On Thu, Mar 19, 2009 at 5:56 AM, Gábor Melis <mega@retes.hu> wrote:
On Lunes 16 Marzo 2009, Jean-Claude Beaudoin wrote:
> On Mon, Mar 16, 2009 at 1:34 AM, Juho Snellman <jsnell@iki.fi> wrote:
> >
> > Yes, it switches stacks in call_into_lisp_first_time.

In case you could find this information useful here is the end result of this
situation as far as CL+J is concerned.

I ended up writing SBCL specific code for the Linux version that will, hopefully,
prevent an attempt to initialize the JVM from the "initial thread". And since I
cannot suspend the "initial thread" in all circumstances I also wrote some more
code that should prevent any attempt to attach the JVM to the "initial thread".
But this is a bit like living under a Damocles sword since any failure in these
defenses will result in a core dump...

I can always hope to identify, in some future, a JVM replacement that is not as
sensitive to stack manipulations as Sun's code.

Thank you all for your help,

Jean-Claude Beaudoin