in 32-bit code (0x40afb052).
In 32-bit mode.
0x40afb052 (__wine_unimplemented+0x52 [user32.spec.c:42] in user32.dll.so): 
jmp0x40afb04c (__wine_unimplemented+0x4c [user32.spec.c:42] in user32.dll.so)
50      void __wine_stub_user32_dll_14(void) { 
__wine_unimplemented("BlockInput"); }

Kevin

[Uwe Bonnes]

> > > > > "Kevin" == Kevin DeKorte  writes:

Kevin> This mornings CVS of wine breaks Lotus Notes 5.0.11 This is the
    Kevin> message I get from wine-dbg

Kevin> Unhandled exception: unimplemented function
    Kevin> user32.dll.SendIMEMessageExA called in 32-bit code (0x40afb052).
    Kevin> In 32-bit mode.  0x40afb052 (__wine_unimplemented+0x52
    Kevin> [user32.spec.c:42] in user32.dll.so): jmp0x40afb04c
    Kevin> (__wine_unimplemented+0x4c [user32.spec.c:42] in user32.dll.so)
    Kevin> 50 void __wine_stub_user32_dll_14(void) {
    Kevin> __wine_unimplemented("BlockInput"); }

Are you sure you didn't change anything else? I doubt that wine itself will
call the unimplemented function SendIMEMessageExA. So either your Notes
setup/native dll setup changed or wine was fixed in some way that another
codepath is taken now that calls SendIMEMessageExA. Perhaps try to find out
where SendIMEMessageExA is called. depends.exe moght be of help/.

Anyways: SendIMEMessageExA seems undocumented:
Peace, another Win compatibility approach
http://chiharu.hauN.org/peace

has 
/*
 * Old IME API?
 */
EXTERN_C LRESULT WINAPI
SendIMEMessageExA(HWND p1, LPARAM p2)
{
        ADP("SendIMEMessageExA [not implemented]");
        /* XXX */
        return TRUE;
}

as stub implementation. Do you know how to integrate this in wine?
If not, tell me to send you a patch for testing.

It might however be that Notes _needs_ this function working or otherwise
needed functions will not work. We have to see.

Bye

[Kevin DeKorte]

Uwe,

Nope did not change anything, and in fact I rolled back to Wine 5/1/2003 and 
Notes works again.

Kevin

On Monday 05 May 2003 11:27 am, Uwe Bonnes wrote:
> > > > > > "Kevin" == Kevin DeKorte  writes:
> Kevin> This mornings CVS of wine breaks Lotus Notes 5.0.11 This is the
> Kevin> message I get from wine-dbg
>
>
> Kevin> Unhandled exception: unimplemented function
> Kevin> user32.dll.SendIMEMessageExA called in 32-bit code (0x40afb052).
> Kevin> In 32-bit mode.  0x40afb052 (__wine_unimplemented+0x52
> Kevin> [user32.spec.c:42] in user32.dll.so): jmp0x40afb04c
> Kevin> (__wine_unimplemented+0x4c [user32.spec.c:42] in user32.dll.so)
> Kevin> 50 void __wine_stub_user32_dll_14(void) {
> Kevin> __wine_unimplemented("BlockInput"); }
>
> Are you sure you didn't change anything else? I doubt that wine itself will
> call the unimplemented function SendIMEMessageExA. So either your Notes
> setup/native dll setup changed or wine was fixed in some way that another
> codepath is taken now that calls SendIMEMessageExA. Perhaps try to find out
> where SendIMEMessageExA is called. depends.exe moght be of help/.
>
> Anyways: SendIMEMessageExA seems undocumented:
> Peace, another Win compatibility approach
> http://chiharu.hauN.org/peace
>
> has
> /*
> - Old IME API?
>
> */
> EXTERN_C LRESULT WINAPI
> SendIMEMessageExA(HWND p1, LPARAM p2)
> {
>         ADP("SendIMEMessageExA [not implemented]");
>         /* XXX */
>         return TRUE;
> }
>
> as stub implementation. Do you know how to integrate this in wine?
> If not, tell me to send you a patch for testing.
>
> It might however be that Notes _needs_ this function working or otherwise
> needed functions will not work. We have to see.
>
> Bye

[Uwe Bonnes]

> > > > > "Kevin" == Kevin DeKorte  writes:

Kevin> Uwe, Nope did not change anything, and in fact I rolled back to
    Kevin> Wine 5/1/2003 and Notes works again.

Any chance that you try the approaches I mentioned?

Kevin> On Monday 05 May 2003 11:27 am, Uwe Bonnes wrote:
    Kevin> ...
Bye

[Sylvain Petreolle]

Uve, look at
http://www.winehq.com/hypermail/wine-cvs/2003/05/0005.html.
(03/05/02)
This patch adds user32.dll.SendIMEMessageExA 
as a stub in user32.spec.
> Any chance that you try the approaches I mentioned?
> Kevin> On Monday 05 May 2003 11:27 am, Uwe Bonnes wrote:
> Kevin> ...
>
> Bye

=====
Sylvain Petreolle (spetreolle at users dot sourceforge dot net) ICQ #170597259
No more War !

"What if tomorrow the War could be over ?" Morpheus, in "Reloaded".

For the Law of Oil and Fire, Im an European that lives in France.
For all my Brothers and friends, Im a human living on Earth.

___________________________________________________________
Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com

[Uwe Bonnes]

> > > > > "Sylvain" == Sylvain Petreolle  writes:

Sylvain> Uve, look at
    Sylvain> http://www.winehq.com/hypermail/wine-cvs/2003/05/0005.html.
    Sylvain> (03/05/02) This patch adds user32.dll.SendIMEMessageExA as a
    Sylvain> stub in user32.spec.

Oh.

Probably the function is called via GetProcAddress. If the function is not
listed in our spec file, GetProcAdress will fail. If it is listed,
GetProcAddress will succeed and when the prgram jumps to that address, a
crash will happen.

This makes me wonder if we should not return Failure toe GetProcAddress if a
stub-only function is requested. Investigating...

Bye

PS: Uve->Uwe and please no 3 mails ( to the list, to me and to me as CC)

[Sylvain Petreolle]

Im curious to see what adress GetProcAdress gets when asked for a stub.
> Probably the function is called via GetProcAddress. If the function
> is not
> listed in our spec file, GetProcAdress will fail. If it is listed,
> GetProcAddress will succeed and when the prgram jumps to that
> address, a
> crash will happen.

Look at the recipients you sent your mail :

De: "Uwe Bonnes"  | Ce mail
est un spam | Ajouter au carnet d'adresses 
À:  "Kevin DeKorte"  
CC: "Uwe Bonnes" ,
"[e-mail]"

It comes from you, is sent to wine-devel and to you. Hitting reply to
all sent my reply to you (as sender) and you + wine-devel as CC.
> PS: Uve->Uwe and please no 3 mails ( to the list, to me and to me as
> CC)

Argh silly me! Writing "Uve" after 7 years of german :)

=====
Sylvain Petreolle (spetreolle at users dot sourceforge dot net) ICQ #170597259
No more War !

"What if tomorrow the War could be over ?" Morpheus, in "Reloaded".

For the Law of Oil and Fire, Im an European that lives in France.
For all my Brothers and friends, Im a human living on Earth.

___________________________________________________________
Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com

[Uwe Bonnes]

> > > > > "Sylvain" == Sylvain Petreolle  writes:

Sylvain> Im curious to see what adress GetProcAdress gets when asked for
    Sylvain> a stub.

Wine happily returns a valid function adress, also obvious the same at least
for different stubbed functions in the same dll.  I searched for a way to
return perhaps some flag in  LdrGetProcedureAddress for that case, but I was
not able to find something. I didn't find a way how LdrGetProcedureAddress
could get that address and compare with the function address calculated for
the requested function.

Perhaps Eric or some other guru can...

I propose that GetProcAddress evaluates that flag and should return 0 for
stubs.

Bye

[Francois Gouget]

On Tue, 6 May 2003, Uwe Bonnes wrote:
[...]
> I propose that GetProcAddress evaluates that flag and should return 0 for
> stubs.

This may fix this problem but may also cause a lot of others: cases
where the application checks the value returned by GetProcAddress but
doesn't actually call the function (or only calls it in some rarely used
code path).

I think it would be better to transform SendIMEMessageExA into a
semi-stub. Hopefully that will be enough.

[Sylvain Petreolle]

Of course, its returning 0xdeadbeef :)

--- Uwe Bonnes  a écrit :
> > > > > > "Sylvain" == Sylvain Petreolle  writes:
> Sylvain> Im curious to see what adress GetProcAdress gets when
>
> asked for
>     Sylvain> a stub.
>
> Wine happily returns a valid function adress, also obvious the same

=====
Sylvain Petreolle (spetreolle at users dot sourceforge dot net) ICQ #170597259
No more War !

"What if tomorrow the War could be over ?" Morpheus, in "Reloaded".

For the Law of Oil and Fire, Im an European that lives in France.
For all my Brothers and friends, Im a human living on Earth.

___________________________________________________________
Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com

[Uwe Bonnes]

> > > > > "Sylvain" == Sylvain Petreolle  writes:

Sylvain> Of course, its returning 0xdeadbeef :) 
No longer. Wine sets up  a special exception now to display more information
when the function is accessed.

Bye

[Alexandre Julliard]

Uwe Bonnes  writes:
> Wine happily returns a valid function adress, also obvious the same at least
> for different stubbed functions in the same dll.  I searched for a way to
> return perhaps some flag in  LdrGetProcedureAddress for that case, but I was
> not able to find something. I didn't find a way how LdrGetProcedureAddress
> could get that address and compare with the function address calculated for
> the requested function.
>
> Perhaps Eric or some other guru can...
>
> I propose that GetProcAddress evaluates that flag and should return 0 for
> stubs.

There's no reason to do that; the whole point of declaring a stub is
to provide a function address. If we don't want that function at all
it should simply be commented out in the spec file.

additional user32 .spec entries was:Re: CVS - Notes Breakage

[Uwe Bonnes]

> > > > > "Alexandre" == Alexandre Julliard  writes:

Alexandre> Uwe Bonnes  writes:
    >> Wine happily returns a valid function adress, also obvious the same
    >> at least for different stubbed functions in the same dll.  I searched
    >> for a way to return perhaps some flag in LdrGetProcedureAddress for
    >> that case, but I was not able to find something. I didn't find a way
    >> how LdrGetProcedureAddress could get that address and compare with
    >> the function address calculated for the requested function.
    >> 
    >> Perhaps Eric or some other guru can...
    >> 
    >> I propose that GetProcAddress evaluates that flag and should return 0
    >> for stubs.

Alexandre> There's no reason to do that; the whole point of declaring a
    Alexandre> stub is to provide a function address. If we don't want that
    Alexandre> function at all it should simply be commented out in the spec
    Alexandre> file.

Andi,

as your patches causes breackage, please revisit with Alexandre's rule
applied.

Bye

Notes Breakage 2003-05- 08

[Kevin DeKorte]

This is still broken in the 5/8/2003 release of WINE.

Kevin

On Monday 05 May 2003 10:40 am, Kevin DeKorte wrote:
> This mornings CVS of wine breaks Lotus Notes 5.0.11
>
> This is the message I get from wine-dbg
>
> Unhandled exception: unimplemented function user32.dll.SendIMEMessageExA
> called in 32-bit code (0x40afb052).
> In 32-bit mode.
> 0x40afb052 (__wine_unimplemented+0x52 [user32.spec.c:42] in user32.dll.so):
> jmp0x40afb04c (__wine_unimplemented+0x4c [user32.spec.c:42] in
> user32.dll.so) 50      void __wine_stub_user32_dll_14(void) {
> __wine_unimplemented("BlockInput"); }
>
> Kevin

Notes Breakage 2003-05- 08

[Sylvain Petreolle]

what about sending a tiny stub implementation ?

--- Kevin DeKorte  a écrit :
> This is still broken in the 5/8/2003 release of WINE.
>
> Kevin
>
> On Monday 05 May 2003 10:40 am, Kevin DeKorte wrote:
> > This mornings CVS of wine breaks Lotus Notes 5.0.11
> >
> > This is the message I get from wine-dbg
> >
> > Unhandled exception: unimplemented function
>
> user32.dll.SendIMEMessageExA
> > called in 32-bit code (0x40afb052).
> > In 32-bit mode.
> > 0x40afb052 (__wine_unimplemented+0x52 [user32.spec.c:42] in
>
> user32.dll.so):
> > jmp0x40afb04c (__wine_unimplemented+0x4c [user32.spec.c:42] in
> > user32.dll.so) 50      void __wine_stub_user32_dll_14(void) {
> > __wine_unimplemented("BlockInput"); }
> >
> > Kevin

=====
Sylvain Petreolle (spetreolle at users dot sourceforge dot net) ICQ #170597259
No more War !

"What if tomorrow the War could be over ?" Morpheus, in "Reloaded".

For the Law of Oil and Fire, Im an European that lives in France.
For all my Brothers and friends, Im a human living on Earth.

___________________________________________________________
Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com

Notes Breakage 2003-05- 08

[Uwe Bonnes]

> > > > > "Sylvain" == Sylvain Petreolle  writes:

Sylvain> what about sending a tiny stub implementation ?  --- Kevin

I think writing stub implementations for functions not available in all
windows implementation and where the windows program only load them with
GetProcAdrress and so can cope with the function not available is plain
wrong.

That way we tell the windows program the function _is_ available, so it will
rely on the implementation, but then we provide no implementation...

Andy, do you follow the problems resulting from your last contribution?

Bye