Bug fix in motoi_assign

[David Hakim]

Funny ... this will teach me to read all my emails first and not try to
respond one at a time. The patch I applied (and committed) to the main
branch already was basically this portion of yours:

@@ -766,21 +766,24 @@
env->rval_uc = NULL;
}

- /* FIXME!!! nasty but nice */
-
- if( lval->opcell.opcode == DEREFERENCE_LVAL )
- motoi_dereference_rval(uc_operand(p, 0));
- else
- motoi(uc_operand(p, 0));
-
- v1 = opstack_pop(env);
v2 = opstack_pop(env);

+ if ((lval->type == OP_TYPE &&
+ lval->opcell.opcode == ARRAY_LVAL) ||
+ (lval->type == OP_TYPE
+ && lval->opcell.opcode == DEREFERENCE_LVAL)) {
+ v1 = moto_getVarVal(env,dest);
+ } else {
+ motoi(lval);
+ v1 = opstack_pop(env);
+ }
+
if(motox_try_overloaded_method_op(op,v1,v2,NULL) != MOTO_OK) {
motoi_domath(v1, v2, op);
moto_freeVal(env,v1);
moto_freeVal(env,v2);
}
+
break;
}

I'm still investigating the rest of the patch ... I am a little
hesitant to add another memory tracking hashtable to env but once I
understand the implications of that part of your fix I'll post again.

-Dave

Hi Dave,
the same patch as before, but now with one more bug corrected, in
motoi_assign. Now I can run ary3.moto, from perftest.tar.gz, and in 7
minutes
I get the result! Now the big problem is memory from constructors.
Look at
${
use "codex.util";
int i=0;
int j=0;
for (i = 0; i < 100; i++) {
Vector v = new Vector();
v.add("a");
v.add("b");
v.add("c");
v.add("d");
v.add("e");
v.add("f");
v.add("g");
for (j = 0; j < 1000; j++) {
Integer foo = new Integer(j); <---- This
}
}
}$
memory from the new operations never gets released...
Ciao,
Stefano
<motomem.patch>