fix allp test

tests/node_allp.rs

@@ -38,35 +38,35 @@ fn check_node_allp() {
     // starts with original signal * -0.7
     let mut v = vec![0.7; (2.0 * 44.1_f32).ceil() as usize];
     // silence for 1ms, which is the internal delay of the allpass
-    v.append(&mut vec![0.0; (1.0 * 44.1_f32).floor() as usize - 3]);
+    v.append(&mut vec![0.0; (1.0 * 44.1_f32).floor() as usize - 1]);
 
     // allpass feedback of the original signal for 2ms:
     // XXX: the smearing before and after the allpass is due to the
     // cubic interpolation!
-    v.append(&mut vec![-0.01606, 0.13158, 0.54748]);
+    v.append(&mut vec![-0.03748519, 0.37841395, 0.5260659]);
     v.append(&mut vec![0.51; (2.0 * 44.1_f32).ceil() as usize - 3]);
     // 1ms allpass silence like before:
-    v.append(&mut vec![0.5260659, 0.37841395, -0.03748519]);
+    v.append(&mut vec![0.54748523, 0.13158606, -0.016065884]);
-    v.append(&mut vec![0.0; (1.0 * 44.1_f32).floor() as usize - 6]);
+    v.append(&mut vec![0.0; (1.0 * 44.1_f32).floor() as usize - 4]);
 
     // 2ms the previous 1.0 * 0.7 fed back into the filter,
     // including even more smearing due to cubic interpolation:
     v.append(&mut vec![
-        -0.00035427228,
+        -0.0019286226,
-        0.006157537,
+        0.04086761,
-        -0.005423375,
+        -0.1813516,
-        -0.1756484,
+        -0.35157663,
-        -0.39786762,
+        -0.36315754,
-        -0.3550714,
+        -0.35664573,
     ]);
     v.append(&mut vec![-0.357; (2.0 * 44.1_f32).floor() as usize - 5]);
     v.append(&mut vec![
-        -0.35664573,
+        -0.3550714,
-        -0.36315754,
+        -0.39786762,
-        -0.35157663,
+        -0.1756484,
-        -0.1813516,
+        -0.005423375,
-        0.04086761,
+        0.006157537,
-        -0.0019286226,
+        -0.00035427228,
     ]);
     v.append(&mut vec![0.0; 10]);