Re: Building GCC 3.3 / MingW: The Saga Continues

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Mohan" == Mohan Embar <gnustuff@thisiscool.com> writes:

Mohan> Exception in thread "main" java.lang.UnknownError:
Mohan> Runtime.loadLibrary not implemented
Mohan> I'll look into this later.

On Windows we don't compile libltdl, which is the library we use to
get dynamic loading.  I don't know why, but I assume it is related to
the fact that we don't build libgcj as a DLL on Windows.

Not building libgcj as a DLL is fixable; Anthony Green recently posted
a message about how to do it.  I don't know whether this would make it
possible for Runtime.loadLibrary to work.  It's worth looking in to.

Tom> I think in the long term the plan is to separate out the Windows
Tom> java.net native code into its own files.  My understanding is that
Tom> sharing the code here is not practical.

Mohan> How impractical is this? Could someone educate me? Given the
Mohan> large amount of code sharing made possible with MS's
Mohan> BSD-layer-wannabe implementation, although imperfect, wouldn't
Mohan> creating two parallel code sets create a maintenance headache?

That's what I thought, too, but the more Windows-knowledgeable people
think otherwise.

Mohan> To my knowledge, a formal patch was never created for
Mohan> Ranjit's suggestion:
Mohan> http://gcc.gnu.org/ml/java/2002-12/msg00295.html

I think there was another thread a bit later explaining that this
approach wouldn't work too well in the end.

Tom