Re: -shared option

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Ranjit" == Ranjit Mathew <rmathew4lists at hotmail dot com> writes:

Ranjit> (Which suddenly makes me realise that I haven't got the foggiest 
Ranjit> idea about how GCJ is able to determine class layout, method and 
Ranjit> variable information etc. when it doesn't have the convenience 
Ranjit> of a header file as in C++... Anyone care to enlighten?

It does have "headers", in the form of .class files.
.class files are actually better than headers in several ways.

Ranjit> The point of all this explanation being that if GCJ has
Ranjit> to generate nice code that works with classes defined in a
Ranjit> DLL as well as in a static library *without* the user
Ranjit> being able to indicate it using an attribute, we need
Ranjit> some more magic...

It also occurs to me that gcjh should generate headers that annotate
native methods this way.  My understanding is that you don't have to
do non-native methods since the plan is to have gcj export them
directly.  Is that right?

And, the hand-maintained Object.h and Class.h headers should also be
updated.  And all exported CNI functions should be marked.

A new platform define seems cleanest.

Tom