Re: master 78fc49407b8 1/3: Improve filling of ChangeLog entries

[Dmitry Gutov <dmitry@gutov.dev>]

On 31/01/2024 20:45, Alan Mackenzie wrote:
> Hello, Dmitry.
> 
> On Wed, Jan 31, 2024 at 19:05:57 +0200, Dmitry Gutov wrote:
>> On 31/01/2024 17:32, Alan Mackenzie wrote:
>>> On Wed, Jan 31, 2024 at 17:15:41 +0200, Dmitry Gutov wrote:
>>>> On 31/01/2024 16:01, Po Lu wrote:
>>>>> See any file in CC Mode,
> 
>>>> No shortage of maintainers, you say?
> 
>>> None.
> 
>> One cannot name a package with bus factor of 1 and say it has plenty of
>> people willing to maintain it.
> 
> What's a "bus factor" in this context?

The smallest number of developers who would need to mysteriously 
disappear, for it to become a problem for the project.

> It would appear that one person,
> me, is indeed enough to maintain it.  The rate of bugs reported for it
> has sunk to near zero, possibly because of the release of the tree sitter
> C Mode in Emacs 29.1.

That's good.

>> I don't mean to criticize your work (not knowing the exact tradeoffs),
>> but it's plainly a bad example.
> 
> Bad example of what?

Of a body of Lisp code maintained by different developers, who all 
accepted (and possibly chosen) its current style, thereby justifying it 
as something usable as a standard, rather than only a personal preference.

> [ .... ]
> 
>>> After several days of struggling with named-let, cl-labels, and friends,
>>> I vote for the plain Lisp, even if it does need more lines to express.
>>> It is simply less work.
> 
>> I've never used named-let, and very rarely cl-labels.
> 
> I've never used either, but still need to debug them.  :-(

edebug usually helps, no?

>> The latter is a very simple idea, though: create a bunch of local
>> function definitions. Like nested functions in Python, for example.
> 
> It's a complicated macro.  Why is it needed at all?  Is there anything
> using it that couldn't be conveniently written in plain Lisp?

Having local variable bindings with lambda values is a little messier in 
comparison. For example, the use of cl-labels in 
'comp-collect-rev-post-order' seems easy to read and thus justified. 
Also, it allows having mutually recursive functions, which won't be an 
option in sequential bindings (but I don't see this property taken 
advantage of much).