On 5/15/07, Alek Storm <[EMAIL PROTECTED]> wrote:
On 5/14/07, Mehmet Yavuz Selim Soyturk <[EMAIL PROTECTED] >
wrote:
> Should we not be able to use an object that implements 'invoke' as a
> method of another object? There is some strange behaviour when I try
> to.
I see what you're saying, and it looks like you want a functor. We can't do
exactly what you want in that syntax, because methods are loosely tied to
objects in Parrot. Parrot sees 'Func' as a sub, so it wouldn't make any
sense to tell the sub it's called by itself. Regardless of what
'find_method' returns, the invocant is $P2 when you use PIR syntax. Here's
how to get the effect you want:
.sub main :main
$P0 = newclass 'Func'
$P1 = newclass 'Obj'
$P2 = new 'Obj'
$P3 = find_method $P2, 'some_method'
$P3.$P3()
.end
The patch has been reverted, though, so that won't work until it's fixed and
reapplied.
What I really meant is $P2.$P3() . With Allison's patch both
(*) $P2.'some_method'()
and
(**) $P3 = find_method $P2, 'some_method'
$P2.$P3()
caused the same strange behaviour. I applied your patch, and (**)
worked as I expected. But (*) caused an error:
too many arguments passed (3) - 2 params expected
I think that I found the problem. callmethodcc calls
VTABLE_find_method, which invokes the find_method override, which is a
PIR sub. That causes that interp->current_args gets overwritten.
I have a patch which tries to solve that problem:
--- src/ops/object.ops (revision 18542)
+++ src/ops/object.ops (working copy)
@@ -50,11 +50,20 @@
PMC *method_pmc, *object;
opcode_t *next;
STRING *meth;
+ opcode_t *current_args;
object = $1;
meth = $2;
next = expr NEXT();
+
+ /* If VTABLE_find_method calls a find_method override (which
+ is another PIR sub), interp->current_args gets overwritten.
+ We have to remember it.
+ */
+ current_args = interp->current_args;
method_pmc = VTABLE_find_method(interp, object, meth);
+ interp->current_args = current_args;
+
if (!method_pmc) {
real_exception(interp, next, METH_NOT_FOUND,
"Method '%Ss' not found", meth);
I only did it for the callmethodcc_p_sc opcode though, because I don't
know if it's the right solution. A possible problem: could
current_args get GC'd in the meantime?
--
Mehmet
