> But I must point out typical error (I saw exactly the same error in
Haskell
> HDirect/COM library): ocaml finalization and com release are _essentially
> different_ entities! Sorry for not being sufficiently verbose - i have not
> enough time right now. If You doubt this is an issue - I will definitely
> clarify my point.

Yes, i would actually ear your arguments about that point, because it does
not seems clear to me : the AddRef/Release semantic of COM is a reference
counting to enable the sharing of object and thus determine its time-life
 the last Release() call will trigger effective memory release of the
object ).

For OLE , Release() doesn't not close the Application if it has been put
Visible to the user, but CoUninitiliaze will break the DDE and close it.
That's why i added the primitive "do_not_close" which prevent CoUninitialize
from being called, to enable Applications to remains open even if the ocaml
exit.

> Now only a few words: I've built native olegen version
> (having replaced CoInitialize/CoUninitialize call in DllMain with a call
in
> a static dummy class instance constructor/destructor), and discovered this
> olegen causes segmentaion fault! It turned out to be sufficient simply to
> comment out "i->Release();" in "com_finalize" to make it work.

That behavior should be explained by the following : Release() is called by
the GC, and OCamole perform a full_major cycle on "at_exit" of OCaml. But if
you call CoUnitialize perform having all the COM objects being properly
Released, it can freeze, or simply go on... but then a call to Release()
will cause a seg_fault.

Simply modify ocamole in order to not call "i->Release()" if CoUninitialize
had been made, or ( better ) GC'd all your COM objects before calling
ComUninit.

Nicolas Cannasse


-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners