Strange exceptions in ASP.NET + COM application

[danil.shopyrin]

Hello!

We have some code that is written in COM. It was used in a
"ISAPI-extension server " and everithing was fine. And it is fine at
the moment.

But now this code is used from the ASP.NET server. And we receive some
unknown exceptions. Them occurs after a long period of working (about
of 10,000 stress-test-queries). We can't ever recognize these
exceptions.

For example, yesterday I've received an exception in the method:

STDMETHODIMP CGobiSrv::put_FSCreator(IUnknown* fsCreator)
{
SAFE_METHOD_START("CGobiSrv::put_FSCreator")
AFX_MANAGE_STATE(AfxGetStaticModuleState())
serve->SetFSCreator( fsCreator );
GTRACEL("FSCreator received\n");
return S_OK;
SAFE_METHOD_END
}

We try to catch exceptions using the following code:

#define SAFE_METHOD_START(method_name) \
CMethodExecutionViewer method_ex_viewer( this->GetClassID(),
this->GetInstanceNum(), method_name ); \
try { \

#define SAFE_METHOD_END \
} catch (std::exception& e) { \
method_ex_viewer.ViewException( "standard", e.what() ); \
return E_FAIL; \
} catch (CException* e) { \
const int MAX_ERR_LEN = 256; \
GBuffer<TCHAR> err_msg(MAX_ERR_LEN); \
e->GetErrorMessage( err_msg, MAX_ERR_LEN ); \
method_ex_viewer.ViewException( _TEXT("MFC"), static_cast<TCHAR*>(
err_msg ) ); \
return E_FAIL; \
} catch (_com_error &e) { \
method_ex_viewer.ViewException( _TEXT("COM"), e.ErrorMessage() ); \
return E_FAIL; \
} catch (CAtlException &e) { \
const int MAX_ERR_LEN = 256; \
GBuffer<TCHAR> hr_code(MAX_ERR_LEN); \
ltoa(e, hr_code, 16); \
method_ex_viewer.ViewException( _TEXT("ATL"), static_cast<TCHAR*>(
hr_code ) ); \
return E_FAIL; \
} catch (...) { \
method_ex_viewer.ViewException( "unknown", "" ); \
return E_FAIL; \
} \

and our exceptions are always fall into the catch (...) section.

There is some EnterCriticalSection()/LeaveCriticalSection() that occurs
in SetFSCreator() method because of the fact that some global data is
shared between COM-objects. It's is because we use very legacy code.
Maybe the error is somewhere in interlocking, but why it's working in
the ISAPI-mode?

We use _ATL_APARTMENT_THREADED. We tried to run ASP script with and
without ASPCompat="true". We tried to run on .NET 1.1 and .NET 2.0. The
song remains the same.

It's very hard to debug. Because these exceptions occurs in different
places and after a long time of working.

Can anybody help us?

--
Danil

[Bruce Barker]

you need to set AspCompat=true, to support STA (apartment model) com
objects. also are you calling Marshal.ReleaseComObject(), to release the
object?

note: STA com objected can not be used directly from a webservice.

-- bruce (sqlwork.com)
<da************@gmail.com> wrote in message
news:11**********************@i39g2000cwa.googlegr oups.com...

Hello!

We have some code that is written in COM. It was used in a
"ISAPI-extension server " and everithing was fine. And it is fine at
the moment.

But now this code is used from the ASP.NET server. And we receive some
unknown exceptions. Them occurs after a long period of working (about
of 10,000 stress-test-queries). We can't ever recognize these
exceptions.

For example, yesterday I've received an exception in the method:

STDMETHODIMP CGobiSrv::put_FSCreator(IUnknown* fsCreator)
{
SAFE_METHOD_START("CGobiSrv::put_FSCreator")
AFX_MANAGE_STATE(AfxGetStaticModuleState())
serve->SetFSCreator( fsCreator );
GTRACEL("FSCreator received\n");
return S_OK;
SAFE_METHOD_END
}

We try to catch exceptions using the following code:

#define SAFE_METHOD_START(method_name) \
CMethodExecutionViewer method_ex_viewer( this->GetClassID(),
this->GetInstanceNum(), method_name ); \
try { \

#define SAFE_METHOD_END \
} catch (std::exception& e) { \
method_ex_viewer.ViewException( "standard", e.what() ); \
return E_FAIL; \
} catch (CException* e) { \
const int MAX_ERR_LEN = 256; \
GBuffer<TCHAR> err_msg(MAX_ERR_LEN); \
e->GetErrorMessage( err_msg, MAX_ERR_LEN ); \
method_ex_viewer.ViewException( _TEXT("MFC"), static_cast<TCHAR*>(
err_msg ) ); \
return E_FAIL; \
} catch (_com_error &e) { \
method_ex_viewer.ViewException( _TEXT("COM"), e.ErrorMessage() ); \
return E_FAIL; \
} catch (CAtlException &e) { \
const int MAX_ERR_LEN = 256; \
GBuffer<TCHAR> hr_code(MAX_ERR_LEN); \
ltoa(e, hr_code, 16); \
method_ex_viewer.ViewException( _TEXT("ATL"), static_cast<TCHAR*>(
hr_code ) ); \
return E_FAIL; \
} catch (...) { \
method_ex_viewer.ViewException( "unknown", "" ); \
return E_FAIL; \
} \

and our exceptions are always fall into the catch (...) section.

There is some EnterCriticalSection()/LeaveCriticalSection() that occurs
in SetFSCreator() method because of the fact that some global data is
shared between COM-objects. It's is because we use very legacy code.
Maybe the error is somewhere in interlocking, but why it's working in
the ISAPI-mode?

We use _ATL_APARTMENT_THREADED. We tried to run ASP script with and
without ASPCompat="true". We tried to run on .NET 1.1 and .NET 2.0. The
song remains the same.

It's very hard to debug. Because these exceptions occurs in different
places and after a long time of working.

Can anybody help us?

--
Danil

[danil.shopyrin]

Dear Bruce,

thank you for your reply.

I try to use Marshal.ReleaseComObject(). I check the returm value and
it always equals to 0.

Exceptions are still here. The problem is that I don't know what to
catch. I don't understand what's happened. I afraid that something is
happened with the host aspnet_wp process.

There is anonther question. Is there a way to 'reboot' aspnet_wp
process sometimes? For example, one time for each 1000 requests. It can
help to avoid the problem, if we can't solve them.