Re: JVM access violation in Type Browser

Roger I Martin PhD wrote:

[SNIP - about exception handling]

>>
> Very much like that.  For
> http://cvs.sourceforge.net/viewcvs.py/jawinproject/jawin/cpp/TypeLibWrapper/org_jawin_win32_ITypeInfo.cpp?rev=1.1&only_with_tag=MAIN&view=markup
>
> most need just coarse granularity.  For jawin marhsaling we may want
> fine granularity.

We already have (at least) two helper macros:
   JNI_HR(hr) for checking and throwing a JNIComException if a HRESULT
failed.
and
   HANDLE_JNI_EXCEPTIONS() for catching all the exceptions we throw.

So at least for a first attempt to "catch all the rest", and avoid
native JVM-crashes, I think I will look into applying a catch(...) macro
over the weekend (I just did a quick search in Stroustrup's C++ guide,
and it seems he describes a way to retrieve the caught exception, when
using the ...-catch, so it should hopefully be trivial).

>  But maybe after refactoring marshaling? Applying
> java.nio package?
>

We have to get to that sometime, but what are your priorities with
regards to the event support, John Pompeii are asking for, and Robert
and Vitaly have proposed a solution for. In my opinion this is more
needed (since the refactoring to java.nio does not actually add new
features, and the existing implementation works - but I agree that
perhaps the java.nio can make it faster).

If you want to look at the event support first, I think I can create the
code for the native test in next week (as sketched in another mail
yesterday), then you, Robert and Vitaly "just" need to do the hard part,
doing the actual implementation :-)

Best Regards
Morten