Fix out of bounds integer behavior in full

Author: ndgrigorian

dpctl/tensor/_ctors.py

@@ -1101,6 +1101,8 @@ def full(
         fill_value = int(fill_value.real)
     elif fill_value_type is complex and np.issubdtype(dtype, np.floating):
         fill_value = fill_value.real
+    elif fill_value_type is int and np.issubdtype(dtype, np.integer):
+        fill_value = _to_scalar(fill_value, dtype)
 
     hev, _ = ti._full_usm_ndarray(fill_value, res, sycl_queue)
     hev.wait()

dpctl/tests/test_usm_ndarray_ctor.py

@@ -1477,6 +1477,19 @@ def test_full_strides():
     assert np.array_equal(dpt.asnumpy(X), Xnp)
 
 
+def test_full_gh_1230():
+    q = get_queue_or_skip()
+    dtype = "i4"
+    dt_maxint = dpt.iinfo(dtype).max
+    X = dpt.full(1, dt_maxint + 1, dtype=dtype, sycl_queue=q)
+    X_np = dpt.asnumpy(X)
+    assert X.dtype == dpt.dtype(dtype)
+    assert np.array_equal(X_np, np.full_like(X_np, dt_maxint + 1))
+
+    with pytest.raises(OverflowError):
+        dpt.full(1, dpt.iinfo(dpt.uint64).max + 1, sycl_queue=q)
+
+
 @pytest.mark.parametrize(
     "dt",
     _all_dtypes[1:],