Marshal bug

[Duane Clark]

Howdy,

This patch:
http://www.winehq.org/hypermail/wine-cvs/2004/05/0283.html

causes a crash within Actel Designer (commercial FPGA design software). 
Reverting the patch with current CVS fixes the application.

A trace and crash dump look like
> wine /c/Actel/bin/designer.exe
> trace:ole:DllMain 0x40ea0000 0x1 0x1
> trace:ole:DllMain (0x41030000,1,0x1)
> trace:ole:OleInitialize ((nil))
> trace:ole:CoInitializeEx ((nil), 2)
> trace:ole:CoInitializeEx () - Initializing the COM libraries
> trace:ole:RunningObjectTableImpl_Initialize ()
> trace:ole:OleInitialize () - Initializing the OLE libraries
> trace:ole:OLEClipbrd_Initialize ()
> fixme:ole:CoRegisterMessageFilter stub
> trace:ole:RegisterDragDrop (0x10040,0x41fc1ddc)
> trace:ole:UuidCreate {8b881308-d111-11d8-8790-002078041c53}
> trace:ole:__CLSIDFromStringA {A111826C-DB40-11d2-8D34-0008C7523E29} -> 
> 0x4073f218
> trace:ole:WINE_StringFromCLSID 
> 0x4073f218->{A111826C-DB40-11D2-8D34-0008C7523E29}
> trace:ole:CoLoadLibrary (L"C:\\Actel/bin/comdesign.dll", 0)
> trace:ole:CoCreateFreeThreadedMarshaler (0x41fc6a60 0x41fc6a74)
> trace:ole:__CLSIDFromStringA {4B95E7E3-6EC5-11D2-BA21-0008C7A01E5D} -> 
> 0x4073f870
> trace:ole:WINE_StringFromCLSID 
> 0x4073f870->{4B95E7E3-6EC5-11D2-BA21-0008C7A01E5D}
> trace:ole:CoLoadLibrary (L"C:\\Actel/bin/timingui.dll", 0)
> trace:ole:__CLSIDFromStringA {F70AA652-C79A-11d2-BA4B-0008C7A01E5D} -> 
> 0x4073f66c
> trace:ole:WINE_StringFromCLSID 
> 0x4073f66c->{F70AA652-C79A-11D2-BA4B-0008C7A01E5D}
> trace:ole:CoLoadLibrary (L"C:\\Actel/bin/taiming.dll", 0)
> err:toolbar:TOOLBAR_GetImageListForDrawing bitmap for ID 0, index 0 is 
> not valid, number of bitmaps in imagelist: 0
> err:toolbar:TOOLBAR_GetImageListForDrawing bitmap for ID 0, index 0 is 
> not valid, number of bitmaps in imagelist: 0
> err:toolbar:TOOLBAR_GetImageListForDrawing bitmap for ID 0, index 0 is 
> not valid, number of bitmaps in imagelist: 0
> err:toolbar:TOOLBAR_GetImageListForDrawing bitmap for ID 0, index 0 is 
> not valid, number of bitmaps in imagelist: 0
> trace:ole:CoMarshalInterThreadInterfaceInStream 
> ({27a4f276-6452-11d2-ba1d-0008c7a01e5d}, 0x43932bf8, 0x4073f9e8)
> trace:ole:CoMarshalInterface (0x402b4180, 
> {27a4f276-6452-11d2-ba1d-0008c7a01e5d}, 0x43932bf8, 3, (nil), 0)
> trace:ole:_StubMgrThread Stub Manager Thread starting on 
> (\\.\pipe\WINE_OLE_StubMgr_00000008)
> trace:ole:CoCreateFreeThreadedMarshaler (0x41fc72f0 0x41fc7348)
> trace:ole:IiFTMUnknown_fnQueryInterface
> trace:ole:FTMarshalImpl_AddRef
> fixme:ole:FTMarshalImpl_GetUnmarshalClass (), stub!
> fixme:ole:FTMarshalImpl_MarshalInterface (), stub!
> trace:ole:FTMarshalImpl_Release
> trace:ole:CoInitializeEx ((nil), 2)
> trace:ole:CoGetInterfaceAndReleaseStream (0x402b4180, 
> {27a4f276-6452-11d2-ba1d-0008c7a01e5d}, 0x46cb413c)
> trace:ole:CoUnmarshalInterface 
> (0x402b4180,{27a4f276-6452-11d2-ba1d-0008c7a01e5d},0x46cb413c)
> trace:ole:WINE_StringFromCLSID 
> 0x46cb4014->{402B4180-FA0C-40EE-EBF8-EE40F0CE5360}
> trace:ole:CoGetClassObject
>          CLSID:  {402b4180-fa0c-40ee-ebf8-ee40f0ce5360},
>          IID:    {00000001-0000-0000-c000-000000000046}
> warn:ole:CoGetClassObject class {402B4180-FA0C-40EE-EBF8-EE40F0CE5360} 
> not registred
> trace:ole:WINE_StringFromCLSID 
> 0x46cb4014->{402B4180-FA0C-40EE-EBF8-EE40F0CE5360}
> trace:ole:WINE_StringFromCLSID 
> 0x46cb4014->{402B4180-FA0C-40EE-EBF8-EE40F0CE5360}
> fixme:ole:CoCreateInstance no classfactory created for CLSID 
> {402b4180-fa0c-40ee-ebf8-ee40f0ce5360}, hres is 0x80040150
> fixme:ole:CoUnmarshalInterface Failed to create instance of unmarshaller 
> {402b4180-fa0c-40ee-ebf8-ee40f0ce5360}.
> wine: Unhandled exception (thread 000d), starting debugger...
> fixme:console:SetConsoleCtrlHandler (0x4066a5e0,1) - no error checking 
> or testing yet
> WineDbg starting on pid 0x8
> Unhandled exception: page fault on read access to 0x00000000 in 32-bit 
> code (0x6051b58f).
> In 32 bit mode.
> err:dbghelp_msc:pdb_process_file -Unable to peruse .PDB file 
> e:\users\ChiJ\objs\pc_4.0.3.5\timingui\msobj\timingui.pdb
> 0x6051b58f: movl        0x0(%eax),%ecx
> Wine-dbg>bt
> Backtrace:
> =>1 0x6051b58f (0x46cb4160)
>    2 0x40374e57 THREAD_Start+0xef(ptr=0x402b3dd8) [thread.c:107] in 
> kernel32 (0x46cb4234)
>    3 0x400a2f23 start_thread+0x147(info=0x402b05e0) [thread.c:199] in 
> ntdll (0x46cb4a70)
>    4 0x40043484 start_thread+0x64 in libpthread.so.0 (0x46cb4a94)
>    5 0x420df147 __clone+0x57 in libc.so.6 (0x00000000)
> Wine-dbg>

[Mike Hearn]

What happens if you reinstall the app or reregister all the DLLs shipped
with it (using regsvr32) ? Does that fix it ?

thanks -mike

[Duane Clark]

Mike Hearn wrote:
> What happens if you reinstall the app or reregister all the DLLs shipped
> with it (using regsvr32) ? Does that fix it ?

Because of some reconfiguration on my end, I performed a new clean 
install in a fresh c: and fresh ~/.wine/config just a couple of days 
ago, with a CVS from then.

[Duane Clark]

Duane Clark wrote:
> Mike Hearn wrote:
> > What happens if you reinstall the app or reregister all the DLLs shipped
> > with it (using regsvr32) ? Does that fix it ?
>
> Because of some reconfiguration on my end, I performed a new clean 
> install in a fresh c: and fresh ~/.wine/config just a couple of days 
> ago, with a CVS from then.

If I had some idea of what is going on in the traces, I would attempt to 
do some debugging. Do you know of any "Marshalling for dummies" or some 
such thing? Currrently the traces are complete jibberish to me.

Is this likely to be a result of the unimplemented functions in the traces?
fixme:ole:FTMarshalImpl_GetUnmarshalClass (), stub!
fixme:ole:FTMarshalImpl_MarshalInterface (), stub!

[Mike Hearn]

On Fri, 09 Jul 2004 10:17:07 -0700, Duane Clark wrote:
> If I had some idea of what is going on in the traces, I would attempt to
> do some debugging. Do you know of any "Marshalling for dummies" or some
> such thing? Currrently the traces are complete jibberish to me.
>
> Is this likely to be a result of the unimplemented functions in the
> traces? fixme:ole:FTMarshalImpl_GetUnmarshalClass (), stub!
> fixme:ole:FTMarshalImpl_MarshalInterface (), stub!

Hmm, well, OLE marshalling isn't something you can explain in 5 minutes
... I keep meaning to do this but if you like I'll post a complete (to my
knowledge) explanation of how it all works say over the weekend or on
Monday.

But to get you started, I don't know if the FTM stubs are a problem or
not, but this bit almost certainly is:

trace:ole:CoUnmarshalInterface (0x402b4180,{27a4f276-6452-11d2-ba1d-0008c7a01e5d},0x46cb413c)
trace:ole:WINE_StringFromCLSID 0x46cb4014->{402B4180-FA0C-40EE-EBF8-EE40F0CE5360}
trace:ole:CoGetClassObject
         CLSID:  {402b4180-fa0c-40ee-ebf8-ee40f0ce5360}, 
	 IID:    {00000001-0000-0000-c000-000000000046}
warn:ole:CoGetClassObject class {402B4180-FA0C-40EE-EBF8-EE40F0CE5360} not registered

Basically as part of unmarshalling an interface, a "proxy/stub class
factory" is activated. This is a COM object that implements
IPSFactoryBuffer:

http://msdn.microsoft.com/library/default.asp?url=/library/en-us/com/htm/cmi...

which as you can see can then be used to construct proxy and stub objects.
Therefore if this object cannot be activated for whatever reason,
unmarshalling cannot continue and we bail out, handing the app a
null vtable which it then dereferences.

So you need to figure out why {402b4180-fa0c-40ee-ebf8-ee40f0ce5360} isn't
being registered, which may involve figuring out what it actually is :)

Good luck! If a more complete marshalling explanation would be useful, let
me know and I'll see what I can do in a few days.

thanks -mike

[Duane Clark]

Mike Hearn wrote:
> ...
> So you need to figure out why {402b4180-fa0c-40ee-ebf8-ee40f0ce5360} isn't
> being registered, which may involve figuring out what it actually is :)

Okay, I'll see if I can make some sense out of that.
> Good luck! If a more complete marshalling explanation would be useful, let
> me know and I'll see what I can do in a few days.

There is no hurry, since it works fine if I back out the one patch. But 
yes, I certainly would appreciate a more complete explanation.

[Robert Shearman]

On Fri, 2004-07-09 at 21:18, Duane Clark wrote:
> Mike Hearn wrote:
> > ...
> > So you need to figure out why {402b4180-fa0c-40ee-ebf8-ee40f0ce5360} isn't
> > being registered, which may involve figuring out what it actually is :)
>
> Okay, I'll see if I can make some sense out of that.
> > Good luck! If a more complete marshalling explanation would be useful, let
> > me know and I'll see what I can do in a few days.
>
> There is no hurry, since it works fine if I back out the one patch. But 
> yes, I certainly would appreciate a more complete explanation.

I am 100% certain that the patch is correct. As Mike explained it seems
to have uncovered a bug or installation issue that wasn't triggered
before because we faked the marshalling. I don't think it is a good idea
to back the patch out of cvs as it slightly improves InstallShield.

Rob

[Mike Hearn]

On Sat, 10 Jul 2004 18:24:00 +0100, Robert Shearman wrote:
> I don't think it is a good idea
> to back the patch out of cvs as it slightly improves InstallShield.

How does it improve InstallShield? I only started looking at IS stuff
again after the patch went in, so I don't have anything to compare it
against.

thanks -mike