Differences in Exception Handling Behavior Under /CLR
Basic Concepts in Using Managed Exceptions discusses exception handling in managed applications. In this topic, differences from the standard behavior of exception handling and some restrictions are discussed in detail. For more information, see The _set_se_translator Function.
Jumping Out of a Finally Block
In native C/C++ code, jumping out of a __finally block using structured exception handling (SEH) is allowed although it produces a warning. Under /clr, jumping out of a finally block causes an error:
// clr_exception_handling_4.cpp
// compile with: /clr
int main() {
try {}
finally {
return 0; // also fails with goto, break, continue
}
} // C3276
Raising Exceptions Within an Exception Filter
When an exception is raised during the processing of an exception filter within managed code, the exception is caught and treated as if the filter returns 0.
This is in contrast to the behavior in native code where a nested exception is raised, the ExceptionRecord field in the EXCEPTION_RECORD structure (as returned by GetExceptionInformation) is set, and the ExceptionFlags field sets the 0x10 bit. The following example illustrates this difference in behavior:
// clr_exception_handling_5.cpp
#include <windows.h>
#include <stdio.h>
#include <assert.h>
#ifndef false
#define false 0
#endif
int *p;
int filter(PEXCEPTION_POINTERS ExceptionPointers) {
PEXCEPTION_RECORD ExceptionRecord =
ExceptionPointers->ExceptionRecord;
if ((ExceptionRecord->ExceptionFlags & 0x10) == 0) {
// not a nested exception, throw one
*p = 0; // throw another AV
}
else {
printf("Caught a nested exception\n");
return 1;
}
assert(false);
return 0;
}
void f(void) {
__try {
*p = 0; // throw an AV
}
__except(filter(GetExceptionInformation())) {
printf_s("We should execute this handler if "
"compiled to native\n");
}
}
int main() {
__try {
f();
}
__except(1) {
printf_s("The handler in main caught the "
"exception\n");
}
}
Output
Caught a nested exception
We should execute this handler if compiled to native
Disassociated Rethrows
/clr does not support rethrowing an exception outside of a catch handler (known as a disassociated rethrow). Exceptions of this type are treated as a standard C++ rethrow. If a disassociated rethrow is encountered when there is an active managed exception, the exception is wrapped as a C++ exception and then rethrown. Exceptions of this type can only be caught as an exception of type System::SEHException.
The following example demonstrates a managed exception rethrown as a C++ exception:
// clr_exception_handling_6.cpp
// compile with: /clr
using namespace System;
#include <assert.h>
#include <stdio.h>
void rethrow( void ) {
// This rethrow is a dissasociated rethrow.
// The exception would be masked as SEHException.
throw;
}
int main() {
try {
try {
throw gcnew ApplicationException;
}
catch ( ApplicationException^ ) {
rethrow();
// If the call to rethrow() is replaced with
// a throw statement within the catch handler,
// the rethrow would be a managed rethrow and
// the exception type would remain
// System::ApplicationException
}
}
catch ( ApplicationException^ ) {
assert( false );
// This will not be executed since the exception
// will be masked as SEHException.
}
catch ( Runtime::InteropServices::SEHException^ ) {
printf_s("caught an SEH Exception\n" );
}
}
Output
caught an SEH Exception
Exception Filters and EXCEPTION_CONTINUE_EXECUTION
If a filter returns EXCEPTION_CONTINUE_EXECUTION in a managed application, it is treated as if the filter returned EXCEPTION_CONTINUE_SEARCH. For more information on these constants, see try-except Statement.
The following example demonstrates this difference:
// clr_exception_handling_7.cpp
#include <windows.h>
#include <stdio.h>
#include <assert.h>
int main() {
int Counter = 0;
__try {
__try {
Counter -= 1;
RaiseException (0xe0000000|'seh',
0, 0, 0);
Counter -= 2;
}
__except (Counter) {
// Counter is negative,
// indicating "CONTINUE EXECUTE"
Counter -= 1;
}
}
__except(1) {
Counter -= 100;
}
printf_s("Counter=%d\n", Counter);
}
Output
Counter=-3
The _set_se_translator Function
The translator function, set by a call to _set_se_translator, affects only catches in unmanaged code. The following example demonstrates this limitation:
// clr_exception_handling_8.cpp
// compile with: /clr /EHa
#include <iostream>
#include <windows.h>
#include <eh.h>
#pragma warning (disable: 4101)
using namespace std;
using namespace System;
#define MYEXCEPTION_CODE 0xe0000101
class CMyException {
public:
unsigned int m_ErrorCode;
EXCEPTION_POINTERS * m_pExp;
CMyException() : m_ErrorCode( 0 ), m_pExp( NULL ) {}
CMyException( unsigned int i, EXCEPTION_POINTERS * pExp )
: m_ErrorCode( i ), m_pExp( pExp ) {}
CMyException( CMyException& c ) : m_ErrorCode( c.m_ErrorCode ),
m_pExp( c.m_pExp ) {}
friend ostream& operator <<
( ostream& out, const CMyException& inst ) {
return out << "CMyException[\n" <<
"Error Code: " << inst.m_ErrorCode << "]";
}
};
#pragma unmanaged
void my_trans_func( unsigned int u, PEXCEPTION_POINTERS pExp ) {
cout << "In my_trans_func.\n";
throw CMyException( u, pExp );
}
#pragma managed
void managed_func() {
try {
RaiseException( MYEXCEPTION_CODE, 0, 0, 0 );
}
catch ( CMyException x ) {}
catch ( ... ) {
printf_s("This is invoked since "
"_set_se_translator is not "
"supported when /clr is used\n" );
}
}
#pragma unmanaged
void unmanaged_func() {
try {
RaiseException( MYEXCEPTION_CODE,
0, 0, 0 );
}
catch ( CMyException x ) {
printf("Caught an SEH exception with "
"exception code: %x\n", x.m_ErrorCode );
}
catch ( ... ) {}
}
// #pragma managed
int main( int argc, char ** argv ) {
_set_se_translator( my_trans_func );
// It does not matter whether the translator function
// is registered in managed or unmanaged code
managed_func();
unmanaged_func();
}
Output
This is invoked since _set_se_translator is not supported when /clr is used
In my_trans_func.
Caught an SEH exception with exception code: e0000101