ExceptionClear does not work as expected
Hi,
I am wanting to conditionally handle an exception occurring in a general
callback from native code to a java method. My approach is to
query env->ExceptionOccurred() and then if found, call env->ExceptionClear()
to temporarily deactivate the exception, but save a global reference to
the exception. If my handling code does not handle it, then ultimately
I want to call env->Throw() on the global reference to the originally obtained
excetion from ExceptionOccurred().
The problem is that ExceptionClear() seems to destroy the exception object,
and though env->Throw() does not complain, the java application appears to
have thrown a java.lang.Object!!! Apparantly ExceptionClear() destroys the
original exception.
How for example, am I going to be able to wrap the exception in another
exception and throw that if ExceptionClear() destroys the original? Is this
a bug in the JVM? Should I report it to Sun? Or is there an alternative
to ExceptionClear() that allows me to keep the original exception and wrap it.
Andy
[1137 byte] By [
p7ea] at [2007-10-3 0:40:52]
You're right! I tried dumming down my code to match what I think you did
and then it worked for me. In essence, what I was originally doing is:
1. calling ExceptionOccurred() to get "t"
2. calling ExceptionClear()
3. calling gt = NewGlobalRef(t)
4. DeleteLocalRef(t)
5. creating a new C++ struct wrapper for gt with a destructor
that DeleteLocalRef(gt-member)
6. throwing that struct as a C++ exception
7. catching that C++ struct as a reference
8. in the catch code, calling env->Throw(mystruct.gt)
9. return 0 from the native function
allowing the destructor for the C++ call to fire at will and call
DeleteGlobalRef(gt-member)
that last DeleteGlobalRef in the destructor of the C++ wrapper was the
problem. Something weird about my C++ compiler, because when I remove
it from the destructor and do it explicitly as step 10, everything is ok.
Thanks again,
Andy
p7ea at 2007-7-14 17:35:05 >
Don't blame the compiler too much. I am in the same boat, and I also protect the exception with a global ref, then clear the exception, yet I can get the JVM to crash during GC in the VMThread with 1.5_09, and it looks like it's fixed in 1.6. So, first look at the hs* file generated to see where you crash, before you blame the compiler.
