Fix overlap-linear context window crash when context_overlap is 0

create_weights_overlap_linear computed the right-edge ramp with
weights_torch[-context_overlap:], which becomes weights_torch[0:]
(the whole tensor) when context_overlap is 0, while the ramp is a
zero-length tensor. Assigning the empty ramp raised a RuntimeError.

context_overlap == 0 is reachable: the context-window nodes allow a
minimum of 0, and the WAN/LTXV nodes derive it via max(overlap // 4, 0).
Return the all-ones weights early when there is no overlap band to blend.

comfy/context_windows.py

@@ -955,6 +955,8 @@ def create_weights_overlap_linear(length: int, full_length: int, idxs: list[int]
     # based on code in Kijai's WanVideoWrapper: https://github.com/kijai/ComfyUI-WanVideoWrapper/blob/dbb2523b37e4ccdf45127e5ae33e31362f755c8e/nodes.py#L1302
     # only expected overlap is given different weights
     weights_torch = torch.ones((length))
+    if context_overlap <= 0:
+        return weights_torch
     # blend left-side on all except first window
     if min(idxs) > 0:
         ramp_up = torch.linspace(1e-37, 1, context_overlap)

tests-unit/comfy_test/context_windows_test.py

@@ -0,0 +1,29 @@
+import sys
+from unittest.mock import MagicMock, patch
+
+import torch
+
+# comfy.model_management initializes the torch device at import time and requires
+# CUDA (or --cpu), which the unit-test environment does not provide. Stub it so
+# importing comfy.context_windows does not trigger device initialization.
+with patch.dict(sys.modules, {"comfy.model_management": MagicMock()}):
+    from comfy.context_windows import create_weights_overlap_linear
+
+
+class TestCreateWeightsOverlapLinear:
+    def test_zero_overlap_returns_ones(self):
+        # context_overlap == 0 is reachable (the context-window nodes allow min=0,
+        # and the WAN/LTXV nodes derive it via max(overlap // 4, 0)). With no overlap
+        # band there is nothing to blend, so every weight is 1.
+        w = create_weights_overlap_linear(16, 100, list(range(20, 36)), 0)
+        assert w.shape == (16,)
+        assert torch.equal(w, torch.ones(16))
+
+    def test_overlap_ramps_edges(self):
+        w = create_weights_overlap_linear(16, 100, list(range(20, 36)), 4)
+        assert w.shape == (16,)
+        # middle (non-first) window: left edge ramps up to 1 over the overlap width
+        assert w[0] < w[3]
+        assert torch.isclose(w[3], torch.tensor(1.0))
+        # right edge ramps back down
+        assert w[-1] < w[-4]