bug#62158: treesit-end-of-defun error

[Yuan Fu]

> On Mar 20, 2023, at 11:43 AM, Dmitry Gutov <dgutov@yandex.ru> wrote:
> 
> On 14/03/2023 00:04, Yuan Fu wrote:
>> We tried that initially, but end-of-defun doesn’t have the notion of nested 
>> defuns, which leads to problems when end-of-defun-function recognizes nested 
>> defuns. In the following code
>> (defun xxx ()
>>   |
>>   (defun yyy () ...)
>>   (defun zzz () ...)
>>   )
>> If point is at “|” and you call end-of-defun, you’d expect point to move to 
>> the end of yyy, but instead it moves to the end of xxx. That’s because 
>> end-of-defun first runs (beginning-of-defun -1) followed by (end-of-defun 1) 
>> to check if the starting point is in a defun or between two defuns. This is 
>> fine in non-nested defuns, but in this example, the point first goes to the 
>> beginning of xxx, then goes to the end of xxx. And end-of-defun thinks that 
>> we started in a defun and now is at an end of defun, job’s done, and 
>> finishes.
>> The plan is to improve end-of-defun to support nested defuns in Emacs 30. 
>> For now we rebind end-of-defun to treesit-end-of-defun.
> 
> That makes sense, thanks!
> 
> I guess one of the things to try is to call end-of-defun-function first, 
> followed by beginning-of-defun-function. And see if the resulting position is 
> below the original point.

Unfortunately, end-of-defun-function must be called from the beginning of a 
defun, according to its docstring. So that won’t do. We got to upgrade 
beginning/end-of-defun to support nested defuns.

Yuan