Re: emacs-31 29bbd02347d: Fix treesit_cursor_helper_1 for zero-width nodes (bug#81436)

> On Jul 27, 2026, at 12:17 PM, Eli Zaretskii <[email protected]> wrote:
> 
>> From: Michael Albinus <[email protected]>
>> Cc: Yuan Fu <[email protected]>
>> Date: Mon, 27 Jul 2026 18:19:03 +0200
>> 
>> Yuan Fu <[email protected]> writes:
>> 
>> Hi,
>> 
>>> branch: emacs-31
>>> commit 29bbd02347da44599cc9778d9d626fc5c8ca4b2e
>>> Author: Yuan Fu <[email protected]>
>>> Commit: Yuan Fu <[email protected]>
>>> 
>>>    Fix treesit_cursor_helper_1 for zero-width nodes (bug#81436)
>>> 
>>>    * lisp/treesit.el (treesit--some): Add edebug declare.
>>>    * src/treesit.c (treesit_cursor_helper_1): Don't use
>>>    ts_tree_cursor_goto_first_child_for_byte for zero-width nodes.
>> 
>> I'm not sure whether it was *this* commit. However, tree-sitter tests on
>> emba fail now:
> 
> It's definitely that commit.  On my system, the test crashes due to
> infinite recursion.
> 
> I tried to fix that now.  Yuan, please see if I got the logic right.

Thanks Eli! Your fix is right, the idea is that we try the for_byte variant (skipped if start = end), and if that didn’t find a child, try the normal variant. And here’s a trivial patch that hopefully further explains why we need to skip the call to for_byte variant.

Yuan

From 84b3f4e500e783b78e758890f48aa5512c9e9181 Mon Sep 17 00:00:00 2001
From: Yuan Fu <[email protected]>
Date: Tue, 28 Jul 2026 03:13:28 -0700
Subject: [PATCH] ; * src/treesit.c (treesit_cursor_helper_1): Further
 explanation.

---
 src/treesit.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/treesit.c b/src/treesit.c
index 210fae671b3..66644571f83 100644
--- a/src/treesit.c
+++ b/src/treesit.c
@@ -4263,7 +4263,7 @@ treesit_cursor_helper_1 (TSTreeCursor *cursor, TSNode *target,
      Also, ts_tree_cursor_goto_first_child_for_byte can't find
      zero-width nodes (which exist and are legit, e.g., markdown's
      block_continuation), because a zero-width node can't contain a pos
-     (end > pos).  */
+     (end > pos); instead, it'll falsely return the next child.  */
   if (!((start_pos != end_pos
 	 && ts_tree_cursor_goto_first_child_for_byte (cursor, start_pos) != -1)
 	|| ts_tree_cursor_goto_first_child (cursor)))
-- 
2.39.5 (Apple Git-154)