[kaffe] Kaffe Weekly News - 2002 Sep 7

[Patrick Tullmann]

Tim wrote:
> [optimized jit moving JNI functions around patch in queue]
>
> After having some problems with this patch, we think there might be
> a better solution for this problem.  The issue seems to lie more
> with the usage of the program counter to determine whether something
> is JNI or not.  Since the JNI functions only have one exception
> handler per stack frame you only have to store a pointer to the
> stack frame.  This is in contrast to exception handling in java
> where you need the pc and stack frame because the try block doesn't
> have to cover the entire frame.  Anyways, i made the change to the
> janosvm and it seems to work.

I was thinking that this would cause problems because if the exception
handler is associated with a stack frame (and not a pc range) it may
catch things outside the "try block" we're trying to hack up.
However, the exception handler is only valid when its on the chain of
exceptPtrs off the current thread, so that gives us the scoping we
want.

So, I'll volunteer to move Tim's fixes from JanosVM into a vanilla
Kaffe tree and submit a new patch.

-Pat

----- ----- ---- ---  ---  --   -    -      -         -               -
Pat Tullmann                                       tullmann at cs.utah.edu
   Ignorance more frequently begets confidence than does knowledge.