That is indeed exactly what we are talking about.
Based on a really quick scan of the documentation, I had the
impression that it was also motivated in part by references
from native code that are not visible to the GC. I think that's
a different issue, which doesn't arise since JNI has its
own mechanisms for doing similar things.
Hans
> -----Original Message-----
> From: Bart Jacobs [mailto:bart.jacobs@cs.kuleuven.ac.be]
> Sent: Tuesday, May 03, 2005 12:04 PM
> To: Boehm, Hans; 'Thomas Hawtin'; javaMemoryModel@cs.umd.edu
> Subject: RE: JavaMemoryModel: Finalization idioms
>
>
> The motivation usually given for GC.KeepAlive in .NET is the following
> scenario:
>
> class File{
> IntPtr handle;
>
> [DllImport(...)]
> extern void NativeWrite(IntPtr handle, byte b);
>
> [DllImport(...)]
> extern void NativeFree(IntPtr handle);
>
> public void Write(byte b){
> NativeWrite(this.handle, b);
> GC.KeepAlive(this);
> }
>
> ~File(){
> NativeFree(handle);
> }
> }
>
> Without the KeepAlive call, the object might be considered
> unreachable at the start of the NativeWrite call, which might
> cause the handle to be freed concurrently with the NativeWrite call.
>
> It's not always recognized, but I think this is just a
> special case of what this list talks about.
>
> Bart
>
> -----Original Message-----
> From: owner-javamemorymodel@cs.umd.edu
> [mailto:owner-javamemorymodel@cs.umd.edu] On Behalf Of Boehm, Hans
> Sent: Tuesday, May 03, 2005 11:00 AM
> To: Thomas Hawtin; javaMemoryModel@cs.umd.edu
> Subject: RE: JavaMemoryModel: Finalization idioms
>
> I think we really need keepAlive in the library. Perhaps
> java.lang.ref is a good place to put it. If you implement in
> the standard library, with suitable VM cooperation, it should
> be nearly free. .NET has GC.KeepAlive which I believe can be
> used for this purpose, though the primary motivation there
> seems to be different.
>
> Hans
>
>
>
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:01:10 EDT