parser: remove the node instead here.

src/parser/mod.rs

@@ -3073,16 +3073,19 @@ where
         text.drain(..end);
 
         let adjust = spx.consume(end) + 1;
-        // Do we need to account separately for the parent? I _think_ it always
-        // has the same start col to begin with as `sourcepos` itself.
         assert_eq!(
             sourcepos.start.column,
             parent.data.borrow().sourcepos.start.column
         );
 
-        // HACK: See tests::fuzz::echaw9. The paragraph doesn't exist in the source!
-        sourcepos.start.column = usize::min(sourcepos.end.column, adjust);
-        parent.data.borrow_mut().sourcepos.start.column = sourcepos.start.column;
+        // See tests::fuzz::echaw9. The paragraph doesn't exist in the source,
+        // so we remove it.
+        if sourcepos.end.column < adjust {
+            parent.detach();
+        } else {
+            sourcepos.start.column = adjust;
+            parent.data.borrow_mut().sourcepos.start.column = adjust;
+        }
 
         grandparent.data.borrow_mut().value =
             NodeValue::TaskItem(if symbol == ' ' { None } else { Some(symbol) });

src/tests/fuzz.rs

@@ -350,9 +350,7 @@ fn echaw9() {
         "-\t[ ]
",
         (document (1:1-1:14) [
             (list (1:1-1:14) [
-                (taskitem (1:1-1:14) [
-                    (paragraph (1:14-1:14)) // Unsatisfying. See parser::process_tasklist.
-                ])
+                (taskitem (1:1-1:14))
             ])
         ]),
     );