Thanks for giving it a spin. Your test case is a good one, but there
seem to be logic errors.
This line:
site.setTarget(target);
should probably be
site.setTarget(this);
I'm surprised you don't get an error at that point. Your target takes
2 string args, while the indy site takes just one. I think the call
to CallSite.checkTarget should be throwing an error at this point.
(This is also where static typing would help.)
The crash happens, perhaps, because the JVM is trying to raise a
WrongMethodType and the assembly code has left the stack a little bit
off-balance. Our error-reporting paths need work. (ASAP, since this
will be lots of people's experience!)
It is probably good style to close the CallSite into the "method not
found" handler, but in that case you don't need to redundantly close
in the CallSite.name.
Thanks!
-- John
On May 18, 2009, at 1:25 PM, Yehuda Katz wrote:
I'm trying to implement a simple method missing using invokeDynamic.
I've been able to make it work using static methods or static
methods that proxy to virtual methods (but not directly to virtual
methods). However, I'm trying to add support for including the name
of the method in the call to MethodMissing.
Based on the Fidget demo, it seemed like the way to go would be to
create a custom JavaMethodHandle that encapsulated the logic. I
wasn't able to get it working (mostly in the setTarget step).
However, during some experiments, I encountered a segfault in the
JVM, so I figured I'd report it. The details (including the minimal
test case) are at http://gist.github.com/113717.
That said, I'm curious about the appropriate way to wire this up.
I'm rather new to all of this, so I'm probably missing something
obvious.
_______________________________________________
mlvm-dev mailing list
mlvm-dev@openjdk.java.net
http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
