On Tue, 15 Jul 2008, Michael Abbott wrote:
> Only half of my patch for this bug has gone into trunk, and without the
> rest of my patch there remains a leak.

I think I might need to explain a little more about the reason for this
patch, because obviously the bug it fixes was missed the last time I
posted on this bug.

So here is the missing part of the patch:

> --- numpy/core/src/scalartypes.inc.src  (revision 5411)
> +++ numpy/core/src/scalartypes.inc.src  (working copy)
> @@ -1925,19 +1925,30 @@
>          goto finish;
>      }
>
> +    Py_XINCREF(typecode);
>      arr = PyArray_FromAny(obj, typecode, 0, 0, FORCECAST, NULL);
> -    if ((arr==NULL) || (PyArray_NDIM(arr) > 0)) return arr;
> +    if ((arr==NULL) || (PyArray_NDIM(arr) > 0)) {
> +        Py_XDECREF(typecode);
> +        return arr;
> +    }
>      robj = PyArray_Return((PyArrayObject *)arr);
>
>  finish:
> -    if ((robj==NULL) || (robj->ob_type == type)) return robj;
> +    if ((robj==NULL) || (robj->ob_type == type)) {
> +        Py_XDECREF(typecode);
> +        return robj;
> +    }
>      /* Need to allocate new type and copy data-area over */
>      if (type->tp_itemsize) {
>          itemsize = PyString_GET_SIZE(robj);
>      }
>      else itemsize = 0;
>      obj = type->tp_alloc(type, itemsize);
> -    if (obj == NULL) {Py_DECREF(robj); return NULL;}
> +    if (obj == NULL) {
> +        Py_XDECREF(typecode);
> +        Py_DECREF(robj);
> +        return NULL;
> +    }
>      if (typecode==NULL)
>          typecode = PyArray_DescrFromType(PyArray_@TYPE@);
>      dest = scalar_value(obj, typecode);

On the face of it it might appear that all the DECREFs are cancelling out
the first INCREF, but not so.  Let's see two more lines of context:

>      src = scalar_value(robj, typecode);
>      Py_DECREF(typecode);

Ahah.  That DECREF balances the original PyArray_DescrFromType, or maybe
the later call ... and of course this has to happen on *ALL* return paths.
If we now take a closer look at the patch we can see that it's doing two
separate things:

1. There's an extra Py_XINCREF to balance the ref count lost to
PyArray_FromAny and ensure that typecode survives long enough;

2. Every early return path has an extra Py_XDECREF to balance the creation
of typecode.

I rest my case for this patch.